Effects of calcium–magnesium carbonate and calcium–magnesium hydroxide as supplemental sources of magnesium on microbial fermentation in a dual-flow continuous culture

Abstract Supplemental sources of Mg can also aid in ruminal pH regulation due to their alkaline properties. Magnesium oxide (MgO) is the most common source of Mg for ruminants and can help controlling ruminal pH; however, the alkaline potential of other sources of Mg has not been evaluated. We aimed to evaluate the inclusion of calcium–magnesium carbonate (CaMg(CO3)2) and calcium–magnesium hydroxide (CaMg(OH)4) alone or in combination as supplemental sources of Mg in corn silage-based diets and its impact on ruminal microbial fermentation. We hypothesized that inclusion of CaMg(OH)4 would allow for ruminal fermentation conditions resulting in a greater pH compared to the inclusion of CaMg(CO3)2. Four treatments were defined by the supplemental source of Mg in the diet: 1) Control (100% MgO, plus sodium sesquicarbonate as a buffer); 2) CO3 [100% CaMg(CO3)2]; 3) OH [100% CaMg(OH)4]; and 4) CO3/OH [50% Mg from CaMg(CO3)2, 50% Mg from CaMg(OH)4]. Nutrient concentration was held constant across treatments (16% CP, 30% NDF, 1.66 Mcal NEl/kg, 0.67% Ca, and 0.21% Mg). Four fermenters were used in a 4 × 4 Latin square design with four periods of 10 d each. Samples were collected for analyses of nutrient digestibility, soluble Mg, VFA, and NH3, while pH was measured at 0, 1, 2, 4, 6, 8, and 10 h post morning feeding to estimate % time when pH was below 6 (pH-B6) and area under the pH curve for pH below 6.0 (pH-AUC). Bacteria pellets were harvested for 15N analysis and estimates of N metabolism. Treatment effects were analyzed with the mixed procedure of SAS, while effects of using either CaMg(CO3)2 or CaMg(OH)4 as Mg source in comparison to Control treatment were evaluated by orthogonal contrasts. Similar pH-related variables were observed for Control, OH, and CO3/OH treatments, which had smaller pH-AUC and pH-B6 than CO3 (P ≤ 0.01). Butyrate molar proportion was greater in Control and CO3/OH than in CO3 and OH (P = 0.04). Orthogonal contrasts showed lower flow of bacterial N (P = 0.04), lower butyrate molar proportion (P = 0.08) and greater pH-AUC (P = 0.05) for diets with CaMg(CO3)2 in comparison with the Control. Concentration of soluble Mg in ruminal fluid (P = 0.73) and nutrient digestibility (P ≥ 0.52) were similar across treatments. Under the conditions of this experiment, using CaMg(OH)4 alone or combined with CaMg(CO3)2 allowed for a less acidic ruminal fermentation pattern than a diet with only CaMg(CO3)2.


INTRODUCTION
Animals require Mg as an enzymatic cofactor involved in energy metabolism, DNA synthesis and cell growth; and also for adequate functioning of muscle cells, signal transmission of nerves and structural component of bones (Schonewille and Beynen, 2005).
In ruminants, maintaining an adequate status of Mg is greatly dependent on ruminal absorption of this nutrient, which relies on the solubility of Mg sources (NRC, 2001;Goff, 2018). In this regard, magnesium oxide (MgO) is the most common supplemental source of Mg due to its ruminal solubility and absorption (NRC, 2001).
Because of their alkaline properties, supplemental sources of Mg can aid in ruminal pH control. Supplementation with MgO has shown effective to prevent a decline in ruminal pH, fiber digestibility, and milk fat production; accounting for similar effects than those observed with sodium bicarbonate buffer (Erdman et al., 1982;Bach et al., 2018). However, much less is known about other sources of Mg that could potentially provide an alkalinizing effect in the rumen.
Calcium magnesium carbonate (CaMg(CO 3 ) 2 ) has been fed as a source of Mg to ruminants, with an estimated bioavailability ranging from lower (Ammerman et al., 1972;Rahnema and Fontenot, 1983) to similar (Leno et al., 2017) to that one observed for MgO; however, its effects on ruminal fermentation are still not clear. Furthermore, to the best of our knowledge calcium magnesium hydroxide (CaMg(OH) 4 ), resulting from the hydration of calcium magnesium oxide, has not been evaluated yet as a source of Mg for ruminants. However, due to its alkaline nature, it may influence ruminal pH when used as a supplemental source of Mg.
According to their chemical composition, theoretical neutralizing capacities of CaMg(CO 3 ) 2 and CaMg(OH) 4 correspond to 21.7 and 60.4 mEq/g, respectively; indicating a greater alkalinizing potential for CaMg(OH) 4 . Based on these theoretical values, we performed a preliminary test of reactivity in acetic acid according to Goff (2018). In agreement with theoretical neutralizing capacity, estimated reactivity of CaMg(OH) 4 was approximately three times greater than CaMg(CO 3 ) 2 reactivity. Therefore, because of its chemical properties, when included in diets for ruminants, CaMg(OH) 4 may provide a greater alkaline effect than CaMg(CO 3 ) 2 .
In this experiment, we evaluated supplementation of different inorganic sources of Mg at the levels recommended by NRC (2001) for lactating dairy cows' diets. Our objective was to evaluate CaMg(CO 3 ) 2 and CaMg(OH) 4 alone or in combination as supplemental sources of Mg and their impact on microbial ruminal fermentation in a continuous culture system. We hypothesized that inclusion of CaMg(OH) 4 would allow for ruminal fermentation conditions resulting in a greater pH compared to the inclusion of CaMg(CO 3 ) 2 .

Experimental Design and Diets
The University of Florida Institutional Animal Use and Care Committee approved all the procedures for animal care and handling required for this experiment. Four fermenters of a dual-flow continuous culture system, with an average volumetric capacity of 1.82 L each, were used for this experiment in a 4 × 4 Latin square design with treatments defined by the supplemental source of Mg in the diet, as follows: 1) Control: 100% supplemental Mg from MgO, plus 0.6% sodium sesquicarbonate; as a positive control treatment; 2) CO 3 : 100% supplemental Mg from CaMg(CO 3 ) 2 ; 3) OH: 100% supplemental Mg from CaMg(OH) 4 ; and 4) CO 3 /OH: 50% supplemental Mg from CaMg(CO 3 ) 2 and 50% from CaMg(OH) 4 . As shown in Table 1, experimental diets were formulated to provide the same concentration of nutrients regardless of treatment and according to the NRC (2001) recommendations for lactating Holstein cows with 680 kg body weight, and milk production of 45 kg/d with 3.5% fat, 3.0% protein, and 4.8% lactose.
Nutrient composition of feeds was determined in samples ground through a 1 mm screen in a Wiley mill (model no. 2; Arthur H. Thomas Co., Philadelphia, PA). Major feed ingredients for experimental diets fed to the fermenters (corn silage, corn grain, soybean meal, and grass hay) were ground through a 2 mm screen using the same type of mill described above. Before grinding, corn silage was dried for 72 h at 60 °C in a forced-air oven (Heratherm, Thermo Scientific, Waltham, MA) until %DM was approximately 90% allowing for proper grinding and storage.
The sources of Mg evaluated in this experiment contained 48.2% Mg (MgO), 13.3% Mg and 20.6% Ca (CaMg(CO 3 ) 2 ), and 18.5% Mg and 24.2% Ca (CaMg(OH) 4 ), on a dry basis. Reactivity of Mg sources was evaluated according to Goff (2018) as an indicator of alkalinizing potential and solubility. Briefly, for each source, 3.0 g were weighed into a 225 mL plastic container, then 40 mL of a commercial Translate basic science to industry innovation 5% dilution of acetic acid were slowly added and shaken by hand for 15 s. Then, the mix was let sit for 15 min and shaken for 15 s again. Finally, 30 min after acetic acid was added, the pH of the mix was measured with a portable pH meter probe (Thermo Scientific Orion Star A121) and used as an indicator of reactivity. Each source was analyzed by triplicate and the average final pH was 3.52, 3.34, and 12.22 for MgO, CaMg(CO 3 ) 2 , and CaMg(OH) 4 , respectively.

Dual-Flow Continuous Culture System Operation
A dual-flow continuous culture system originally described by Hoover et al. (1976) and evaluated by Brandao and Faciola (2019) and Brandao et al. (2020) was used for this experiment. Ruminal fermentation is simulated in this system through continuous agitation (100 rpm), infusion of N 2 gas to displace oxygen, constant temperature (39 °C), and infusion of artificial saliva (Weller and Pilgrim, 1974) with 0.40 g/L of urea, at 3.05 mL/min to individually regulate passage rates of liquid (11%/h) and solid (5.5%/h) effluents of digesta.
This experiment consisted of four fermentation periods of 10 d each (40 d of fermentation total). Fermenters were inoculated on day 1 of each fermentation period with fresh ruminal contents collected from two cannulated Holstein cows in midlactation (108 ± 9 DIM on average) fed twice a day a total mixed ration with 38% corn silage, 19% ground corn, 13% soybean meal, 11% cottonseed, 9% citrus pulp, 8.5% mineral premix, and 1.5% palmitic acid supplement (on a DM basis) from 3 wk before the start and until completion of the experiment. Approximately 1 h after morning feeding, ruminal contents were manually collected from each cow and strained through two layers of cheesecloth, transferred into prewarmed thermos jars, and immediately transported to the lab. Each fermenter was pre-warmed and under the continuous flush of N 2 gas at the moment of inoculation when it was filled with a 50:50 mix (v/v) of ruminal contents from both cows.
Each fermenter was provided with its corresponding experimental diet (106 g DM/d) distributed into two identical portions of 53 g DM at 0800 and 1800 h. On d 5 of each period, a pulse dose of 0.1733 g ( 15 NH 4 ) 2 SO 4 10.2% atom excess (Sigma Aldrich Co.) was provided to each fermenter to create a steady-state of 15 N, and then ( 15 NH 4 ) 2 SO 4 a Experimental treatments based on supplemental source of Mg: "Control" = 100% as MgO + sodium sesquicarbonate added as a buffer; "CO 3 " = 100% as CaMg(CO 3 ) 2 ; "OH" = 100% as CaMg(OH) 4 ; "CO 3 /OH" = 50% as CaMg(CO 3 ) 2 /50% as CaMg(OH) 4 . b Expressed as a percent of DM unless otherwise stated.
was continuously added as a marker in artificial saliva at a rate of 0.077g/L until the end of that fermentation period. From d 8 to d 10 of fermentation, containers of solid and liquid digesta effluent were kept in an ice-cold water bath and digesta temperature maintained below 2 °C to prevent any further microbial fermentation of nutrients happening outside of the fermenters. On day 10 upon completion of each period, all fermenters were put apart and cleaned, disposable components of the system were replaced, and fermenters were reassembled, and rerandomized into experimental treatments for the following period.

Collection of Data and Samples
Data of pH and daily effluents of digesta, and samples for analyses of nutrient digestibility, VFA, ammonia N (NH 3 -N), the concentration of soluble Mg, and 15 N concentration (N metabolism), as shown below.
Nutrient digestibility. Effluent samples (liquid and solid) were collected on days 8, 9, and 10 of each period for estimates of ruminal digestibility of nutrients. Liquid and solid digesta effluents of the same fermenter (representing 24 h of fermentation) were weighed, combined and stored at −20 °C. Samples of digesta from the same fermenter and period were pooled across the three collection days, then a 200 mL composite sample per fermenter per period was freeze-dried, ground with a mortar and pestle, and stored in a plastic container for further analyses to determine ruminal digestibility of OM, CP, and NDF in duplicates.
Ruminal pH. On days 8, 9, and 10 of each period, pH was measured in each fermenter at 0, 1, 2, 4, 6, 8, and 10 h after morning feed provision. These data were used for estimation of daily average pH, time below pH 6.0 (pH-B6) expressed as the percentage of the time during the measurement period that pH was below the threshold of 6.0, and area under the pH curve (pH-AUC) expressed as the area under the curve when the pH was below 6.0 and estimated with the trapezoidal rule (Cardoso et al., 2011). Equations used for pH-B6 and pH-AUC calculations are shown below: where pH a and pH b represent the measured pH values of each interval between timepoints at times t a and t b , respectively. Daily concentrations of Mg, VFA, and NH 3 -N. On days 8, 9, and 10 of each period two 10 mL samples of mixed digesta effluent (solid and liquid) were collected from each fermenter, strained through four layers of cheesecloth, and collected into two separate 15 mL centrifuge tubes. The sample for Mg solubility analysis was collected in an empty 15 mL metal-free centrifuge tube, while the sample for VFA and NH 3 -N was collected in a 15 mL centrifuge tube acidified with 100 µL of 50% H 2 SO 4 , and both samples were stored at −20°C immediately after collection.
Nitrogen metabolism. On day 5 of each period before 15 N enriched saliva started being infused in the system, a 200 mL sample of 15 N enriched saliva and a 200 mL sample of digesta effluent (liquid and solid) from each fermenter were collected to measure the concentration of 15 N in enriched saliva and 15 N background concentration in digesta, respectively. On days 8, 9, and 10, a 200 mL sample of effluent (liquid and solid) was collected from each fermenter for measurement of 15 N concentration in digesta effluent. Also, on day 10 of each period, the whole content of each fermenter was processed for isolation of bacteria according to Krizsan et al. (2010). Briefly, fermenter contents were blended for 30 s, filtered through four layers of cheesecloth, and washed with 400 mL of 0.9% saline solution. The filtrate was centrifuged (Sorvall RC-5B Refrigerated Superspeed Centrifuge, DuPont Instruments) at 1,000 × g for 10 min and the resulting supernatant was centrifuged at 11,250 × g for 20 min to obtain the pellet of bacteria that was resuspended in 200 mL of McDougall's solution, and centrifuged at 16,250 × g for 20 min. Finally, the bacteria pellet was transferred with distilled water into a plastic container for storage.
Immediately after collection, samples of 15 N enriched saliva, 15 N background, digesta effluent, and bacteria, were stored at −20 °C until freeze-dried. Dried samples were first ground with a mortar and pestle and then ball-milled at 25 Hz for 9 min using a Mixer Mill MM400 (Retsch, Newton, PA) for analysis of 15 N.
Volatile fatty acids. Concentrations of acetate, propionate, butyrate, isobutyrate, isovalerate, valerate, and caproate were analyzed by gas chromatography (Agilent 7820A GC, Agilent Technologies, Palo Alto, CA) with a flame ionization detector and a capillary column (CP-WAX 58 FFAP 25 m 0.53 mm, Varian CP7767, Varian Analytical Instruments, Walnut Creek, CA) that was maintained at 110 °C, with injector temperature at 200 °C and detector at 220 °C. Samples were thawed at room temperature and processed according to Ruiz-Moreno et al. (2015) by centrifuging at 10,000 × g for 15 min, mixing the supernatant with a solution of crotonic acid and metaphosphoric acid to freeze overnight, and then centrifuged again at 10,000 × g for 15 min. The supernatant was mixed with ethyl acetate, vortexed and allowed to settle to finally transfer the top layer to a chromatography injection vial for analysis.
Ammonia nitrogen. Analysis of NH 3 -N in ruminal fluid samples was done according to Broderick and Kang (1980). Samples were thawed at room temperature, centrifuged at 10,000 × g for 15 min, and the supernatant used for analysis by the phenol-hypochlorite method in a 96-well flat-bottom plate. Absorbance was measured at 620 nm using a spectrophotometer (SpectraMax Plus 384 Microplate Reader).
Soluble Mg. Analysis of soluble Mg in ruminal fluid samples was performed according to Jittackhot et al. (2004). Briefly, samples were centrifuged at 18 °C at 2,700 × g for 15 min, then the supernatant was centrifuged 20 °C at 30,000 × g for 30 min, and an aliquot was collected for Mg analysis by inductively coupled plasma mass spectrometry (AOAC, 2000; method 985.01).
Percent atom 15 N analysis. Based on expected 15 N concentrations 1, 2, or 4 mg of ball-milled sample (for samples of bacteria, digesta effluent, or background, respectively) were weighed into an 8 × 5 mm pressed standard-weight tin capsule (Elemental microanalysis, Devon, UK) using a Mettler Toledo Excellence Plus XP Micro Balance (Mettler-Toledo GmbH, Laboratory & Weighing Technologies, CH-8606 Greifensee, Switzerland), then 35 µL of K 2 CO 3 solution (10 g/L) were added to each sample and left overnight in a forced-air oven at 40 °C for complete evaporation of NH 3 -N. Finally, % atom 15 N in samples was determined by isotope ratio mass spectrometry (IsoPrime 100, IsoPrime, Manchester, UK) and expressed as the fractional abundance of isotopic fractions ( 15 N/ 14 N) multiplied by 100. Furthermore, these samples were analyzed for DM (AOAC, 1990;method 930.15), ash (AOAC, 1990; method 942.05), and concentration of total N (AOAC, 2000; method 990.03) by rapid combustion with a micro elemental N analyzer (Vario Micro Cube, Elementar, Hanau, Germany).

Calculations for N Metabolism and Digestibility of Nutrients
Total N in digesta effluent was partitioned into three main fractions: NH 3 -N flow, dietary N flow (undigested N), and bacterial N flow. The two latter fractions combined are also known as nonammonia N (NAN).
Flows of NH 3 -N and NAN were determined according to Bach and Stern (1999), while the flow of bacterial N was determined according to Calsamiglia et al. (1996) using the following equations: Additionally, the flow of dietary N and indicators of N metabolism by microorganisms were also determined according to Bach and Stern (1999) as follows: Ruminal digestibility of OM, CP, and NDF was estimated according to Soder et al. (2013) by calculating the proportion of undigested nutrients in the digesta effluent and correcting for the contributions of artificial saliva and bacteria, as shown:

Statistical Analysis
Analysis of variance was performed with the MIXED procedure of SAS 9.4 (SAS Institute Inc., Cary, NC). Data were analyzed as a 4 × 4 Latin square design where the model included treatment as a fixed effect, and random effects of period and fermenter. Additionally, pH data were analyzed as repeated measures using the AR1 covariance matrix structure (based on lowest AIC criterion) and this model also included the fixed effects of time (expressed as hours after feed provision) and time × treatment. Multiple comparisons between treatments: 1) Control; 2) CO 3 ; 3) OH; and 4) CO 3 /OH were evaluated with Tukey's test. Moreover, the following orthogonal contrasts were evaluated: [CaMg(CO 3 ) 2 ] = (CO 3 + CO 3 /OH) vs. Control to compare the effect of CaMg(CO 3 ) 2 as a source of Mg with the use of MgO plus a buffer; and [CaMg(OH) 4 ] = (OH + CO 3 /OH) vs. Control, representing the comparison between the use of CaMg(OH) 4 as a source of Mg and the use of MgO plus a buffer. Significance was declared at P ≤ 0.05, while 0.05 < P < 0.10 was considered as trend.

Ruminal pH and Concentration of Soluble Mg
Results on pH-related variables and concentration of soluble magnesium are shown in Table 2. Repeated measures analysis indicated that interaction treatment × time was not significant; therefore, pH kinetics data are not presented. Average pH of the ruminal fluid in the fermenters was similar across experimental treatments. Similarly, orthogonal contrasts show that the inclusion of either CaMg(CO 3 ) 2 or CaMg(OH) 4 as sources of Mg in the diets, yields similar average pH compared to the Control treatment formulated with MgO and sodium sesquicarbonate.
On the other hand, pH-B6 was affected by treatment (P < 0.01) showing that CO 3 treatment spent a greater percentage of total time below the pH 6.0 threshold when compared to the three other treatments. Consequently, orthogonal contrasts show that using CaMg(CO 3 ) 2 as a source of Mg tends to increase pH-B6 with respect to the Control ([CaMg(CO 3 ) 2 ]; P = 0.06), while pH-B6 with the inclusion of CaMg(OH) 4 in the diet was similar to the Control. The area under the curve below pH 6.0 (pH-AUC) was also affected by treatment (P < 0.01), with the CO 3 treatment having a greater pH-AUC than the other treatments, indicating more acidic conditions for CO 3 in comparison to the other treatments. Similarly, orthogonal contrasts indicated that pH-AUC of the Control treatment with MgO and buffer was smaller than pH-AUC of diets formulated with CaMg(CO 3 ) 2 ([CaMg(CO 3 ) 2 ]; P = 0.05), but was similar to pH-AUC observed in diets with CaMg(OH) 4 as a source of Mg.
Our results show a similar response on pH, pH-B6, and pH-AUC between the diets containing CaMg(OH) 4 as a source of Mg and the Control diet formulated with MgO plus sodium sesquicarbonate. The lack of differences in this case indicates that inclusion of CaMg(OH) 4 as a source of Mg in the diet, allowed for similar pH-related conditions than those observed in a Control diet with a commercial buffer and MgO. Conversely, under the conditions of this experiment, the inclusion of CaMg(CO 3 ) 2 as a source of Mg allowed for a more acidic fermentation than our positive Control diet.
Recently, Leno et al. (2017) reported greater DMI and milk fat yield during the first week of lactation in Holstein cows when CaMg(CO 3 ) 2 was supplemented during the peripartum compared to MgO supplemented cows. Ruminal fermentation was not evaluated in that trial, therefore it is not possible to know whether differences in DMI were related to ruminal pH or just different acceptance of MgO and CaMg(CO 3 ) 2 by the cows, indirectly impacting intake; although greater milk fat concentration may suggest possible changes in ruminal fermentation promoted by CaMg(CO 3 ) 2 . Similarly, greater DMI has been already associated with the positive response in milk production observed in dairy cows supplemented with MgO (Bach et al., 2018), suggesting that the productive response to supplementation with inorganic sources of Mg can be, at least to some extent, driven by changes in intake.
The dual-flow continuous culture system allows to control the intake or supply of dry matter and nutrients. In the case of our experiment, the supply of nutrients was held constant across treatments; eliminating the effect of intake or preference that could be observed in other types of experiments, and consequently isolating the effect of treatments on ruminal fermentation; which in this case indicate that CaMg(OH) 4 had a greater alkalinizing effect than CaMg(CO 3 ) 2 .
We did not observe any effect of treatment on the concentration of soluble Mg in ruminal fluid in our experiment (Table 2), indicating similar solubilization among the sources we evaluated. Decreased ruminal absorption of Mg in sheep (Rahnema and Fontenot, 1983) and lower plasma Mg concentration in cows (Leno et al., 2017), have been reported when animals are fed CaMg(CO 3 ) 2 in comparison to MgO. However, none of these trials evaluated ruminal solubility of Mg; therefore, such differences in absorption or availability of Mg cannot be attributed to differences in ruminal solubility which has been estimated to explain most of the variation observed in the absorption of Mg among different sources (NRC, 2001;Jittakhot et al., 2004). Interestingly, the differences in reactivity between sources, that we observed in acetic acid solution, were not translated into differences in Mg solubility in ruminal fluid. Such discrepancy may be due to factors such as average pH, temperature, and fluctuation in pH, which differ between the acetic acid solution and the ruminal contents in the dual-flow culture system.

Volatile Fatty Acids
Results of total VFA concentration and individual molar proportions are presented in Table 3. The total concentration of VFA and molar proportions of acetate and propionate were similar across treatments while there were no significant differences in the contrasts for these variables either. Butyrate molar proportion was greater in Control and CO 3 /OH treatments than in CO 3 (Trt; P = 0.04); however, butyrate molar proportion in OH did not differ from CO 3 . Orthogonal contrasts indicated that the molar proportion of butyrate tended to be lower for diets supplemented with CaMg(CO 3 ) 2 compared to Control ([CaMg(CO 3 ) 2 ]; P = 0.08). Similarly, the molar proportion of caproate was lower in CO 3 than in the other three treatments (Trt; P = 0.03). Molar proportions of VFA resulting from amino acid degradation are also shown in Table 3. Both isobutyrate (P = 0.06) and isovalerate (P = 0.07) molar proportions tended to be affected by treatments. Inclusion of CaMg(CO 3 ) 2 as an Mg source in the diet promoted a greater molar proportion of isovalerate compared to the Control treatment ([CaMg(CO 3 ) 2 ]; P = 0.01), indicating greater deamination of leucine by ruminal microorganisms (Allison, 1978) when CaMg(CO 3 ) 2 is supplemented; which is reflected in a trend towards a greater BCVFA molar proportion in CaMg(CO 3 ) 2 treated diets compared to Control ([CaMg(CO 3 ) 2 ]; P = 0.09), and maybe indicative of either greater ruminal degradation of the branched-chain amino acid leucine or lesser utilization of isovalerate by ruminal microorganisms (Apajalahti, 2019).
Our results indicate that both the diets with CaMg(OH) 4 as a supplemental source of Mg and the Control treatment with MgO and sodium sesquicarbonate may favor butyrate synthesis and a fermentation profile that could eventually allow for a greater energy supply to the ruminal epithelium (Kristensen and Harmon, 2004); similar to the in vivo response to bicarbonate supplementation on ruminal fermentation and production (Erdman, 1988;Staples and Lough). Moreover, although much less studied, caproate has been reported as a ketogenic VFA (Kristensen and Harmon, 2005) that also has a positive correlation with acetate synthesis in the rumen (Ertl et al., 2015).
In a summary of 82 experiments, Erdman (1988) found that cows consuming diets with corn silage as the source of forage, had lower propionate concentrations in ruminal fluid, and greater acetate concentration and acetate:propionate ratio when supplemented with sodium bicarbonate. Cruywagen et al. (2015) found an increase in ruminal acetate and butyrate, in response to supplementation with either bicarbonate or calcareous marine algae. More recently, Iwaniuk and Erdman (2015) evaluated the effect of the cation-anion difference of the diet on performance of dairy cows supplemented with buffers; their meta-analysis showed a positive association between dietary cation-anion difference and both acetate and butyrate molar percentages. The similar response on butyrate molar proportion observed in our experiment between diets with CaMg(OH) 4 inclusion and Control diet in addition to the response on ruminal pH is consistent with some of these reports in cows supplemented with buffers which seem to stimulate the ruminal synthesis of lipogenic VFA.

Flows of N and N Metabolism
The effects of CaMg(CO 3 ) 2 and CaMg(OH) 4 on the concentration of NH 3 -N were analyzed in samples of digesta effluent and used for estimation of N flows which are shown in Table 4. There were no effects of treatment on NH 3 -N concentration, total N flow, NH 3 -N flow, or NAN flow. None of these variables were altered by CaMg(CO 3 ) 2 or CaMg(OH) 4 inclusion when compared to the Control diet according to the orthogonal contrasts analysis. Nevertheless, within the fraction corresponding to NAN flow, the flow of bacterial N tended to be influenced by treatment (Trt; P = 0.06) and also was lower when diets containing CaMg(CO 3 ) 2 were compared to the positive Control containing MgO plus buffer ([CaMg(CO 3 ) 2 ]; P = 0.04). Moreover, inclusion of CaMg(CO 3 ) 2 in the diet also tended to decrease the efficiency of utilization of available N ([CaMg(CO 3 ) 2 ]; P = 0.06) and digestible OM ([CaMg(CO 3 ) 2 ]; P = 0.08) for synthesis of bacterial N.
As discussed before, our results on isovalerate and BCVFA shown in Table 3 indicate either a greater ruminal degradation of leucine or reduced utilization of isovalerate by ruminal microorganisms, allowing for a greater molar proportion of this VFA when CaMg(CO 3 ) 2 was fed in comparison to Control diet. Based on the results of N metabolism, a possible explanation is that isovalerate accumulates in the ruminal fluid as a consequence of reduced uptake by ruminal bacteria due to a decreased microbial synthesis when CaMg(CO 3 ) 2 is used as the supplemental source of Mg in the diet. Ruminal cellulolytic bacteria require BCVFA as growth factors due to their inability to directly incorporate preformed amino acids which makes them dependent on BCVFA and NH 3 to perform bacterial protein synthesis (Dehority et al., 1967). Then, in this case, a trend towards a greater molar  e ENU = efficiency of N use = (g of bacterial N/g of available N) × 100 (Bach and Stern, 1999).
f Bacterial efficiency = g of bacterial N flow/kg of OM truly digested, according to Calsamiglia et al. (1996). proportion of BCVFA observed in diets containing CaMg(CO 3 ) 2 compared to Control, may be indicative of decreased growth rate of cellulolytic bacteria; however bacterial populations analysis would be required to evaluate this possibility. Decreased bacterial N flow influenced by lower ruminal pH in a dual-flow continuous culture was reported by Calsamiglia et al. (2008) who estimated that 39% of the variation in bacterial N flow was explained by pH. Although they did not evaluate bacterial populations, a possible explanation for limited bacterial growth at lower pH is the greater maintenance requirements and consequently lower growth efficiency of nonstructural carbohydrates-degrading bacteria compared to cellulolytic bacteria (Russell et al., 1992).

Nutrient Digestibility
As shown in Table 5, digestibility of OM, CP, and NDF was similar among treatments. Additionally, the inclusion of either CaMg(CO 3 ) 2 or CaMg(OH) 4 in the diet allowed for similar values of nutrient digestibility than the Control diet, as indicated by orthogonal contrasts. In two reviews summarizing 82 and 41 experiments, respectively, Erdman (1988) and Staples and Lough (1989) reported greater digestibility of DM and NDF or ADF in cows fed corn-silage based diets when supplemented with 1% sodium bicarbonate or sesquicarbonate at 0.75% to 1% of diet DM. Moreover, supplementing MgO at 0.4% to 0.8% of DM increased fat corrected milk yield compared to non-supplemented cows (Staples and Lough, 1989). However, when cows were fed a different source of forage, such as alfalfa hay, or another source of fiber was included in the diet, such as cottonseed hulls, there was no response to buffers supplementation, showing that effectivity of buffers depends on the type of diet and that corn silage based diets with a rather high concentration of starch (>30%) are the ones that show a better response (Erdman, 1988). Our experimental diets met those criteria offering the right conditions for buffer effectiveness, which means that adequate substrate was provided for the synthesis of organic acids from microbial fermentation of starch and sugars, and within that context, the inclusion of CaMg(CO 3 ) 2 and CaMg(OH) 4 had a similar effect on nutrient digestibility, and such effect was also comparable to that one of the Control diets with MgO plus buffer.
Although it is still unclear the mechanism on how ruminal buffers and alkalizers alter ruminal digestibility (Iwaniuk and Erdman, 2015), it has been shown that reducing ruminal acidity by the inclusion of these products in the diets of dairy cows, not only can increase the digestibility of NDF, but also may modify lipid biohydrogenation pattern in the rumen by increasing synthesis of cis-9, trans-11 conjugated linoleic acid (Jenkins et al., 2014); which explains part of the response observed in milk fat synthesis of animals supplemented with ruminal pH modifiers. We did not evaluate lipid biohydrogenation in our experiment; therefore, we can only discard a difference in NDF digestibility between our treatments. Moreover, all the diets in our experiment were formulated to the same concentration of Mg based on the NRC's recommendations (2001). Therefore, our results rule out the confounding effect of a different concentration of Mg among treatments and focus on the effect of source of supplemental Mg, and the inclusion of buffer in the case of the Control treatment; indicating that although it had an impact on the pattern of VFA synthesis, it may have not been such a big change as to modify nutrient digestibility.
Therefore, under the conditions of this experiment, the inclusion of CaMg(OH) 4 in diets as a source of Mg, either alone or combined with CaMg(CO 3 ) 2 , allowed for a less acidic ruminal fermentation pattern than a diet formulated with CaMg(CO 3 ) 2 as the only supplemental source of Mg. Moreover, using CaMg(OH) 4 , either alone or combined with CaMg(CO 3 ) 2 , allowed similar microbial fermentation variables than a positive Control diet formulated with MgO plus sodium sesquicarbonate as a buffer. Further research would be needed to evaluate in vivo response to supplementation with these sources of Mg and also their availability for ruminants.