A shift towards succinate‐producing Prevotella in the ruminal microbiome challenged with monensin

The time‐resolved impact of monensin on the active rumen microbiome was studied in a rumen‐simulating technique (Rusitec) with metaproteomic and metabolomic approaches. Monensin treatment caused a decreased fibre degradation potential that was observed by the reduced abundance of proteins assigned to fibrolytic bacteria and glycoside hydrolases, sugar transporters and carbohydrate metabolism. Decreased proteolytic activities resulted in reduced amounts of ammonium as well as branched‐chain fatty acids. The family Prevotellaceae exhibited increased resilience in the presence of monensin, with a switch of the metabolism from acetate to succinate production. Prevotella species harbour a membrane‐bound electron transfer complex, which drives the reduction of fumarate to succinate, which is the substrate for propionate production in the rumen habitat. Besides the increased succinate production, a concomitant depletion of methane concentration was observed upon monensin exposure. Our study demonstrates that Prevotella sp. shifts its metabolism successfully in response to monensin exposure and Prevotellaceae represents the key bacterial family stabilizing the rumen microbiota during exposure to monensin.

and preventive interventions [8,9].Monensin is used to treat coccidiosis in broilers and ketosis, acidosis and acute bovine pulmonary oedema and emphysema in cattle [1,2,10,11].In the rumen, monensin inhibits acetate formation which results in a lower amount of hydrogen [12].Since hydrogen is needed for methanogenesis, the formation of methane is impaired [13,14].Monensin preferentially complexes and transports Na + across lipid membranes, hereby dissipating electrochemical sodium gradients of bacterial membranes.It is most effective against Gram-positive bacteria [7].The rumen microbiome is a highly diverse ecosystem and consists of several microbial groups covering a broad metabolic capacity [15].Prevotella is one of the major genera in the rumen and it was found to be resilient towards monensin, with Prevotella bryantii B 1 4 exhibiting the highest resilience [16].Former in vivo and in vitro studies of the rumen microbiome treated with monensin showed reduced methane and increased propionate and succinate formation [17,18].Also, a reduced proteolytic activity was observed by Wallace et al. [18].As a consequence, cattle fattening induced by monensin was explained by an increased hepatic gluconeogenesis originating from the elevated propionate levels [4,19].
However, the detailed metabolic routes affected by monensin treatment and the molecular mechanisms leading to resilience of Prevotella sp.against monensin remained unclear.
In the past, effects of monensin were studied using the Rusitec or similar in vitro systems focusing on a narrow range of specific microbial groups and metabolites [20][21][22][23][24]. Former Rusitec experiments revealed adaptation phases between 2 and 13 days, with a tendency closer to 2 days, after which the treatment test with an established microbial consortium was started [25].In the present study, the adaptation period was adjusted to 1 week and the use of a multi-omics approach revealed a novel mechanism for the resilience against monensin in Prevotella: the dissipation of the Na + -gradient by monensin is counteracted by the increased activity of a Na + -translocating membrane protein complex which reduces fumarate to succinate.The implications of this catabolic route on the cellular and metabolic functions of the ruminal bacteria challenged with monensin are highlighted in the present study.

Diet and treatment preparation
The total mixed ratio (TMR) included 40% dried maize silage, 20% dried grass silage, 20% soybean meal extract and 20% dried corn grains (dried for 48 h at 60 • C and percentage refers to weight).All ingredients were milled through a sieve of 0.5 mm (SM1, Retsch) and mixed for 15 min with the M200 (GLP, Gebrüder Lödige Paderborn).Nylon bags (100 × 50 mm, pore size = 0.1 mm) were filled with 15 ± 0.01 g of TMR and were sealed with a cable tie.Filled bags were stored in dry and dark conditions at 8 • C until further usage.During the treatment phase, 1 ml of a monensin stem solution (280 mg monensin/45 ml ethanol) was added to the TMR in nylon bags to gain in a final concentration of 415 mg monensin/kg TMR.The control treatment included the TMR only with ethanol.Ethanol was evaporated in all bags before they were placed in the reactor vessel.The buffer was prepared 1 day before usage [26].

Inoculation and Rusitec design
Four lactating Jersey cows were fed ad libitum with straw and TMR, with same composition as described before, for at least 1 week before rumen fluid sampling.Sampling was done through the rumen fistula the buffer [20], was introduced into the vessel by a peristaltic pump (IPC, ISMATEC) with a flow of approximately 31.2 ml/h, resulting in a 75% volume exchange within 1 day.Produced gas and superfluous liquid eluted via the overflow into a cooled gas trap (ca.0 • C) separating liquid and gas phase.The liquid phase was directed into waste containers and discarded, while the produced gas amount was channelled to the gas meter (BlueVCount, BlueSens), which exhausted the measured gas into a gas proof bag (Plastigas, Linde).Total gas emission was measured online from day 1 until day 17.The technical setup is illustrated in Figure 1A.

Parameter measurement and sampling
Daily temperature measurements via a rod thermometer (TFA Dostmann) were required to maintain and mimic the optimal rumen temperature.The volume of the overflow in the waste vessels was measured daily to evaluate the continuous buffer supply.Total gas production was determined by the gas meters (BlueVCount, BlueSens).
Methane concentration of accumulated gas was daily determined by channelling the accumulated gas through the Advanced Gasmitter During the first 2 days of treatment and elution phase, samples for metabolome analyses were taken in an interval of 12 instead of 24 h.

Experimental setup
The experiment was divided into three phases starting with: adaptation, treatment and elution.Three reaction vessels were set up for either monensin presence (MON) or absence (CON).In the adaptation phase (first 7 days), ruminal microorganisms adapted primarily to the in vitro conditions without monensin.Monensin was supplemented to the MON group only during the treatment phase (days 7-14).Elution phase was defined as days 15-18 (Figure 1B).

Gas slope calculations
Gas production behaviour was monitored by calculating the linear slope by hour, using measurements of every minute starting from day 3 to 17. Slope calculation was performed by using Excel (Microsoft Office, v.2016).

Protein extraction and purification
Solid sample material from nylon bags were retrieved from three reactors per treatment group (CON, MON) at selected time points (7, 8, 11, Pellets were stored at -20 • C until protein extraction.Protein extraction and purification was performed as described by Trautmann et al.
[28] and modified by adapting the amount of precipitated protein to 80 μg.
Default settings were used to obtain label-free quantification (LFQ) values of proteins.Protein identification was based on a freely available rumen metagenomic database [31].Functional annotation was conducted using EggNOG [32] and the predetermined labels from Stewart et al. [31].LFQ values of proteins with the same KEGG orthology ID were summed up for KEGG pathway analysis of the total quantified microbiome for each time point.Comparative analysis of protein numbers was always performed by subtracting the number of quantified proteins of the monensin group from the control group (CON -MON).For comparative analysis with quantitative protein measurements, LFQ ratios between monensin versus control group, were used to display a fold-change within the ratio (MON/CON).Pearson correlation was performed with relative abundance values in Excel (Microsoft Office, v.2016).Proteins aligning in number and abundance with treatment duration and monensin concentration underwent a closer investigation.For KEGG mapper analysis, at least three of five time points required to be elevated in CON or MON in terms of summed LFQ values and number of proteins for a certain KEGG orthology ID.Additionally, a similar pattern as the monensin concentration and the experimental phases were elucidated for further investigation and demonstration.

Data analysis
Principal coordinate analysis (PCoA) was performed using a Bray-Curtis similarity matrix with standardized LFQ values in PRIMER 6 & PERMANOVA+ (v.1.0.6).Differential analysis in number were always calculated by subtracting the number of quantified peptides or proteins of the monensin group from the control group.LFQ-values from the monensin group were divided by the control group for each time point and formed a ratio for quantitative analysis.COG classes were additionally analyzed using the number of proteins from CON minus the number of MON, which shows information about differences of the absolute quantity.Statistical analysis were conducted using the least significant difference (LSD) test in Infostat (v.2008) [33].

Gas and methane emission
The gas production rate and the methane concentration of the monensin reactors decreased immediately after the first monensin supply (Figure S1a, b).A significant difference in methane concentration between monensin reactors (MON) and control reactors (CON) was detected (p < 0.05, Figure S1b, c).The difference in methane concentration remained until the end of the elution phase (17 days), where methane concentrations increased slightly in MON.The total amount of daily gas production was fluctuating between mean volumes of 600 and 1000 ml/d during the experiment in all vessels (Figure S1c).

Fermentation and metabolites
The pH value of the medium fluctuated around 6.9 and was hardly affected by the treatment (Figure S2).An overall increase of fermentation products was measured at time point 7.5 d in MON, when monensin had already been introduced for 12 h (Figure 2).S3).During the first 24 h of the monensin supply, multiple compounds including low concentrated metabolites decreased (Figure S4).In the elution phase, propionate and valerate increased significantly (p < 0.05) in MON (Figure 2).Low concentrated fermentation products such as formate and succinate showed fluctuations in MON along the complete experimental period (Figure S4).A declining trend in concentrations of isobutyrate (Figure 3A) and isovalerate (Figure 3B) was shown during the treatment phase in MON, which aligned with the ammonium and amine concentration (Figure 3C).These proteolytic side products, especially ammonia, decreased significantly at days 11, 14, and 17 from about 3.5 mM to 1.5-2.0mM (Figure 3C).

Monensin in reactor effluent
Relative monensin quantifications showed a linear increase of monensin in the reactor fluid starting from day 8 (Figure S5).The maximum monensin concentration was most likely achieved after the last monensin supplementation between days 13 and 14.An approximate monensin elution of 200% per day can be calculated when monensin supplementation was ceased after days 13-17.

Metaproteome is altered by monensin exposure
In total 9159 protein groups were identified, and out of this 7149 protein groups were quantified with LFQ values at the time points 0, 7, 8, 11, 14, and 17 days.A peptide identification rate of 19% (total peptides found: 28,419) was achieved with the database from Stewart et al. [31].
A core proteome of 2532 protein groups were found among all time points and both treatment groups (Table S1).Measurements at day 0 were excluded for direct comparison between both groups.The distribution of metaproteomic datasets of the reactor samples based on time points and groups is depicted in Figure 4. Two sample clusters with 80% Bray-Curtis similarity could be identified.The first cluster, including three samples from MON group, was characterized by an increase of proteins related to membrane biogenesis, inorganic transport and metabolism and intracellular trafficking.Among all detected proteins, about 32% were assigned to Prevotella, showing the dominance of this genus in the active community.Proteins affiliated to Selenomonas (6.2%), Lachnospiraceae (4.4%) and Treponema (3.9%) followed with a descending average contribution (Table S2).
LFQ values were used to calculate the ratio of the metaproteomes between MON and CON at respective time points.Bacterial taxons identified on single protein groups were divided into monensin fast and slow-responding groups.Slow responders were defined by a correlation coefficient (R 2 ) >0.6 with the relative monensin concentration and MON/CON ratio for taxons that had a contribution of at least eleven peptides which was set as a threshold for this dataset (Table S3).Fast responding taxa changed their ratio exponentially from day 7 compared to the mean of the residual time points with a greater magnitude (≥2-fold).Microorganisms which decreased remained diminished until the end of the elution phase in MON (Figure 5).Proteobacteria divided in γ-Proteobacteria, Aeromonadales and Succinatimonas contributed with at least seven specific peptides.Eubacterium belonged also to the fast-responding microorganisms, which declined with a smaller extent after monensin exposure.Proteins related to other taxa showed similar trends in a slow and fast responding manner but they were lacking a sufficient number of representing peptides.

Prevotella species are less affected by monensin
The highest number of quantified proteins with the used database from Stewart et al. [31] belonged to the dominant genus Prevotella (Table S2) and proteins of seven Prevotella species were identified with at least ten species-specific peptides (Figure S6).The unambiguous species assignment of the peptides was checked by a BLAST search against NCBInr to avoid discrepancies due to the limits of the used database.Prevotella species with most unique peptides were P. multisaccharivorax, Prevotella ruminicola, and P. bryantii.Latter two species were two-to three-fold more abundant in MON from day 11 on.Peptides assigned to P. multisaccharivorax were more abundant in CON, while at day 11 and 17 they were elevated in MON.Most proteins of Prevotella species showed a gradual elevation to its maximum at day 17, when monensin was absent for 3 days.In CON, a decline of Prevotella related proteins was detected, except for proteins from P. ruminicola and P. copri (Figure S6).Linear trends in MON were only found for proteins of Prevotella dentalis and P. copri.Proteins of P. brevis (formerly P. ruminicola 23 [34]) increased about 3-fold in the MON at day 11 and decreased at day 14 and 17.Proteins of P. bryantii also increased in MON at day 11 about 2-fold and decreased proportional in both groups over time.Proteins of Prevotella sp.S7-1-8 remained mostly unchanged until day 17, where an approximate 2.8-fold decline was seen in CON.These predominant Prevotella species were represented by proteins from COG classes for carbohydrate metabolism (G), translation (J), membrane and cell wall biogenesis (M) and inorganic transport (P).Although Prevotella proteins were dominant here, it must be clarified that these results were generated with the currently most suitable database for ruminants.This may lack the detection of rare microbes as the number of proteins affiliated with single taxa is rather low in the database.

Functional metaproteome influenced by monensin
The variation of the metaproteomic data and the evolution of the bacterial functions among the control and monensin incubations over time were also detectable by the analyses of the COG classes (Figure 6).Proteins belonging to COGs of intracellular trafficking and secretion (U), cell wall/membrane/envelope biogenesis (M) and  6).The LFQ ratio between MON and CON was slightly lower in proteins assigned to COG class G of the carbohydrate metabolism and transport (Figure 6).The alternate unweighted perspective showed a tremendous depletion of the total number of these proteins in MON from day 11 to 17 (Figure S7).Deeper insight was provided via KEGG pathway analysis of the quantified proteins resulting in 1215 unique KEGG orthologues (KO).Major changes were detected in the starch and sucrose metabolism (ko00500), the amino acid metabolism (ko01230) and the carbon metabolism (ko01200).KO that appeared in quantitative differences in either the CON or MON were shown in the KEGG mapper for the carbon metabolism (Figure S8).Several enzymatic reactions belonging to glycolysis, amino acid biosynthesis and fatty acid metabolisms were mainly elevated in CON during the experiment (Figure S8).An elevated number and quantity of proteins belonging to the TCA cycle (Figure 7A) was found in MON at day 11, 14 and 17.The main fermentation product of the TCA cycle is succinate, which on average was always slightly higher in MON (Figure 7B).Elevated enzymes of the TCA cycle and especially the quinol:fumarate reductase (QFR) originated mainly from Prevotella species (Table S4).
Elevated enzyme activity of the fumarate reductase was also found in a monensin supplemented Rusitec study pointing to a prevalence of less monensin-sensitive succinate producers [18].
A decrease of multiple proteins that were involved in tryptophan biosynthesis was identified in MON with a minimum at day 14 (Figure 3E).Proteins assigned to amino acid metabolism showed a similar trend to the depleted products of proteolysis (Figure 3B,D).
Microorganisms such as P. ruminicola, Lactobacillus mucosae, Mitsuokella jalaludinii, Treponema sp.JC4, P. dentalis correlated with proteolytic side products in MON with an absolute average ≥ 0.6 and a minimum of 10 peptides per taxon (Table S3).

Monensin leads to reduced fibre digesting and assimilating capabilities of the microbiome
The downstream functional analyses revealed 1401 proteins with one or more carbohydrate-active enzyme (CAZyme) IDs out of

CAZymes (Table S1).
A gradual increase of proteins belonging to inorganic transport and metabolism (COG: P) was identified in MON over time (Figure 6, Figure S7).This general change was observed also on the individual level of iron-transporting proteins.COG P included also ATP-binding proteins from multiple sugar transport system (K10112), such as maltose/maltodextrin transport system permease proteins (K10109) during monensin exposition (Table S1).

Proteolytic activities influenced by monensin
Diminished ammonium and branched-chain fatty acid (BCFA) levels indicated a reduced proteolytic activity in the monensin-supplemented reactors (MON).Species such as P. bryantii and P. ruminicola were found to be hyper-ammonia producing and have been associated with protein degradation in the rumen [35,36].Despite Prevotella is known for their prominent proteolytic activity [37], correlations with proteolytic side products were found with a sufficient amount of peptide evidence (at least 11 unique peptides/time point) in P. ruminicola and P. dentalis under MON (Table S3).Correlations with proteolytic products showed that individual species either prefer BCFA or ammonia formation.A decreased diversity and quantity of proteins assigned to amino acid metabolism was found in MON, pointing towards a hampered tryptophan metabolism in Prevotella species and demonstrating the intermediate step to prevent acute bovine pulmonary edema by monensin [1].The lack of BCFAs inhibited most likely the formation of branched-chain amino acid (BCAA) due to the structure similarity of BCAA to isovalerate and isobutyrate [28].We, therefore, agree with the statement of Wallace et al. [38] that a smaller part of high-active rumen microorganisms are most likely responsible for the proteolysis [38].Proteolytic activity may also be uncoupled from the abundance of bacterial species but regulated by the availability of cofactors, such as NADH.However further research is required to verify the major key players for proteolysis in rumen microbiome.

Reduced fibre degradation in the presence of monensin
Functional changes in the metaproteome were induced by the first monensin supplementation at time point 8 d and continued for at least 24 h.The elevation of proteins assigned to translation and transcription was often seen in proteomic studies with antibiotic supplementation or in microorganisms under stressful conditions [39].
The large drop in proteins for carbohydrate metabolism (COG G) gave rise to suspicion, that carbohydrate degradation is reduced in the variety of degradation as shown by Poos et al. [40] who observed a diminished dry matter digestibility via faecal analysis after monensin supplementation in lambs [40].Haimoud et al. (1995) documented similar findings for ruminal fibre digestion under monensin in cows but total fibre digestion remained unchanged when compared to control [41].The current study underlined findings of reduced fibre digestion by a drop of bacterial glycoside hydrolases GH1, GH3, GH51, GH53, GH78 and GH95, which contribute to plant fibre deconstruction [42].
Unfortunately, the contribution of CAZymes expressed by protozoa or fungi were excluded in this analysis as the used database lacks the F I G U R E 5 Time course of relative peptide abundances from Rusitec consortia with or without monensin treatment reveal slow and fast responding microorganisms.Slow responding microbes appeared with a linear relation to monensin concentration, wherefor R 2 ≥ 0.6 was used.Fast responding microbes had an exponential relation to monensin concentration, which is why fold-change ≥ 2 between ratio of time point 7 d and the average of rations from 8 to 17 d was used.More abundance in MON is displayed in red while for CON in blue.Taxon assignation was indicated with the minimal and maximal number of contributed peptides over time in brackets.Ratio was calculated by dividing the relative abundance of the MON by CON.Trends for MON (red) and the CON (blue) either increased ( ), decreased ( ) or remained unchanged over time (missing arrow) respective protein entries.The decline of polysaccharide degradation appears to be connected to changes of the monomer uptake.Results indicated a reduced presence of ATP-binding protein of the multiple sugar transport systems that enforces the perspective of a diminished sugar uptake too, which was also seen under monoculture conditions [43].

Monensin affects fermentation activities
Proteins contributing to glycolysis were also lowered, meaning that either a switch to hibernation or an alternative energy generation via anaerobic respiration is made as described for P. bryantii B 1 4 [44].
Reduced glycolysis induced by monensin was also described in studies with P. bryantii [43] and Streptococcus faecalis [45], which reveal a reduction of the whole ruminal carbohydrate degenerative system.
Analyses of the fermentation products showed significantly enhanced concentrations of succinate and propionate at late MON incubations.
This can be directly linked to the protein abundances of the respective metabolic pathways.An incorporation of propionate into valerate can explain the similar trend at the elution phase since both are linear organic acids with an odd number of carbon atoms [28].Protein abundance of the succinate-producing genus Succinatimonas [46] showed that this bacterium belonged to the fast-responders, which decreased almost 5-fold, as soon as monensin was introduced (8 d).Thus, the formation of succinate has to be linked to other succinate-producers as the level of succinate was constant and even increased during the treatment phase in MON.Similar SCFA ratios were observed by Kaplan Shabtai et al. [47] showing that the succinate-dependent propionate production is associated with Succinivibrionaceae, Bulleidia and Prevotella [47].Prevotella appears to be the main candidate since Succinivibrionaceae were depleted in the present study and proteins for Bulleidia were not identified which is due the limited amount of only six proteins in the database.
In vitro and in vivo studies also demonstrated monensin tolerance and resistance of several Prevotella species [6,48,49], which largely contributed to the ruminal microbiome under monensin supplementation [50].This was verified by the predominance of protein Treatment ___EluƟon___ F I G U R E 6 Shift in relative abundance of annotated proteins in solid phase of Rusitec consortia with (MON) or without monensin (CON) treatment over time.A higher relative abundance in MON is displayed in red while for CON in blue.Summed label-free quantification values of clusters of orthologous groups (COGs) from control and monensin group were divided (MON/CON) and resulted in a ratio, which is displayed in the boxes.If in one group no proteins were detected (e.g., in CON), the opposite group (e.g., MON) was displayed in the cell instead of the ratio.Right column indicates total amount of proteins detected.Colour code of heat map is illustrated at the bottom of the differential and the relational approach.PTM in the COG description stands for post-translational modifications.One protein can belong to multiple COGs, which leads to a higher number sum of total counts than proteins are quantified respiratory supercomplex which oxidizes NADH and reduces fumarate under formation of an electrochemical Na + -gradient [44].Furthermore, less sugar is assimilated in Prevotella under monensin [43], which point to an increased metabolic shift to succinate production in order to maintain its concentration.By shifting its metabolism from acetate to succinate production [16], a conversion from succinate into propionate becomes very likely and beneficial for maintaining the monensin-disrupted sodium gradient [43].Previous studies and current findings also indicated an increase of propionate during monensin supplementation [51], whereas a significant elevated propionate concentration was only determined after monensin supplementation was stopped.Defence mechanisms seem to require an adjustment time longer than 24 h and non-increasing monensin concentrations.
P. ruminicola was shown to produce elevated propionate levels under monensin [16] and propionate production was only occurring in presence of the cofactor vitamin B 12 [52].An inhibition of vitamin B 12 synthesis by monensin appears to be less likely since the microbialborne vitamin B 12 concentration in the blood serum of monensin treated lambs remained unchanged [53].Some species of the genus Pre-votella seem to use the niche created by monensin and act as the main trader for metabolites like acetate and succinate [54].This perspective is supported by the high abundance of Prevotella, the elevated amount of succinate [16,54] and the detected increase of succinate forming pathways under monensin in the present study.
Further ruminal propionate producers are Rikenellaceae and Ruminococcaceae [50], which showed a decreasing quantity with progressing time after the initial day of treatment.Some Succinivibrionaceae species were shown to be resistant towards monensin [55], but the current findings demonstrated a fast-responding sensitivity towards monensin.The presented data showed a decrease of Butyrivibrio sp.specific proteins in MON, which may have caused an inhibition of the propionate consumer Butyrivibrio fibrosolvens leading to a propionate enrichment by monensin.Schären et al. (2017) assumed that monensin is leading rather to an increase of the substrate succinate than affecting the propionate producers [50].According to this, Prevotella species are assumed to be the main mediator for the increase of propionate during the elution phase, where the dwindling antibiotic pressure enabled countermeasures of Prevotella against

Treatment ___EluƟon___
F I G U R E 8 Monensin exposure of the microbial consortium affects fibrolytic CAZyme families.More abundance in MON is displayed in red while for CON group in blue.Ratio was calculated by dividing the LFQ-value of MON from the CON.CAZyme families were assigned to their targets of plant fibre degradation monensin.Furthermore, a negative correlation trend in MON for succinate and Prevotellaceae was found (R = -0.94),while succinate correlated positively with Lachnospiraceae and Megasphaera (R = 0.94).
Those two conspicuous trends align with the production of propionate by Megasphaera, which belongs to the family Veillonellaceae [56], and butyrate by a certain Lachnospiraceae [57].The positive relationship between propionate with Lachnospira and Shuttleworthia described by Xue et al. was rediscovered in the correlation analysis of the current study and verified "that some species belonging to this genus may positively interact with propionate-producing taxa" [58].Another propionate producer is Acidaminococcaceae [59], which together with Veillonellaceae is correlating (R ≥ 0.8) with propionate and succinate resulting in the responsive families for propionate production under monensin exposition.Species from the family Veillonellaceae were shown to synthetize propionate out of succinate and lactate [60].
Assuming that the lactate pathway was inhibited due to monensin, the production of propionate was mainly carried out by succinate.
However, the exact process of succinate production by Prevotella and the subsequent assimilation via Megasphaera and Veillonellaceae species over time has to be shown in a co-cultivation experiment with monensin.Veillonellaceae appeared to be insensitive towards monensin up to concentrations, which were exceeded in this experiment [55].The intermediate sampling points 14.5 d and 15.5 d showed an elevation of succinate in MON that is probably due to the monensin binding or elution after the supplementation.
Methane production was reduced in presence of monensin, which was already demonstrated in previous studies [18,51].The study of Wallace et al. (1981) indicated a drop in methane production during continuous monensin administration independently of the applied dosage [18].Their highest added dosage of monensin per day was similar to the applied dosage of the current study (6.2 mg/d).The decline of methane was associated with reduced formation of acetate, increased formation of propionate and an elevated presence of the genus Prevotella [61].Those findings align with the observation of Prevotella being a key player in the rumen which is responsible for the metabolic shift by redirecting reducing equivalents from CO 2 reduction (and methane formation) to fumarate reduction (and succinate production).This paves the way for increased propionate formation from succinate [44].In summary, our study demonstrates that the ionophoric effect of monensin induces the modulation of metabolic routes in rumen bacteria without causing a major phylogenetic reconstruction.

Accumulation and elution of monensin
In a previous study, we assumed that monensin is bound and accumulated in extracellular polysaccharides (EPS) but at a certain point monensin is released due to a deconstruction of EPS via CAZymes [43].
Together with a periodical supplementation of monensin, this situation results in a prolonged effect of monensin, meaning that the rumen microorganisms require more time to recover despite monensin supplementation was stopped.The current data does not provide a full explanation and illustration of the monensin accumulation and elution in a ruminal in vitro system, but it shows that there is an undiscovered field of how long drugs remain in the gastro-intestinal tract before being washed out.The consequences for a prolonged retention of monensin are difficult to assess, however, current data showed that a reduction or even eradication of certain low-abundant microbes is a probable outcome.

ASSOCIATED DATA
The mass spectrometry proteomics data have been deposited to the ProteomeXchange Consortium via the PRIDE [62] partner repository with the dataset identifier PXD034099.
using a tube connected to a vacuum pump to sample rumen fluid, rumen solid material was sampled manually.Rumen fluid from all cows was Significance of the study Monensin is known to inhibit the growth of bacteria by dissipating electrochemical sodium gradients of bacterial membranes.The use of monensin as a therapeutic agent leads to structural and functional changes of the rumen microbiome and a concomitant shift in the availability of fermentation products for the host.The mechanistic details are investigated in the present study revealing a resilience mechanism of members of Prevotellaceae which is prominent family in the rumen microbiome.The observed shift of the fermentation metabolism from acetate to succinate enables Prevotella to survive the monensin treatment and to stabilize the rumen functionality.combined and filtered through two layers of cheesecloth at a temperature of 39 • C. Six thermostatic reaction vessel (Gassner Glastechnik GmbH) were run in parallel, each was filled with 500 ml rumen fluid and 500 ml of buffer with same composition as in Wischer et al. [20].Two nylon bags were placed into the agitating cage inside the filled vessel.One bag contained TMR and was replaced by a fresh bag every 48 h.The second nylon bag contained an equal amount of ruminal solid phase from each cow (totally -change 60 g) and was replaced after 24 h by a fresh TMR bag.The change of the TMR bag of every 24 h simulates the feed intake of the ruminant and additionally the fibreattached microbes are allowed to colonize the newly introduced solid matter.The vessel content was kept at 39 • C ± 0.5 • C and mixed by agitating the feed container up-and downwards (12 rpm) to simulate the peristaltic movement of the rumen.Artificial saliva, represented by

F I G U R E 1
Experimental setup.(A) Technical setup of the rumen simulation technique (Rusitec) with an indication of sampling material and gas measurement, including methane.(B) Time points of sampling for respective analysis are marked by a square (□).Striped squares indicate that total mixed ratio was sampled as solid material for the respective analysis.Daily addition of monensin in the treatment phase is indicated by black diamonds above the phase description (♦) (Pronova).Liquid phase of the reaction vessel (18 ml) was retrieved via a syringe for metabolic analysis.A FiveEasy pH meter (Mettler Toledo) was used to determine pH within a sample tube directly after the collection.TMR was sampled after 48 h of incubation from the nylon bag to investigate solid-phase adherent microbes with proteomic analyses.Sampling and daily TMR exchange was always carried out under constant CO 2 gassing on the exposed reactor content to maintain anaerobic conditions.Samples were transferred directly to ice before storage at -80 • C. Time points for sampling are illustrated in Figure 1B.

30 min, 4 •
C) and the soluble part was transferred into an NMR-tube (d = 5 mm). 1 H-NMR was performed subsequently with a 600 MHz spectrometer (Avance IIIHD, Bruker Corporation) equipped with a Cryoprobe Prodigy BBO.1D 1 H-NMR experiments based on the first increment of a 2D NOESY pulse sequence (1D NOESY) with 16 scans, 100 ms mixing time, 12 ppm sweep width and 4 s acquisition time (about 57k data points) were run at room temperature with Topspin software (Version 3.6.1,Bruker Corporation).Processing (phase correction and baseline correction, manually performed) and metabolite quantification was conducted by using the Chenomx NMR suit (Chenomx Inc., v.8.6).

14, and 17
days), while time point 0 days was a unique sample (fresh rumen material).These single solid samples of each treatment group were pooled for each time point to a total mass of 4 g solid sample material and incubated with 35 ml methylcellulose buffer (200 mM NaCl, 50 mM Tris-HCl, 0.1% (w/v) methylcellulose, pH 8) under 30 min rotation at 4 • C with subsequent vigorous mixing.The samples were treated in an ultrasonic bath (RT = 20 ± 1 • C) for 1 min and vacuum filtrated through a 2-layered sterile cheesecloth via a strainer and washed twice with 25 ml of rinsing buffer (200 mM NaCl, 50 mM Tris-HCl, pH 8).Samples were centrifuged (200 × g, 10 min, 4 • C), and the resulting supernatants were centrifuged once again (10,000 × g, 15 min, 4 • C).Obtained pellets were washed twice with 5 ml wash buffer (50 mM Tris-HCl, 0.1 g/L chloramphenicol, 1 mM PMSF, pH 7.5).
Nano-LC-ESI-MS/MS experiments were performed on an EASY-nLC 1200 (Thermo Fisher Scientific, Germany) coupled to a Q-Exactive HF mass spectrometer (Thermo Fisher Scientific, Germany) equipped using a NanosprayFlex source (Thermo Fisher Scientific, Germany) in the University of Hohenheim, Stuttgart, Germany.Tryptic digested peptides were injected to a NanoEase analytical column (NanoEase M/Z HSS C18 T3, 1.8 μm 100 Å 75 μm × 250 mm column, Waters GmbH, Germany) at a constant temperature of 35 • C by a PRSO-V2 column oven (Sonation GmbH, German) and a flow of 250 nl/min.A gradient with solvent A (0.1% formic acid) and B (0.1% formic acid and 80% acetonitrile) was run for 140 min (0-115-130-140 min) with increasing solvent B (1-50-56-95%).The Orbitrap of the Q-Exactive HF enabled a resolution of 60,000 with a range from 200 to 2000 m/z and a TopN of 20.Tandem MS spectra were generated for the 20 most abundant peptide precursors using high-energy collision dissociation (HCD) at a resolution of 15,000, normalized by a collision energy of 27.Lock-mass ions from the ambient air were used for internal calibration [29].

F I G U R E 2
Average metabolite difference between control (CON, blue) and monensin (MON, red) group from Rusitec supernatant determined by 1H-NMR.Displayed are the SCFAs: acetate, propionate, butyrate and valerate.Error bars indicate standard deviation (n = 3) while asterisk (*) indicates a significant difference between both groups (LSD-Fisher, p < 0.05)Meaning that slow responders seem to react corresponding to the amount of monensin in the Rusitec.A gradual decrease over time was seen for proteins affiliated to Actinobacteria in MON, which was verified on deeper taxonomical ranks in the genus Bifidobacterium (Figure5).Linear trends in MON | CON (↘|↗) showed the effect of monensin on the amount of Bifidobacteriaceae related proteins, which were contributing about 0.3% to the total LFQ-value at family level.Other slow responding families were Acidaminococcaceae, increased slightly within MON, and Ruminococcaceae which increased in CON.

F I G U R E 3
Monensin treatment results in a decrease in proteins involved in proteolysis and tryptophan biosynthesis.Mean proteolysis indicators (n = 3) with standard deviations and asterisks (*) as significant differences (LSD-Fisher; p < 0.05) between CON (blue) and MON (red) were shown like (A) Isobutyrate (B) Isovalerate and (C) Ammonia and amines.(D) Phenylalanine, tyrosine and tryptophan biosynthesis KEGG map (ko00350) with blue marked enzymes elevated in quantity (LFQ-values) and number in CON (blue) during treatment and elution phase.(E) Ratio of summed LFQ-values based on EC corresponding to KEGG map ko00350 inorganic transport (P) were quantitatively predominant in MON (Figure

4
Principal coordinate plot of proteins obtained from solid samples retrieved from Rusitec consortia with or without monensin treatment.7149 quantified proteins were standardized by total underwent a resemblance by Bray Curtis similarity.Dashed line indicates sample clustering with 80% similarity.The legend displays the CON (blue) and MON (red) conditions with different symbols for each time point.Adaptation phase (7 d), treatment phase (8, 11, 14 d) and post-treatment phase (17 d).Selected cluster of orthologous groups (COG) are indicated as an elevated (↑) unique feature for certain samples and clusters.COG: membrane biogenesis (M); inorganic transport and metabolism (P); intracellular trafficking and secretion (U) 135 unique CAZyme families.Glycoside hydrolase (GH) 16 was the largest CAZyme family including 19 identified proteins, aligned with a decrease of the quantitative ratio over time (R = 0.73) and were mostly represented by Prevotella species.Changes of protein abundance of CAZymes contributing to plant fibre degradation were listed in Figure 8. Day 14 marked the lowest ration for CAZyme families in MON among all time points, indicating a minimum of GH abundance.Carbohydrate esterases (CE) 1, CE12 and GH16 increased in MON over time, whereas most of the GHs decreased along the treatment (Figure 8).Most of the selectively presented CAZymes were targeting cellulose β-1,4/1,3 glucan linkages (n = 36), followed by pectin linkages (n = 42), β-1,4 arabinoxylan (n = 27) and β-1,4 xyloglucan (n = 8).P. ruminicola were majorly assigned to 21 proteins from GH35, followed by Prevotella sp.ne3005 with 11 proteins.The species with the most diverse set of CAZymes was Prevotella multisaccharivorax DSM 17128 with 28 different CAZymes families found in the present metaproteomics dataset followed by Selenomonas bovis with 19 diverse counts from genus Prevotella in the present study.Prevotella was able to consolidate its ecological niche within the monensin environment and was able to utilize the provided nutrients for acetate and succinate biosynthesis.In P. bryantii B 1 4, succinate is the product of a E 7 Enhanced succinate production in Rusitec consortium exposed to monensin.(A) Highest protein number and abundance in group.Proteins of groups (CON, MON) exceeding number and summed quantity were displayed per time points.In case of a dash number or summed quantity was lacking.Red reaction arrows point out reactions elevated mainly in the monensin group.KEGG orthology numbers are indicated above the boxes.Enzyme commission number is located close to the reaction arrows.(B) Mean succinate concentrations with error bars standard deviations and asterisk as indicator for significance (p < 0.05, LSD-Fisher)