The effects of substrate dispersion method, substrate type, and their interaction were evaluated in a 3 × 3 factorial arrangement with three dispersion methods (samples incubated directly dispersed in the medium, within F57 bags, and within non-woven bags, NWT) and three types of substrates with different forage:concentrate ratio (low-quality tropical grass hay only, 100F; a low-quality tropical grass hay and concentrate mixture in a 50:50 ratio, mixture; and concentrate mixture only, 100C, all on a dry matter (DM) basis). The three substrates were used to cover a large variability in chemical composition (low, average, and highly fermentable) and to generate different in vitro fermentation end-products in the bottle during incubation.

Low-quality tropical grass (Brachiaria brizantha cv. BRS Piatã) was collected from a dry season cutting of the forage available in a pasture located in the Midwest region of Brazil. The concentrate feed consisted (DM basis) of a mixture of ground corn (838 g/kg) and soybean meal (162 g/kg). Forage and concentrate were ground in a knife mill (Willey Type Mill, TE-650; Tecnal, Piracicaba, SP, Brazil) to pass through a 2-mm screen. After that, half of each ground sample was ground again to pass through a 1-mm screen. The samples ground through 1-mm sieves were analyzed for DM (dried overnight at 105 °C; method INCT-CA number G-003/1), ash (complete combustion in a muffle furnace at 600 °C for 4 h; method INCT-CA number M-001/1), nitrogen (Kjeldahl procedure; method INCT-CA number N-001/1), and neutral detergent fiber corrected for ash and protein (NDFap, using a heat-stable α-amylase, omitting sodium sulfite and correcting for residual ash and protein; method INCT-CA number F-002/1) by chemical analytical methods according to the standard analytical procedures of the Instituto Nacional de Ciência e Tecnologia de Ciência Animal (Detmann et al., 2012). Low-quality tropical grass hay and concentrate contained (g/kg DM): 926 and 894 g DM/kg, 952 and 980 organic matter (OM), 25 and 150 crude protein (CP), 678 and 143 NDFap, 12 and 625 starch, respectively.

Fresh rumen fluid (pH = 7.07 ± 0.144) was collected manually from the cranial, ventral, and caudal areas of two cannulated Holstein steers (442 ± 27.6 kg of body weight; housed at a metabolism barn) before morning feeding to obtain stable rumen microbial cultures. These two steers grazed Marandu palisade grass (Brachiaria brizantha) pasture and received 4 kg/day of commercial concentrate provided twice daily (09:30 and 17:30 h) to establish about 400 g concentrate kg^-1 diet, on a DM basis (160 g CP/kg, based on ground corn, soybean meal, cottonseed cake, vitamin, and trace mineralized salts; Fortuna Nutrição Animal, Nova Canaã do Norte, MT, Brazil). The rumen fluid collected from each of the steers was mixed in equal amounts, filtered through three layers of cheesecloth, and used as the inoculum in the in vitro rumen fermentation.

The in vitro batch fermentation was carried out using a commercial wireless gas production (GP) apparatus (Ankom^RF GP System, Ankom Technology^®, Macedon, NY, USA) consisting of 24 bottles equipped with pressure sensors and wireless connection to a computer. Four incubation runs were conducted in four successive weeks. Each treatment (all combinations between the three dispersion methods and three substrates) plus blanks (bottles containing only the rumen fluid and the buffer solution) was run in duplicate across four successive runs (blocks). Samples of substrates milled through a 1-mm screen were weighed directly into bottles (310 mL), incubated within F57 filter bags (Ankom Technology Corp., Macedon, NY), or incubated using manufactured NWT filter bags (100 g/m²) of 4.0 × 4.5 cm (Valente et al., 2011) that were sealed and placed into the bottles. The F57 and NWT filter bags presented 36 cm² of surface, and samples were put in the bags following the ratio of 20 mg DM/cm² of surface (Nocek, 1988). Both types of filter bags were previously washed with commercial neutral detergent, water, and acetone, and were dried in an oven at 55 ℃ before samples were weighed and sealed. This washing procedure was done to remove the surfactant added during the manufacturing process of the F57 bag, avoiding microbial digestion inhibition (Krizsan et al., 2012).

The buffer solution was prepared according to McDougall (1948), heated at 39 °C, purged continuously with nitrogen (N₂) for 30 min to maintain anaerobic conditions, and pH was adjusted to 6.8 with orthophosphoric acid (Carreño et al., 2015). The anaerobic buffered medium (40 mL) and the rumen inoculum (10 mL) were mixed (buffer:rumen fluid ratio of 4:1), dispensed into each bottle, pre-flushed with N₂ (Benedeti et al., 2018; Patra and Yu, 2013), and incubated for 48 h at 39 ± 0.5 °C. The bottles were shaken gently for proper mixing of feeds or bags with microbial cultures.

The data acquisition software (Gas Pressure Monitor, Ankom technology, NY, USA) was set to monitor the cumulative pressure every 5 min, and data was recorded every 30 min for 48 h. Valves were set to automatically release the gas when the pressures reached 3.4 kPa (Tagliapietra et al., 2012). After correction for the blanks, the recorded cumulative gas pressure was converted to mL of gas produced under the manufacturer’s instructions using Avogadro’s law (gas volume ( m L ) = gas pressure × [ V / ( R × T ) ] × 22.4 × 1 , 000 , in which V is headspace volume in the bottle in L = 0.26, R is the gas constant 8.314472 L kPa/K/mol, and T is the temperature in Kelvin).

Gas samples were drawn using a 10-mL gas-tight syringe from each bottle headspace at 24 and 48 h of incubation and stored in a 10-mL vacuum tube. Gas chromatograph (EZChrom Elite software interface, Model 7820A, Agilent Technologies Brazil, Barueri, SP, Brazil) was used for the quantification of CH₄ and CO₂. The gas chromatograph equipment was equipped with two six-way valves, one being used for the sampler system connected to a loop of 0.5 mL. Injector split-splitless type was used in split mode to 50:1 at 120 ℃. The separation system consisted of two columns. The first column was a HP- Plot/Q 30 m × 0.530 m m × 40.0 m m (Agilent Technologies Brazil). The detection system comprised a thermal conductivity detector at 250 ℃, with 25 mL H₂/min as flow reference. The second column was a Molesieve HP-30 m × 0.530 mm × 25.0 mm, using H₂ as carrier gas at a flow rate of 8.3 mL/min, with flame ionization detector at 270 ℃ and 15 mL/min H₂ flow rate, and 350 mL/min synthetic air flow. The gas chromatograph was also equipped with a methanizer maintained at 375 ℃, which allows detecting very low concentrations of CO₂. The oven temperature was maintained at 55 ℃. The calibration curves were performed with reference standards for CH₄ concentrations, as follows: 0 mL/L, 50.5 mL/L, 102 mL/L, 147 mL/L, and 201 mL/L; and for CO₂ as follows: 0 mL/L, 202 mL/L, 397 mL/L, 583 mL/L, and 799 mL/L.

After 48 h of incubation, fermentation was stopped by placing serum bottles in ice-water bath for 15 min. For the substrates that were dispersed directly into the buffered ruminal fluid, the contents of the bottles were filtered under vacuum through glass crucibles with a sintered filter (coarse porosity no. 1, pore size 100 to 160 μm). The fluid pH was measured with a pH meter (PG 2000 Portable pH meter; Gehaka, São Paulo, SP, Brazil) by submerging the probe approximately 3 cm in the fermentation media. An aliquot of fermentation media (1.2 mL) was combined with 0.4 mL of 250 g/L metaphosphoric acid and frozen for subsequent analysis of volatile fatty acids (VFA). Following, to estimate the apparently undigested DM residue, fermentation residues in grass crucibles and bags were oven-dried (55 °C/24 h and 105 °C/16 h, sequentially), placed in a desiccator, and weighed. The DM incubation residues in grass crucibles and filter bags were ashed at 600 °C for 4 h to estimate OM disappearance. Blank bags were also dried and ashed for the correct weight.

Sample VFA profile was obtained using a high-performance liquid chromatograph (Varian Pro Star 325 LC) equipped with an ultraviolet detector. The separation system consisted of a C18 Supelco reverse phase column (250.0 × 4.6 mm, 5 μm). The elution was carried out with an isocratic mobile phase consisting of an aqueous solution of metaphosphoric acid (HPO₃) (eluent A, pH adjusted to 2.5) and acetonitrile (eluent B). The gradient was constituted of 920 mL/L A and 80 mL/L B, with a flow of 1 mL/min and a running time of 8 min. The injection volume was 20 μL and detection at 210 nm. Quantification was obtained using a calibration curve with external standards.

Gas production was calculated as the volume of gas (mL) produced after 24 h (GP24) and 48 h of incubation (GP48) divided by the amount of DM incubated (DMi, g). The fraction of in vitro DM (IVDMD) and in vitro OM (IVOMD) digestibilities for each bottle was calculated by subtracting the DM and OM residue weight (corrected for the blank) from the DM and OM substrate weight and dividing by the weight of substrate, respectively. The partitioning factor at 48 h of incubation (PF; a measure of fermentation efficiency) was calculated as the ratio of OM degradability in vitro (OMd, mg) to the volume (mL) of GP at 48 h (i.e., OMd/total GP), according to Blümmel et al. (1997b). The change of total VFA concentration in the medium during incubation was calculated by subtracting the total VFA concentration of the initial medium (rumen fluid and buffer) from the total VFA concentration in the medium after 48 h of incubation (ΔVFA). Apparent change of VFA per g OM incubated (ΔVFA/OMi) or g OM digested (ΔVFA/OMd) were calculated by multiplying total VFA by the volume of medium (50 mL) and dividing by the weight of substrate of OM incubated or OM digested, respectively. Acetic, propionic, and butyric acids were expressed as molar proportion relative to total VFA (mol/100 mol).

Because the present experiment was conducted without the use of tight bags for gas collection to measure the volume and concentration in vented gas, we used the following equation described by Cattani et al. (2016) to estimate methane production ( M P , m L ) : M P = H S C H 4 × ( 260 + 0.0057 × G P 2 ) , in which HCH₄ (L/L) is the CH₄ proportion in the headspace and 260 (mL) is the total headspace volume. The same equation was also used to calculate CO₂ production, considering that CO₂ and CH₄ are in equilibrium in the headspace (i.e., the proportions of gases in the vented gas are the same as in the headspace), and the outflow of CO₂ and CH₄ produced is relative to their concentrations in the headspace (Ramin and Huhtanen, 2012). Methane and CO₂ yield were expressed relative to DM incubated and OM digested (mL/g). Proportion of CH₄ and CO₂ (mL/L) in the fermentation gas was calculated by multiplying the production of the relative greenhouse gas by 1,000 and dividing by GP.

To assess biological values, gas curves were fitted to a dual-pool logistic equation derived on the assumption that the rate of gas production is proportional to both the accumulated microbial mass and to the amount of digestible substrate remaining (Schofield et al., 1994). This equation is: V = V F / { 1 + exp ⁡ ( 2 + 4 × ( S R F × ( L a g − t ) ) ) } + V S / { 1 + ( exp ⁡ ( 2 + 4 × ( S R S × ( Lag − Time ) ) ) } , in which V = gas volume produced up to the specific time t (mL); V_F and V_S = maximum gas volumes (mL) achieved from complete digestion of the first and second pool (i.e., faster- and slower-digesting fractions, respectively); SR_F and SR_S = specific rate of digestion (h^-1) for these fractions; and Lag = lag time (h).

All statistical analyses were performed using the MIXED procedure of SAS (SAS Inst., Inc., Cary, NC), with the mean of the bottles within substrates in each incubation run (true replicate) as the experimental unit. Means for all variables were obtained using the LSMEANS statement. The model was as follows:

in which μ is the general mean, I_i is the random effect of incubation i, M_k is the fixed effect of dispersion method k, S_j is the fixed effect of substrate j, (M × S)_jk is the fixed effect of interaction between substrate type and dispersion method, and ɛ_ijk is the random error.

For GP, CH₄, and CO₂ proportion, emission, and yield, data were reported at 24 and 48 h of incubation. For total gas production over time data, logistic nonlinear functions for two pools and a discrete lag (Schofield et al., 1994) were adjusted for substrates (forage:concentrate ratio of 100:0, 50:50, and 0:100), as well as incubation methods, to compare possible differences in fermentation profiles. Treatment mean values were calculated using the LSMeans statement in SAS and, when the interaction between factors was significant (P<0.05), they were separated using PDIFF option with Tukey’s adjustment. Orthogonal polynomial contrasts were used to examine linear and quadratic effects of increasing concentrations of forage when the effects of substrate were found significant (P<0.05). Pearson’s correlation of variables at 48 h of fermentation among methods were analyzed using PROC CORR of SAS.

There were interactions between different dispersion methods and substrate forage:concentrate ratio (P<0.02) on GP24, GP48, IVDMD, IVOMD, and PF (Figure 1, panel A–E), while no effect interaction was observed on pH (Figure 1F) (P = 0.80). There were no differences between methods (P≥0.08) for GP24 when the 100F ratio substrate was incubated (averaging, 71.6 mL/g DMi). The incubation of mixture and 100C substrates promoted greatest GP24 in DIS followed by NWT and F57 (P≤0.03). The method × substrate interaction showed similar effects for GP48 and GP24 (P≤0.04), except those when the 100F ratio substrate was incubated in NWT, which had greater gas production than in DIS and F57 (P<0.01).

Figure 1 White, gray, and black bars correspond to samples that were incubated directly in the medium, within F57 and NWT bags, respectively. Error bars represent standard error of the mean. LIN and QUA represent linear and quadratic effects of forage:concentrate ratios, respectively. For substrate dispersion method treatment effect, means within forage:concentrate ratios with different letters (a–c) differ (P<0.05).

The in vitro digestibilities (IVDMD and IVOMD) presented similar results (Figures 1C and D, respectively). The IVDMD and IVOMD were about 82 and 88% greater (P<0.001), respectively, when 100F substrate was incubated in NWT than in DIS and F57. With mixture substrate incubation, there were no differences for in vitro digestibilities between DIS and NWT (P≥0.58; on average, 578 and 585 g/kg for IVDMD and IVOMD, respectively), but they were greater than F57 (P<0.01; 494 and 499 g/kg, respectively). The in vitro digestibilities of concentrate substrate (100C) were enhanced in DIS, intermediate in NWT, and decreased in F57 (P<0.01). For IVOMD, these values were 918, 838, and 722 g/kg, respectively.

The PF was 46% greater in NWT method (P<0.05) than in DIS and F57, and both presented similar results (P = 0.26) when 100F substrate was incubated. When mixture substrate was incubated, PF was greater in NWT and smallest in DIS (P = 0.04; 1.70 and 1.52 mg OMd/mL, respectively), and F57 presented value similar to them (P≥0.08; 1.67 mg OMd/mL). With the concentrate incubation (100C), differences became less evident, without difference among methods (P≥0.11). The pH was affected only by substrate (P<0.001). As expected, increasing forage inclusion linearly increased pH from 6.00 to 6.52 (P<0.001).

A method × substrate interaction was found for CH₄ yield relative to DM incubated (DMi) (Figure 2B), but not for CH₄ proportion (Figure 2A) and yield relative to OMd (Figure 2C) at 24 h of incubation. There was no difference between methods in CH₄/DMi (P≥0.52) when 100F substrate was incubated. But for the other evaluated substrates, DIS produced about 2.4 times greater CH₄/DMi compared with filter bags (P<0.001). There were method effects on CH₄ proportion and CH₄/OMd (P≤0.03), with estimates about 67% greater when substrates were incubated in DIS than into filter bags (P≤0.01), but there were no differences between F57 and NWT bags (P≥0.06). Interactions were observed between the substrate dispersion method and forage:concentrate ratio of substrates with respect to CO₂ yield relative to OMd (P = 0.04), but not to CO₂ proportion (Figure 2C; P = 0.60), and CO₂ yield relative to DMi (Figure 2F; P = 0.31). When 100F substrate was incubated, CO₂ yield relative to OMd was greater in DIS, intermediate in F57, and lower in NWT method (P≤0.01). For the other substrates, methods provided similar results (P≥0.07). The CO₂ yield relative to DMi linearly increased when forage:concentrate ratio decreased (P<0.01).

Figure 2 White, gray, and black bars correspond to samples that were incubated directly in the medium, within F57 and NWT bags, respectively. Error bars represent standard error of the mean. LIN represents linear effect of forage:concentrate ratios. For substrate dispersion method treatment effect, means within forage:concentrate ratios with different letters (a–c) differ (P<0.05).

At 48 h of incubation, there were interactions between substrates and dispersion method (P≤0.01) on CH₄ proportion (Figure 3A, mL/L) and yield relative to DM incubated (Figure 3B, mL/g DMi), but not on yield relative to OM digested (Figure 3C, mL/g OMd; P = 0.47). The CH₄ proportion values were slightly altered (P<0.001) by the methods (DIS > filter bags). When 100F substrate was incubated, methods presented similar CH₄ yield relative to DMi (P≥0.12). However, differences between methods in CH₄ yield/DMi became more pronounced when substrate fermentability increased, in which DIS showed greater CH₄ yield/DMi than when incubated in the F57 and NWT bags (P<0.001; 13.2 vs. 8.9 mL/g DMi). There was method effect on CH₄ yield relative to OMd (P<0.001), with lower values found with the incubation substrates within bags than in DIS (14.6 vs. 32.2 mL/g OMd). Additionally, CH₄ yield/OMd linearly decreased in response to decrease of forage:concentrate ratio (P<0.001). Regarding CO₂ output at 48 h of incubation (Figure 3D–F), greater proportion and yield relative to OMd (P≤0.04) were noted for DIS compared with incubation within filter bags (118 vs 98 mL/L and 47.6 vs. 32.2 mL/g OMd, respectively). With decreasing forage:concentrate ratio of the substrates, CO₂ yield/OMd linearly decreased (P<0.001). There was interaction (P<0.01) between substrate and dispersion method for CO₂ yield relative to DM incubated. The CO₂ yield/DMi did not differ (P≥0.60) between the three incubation methods when the substrate contained forage only, but it was 3.4 times greater in DIS than in filter bags when the substrate contained concentrate only (P<0.001). When the mixture substrate was incubated, CO₂ yield relative to DMi was greater in DIS than in F57 (P = 0.05), but NWT was similar to DIS and F57 (P≥0.07).

Figure 3 White, gray, and black bars correspond to samples that were incubated directly in the medium, within F57 and NWT bags, respectively. Error bars represent standard error of the mean. LIN represents linear effect of forage:concentrate ratios. For substrate dispersion method treatment effect, means within forage:concentrate ratios with different letters (a,b) differ (P<0.05).

There were no method × substrate interactions (P≥0.06) on kinetic variables of gas production (Figure 4A–E). The maximum gas volumes and specific rates of digestion for first and second pools were affected by substrates (P≤0.01), but in opposite ways. While maximum gas volumes and specific rates of digestion for first pool (fast degradation) increased linearly (P<0.01), these variables for second pool (slow degradation) depressed linearly (P<0.01) with each reduction in forage:concentrate ratio (i.e., increase of substrate fermentability). No effects were observed on the lag time (P≥0.43), averaging 2.37 h. Additionally, only specific rates of digestion for first pool differed between methods (P<0.01). The DIS had greatest specific rates of digestion for first pool than that into bags (0.035 vs. 0.024 h^-1; P≤0.02), but F57 was similar to NWT (P = 0.14).

Figure 4 White, gray, and black bars correspond to samples that were incubated directly in the medium, within F57 and NWT bags, respectively. Error bars represent standard error of the mean. LIN represents linear effect of forage:concentrate ratios.

There were no interactions between method × substrate (P>0.26) on molar proportions of individual VFA (i.e., acetic, propionic, and butyric), acetic:propionic ratio, total VFA, and Δ total VFA (total VFA post-incubation – total VFA pre-incubation; Figure 5, panel A–E). Dispersion methods had effects on molar proportion of propionic acid and acetic:propionic ratio (P≤0.04). The propionic acid (mol/100 mol) was greater (P≤0.03) when substrates were incubated within bags than in DIS (8.11 vs. 7.25 mol/100 mol), but no differences were found for F57 and NWT (P = 0.92). In an opposite way, acetic:propionic acid ratio showed lower values (P≤0.03) when substrates were incubated into bags than in DIS (3.20 vs. 4.21), but without differences between F57 and NWT (P = 0.57). Substrate affected (P<0.01) molar proportion of acetic and propionic acids, acetic:propionic acid ratios, total VFA, and ΔVFA in the medium. Acetic molar proportion and acetic:propionic ratio decreased linearly (P<0.01), while molar proportion of propionic acid, total VFA, and ΔVFA increased linearly (P<0.001) with each decrease in forage:concentrate ratio.

Figure 5 Δ Total VFA = change of total VFA concentration in the medium. White, gray, and black bars correspond to samples that were incubated directly in the medium, within F57 and NWT bags, respectively. Error bars represent standard error of the mean. LIN represents linear effect of forage:concentrate ratios.

There were no interaction effects (P≥0.22) between method and substrate on ΔVFA expressed in relation to OM incubated or digested, as well as on CH₄ and CO₂ in relation to ΔVFA (P≥0.06) (Figure 6). The ΔVFA expressed in relation to OM incubated was substrate dependent (P<0.001), wherein decreasing the forage:concentrate ratio resulted in a positive linear response (P<0.001). The CH₄/ΔVFA was affected by the method (P = 0.03). The NWT presented similar CH₄/ΔVFA to DIS (P = 0.13), which was equal to F57 (P = 0.22).

Figure 6 ΔVFA/OMi = apparent total volatile fatty acids (VFA) output per g organic matter (OM) incubated; ΔVFA/OMd = apparent total VFA output per g OM digested; CH₄/ΔVFA = methane output/unit of total VFA output; and CO₂/ΔVFA = carbon dioxide output/unit of total VFA output. White, gray, and black bars correspond to samples that were incubated directly in the medium, within F57 and NWT bags, respectively. Error bars represent standard error of the mean. LIN represents linear effect of forage:concentrate ratios.

The GP48, IVDMD, IVOMD, PF, molar proportion of propionic acid, acetic:propionic acid ratio, CH₄/OMd, and CO₂/OMd evaluated in DIS method were positively correlated (P<0.05), but molar proportion of butyric acid, ΔVFA, ΔVFA/OMi, CH₄ and CO₂ proportions, CH₄/ΔVFA, and CO₂/ΔVFA were not correlated (P>0.05) with these variables observed in either F57 and NWT (Table 1). The pH, acetic molar proportion, and ΔVFA/OMd for DIS were positively correlated (P<0.05) with these variables evaluated within F57 bags. Nevertheless, these latter variables were not correlated (P>0.05) when substrates were incubated within NWT bags, while CH₄/DMi, and CO₂/DMi were positively correlated with those evaluated by DIS.

Table 1

Item	F57	NWT
GP48	0.92***	0.90***
IVDMD	0.97***	0.98***
IVOMD	0.98***	0.99***
PF	0.92***	0.85***
pH	0.92***	0.52NS
Acetic acid	0.61*	0.34NS
Propionic acid	0.81**	0.85**
Butyric acid	0.30NS	–0.11NS
Acetic:propionic acid	0.87***	0.71*
ΔVFA	0.35NS	0.45NS
ΔVFA:OMi	0.35NS	0.45NS
ΔVFA:OMd	0.93***	–0.17NS
CH4	0.58NS	0.32NS
CO2	0.65NS	0.64NS
CH4/DMi	0.51NS	0.78*
CO2/DMi	0.45NS	0.79**
CH4/OMd	0.73*	0.92**
CO2/OMd	0.71*	0.84*
CH4/ΔVFA	0.26NS	0.20NS
CO2/ΔVFA	0.32NS	0.33NS

F57 - Ankom^® filter bag; NWT - non-woven fabric; GP48 - gas production at 48 h; IVDMD - in vitro dry matter digestibility; IVOMD - in vitro organic matter digestibility; PF - partitioning factor; ΔVFA - total volatile fatty acids; OMi - organic matter incubated; OMd - organic matter digested; DMi - dry matter incubated; NS - not significant. * P<0.05; ** P<0.01; *** P<0.001.

Our objective was to examine the dispersion method effects in an in vitro gas production assay on estimates of enteric greenhouse gas emissions (CO₂ and CH₄), digestibility, gas production kinetic variables, and VFA profile from substrates with different fermentabilities (low, average, and high; i.e., 100F, mixture, and 100C, respectively). Traditionally, substrates in an in vitro rumen incubation are directly dispersed in the medium. In this current study, different textiles were utilized to make simultaneous determination of in vitro digestibility easier by avoiding transfers of sample residues after incubation when direct dispersion in the medium is used. Different forage:concentrate ratios and different dispersion methods were compared to identify differences in CO₂ and CH₄ output among them. Effects of these factors on gas production, degradability, greenhouse gas yield, and other fermentation characteristics were studied using three forage:concentrate ratio that varied widely in fermentability at 48 h of incubation (GP: 101 to 262 mL/g DMi; IVDMD: 183 to 899 g/kg; CH₄ yield: 9.2 to 42.9 mL/g OM digested; CO₂ yield: 22.9 to 67.0 mL/g OM digested; change of total VFA/g OM digested: 34.1 to 72.6 mmol/g).

After 48 h of fermentation, the measured GP of the various evaluated substrates ranged from 108 to 226 mL/g DM, suggesting that a steer consuming 10 kg/d DM might produce between 1,080 and 2,260 L/d of gas. The CH₄ and CO₂ yield from fermentation of these substrates ranged from 6.0 to 13.7 mL/g DM and 10.6 to 24.9 mL/g DM, respectively, suggesting that for a DM intake of 10 kg/d, a steer may produce 60 to 137 L/d of CH₄ and 106 to 249 L/d of CO₂.

Incubating substrates within filter bags has been widely applied in vitro (Eun and Beauchemin, 2007; Krizsan et al., 2012; Tagliapietra et al., 2012; He et al., 2013; Ramin et al., 2013) and compared to dispersing the substrates into the medium. This approach has the advantage of being able to simultaneously determine in vitro digestibility without the need to capture residues after incubation. In addition, the filter bag system was found to be more advantageous than the conventional system, which uses crucibles, owing to its lower labor requirement and cost (Cherney, 2000), also avoiding losses of residues during the filtration process. However, Tagliapietra et al. (2012) reported lower IVDMD, while He et al. (2016) found greater IVDMD when feedstuffs were incubated in filter bags. In this sense, GP could be associated with these respective changes, as demonstrated by the similar results at this present study (Figures 1A–D).

The poor exchange of the fluid inside the bag with the medium promotes VFA accumulation, which may decrease the pH (Marinucci et al., 1992) and negatively affect the fibrolytic enzyme activities of microorganisms attached to substrate particles within the filter bags in an in vitro bottle (Krizsan et al., 2012). In fact, independent of the substrate type incubated, it resulted in an altered fermentation within bags with greater propionic acid produced and lower acetic:propionic ratio. Consequently, although overall pH and total VFA production of the medium at 48 h of incubation were not different, probably there was an inhibitory level of pH inside the bag associated with the physical constraint that could hinder the exchange of microorganisms, nutrients, buffer, and fermentation end-products through the small pores of the bags (Krizsan et al., 2012). These associated effects would explain the lower specific rate of digestion, lesser numeric gas volume of fast pool, and CH₄/ΔVFA verified for F57 and NWT bags compared with substrates dispersed in the in vitro medium in this study.

In general, substrates incubation within bags resulted in lower proportion of CO₂ and CH₄ in the fermentation gas, and yield relative to DMi or OMd at 48 h of incubation, but not relative to ΔVFA. The CO₂ and CH₄ emissions are calculated by multiplying GP by the respective gas proportion, and the division by incubated DM and digested OM only corrects for the incubated and fermentable amount of the substrate, respectively. In this sense, numerator variables such as GP (184 vs. 159 mL), CH₄ proportion (84 vs. 44 mL/L), and CO₂ proportion (118 vs. 98 mL/L at 48 h) were significantly greater when the substrates were dispersed into the medium than within bags. Nevertheless, denominator variables (DMi and OMd, which were 0.503 vs 0.505 g and 0.363 vs. 0.260 g, respectively) showed small difference, resulting in lower enteric greenhouse gas yield relative to DMi and OMd for the incubation within bags.

As expected, gas production, in vitro digestibilities, as well as enteric CO₂ and CH₄ proportion and production (mL/g DMi) increased linearly with the decreased proportion of F:C ratio of the substrates. On the other hand, pH, enteric CO₂ and CH₄ yield relative to OMd, and acetic:propionic ratio linearly decreased with the increase of F:C ratio. These results verified in CH₄ output agree with findings from in vitro studies comparing different substrates (Yan et al., 2000; Klevenhusen et al., 2008; Navarro-Villa et al., 2011). In general, there was an increase in CH₄ yield relative to substrate incubated as the quantity of starch in the diet increased; such an effect can be largely explained by the increasing amount of fermentable substrate with greater quality feeds. Otherwise, when the output is expressed on disappeared basis (i.e., DM or OM digested), CH₄ yield decreased as the quantity of fermentable substrate in the diet increased.

The incubation of low-fermentable substrate (100F) within filter bags showed differences for GP48 in in vitro digestibilities and PF, in which NWT presented higher values than when the substrates were dispersed in bottles and within F57. We observed in dispersed method that some feed particles of the forage were floating on the top of in vitro medium and adhered to the side of the bottles. Consequently, this substrate would not come in direct contact with microorganisms, thus, negatively affecting digestibility and GP. Similar effects were described by He et al. (2016). The NWT and F57 bags were kept under the surface of the buffered rumen fluid, but only the first bag had greater digestibility, GP48, and PF (calculated from both variables) compared with the dispersed method. In that case, greater estimates only in NWT when 100F substrate was incubated was unexpected, because GP and digestibilities were not changed or lower for NWT compared with DIS when the other substrates were incubated.

The incubation of mixture and 100C (average- to high-fermentable substrate) into filter bags consistently decreased GP, IVDMD, and IVOMD as compared with direct dispersion in the medium, although NWT had greater values than F57 (Figures 1A–D). The small pore size of the bags could constrain the entry of some ruminal protozoa, resulting in reduced starch degradability (Mendoza et al., 2014). This lower feed digestibility may arise from the inability of microbes to readily gain access to substrates within the bags, lowering digestion (Krizsan et al., 2012). According to Valente et al. (2011), superficial structures of F57 and NWT are similar and both present irregular geometrical arrangements, being formed by deposition of synthetic fiber without weaving. From this physical consideration, it could be inferred that their porosities could affect mainly the inflow of microorganisms and outflow of fermentation products. Very low porosities can compromise the withdrawal of fermentation gases from inside the bags, which can restrict microbial access to the substrate (Udén and Van Soest, 1984). In fact, independent of substrate type, CO₂ and CH₄ yield relative to OMd (Figures 3 C and F) or ΔVFA at 48 h (Figures 6 C and D) were lower when feeds were incubated within bags than dispersed in the medium. However, NWT presented greater GP48 and digestibilities than F57, showing that there could be differences between fermentation within these different textiles.

Methane yield relative to OM digested (CH₄/OMd) is the preferred unit for expressing in vitro CH₄ output (Navarro-Villa et al., 2011; Yáñez-Ruiz et al., 2016). In this sense, we may also consider CO₂/OMd the better unit to express CO₂ output, as it is related with the amount of substrate degraded. The linear decrease in CO₂ and CH₄ yield, according to improvement in substrate fermentability, appeared to be associated with greater acetate:propionate ratio, although the extent of fermentation measured by ΔVFA/OMd was not affected. The CH₄ yield relative to OMi was greater when the substrates with average or high fermentation (mixture and 100C, respectively) were incubated directly dispersed in the medium compared with when they were in filter bags. Ramin et al. (2013) reported that the proportion of CH₄ (mL/L) was greater for substrates incubated within filter bags compared with feeds directly dispersed in the medium. On the other hand, in this study we observed a general lower CH₄ and CO₂ proportion to total gas at 24 and 48 h of incubation and positive correlation (r = 0.71 to 0.92) between enteric CH₄ and CO₂ yield relative OMd from substrates incubated within filter bags versus directly dispersed in the in vitro medium (Table 1). Ramin et al. (2013) attributed the greater CH₄ proportion to the difference in the extent of fermentation and fermentation pattern compared with the dispersion on the in vitro medium reflecting a biased measurement. Based on the weak relationship between enteric CH₄ (mL/g of DM) from substrates incubated in F57 bags versus directly dispersed in the medium, these authors concluded that it is not a relevant filter bag for ranking feedstuffs.

Several factors, including source and collection time of inoculum, inoculum size and preparation, apparatus design, incubation length, headspace gas pressure, and medium composition including buffer components, trace elements, nitrogen concentrations, and reducing agents (Cone et al., 1996; Rymer et al., 2005), have been evaluated for their effect on in vitro gas production. One of the main differences between this present study and that conducted by Ramin et al. (2013) could be the headspace gas composition, which was CO₂ in that study and N₂ in our experiment. In addition, we include the evaluation of NWT bags that has been used with success in some in vitro ruminal fermentation systems to evaluate digestibilities (Silva et al., 2017; Camacho et al., 2019). A recent study reported that greater CO₂ concentration in the headspace would result in a greater dissolved CO₂ concentration in the media (Patra and Yu, 2013). Consequently, there is a greater CH₄ production corresponding to the CO₂ headspace. This might be attributed to the immediate and increased availability of CO₂ in the culture media as acceptor and H₂ (or formate) as electron donor for the methanogenesis pathway, associated with the increase in growth and activity of methanogen microorganisms. Furthermore, although CH₄ production was greater for CO₂ headspace than for N₂ headspace, total GP was lower for CO₂ than for N₂. These results could partially explain the differences found in this present study compared with Ramin et al. (2013). However, more studies are necessary to confirm this hypothesis, because we did not evaluate gas composition in headspace.

As CH₄ output generally decreases with the increasing amount of fermentable substrate found in higher-quality feeds (Klevenhusen et al., 2008; Navarro-Villa et al., 2011), we expected it when evaluating the three different substrates (i.e., 100F > mixture > 100C). Considering the enteric greenhouse yields relative to OMd for DIS method (24 and 48 h of incubation), they were greater for 100F than for mixture and 100C substrates, while these last two were similar. The enteric greenhouse yields relative to OMd evaluated in filter bags were positively correlated with those estimated at DIS method. However, when substrates were incubated within bags, the enteric greenhouse yields for the different substrates were similar at 24 h, but the results observed for DIS at 48 h of incubation was equal only to F57 (i.e., 100F > mixture = 100C substrate). However, CO₂ and CH₄ yield were, on average, 27 and 54% lower when substrates were incubated in F57 bags than in DIS. In this sense, we can infer that filter bags utilization in an in vitro system to rank CO₂ and CH₄ yield from feeds with different fermentabilities may be biased.

Considering that the feeds directly dispersed in the medium is the standard method of in vitro evaluation, we compared the relationships obtained with DIS with those obtained with substrates incubated within filter bags. In this way, considering the correlation coefficients, incubation into NWT bags showed more altered in vitro fermentation characteristics than that incubated within F57 bags. This can be justified by the absence of correlation for pH, acetic molar proportion, and ΔVFA/OMd between DIS and NWT methods (Table 1). The VFA profile and pH are important regarding the availability of H₂ for greenhouse production (Argyle and Baldwin, 1988; Bannink et al., 2006; Lana et al., 1998; Murphy et al., 1982). The greater propionic acid and lower acetic molar proportions are associated with lower H₂ production and thus lower CH₄ output (Demeyer and Van Nevel, 1975). However, absence of correlation between NWT and DIS methods for pH, acetic molar proportion, and ΔVFA/OMd suggests that the quantity of VFA produced from evaluated substrates with these methods, and thus the extent of fermentation, had no effect on CO₂ and CH₄ yield. This is supported by the positive correlation found in CO₂ and CH₄ yield (g/DMi and g/OMd)_, while ΔVFA/OMd was not correlated between DIS and NWT.