Scrophularia striata was purchased from a local herbalist’s shop in Ilam, a province in the west of Iran and was authenticated as S. striata Bioss family (Scrophulariaceae) with a voucher specimen being preserved in the pharmacy herbarium (TEH No: 6916) at Tehran University of Medical Science, Tehran, Iran. Whole plants containing stem and seed were ground in a centrifugal mill with 0.5 mm screen. Ground plant material was mixed with 600 ml L^-1 aqueous ethanol solution at the ratio 1:25 (w:v) for 1 h in sonicator bath with cooling system (Sonorex RK 156H, Bandelin, Berlin, Germany). The mixture was filtered through filter paper and concentrated in a vacuum at 40°C by a rotary evaporator (Buchi, Flawil, consisting of water bath B-480, Rotovapor R-124, and Vacuum Controller B-72, Switzerland). After ethanol evaporation, the particulate was transferred to beaker flask, frozen at -20°C and then freeze-dried according to previously described methods (Goel et al., 2008).

The experiment was designed as a completely randomized design with 2 experimental runs, 4 treatments, and 3 replicates per run, resulting in six independent measurements per treatment. The treatments consisted of either no additives (control), 0.83 mg g^-1 DM of substrate monensin sodium salt (monensin; Sigma Chemical Co. Ltd.), 40 mg g^-1 DM (S. striata-Low), or 80 mg g^-1 DM (S. striata-High) of dried of S. striata extract. Monensin and S. striata were dissolved in 1 ml ethanol [90% (v/v), TechniSolv^®, pure] and an equal volume of ethanol was added to the control treatment group (Cardozo et al., 2005).

All runs were comprised of 12 fermenters and lasted for 10 days, with sampling on the last 5 days of each run based on a 4 day adaptation period for the rumen microbes (Soliva and Hess, 2007). Initial inocula for each experimental run was collected from three non-lactating rumen-fistulated Holstein cows, housed at the Dairy Research Station of the University of Veterinary Medicine Vienna (Pottenstein, Austria). The use of donor cows was approved by the institutional ethics committee of the University of Veterinary Medicnie (Vetmeduni). Donor cattle were fed hay and grass silage ad libitum and free access to drinking water. Animals were cared for according to the Austrian guidelines for animal welfare (Federal Ministry of Health, 2004). Rumen contents of the three cows were mixed in equal parts, the liquor was first filtered through four layers of medical gauze (1-mm pore size) before adding to the fermenters and the solid digesta was filled (∼60 g) into nylon bags and placed into each fermenter on the first day of each experimental run according to the previously published methodology (Klevenhusen et al., 2015). In brief, each fermenter was filled with 100 ml artificial saliva and 600 ml rumen fluid on the first day of each experimental run (McDougall, 1948). New substrate was provided to each fermenter daily, based on a 50:50 forage-to-concentrate dairy ration (Supplementary Table S1) using nylon bags (140 mm × 70 mm, 150-μm pore size, Linker Industrie–Technik, Kassel, Germany). After 24 h of incubation, the bags containing solid digesta were replaced with fresh bags containing the experimental diet. Accordingly, every feed bag was incubated for 48 h, and then replaced. Artificial saliva buffer was continuously infused into each vessel (357 ± 19.2 ml/d) through a 12 channel peristaltic pump (Model ISM 932D, Ismatec, Index Health and science GmbH, Wertheim, Germany) in accordance with Soliva and Hess (2007). Feed additives were pipetted into the fermenter fluid every day at the time of feeding (0900 h) starting on day 0. After each feeding procedure the system was flushed with N₂ gas for 3 min to maintain anaerobic condition inside the vessels. Daily effluent and fermentation gasses were collected in a 1 L volumetric flask located in an ice-filled box and reusable gas-tight aluminum bags (TECOBAG 8L, Tesseraux ContainerGmbH, Burstadt, Germany), respectively. On sampling days, directly before feed bag exchange, fermenter fluid samples were collected using a syringe equipped with a plastic tube. Part of the fluid samples were immediately analyzed for rumen fermentation parameters and aliquots were stored in separate tubes at -20°C for SCFA and ammonia measurements. Additional fluid samples were snap frozen in liquid nitrogen and stored at -80°C for DNA extraction. The feed residue bags were hand-washed with running water until the water was clear and kept at -20°C for chemical composition analysis.

Immediately after sampling pH and redox potential (seven Multi^TM, Mettler-Toledo GmbH, Schwerzenbach, Switzerland) of the rumen fermenter liquid were measured. Fermentation gasses were also measured every 24 h during the sampling period. Gas volume was determined using the water displacement technique (Soliva and Hess, 2007), and the concentration of CH₄ was measured using a portable infrared detector (ATEX biogas Monitor Check BM 2000, Ansyco, Karlsruhe, Germany). In brief, the biogas monitor is attached to the gas collection bag after a 24 h collection period, the sample is measured through pumping of gas into the monitor for 30 s to provide a consistent reading as per the manufacturer’s instructions.

Concentration and composition of SCFA (acetate, propionate, n-butyrate, isobutyrate, n-valerate, isovalerate, and caproate) were analyzed by GC, as described previously (Qumar et al., 2016). Briefly, incubation fluid samples were thawed and then centrifuged at 20,000 × g for 25 min at 4°C. The supernatant (0.6 mL) was transferred into a fresh tube and 0.2 mL of HCl (1.8 mol L^-1) was added, followed by 0.2 mL of the internal standard (4-methylvaleric acid, Sigma-Aldrich Co. LLC., St. Louis, MI, United States). After centrifugation of the mixture at 20,000 × g for 25 min at 4°C, the clear supernatant was transferred into the GC vial. Analysis of SCFA concentrations was conducted via a gas chromatography apparatus (GC Model 8060 MS 172 DPFC, No.: 950713, Fisons, Rodano, Italy) which was equipped with a flame-ionization detector and a 30 m × 0.53 mm ID × 0.53 μm df capillary column (Trace TR Wax, Thermo Fisher Scientific, Waltham, MA, United States). Injector and detector were at temperatures of 170°C and 190°C, respectively. Helium was used as carrier gas with a flow rate of 6 mL min^-1. Stratos Software (Stratos Version 4.5.0.0, Polymer Laboratories, Church Stretton, Shropshire, United Kingdom) was utilized for generation and evaluation of chromatograms.

The concentration of ammonia was determined using the indophenol reaction (Weatherburn, 1967). The fermenter fluid samples were thawed at room temperature and subsequently centrifuged at 15,115 × g for 10 min. The clear supernatant was diluted with distilled water to obtain a concentration within the standard calibration curve. Ammonia and phenol were oxidized by sodium hydroxide in the presence of sodium nitroprusside and dichloroisocyanuric acid. The absorbance was measured at 655 nm using U3000 Spectrophotometer (INULA GmbH, Vienna, Austria) after 90 min of reaction.

Analysis of the residual feeds after 48 h of incubation and the original feed samples were conducted in duplicate. Prior to analysis, samples were oven-dried at 65°C for 48 h and ground using through a 0.75 mm sieve (VDLUFA, 2007). The DM of feeds and feed residues was measured after oven drying at 104°C for 24 h and ash was analyzed by combustion of samples over night at 580°C. Crude protein was measured by Kjeldahl method (VDLUFA, 2007). Neutral detergent fiber (aNDFom; Udén et al., 2005) and acid detergent fiber (ADFom) contents were analyzed using Fibretherm FT12 (C. Gerhardt GMbH and Co. KG, Konigswinter, Germany) based on VDLUFA (2007). Heat stable α-amylase used in the NDF procedure and all fiber fractions were expressed exclusive residual ash. Disappearance of nutrients as a measure of degradability was calculated from the difference between content of nutrients in the diet before and after 48 h incubation in the Rusitec system.

Frozen rumen fluid samples were thawed on ice, vortexed shortly to mix the sample, and then samples from d 5 to 10 were pooled per fermenter per run (n = 6 per treatment). From the pooled sample, 1 mL subsample was used for DNA isolation. DNA isolation was performed using the QIAamp Fast DNA Stool Mini Kit (QIAGEN, Hilden, Germany) and samples were mixed with 1 mL of InhibitEX buffer provided in the kit and heated at 95°C for 5 min to ensure proper lysis of bacteria. After heating, additional enzymatic (mutanolysin, 2.5 U μL^-1; lysozym, 100 mg mL^-1; proteinase K, 20 mg mL^-1) and mechanical lysis (0.4 g sterile ceramic beads, diameter 1.4 mm) using a bead-beater (FastPrep-24 5G, MP Biomedicals, LLC, Santa Ana, CA, United States) was performed to disrupt bacterial cells according to previously published procedures (Kong et al., 2010). This was followed by chemical removal of cell debris and PCR inhibitors by centrifugation and supernatants were transferred to fresh tubes for column based isolation of total genomic DNA using the QIAcube robotic workstation. Total genomic DNA was eluted in 200 μL of low-salt buffer. The isolated DNA concentration was determined by a Qubit 2.0 Fluorometer (Life Technologies, Carlsbad, CA, United States) using the Qubit dsDNA HS Assay Kit (Life Technologies) and stored at -20°C until further analysis.

Using qPCR, the abundance of Protozoa, Bacteria, Archaea, and Fungi present in Rusitec samples were analyzed on a Stratagene Mx3000P real-time PCR System (Agilent Technologies, Santa Clara, CA, United States) using established primer pairs (Table 1). DNA samples and negative controls were assayed in duplicate in a 20 μL reaction mixture containing 10 μL the Fast-Plus EvaGreen Master Mix with Low ROX (Biotium, Hayward, CA, United States), 1 μL of each primer (Protozoa 400 nM; Bacteria 100 nM; Archaea and Fungi 200 nM), 7 μL of nuclease-free water and 1 μL DNA template (4 ng). The amplification program included an initial denaturation step at 95°C for 10 min, followed by 40 cycles of 95°C for 10 s, primer annealing at 60°C for 30 s, and elongation at 72°C for 30 s. Fluorescence was measured at the last step of each cycle. Melting curve analysis was performed to determine the specificity of the amplification. The dissociation of PCR products was monitored by slow heating with an increment of 0.1°C s^-1 from 55 to 95°C, with fluorescence measurement at 0.1°C intervals. Standard curves for all primers were generated using 10-fold serial dilutions (10⁸–10³ molecules μL^-1) of the purified and quantified PCR products generated by standard PCR using DNA from the pooled rumen fluid from both runs of the present experiment. The PCR efficiency was calculated: E = 10^(-1/slope)-1 (Table 1). Results were analyzed using the associated software (Stratagene MxPro, QPCR Software, version 2.00).

One 40 μL aliquot of each DNA sample was sent for amplicon sequencing using a MiSeq Illumina sequencing platform and paired-end technology (Microsynth AG, Balach, Switzerland). Sequencing targeted the V3–V5 hypervariable region of the 16S rRNA gene using the primer set 357F-HMP (5′-CCTACGGGAGGCAGCAG-3′) and 926R-HMP (5′-CCGTCAATTCMTTTRAGT-3′) to produce an amplicon size of ∼523 bp (Peterson et al., 2009). In brief, libraries were constructed by ligating sequencing adapters and indices onto purified PCR products (Nextera XT Sample Preparation Kit, Illumina, San Diego, CA, United States) according to the recommendations of the manufacturer. Equimolar amounts of each of the libraries were pooled together and sequenced on an Illumina MiSeq Personal Sequencer and the resulting paired ends were stitched together by Microsynth (Balach, Switzerland) resulting in a total of 3.8 million unfiltered reads with an average of 503 nt in length. Sequence quality control was performed using the QIIME pipeline (Caporaso et al., 2010). Sequences were first quality filtered following (Q = 20; 27) and then screened and filtered for chimeras using USEARCH and its database (USEARCH v8.1; Bokulich et al., 2013). Clustering and alignment were performed using UCLUST (Edgar, 2010) and OTUs were defined using PyNAST (Caporaso et al., 2010) and the SILVA database (v123; accessed July 19, 2017; Quast et al., 2012; Yilmaz et al., 2013). The degree of similarity between sequences was defined as 97% to obtain OTU. Any OTUs which clustered with less than 10 reads were manually removed. A total of 969,155 sequences (mean sequences per sample 38,504) clustered into 2,727 OTUs for further analysis. All OTUs with a relative abundance greater than 0.1% (135 OTUs) were taken in consideration for statistical analysis and statistically derived means for relative abundance were graphed in excel. For calculation of the non-parametric species richness estimators Chao1 and the observed OTUs per sample were used. Beta diversity analysis was performed using weighted Unifrac dissimilarity metrics and the principal coordinate analysis (PCoA) plotting in QIIME with rarefaction at 20,133 sequences, based on the lowest number of sequences in a sample. Sequencing data are available in BioProject SRA database under the accession number SRP140677.

Analysis of variance was conducted using the PROC MIXED procedure of SAS (Version 9.2, SAS Institute, Cary, NC, United States). Treatments were considered as fixed effects, while experimental run and fermenter within run were defined as random effects. Degrees of freedom estimation was done using Kenward-Rogers. All values were reported as least squares means. Comparison between the control and treatments was carried with Dunnett–Hsu. Differences were declared as significant at P ≤ 0.05. Correlation analysis using Pearsons correlation coefficient was performed using the PROC CORR procedure of SAS. Following Hinkle et al. (2003), the r was interpreted as follows: 0.00–0.30 as negligible; 0.30–0.50 as low; 0.50–0.70 as moderate; 0.70–0.90 as high, and 0.90–1.00 as substantial.

The pH in the fermenters remained constant throughout both experimental runs (Supplementary Figure S1 and Table 2). However, redox potential in the fermenters was lower for all treatments when compared to the control (P = 0.007). Analysis showed that the ammonia concentration decreased (P < 0.001) with monensin and S. striata supplementation. The extent of ammonia reduction was 35, 26, and 52%, respectively, by monensin, S. striata-Low and S. striata-High treatments. Total SCFA production was decreased by monensin, while S. striata supplementation increased it compared with control (P = 0.001). All individual SCFA were affected by supplementation (Table 2). Monensin decreased acetate production by 9%, whereas S. striata had no effect in comparison with control. Propionate was increased by 186, 83, and 35% by feeding monensin, S. striata-High and S. striata-Low, respectively. As a result, monensin and S. striata-High had a lower acetate to propionate ratio (A:P) in comparison to control (P < 0.001). Supplementation decreased iso-butyrate concentrations in comparison to control (P < 0.001). Butyrate was decreased by monensin and S. striata-High (P < 0.001) in comparison to control. In contrast, the addition of S. striata-Low increased the concentration of iso-valerate above that of the control group (Table 2). Monensin decreased the proportion of iso-valerate to 8.1% of the control concentration and increased of n-valerate by 97% (P < 0.001). Supplementation with monensin, S. striata-Low and S. striata-High caused a drop in methane production by 67, 29, and 36%, respectively (P < 0.003). The DM, OM, and ADF degradation were not affected by supplementation (Table 2). However, S. striata-High decreased CP degradation (P = 0.03) and all additives decreased NDF degradability (P = 0.02) compared to the control group.

Results of qPCR analysis are presented in Figure 1. Supplementation did not affect the abundance of total Bacteria and Fungi populations compared to control. Monensin remarkably lowered the gene copy number of Protozoa (1.00 × 10²), though it was significant only when compared to S. striata-High group (3.15 × 10⁵; P = 0.05). Copy numbers of total Archaea were lower with S. striata-High and monensin compared to S. striata-Low (P = 0.001), and tended to be lower with monensin compared to control (P = 0.08).

Statistical analysis of both the intra- (alpha diversity; Table 3) and inter- (beta diversity; Figure 2) sample diversity was performed. Despite no differences in the total numbers of sequences found within a sample (P = 0.91), there were differences between treatments in the number of OTUs (P < 0.001) and the Chao1 index (P < 0.001). For both indices, the diversity within a sample was highest in the control, followed by S. striata-Low, S. striata-High, and monensin showed the least microbial diversity. PCoA showed a clear clustering of microbial communities for each treatment group. Principle component 1 described 40.8% of the diversity between samples between original rumen fluid and experimental treatments, with the monensin treatment clustering independently (Figure 2). Principle component 2 represented 30.2% of the diversity with a separation of the S. striata samples from the control and monensin groups. Principle component 3, representing 10.3% of the variation between samples, was related to the original pooled rumen fluid sample.

In total, 15 phyla, predominated by Bacteroidetes, Firmicutes, and Proteobacteria, could be identified into 42 genera. All phyla and all genera were significantly affected by treatments (Figure 3). Supplementation with S. striata-High increased the total Bacteroidetes and reduced the Firmicutes taxa, whereas monensin increased Fusobacteria, Synergistetes, and Proteobacteria phyla (Figure 3A). The addition of all additives decreased the relative abundance of Elusimicrobia and TM7 in comparison to control. Fusobacterium and Firmicutes were, respectively, 4 and 13% more abundant in the monensin in comparison to the S. striata-High supplementation (Figure 3). At the genera level, Prevotella was the most abundant genus ranging from 38.5% in control to 46.1% in S. striata-High and was significantly increased with S. striata supplementation. Bacteroidetes-associated genera represented the largest proportion of the in vitro microbiota despite the fact that samples were taken from the rumen liquid (Figure 3B). Firmicutes had the largest diversity in identified genera with a significant effect of supplementation (Figure 3C). A drastic increase in Fusobacterium spp. was found in monensin supplementation compared to the other treatments (Figure 3D). Direct comparison of the monensin and S. striata-High treatments at the phyla level for significant differences in the least square means showed of the 15 phyla, eight were significantly different between the two treatment groups (Figure 4). Of the 30 most abundant OTUs, 29 were significantly affected by the supplementation treatment. Bacteroidetes and Firmicutes accounted for 16 and 8 of the OTUs, respectively (Figure 5). Three OTUs were identified to the species level, Prevotella copri-like OTU2, Bacteroides uniformis-like OTU10 and Fibrobacter succinogenes-like OTU24. In comparison to the control, OTU2 was 4.2-fold and four-fold increased in monensin and S. striata-High, respectively (P < 0.001). OTU10 accounted for 5.13% of the relative microbial abundance in the monensin supplemented group but was less than 0.6% of the abundance in the other treatments. F. succinogenes-like OTU24 was increased in both the S. striata-Low (1.05%) and S. striata-High (0.87%) supplemented groups compared to both control and monensin.

Several correlations between rumen fermentation parameters and microbial communities significantly altered by the supplementations were found (Table 4). Fusobacterium spp. showed the greatest number of highly (0.70 < r < 0.90) negative correlations, including methane, acetate, butyrate, isovalerate and total SCFA production (P < 0.001) and positive correlations to propionate (P < 0.001) and valerate (P < 0.001). Similarly, Bacteroides (phyla Bacteroidetes) and Succinivibro (phyla Proteobacteria) genera also showed an inverse relationship to methane production (r = -0.76 and r = -0.71, respectively) and a positive correlation to valerate concentrations in vitro (P < 0.001). Furthermore, Bacteroides spp. was also highly correlated with decreases in butyrate and isovalerate, similar to Fusobacterium spp. Succinivibrio showed similar correlations as Bacteroides (Table 4). The strongest correlation was that of Fibrobacter spp. with total SCFA concentrations (r = 0.86; P < 0.001), which was supported by additional positive correlations with acetate and isovalerate concentrations. Lachnospira spp. showed a highly negative correlation to ammonia concentrations (P < 0.001).