Initial effects of supplemental forages and feedstuffs on bovine rumen ecology in vitro as determined by DNA-based molecular procedures

ABSTRACT

This research aided in determining the impacts on rumen microbial ecology when supplemental forages and feedstuffs were added for 48 h after an initial 24-h in vitro rumen fermentation of orchardgrass (Dactylis glomerata L.) hay. Short-term shifts in bovine rumen community structure (bacteria, archea, protozoa, and fungi) resulting from each separate forage and feedstuff addition were measured using unique operational taxonomic units (OTU)s contained in terminal-restriction fragment length polymorphism (T-RFLP) profiles. The Tukey Vacuum Cleaner Analysis (TVCA) model of bacteria accounted for 66.3% of the treatment variance and OTUs clustered into 3 distinct groupings. The model developed for TVCA analysis of archaea accounted for 76.6% of the treatment variance and OTUs clustered into 2 distinct groupings. The TVCA treatment variance of protozoa accounted for 75.8% and OTUs clustered into 3 distinct groupings. Additions of various forages and feedstuffs did not result in shifts in fungal community structure in the short-term experiment. Our results demonstrate that the rumen ecology using initial T-RFLP profiles in vitro of bacteria, archaea, and protozoa and a TVCA analysis can identify pattern groupings of forages and feedstuffs. Groupings of forages and feedstuffs can help refine supplementation for improved nutritional management of ruminants.

KEYWORDS:

• Forages
• bovine
• rumen in vitro
• rumen ecology
• T-RFLP
• VFA

Introduction

A diverse array of forages including herbages (grasses), forbs (broadleaf legumes, weeds, and herbs), and browse is used in ruminant livestock production systems (Turner and Belesky 2010). Supplements including energy and protein-rich forages and feedstuffs are used to meet dietary requirements and improve nutrient-use efficiency in ruminants such as cattle (Bos sp), sheep (Ovis aries), and goats (Capra hircus). Energy supplements include corn (Zea mays L.) and various cereal grains. Protein supplements include high crude protein (CP) legumes, such as alfalfa (Medicago sativa L.), and high protein feedstuffs and by-product feeds, such as whole cottonseed (Gossypium hirsutum L.). Some forages contain secondary plant metabolites that can influence grazing behavior, selectivity, feed intake (Acamovic and Brooker 2005), and rumen microbial populations (Wallace et al. 2002). Many of these metabolites are being re-evaluated to identify opportunities to influence ruminant performance through improved feed intake, digestion, nutrient-use efficiency, and health (Rochfort et al. 2008). Improved nutrient-use efficiency in grazing livestock can help to reduce methane emissions, enhance carbon sequestration, and reduce fertilizer inputs in grassland agriculture (Abberton et al. 2007).

The rumen of cattle contains a complex mixture of anaerobic bacteria, archea, protozoa, and fungi that responds to short- and long-term changes in diet. High starch feedstuffs can shift the ruminal microbial population from fiber-degrading to a starch-degrading microbial population. Excessive amounts or sudden diet changes with limited rumen microbial population adaptation to high starch diets is detrimental to maintaining ruminant health (i.e. rumen acidosis). Usually when supplementing feedstuffs high in starch, the supplement is generally limited to 0.25 - 0.5% of body weight (BW) to maintain a diverse microbial ecosystem for sustained fiber utilization by the animal and improved forage intake and overall performance (Horn and McCollum 1987; Bowman and Sanson 1996). Protein supplementation is usually required when forage chemical composition contains < 8% CP. Both supplementation strategies are aimed at maintaining a desirable energy (as total digestible nutrients; TDN) to protein ratio in the diet (TDN:CP ratio ∼7.0; Moore et al. 1999) to maximize microbial protein synthesis, especially from bacteria, in the rumen for improved nutrient-use efficiency and ultimately animal performance.

Metagenomics involves the analysis of natural assemblies of organisms based on the DNA signature (Handelsman 2004). These techniques can include high throughput sequencing platforms (next generation sequencing; NGS), but there is evidence that terminal-restriction fragment length polymorphism (T-RFLP) data is comparable (de la Fuente et al. 2014). Specifically, the T-RFLP fingerprint technique provides a snapshot of rumen microbiota and can be used to compare relative differences between or among treatments (de la Fuente et al. 2014; Fang et al. 2018). This information can be used to better understand how rumen communities adapt to diet and dietary supplements (Firkins et al. 2008).

The rumen fluid from two different beef steers was mixed together and used in this experiment. Our goal was not to evaluate or relate the rumen microbiome of each animal, but to use an initial rumen fluid inoculum in vitro that was representative of an orchardgrass forage-based diet. Our aim was to profile the modifying effects from different supplemental forages or feedstuffs on the resultant rumen community pattern in vitro to help group feedstuffs with similar fermentation patterns. Our specific objective was to use T-RFLP fingerprinting to profile bovine rumen community structure 48 h after addition of a supplemental forage or feedstuff to a 24-h basal cool-season grass hay fermentation in vitro.

Materials and methods

Rumen fluid and sample feedstuffs

All applicable international, national, and/or institutional guidelines for the care and use of animals were followed. All animal care and use procedures were reviewed and approved by the Institutional Animal Care and Use Committee, Appalachian Farming Systems Research Center, Beaver, WV, USA. Rumen fluid donor steers were cared for in accordance with the standards of the Guide for the Care and Use of Agricultural Animals in Agricultural Research and Teaching (FASS 2010).

Rumen fluid was collected and combined from two ruminally cannulated beef steers and transported to the lab in closed containers. A grab sample of particulate material was also included with the raw rumen fluid (initial pH∼6.8) to account for the particulate-adhering microbes. Rumen fluid, McDougall’s (1948) buffer solution and in vitro rumen fermentation incubations were maintained under anaerobic conditions at 39°C throughout the experiment. Carbon dioxide (CO₂) was bubbled through the rumen fluid to minimize oxygen and to agitate the rumen fluid containing particulate material as to help release microbes adhering to fiber in order to create a more heterogenous rumen fluid inoculum. The rumen in vitro dry matter disappearance (IVDMD) of forage and feedstuffs was determined by using the Tilley and Terry (1963) procedure as modified by Moore (1970). All tubes contained 0.6 g (dry weight basis) orchardgrass (Dactylis glomerata L.; OGR) hay and were inoculated with 50 ml of 2:3 mixture of rumen fluid to McDougall’s buffer on d 1, fitted with a stopper containing a one-way valve to vent excess gas, and incubated in a water bath maintained at 39°C. All tubes were incubated for 24 h to establish the initial baseline fermentation. This helped to create an OGR baseline fermentation with an established short-term rumen ecological community. A BLANK was included which contained rumen fluid plus buffer solution (no substrate).

In order to model the offering of supplemental feedstuffs to ruminants already being fed OGR hay, supplemental feedstuffs (dried plant materials and energy feeds; 0.40 g on a dry weight basis) listed in Table 1 were added to tubes on h 24 to maintain a 1g-total amount (dry weight basis) of substrate in each tube at a ratio of 60:40 (OGR: supplemental feedstuff ratio). Tubes were incubated for an additional 48 h and were swirled while maintaining anaerobicity 3 times per d. Each treatment (supplemental forages, feedstuffs, and blank) was represented by four parallel replicate tubes (each replicated set contained all treatments plus a blank).

Table 1. Dried plant samples, hays, and feedstuffs used in experiment.

Supplemental Feedstuff^a	Abbreviation	Scientific Name
Orchardgrass	OGR	Dactylis glomerata L.
Corn grain	CORN	Zea maize L.
Whole cottonseed	WCS	Gossypium hirsutum L.
Alfalfa	ALF	Medicago sativa L.
Red clover	RCL	Trifolium repens L.
Crimson clover	CCL	Trifolium incarnatum L.
Peanut hay	PNT	Arachis hypogaea L.
Rhizoma peanut	RZP	Arachis glabrata Benth.
Birdsfoot trefoil	BFT	Lotus corniculatus L.
Sericea lespedeza	SLZ	Lespedeza cuneata (Dum. Cours.) G. Don
Chicory	CHIC	Cichorium intybus L.
Chicory cv. Puna	CPUNA	Cichorium intybus L.
Chicory cv. Forage Feast	CFEAST	Cichorium intybus L.
Artemisia (leaves)	ART	Artemisia annua L.

^acv = cultivar.

After the supplemental forage or feedstuff was added and the additional 48 h of incubation completed, each 100-ml tube was removed from the water bath, thoroughly mixed via vortex for 20 sec, contents filtered through 2-layers of cheesecloth in a funnel, and the liquid captured in a 50-ml screw-top centrifuge tube. Tubes were centrifuged using a Jouan KR22i centrifuge with AK500-11 fixed angle rotor with inserts for 50-ml conical, screw-cap centrifuge tubes spun at 9400 × g for 10 min at 5°C to isolate the microbial mass. The vortexing of tubes prior to filtering helped to dislodge any fiber adhering microbes for a more homogenous sample.

Following centrifugation, supernatant was decanted into a 20-ml vial for volatile fatty acid (VFA) analysis; the remainder of the supernatant was decanted into a 60-ml Nalgene bottle maintained on ice for determination of pH using a hydrogen ion electrode (Fisher Accumet Basic-AB15; Thermo Fisher Scientific, Hampton, NH, USA) with calibrated probe and recorded within 1 h after collection.

All vials and bottles were placed on ice until moved and stored in the cooler (3^o C) or freezer (−20°C). Centrifuge tubes containing microbial pellets were frozen at −20°C until genomic DNA (gDNA) was isolated.

Composition of forages and feedstuffs

Forages and feedstuffs used in the experiment were ground to pass a 1-mm screen in a Wiley mill or a 0.5-mm screen in a cyclone mill. The 1-mm ground subsamples of feedstuffs were analyzed for dry matter (DM) and ash (AOAC 1990); neutral detergent fiber (NDF; procedure included amylase; final values contained ash residues) and acid detergent fiber (ADF; final values contained ash residues) (Goering and Van Soest 1970; Van Soest et al. 1991) using a Fiber Analyzer (Ankom Technology Corp., Fairport, NY, USA); and IVOMD (Tilley and Terry 1963; Moore 1970). A set of 0.5-mm ground subsamples were analyzed for total N using an elemental analyzer (Flash EA, 1112 Series, Thermo Electron Corp., Rodano, MI, Italy) and CP calculated as total N% × 6.25. Total non-structural carbohydrates (TNC) were determined by an automated hydrolysis method (Denison et al. 1990).

The supernatant samples were thawed, a one ml-subsample was centrifuged at 10,000 × g for 10 min, then the supernatant was removed and mixed with metaphosphoric acid, cooled on ice, and centrifuged at 10,000 × g for 10 min (Goetsch and Galyean 1993). The VFA concentrations were then quantified via gas liquid chromatography (Hewlett-Packard GC 6890, Hewlett-Packard Co., Wilmington, DE, USA) using a Supelco glass column (Supelco/Sigma-Aldrich Inc., Bellefonte, PA, USA) with packing material as outlined in the 1988 Supelco Bulletin 856B for rumen fluid VFA.

gDNA extraction

Samples were thawed and mixed thoroughly after which gDNA was extracted from a 250-mg subsample using a PowerSoil DNA isolation kit (MoBio Laboratories Inc., Carlsbad, CA, USA) (Johnson et al. 2009) following the manufacturer’s instructions with the exception that the final volume eluted was 60 μL. The gDNA concentration in all samples was determined using a CARY 100 UV/VIS spectrophotometer (Mulgrave, Australia) after which samples were stored at −20°C until amplification via polymerase chain reaction (PCR).

PCR amplification for T-RFLP analysis

Genomic DNA from bacteria, archaea, protozoa, and fungi were each amplified separately using nested primer sets. A nested PCR approach is necessary when copy numbers of the gDNA of interest are low or in regions difficult to amplify due to non-specific binding of primers to amplicons requiring secondary PCR amplification to obtain sufficient amplicons for analysis. The secondary primer set is designed to target a smaller region within the amplicon obtained from the initial PCR reaction. All primer pairs were previously determined to provide coverage of organisms present in rumen fluid (see individual primer references for details). In this experiment, the second amplification was performed using 1 μL of the first PCR product as template. The forward primer for the second PCR reaction was labeled on the 5′ end with HEX (hexachloro-fluorescein) and FAM (5-carboxy-fluorescein) fluorescent markers [1.5 μL (10 μM)] for terminal fragment tracking. The PCR reactions containing gDNA and primary PCR amplicons were amplified using a GeneAmp9600 Thermocycler (Perkin Elmer, Norwalk, CT, USA). The thermocycler conditions used for amplification were the same irrespective of the template organism or PCR master mix. Conditions included an initial 2-min denaturing step at 94° C, followed by 40 cycles of a 92° C-denaturing step for 45 sec, a 55° C-annealing step for 1 min, a 72° C-extension step for 2 min, and a final 10-min extension step at 72° C.

Master mix for bacteria

The 16S ribosomal DNA (rDNA) was amplified from the isolated gDNA samples using a set of universal bacterial-primers that target 16S rDNA primers, 8F (5′-AGA GTT TGA TCM TGG CTC AG-3′) and 1492R (5′-GGT TAC CTT GTT ACG ACT T-3′) (Johnson et al. 2009). The gDNA sample was amplified in a 50-μL PCR reaction that combined 50–200 ng of gDNA, 25 μL of AccuPrime SuperMixII (Invitrogen, Carlsbad, CA, USA), 1.5 μL (10 μM) of reverse primer, and 1.5 μL (10 μM) of forward primer, and PCR grade water sufficient to bring the final volume to 50 μL. Bacterial re-amplification employed nested primer sets 530F (HEX 5′-GTC CCA GCM GCC GCG G-3′) (Winker and Woese C 1991) and 1392R (FAM 5′-ACG GGC GGT GTG TRC-3′) (Scully et al. 2005).

Master mix for archea

Initial 16S rDNA amplicons were obtained from gDNA samples using the archaeal-specific primers Ar109F (5′-ACK GCT CAG TAA CAC GT-3′) (Grosskopf et al. 1998) and AR1383R (5′-CGG TGT GTG CAA GGA GCA-3′) (Shlimon et al. 2004). The second reaction used to amplify the amplicon from the above reaction contained primers, Ar349F (HEX 5′-GYG CAS CAG KCG MGA AW-3′) (Hristov et al. 2012) and Ar912F (5′-CTC CCC CGC CAA TTC CTT TA-3′) (Lueders and Friedrich 2000). The forward primer was labeled on the 5′ end with a HEX fluorescent marker for terminal fragment tracking.

Master mix for protozoa

Procedures as outlined above for bacterial amplification were used. The initial 16S rDNA amplicon was amplified from the isolated gDNA samples using the protozoal-specific primer set Euk82F (5′-AAA CTG CGA ATG GCT C-3′) and MedlinB (5′-TGA TCC TTC TGC AGC AGG TTC ACC TAC-3′) (Skillman et al. 2006). The second amplicon was obtained using the forward primer Euk342F (HEX 5′-CTT TCG ATG GTA GTG TAT TGG ACT AC-3′) (Egan 2005) for terminal fragment tracking and reverse primer MedlinB (5′-TGA TCC TTC TGC AGC AGG TTC ACC TAC-3′).

Master mix for fungi

Initial 18S rDNA from fungi were amplified from isolated gDNA samples using primers NS1F (5′-GTA GTC ATA TGC TTG TCT C-3′) (Fliegerova et al. 2006) and SR6R (5′-TGT TAC GAC TTT TAC TT-3′) (Vilgalys and Hester 1990). The second amplification used a nested set of primers NS19F (HEX 5′-CCG GAG AAG GAG CCT GAG AAA C-3′) and NS6R (5′-GCA TCA CAG ACC TGT TAT TGC CTC-3′) (White et al. 1990; James et al. 2000) the forward primer was labeled on the 5′ end with a HEX fluorescent marker for terminal fragment tracking and used.

Enzymatic digestion

All PCR products were cleaned using a Qiagen PCR Purification Kit (Qiagen, Valencia, CA, USA). The concentrations and purity of each individual PCR product were verified using A₂₆₀/A₂₈₀ ratios measured on a spectrophotometer. Each tube contained 3 μg of re-amplified PCR reaction, 1.5 μL of MspI restriction enzyme, cuts at recognition site C^CGG (Invitrogen, Carlsbad, CA), 5 μL of buffer, 1 μL BSA, and enough PCR-grade water to bring final volume to 50 μL. The restriction enzyme was selected to optimize genus-specific terminal restriction fragments (T-RF) associated with phylogenetically informative species. Digestion reactions were incubated overnight in a 37° C water bath and heat inactivated by immersing in a 60° C water bath for 10 min. Digested amplicons were desalted using AmiconUltra 30 kDA SpinFilters (Millipore, Billerica, MA, USA) and stored at −80° C until gene analysis.

Gene analyses and indices

Digested and desalted fungal, archaeal, bacterial, and protozoal samples were shipped to the Genomic Technology Support Facility at Michigan State University (East Lansing, MI, USA) for analysis on an Applied Biosystems Prism 3130 Genetic Analyzer (Applied Biosystems, Foster City, CA, USA). The fragment data files were analyzed and filtered using GeneMarker V1.6 (SoftGeneTics LLC, State College, PA, USA) so only information associated with the tagged PCR products remained. This technique permits an automated quantification of the fluorescence signal intensities of the individual terminal restriction fragments in a given T-RFLP profile (i.e. community fingerprint). The resulting fragment patterns were exported to a MicroSoft Excel-based software package T-Align (Smith et al. 2005) for further determination of gene indices.

Each T-RFLP profile community was compared using T-Align. A unique terminal restriction fragments (T-RF) in a given profile that varied by 0.5 bp or less among all the T-RFLP community profiles was considered the same fragment. A file was generated that showed the presence or absence of a T-RF, its average size, and the percentage fluorescence intensity of total fluorescence in each profile. Estimates S (Colwell 2019), a MicroSoft Excel software add-in program, was employed to calculate diversity indices for the T-RFLP profiles using output files imported from T-Align. These indices include the Shannon-Weaver index (H′), Simpson diversity (reported as 1-D), the reciprocal Simpson (1/D), and evenness (E). The diversity equations used have been defined and described previously (Krause et al. 1995) and represent diversity measures based on fragment numbers in each treatment.

Statistical analyses

The VFA and diversity indices (H′, 1-D, 1/D, and E) data were analyzed according to a one-way analysis of variance using ANOVA procedure (SAS Institute, Cary, NC, USA). Sources of variations in the model included supplemental feedstuff and replicate (n = 4). Significance was reported at P < 0.05, unless otherwise stated.

Statistical analyses of T-RFLP profiles

Statistical analyses were performed with the GLM model of the SAS Statistical Package Version 9.1 (SAS Institute, Cary, NC, USA). Supplemental forages and feedstuffs were fixed effects and T-RFLP profiles random. Data from T-RFLP were transformed ((X + 0.5)^0.05) before analysis to satisfy the assumption of normality and equality of variances. Analysis of variance tests were performed on univariate data of T-RFLP profiles for each individual community (bacterial, archaeal, protozoal, and fungal), and mean comparisons were performed using Fisher’s protected least significant difference (LSD;α = 0.10). Interaction between supplemental feedstuffs and microbial populations were carried out on the DNA similarity search peak height data matrices using a modified Additive Main Effects and Multiplicative Interaction (AMMI) using a model described by Zobel and Wallace (1995). The model was modified to use the two-factor analysis of variance model proposed by Wright (1971) supplemented by Singular Value Decomposition of the residual (modified Principal Component Analysis). Non-significant terms were eliminated from the model and the analysis was re-run to produce the final tables (variance components were ordered by size) of variance; largest first; Tukey’s vacuum cleaner analysis (TVCA; Tukey 1962a, 1962b; Lee and Johnson 2006).

Results

Feedstuff chemical composition

Composition of supplemental forages and feedstuffs used in this experiment is shown in Table 2. Generalized trends in the chemical composition of these forages and feedstuffs were: (1) the total N (CP) was greatest for ALF, ART, and CCL relative to the other forages and feedstuffs; (2) the NDF was generally highest for OGR, RZP, and SLZ; (3) the ADF was generally least for ART and CCL when compared to the other feedstuffs; (4) the soluble carbohydrates were greatest in CORN, BFT, and PPN relative to other feedstuffs; and (5) the overall in vitro organic matter disappearance (IVOMD) of pure feedstuffs was generally greatest for CORN, CCL, BFT, and CHIC in comparison to other feedstuffs.

Table 2. Chemical composition (dry matter basis) of amendments^a that include dried plant samples, hays, and feedstuffs used in the experiment.

Item^b	OGR	CORN	WCS	ALF	RCL	CCL	BFT	SLZ	PNT	RZP	CHIC	CPUNA	CFEAST	ART
DM,%	89.7	86.9	89.6	93	90	94.5	86.5	88	90.5	90.9	86.1	87.2	89.9	92.1
Total N, %	1.9	1.4	3.8	4.3	3	3.4	2.2	1.7	2.1	2.4	2.4	2.6	2.6	3.6
NDF, %	67.2	10.8	47.3	38.2	41.6	27.5	41	54.9	31.4	67	38.3	35.2	31.6	25.4
ADF, %	37.6	3.7	37.9	26.2	33	18.5	30.5	43.6	26.8	50.4	31	27.7	27.2	14
TNC, %	9.3	75	8.7	2.4	4.4	8.7	15.4	5.3	18.6	4.3	8.7	ND	ND	8.3
IVOMD, %	55.6	87.9	36.1	60.8	44	73.3	73.8	20.4	59.7	46.4	73.9	ND	ND	60.8

^aORG = orchardgrass; CORN = corn grain; WCS = whole cottonseed; ALF = alfalfa; RCL = red clover; CCL = crimson clover; BFT = birdsfoot trefoil; SLZ = sericea lespedeza; PNT = perennial peanut (forage); RZP = rhizome peanut (forage); CHIC = chicory; CPUNA = chicory (cv Puna); CFEAST = chicory (cv Forage Feast); ART = artemisia (leaves).

^bDM = dry matter; N = nitrogen; NDF = neutral detergent fiber; ADF = acid detergent fiber; TNC = total non-structural carbohydrates; IVOMD = in vitro organic matter disappearance; ND = not determined.

Diversity indices

There were no differences (P > 0.01) in the gene indices for bacteria, archaea, or fungal (data not shown) in the short-term rumen ecological community established by the OGR baseline. Gene index differences (P < 0.01) for protozoa are presented in Table 3. Mean richness and percent richness indices of CORN, ORG, and BLANK were similar; all were higher (P < 0.001) when compared to the other treatments. The BLANK represented fresh rumen fluid with no additions of OGR or other supplements used in the experiment. Shannon’s index mean for ART, BFT, CCL, CFEAST, CHIC, CPUNA, PNT, RZP, SLZ, and WCS were similar; all were higher (P < 0.001) when compared to ALF, BLANK, CORN, ORG, and RCL. Simpson’s index means for ART, BFT, CCL, CFEAST, CHIC, CPUNA, and RCL were similar; all were higher (P < 0.001) compared to the remaining feedstuffs.

Table 3. gDNA indicies for Protozoa. Means ± 1 SEM^g.

Supplemental Feedstuff	Tmt^h	Richness	Shannons	%Richness	Simpsons
Orchardgrass	OGR	16.9 ± 1.3^a,c	2.86 ± 0^b	93.8 ± 7.2^a	1.00 ± 0.01^b
Corn grain	CORN	17.2 ± 1.6^a	2.85 ± 0.01^b	95.7 ± 9.0^a	1.00 ± 0.01^b
Whole cottonseed	WCS	14.4 ± 0.7^b	2.87 ± 0.01^a	79.8 ± 3.9^b	1.01 ± 0^a,f
Alfalfa	ALF	14.9 ± 1.0^b,c	2.86 ± 0.01^b	82.5 ± 5.6^b	1.01 ± 0.01^b,c,d,e,f
Red clover	RCL	14.8 ± 0.8^b,c	2.86 ± 0^b	82.3 ± 4.4^b	1.01 ± 0.01^b,c,f
Crimson clover	CCL	14.5 ± 1.0^b	2.87 ± 0.01^a	80.8 ± 5.3^b	1.01 ± 0^a
Perennial peanut	PNT	13.6 ± 0.3^b	2.87 ± 0^a	75.7 ± 1.4^b	1.02 ± 0^a
Rhizoma peanut	RPN	13.4 ± 0.1^b	2.88 ± 0^a	74.3 ± 0.7^b	1.02 ± 0^a,c
Birdsfoot trefoil	BFT	13.3 ± 0.2^b	2.88 ± 0^a	73.6 ± 1.2^b	1.02 ± 0^a
Sericea lespedeza	SLZ	13.4 ± 0.2^b	2.88 ± 0^a	74.6 ± 1.0^b	1.02 ± 0^a,c
Chicory	CHIC	13.9 ± 0.7^b	2.87 ± 0.01^a	77.2 ± 3.6^b	1.02 ± 0^a
Chicory-Puna	CPUNA	13.3 ± 0.0^b	2.88 ± 0^a	73.6 ± 0^b	1.02 ± 0^a
Chicory-Feast	CFEAST	13.5 ± 0.2^b	2.88 ± 0^a	75.0 ± 0.8^b	1.02 ± 0^a,e,f
Artemisia (leaf)	ART	13.5 ± 0.3^b	2.88 ± 0^a	75.0 ± 1.4^b	1.02 ± 0^a,d
BLANK	BLANK	17.0 ± 1.0^a	2.86 ± 0^b	94.2 ± 5.6^a	1.00 ± 0^b
P level		< 0.001	< 0.01	< 0.001	< 0.001

^a,b,c,d,e,fMeans within a column with unlike letters differ at the P level listed.

^gSEM = standard error of the mean.

^hTmt = treatment.

Interaction analysis

The data analysis is based on the variation of T-RF peak heights in each forage and feedstuff treatment relative to the 1-gm ORG baseline with the established short-term rumen community. Two assumptions were made in reporting the results: (1) the supplemental feedstuffs are independent and the T-RF are dependent; (2) all results are relative to the OGR baseline. Specifically, data are modeled as the variance around the T-RF values obtained from the ORG treatments.

Tukey’s vacuum cleaner analysis

Peak heights of each set of community profiles (bacteria, archaea, protozoa, and fungi) were analyzed via TVCA separately and are presented in the order of the AMMI analyses (Tables 4, 6, 8 and 9) followed by tables of mean interactions for forage and feedstuff treatments (Tables 7, 9, and Supplemental Table S3) and mean interactions for individual T-RF (Supplemental Tables S1, S2, and S4). Culman et al. (2008) have successfully applied AMMI analysis to T-RFLP data. The clustering of the T-RFLP profiles resulting from forage and feedstuff applications within each community (bacteria, archaea, protozoa, and fungi) is presented in Table 10.

Table 4. TVCA^a for Bacteria. Row (effects) are in Sums of Squares order (except the error term). The percentage of Tmt^b variance due to model effect is shown in the last column. Grand Mean = 0.81579.

Source		df	SS	Mean Square	P level	Sig.	% Tmt
Total		1919	94.19	0.0491
Tmt		479	27.05	0.0565	0.0044	**
	T-RF^c	31	12.43	0.4011	0.0000	***	45.97
	IAS 1^d	30	2.72	0.0907	0.0002	***	10.06
	Feedstuff	14	1.67	0.1191	0.0003	***	6.16
	Feed Regrs^e	13	1.11	0.0852	0.0147	*	4.09
Error		1831	76.26	0.0417

*P < 0.05; **P < 0.01; ***P < 0.001.

^aTVCA = Tukey’s vacuum cleaner analysis.

^bTmt = treatment.

^cT-RF = unique terminal restriction fragment.

^dIAS = interaction axis score.

^eRegrs = regression.

Table 5. Feedstuff mean and interaction for bacteria.

Feedstuff^c	Mean	LSD^d	Slope	IAS1^e
ALF	0.7988	b	−0.18658	0.1415
ART	0.8433	a	0.06491	0.1334
BFT	0.8598	a	0.23175	−0.4948
BLANK	0.8787	a	0.01524	0.3653
CHIC	0.8178	b	−0.20838	0.2507
CFEAST	0.7988	b	0.336410	−0.1134
CPUNA	0.7895	b	−0.43623	0.1926
CORN	0.7786	b	−0.06984	0.0338
CCL	0.8031	b	0.30396	−0.0127
OGR	0.8231	b	−0.45665	0.0720
PNT	0.7948	b	0.66707	−0.4089
RCL	0.7879	b	0.19438	−0.0607
RZP	0.7826	b	−0.23410	0.1198
SLZ	0.8408	a	0.00682	−0.3095
WCS	0.8393	a	−0.22875	0.0910
Grand Mean	0.8158	LSD05 = 0.0529

^a,bCorrected means without common letters differ P = 0.0529.

^cALF = alfalfa; ART = artemisia (leaves); BFT = birdsfoot trefoil; BLANK = blank; CCL = crimson clover; CFEAST = chicory (cv Forage Feast); CHIC = chicory; CPUNA = chicory (cv Puna); CORN = corn grain; ORG = orchardgrass; PNT = perennial peanut (forage); RCL = red clover; RZP = rhizome peanut (forage); SLZ = sericea lespedeza; WCS = whole cottonseed.

^dLSD = least significant difference.

^eIAS = interaction axis score.

Table 6. TVCA^a for Archea. This is a typical table for AMMI^b analysis. Row (effects) are in Sums of Squares order (except the error term). The percentage of treatment variance due to model effect is shown in the last column. Treatment Grand Mean = 0.87596.

Source	df	SS	MS	Prob.	Sig.	% Tmt
Total	3975	269.45	0.0678
Tmt^c	993	121.57	0.1224	0.00000	***
T-RF^d	70	84.32	1.2046	0.00000	***	69.4
IAS^e	69	4.76	0.0689	0.00539	**	3.9
feedstuff	13	3.99	0.3070	0.00000	***	3.3
Error	3823	176.39	0.0461

**P < 0.01; ***P < 0.001.

^aTVCA = Tukey’s vacuum cleaner analysis.

^bAMMI = additive main effects and multiplicative interaction.

^cTmt = treatment.

^dT-RF = unique terminal restriction fragment.

^eIAS = interaction axis score.

Table 7. Feedstuff mean and interaction for archaea.

Feedstuff^d	Mean	LSD^e	IAS^f
ALF	0.8903	a,b	0.0153
ART	0.8829	a,b	0.6111
BFT	0.8868	a,b	0.1986
CHIC	0.8668	b	−0.2530
CFEAST	0.8513	b,c	0.2098
CPUNA	0.9185	a	−0.0454
CORN	0.8709	b	−0.4598
CCL	0.8555	b,c	0.2439
ORG	0.7851	d	0.0011
PNT	0.8969	a	0.2195
RCL	0.8611	b	−0.2331
RZP	0.8872	a,b	−0.4361
SLZ	0.9026	a	−0.0132
WCS	0.9074	a	−0.0587
Grand	0.87596	LSD05	0.03664

^a,b,cCorrected means with unlike letters differ (P < 0.05).

^dALF = alfalfa; ART = artemisia (leaves); BFT = birdsfoot trefoil; BLANK = blank; CCL = crimson clover; CFEAST = chicory (cv Forage Feast); CHIC = chicory; CPUNA = chicory (cv Puna); CORN = corn grain; ORG = orchardgrass; PNT = perennial peanut (forage); RCL = red clover; RZP = rhizome peanut (forage); SLZ = sericea lespedeza; WCS = whole cottonseed.

^eLSD = least significant difference.

^fIAS = interaction axis score.

Table 8. TVCA^a for Protozoa. Row (effects) are in Sums of Squares order (except the error term). The percentage of treatment variance due to model effect is shown in the last column. Treatment Grand Mean = 0.80530.

Source		df	SS	MS	Prob.	Sig.	% Tmt
Total		1079	50.79	0.0471
Tmt^b		269	15.73	0.0585	0.0009	***
	Feedstuff	14	6.23	0.4450	0.0000	***	39.6%
	T-RF^c	17	2.30	0.1352	0.0000	***	14.6%
	T-RF Regrs^d	16	1.91	0.1195	0.0000	***	12.2%
	IAS 1^e	16	1.48	0.0926	0.0014	**	9.4%
Error		1016	38.87	0.0383

**P < 0.01; ***P < 0.001.

^aTVCA = Tukey’s vacuum cleaner analysis.

^bTmt = treatment.

^cT-RF = unique terminal restriction fragment.

^dRegrs = regression.

^eIAS = interaction axis score.

Table 9. TVCA^a for Fungi. Results obtained by removing non-significant factors from the analysis model and rerunning the TVCA analysis. Grand Mean = 0.83083.

Source		df	SS	MS	Prob.	Sig.
Total		239	12.3369	0.0516
Tmt^b		59	4.1727	00.0707	0.0139	*
	Feedstuff	14	0.6139	0.0438	0.4891
	T-RF^c	3	2.3253	0.7751	0.0000	***
	Feed x T-RF	42	1.2334	0.0293	0.9513
	Joint Regrs^d	1	0.0038	0.0038	0.7699
	Feed Regrs	13	0.6966	0.0535	0.2961
	T-RF Regrs	2	0.0125	0.0062	0.8707
	IAS 1^e	13	0.3972	0.0305	0.7870
	IAS 2	13	0.1231	0.0094	0.9986
Error		180	8.1642	0.0453
Grand mean		0.83083
JOINT	REGRESSION		K =	0.8086

*P < 0.05; ***P < 0.001.

^aTVCA = Tukey’s vacuum cleaner analysis.

^bTmt = treatment.

^cT-RF = unique terminal restriction fragment.

^dRegrs = regression.

^eIAS = interaction axis score.

Table 10. AMMI^a model analysis of feedstuff^b grouping patterns (action with T-RF^c) for an individual rumen ecological group (bacteria, archaea, protozoa, and fungi).

Group	Pattern 1	Pattern 2	Pattern 3	Weak
Bacteria	RCL	CHIC	ORG	CORN
	PNT	RZP	CPUNA	ALF
	BFT	WCS		CCL
	CFEAST	ART		SLZ
Archaea	PNT	CORN		ORG
	ART	RCL		ALF
	CCL	CHIC		WCS
	BFT	RZP		CPUNA
	CFEAST			SLZ
Protozoa	ORG	RCL	PNT	RZP
	CORN	ALF	ART	BFT
	CHIC	WCS	CPUNA	SLZ
		CCL		CFEAST
Fungi				All

^aAMMI = Additive Main-effects and Multiplicative Interaction. Model generated by adding grand mean to the additive variances of each individual sub-factor then adding the regressions variances and the interaction variance products: Y = μ + α + β + καβ + εα + οβ + γσ, where μ = grand mean, α = supplemental feedstuff additive variance, β = T-RF additive variance, κ =Tukey one degree of variance coefficient, ε = slope of T-RF regression, ο = slope of supplemental feedstuff regression, γ = interaction score supplemental feedstuff, and σ = interaction score for T-RF.

^bALF = alfalfa; ART = artemisia (leaves); BFT = birdsfoot trefoil; BLANK = blank; CCL = crimson clover; CFEAST = chicory (cv Forage Feast); CHIC = chicory; CPUNA = chicory (cv Puna); CORN = corn grain; ORG = orchardgrass; PNT = perennial peanut (forage); RCL = red clover; RZP = rhizome peanut (forage); SLZ = sericea lespedeza; WCS = whole cottonseed.

^cT-RF = unique terminal restriction fragment.

Bacteria

The TVCA (Table 4) includes a row of feedstuff regressions that are associated with T-RF means. Removal of PNT treatment from the data set eliminates the regression significance, demonstrating that the positive slope of the PNT regression is responsible for this significant interaction. Specifically, the DNA similarity search peak value of all T-RF is increased with PNT feedstuff, and T-RF with higher means are increased to a greater degree by addition of PNT. When CPUNA and ORG are removed from the data set, the regression becomes non-significant most likely due to the negative response slopes associated with these treatments (Table 5). The interaction of these two feedstuffs was opposite to that of the PNT treatment. Specifically, a decrease in peak height with increasing T-RF mean. The ART, BFT, SLZ and WCS treatments had the highest peak heights of the forage and feedstuffs (Table 5). In the bacteria data, T-RF 2, 4, 5, and 24 have the highest average peak heights, while T-RF 29 has the lowest (Table S1).

Interaction axis significance is predicated by PNT feedstuff addition, and T-RF 13. The removal of either of these from the dataset removes the significance of the interaction axis, thus indicating that they have a broad interaction with the opposing factor elements (positive with some, negative with others, and slight with the rest). If CHIC and T-RF14 or BFT and T-RF 21 are removed, the interaction axis also becomes non-significant, indicating specific interactions (i.e. CHIC specifically suppresses T-RF 14 and BFT specifically suppresses T-RF 21).

In the full model, there were three distinct pattern groupings (1) BFT, PNT, RCL, CFEAST; (2) CHIC, RZP, WCS, ART; and (3) CPUNA, ORG. The remaining feedstuffs slightly mimic one of these three patterns, but at very low variance (Weak; Table 10). The developed TVCA model accounted for 66.3% (Table 4) of the treatment variance and demonstrated that BFT and PNT dramatically increase the peak height of T-RF 2, 5, 12, 14, and 24, while PNT suppressed the peak heights of T-RF 6-11, 16-18, 23, 26, 28, and 29. Feedstuffs CHIC and RZP suppressed peak heights of T-RF 1, 3, 12, 14-15, 22, 27, and 30–32 whereas CHIC increased the peak heights of T-RF 13 and 21. In contrast, addition of CHIC and RZP reduced the peak heights of T-RF 12, 13, and 14 and additions of BFT and PNT has the opposite effect.

Archaea

The TVCA demonstrated a typical AMMI analysis pattern (i.e. significant additive variation within the two sets of factors and one significant interaction axis). The feedstuff ORG decreased T-RF peak height of Archaea (lowest average additive variance T-RF peak height by 2 LSDs; Table 6). The dominant factor feature in the additive feedstuff variance was the low value of T-RF peak heights in the ORG treatment (Table 7), and higher values for T-RF 37, 38, and 67 (Table S2). The T-RF peak heights 37, 38 and 67 are high in additive variance (Table S2). The interaction component was characterized by strong opposite interactions of the feedstuffs CORN and ART; and T-RF 4, any of which when eliminated from the data set, increased the probability above 0.05. This indicates that these factor elements have a broad interaction with the elements of the opposing factor (positive with some, negative with others, and slight with the rest).

The signs of the Interaction Scores (IAS) were used to determine the nature of the interaction [e.g. feedstuff CORN interacted negatively with the Archaea represented by T-RF 4 (−0.45812 * 0.38769 = −0.17761)]. Thus, the average DNA similarity search peak for T-RF 4 when CORN was used [0.87618 + (−0.00618) + 0.10832 = 0.97832] decreased by 0.17761 (0.97832–0.17761 = 0.80071) to a value that is less than the grand mean of 0.87618. In contrast, the average DNA similarity search peak height for T-RF 4 increased after addition of ART. The interaction of feedstuff CORN with the average DNA similarity search of T-RF peak heights revealed an inhibitory effect when the IAS was positive and a stimulatory effect when the IAS was negative, while feedstuff ART addition resulted in the opposite response.

By using paired elimination of characters from the data set, it was determined that the removal of RCL and T-RF 1; RZP and T-RF 1; and CHIC and T-RF 1, each eliminated the significant interaction axis. These pairs represented specific interactions (in this case negative interactions), where ART, CORN and T-RF 4 generated broad interactions.

The TVCA procedure was designed to remove noise from data sets. The simplified reduced full model accounted for 76.6% of the treatment variance in archaea T-RFLP profiles, and it was found that there were distinct patterns in the model (Table 10). Pattern 1 represented the response of T-RF to treatments ART, CCL, CFEAST, and to a lesser extent, to BFT and PNT. Pattern 2 represented the response to T-RF to treatments CORN, RZP, and to a lesser extent RCL and CHIC additions. The third pattern represented the response of T-RF to treatments ORG, CPUNA, WCS, SLZ, and ALF, all of which affected the T-RF equally (Weak; Table 10). Because the last four treatment additions of the third pattern gave similar peak height values (Table 7), it was assumed that ORG generally suppressed archaeal development. The strong interaction variance of ART addition raised the model variances of T-RF 1, 4, 49, 50, 53, and 69 well above those of the other feedstuffs, while strongly suppressing T-RF 12, 20, and 22. The CORN addition had exactly the opposite response pattern. Variance in T-RF patterns after addition of BFT, RCL, and ALF mimicked to a lesser effect that of the ART treatment.

Protozoa

The model derived from TVCA analysis accounted for 75.8% of the treatment variance in T-RF peak heights (Table 8) and described 3 distinct patterns: ORG, CORN, and CCL (pattern 1); RCL, ALF, WCS, CHIC (pattern 2); PNT, ART, and CPUN (pattern 3); while pattern 4 consisted of BFT, RZP, CFEA, and SLZ, but represented low or intermediate T-RF values with no distinctive peaks (Table 10). The CORN and ORG feedstuffs had the greatest effect on increased T-RF peak heights (Table S3) and the mean peak heights of T-RF 18 was greater than that of any other T-RF (Table S4). In the protozoa data set, there was a significant T-RF regression associated with the feedstuff means. The average peak height values of T-RF 7 and 8 increased with feedstuff additions relative to the remaining T-RF while peak height values of T-RF 14, 15, 16, and 17 decreased with feedstuff additions (Table S4). If WCS, RCL, or T-RF 15 were eliminated from the data set, then the significant interaction axis was removed. Therefore, these factor elements had a broad interaction with the elements of the opposing factor (positive with some, negative with others, and slight with the rest). Conversely, if ALF and T-RF 14 were removed, then significant interaction axis was also removed; this effect represented a specific interaction where ALF selectively suppressed the peak height of T-RF 14.

Patterns are presented in Table 10. In pattern 1, both ORG and CORN additions increased T-RF peak heights with the exception of T-RF 14, 15, and 17 (Table S4). Both feedstuffs suppressed T-RF 17, but only ORG suppressed T-RF 15. In the non-distinct group, BFT suppressed the peak height of T-RF 15 while the RCL group (pattern 2) stimulated it. The RCL group (pattern 2) suppressed peak heights of T-RF 2, 3, 10 and 14, but stimulated T-RF 5 and 11 to nearly the same degree as ORG and CORN additions (pattern 1). The BFT addition (pattern 4; Weak) generally suppressed T-RF peak heights except for T-RF 1, 14, 16, and 18. The PNT group (pattern 3) followed a similar less distinct pattern but did not suppress T-RF 15 peak height. The CHIC (pattern 1) and CCL groups (pattern 2) were intermediate and were distinct from one another. The CCL (pattern 2) mirrored ORG pattern 1 while the CHIC treatment mirrored the RCL group (Table 10).

Fungi

There were no significant mean interactions with forage and feedstuff treatments or variation among T-RF within the fungal community. Therefore, only the AMMI analysis (Table 9) is presented. The DSS revealed no interactions. All pattern groupings for the forages and feedstuffs were Weak (Table 10).

pH and VFA profiles

The pH and VFA concentrations are presented in Table 11. The pH at the end of the 48-h incubation ranged from 6.4–7. The pH was lowest for CORN compared to the other feedstuffs. Total VFA concentrations were lowest for BLANK compared to all feedstuffs. The OGR, WCS, ALF, and SLZ tended to have lower total VFA in comparison to the majority of the feedstuffs.

Table 11. Volatile fatty acid concentrations from in vitro rumen fermentation of feedstuffs and forages used in the experiment. Data are Mean ± 1 SEM^h.

Tmt^i	Total VFA, mmol	Acetic,%	Propionic, %	I-Butyric, %	Butyric,%	I-Valeric, %	Valeric, %	pH	A:P Ratio	A:B Ratio
OGR	115.1 ± 11.9d,e,f	63.9 ± 2.8d	28.6 ± 1.3a	1.1 ± 0.1b,c,d	5.4 ± 1.7b	0.5 ± 0.1c	0.5 ± 0.1d,e,f	6.9 ± 0.1c,d,e	2.3 ± 0.2e	14.7 ± 3.1d,e
CORN	130.9 ± 1.1a,b,c	63.3 ± 0.4d	28.0 ± 0.5a,b	1.0 ± 0.1d	6.7 ± 0.2a	0.5 ± 0.03c	0.5 ± 0.01c,d	6.4 ± 0.05g	2.3 ± 0.1e	9.5 ± 0.2f
WCS	107.6 ± 6.0f	65.2 ± 0.2c,d	28.7 ± 0.1a	1.2 ± 0.1b	3.7 ± 0.1e,d	0.6 ± 0.03b	0.6 ± 0.03a,b	7.1 ± 0.02b,c	2.3 ± 0.02e	17.9 ± 0.4b,c
ALF	118.6 ± 5.3c,d,e,f	67.9 ± 0.5b,c	26.1 ± 0.4c,d	1.2 ± 0.1b,c	3.5 ± 0.1e,d	0.7 ± 0.04b	0.6 ± 0.02a,b	7.0 ± 0.1b,c,d	2.6 ± 0.1c,d	19.5 ± 0.6a,b,c
RCL	123.6 ± 1.8b,c,d,e	68.5 ± 0.5a,b	26.4 ± 0.5b,c,d	1.0 ± 0.03d	3.1 ± 0.1e	0.5 ± 0.01c	0.5 ± 0.01d,e,f	6.9 ± 0.04c,d,e	2.6 ± 0.1c,d	21.9 ± 0.5 a
CCL	133.9 ± 0.9a,b	67.2 ± 0.6b,c	27.3 ± 0.6a,b,c	1.0 ± 0.03d	3.5 ± 0.1e,d	0.5 ± 0.02c	0.6 ± 0.02b,c	6.9 ± 0.01c,d,e,f	2.5 ± 0.1d,e	19.2 ± 0.6a,b,c
BFT	133.5 ± 1.1a,b	67.7 ± 0.7b,c	26.4 ± 0.8b,c,d	0.9 ± 0.1d	4.0 ± 0.03c,d,e	0.5 ± 0.03c	0.5 ± 0.01c,d,e	6.7 ± 0.02f	2.6 ± 0.1c,d	16.8 ± 0.2c,d,e
SLZ	110.6 ± 3.6e,f	68.6 ± 0.5a,b	25.9 ± 0.4c,d,e	1.0 ± 0.03b,c,d	3.5 ± 0.1e,d	0.5 ± 0.02c	0.5 ± 0.1d,e,f	7.0 ± 0.03b,c,d	2.7 ± 0.1c,d	19.5 ± 0.5a,b,c
PNT	130.3 ± 2.2a,b,c	67.5 ± 0.7b,c	26.7 ± 0.6b,c,d	0.9 ± 0d	3.9 ± 0.1c,d,e	0.5 ± 0.02c	0.5 ± 0.01d,e	6.8 ± 0.05e,f	2.5 ± 0.1c,d	17.5 ± 0.7c,d,e
RZP	125.3 ± 5.8a,b,c,d	67.7 ± 1.1b,c	26.6 ± 0.8b,c,d	1.0 ± 0.03b,c,d	3.6 ± 0.3e,d	0.5 ± 0.02c	0.5 ± 0.02d,e,f	6.9 ± 0.1d,e,f	2.6 ± 0.1c,d	19.0 ± 1.6a,b,c
CHIC	136.0 ± 3.0a,b	69.3 ± 0.6a,b	25.3 ± 0.5d,e	1.0 ± 0.1c,d	3.4 ± 0.1e	0.5 ± 0.04c	0.5 ± 0.03e,f,g	6.8 ± 0.1e,f	2.7 ± 0.1b,c	20.7 ± 0.7a,b
CPUNA	137.5 ± 3.6a,b	67.4 ± 0.3b,c	27.0 ± 0.3a,b,c,d	1.0 ± 0.02d	3.7 ± 0.1c,d,e	0.5 ± 0.02c	0.5 ± 0.02d,e,f	6.7 ± 0.1e,f	2.5 ± 0.04d	18.2 ± 0.4b,c
CFEAST	138.9 ± 1.1 a	67.6 ± 0.1b,c	27.0 ± 0.1a,b,c,d	0.9 ± 0.02d	3.5 ± 0.04e,d	0.5 ± 0.02c	0.5 ± 0.02e,f	6.7 ± 0.1f	2.5 ± 0.01d	19.3 ± 0.2a.b.c
ART	126.3 ± 3a,b,c,d	69.8 ± 0.3a,b	24.3 ± 0.3e	1.0 ± 0.04d	3.8 ± 0.1c,d,e	0.5 ± 0.02c	0.7 ± 0.01a	6.8 ± 0.7d,e,f	2.9 ± 0.04b	18.6 ± 0.6a,b,c
BLANK	2.4 ± 0.02 g	71.2 ± 0.2a	20.6 ± 0.1f	2.0 ± 0.1a	4.9 ± 0.1b,c,d	0.9 ± 0.01a	0.4 ± 0f,g	7.6 ± 0.1a	3.5 ± 0.03a	14.5 ± 0.2d,e
P value	< 0.001	< 0.001	< 0.001	< 0.001	< 0.001	< 0.001	< 0.001	≤ 0.001	< 0.001	< 0.001

^{a,b,c,d,e,f,g}Means within a column without a common letter differ at the P value listed.

^hSEM = standard error of the mean; VFA = volatile fatty acid; A:P ratio = acetic to propionic ratio; A:B ratio = acetic to propionic ratio.

ⁱORG = orchardgrass; CORN = corn grain; WCS = whole cottonseed; ALF = alfalfa; RCL = red clover; CCL = crimson clover; BFT = birdsfoot trefoil; SLZ = sericea lespedeza; PNT = perennial peanut (forage); RZP = rhizome peanut (forage); CHIC = chicory; CPUNA = chicory (cv Puna); CFEAST = chicory (cv Forage Feast); ART = artemisia (leaves).

Acetate concentration in the fresh rumen fluid BLANK was higher (P < 0.001) than ALF, RZP, BFT, CFEAST, PNT, CPUNA, CCL, WCS, OGR, and CORN; CORN, OGR and WCS were similar. The acetate concentration in BLANK was similar to ART, CHIC, SLZ, and RCL. Acetate was similar for ART, CHIC, SLZ, RCL, ALF, RZP, BFT, CFEAST, PNT, CPUNA and CCL.

Propionate concentrations were higher (P < 0.001) for WCS and OGR when compared to all treatments except CORN, and CCL. The propionate produced from supplemental CORN, CCL, CFEAST, CPUNA, PNT, RZP, RCL and BFT were similar, but higher (P < 0.001) than ART and BLANK. Propionate produced from added BFT, ALF, SLZ, and CHIC were similar, but higher (P < 0.001) than ART and BLANK.

Butyrate concentration was highest (P < 0.001) for CORN compared to the other treatments. The OGR was higher (P < 0.001) than WCS, RZP, SLZ, CCL, CFEAST, ALF, CHIC and RCL. The BLANK was similar to OGR, BFT, PNT, ART, CPUNA, WCS, RZP, SLZ, CCL, CFEAST, ALF, CHIC, and RCL.

Iso-butyrate was highest (P < 0.001) for BLANK compared to all other treatments. The WCS was similar to ALF, OGR, SLZ, and RZP, but higher (P < 0.001) than the remaining treatments while ALF was similar to OGR, SLZ, RZP and CHIC, but higher (P < 0.001) than the remaining treatments. The OGR, was similar to SLZ, RZP, CHIC, ART, RCL, CCL, CPUNA, CORN, CFEAST, BFT and PNT.

Valerate in ART was similar to ALF and WCS, but higher (P < 0.001) than the other treatments. ALF was similar to WCS and CCL and higher (P < 0.001) than the remaining treatments. The CCL, CORN, and BFT were similar.

Iso-valerate followed the trend of BLANK > ALF, WCS > remaining treatments.

The acetate:propionate (A/P) ratio of the BLANK was greater (P < 0.001) compared to all the other treatments. Feedstuff CHIC and ART had similar ratios. The A/P ratio in CHIC, SLZ, ALF, RCL, BFT, RZP, PNT were similar, but greater (P < 0.001) than WCS, CORN, and OGR which were similar.

The acetate:butyrate (A/B) ratio for RCL was similar to CHIC, SLZ, ALF, CFEAST, CCL, RZP, and ART, but greater (P < 0.001) than the remaining treatments. The CORN had the lowest (P < 0.001) A/B ratio compared to the other treatments while BFT, OGR, and BLANK were similar to PNT, but lower than all treatments except CORN.

Discussion

Potential effects of forages and feedstuffs on rumen diversity

Forages and supplemental feedstuffs used in the experiment represent typical conserved forages and feedstuff supplements to cattle diets or forages grazed by cattle, except for Artemisia annua, and represent variations in protein, soluble carbohydrates, fiber, and secondary plant metabolites.

Orchardgrass (OGR) is a cool-season perennial grass (Christie and McElroy 1995) while corn grain (CORN) is a starchy feed grain (Theurer 1986). Whole cottonseed (WCS) is a by-product feedstuff of the cotton ginning industry often used as a high protein feed supplement for ruminants and contains condensed tannins (Terrill et al. 1992; Feng et al. 1993). Alfalfa (ALF), red clover (RCL), crimson clover (CCL), peanut hay (PPN), rhizome peanut hay (RZP), birdsfoot trefoil (BFT) and sericea lespedeza hay (SLZ) are all classified as high protein legume forages (McGraw and Nelson 2003; Sollenberger and Collins 2003). Chicory (CHIC), chicory cv. Puna (CPUNA), and chicory cv. Forage Feast (CFEAST) are classified as forbs and are generally used for grazing; each cultivar differs in the amount of secondary plant compounds (i.e. sesquiterpene lactones, Foster et al. 2011; and condensed tannins, Terrill et al., 1992). Artemesia (ART) is a forb containing secondary plant compounds (i.e. sesquiterpene lactones; Jessing et al. 2014).

Stimulation or suppression of T-RF or peak heights in our study is most likely a result of specific nutrients or secondary metabolites contained in the feedstuffs affecting the rumen ecology. The secondary metabolites were not quantified in the specific feedstuffs or hays used in the present study, but the discussion included here is relevant, published research to support overall trends of these metabolites.

In our study, treatments receiving addition of CORN and OGR had greater diversity of protozoa compared to the other treatments. Addition of rapidly fermentable carbohydrates (i.e. starch in CORN) likely stimulated protozoa; the OGR treatment represented the fresh microbial and protozoan populations in the rumen fluid used in the current experiment. Kocherginskaya et al. (2001) reported greater microbial diversity and richness in grain-based diets compared to hay diets. In addition, results of the present study represent rapid (48 h) utilization of added nutrients by microorganisms and protozoa, but do not represent long-term adaptation of overall rumen populations in vivo. Thus, only initial response of the rumen to changes are represented in vitro. Because of the synergism and antagonism among the rumen microbes and protozoa, the ability to quantify and identify the specific relationships/interactions among microbial groups is difficult (Kamra 2005). Short-term in vitro of grass-legume mixture have been used to evaluate rumen fermentation parameters (Tilley and Terry 1963; Niderkorn et al. 2011) Overall livestock fed high-grain (starch) diets have lower bacterial diversity, especially fibrolytic microbes (Belanche et al. 2012). Mainly amylolytic bacteria (and protozoa and fungi to a lesser degree) degrade starch in the rumen (Huntington 1997). The community species richness (bacteria) in the rumen was higher when alfalfa hay compared to triticale (Triticum aestivum) straw was fed to cows (Kong et al. 2010). These authors suggest that the greater density and complexity of overall nutrients available in alfalfa vs. triticale was responsible for the higher rumen bacterial species richness. Thus, overall chemical composition differences among feedstuffs is an important factor impacting rumen microbial ecology profiles. Zhu et al. (2018) reported that the relative bacterial abundance of Ruminococcus, Butyrivibrio, Clostridium, Coprococcus, and Pseudobutyrivibrio were not impacted when dairy cattle were transitioned from low-grain (pre-partum) to high-grain (post-partum) diets. These researchers also reported that there were distinct changes in the relative rumen abundance of Prevotella, Ruminococcaceae, and Succinibibrionaceae as well as in profiles of VFA.

Bacteria

Primary fibrolytic bacteria in the rumen include Fibrobacter succinogenes, Ruminococcus albus and R. flavefaciens while F. succinogenes and R. flavefaciens are associated with the particulate fraction in the rumen (Firkins and Yu 2006). Fang et al. (2018) reported that F. succinogenes concentration and overall NDF and ADF concentrations were higher in perennial ryegrass-white clover fermentation cultures compared to herbal plintain and fodder beet forage rumen fermentation cultures using mixed rumen fluids from different cattle and PCR techniques. In a hay feeding study with cattle, total counts of ruminal bacteria were greatest when fed alfalfa, intermediate for switchgrass (Panicum virgatum L; warm-season grass), and least for smooth bromegrass (Bromus inermis Leyss.; cool-season grass) (Jung and Varel 1988). Bowman and Firkins (1993) reported that red clover had more cellulolytic bacteria associated with forage particles in early rumen incubation in situ compared with a cool-season grass (orchardgrass) and a warm-season grass (gamagrass; Tripsacum dactyloides [L.] L). In our study, the initial rumen fluid along with a sample of the particulate fraction were collected from ruminally cannulated steers. The collected rumen fluid was agitated to weaken attachment of bacteria to fiber with the intent of obtaining a more homogenous inoculum for tubes.

Secondary plant metabolites can suppress or stimulate rumen microbial populations. Growth of the rumen hyper-ammonia producing bacteria Clostridium sticklandii has been shown to be inhibited by red clover (Flythe and Kagan 2010). Red clover contains the enzyme polyphenol oxidase which protects plant protein from degradation in the rumen (Lee 2014). Min et al. (2002) reported that condensed tannin in birdsfoot trefoil offered to sheep decreased populations of the proteolytic rumen bacteria Clostridium proteoclasticum B316, Eubacterium sp. C12b, Streptococcus bovis B315, and Butyrivibrio fibrisolvens C211a. Sun et al. (2008) reported that Lachnospira multiparus and Fibrobactor succinogenes but not Ruminococcus flavefaciens or Butyrivibrio hungatei degraded fresh forage chicory in vitro; L. multiparus degraded high pectin fiber in chicory faster than F. succinogenes. Chicory (CHIC, CPUNA and CFEAST in the present study) contains sesquiterpene lactones which are reported to be antifungal, antibacterial, and antiprotozoal (Picman 1986). Plants containing sesquiterpene lactones have been reported to have anthelmintic (deworming) activity against Haemonchus contortus (a gastrointestinal nematode of sheep and goats) eggs in vitro (Foster et al. 2011).

Protozoa

Rumen protozoa are generally more sensitive to dietary changes than bacteria. Protozoa are difficult to isolate from the rumen environment, and are generally found in higher numbers when diets contain more readily fermentable carbohydrates (Yokoyama and Johnson 1988). Fang et al. (2018) reported the protozoa abundance in in vitro fermentation cultures using mixed rumen fluids from different cattle and PCR techniques were greater for fodder beet (Beta vulgaris L.) compared to herbal plantain (Plantago lanceolata L.) and perennial ryegrass (Lolium perenne L.)-white clover forages. Fodder beet in that experiment had the highest water soluble cabohydrates compared to the other two forages. In our study, CORN supplementation (starch and a more readily fermentable carbohydrate) resulted in a greater protozoan Richness index. However, protozoa alone are not responsible for degrading the cell wall fraction (Lee et al. 2000). Cellulolysis of orchardgrass cell walls by bacteria was reported to be highest during early stages (Lee et al. 2000). Using rumen fluid collected from animals offered different forages (cool- vs warm-season grasses) and supplemental grain would likely have influenced our results. Additionally, dietary fat and dietary oil and essential oils have been shown to suppress protozoal numbers (Loor et al. 2004; Hristov and Jouany 2005) which may be the case with oil-containing. WCS supplementation in our study. Saponins in plants (high in ALF) would not have eliminated all protozoa in the rumen, but probably reduced total protozoal numbers (Lu and Jorgensen 1987) and specific protozoal species (Ivan et al. 2004). Saponins in alfalfa (Sen et al. 1998) is dependent upon cultivar and growth stage (including number of leaves in early growth stages) (Hanson et al. 1963).

Archaea

Bacteria in the genus Methanobrevibacter are primarily responsible for methane production in the rumen (Miller 1995). Ouwerkerk et al. (2008) suggested that grain feeding reduced rumen methanogen species diversity in Bos indicus cattle. Sheep offered alfalfa pellets had a more diverse Archaea community than Bos indicus cattle and may have been the result of differences in host-animal eating behavior or greater pectin (saponins) in alfalfa that supports Archaeal growth (Ouwerkerk et al. 2008). Rumen methanogens were directly inhibited by condensed tannins in Lotus pedunculatus (Tavendale et al. 2005). Condensed tannins were shown to indirectly reduce hydrogen production via decreased forage digestibility in the rumen (Tavendale et al. 2005). Orchardgrass in the present study significantly reduced peak heights of T-RF of archaea. These results may be due to increases in propionate and butyrate production resulting from competition for hydrogen in the rumen (Abberton et al. 2007). Propionate and butyrate concentrations in OGR in the present study were higher in comparison to many of the other feedstuffs used in the study. The aspect of using orchardgrass pasture and hays in livestock production to help reduce methane emissions needs further investigation.

Fungi

Rumen fungi have cellulolytic and hemicellulolytic activity (Bauchop 1981). Although the anaerobic fungi make up a very small percentage of the total microbial population they are thought to act as the initial colonizers of plant material and play an important role in plant cell wall weakening due to their high hemicellulase and cellulase activity and tend to be reduced on high concentrate diets. Cellulolysis of orchardgrass cell wall by the bacterial fraction was highest during early incubation stages, but cellulolysis by the fugal fraction was highest during the later stages of incubation (Lee et al. 2000). In the current study there was no variation in the fungal levels among the feedstuffs, whereas in contrast, Fang et al. (2018) reported that anaerobic fungi levels were higher in perennial ryegrass-white clover forage samples following in vitro rumen fermentation compared to herbal plantain or fodder beet forages. These authors further related the higher fungi levels to higher fiber concentration in the ryegrass-white clover forages compared to plantain or fodder beet forage samples. Specific fungi effects in the current study were not evident.

pH and volatile fatty acids

The range in pH (6.4–7) was considered conducive to rumen microbial ecology activity (Roger et al. 1990). In an in-vitro study, Mouriño et al. (2001) reported that rumen pH below 6.0-6.3 (considered a critical value) impaired fiber breakdown. A low rumen pH reduced primary fibrolytic bacteria (Erdman 1988). As VFA are produced by the rumen ecosystem in vivo or in vitro can reduce the overall pH if the VFA are not buffered or removed from the system. In the present study the pH was buffered with McDougall’s buffer which should have been effective for approx. 4 d. Volatile fatty acid concentrations and ratios are the result of a mixed community in the rumen that represents interactions among bacteria, archaea, protozoa, and fungi which creates a unique VFA profile dependent on the overall diet or feedstuff being fermented. The main fermentation products of bacterial spp. include acetate, butyrate, propionate, valerate, and iso-butyrate, and iso-valerate while the main fermentation products of protozoal spp. are acetate and butyrate (Van Soest 1983). Cheng et al. (2009) reported that acetate with minimum quantities of lactate and formate were the main end products when rumen-sourced anaerobic fungi and methanogens digested lignocellulose in batch culture.

Forage-based diets typically result in the production of acetate as the predominant VFA while grain-based (CORN or high carbohydrate) diets result in the production of propionate as the predominant VFA in the rumen. High concentrate (starch as in corn) in ruminant diets typically have the lowest A:P ratio (Bannink et al. 2008; Dijkstra et al. 2012). Zhu et al. (2018) reported that when dairy cows were transitioned from low-grain to high-grain diets, the change in specific bacterial abundance resulted in changes in VFA profiles. Forage-based diets with supplemental fats generally inhibit protozoa and increase rumen propionate concentrations in the rumen (Firkins 1996). In the present study, addition of WCS (typically contains high oil; Moore et al. 1986) did not increase propionate concentrations over those of ORG or CORN, but all three were higher in propionate relative to the other supplemental feedstuffs. Saponins in plants such as alfalfa have been reported to reduce total short-chain VFA but increase the molar proportions of propionate to decrease the acetate to propionate ratio (Lu and Jorgensen 1987). Saponins can target specific rumen bacteria, archea, protozoa, and fungi to impact H ion flow thus impacting VFA and methane production (Patra and Saxena 2009). In the present study total VFA and propionate concentrations in ALF were not different from other protein legume feedstuffs suggesting that saponins did not have an impact on VFA. Waghorn and Shelton (1995, 1997) reported 1.8% condensed tannin in the dietary DM had minimal effect on rumen fermentation, mainly on minor VFA (isobutyrate, valerate, and isovalerate). In the present study, SLZ (condensed tannin containing; Muir et al. 2018), and WCS (oil; Moore et al., 1986, and condensed tannin containing; Yu et al., 1993) resulted in similar VFA patterns while the BFT (contains condensed tannin; Min and Hart 2003) treatment exhibited higher overall VFA concentrations. Condensed tannins in SLZ forage have been reported to have a higher affinity for protein than tannins in BFT forages (Aerts et al. 1999). Thus, in our study, condensed tannins in SLZ could bind protein more tightly than condensed tannins in BFT resulting in the protein being protected from rumen fermentation and may have resulted in lower VFA concentrations overall.

Conclusions

The information presented here is an application of the T-RFLP method to a rumen in vitro digestibility technique commonly used to evaluate forage and feedstuff chemical composition for ruminants. Our results demonstrate that the ecological profile resulting from T-RFLP methods and then refined using a simplified TVCA analysis can be used to determine effects of feedstuffs by identifying groupings of T-RF of rumen bacteria, archaea, and protozoa from rumen fermentation in vitro. Nutritionists and producers typically provide supplemental feedstuffs and mixed rations to correct nutrient deficiencies for improved overall feed- and forage-use efficiency and improved animal growth and performance, all aimed at improving economic return. Information in the present study could be used to allot forages and feedstuffs into ecological groups with similar taxonomic patterns. Ecological profiling could potentially be used to create forage mixtures and feedstuff supplements (energy, protein, or a combination) for improved nutritional management and nutrient use in livestock, However, more research is needed on integration of ecological groupings based on a single feedstuff having commonality with two or more of the bacterial, archeal, protozoal, and fungal groups.

Acknowledgments

The authors thank Jerry M. Carter, Jeffrey B. Ellison, Kenneth N. Harless, J. Mark Peele (posthumously), Joyce M. Ruckle, and John P. Snuffer for their invaluable laboratory support and efforts. The authors are grateful to Drs. David Burner, Kim Cassida, Jorge Ferreira, Joyce Foster, Will Getz (posthumously), and Tom Terrill for supplying some of the plant samples used in this study. Mention of trade names or commercial products in this publication is solely for the purpose of providing specific information and does not imply recommendation or endorsement by the US Department of Agriculture. The US Department of Agriculture (USDA) prohibits discrimination in all its programs and activities on the basis of race, color, national origin, age, disability, and where applicable, sex, marital status, familial status, parental status, religion, sexual orientation, genetic information, political beliefs, reprisal, or because all or part of an individual’s income is derived from any public assistance program. USDA is an equal opportunity provider and employer.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Additional information

Funding

Funding for this project was provided from Congressional appropriations to the United States Department of Agriculture, Agricultural Research Service.