Natural Plant-Based Additives Can Improve Ruminant Performance by Inuencing the Rumen Microbiome

Background The use of synthetic compounds as growth promoters in animal production, is now limited or even banned by health agencies globally due to human safety concerns. In feedlot cattle, when using high grain diets, it is necessary to supplement the diet with compounds capable of modulating the rumen in order to reduce the incidence of acidosis and improve growth. In this context, natural substances have become promising substitutes. The objective of this study was to evaluate the effects of a natural additive blend (NA) on animal performance, the rumen microbiome and ingestive behavior in 40 young bulls. Results The initial and nal average body weight was similar (P > 0.05) for all diets, although average daily gain increased linearly (P < 0.01) when NA was fed. However, feed eciency improved linearly (P < 0.05) by including NA in the diet. Principal volatile fatty acid: acetic, butyric, isovaleric and valeric decreased linearly (P < 0.02) following NA addition. Similarly, NA addition linearly decreased (P < 0.02) the acetate/propionate ratio. The propionate and isobutyric acid concentrations showed a positive quadratic effect (P < 0.05). Furthermore, NA addition reduced ammonia concentrations (P < 0.001) and ruminal pH was not affected (P > 0.05) by the diets. The rumen microbiome was signicantly different between beef cattle fed the different treatments (P < 0.05), with a reduction in the archaea, and within the Clostridium, Robinsoniella, Acidaminococcus, Acetitomaculum, Succinimonas and Weissella (P < 0.05) seen when NA was fed. The functional capacity of the rumen microbiome was affected following NA supplementation. Overall, we observed Aldehyde oxidase/xanthine dehydrogenase, molybdopterin binding; RecG, N-terminal antiparallel four helix bundle; Transposase, ISC1217; Restriction endonuclease, type II, XamI; Acyl-protein synthetase, LuxE; ABC-2 transporter; which could be related to the natural additives mechanism


Feeding behavior activities
There were no effects of NA blend addition to bull diets on rumination, feed intake, water intake and idle time (P > 0.05; Table 2). ¹CON = control (without natural additives); 2 NA15 -addition of 1.5 g/animal/day of natural additives; 3 NA30 -addition of 3.0 g/animal/ day of natural additives; 4 NA45 -addition of 4.5 g/animal/day of natural additives; 5 NA60 -addition of 6.0 g/animal/day of natural additives. Naturals additives contained clove leaf essential oil, castor and cashew functional oils and a commercial blend composed of vanillin, eugenol and thymol; 6 Standard error of means; 7 Linear effect; 8 Quadratic effect. Table 2 Feeding behavior from young bulls with and without natural additive addition to the diet Animal performance The initial body weight and nal body weight (FBW) were similar for all diets (P > 0.05), nonetheless average daily gain (ADG) of bulls increased linearly (P < 0.01) when the NA blend was added in the diets ( Table 3). The addition of NA in the diets had no effect (P > 0.05) on Dry Matter Intake (DMI) (kg/day -9.9 or kg/100 kg body weight -2.3%). However, feed e ciency improved linearly (P < 0.04) with the addition of NA to the diets (Table 3). In addition, the HCW (Hot Carcass Weight) and HCD (Hot Carcass Dressing) did not differ between cattle fed with blend of NA (P > 0.05; Table 3).  Table 3 Animal performance of young bulls with and without natural additive addition to the diet The addition of a blend of NA affected rumen fermentative characteristics and resultant VFAs produced (Table 4). The major VFAs: acetate, butyrate, isovalerate, and valerate were reduced linearly when animals were fed NA (P < 0.05). Similarly, NA addition in the diets linearly reduced (P < 0.02) the acetate/propionate ratio. NA supplementation of diets resulted in a quadratic effect on propionate and isobutyric acid concentrations (P < 0.05). Furthermore, animals supplied with NA had linear reductions in rumen methane concentration (P < 0.001). Ammonia concentration had a quadratic effect following NA blend supplementation of bull diets (P < 0.001). The ruminal pH was not affected (P > 0.05) by inclusion of NA in the diets (Table 4).  Table 4 Ruminal parameters of young bulls with and without natural additive addition to the diet Rumen bacterial diversity and abundance In our study, the major phyla present in the rumen were Bacteroidetes (47%) and Firmicutes (36%; Figure  1). Bacteroidetes (P < 0.05) were reduced when NA was included in the diet. A quadratic response was seen for Candidatus Saccharibacteria, Chytridiomycota, Elusimicrobia, Eukaryota Unassigned, Fibrobacteres, Firmicutes, Spirochaetes, Synergistetes and Tenericutes (P < 0.05). Source data are included in supplementary material (Table S1).

Methanogen diversity and abundance
Archaeal abundance was reduced on the whole with the inclusion of NA in the bull diets (P < 0.05; Table 5). The families Methanobacteriaceae and Methanomicrobiaceae (P < 0.05); orders Methanomicrobiales, Methanobacteriales and Methanomassiliicoccales (P < 0.05) and the genera Methanobrevibacter and Methanosphaera, showed a signi cant decrease with NA supplementation, whilst the genus Methanomicrobium showed a tendency to be present at lower abundance (P = 0.051). Furthermore, on a species level, a decrease in Methanobrevibacter ruminantium, Methanobrevibacter sp D5 and Methanobrevibacter sp G16 was seen following NA supplementation of bull diets (P < 0.05). ¹CON = control (without natural additives); 2 NA15 -addition of 1.5 g/animal/day of natural additives; 3 NA30 -addition of 3.0 g/animal/ day of natural additives; 4 NA45 -addition of 4.5 g/animal/day of natural additives; 5 NA60 -addition of 6.0 g/animal/day of natural additives. Naturals additives contained clove leaf essential oil, castor and cashew functional oils and a commercial blend composed of vanillin, eugenol and thymol; 6 Standard error of means; 7 Linear effect; 8 Quadratic effect; f_ = family taxonomy, g_ genus taxonomy; o_ = order taxonomy; s_= species taxonomy.

Gene Network correlations
We observed close to 13,000 functionally annotated genes in total across the experimental samples using shotgun metagenomics and 28 were signi cantly differentially abundant when the bull diet contained NA ( Fig. 4; Table S4). Functional annotation data showed signi cantly biological responses due to the NA addition whereas mostly related to protection against foreign attack to DNA and DNA maintenance, replication and repair (Restriction endonuclease, type II, XamI; Restriction endonuclease, type II, EcoRV; Host-nuclease inhibitor protein Gam; RecG, N-terminal antiparallel four helix bundle; Type IV secretion system protein TraG/VirD4; Type IV secretion system, VirB10 / TraB / TrbI and Transposase, ISC1217). There were also functional process associated with membrane protection and maintenance  Fig. 6).

Discussion
In this study we evaluated the mechanism of action of a commercially available blend of essential oil, at increasing concentrations, on the rumen microbiome and host phenotype. Feeding behavior of ruminants is dependent on diet and the environment [21], and as expected, no differences were observed between treatments in this study. On average, animals spent 336 minutes at the feeder, 236 minutes ruminating, 35 minutes drinking water and the remaining at rest. Beef cattle tend to spend an average of 400 minutes eating and 300 ruminating when nished in feedlot [21]. Fiber content is a known factor in uencing time spent ruminating and consequently in water ingestion due to the stimulus on the salivary glands [22]. The observed values in this study provide evidence of a healthy rumen, which is supported by the pH values, which are higher than 6.90 for all treatments. Ornaghi et al. [20], also observed similar feeding behavior when young bulls were fed diets with essential oils and 70:30 concentrate to roughage ratio. Moreover, Zotti et al. [23], fed monensin (included at 30 mg/kg or 40 mg/kg) and functional oils (blend of castor oil and cashew nut shell liquid included at 400 mg/kg) to a high concentrate diet (92.25% concentrate) with 12 steers and observed no effects on feeding behavior parameters.
Essential oils are volatile and odorant compounds which can impact the palatability of the diet, positively or negatively [13], nonetheless we found no effects on DMI in this study. Our results are in agreement with those from Valero et al. [8], whereby bulls fed with 3 g/animal/day of ricinoleic acid (extracted from castor oil seed), anacardic acid, cardanol and cardol (extracted from the cashew nut shell liquid) during nishing had similar DMI (kg/day). On other hand, Yang et al. [24] reported an increase in DMI when cinnamaldehyde (0.4, 0.8 and 1.6 g/day per animal) was fed to feedlot cattle during 28 days of observation. These variations might be related to the differing effects of the essential oils in isolation as opposed to presence in a mixture.
Secondary metabolites extracted from plants often have antimicrobial properties [25,26]. In our study, the main compounds present in the blend were: eugenol, vanillin, thymol, cardol, cardanol, ricinoleic acid, which can modulate the rumen fermentation and reduce methanogens abundance [27]. These compounds may improve the animal performance by modulating rumen fermentation [8,10,20]. Indeed, the ADG and feed e ciency increase linearly when NA were added to the diets. Furthermore, acetate, butyrate, isovaleric, valeric, and ammonia concentration were reduced when NA were added to the diets.
Ornaghi et al. [20], also reported a signi cant increase in ADG using NA (clove essential oil and cinnamon essential oil in two different doses 3.5 and 7.0 g/animal/day) in the diet of young bulls nished in feedlot. However, most studies using NA are in vitro, and in vivo experiments are still scarce in literature. VFAs provide energy for the ruminant maintenance and to produce milk and meat. Nearly 252 kcal are necessary to produce 1 mol of acetate, compared to 62 kcal net gain to produce propionate [28], which also release free hydrogens used to produce methane by archaea (methanogens). We observed a reduction of Acetitomaculum, an important acetogenic bacterial genus, which utilizes monosaccharides to produce acetate, and is often found when cattle are fed high grain diets [29]. We also observed a reduction of the Acidaminococcus genus, which have acetate as major end-product [30]. Reducing the production of acetate can be positive to reduce environmental impact of beef cattle production as more energy is available to the animal as opposed to being lost in the form of methane.
Methanogens are commonly found in association with protozoa [31], which use hydrogenosomes to produce methane. In this study, the use of NA linearly reduced acetate and the archaeal population, that likely reduced methane production suggested by the reduction in archaea abundance. This decrease in the archaeal population post NA supplementation of diets could be due to hydrophobicity of phenolic compounds present in the NA, allowing permeation of the phospholipidic membrane resulting on cell lysis [32; 33]. Khorrami et al. [34] supplemented thyme and cinnamon essential oils (500 mg/kg DM) into ruminant diets and evaluated rumen fermentation and observed decreased protozoal and methanogens abundance, thus corroborating our data. Macheboeuf et al. [35], studied the production of methane in vitro following the inclusion of essential oils from ve plants: Thymus vulgaris, Origanum vulgare, thymol chemo-type of O. vulgare, Cinnamomum verum, and Anethum graveolens); and three pure compounds: thymol, carvacrol, and cinnamaldehyde, and observed a decrease of methanogenesis up to 76% with the highest doses. Patra and Yu [6], also provided evidence for the inhibition of methanogenesis and decreases in protozoal density following addition of ve essential oils from clove, eucalyptus, garlic, origanum and peppermint oils and using three different doses in vitro (0.25, 0.50, and 1.0 g/L).
The effects of the NA blend on propionate production was quadratic and showed the maximum concentration at 4.5 level of natural mix addition. Propionate is the principal precursor of liquid glucose and is related to gluconeogenesis. In addition, production of propionate causes a net gain of around 62 kcal of energy, therefore propionate is bene cial for ruminant production. There was a linear decrease of butyrate following the supplementation of NA to the diet of bull diets. Butyrate can inhibit propionate absorption, therefore is not as bene cial as an energy source for the ruminant [36]. Watanabe et al. [37], observed reduction of butyrate, acetate and methane production when raw cashew nut shell liquid was added to in vitro cultures. It is therefore important to highlight the dose-type dependent effect of the natural additives, which are enhanced when administered as a blend.
NA had a quadratic effect on ruminal ammonia concentration and was higher in bulls fed the control diet compared with those fed NA (21.82 vs 4.78 mg/dL). This lower production may be related to the reduction in hyper-ammonia bacterial abundances, for example the Clostridium genus abundance was signi cant lower compared to the control diet. The Clostridium genus is one of the major ammonia producers and is highly affected by NA [39]. Furthermore, the genus Acidaminococcus and Robinsoniella were linearly reduced. The genus Acidaminococcus produces ammonia as the major end product through glutamate fermentation [30]. The genus Robinsoniella is correlated with high ruminal ammonia concentration and with methanogens, which is due to a re ection of metabolic interaction among microbial consortium [40]. Thus, abundance decreases for both genera could impact the microbial consortium leading to lower methane production. Furthermore, the potential antimicrobial power of NA can be potentiated when the ruminal pH is low as in the grain diets such as in this study [39].
Furthermore, this decrease likely increases absorption of amino acids that are not broken to ammonia, which will be available for absorption in the gut [35]. In contrast, Jesus et al. [41], observed no signi cant effect on ruminal ammonia but an increase in propionate and lower blood urea concentration, suggesting a potential rumen fermentation shift, when a commercial blend (cashew nut shell liquid and castor oil) was fed to dairy cattle, these responses might be related to the animal basal diet. Recently, Cobellis et al. [17], reported that some essential oils can affect VFA production in the rumen but that it is dose and compound dependent, thus, they have speci c effects on the rumen microbiome. As the rumen microbiome present a higher variability, some biological role can interact with the results of this study such as animal effect.
In terms of gene network interactions and function of the rumen microbiome, we found that Glycyl Radical and Peptidase function, were positively correlated to each other. The Ruminococcaceae family undergo changes with the inclusion of NA and had a positivel correlation with the abundance of protein Glycyl Radical genes, which are found to contribute to environmental resilience, and are also potentially related with VFA production [42]. The abundance of Prevotellaceae was negatively correlated with Ruminococcaceae; the two major bacterial families found in our study. Both families are known to compete for the same niche in the rumen [43] perhaps explaining their negative correlations to each other. Blautia tended to increase linearly, even in a low concentration. This taxon can improve polysaccharides utilisation, improving the rumen fermentation [44]. Some Blautia species can consume H 2 increasing the acetogenesis, which can lead to competition with the methanogens [45]. Nonetheless, the Peptococcaceae family was reduced using the blend of NA. This family is a producer of H 2 from amino acids or carbohydrates fermentation. The impact on Blautia genus and Peptococcaceae family might be a secondary cause of the methanogens reduction as the competition for substrates and H 2 lower production can reduce the archaea abundance [46].
There is no doubt that the rumen is a complex environment [47]. Understanding the abundance of the microbes is and their function is nonetheless crucial when investigating the mechanisms of action of a novel additive and to ensure no detrimental effects are encountered. In this study, we show that the essential oil blend used affected the rumen microbiome, potentially through disruption of bacterial cell membranes and breakdown in DNA replication [17,18,26,38]. Important bacterial defense mechanisms used by microbes were observed in our study, such as DNA replication and protection against attack from outsider metabolites, being mostly from membrane sites in response to encountering the blend of essential oils. Furthermore, one of the major protein annotations in our study was the ABC transporter group, the key role of this protein is translocating molecules across the membrane to the maintenance of the cell, followed by multidrug or antimicrobial e ux pumps [30]. This protein was affected and decreased by NA addition. [30]. We also noted some DNA restrictions modi cation mechanisms used for protection of bacterial and archaea against invading foreign DNA were reduced by NA addition, both Restriction endonuclease, type II XamI and EcoRV, to date the difference between them are in the mode of recognition process and cleavage [48].

Conclusions
In our study, the blend of natural additives improved animal performance by bene cial modulating the rumen microbiome. Furthermore, our data suggest that methane emissions may be decreased with NA levels from 3 g/animal/day addition in this study, suggested by the archaeal reduction. Ammonia rconcentrations were also reduced which is also of major bene t for the environment. Also, we can conclude that the level 4.5 g/animal/day in this study had improved animal performance, thus, may replace the use of antibiotics in beef cattle nished in a feedlot with high grain diets. These positive results are mainly a consequence of the ability of the NA blend to bene cially modulate the rumen microbiome.

Animals and diets
A total of 40 (½ Angus vs ½ Nellore) young bulls of 16 ± 2.2 months of age and with a body weight of 385.8 ± 20.7 kg were used in this study. A 14-d adaptation period before starting the experiment was used, during which the concentrate was gradually increased for the animals. The bulls were weighed every 28 days at a trunk balance (Beckehauser Cia. Paranavaí city, Paraná, South Brazil).
Bulls were fed with a basal diet comprised of 70% concentrate and 30% corn silage offered ad libitum for 62 days (Table 1), and feed intake was recorded individually every day for 5% leftovers. Following day 62 in the feedlot, the animals were weighed after 16 hours of fasting (482 ± 31.9 kg) and transported to a commercial slaughterhouse (Campo Mourão city, Paraná, South Brazil). The truck stocking density was 0.8 ± 0.2 bulls/m 2 , and the transport distance was less than 90 km. The bulls were slaughtered following the usual practices of the Brazilian beef industry. The animals were stunned using a captive-bolt pistol. Then, they were bled through exsanguinations by cutting the neck vessels, and the head hide, viscera, tail, legs, diaphragm and excess internal fat were removed. Afterwards, the carcasses were divided medially from the sternum and spine, resulting in two similar halves, which were weighed to calculate the hot carcass weight (HCW). Then, the half-carcasses were washed, weighed, identi ed and stored in a chilling chamber at 4 °C, where they remained for a 24 h period and drip loss measured by the difference between the hot carcass weight and the carcass weight observed 24 hours later after chilling.
The hot carcass dressing (HCD) percentage was de ned as the hot carcass weight divided by the FBW 16 hours before slaughter and calculated by using the equation: HCD = (HCW/FBW) x 100.

Diet chemical analyses
The dry matter (DM) content of the ingredients was determined by oven-drying at 65 °C for 24 h and then drying at 135 °C for 3 h (Method 930.15) [49]. The organic matter (OM) content was calculated as the difference between the DM and ash contents, with ash determined by combustion at 550 °C for 5 h [49]. The N content in the samples was determined by the Kjeldahl method (Method 976.05) [49]. The neutral detergent bre (NDF) and acid detergent bre (ADF) contents were determined using the methods described by Van Soest et al. [21], using heat stable α-amylase and sodium sulphite for the NDF procedure, and residual ash. The factor of 0.82 was used to convert metabolizable energy requirement to digestible energy requirements, and the factor 4.1868 was used to convert total digestible nutrients requirement to megajoules (NRC, 2000).

Feeding behavior
In order to evaluate diet effects on feeding behavior, the young bulls were subjected to two periods of 24 h of observation using ve-minute intervals and three trained evaluators. A total of 288 observations were performed for each animal. Animals were adapted to feeding behavior evaluation for ve days prior to the start of evaluations. Water and feed intake, and rumination and idle periods were obtained by the sum of 288 observations (minute/day). Observations were performed without interfering with the animal's routine. The water intake was considered when animals were at the individual water reservoir, and feed intake was considered when animals were at the feeder. Rumination was considered when animals were chewing a bolus. Idle was considered when animals were not performing any of the activities described previously [50].

Rumen sampling
Fresh rumen content was collected at the end of the experimental period (5 days before the slaughter) 4 h before animals were fed, from 25 animals chosen at random (5 on each treatment). Rumen contents were sampled by a trained veterinarian using an esophageal probe and vacuum pump. Rumen liquor (50 mL) were sampled from the ventral region of the rumen and was then strained through two layers of muslin.
The pH was recorded immediately using a pH meter (Hanna instruments model HI99163, Romaria -Brazil); the electrode was previously calibrated and then inserted into the rumen uid. Sub-samples used to evaluate volatile fatty acids (VFA) and ammonia concentrations were preserved by the addition of trichloroacetic acid (25%; v/v) following storage in ultra-freezer (− 80 °C). Sub-samples used to evaluate protozoal count were preserved using formaldehyde (4%; v/v/).

Rumen microbiome diversity, function and gene network correlations
Taxonomic and functional analysis data were assessed with MGnify (http://www.ebi.ac.uk/metagenomics) following the pipeline version 5.0. Differential abundances of gene functional categories were assessed between dietary treatments using DESeq2 [52]. The input for correlation analysis was performed with the normalized counts taken over all samples from the internal normalization calculated by DESeq2. We applied a P-value cut-off of 0.01 to the resulting domain predictions and counted the number of gene functional which were assigned domains using volcano plots to the differences between control diet and the treatments. Correlations between datasets (biological taxonomy and functional annotation) were calculated using Pearson's rank correlation using R software and visualized with ggplot package. The differences were considered signi cant at Bonferroni corrected p-value < 0.05. After the correlation procedure and p adjusted values the results were used to develop the functional annotation of proteins and biological taxonomy network using standard procedures of the software Cytoscape.

Statistical analyses
In the current study, only microbial taxa with a relative abundance higher than 10 reads were considered and used for the analysis. Bacterial abundance pro les were summarized at phyla, family and genus levels, and archaeal communities were summarized to species level. Relative abundances of microbial taxa were normalized to the lowest reads number for bacteria, and then compared among diet using analysis of variance (ANOVA) and the MIXED procedure to determine the linear and quadratic effects and assess the effects of the treatment control versus blend of NA. All performance data were tested for normality and showed a normal distribution. The data were analyzed using ANOVA and by use of regression equations using the MIXED procedure. In all statistical analyses, the diet was considered a xed effect, and the animals considered a random effect.   Relative abundance of rumen microbiota based on phyla level and taken from young bulls nished in a feedlot and fed with and without natural additives. Sequences that represented < 10% in a sample were combine in others (blue) to aid the visualization.

Figure 2
Relative abundance of rumen microbiota on family level of young bulls nished in feedlot and fed natural additives. Sequences that represented < 10% in a sample were combine in others (blue) to aid the visualization.
Page 27/30 Figure 3 Relative abundance of rumen microbiota on a genera level and taken from young bulls nished in feedlot and fed with and without natural additives. Sequences that represented < 1.5% in a sample were combine in others (blue) to aid the visualization.

Figure 4
Volcano plot of rumen microbial genes following shotgun metagenomic sequencing of samples obtained from young bulls nished in the feedlot and fed with and without natural additives. Black dots represent non-significantly differentially expressed proteins, green dots represent proteins significantly differentially expressed at pFDR < 0.05 while red dots represent the most significantly differentially expressed proteins; A -Control diet versus Na15 (addition of 1.5 g/animal/day of natural additives), B -Control diet versus Na30 (addition of 3.0 g/animal/day of natural additives), C -Control diet versus Na45 (addition of 4.5 g/animal/day of natural additives), D -Control diet versus Na60 (addition of 6.0 g/animal/day of natural additives).

Figure 5
Gene network correlation between rumen diversity and gene functional annotation P < 0.05; light blue nodes) and biological taxonomy family abundance (pink nodes) of young bulls nished in feedlot and fed natural additives. The nodes size is related to the number of directed edges. Green lines are positive correlation (r2 = > 0.5) and red lines negative correlation (r2 = < -0.5). Family taxonomy abundance with signi cant effect between treatments (P < 0.05; red nodes)