Can dietary magnesium sources and buffer change the ruminal microbiota composition and fermentation of lactating dairy cows?

Abstract

Magnesium oxide (MgO) is one of the most used Mg supplements in livestock. However, to avoid relying upon only one Mg source, it is important to have alternative Mg sources. Therefore, the objective of this study was to evaluate the effects of the interaction of two Mg sources with buffer use on the ruminal microbiota composition, ruminal fermentation, and nutrient digestibility in lactating dairy cows. Twenty lactating Holstein cows were blocked by parity and days in milk into five blocks with four cows each, in a 2 × 2 factorial design. Within blocks, cows were assigned to one of four treatments: 1) MgO; 2) MgO + Na sesquicarbonate (MgO+); 3) calcium–magnesium hydroxide (CaMgOH); 4) CaMgOH + Na sesquicarbonate (CaMgOH+). For 60 d, cows were individually fed a corn silage-based diet, and treatments were top-dressed. Ruminal fluid was collected via an orogastric tube, for analyses of the microbiota composition, volatile fatty acids (VFA), lactate, and ammonia nitrogen (NH₃–N). The microbiota composition was analyzed using V4/16S rRNA gene sequencing, and taxonomy was assigned using the Silva database. Statistical analysis was carried out following the procedures of block design analysis, where block and cow were considered random variables. Effects of Mg source, buffer, and the interaction between Mg Source × Buffer were analyzed through orthogonal contrasts. There was no interaction effect of the two factors evaluated. There was a greater concentration of NH₃–N, lactate, and butyrate in the ruminal fluid of cows fed with CaMg(OH)₂, regardless of the buffer use. The increase in these fermentation intermediates/ end-products can be explained by an increase in abundance of micro-organisms of the genus Prevotella, Lactobacillus, and Butyrivibrio, which are micro-organisms mainly responsible for proteolysis, lactate-production, and butyrate-production in the rumen, respectively. Also, dietary buffer use did not affect the ruminal fermentation metabolites and pH; however, an improvement of the apparent total tract digestibility of dry matter (DM), organic matter (OM), neutral fiber detergent (NDF), and acid fiber detergent (ADF) were found for animals fed with dietary buffer. In summary, there was no interaction effect of buffer use and Mg source, whereas buffer improved total tract apparent digestibility of DM and OM through an increase in NDF and ADF digestibility and CaMg(OH)₂ increased ruminal concentration of butyrate and abundance of butyrate-producing bacteria.

Magnesium oxide is the most used source of magnesium (Mg) in livestock diets; however, dairy operations can benefit from other Mg sources, such as calcium–magnesium hydroxide (CaMg(OH)₂). This study demonstrated that CaMg(OH)₂ can improve ruminal fermentation by modulating ruminal microbiota composition, regardless of ruminal buffer use.

Introduction

Magnesium (Mg) plays a key role in dairy cows’ metabolism, it is essential in bone formation, nerve function, muscle function, and enzymatic reactions (Katz and Miledi, 1967; Rude et al., 1999; Gordon et al., 2000). Supplementation with Mg is fundamental to maximize livestock production, especially milk yield (NASEM, 2021). In livestock, the main sources of dietary Mg supplementation consist of metal oxides, such as Mg oxide (MgO; (Goff, 2018; NASEM, 2021). Recently, efforts have been made to find alternative sources of Mg that can be used in diet supplementation, such as Mg hydroxides (Mg(OH)₂) and carbonates (Mg(CO₃)₂; Arce-Cordero et al., 2021a; Agustinho et al., 2022; Lobo et al., 2023). Since most MgO is from China and the need of reducing pollution has forced the closure of several manufacturing plants in the country in the past few years, searches for alternatives to MgO have driven the Mg market.

Oxides and hydroxides of Mg are a class of oxygen-bearing metals, in which Mg is bound to oxygen or hydroxyl group (OH), respectively. Both oxides and hydroxides metals have shown antimicrobial activity (Kumar et al., 2011; Raghunath and Perumal, 2017; Eivazzadeh-Keihan et al., 2021). He et al. (2016) reported that MgO can kill bacteria by generating cellular oxidative stress, disrupting the cell membrane integrity. On the other hand, hydroxides released by metal hydroxide ionization alter the respiratory electron transport chain, located in the bacterial cell membrane, by reacting with the H⁺ available for adenosine triphosphate production, which reduces the H⁺ available and consequently reduces bacterial ATP production, reducing bacterial growth (Tan et al. 2018, 2020). Therefore, it is possible that Mg oxides and hydroxides have different effects on ruminal bacteria.

Because oxides and hydroxides sources of Mg have been used as feed supplements for ruminants, it is important to understand their effects on the fermentation kinetics and ruminal microbiota composition. Arce-Cordero et al. (2021a, 2021b) and Agustinho et al. (2022) studied the effects of different sources of Mg on ruminal fermentation kinetics using an in vitro system. More specifically, Mg hydroxide sources increased the average pH and butyrate concentration in vitro (Arce-Cordero et al., 2021b). Also, Arce-Cordero et al. (2022) reported that a greater relative abundance of Butyrivibrio and Lachnospiraceae genera were observed when diets were supplemented with hydroxide and oxide sources of Mg, compared to carbonate sources. To our knowledge, there has been only one published study that evaluated, in vitro, the impact of Mg sources on ruminal fermentation and the microbiota composition (Arce-Cordero et al., 2022); therefore, in vivo studies are warranted to better understand the effects of different Mg sources on the ruminal microbiota composition and fermentation as well as on nutrient digestion.

Ruminal buffers, such as sodium bicarbonate and sesquicarbonate are commonly fed to high-producing dairy cows to avoid ruminal acidosis. However, different Mg sources have varied alkalization activity, which could interact with dietary ruminal buffers, potentially impacting nutrient digestion and milk production in dairy cows (Erdman et al. 1980, 1982). In our companion study, we reported milk production, ­composition, and Ca and Mg balance in lactating Holstein cows fed with different sources of supplemental Mg (Lobo et al., 2023). In the current study, we focused on ruminal fermentation, microbial protein synthesis, the ruminal microbiota composition, and apparent nutrient digestion. Therefore, the objective of the current study was to evaluate the effects of different sources of supplemental Mg and potential interactions of Mg source with buffer use on the ruminal microbiota composition and fermentation, microbial protein synthesis, and apparent nutrient digestion in lactating dairy cows.

Materials and Methods

The Institutional Animal Use and Care Committee approved this project that was conducted at the University of Florida Dairy Research Unit (Alachua, FL).

Cows and experimental design

The details related to cows, diet, and experimental design were described in our companion study (Lobo et al., 2023). Briefly, cows were trained to use the Calan-gates system prior to the beginning of the trial. A group of 60 high-producing Holstein dairy cows was used. The animals were assigned to blocks, based on parturition and days in milk (DIM) of their current lactation. A total of 15 blocks of four animals were established. Each animal within a block was randomly assigned to one of the four experimental treatments. Dry matter intake and production parameters were measured from all 60 animals, and it is available in our companion paper (Lobo et al., 2023). A subsample of the cows was randomly selected (five blocks, N = 20 cows; five for each treatment) for this metabolic trial, to study the microbiota composition, ruminal fermentation, nutrient digestibility, and microbial protein synthesis. The subgroup of cows consisted of 12 multiparous and 8 primiparous cows with an initial average of 96 ± 32 DIM, body weight of 628 ± 90 kg, and milk production of 39.3 ± 5.8 kg/d.

Diets were balanced to be isonitrogenous (16.3% of crude protein [CP]), isocaloric (1.73 Mcal Kg⁻¹ NE_l), and to have the same composition of macro- and micro-minerals (0.70% of Ca, 0.46% of P, 0.27% of Mg, 0.07% of S, 0.37% Cl, and 1.19% of K), following the recommendations of NRC (2001) for midlactation cows with similar body weight and milk production of the group used. The treatments were arranged in a 2 × 2 factorial design, where the first factor was the source of Mg, and the second factor was the use of a ruminal buffer (Na sesquicarbonate). The arrangement of treatments was: 1) Mg oxide (MgO); 2) Mg oxide + Na sesquicarbonate (MgO+); 3) Ca–Mg hydroxide (CaMgOH); and 4) Ca–Mg hydroxide + Na sesquicarbonate (CaMgOH+). The basal diet supplied 75% of the Mg required and the remaining was supplemented with the treatments. In addition, ruminal buffer was added at the ratio of 60 g/kg of TMR in a DM basis. Treatments were offered as top-dress and the animals were fed diets for 60 d (20 d of adaptation to the diet and 40 d for sample collection). Before the morning feeding, the individual feed bunks were emptied, and the weight of the orts was recorded. The daily diet offered was adjusted to allow at least 10% of orts.

Sample collection

Collection of feces and urine, from the 20 randomly selected cows, was carried out by spot urine and feces sampling at 0400, 1200, and 2000 hours on days 28, 38, 48, and 58, and at 0800, 1600, and 2400 hours on days 29, 39, 49, and 59 of the experiment. Urine was collected by massage of the perineal area. A daily composite sample of urine was collected for each cow and stored in a 50 mL tube at −20 ˚C for nitrogen and creatinine analysis. In another 50 mL tube, urine was diluted in a 1:4 (v:v) ratio with 1N H₂SO₄, a subsample of urine from each daily timepoint was transferred to the tube, and a daily acidified composite sample was stored at −20 ˚C for analysis of allantoin and uric acid.

Feces were collected directly from the rectum of the animals at the same timepoints as the urine collections as previously described. Within 1h of collection the material was homogenized, and a subsample of 150 g was collected and pooled into a two-day pool (days 28/29; 38/39; 48/49; and 58/59) and then stored at −20 ˚C. Feed and orts were collected for 4 d for intake estimation; 2 d before feces/urine collection, and 2 d after feces/urine collection. Right after collection, the material was weighed and placed in an oven for 72 h at 60 ˚C and stored in paper bags.

Ruminal contents from the randomly selected cows were collected 4 h after morning feeding using a vacuum pump and an orogastric tube, on days 27, 37, 47, and 57 of the experiment. The first 100 mL of ruminal fluid was discarded to avoid saliva contamination. From the second amount of ruminal content collected, a sample was filtered through four layers of cheesecloth. An aliquot of 10 mL was collected and acidified with 100 µL of 50% H₂SO₄ (v/v), and the acidified sample was stored at −20 ˚C for volatile fatty acid (VFA) and ammonia (NH₃–N) analysis. At the same time, two sets of 2 mL samples were collected, one was stored at −20 ˚C for lactate analysis, and the other was snap-frozen in liquid nitrogen and then stored at −80 ˚C for DNA extraction and assessment of the bacterial community that is primarily planktonic but does also contain bacteria attached to the small digesta particles that were collected by the stomach tubing procedure.

Chemical analysis

Daily urinary excretion was estimated based on the creatinine concentration in the urine, according to Chizzotti et al. (2008), considering a daily excretion of 0.212 mmol of creatinine per kg of body weight. Creatinine was analyzed using the DetectX Urinary Creatinine Detection Kit (K002-H5, Arbor Assay, Michigan, USA), intra- and interassay co-efficients of variation (CV) were 2.8% and 2.3%. Urinary nitrogen concentration was measured using a mass spectrometer (IsoPrime 100, IsoPrime), according to method 990.03 (AOAC, 2000). Purine derivatives (allantoin and uric acid) were measured using the colorimetric method described by Chen and Gomes (1992).

Feces were thawed at 4 ˚C, and then placed in a forced-air oven for 72 h at 60 ˚C for drying. Posteriorly, dried feed, orts, and feces were ground in a Wiley mill (model N°2; Arthur H. Thomas Co., Philadelphia, PA) to pass through a 2 mm screen. Each ingredient/feces were homogenized, one sample was taken and ground to pass through a 1 mm screen to determine the chemical composition. Samples of ingredients, orts, and feces were analyzed for dry matter (DM) content in a forced air oven at 105 ˚C according to method No. 930.15 (AOAC, 1990). Ash was determined by combustion at 600 ˚C for 6 h in a muffle furnace, according to method no. 942.05 (AOAC, 1990). Total N was analyzed using a mass spectrometer (IsoPrime 100, IsoPrime), according to the method 990.03 (AOAC, 2000) and multiplied by a factor of 6.25 to estimate the CP of the sample. Ash-free neutral detergent fiber (aNDFom) was determined, according to Mertens et al. (2002), using thermostable α-amylase and sodium sulfite modified for Ankom200 Fiber Analyzer (Ankom Technology, Macedon, NY). Ash-free acid detergent fiber (aADFom) was determined according to method No. 973.18 (AOAC, 1990), sequentially to aNDFom, the ash correction was carried out by combustion of the remains at 600 ˚C for 6 h in a furnace of ADF residues.

Indigestible NDF (iNDF) was used as an internal marker for fecal output. Briefly, two sets of 1 g samples ground at 2 mm were weighed and placed in an F57 Ankom bag (Ankom Technology). Two lactating cannulated dairy cows were used to do in situ incubation. Bags were heat-sealed, and one bag from each sample was placed in the rumen of each cannulated lactating dairy cows; bags were incubated for 240 h. After incubation, the bags were washed until recovery of clear water and dried at 60 ˚C for 72 h. Analysis for aNDFom was carried out following the procedure described by Mertens et al. (2002) with the addition of thermostable α-amylase and sodium sulfite in an Ankom200 Fiber Analyzer (Ankom Technology). Apparent total tract DM digestibility (DMD, %) was calculated as:

$$DMD=1-\frac{dietiNDF,\%}{FecesiNDF,\%}$$

Fecal output (Fout, kg) was calculated based on the DM intake (kg) and DMD as:

$$Fout=1-\frac{DM\cdot intake\times (100-DMD)}{100}$$

The estimation of the co-efficient of apparent digestibility of the other nutrients (apDN, %), such as OM, CP, NDF, and ADF was calculated as:

$$apDN=100-\left(\frac{Fout\times Fecalnutrientconcentration}{DMintake\times Dietnutrientconcentration}\times 100\right)$$

Nonacidified and frozen ruminal fluid was analyzed for lactate concentration using a commercial kit (D-Lactic acid/ L-Lactic acid kit, R-Biopharm AG). Samples were processed and analyzed according to the instructions provided by the manufacturer. Absorbance was measured in a spectrophotometer (SpectraMax Plus 384 Microplate Reader, Molecular Devices, San Jose, CA). Acidified rumen fluid samples that were collected for VFA and NH₃–N concentration analysis were centrifuged (Sorvall LYNX 4000 Centrifuge, Thermo Scientific, GA) at 10,000 × g for 15 min at 4 ˚C. NH₃–N concentration was determined in duplicate according to Broderick and Kang (1980) and adapted to a plate reader by using 2 µL of the sample, 100 µL of phenol solution, and 80 µL of hypochloride solution in each well of the microplate. Absorbance was measured in a spectrophotometer (SpectraMax Plus 384 Microplate Reader, Molecular Devices, San Jose, CA) at 620 nm.

Samples to determine VFA were further processed, according to Ruiz-Moreno et al. (2015), where solution of 2 g/L (w/v) of crotonic acid and 25% (w/v) of metaphosphoric acid was added in a 1:5 ratio (crotonic and metaphosphoric acid to sample) to the supernatant and frozen overnight. The sample was then centrifuged at 10,000 × g for 15 min at 4 ˚C. The supernatant was recovered, and ethyl acetate was added in a 2:1 ratio to the supernatant, vortexed, and allowed to separate into layers, according to Ruiz-Moreno et al. (2015) with the following modifications. The top layer was transferred to a chromatography vial. The concentration of acetate, propionate, butyrate, isobutyrate, isovalerate + 2-methylbutyrate, valerate, and caproate was determined by gas chromatography (Agilent 7820A GC, Agilent Technologies, Shanghai) using a flame ionization detector and a capillary column (CP-WAX 58 FFAP 25 m 0.53 mm, Varian CP7767, Varian Analytical Instruments, CA) at 110 ˚C with injector temperature at 200 ˚C and detector at 220 ˚C.

Microbiota composition analysis

The ruminal fluid samples stored at -80 ˚C were thawed at room temperature and DNA extraction was carried out using the fecal and soil microbiome DNA miniprep kit (D6010, Zymo Research Corporation, CA), which included both mechanical and chemical cell disruption. The DNA quality of the extracted DNA was evaluated using 1.5% agarose gel electrophoresis. Processing of DNA and amplification of region V4 of the 16S rRNA gene was performed according to Kozich et al. (2013). Briefly, amplification was obtained with PCR in a C1000 Touch Thermal Cycler (Bio-Rad, CA). The V4 region of the 16S rRNA gene was amplified by dual-index universal bacterial primers (forward: GTGCCAGCMGCCGCGGTAA; reverse: GGACTACHVGGGTWTCTAAT) through an initial denaturation of 5 min under 95 ˚C, followed by 30 cycles of 30 s at 95 ˚C, 30 s at 55 ˚C, 1 min at 72 ˚C, and 5 min for final elongation at 72 ˚C. Forward and reverse primers, as well as small DNA fragment contaminants, were removed using a 1% low-melting agarose gel extraction kit (National Diagnostics, Atlanta, GA). Amplicons were then purified and normalized using a SequalPrep plate kit (Invitrogen, CA), and the DNA concentration was measured with a Qubit fluorometer. Adapters were added to the amplicons, and the DNA library was constructed by equally pooling all the amplicons together and using quantitative real-time PCR for quality check. Sequencing was performed using a MiSeq reagent kit V2 (2 × 250 cycles run; Illumina) in an Illumina MiSeq platform at the Interdisciplinary Center for Biotechnology Research of the University of Florida (Gainesville, FL). All sequences were deposited in the Sequence Read Archive of the National Center for Biotechnology Information (https://www.ncbi.nlm.nih.gov/sra) under access no. PRJNA900619.

Data processing and statistical analysis

Process and statistical analyses of the microbiota composition data were carried out in R (RStudio 3.0.1, https://www.r-project.org/). Sequenced amplicons were processed using the DADA2 package of R (Callahan et al., 2016) and taxonomy was assigned using the 16S rRNA SILVA v. 138 database (Quast et al., 2013) and phyloseq package of R (McMurdie and Holmes, 2013). Briefly, paired-end reads were demultiplexed and quality profiles of both forward and reverse reads were separately inspected, filtered, and trimmed based on the quality scores. Forward reads were trimmed on position 30 and 240 base pair and reverse reads were trimmed on position 30 and 180 base pair. The forward and reverse reads were merged, and another quality check was carried out to remove chimeras (Table S1, https://osf.io/q3yjs/?view_only=2110f1f0769d431f8ebda596238b7386), additionally, singletons were removed.

After the taxonomy assignment, the microbiome package was used to merge the reads at the phylum level, and the ratio of the three main phyla was calculated (Bacteroidota:Firmicutes; Bacteroidota:Proteobacteria; and Proteobacteria:Firmicutes). For the calculation of diversity indexes, the data was rarefied. The calculation of alpha diversity (Chao1, Shannon, and Simpson), evenness (Pielou, Camargo, and Simpson), absolute dominance, and rarity (low abundance and rare abundance) was carried out using phyloseq and microbiome (Lahti and Shetty, 2017) package of R. To verify whether the number of sequences obtained per sample was representative of the microbial community, coverage was calculated according to Good (1953), and it is presented on Table S1 (https://osf.io/q3yjs/?view_only=2110f1f0769d431f8ebda596238b7386).

To improve the accuracy of microbiota composition analysis, a second denoising algorithm was applied (Roesch et al., 2020). This algorithm is called Prevalence Interval for Microbiome Evaluation (PIME) and allows for the separation of noise across samples from biologically significant findings. In PIME, taxa were filtered through random forest classifications and noise removed through prevalence intervals, from which taxa that were not shared, given an ideal prevalence interval within an independent treatment group were removed for better visualization of community differences. The error rate calculated in PIME is presented in Table S2 (https://osf.io/q3yjs/?view_only=2110f1f0769d431f8ebda596238b7386) and a prevalence of 60% was used in the further analysis.

The dissimilarity analysis was carried out using the filtered data from PIME (prevalence of 60%) and the function ordinate from phyloseq using the Bray-Curtis and Euclidian distance. To test the treatment effect, PERMANOVA analysis from the vegan package (Oksanen et al., 2020) was carried out using 2,000 permutations and adonis2 function for analysis and partitioning sums of squares using dissimilarities, which is based in the principles of McArdle and Anderson (2001). The model used source, buffer, and interaction source × buffer as fixed effects.

The differential abundance at phylum and genus level was analyzed using ALDEx2 package (Fernandes et al., 2013). ALDEx2 is a package that can analyze the differential abundance between two groups of the studied cohort, two separate analyses were carried out, first to compare the two contrasting groups that were fed MgO versus CaMgOH, and then a second analysis was carried out to compare the buffer use (No vs. Yes). The differential abundance analysis was carried out using CLR transformed data and the relative abundance for each taxa was calculated using the Phyloseq package and included in the output table. Wilcoxon rank test was used to test the hypothesis, and significance was declared when P-value adjusted for Benjamini–Hochberg correction was ≤0.05 and the median effect size was ≥1. According to Gloor et al. (2016), the effect size is more robust than P-value in the differential abundance analysis and will likely correspond to biological relevance. In addition, effective size was calculated as the ratio between the between-group difference divided by within-group dispersion.

The experiment followed a randomized complete block design with the cow as the experimental unit. Cows were blocked according to the parity (primiparous or multiparous) and DIM of the current lactation. Within each block, cows were assigned randomly to one of the four treatments. Therefore, two blocks of four primiparous cows each and three blocks of four multiparous cows each were selected, as described before.

The normality of residuals and homogeneity of variance were examined for each continuous dependent variable using the Shapiro–Wilk test from the UNIVARIATE procedure of SAS 9.4. Data of VFA concentration, digestibility, and N balance were collected over time and analyzed as repeated measurements. The most appropriate covariance structure was selected for each model based on the spacing of measurements and the smallest corrected Akaike’s information criterion. Statistical analyses were performed using the MIXED procedure of SAS 9.4 using the model:

$${\text{Y}}_{\text{ijklm}}=\mu +{\text{C}}_{\text{i}}+{\text{T}}_{\text{j}}+{\text{P}}_{\text{k}}+{\text{T}}_{\text{j}}\times {\text{P}}_{\text{k}}+B_m+C{(B)}_{lm}+e_{ijklm}$$

Y_ijklm is the observation ijklm; µ is the overall mean; C_i is the true covariate; T_j is the fixed effect of treatment (j = 1 to 4); P_k is the fixed effect of period (k = 1 to 4); T_j × P_k is the interaction between T_j and P_k; B_m is the random effect of block (m = 1 to 15); C(B)_lm is the random effect of cow (l = 1 to 4) nested within block (m = 1 to 15); and e_ijklm is the random residual. The ratio of bacterial phyla and the alpha diversity indexes evaluated, were analyzed similarly with the exception that the covariate, period, and interaction between treatment × experimental period were not included in the model as samples were pooled across experimental periods. Following the model:

$${\text{Y}}_{\text{ijk}}=\text{μ}+{\text{T}}_{\text{i}}+B_j+C{(B)}_{\text{kj}}+e_{\text{ijlm}}$$

Y_ijk represents the observation ijk; µ represents the overall mean; T_i is the fixed effect of treatment (i = 1 to 4); B_j is the random effect of block (j = 1 to 5); C(B)_kj is the random effect of cow (k = 1 to 4) nested within block (j = 1 to 5); and e_ijk is the random residual. The main effects (Mg source and ruminal buffer) and their interaction were analyzed through orthogonal contrasts also in SAS, and differences were considered significant if P ≤ 0.05 and considered a tendency if 0.05 < P ≤ 0.10.

Results

Ruminal fermentation, apparent digestibility, and microbial protein synthesis

Fermentation metabolites are presented in Table 1. There was an interaction effect of source of Mg and buffer use on total VFA molar concentration (P = 0.04). There was an effect of buffer use on butyrate molar proportion (P = 0.05), where animals fed with buffer had a greater molar proportion of butyrate than animals with no buffer supplementation (10.0% vs 9.10%, respectively). Cows that were fed with CaMgOH had a greater ruminal NH₃–N concentration (P = 0.03) and greater ruminal molar proportion of butyrate (P = 0.03) and caproate (P = 0.02), than cows fed with MgO. There was no effect of buffer or Mg source on molar proportion of acetate, propionate, isobutyrate, isovalerate + 2-methylbutyrate, valerate, and molar ratio of acetate to propionate.

Table 1.

Effects of Mg source and ruminal buffer use on ruminal fermentation in high producing dairy cows (N = 20)

Parameter1	Treatment2				SEM3	P-value4
	MgO	MgO+	CaMgOH	CaMgOH+		Source	Buffer	S*B
Mmol/liter
Total VFA	93.2	103.3	102.1	99.6	3.07	0.42	0.25	0.04
Lactate	0.24	0.27	0.28	0.27	0.01	0.07	0.32	0.14
NH3–N	0.12	0.12	0.13	0.13	0.01	0.03	0.77	0.85
Molar proportion
Acetate	61.1	60.9	61.1	62.5	1.08	0.40	0.58	0.42
Propionate	26.5	26.4	26.0	23.1	1.36	0.16	0.28	0.29
Isobutyrate	0.68	0.58	0.61	0.63	0.05	0.88	0.48	0.26
Butyrate	8.86	9.33	9.41	10.75	0.42	0.03	0.05	0.32
Isovalerate + 2-methylbutyrate	1.41	1.24	1.19	1.26	0.11	0.41	0.70	0.31
Valerate	1.41	1.35	1.38	1.37	0.08	0.96	0.70	0.78
Caproate	0.11	0.20	0.26	0.30	0.05	0.02	0.21	0.63
A:P	2.37	2.35	2.43	2.85	0.19	0.15	0.31	0.25

¹mmol, milli moles; A:P = molar ratio of acetate to propionate.

²MgO+, magnesium oxide and sodium sesquicarbonate; CaMgOH+, calcium–magnesium hydroxide and sodium sesquicarbonate;

³SEM, standard error of the mean.

⁴Values from orthogonal contrast are significantly different if P ≤ 0.05 and tendency if 0.05 < P ≤ 0.10.

Nutrient intake and total tract apparent digestibility co-efficient of nutrients are presented in Table 2. There was no effect of the interaction between source of Mg and buffer supplementation on either intake or digestibility co-efficient of nutrients. Also, there was no effect of Mg source on the nutrient intake and digestibility co-efficient evaluated. There was an effect of buffer supplementation on the digestibility co-efficient of the DM (P = 0.05), OM (P = 0.04), NDF (P = 0.04), and ADF (P = 0.03).

Table 2.

Effects of Mg source and ruminal buffer use on intake and apparent total tract digestibility of nutrients in high producing dairy cows (N = 20)

Parameter	Treatment1				SEM2	P-value3
	MgO	MgO+	CaMgOH	CaMgOH+		Source	Buffer	S*B
Intake
OM	20.0	21.8	22.2	19.7	1.21	0.94	0.79	0.11
CP	3.42	3.73	3.75	3.40	0.20	0.89	0.92	0.12
aNDFom	5.53	6.03	6.16	5.47	0.34	0.92	0.79	0.11
aADFom	3.08	3.38	3.44	3.06	0.19	0.92	0.85	0.10
Digestibility co-efficients
DM	69.1	70.6	68.8	70.7	0.98	0.95	0.05	0.79
OM	71.4	73.2	71.1	73.6	1.21	0.96	0.04	0.66
CP	71.7	72.8	71.1	71.4	1.90	0.50	0.65	0.76
aNDFom	35.5	40.3	35.7	42.5	2.77	0.60	0.04	0.67
aADFom	34.4	39.4	35.6	42.8	2.87	0.35	0.03	0.69

¹MgO+, magnesium oxide and sodium sesquicarbonate; CaMgOH+, calcium–magnesium hydroxide and sodium sesquicarbonate;

²SEM = standard error of the mean.

³Values from orthogonal contrast are significantly different if P ≤ 0.05 and tendency if 0.05 < P ≤ 0.10.

The microbial protein synthesis and nitrogen balance parameters are presented in Table 3. There was an effect of buffer on total urinary excretion (P < 0.01) and uric acid excretion (P < 0.01), where animals fed with buffer excreted greater amounts of both compounds. There was no effect of interaction of the Mg source and buffer use on the parameters evaluated. Also, effects of Mg source were not observed on the studied parameters.

Table 3.

Effects of Mg source and ruminal buffer use on nitrogen metabolism and microbial protein synthesis in high producing dairy cows (N = 20)

Parameter1	Treatment2				SEM3	P-value4
	MgO	MgO+	CaMgOH	CaMgOH+		Source	Buffer	S*B
N intake, g/d	547	597	600	543	31.8	0.98	0.92	0.12
Milk excretion
Milk total N, g/d	185	172	171	168	10.2	0.40	0.45	0.64
Milk total N, % of N intake	32.7	28.1	30.3	30.8	1.84	0.93	0.27	0.18
Urinary excretion
Urine excretion, kg/d	22.1	32.8	24.1	30.4	2.67	0.93	< 0.01	0.35
Urine total N, g/d	229	241	230	234	12.8	0.73	0.46	0.69
Urine total N, % of N intake	41.0	41.5	39.4	42.1	2.05	0.83	0.46	0.60
Allantoin, mmol/d	365	335	345	340	23.7	0.71	0.39	0.52
Uric Acid, mmol/d	40.5	51.2	38.4	49.6	3.7	0.61	< 0.01	0.95
Total PD, mmol/d	405	384	382	393	26.7	0.76	0.81	0.46
Absorbed PD, mmol/d	400	341	372	388	27.9	0.75	0.42	0.17
MPS, g of N/d	291	248	270	282	20.3	0.75	0.42	0.17
MPS, kg of CP/d	1.82	1.55	1.69	1.76	0.13	0.75	0.42	0.17
Fecal N excretion	6.47	6.63	7.17	6.08	0.38	0.84	0.24	0.12
Fecal N, g/day	158	159	171	158	13.3	0.61	0.63	0.54
Fecal N, % of N intake	28.3	27.2	28.1	27.9	1.81	0.85	0.58	0.68
Manure N, g/d	389	398	401	388	22.4	0.96	0.88	0.51
Manure N, % of N intake	69.2	68.7	66.0	68.7	2.71	0.54	0.69	0.54

¹PD, purine derivatives; MPS, microbial protein synthesis estimated according to Chen and Gomes (1992).

²MgO+, magnesium oxide and sodium sesquicarbonate; CaMgOH+, calcium–magnesium hydroxide and sodium sesquicarbonate.

³SEM, standard error of the mean.

⁴Values from orthogonal contrast are significantly different if P ≤ 0.05 and tendency if 0.05 < P ≤ 0.10.

Ruminal microbiota composition analysis

The sequence coverage of the dataset was ≥99% (Table S1, https://osf.io/q3yjs/?view_only=2110f1f0769d431f8ebda596238b7386) and more than 83% of the reads were assigned taxonomy to genus level (Table S3, https://osf.io/q3yjs/?view_only=2110f1f0769d431f8ebda596238b7386). The microbiota composition of the rumen environment of the cows studied were mainly composed of micro-­organisms of the phylum Bacteroidota (43.0%), followed by the phylum Proteobacteria (25.4%), and Firmicutes (23.8%), which together accounted for more than 90% of the reads extracted from ruminal samples of lactating dairy cows ­(Figure 1).

Figure 1.

Average microbial composition of the rumen of lactating dairy cows (N = 20) at phylum level by experimental treatment (CaMgOH+ = calcium–magnesium hydroxide and sodium sesquicarbonate supplementation; and MgO+ = magnesium oxide and sodium sesquicarbonate supplementation.

The diversity indexes and phylum ratio are presented in Table 4. There was no interaction effect of the two main factors in the diversity of the ruminal microbiota ­composition of lactating dairy cows. Buffer supplementation tended to increase the evenness of species when calculated using the Simpson index (P = 0.08). The alpha diversity tended to increase in lactating dairy cows fed with CaMgOH when estimated using Shannon (P = 0.08) and Simpson (P = 0.10) indexes. A tendency to increase the species evenness was observed for animals fed with CaMgOH when estimated using the Pielou (P = 0.08) and Simpson (P = 0.06) indexes. Also, species absolute dominance (P = 0.07) and rarity (P = 0.04) were slightly increased in animals fed with CaMgOH.

Table 4.

Effects of Mg source and ruminal buffer use on ecological alpha diversity, evenness, and phylum ratio (Bacteroidota, Firmicutes, and Proteobacteria) in the rumen environment of high producing dairy cows (N = 20)

Parameter	Treatment1				SEM2	P-value3
	MgO	MgO+	CaMgOH	CaMgOH+		Source	Buffer	S*B
Alpha diversity indexes
Chao1	417	424	473	471	30.5	0.13	0.94	0.89
Shannon	4.22	4.27	4.38	4.74	0.15	0.08	0.23	0.38
Simpson	0.91	0.91	0.93	0.95	0.01	0.10	0.45	0.43
Evenness indexes
Camargo	0.22	0.23	0.23	0.27	0.01	0.13	0.16	0.30
Pielou	0.70	0.70	0.72	0.77	0.02	0.08	0.24	0.38
Simpson	0.03	0.03	0.03	0.07	0.01	0.06	0.08	0.15
Dominance indexes
Absolute	5,917	5,788	5,329	3,815	603	0.07	0.24	0.31
Rarity indexes
Low abundance	0.21	0.22	0.22	0.24	0.02	0.28	0.33	0.60
Rare abundance	0.20	0.19	0.21	0.27	0.02	0.04	0.24	0.13
Ratio
Firmicutes:Bacteroidota	2.03	1.83	1.83	1.66	0.14	0.10	0.09	0.89
Proteobacteria:Bacteroidota	1.52	1.62	1.66	3.29	0.32	0.02	0.03	0.05
Firmicutes:Proteobacteria	1.34	1.01	1.11	0.56	0.10	< 0.01	< 0.01	0.32

¹MgO+, magnesium oxide and sodium sequicarbonate; CaMgOH+, calcium–magnesium hydroxide and sodium sesquicarbonate.

²SEM, standard error of the mean.

³Values from orthogonal contrast are significantly different if P ≤ 0.05 and tendency if 0.05 < P ≤ 0.10.

The population ratios of the three main bacterial phyla are presented in Table 4. There was an interaction effect of the Protobacteria to Bacteroidota ratio (P = 0.05), showing that the ratio did not change on animals fed with MgO whether or not they received buffer supplementation. However, when cows were fed CaMgOH, supplementation of buffer increased the populational ratio of Proteobacteria to Bacteroidota. There was a reduction in the Firmicutes to Bacteroidota ratio (P = 0.09) and Firmicutes to Proteobacteria ratio (P < 0.01) on animals fed buffer. Also, there was a reduction in the Firmicutes to Bacteroidota ratio (P = 0.10) and Firmicutes to Proteobacteria ratio (P < 0.01) in animals fed with CaMgOH when compared to animals fed with MgO.

Beta diversity analysis is presented in Figure 2. The principal coordinates analysis (PCoA) of the ruminal microbiota composition of lactating dairy cows was carried out using Bray–Curtis distance matrix. In this analysis, 48.1% of the variation was accounted in the first axis, and 14.6% was accounted in the second axis (Figure 2). There was no visual cluster formation; however, the PERMANOVA analysis depicted the effects of both Mg source (P = 0.02) and buffer use (P = 0.02).

Figure 2.

Principal coordinates analysis (PCoA) after PIME denoising based on Bray–Curtis distance matrix (A) and Euclidean distance matrix (B) showing the differences between microbial communities at ASV level. Each point represents a microbial community from one sample; blue represents animals fed with MgO and red represents the animals fed with CaMgOH; triangle and circle represent the animals supplemented or not with sodium sesquicarbonate. PERMANOVA results: Bray–Curtis (Mg source = 0.02; Buffer use = 0.02; and Interaction = 0.11) and Euclidean distance (Mg source = 0.05; Buffer use = 0.18; and Interaction = 0.28).

Differential abundance

The differential abundance analysis and relative abundance at the phylum level for the Mg source effect is presented in Figure 3A and Table S4 (https://osf.io/q3yjs/?view_only=2110f1f0769d431f8ebda596238b7386). We can observe that at the phylum level, there was an increase in CLR transformed median (P_adj = 0.02 and effect size = −0.93, Table S4, https://osf.io/q3yjs/?view_only=2110f1f0769d431f8ebda596238b7386) of Chloroflexi phylum on animals fed with CaMgOH when compared to animals fed with MgO (−2.95 vs, −7.29). The differential abundance analysis at the genus level for the factor Mg Source is presented in Figure 3B and Table S5 (https://osf.io/q3yjs/?view_only=2110f1f0769d431f8ebda596238b7386). We can observe that there was an increase in the genera Prevotellaceae_UGG-001 (P_adj = 0.02, effect size = 1.20), Lactobacillus (P_adj < 0.01, effect size = 1.48), and Butyrivibrio (P_adj < 0.01, effect size = 1.09) of the rumen environment of lactating dairy cows fed with CaMgOH. Also, for the same cows, a reduction in the abundance of the genera Erysipelotrichaceae_UCG-009 (P_adj < 0.001, effect size = 1.73), FD2005 (P_adj < 0.001, effect size = 1.93), and Lachnospira (P_adj < 0.01, effect size = 1.35) was observed.

Figure 3.

Bland–Altman plot (MA plot) at phylum (A) and genus (B) level, contrasting abundance of the rumen microbiota composition of cows fed with MgO vs. CaMgOH, using ALDEx2 package and PIME filtered data. Each point of the graphic represents one bacterial community. Red points had P-value of Wilcoxon rank test adjusted for Benjamini–Hochberg smaller or equal to 0.05 and effect size greater or equal to 1; orange points had P-value of Wilcoxon rank test adjusted for Benjamini–Hochberg smaller or equal to 0.05 and effect size smaller than 1; blue points represent nonsignificant abundance micro-organisms.

The differential abundance analysis at the phylum level for the buffer effect is presented in Figure 4A and Table S7 (https://osf.io/q3yjs/?view_only=2110f1f0769d431f8ebda596238b7386). There was an increase in micro-organisms assigned to the phylum Spirochaetota (P_adj = 0.03) and Planctomycetota (P_adj = 0.04) in the rumen fluid of animals supplemented with ruminal buffer when compared to not supplemented animals. The differential abundance analysis at the genus level for the buffer use is presented in Figure 4B and Table S8 (https://osf.io/q3yjs/?view_only=2110f1f0769d431f8ebda596238b7386). We observed a reduction in the sequences assigned to the genus Roseburia (P_adj = 0.06, effect size = 1.20) and Syntrophococcus (P_adj = 0.03, effect size = 1.08) in lactating dairy cows fed the ruminal buffer.

Figure 4.

Bland–Altman plot (MA plot) at phylum (A) and genus (B) level, contrasting abundance of the rumen microbiota composition of cows fed with ruminal buffer (No vs. Yes), using ALDEx2 package and PIME filtered data. Each point of the graphic represents one bacterial community. Red points had P-value of Wilcoxon rank test adjusted for Benjamini–Hochberg smaller or equal to 0.05 and effect size greater or equal to 1; orange points had P-value of Wilcoxon rank test adjusted for Benjamini–Hochberg smaller or equal to 0.05 and effect size smaller than 1; blue points represent nonsignificant abundance micro-organisms.

Discussion

Since there were no interaction effects observed within main factors (Mg source and ruminal buffer use), only the main effects will be discussed in this section.

Effect of Mg sources

The main source of Mg used in dairy operations is the MgO (NASEM, 2021); however, lately there has been increased interest in finding alternative sources of Mg (Leno et al., 2017; Arce-Cordero et al. 2021a, 2021b; Agustinho et al., 2022). Some of the previous work from our research group, showed that CaMg(OH)₂ had a similar fermentation profile (pH and most VFAs) when compared to the MgO (Arce-Cordero et al. 2021a, 2021b). Also, there was no interaction of Mg source and buffer use (Arce-Cordero et al., 2021b; Agustinho et al., 2022). However, these experiments were carried out in a dual-flow continuous culture system and further investigation considering the whole gastrointestinal tract and animal ­physiology was needed for validation of alternative Mg sources in vivo.

The results of the present experiment indicated that ruminal fermentation was slightly different for animals fed CaMgOH due to increased molar concentration of lactate and ammonia and molar proportion of butyrate and caproate when compared to animals fed MgO, regardless of the buffer use. Lactate and ammonia are intermediate metabolites of ruminal fermentation and are produced by lactate-producing bacteria (Hernandez et al., 2008) and by the deamination of amino acids (Satter and Roffler, 1975), respectively. These intermediate products can be further fermented to produce VFAs (such as acetate, propionate, and butyrate) by a wide variety of lactate-utilizing bacteria (Mackenzie, 1967) and used for microbial protein synthesis (Satter and Roffler, 1975), respectively.

Along the same lines, we observed an increase in abundance of both Lactobacillus and Butyrivibrio in the rumen of dairy cows fed with CaMgOH, which could have led to the increases observed in lactate and butyrate concentrations. Lactobacillus is a well-studied genus of the rumen. They have been reported to ferment hexoses to several byproducts, mainly lactate (Hammes and Hertel, 2006; Monteiro et al., 2022). In the rumen, it is well-established that Megasphaera, Selenomonas, and Veillonella, are micro-organisms capable of further fermenting lactate into VFAs (Hungate, 1979; Counotte et al., 1981; Hashizume et al., 2003), especially Megasphaera, which has been reported to produce butyrate (Hashizume et al., 2003; Duncan et al., 2004). Thus, a reduction in abundance of unidentified lactate-utilizers may have caused an increase in lactate concentration, which may favor butyrate-producing bacteria and consequently explain the greater butyrate concentration. However, in this study, none of these micro-organisms had a greater abundance in the rumen of cows fed with CaMgOH, but still, this hypothesis should not be discarded, and further investigation is warranted (Ravelo et al., 2022).

Butyrivibrio is a well-characterized genus of the mammalian gastrointestinal tract, which utilizes plant carbohydrates, such as xylans, and produce several VFAs, such as butyrate (Cotta and Forster, 2006). Although xylans and butyrate have been reported to be predominantly the major substrate and end-product of fermentation for this genus, Shane et al. (1969) isolated 19 strains belonging to the genus Butyrivibrio from sheep rumens. The authors characterized that all the isolates utilized L-arabinose, galactose, sucrose, cellobiose, pectin, starch, and xylan as carbon sources and produced butyrate, formate, hydrogen ions, and at least traces of ethanol.

Another interesting finding from the study of Shane et al. (1969), was that they could divide the 19 isolates into two main groups, one that was able to produce considerable amounts of lactate and remove acetate from the medium during cellobiose fermentation (group 1), while the other produced acetate but had little or no lactate production (group 2). Together with our findings, these results indicate that CaMgOH may favor the proliferation of Lactobacillus, and Butyrivibrio such as those similar to group 1 reported by Shane et al. (1969). A greater abundance of these micro-organisms may have been the reason for a greater concentration of lactate and butyrate in the rumen of dairy cows.

Butyrate is one of the main VFAs produced in ruminal fermentation, and it is a key component of the development of papillae and ruminal epithelium. Butyrate is the main source of energy for the epithelial cells due to a high expression of butyryl-CoA synthetase. This enzyme catalyzes the activation of butyrate to butyryl-CoA, which is eventually transferred into the mitochondria, and will be oxidized to produce ATP (Górka et al., 2011; Diao et al., 2019). Thus, even a small increase in the butyrate molar proportion, as described before, could be beneficial to dairy cows due to an increase in the availability of energy to the ruminal epithelium and consequently potential improvements of translocation of end-products of the fermentation to the bloodstream.

Another function of ruminal micro-organisms is the degradation of protein. It is well- documented in the literature that ruminal micro-organisms can breakdown proteins and deaminate amino acids into ammonia (Tamminga, 1979; Schwab and Broderick, 2017; Liu et al., 2019). Our results indicated that a greater ammonia concentration was present in cows fed with CaMgOH, which suggests a greater degradation and consequently deamination of amino acids in the rumen. This result is supported by an increase of Prevotellaceae UGG-001 for those cows fed with CaMgOH. Prevotella is a well-known genus with several species that exhibit peptidase activity (Wallace and Brammall, 1985; Wallace and McKain, 1991; Wallace et al., 1997). The most well-known Prevotella species of the rumen is Prevotella ruminicola 23, which has the ability of utilizing both ammonia and peptides as nitrogen sources for growth (Pittman and Bryant, 1964; Pittman et al., 1967; Kim et al., 2017). These micro-organisms could increase the ruminal protein degradation, leaving these compounds available for microbial growth or intestinal absorption for the animals fed with CaMgOH. Despite the effects of Mg sources on the rumen microbiota composition not being widely studied and understood, the results from this study support the hypothesis that modulation of the ruminal microbial population through mineral source is feasible; thus, warranting further exploration.

Mg oxides and hydroxides are a class of oxygen-bearing metals, where a metal is bound to oxygen or hydroxyl group (OH), respectively, both of which have demonstrated antimicrobial activity (Kumar et al., 2011; Raghunath and Perumal, 2017; Eivazzadeh-Keihan et al., 2021). Primarily, MgO induces the production of reactive oxygen species (ROS), which induces the expression of oxidative stress defense genes, cell morphology changes, and membrane leakage of some bacterial cells (He et al., 2016).

Differently, metal hydroxides act in the respiratory electron transport chain located in the bacterial cell membrane. Bacterial cells transport H⁺ from the cytoplasm to the extracellular matrix to generate an electrochemical proton gradient that is used to produce adenosine triphosphate (ATP), Boyer (1997). When dissolved, the hydroxyl group of the metal hydroxides can react with the H⁺ available, creating a localized alkaline micro-environment (Tan et al. 2018, 2020). With a reduction of H⁺ available, less ATP is produced, resulting in a reduction of the energy available to the microbial cell to carry out metabolic processes. Furthermore, the alkaline micro-environment can also cause hydrolysis of cellular phospholipids, consequently compromising the cellular integrity (Qin et al., 2015; Tan et al., 2018).

The reported studies that evaluated the effect of oxygen-bearing metals, such as Mg oxide or hydroxide, on micro-organisms were developed using aerobic pathogen micro-organisms or aerobic micro-organisms from other environments, such as soil micro-organisms. Therefore, there is a lack of research on the ruminal microbiota composition and its interaction with oxygen-bearing metals in anaerobic environments. From the best of our knowledge, there is no published study evaluating this interaction on pure culture of ruminal micro-organisms.

Effects of buffer use

Modern lactating dairy cows have an increased energy demand due to their high milk production (NASEM, 2021), dairy operations had to switch to diets with a greater energetic concentration to supply the demand of these animals. To increase energy concentration, easily digestible carbohydrates, such as starch, are needed, which in high concentrations could have deleterious effects in the ruminal environment and fermentation. Several studies have shown that on high starch diets, ruminal buffer is beneficial to cow health and production (Erdman et al. 1980, 1982; Iwaniuk and Erdman, 2015).

Kalscheur et al. (1997) studied the interaction effect of the forage concentration and buffer use on ruminal metabolism of lactating dairy cows. The researchers observed that on low forage diets (22% of NDF), the use of buffer improved the ruminal environment by increasing the ruminal pH; however, on high forage diets (35% of NDF), the beneficial effect of the buffer was lost. The researchers suggested that the high forage diets had lower VFA production, which would reduce the risk of low pH, thus no additional effect of buffer was observed. These results corroborate with the results from our study, there was no effect of buffer inclusion on the ruminal fermentation parameters evaluated, such as pH and molar concentration and proportion of the VFAs. These results could indicate that even with a high level of starch (32%), the experimental diets had adequate fiber concentrations (27%). Thus, the additive effect of buffer was not observed. Boerner et al. (1987) carried out a study using ruminal, duodenal, and ileal cannulated animals in a 2 × 3 factorial arrangements of treatments, where the first factor was the level of concentrate in the diet (50% and 90%) and the second factor was the buffer use (control, sodium bicarbonate, and sodium sesquicarbonate). These researchers observed that the use of buffers could improve total tract digestibility of nutrients and sodium sesquicarbonate shifted the digestibility of starch from the rumen to the small intestine, regardless of the concentrate level.

Conclusions

Compared to Mg oxide, Mg hydroxide increased the molar concentration of ammonia and lactate, and the molar proportion of butyrate in the rumen of lactating dairy cows, regardless of buffer supplementation in the diet. These compounds are intermediate and end-products of ruminal fermentation, likely carried out by micro-organisms of the genera Prevotella, Lactobacillus, and Butyrivibrio. Also, an improvement in total tract apparent digestibility of dry matter, organic matter, NDF, and ADF was observed when dietary buffer was used, regardless of Mg source. Therefore, based on the findings from this study, Mg hydroxide may replace Mg oxide without detrimental effects on nutrient digestibility and ruminal fermentation, regardless of buffer use. Future studies should be focused on evaluating the effects of different Mg sources on individual ruminal bacteria populations.

Glossary

Abbreviations:

aADFom

acid detergent fiber added amylolytic enzyme and corrected for ash

ADF

acid detergent fiber

aNDFom

neutral detergent fiber added amylolytic enzyme and corrected for ash

ATP

adenosine triphosphate

Ca

calcium

Cl

chloride

CP

crude protein

CV

co-efficient of variation

DIM

days in milk

DM

dry matter

H⁺

ion hydrogen

H₂SO₄

sulfuric acid

iNDF

indigestible neutral detergent fiber

K

potassium

Mcal

mega calorie

Mg(CO₃)₂

magnesium carbonate

Mg(OH)₂

magnesium hydroxide

Mg

magnesium

MgO

magnesium oxide

N

nitrogen

NDF

neutral detergent fiber

NEl

net energy for lactation

NH3-N

ammoniacal nitrogen

OH

hydroxyl group

OM

organic matter

P

phosphorus

PIME

prevalence interval for microbiome evaluation

S

sulfur

VFA

volatile fatty acid

w

weight

Conflict of interest statement

The authors declare no real or perceived conflicts of interest.