Supplementing branched-chain volatile fatty acids in dual-flow cultures varying in dietary forage and corn oil concentrations. I: Digestibility, microbial protein, and prokaryotic community structure

Branched-chain amino acids are deaminated by amy-lolytic bacteria to branched-chain volatile fatty acids (BCVFA), which are growth factors for cellulolytic bacteria. Our objective was to determine the dietary conditions that would increase the uptake of BCVFA by rumen bacteria. We hypothesized that increased forage would increase cellulolytic bacterial abundance and incorporation of BCVFA into their structure. Supple-mental polyunsaturated fatty acids, supplied via corn oil (CO), should inhibit cellulolytic bacteria growth, but we hypothesized that additional BCVFA would alleviate that inhibition. Further, supplemental BCVFA should increase neutral detergent fiber degradation and efficiency of bacterial protein synthesis more with the high forage and low polyunsaturated fatty acid dietary combination. The study was an incomplete block design with 8 dual-flow continuous cultures used in 4 periods with 8 treatments (n = 4 per treatment) arranged as a 2 × 2 × 2 factorial. The factors were: high forage (HF) or low forage (LF; 67 or 33%), without or with supplemental CO (3% dry matter), and without or with 2.15 mmol/d (which included 5 mg/d of 13 C each of BCVFA isovalerate, isobutyrate, and 2-methylbutyr-ate). The isonitrogenous diets consisted of 33:67 alfalfa: orchardgrass pellet, and was replaced with a concentrate pellet that mainly consisted of ground corn, soybean meal, and soybean hulls for the LF diet. The main effect of supplementing BCVFA increased neutral detergent fiber (NDF) degradability by 7.6%, and CO increased NDF degradability only in LF diets. Supple-mental BCVFA increased bacterial N by 1.5 g/kg organic matter truly degraded (6.6%) and 0.05 g/g truly degraded N (6.5%). The relative sequence abundance

Although increased NDF degradability and microbial N are commonly expected responses from BCVFA supplementation (Andries et al., 1987), research is needed to explain inconsistency among responses.For example, in a diet designed to induce milk fat depression, supplementation of BCVFA plus valerate alleviated milk fat depression, but it did not influence NDF degradability or N metabolism (Copelin et al., 2021).When 2-methylbutyrate was supplemented at 15 g/d to a 60:40 forage: concentrate or 40:60 forage: concentrate diet fed to Simmental steers (Wang et al., 2018), degradability of NDF increased, but the response was greater with the 60% forage diet.Roman-Garcia et al. (2021b) noted that BCVFA supplementation improved NDF degradability even when pH decreased and solids passage rate was increased.However, Roman-Garcia et al. (2021c) detected no change in efficiency of microbial protein synthesis (EMPS), even with increased relative abundance of key cross-feeding partners Fibrobacter succinogenes and Treponema, which both require BCVFA.
In comparison to the positive responses of BCVFA on cellulolytics, PUFA supplementation at high enough concentration can have bacteriostatic effects.In their meta-regression, Weld and Armentano (2017) estimated that NDF digestibility would decrease 1.3 percentage units with the addition of 3% CO to the diet.The main fatty acids (FA) of corn oil (CO) are 52% linoleic, 30% oleic, and 13% palmitic (Gunstone, 1996).The inclusion of linoleic acid inhibited growth of cellulolytic and some butyrate-producing bacteria (Maia et al., 2007).In contrast, oleic and palmitic acids supplemented at 1.5% DM increased total-tract NDF digestibility linearly with increasing oleic: palmitic ratio of FA (de Souza et al., 2021).The mode of action for either suppression of NDF degradability with linoleic acid or improvement with oleic and palmitic acid supplementation is not fully understood.
Our objective was to investigate the effect of BCVFA supplementation on EMPS under varying dietary conditions in dual-flow continuous cultures.Higher forage: concentrate was hypothesized to increase the benefit of BCVFA because of more fibrous substrate, but we also hypothesized that cellulolytic bacteria would benefit from BCVFA supplementation in lower forage diets because of their increasing competition with amylolytics for BCVFA precursors.A greater proportion of cellulolytic bacteria with high forage diets was projected to increase the conversion of BCVFA into bacterial components, so we hypothesized less BCVFA would be measured in outflow.Finally, the addition of PUFA was expected to be bacteriostatic, especially toward cellulolytic populations.We hypothesized for an interaction in which supplementation of CO without BCVFA would impede NDF degradability, especially with lower forage, but BCVFA addition would alleviate that PUFA inhibition.

Experimental Design and Treatments
The experiment was a 2 × 2 × 2 factorial arrangement of 8 treatments with 8 anaerobic dual-flow continuous culture vessels and 4 periods in a randomized incomplete block (n = 4).The vessels were randomly assigned to a treatment with either BCVFA supplementation (2.15 mmol/d isobutyrate, 2.15 mmol/d isovalerate, and 2.15 mmol/d 2-methylbutyrate) or no supplemental BCVFA, high forage: concentrate diet (HF,67:33) or low forage: concentrate diet (LF, 33:67), and either supplemental CO (additional 3% of DM) or no additional fat supplement.Roman-Garcia et al. (2021b) previously justified 2 mmol/d dosages of each BCVFA.Because of slightly higher intakes in our study, the dose was scaled to 2.15 mmol/d of each BCVFA to maintain the same ratio of BCVFA mmol/d to DMI.The diets (Table 1) were formulated with Spartan Ration Evaluator (v 3.0.3,Michigan State University, Department of Animal Science).The forage: concentrate ratios were determined to maximize forage NDF (35.2% DM) in the HF diets and minimizing starch (16.7% DM), and the LF were designed to minimize forage NDF (17.6% of DM) and maximize starch (31.7% of DM).All 4 diets were designed to be 15.0%CP by adding more soybean meal with the low forage concentrate pellets.The supplemental CO ranges from 50 to 60% linoleic and 20 to 30% oleic acids (Gunstone, 1996).Supplementation of CO at 3% DM would provide approximately a 1.5% DM dose of linoleic acid.Previous research on biohydrogenation have used primarily soybean oil or CO at 1.5 to 4% DM (Griinari et al., 1998;Rico and Harvatine, 2013).

Continuous Culture Operation
Two lactating Jersey cows were fed a standard lactation diet averaging 39.0% corn silage, 13.3% alfalfa hay, 20.0% ground corn, 9.9% wet brewer's grains, 7.7% soybean meal, 4.6% AminoPlus (AGP Ag Processing Inc.), 0.9% Megalac (Church and Dwight), 0.9% molasses, and the remaining 3.4% containing vitamins, minerals, and additives (no ionophores or other additives that would affect rumen microbial function; data not shown).Cows were housed and rumen contents sampled according to Institutional Animal Care and Use Committee standards at the Waterman Dairy in Columbus, Ohio.A total of 8 L of rumen contents were manually sampled through ruminal cannulas and squeezed through 2 layers of cheesecloth; the liquid was immediately transferred into prewarmed, insulated containers maintained at 39°C.Rumen fluid was pooled per period and then inoculated into fermenters at 50% of fermenter working volume, with the remainder being filled with 39°C reduced buffer just before adding rumen fluid.The buffer was made according to Weller and Pilgrim (1974), but 20 mg/dL of urea was added to maintain NH 3 -N concentrations above 5 mg/dL at all times postfeeding.The rumen fluid was added at 25% of the working volume in numerical order (fermenter 1 to fermenter 8) and then the final 25% was added in reverse (fermenter 8 to fermenter 1) to maintain uniformity.Treatments were randomized to fermenters.Clarified rumen fluid (centrifuged at 15,000 × g, 4°C, 15 min, and then autoclaved) was added to the buffer (1 L clarified rumen fluid: 20 L buffer) for the first day of each period to improve microbial adaptation to fermenters.Each period of the experiment was 12 d, with 8 d of acclimation and 4 d of sampling.Buffer was continuously infused into the continuous cultures at a rate of 10%/h on a volume basis; the filtrate outflow was 5%/h, and the overflow passage was 5%/h.Thus, liquid and solid passage rates of 10 and 5%/h, respectively, were maintained as in Roman-Garcia et al. (2021b).The temperatures of the vessels were maintained at 39°C, and the pH of the buffers being infused were adjusted, as described above for BCVFA, to remove differences among treatments.To minimize treatment effects on vessel pH, the LF buffer was adjusted about 0.1 pH units higher than the HF buffer with 10 N NaOH.All fermenters were checked hourly on d 1 and 2 of each period to maintain the pH of all vessels between 6.8 prefeeding and 6.0 at the nadir postfeeding (see details below).In subsequent days, the pH was stabilized, as verified by consistent monitoring.

Treatment and Isotope Supplementation
The cultures were fed 50 g DM/feeding and dosed their respective treatments (control buffer vs. BCVFA treatment) 2 times a day at 12-h intervals.Previously, because the BCVFA were prepared as acids, pH decreased by 0.06 units with BCVFA supplementation (Roman-Garcia et al., 2021b), whereas we adjusted the pH of the BCVFA treatment buffer with 10 N NaOH to minimize differences.After background sampling (d 5) was completed, [2-methyl-13 C]-2-methylbutyrate, [2,3-13 C 2 , 2-methyl-13 C]-isobutyrate, and [2,3,4-13 C 3 , 3-methyl-13 C]-isovalerate (Cambridge Isotope Laboratories, Inc.) were bolus-dosed and replaced a portion of the respective unlabeled source to provide 5 mg/d of 13 C from each BCVFA while maintaining the same molar dose of total (labeled and unlabeled) BCVFA.Additionally, 15 N-enriched (NH 4 ) 2 SO 4 was mixed into the buffer to continuously infuse 50 mg/d.All labeled sources were provided from d 6 through 12.

Sample Collection and Processing
Data collection was between the morning and evening feedings only.On d 3 to 7, immediately before feeding at h 0 and every hour thereafter, pH was measured for statistical analysis.Reduction/oxidation potential (ORP) was measured at the same time points as pH with a Pinpoint probe that was used according to directions of the manufacturer (American Marine Inc.).At h 12, before feeding, the total filtrate and overflow outputs were collected and weighed separately to measure filtrate, solids, and total outflow rates.
Effluents from the continuous cultures were collected on ice to prevent microbial activity outside of the dual-flow system.After combining filtrate and overflow after each 12-h feeding, duplicate 250-mL samples were subsampled from the total mixed outflow and pooled daily.One subsample was dosed with 8 mL of 5 N HCl to bring the pH down to 2 and stored at 4°C for 24 h for bacterial pellet recovery.The next day, particle-associated bacteria were dissociated from feed so that our bacteria pellet included liquid-and particle-associated bacteria (Whitehouse et al., 1994).The pooled pellet was analyzed for bacterial N and C by elemental analysis and 15 N and 13 C enrichment by isotope-ratio mass spectrometry (Roman-Garcia et al., 2021a).Supernatant during the processing of bacterial pellets was used for a colorimetric ammonia analysis (Chaney and Marbach, 1962).Another 20-mL sample of the supernatant was mixed with 5 mL of 50% TCA solution, and ammonia was diffused to filter disks before isotope-ratio mass spectrometry assay for 15 N enrichment (Hristov et al., 2001).
In the collection period (d 9 to 12), the continuous cultures were sealed for collection of gas data except when fed twice daily.A Micro-Oxymax Respirometer (Columbus Instruments Inc.) continuously monitored 4 vessels for CH 4 and H 2 emission rate and accumulation during the collection period (n = 2).Samples for VFA analysis were taken from the filtrate effluent line at 0.5, 1, 2, 4, 8, and 12 h in relation to the time of BCVFA dosage and feeding.The samples were processed and analyzed for VFA by GLC as described by Roman-Garcia et al. (2021a).At 1, 2, 4, 8, and 12 h, a 20-mL sample from the liquid effluent line was taken and analyzed for aqueous H 2 [herein referred to as H 2 (aq)] as described by Wenner et al. (2020).
Samples were collected for background and postdosing of isotopes and analyzed as described previously (Roman-Garcia et al., 2021b;Roman-Garcia et al., 2021c).A 5-mL sample of the total effluent from each 12-h interval was processed to measure total VFA production.Additional subsamples that were approximately 15% of the total effluent volume for each 12-h period were frozen, later pooled over the collection period, and dried at 55°C.A subsample of the effluent was dried at 105°C overnight and then ashed at 550°C overnight to determine DM and OM.Residual NDF in the effluent was determined as described by Van Soest et al. (1991) using amylase and sulfite.The effluent was also analyzed for 15 N and 13 C enrichments and N and C composition as previously described.Samples of diet ingredients were also analyzed for DM, OM, NDF, and N with the same methods.The feed samples were also analyzed by Cumberland Valley Analytical for starch (Hall, 2009), ether extract (AOAC International, 2000;method 2003.05) and FA (Sukhija and Palmquist, 1988).

Mitchell et al.: BRANCHED-CHAIN VOLATILE FATTY ACIDS AND RUMEN BACTERIA
Additionally, duplicate 30-mL subsamples from the effluent were frozen at −80°C before pooling the 12-h samples within period.Genomic DNA was extracted using repeated bead beating plus column purification (Yu and Morrison, 2004) followed by sequencing of the V4-V5 hypervariable region with the Illumina MiSeq platform (Illumina Inc.).Computational analysis was performed as described by Faulkner et al. (2017) except that amplicon sequence variants (ASV) were recovered rather than operational taxonomic units, as described in Lee et al. (2021), and for Bray-Curtis dissimilarity, which was analyzed as described by Wenner et al. (2020).

Calculations
After adjusting the 15 N enrichment for background (sampled before isotope dosing), the atom percentage excess (APE) was used in the calculations described by Roman-Garcia et al. (2021b).Bacterial N flow was calculated as total N flow × 15 N APE of effluent/ 15 N APE of bacteria.The bacterial N derived from NH 3 -N was calculated as 15 N APE of bacteria/ 15 N APE of diffused ammonia.The NH 3 -N flow was the product of NH 3 -N concentration in fluid and the total fluid effluent flow.The dietary RUP proportion was estimated as nonammonia-nonbacterial N (NANBN) flow divided by the nonammonia-N (NAN) flow and converted to a percentage.The RDP was estimated as RUP subtracted from dietary CP.Net VFA production was the measured VFA production minus the amount of BCVFA dosed.Flow of C from effluent was the product of effluent percent C and effluent DM flow.The flow of bacterial C was calculated by multiplying the flow of bacterial N by the ratio of C: N in the bacteria pellets.Recovery of 13 C was calculated by multiplying the 13 C APE/100 of effluent or bacteria pellet by the respective C flow.Dose recovery was the micrograms of 13 C recovered in bacterial C per the milligrams of 13 C dosed.

Statistical Analysis
Daily collected VFA production, digestibility, and flow measurements were analyzed using PROC MIXED in SAS 9.4 (SAS Institute Inc.) according to this model: where Y ijtlk is the dependent variable; µ is the overall population mean, D i is the fixed effect of diet (i = HF or LF), C j is the fixed CO supplementation (j = − CO or + CO), B t is the fixed effect of BCVFA supplementation (t = − BCVFA or + BCVFA), (D × C) ij , (D × B) iw , (C × B) jw , (D × C × B) ijw are the respective interactions of the main effects; f l = is the random effect of lth vessel (l = 1 to 8); p k is the random effect of kth period (k = 1 to 4); and e ijtlk is the random error.Recovery of 13 C dose was analyzed with the same model, but an additional covariate was included because LF had a higher calculated RDP (as explained above) than HF, which would influence BCVFA precursor availability; therefore, BCVFA production (mmol/d) was included as the covariate to account for differing degraded BCAA precursor availability.
Hourly data were analyzed using the same model but including the REPEATED (repeated measures) statement to assess the fixed effect of hour (T h ) and all its interactions with the other main effects.The heterogeneous first-order autoregressive covariance structure (hourly pH and ORP) or the spatial power [for unequally spaced VFA, H 2 (aq), and gas production] covariate structures were used based on the lowest Corrected Akaike Information Criterion.If there was an interaction with time, treatment means per hour were reported and contrasted at each time using the SLICE statement.Differences were declared at P ≤ 0.05, and trends were 0.05 < P ≤ 0.10.

pH and Reduction/Oxidation Potential
We detected no main effect differences on pH (P > 0.41), which demonstrates the success of buffer adjustment with different forage: concentrate.Diet and CO tended to interact (P = 0.09) for pH.In the HF -CO, HF + CO, LF -CO, and LF + CO treatments, mean pH was 6. 37, 6.41, 6.41, and 6.40, respectively (data not shown); thus, these changes are deemed inconsequential to our objectives.We observed a diet × hour interaction for pH, which is shown in Figure 1 (panel a).Only at h 0 (P = 0.02), 9 (P = 0.07), and 12 (P = 0.04) was the pH from LF treatments greater than that of HF by 0.03 units, which again was deemed inconsequential.In contrast with pH, we detected main effect differences for ORP (data not shown).Decreased forage increased (P < 0.01) ORP (−318 mV for HF vs. −310 mV for LF).When CO was supplemented, ORP decreased (P = 0.01) from −313 without CO compared with −315 with CO.There was also a diet × hour interaction (P = 0.05) for ORP (Figure 1, panel b).At all time points, LF had a higher ORP than HF, but diet differences were less at h 0, 1, and 12 h postfeeding, which is when degradation of new feed was minimal.

Nutrient Degradation and Flow of Nitrogen
The overflow and liquid passage rate ranged from 4.95 to 5.09%/h and 9.81 to 9.93%/h, respectively, and were not affected by any main effect or interactions (P > 0.20, data not shown).The LF diet had 6.7% greater true OM degradability than HF (P < 0.01, Table 2).Apparent starch degradability also tended to increase (P = 0.10) from 98.7% with HF to 99.2% with LF.We detected a trend (P = 0.10) for the main effect  ) or low forage (LF) diets, no supplemental fat (− CO) or supplemental corn oil (+ CO), and either no supplemental branched-chain VFA (-BCVFA) or supplemental BCVFA (+ BCVFA).Data are means from all 4 experimental periods taken on d 3 to 7 of the period.For pH, there was a diet × hour interaction (P = 0.07), but pH only differed by diet at h 0, 9, and 12 when pH was higher (P < 0.08, designated by * or P < 0.05 designated by **) with the main effect of LF compared with HF.For ORP, there was a diet × hour interaction (P = 0.05), but ORP was higher (P < 0.05; designated by **) for LF than HF at all time points.The SEM is pooled across treatment per time and shown for only 1 series per plot.
of BCVFA supplementation to increase NDF degradability by 3.0 percentage units (i.e., 7.6%).We observed no interactions with BCVFA supplementation.We detected a trend (P = 0.07) for an interaction of diet × CO, whereby CO supplementation in the LF diet increased (P = 0.02) NDF degradation by 6.5 percentage units, but CO supplementation with HF did not influence NDF degradation (P = 0.98).
We observed a tendency for a 3-way interaction for total N flow (P = 0.07, Table 2).The flow of NH 3 -N only differed by diet, and flow NH 3 -N flow decreased (P < 0.01) from 0.349 g/d with HF to 0.251 g/d with LF.The concentration of NH 3 -N decreased (P < 0.01) from 8.38 mg/dL with HF to 6.15 mg/dL with LF.For the flow of NAN, there was also a tendency for a 3-way interaction (P = 0.07) resulting from the 3-way interaction (P = 0.02) of bacterial N, which is a component of NAN.Bacterial N flow increased with BCVFA supplementation, mainly due to an increase when supplemented in the HF -CO diet from 1.14 g/d (-BCVFA) to 1.38 g/d (+ BCVFA).Numerically bacterial N flow increased with BCVFA supplementation with LF -CO and LF + CO, but it numerically decreased with BCVFA supplementation with the HF + CO.Bacterial C flow also had a 3-way interaction (P < 0.01), which was similar to bacterial N flow.Supplementation of BCVFA increased bacterial C flow in the HF -CO, LF -CO, and LF + CO treatments, but bacterial C flow was decreased when BCVFA was supplemented to HF + CO.The flow of NANBN decreased (P < 0.01) from 0.710 g/d with HF to 0.639 g/d with LF.Therefore, the 3-way interaction of total N flow is due to the 3-way interaction of bacterial N flow.The ratio of bacterial C: N decreased (P < 0.01) from 5.30 for HF to 5.07 with LF, but the bacterial N that originated from NH 3 -N was not influenced by any main effects (P > 0.14).The RUP (i.e., NANBN / NAN × 100) decreased (P < 0.01) from 36.3% with HF to 33.5% with LF; RDP (i.e., 100 -RUP) increased from 63.7% with HF to 66.5% with LF.Bacterial N per kilogram truly degraded OM was greater (P < 0.01, Table 2) with the main effect of HF (25.1 g/kg) compared with LF (23.2 g/kg), and it also increased (P = 0.03) with the main effect of supplementing BCVFA from 23.4 to 24.9 g/ kg of OM truly degraded.Supplementation of BCVFA increased (P = 0.02) bacterial N by 0.05 g/g of truly degraded N.

Net Production of VFA, Methane Emission, and Aqueous Hydrogen
For total VFA net productions (Table 3), we detected an interaction of diet × BCVFA (P = 0.05).We observed a 2.87% decrease in net production in the HF treatments when BCVFA was supplemented.However, when BCVFA was supplemented with LF, net production decreased by about 11.4%.Acetate and propionate production decreased (P ≤ 0.02) by 11.4 and 5.53% with the main effect of BCVFA supplementation.Propionate production increased (P < 0.01) by 27.4% with LF compared with HF.Acetate: propionate decreased (P < 0.01) from 3.51 with HF to 2.78 with LF and decreased (P < 0.09) from 3.25 without CO to 3.03 with CO (all main effect differences).For butyrate, valerate, and caproate production, we detected trends for interactions (P ≤ 0.10) of diet × BCVFA.When BCVFA was supplemented with LF treatments, butyrate and valerate production decreased more than when BCVFA was supplemented to HF.The trend (P = 0.10) for a diet × BCVFA in caproate production was the opposite; supplementation of BCVFA in the HF treatments increased caproate production but decreased caproate in the LF treatments.The concentrations of the major and minor VFA in the vessels over time are shown in Supplemental Figures S1, S2, and S3 (https://doi.org/10.6084/m9.figshare.22300714.v1,Mitchell et al., 2023).
Net production of BCVFA is the total daily production minus their daily sum of both 12-h doses (Table 3).Isobutyrate (P = 0.04), 2-methylbutyrate (P < 0.01), and isovalerate (P = 0.09) had interactions with diet in which supplementation of BCVFA decreased the net production by twice or more in the LF diet compared with the decreased net production in the HF diet, which is similar to total VFA production described previously.Total BCVFA net production also had a similar interaction of diet with BCVFA (P = 0.01).The concentrations of individual BCVFA and total BCVFA over time are shown in Supplemental Figures S4 and  S5 (https://doi.org/10.6084/m9.figshare.22300714.v1,Mitchell et al., 2023).
Compared with the main effect of HF, LF increased (P < 0.01) H 2 emission (Table 3).This increase is supported by H 2 (aq) concentrations (Supplemental Figure S6, https://doi.org/10.6084/m9.figshare.22300714.v1,Mitchell et al., 2023) and H 2 production rate expressed over time (Supplemental Figure S7, https://doi.org/10.6084/m9.figshare.22300714.v1,Mitchell et al., 2023) because of diet × time interactions (P < 0.01).The LF H 2 (aq) concentrations and H 2 production rates were only greater than HF during the first 4 h after feeding after which treatments converged as remaining degradable substrate declined.Methane production (Table 3) did not differ by diet (P = 0.48), but we detected trends for diet × BCVFA (P = 0.10) and CO × BCVFA (P = 0.09) interactions.With BCVFA supplementation, CH 4 production numerically increased in the HF diet from 91.9 to 102 mmol/d, whereas with   LF, the supplementation of BCVFA decreased CH 4 production from 114 to 82.6 mmol/d.In contrast, BCVFA supplementation without CO decreased CH 4 production from 111 to 87.9 mmol/d but did not affect CH 4 production with CO.Methane production rate (Supplemental Figure S7) also had diet × time and diet × CO × time interactions (P = 0.02).The differences by diet were within h 3 to 7 because CH 4 production rate peaked later for HF than LF and decreased at a slower rate than that for LF.

Isotope Recovery
The 13 C from dosed BCVFA as recovered in the total effluent includes bacteria, recycled products of microbial metabolism not retained in bacteria, and unmetabolized dose.(Table 3).We observed no difference (P ≥ 0.14) in the recovery of dosed 13 C from BCVFA from the total effluent flow, but recovery with HF was numerically greater than with LF.Recovery in bacterial C outflow decreased (P = 0.03) from 144 µg/ mg with HF to 98.9 µg/ mg with LF.Therefore, recovery of nonbacterial 13 C in total effluent did not differ (P > 0.33) by treatment, either, and ranged from 787 to 703 µg/ mg.Recoveries of 13 C in individual branchedchain lipids and BCAA will be reported in Mitchell et al. (2023a,b).

Prokaryotic Profile Changes
Diet overwhelmingly influenced α diversity indices (Table 4).The main effect of LF decreased (P < 0.04) most indices compared with the main effect of HF.Only Simpson's Index was unaffected by treatment.Good's Coverage ranged from 0.996 to 0.977.For β diversity, shown with a Bray-Curtis distance matrix plot in Figure 2, only diet (P < 0.01, panel a) and period (P < 0.01, panel b) supported major shifts discussed for numerous taxa reported subsequently and the importance of initial inoculation that differed among periods, respectively.
The relative percentages of ASV were combined into genus and phylum ranks (Table 4).Not all ASV can be annotated or met the criterion (at least 0.5% relative abundance for an observation) by our approach.However, when discernible, if sequences from only 1 ASV made up the entire genus or if a single genus made up the entire phylum, then only the lowest taxonomic rank was shown to avoid duplication.Bacteroidetes, the phylum in highest abundance, decreased (P < 0.01) from 56.3% with HF to 41.9% of total sequences with LF.The genus BF311 (within the Bacteroidetes phylum) decreased (P < 0.01) with decreased forage.Additionally, supplementation of BCVFA with HF tended to decrease (P = 0.06, diet × BCVFA) the relative abundance of BF311.The major genus within phylum Bacteroidetes was Prevotella, which was not affected by diet (P = 0.28), although we detected a shift among the species resolved within Prevotella.The characterized P. ruminicola and one uncharacterized ASV belonging to Prevotella increased (P ≤ 0.04) with LF, whereas the other predominant uncharacterized Prevotella ASV decreased in relative sequence abundance.The CF231, YRC22, and another uncharacterized genera decreased (P < 0.01) for LF versus HF.The phylum Fibrobacteres also decreased (P < 0.01) from 0.31% with HF to 0.14% with LF, as also documented for Fibrobacter succinogenes, which was the only annotated ASV in this phylum.
Diet did not influence Firmicutes relative abundance (Table 4), but this is a diverse phylum that contains the order Clostridiales and many genera within that order that did shift by diet.Butyrivibrio decreased (P < 0.01) with decreased forage and decreased (P = 0.10) with supplemented CO.A trend for a diet × CO interaction (P = 0.10) was detected for genus Ruminococcus from which relative abundance was lower with HF -CO versus HF + CO but decreased when CO was supplemented with LF diets.This was mainly due to a shift with R. albus, which also had a diet × CO interaction (P = 0.03) explained by increased abundance with CO supplementation in the HF diet but decreased relative abundance with CO supplementation in the LF diet.Relative abundance of R. bromii and R. flavefaciens had opposite trends for diet × BCVFA interactions (P < 0.10).The relative sequence abundance of R. bromii increased with BCVFA supplementation in the HF diet but decreased with BCVFA supplementation with LF, whereas R. flavefaciens decreased with BCVFA supplementation in the HF diet but increased with BCVFA supplementation with LF.We observed few main effects of CO on relative abundance.However, Anaerovibrio increased (P < 0.01) with LF compared with HF and increased (P = 0.02) when CO was supplemented.
The phylum Proteobacteria increased (P < 0.01) in relative sequence abundance from 5.79% with HF to 18.7% with LF; much of this shift occurred in an unknown genus in the Succinivibrionaceae family, which increased from 4.24% with HF to 10.1% with LF (Table 4).The abundance of Ruminobacter increased (P < 0.01) with additional CO but did not change with forage (P = 0.26).Genus Succinimonas, which consisted only of S. amylolytica, increased (P < 0.01) with supplemented CO and increased (P = 0.01) with decreased forage.Genus Succinivibrio also increased (P = 0.01) with decreased forage.The phylum SR1 decreased (P < 0.01) with decreased forage.(-CO) or supplemental corn oil (+ CO), and either no supplemental branched-chain VFA (-BCVFA) or supplemental BCVFA (+ BCVFA) The abundances of Spirochaetes, TM7, or Verrucomicrobia phyla did not change by diet (Table 4).Within Spirochaetes, relative sequence abundance of Treponema decreased (P = 0.05) with CO, but no genera or species abundance within Spirochaetes, TM7, or Verrucomicrobia were influenced by any dietary conditions (P > 0.18).The relative abundance of phylum Tenericutes decreased (P < 0.01) with decreased forage.The unknown phylum had a trend for a diet × CO interaction (P = 0.06) in which supplemental CO decreased the abundance moderately with LF, but more so when CO was supplemented with the HF diet.

DISCUSSION pH and Reduction/Oxidation Potential
Although by only 0.008 mV, there was a highly significant decrease in ORP between HF than LF despite the successful minimization of pH differences under different diet conditions.Roman-Garcia et al. (2021b) fed the same diets but altered pH of buffers infused into continuous cultures and reported that lower pH increased ORP (became less negative) by about 0.028 mV.Both pH and ORP are influenced by proton-coupled reduction/oxidation reactions (Krishtalik, 2003).As with pH, measurement of ORP depends on the location in the rumen and the sampling method (Huang et al., 2018).The measurement of ORP is sensitive to air contamination, further complicating its measurement.The ORP has been positively correlated with DMI and the concentrate proportion of the diet but negatively correlated with pH, proportion of acetate, and H 2 production (Baldwin and Emery, 1960;Huang et al., 2018;Kelly et al., 2022).The ORP is reflective of intracellular ORP status (Liu et al., 2013).The NAD + :NADH regulates fermentation pathways and influences acetate: propionate and H 2 formation via hydrogenases (van Lingen et al., 2016).

Nutrient Degradation
Dual-flow continuous cultures have very high starch degradability because of settling of starch granules below the overflow port, as explained by Roman-Garcia et al. (2021b).Because of increased starch inclusion, OM degradability also was higher with LF.Though greater inclusion of starch inhibits NDF degradability, increasing starch is also positively associated with increased apparent ruminal starch digestibility (White et al., 2016).Degradability of NDF should be higher with LF diets because of increased inclusion of NDF sources such as soybean hulls that are more degradable than forage NDF.
We expected that CO, especially with the LF diet, would have a bacteriostatic effect on cellulolytics because there is less fibrous surface area to disperse PUFA (Jenkins et al., 2008), but the interaction of diet × CO alleviated a limitation in NDF degradability in the LF -CO diet.The 3% additional CO, which provided approximately 1.5% of DM as linoleic acid, was not designed to be excessive but still has decreased total-tract NDF degradability moderately (Weld and Armentano, 2017).Though other studies with similar or lower linoleic FA inclusion depressed milk fat, those results have been with other factors such as low forage diets, which would decrease pH in vivo (Baldin et al., 2018).The effect of low forage is magnified when combined with decreased particle size, which in addition to decreasing effective NDF would increase passage rate from the rumen compared with coarser forage (Ramirez Ramirez et al., 2016).In our study with controlled pH and passage rate, cellulolytic bacteria presumably had less inhibition with additional CO supplementation compared with potential indirect effects in vivo.Moreover, the benefit of a tendency for increased NDF degradability from CO supplementation with LF could be due to the inclusion of oleic and palmitic FA included in corn oil, which has benefited NDF degradability in vivo (de Souza et al., 2021).Mitchell et al. (2023a) discussed the potential for oleic acid to be incorporated into membranes and the need for more fluidity as cellulolytic bacteria degrade cellulose.However, with the HF diet, we detected no increase in NDF degradation with CO, likely due to a lower-than-expected FA increase with the HF diet when CO was added (Table 1).Therefore, the increased NDF degradation with CO + LF probably resulted from greater provision of oleic and palmitic acids than CO supplementation with HF diet.
Increased NDF degradation is a frequently reported benefit of BCVFA supplementation (Liu et al., 2018;Wang et al., 2019;Roman-Garcia et al., 2021a).Roman-Garcia et al. (2021b) reported a 5.2 percentage unit increase in NDF degradability in continuous cultures with no interactions of BCVFA supplementation with the other factors: high or low pH and high or low passage rate.Wang et al. (2018) noted an interaction of diet and 2-methylbutyrate in which the moderateconcentrate diet benefited from supplementation more than the high-concentrate diet in respect to effective degradability of NDF from corn silage measured in situ.In our study, we expected an interaction of forage and BCVFA because BCVFA was likely in greater demand when fermenters were administered the HF diet due to the higher relative abundance of populations with known BCVFA requirements or high proportions of branched-chain lipids in their structure, including genus Butyrivibrio, F. succinogenes, and R. flavefaciens (Stewart et al., 1997).The LF treatments had a higher abundance of P. ruminicola.Although many strains do not require BCVFA, Prevotella brevis was characterized as requiring 2-methylbutyrate (Dehority, 1966).However, with the LF diet, the cellulolytics, even though they are a smaller proportion of the population, likely have more competition for BCVFA with amylolytics some of which do not require them but can use them (Firkins, 2010).In our study, the amylolytic Ruminococcus bromii increased prevalence with increased grain, whereas the cellulolytic R. flavefaciens decreased prevalence.Both require BCVFA, so the opposing interactions of diet with BCVFA suggests that competition for BCVFA probably increased when there was increased overall carbohydrate degradability in the LF treatment (Herbeck and Bryant, 1974;Stewart et al., 1997).The lack of interactions between BCVFA and the other dietary treatments on NDF degradability could indicate a general need for adequate BCVFA under these varying conditions (not just with HF) or could be a result of lack of statistical power to detect interactions.

Bacterial Protein Synthesis
Both bacterial N and C flows had 3-way interactions, but a decrease in the C:N ratio with LF compared with HF would be due to either decreased glycogen storage (decreased C) or increased N concentration from a different bacterial community.Although decreased growth rate from N deficiency typically would increase glycogen, the cycling of glycogen also could prevent its accumulation (Hackmann and Firkins, 2015).A more likely explanation is a shift in bacterial populations, type of cell wall, or a shift in nucleic acid -N: AA -N (Arambel et al., 1982).Notably, Proteobacteria, especially an unknown genus in the Gammaproteobacteria class, increased by a factor of 10 with LF compared with HF, primarily displacing Bacteroidetes.The EMPS differences by diet likely are reflective of amylolytic bacteria having higher maintenance coefficients than fibrolytics (Fox et al., 2004).
We noted few shifts in relative sequence abundance with BCVFA supplementation; therefore, the reported increases in bacterial N outflow were due to an overall increase in bacterial population without major shifts in the prokaryotic profile.Actual abundance of total bacteria (direct count and cultivation by roll tube) and cellulolytic bacteria (cultivation and qPCR targeting Ruminococcus albus, R. flavefaciens, and F. succinogenes) in the rumen increased with supplementation of isobutyrate (Wang et al., 2015), 2-methylbutyrate  (Zhang et al., 2015), and isovalerate (Liu et al., 2014) to steers.Additionally, 2-methylbutyrate supplementation to Simmental steers increased the amount of purine derivatives excreted in urine (indicator of microbial N flow out of the rumen) but to a greater degree with the moderate-concentrate diet compared with a high-concentrate diet (Wang et al., 2018).Changes in forage: concentrate also changed feed BCAA profile; in particular, increased inclusion of corn protein increases availability of Leu.The availability of cytosolic free BCAA was expected to influence the benefits observed with BCVFA supplementation (Firkins, 2021); therefore, the effect of protein supply on microbial N or EMPS benefits with BCVFA in vivo requires further investigation.

Volatile Fatty Acid Production, Gas Production, and Isotope Recovery
When pH was decreased with the same forage: concentrate, Roman-Garcia et al. (2021b) reported decreased acetate: propionate.In addition to differences in pH, decreased forage: concentrate ratio also should increase propionate production by amylolytic bacteria, as demonstrated by our prokaryotic community analysis.For example, P. ruminicola, Succinimonas amylolytica, and genus Succinivibrio increased with LF.Because they all can produce succinate (Stewart et al., 1997), succinate decarboxylation should increase propionate.With supplemental CO, the acetate: propionate ratio tended to decrease.Though we observed very minor influences of CO on the prokaryotic profile, both genus Ruminobacter and S. amylolytica increased with CO.Both also form succinate.Additionally, Coprococcus, which converts lactate to propionate (Sheridan et al., 2022), increased abundance with CO supplementation.
Though LF increased the abundance of succinate producers, there was not an increase in characterized Selenomonas or Succiniclasticum, which are core bacteria associated with the succinate usage niche (Mizrahi et al., 2021).Regardless of which bacteria were responsible, increased propionate production should have been inversely associated with methane emission, which was not affected in the current study.In contrast, the increased H 2 emission for LF versus HF, as supported by elevated H 2 (aq) after feeding, suggests a disruption of methanogenesis or decreased activity of alternate H 2 sinks that might have been available with the higher carbohydrate degradability of LF.Hydrogenotropic bacteria probably have lower affinity for H 2 (aq) than methanogens and therefore allow more H 2 emission (Ungerfeld, 2020).Physical interactions between fibrolytic H 2 producers and archaea (Piao et al., 2014) might have been lessened in these first hours after feeding LF because of the increased abundance of amylolytic bacteria (especially Proteobacteria).
With BCVFA supplementation, the lower acetate and propionate production with both HF and LF is likely a result of increased conversion of degraded C into bacterial C.However, CH 4 production numerically increased when BCVFA was supplemented with HF but decreased when supplemented with LF.Roman-Garcia et al. (2021b) explained increased CH 4 production with BCVFA by the increased NDF degradability but also the potential BCVFA requirements by archaea being met.
Based on the diet formulation program, HF was predicted to have higher RDP than LH (Table 1); however, the opposite was observed (Table 2).Increased abundance of some proteolytic bacteria, such as P. ruminicola, with LF could account for the difference in RDP compared with expectations.The higher concentrations of BCVFA in the vessels could decrease deamination or decarboxylation due to some sort of feedback inhibition or could be regulating gene expression of AA biosynthesis genes, as shown for isovalerate in Prevotella bryantii (Trautmann et al., 2020) and discussed further by Mitchell et al. (2023b).
The BCVFA production from the vessels were used as a covariate in the model to correct for diet differences in diet AA supply (more leucine with LF) and RDP (greater RDP with LF) resulting from greater BCVFA precursor availability with LF versus HF.We detected no differences for isotope recovery in the total effluent.However, with HF treatments, bacteria increased incorporation of 13 C-BCVFA into their cellular components by 45.6% more than bacteria with LF treatments, suggesting that HF increased the relative abundance of populations with known BCVFA requirements.Roman-Garcia et al. (2021c) noted that many populations with BCVFA requirements decreased with low pH, but we observed no difference in isotope recovery in bacteria.The pH was held constant in our study, supporting the forage: concentrate has a greater role than pH (without acidosis) for BCVFA uptake.Treatment differences for 13 C recovered in lipids and AA are detailed in our companion studies (Mitchell et al., 2023a,b).

CONCLUSIONS
Even though net production of BCVFA decreased with lower forage, isotope recovery of the dosed BCVFA did not decrease, supporting greater bacterial uptake of BCVFA with increasing fiber in the diet.Although we expected that supplemental BCVFA would be more beneficial with greater forage inclusion, additional BCVFA increased NDF degradability and EMPS, re-gardless of substrate supply and without shifting bacterial populations.Although many genera shifted up or down within phylum Firmicutes, the total relative abundance of Firmicutes was not affected by forage: concentrate.However, decreasing forage: concentrate decreased Bacteroidetes but increased Proteobacteria (mainly of characterized amylolytic bacteria).A decreased forage: concentrate probably increases the competition between cellulolytics and amylolytics for common growth factors, so providing supplemental BCVFA are projected to improve bacterial protein supply and feed efficiency by lactating dairy cows fed a variety of forages and concentrates.Unexpectedly, supplementing CO increased NDF degradability, perhaps by providing more oleic acid for bacterial membranes, perhaps explaining lack of inhibition in cellulolytic populations of bacteria.
Figure 1.Fermenter pH (panel a) and oxidation/reduction potential (ORP, panel b) over time postfeeding for continuous cultures that had treatments of high (HF)  or low forage (LF) diets, no supplemental fat (− CO) or supplemental corn oil (+ CO), and either no supplemental branched-chain VFA (-BCVFA) or supplemental BCVFA (+ BCVFA).Data are means from all 4 experimental periods taken on d 3 to 7 of the period.For pH, there was a diet × hour interaction (P = 0.07), but pH only differed by diet at h 0, 9, and 12 when pH was higher (P < 0.08, designated by * or P < 0.05 designated by **) with the main effect of LF compared with HF.For ORP, there was a diet × hour interaction (P = 0.05), but ORP was higher (P < 0.05; designated by **) for LF than HF at all time points.The SEM is pooled across treatment per time and shown for only 1 series per plot.

2P
values for main effect means for diet (HF vs. LF), CO, BCVFA or interactions; interactions not shown have P > 0Mitchell et al.: BRANCHED-CHAIN VOLATILE FATTY ACIDS AND RUMEN BACTERIA
Mitchell et al.: BRANCHED-CHAIN VOLATILE FATTY ACIDS AND RUMEN BACTERIA

Table 3 .
Net production of VFA, gas production, and dose recovery in continuous cultures that were administered with either high or low forage diets that varied in corn oil and branched-chain VFA supplementation 2P-values for main effect means for diet (HF vs. LF), CO, BCVFA or interaction; interactions not shown have P > 0.10.3NetVFAproduction was the measured VFA production minus the amount of BCVFA dosed.4Interaction of CO and BCVFA supplementation has P ≤ 0.09.

Table 4 .
Mitchell et al.: BRANCHED-CHAIN VOLATILE FATTY ACIDS AND RUMEN BACTERIA Mitchell et al.: BRANCHED-CHAIN VOLATILE FATTY ACIDS AND RUMEN BACTERIA Alpha diversity indices and prokaryotic profile in continuous cultures that were treated with either high (HF) or low forage (LF) diets, no supplemental fat (-CO) or supplemental corn oil (+ CO), and either no supplemental branched-chain VFA (-BCVFA) or supplemental BCVFA (+ BCVFA) main effect means for diet (HF vs. LF), CO, BCVFA or interactions; interactions not shown have P > 0.10.Phyla are followed by genera and by species when classification is possible; if either a phylum or genus contains a single species, only the relative sequence abundance at the species rank is shown.

Table 4 (
Continued).Alpha diversity indices and prokaryotic profile in continuous cultures that were treated with either high (HF) or low forage (LF) diets, no supplemental fat Mitchell et al.: BRANCHED-CHAIN VOLATILE FATTY ACIDS AND RUMEN BACTERIA