In vitro characterisation of the rumen fermentation pattern of the cell wall fraction from several fibrous sources

Aim of study: To isolate fibre effect from other factors when comparing fibrous sources, the rumen fermentation pattern of extracted cell walls was studied. Material and methods: Cell wall fractions from soybean hulls (SH), sugarbeet pulp (BP), palm kernel cake (PK), oat hulls (OH), dehy drated alfalfa meal (DA) and barley straw (BS) were incubated in four 48 h series. Main results: Cell wall extraction efficiency was ± 0.07 units over the neutral detergent fibre content, except for PK, which recovery was 0.20. Gas produced from BP and SH was higher ( p <0.05) from 6 h. PK behaved similarly to SH from 6 to 24 h but maintained constant thereafter, whereas gas volume from OH was the lowest from 24 to 48 h ( p <0.05). All substrates recorded a maximum rate of gas production at 12 h, except OH, for which fermentation was constant on time. The organic matter disappearance after 48 h incubation agreed with these results, being higher with BP and SH, whereas OH was the lowest ( p <0.05). The proportion of methane in total gas produced was higher in OH than BP at 36 and 48 h ( p <0.05). The highest total VFA concentration was recorded with BP ( p <0.05). Pro pionate proportion was enhanced from BP, BS and SH, and that of butyrate was higher with PK and OH, whereas no differences among substrates were recorded in acetate proportion. Research highlights: Fermentation of the cell wall fraction of fibrous feeds is not directly linked to its chemical composition, not even to its lignin proportion.


Introduction
There is an increasing interest for using non-forage fibre sources, often agricultural by-products, in diets for ruminant feeding, as they can contribute to adjust feed costs compared with forages (Bradford & Mullins, 2012).
These feeds are generally included as ingredients in the compound feed, and although non-forage fibre based diets can maintain or even improve rumen health and performance of dairy cattle under certain conditions (Pereira et al., 1999;Ertl et al., 2015), their use as main sources of fibre is controversial. Their particle size is smaller compared with forages, thus promoting a limited effect as potentially effective fibre, which is necessary for stimulating rumination and modulating rumen pH in high concentrate feeding conditions (Armentano & Pereira, 1997;Grant, 1997).
Because of their variable origin, the composition of non-forage fibre sources is very heterogeneous. Their fibrous fraction can include high proportions of cell wall, such as insoluble polysaccharides (cellulose or hemicelluloses), like in seed hulls or crop and harvest wastes (Hsu et al., 1987;Gasa et al., 1989;DePeeters et al., 1997), as well as of pectin and soluble fibre, like in sugarbeet pulp or citrus pulp (Bampidis & Robinson, 2006;Münnich et al., 2017). Besides, the lignin content in some fibrous feeds may be important. Composition also varies widely in their non-fibrous fraction, with different proportions of protein or even lipids among sources.
The amount, composition and structure of the fibrous fraction of feeds determines the rate and extent of their rumen fermentation (Ford & Elliot, 1989;Chesson, 1993;Miron et al., 2001), in terms of microbial access and activity (Krause et al., 2003;Wang et al., 2018). Further, lignin forms complexes with cellulose and hemicelluloses and affects fermentation of structural polysaccharides (Chesson, 1993;Jung & Deetz, 1993;Wilson, 1994). In addition, other nutrient compounds interact with the fibrous fraction, and must also be considered in order to compare their nutritive value as fibre sources. Thus, released fatty acids from lipolysis may negatively affect microbial activity and thus reduce degradation (Doreau & Ferlay, 1995;Beauchemin et al., 2009). Besides, protein proportion is inversely related to feed fermentation, partly because it is less fermentable than that of carbohydrates (Getachew et al., 1997) but also because the released ammonia binds to the CO 2 produced, underestimating fermentation if measured by gas production (Cone & Van Gelder, 1999). Thus, comparing fibrous substrates without interactions with other feed nutrients that may interfere in their response when included in ruminant diets seems necessary for giving a clear approach to their value as fibre sources. Therefore, this work aimed to study, under in vitro conditions, the rumen fermentation pattern of the cell wall fraction extracted from several forage and non-forage fibrous sources. A previous evaluation of the same non-extracted substrates has been previously published by Ortolani et al. (2020).

Substrates and inocula
The substrates chosen for this comparative study were: soybean hulls (SH), sugarbeet pulp (BP), palm kernel cake (PK), oat hulls (OH), dehydrated alfalfa meal (DA) and barley straw (BS). Chemical composition is given in Table 1. Substrates were ground in a hammer mill (Retsch Gmbh/SK1/417449, Haan, Germany) through a sieve of 1 mm. Cell wall was extracted from ground substrates following the procedure by Smith & Waldo (1969), after boiling in neutral detergent solution (100 mL/g substrate) for 75 min. Residues were extensively washed with distilled water and fixed with acetone, dried and stored until their use as incubation substrates.
Rumen fluid as inoculum was extracted from four adult ewes (54.5 ± 6.8 kg live weight) housed in the facilities of the Servicio de Apoyo a la Experimentación Animal of the Universidad de Zaragoza. Donor animals were fed on 1 kg of a mixed diet composed of (g/kg): alfalfa hay 250; barley straw 250; barley 300; maize 100; soybean meal 100, in a single daily offer at 09:00, from three weeks before the experiment. Before feeding, rumen contents (approximately 300 mL) of each animal were sampled and filtered through a cheesecloth, mixed, collected in thermos flasks and immediately transferred to the lab for incubation. Animal care and procedures for extraction of rumen inoculum were approved by the Ethics Committee for Animal Experimentation (protocol PI48/20). Care and management of animals agreed with the Spanish Policy for Animal Protection RD 53/2013 (BOE, 2013), which complies with EU Directive 2010/63 (EU, 2010) on the protection of animals used for experimental and other scientific purposes.

Experimental procedures
Four incubation runs (48 h) were carried out, in a water bath at 39ºC. Incubation procedures were according to Theodorou et al. (1994) procedures, but without microminerals and resazurin (Mould et al., 2005) In vitro rumen fermentation of cell walls from fibrous sources of bicarbonate buffer in the incubation solution was adjusted as in Amanzougarene & Fondevila (2018) to get a medium pH of 6.5. A total of 28 bottles per run were incubated, with 4 bottles per treatment plus another 4 bottles without substrate considered as blanks of inoculum. Internal pressure of two bottles per treatment was recorded at 2, 4, 6, 8, 10, 12, 16, 24, 36 and 48 h for determination of gas production, and the gas was released. Thereafter, these bottles were opened, and their pH measured (CRI-SON micropH 2001, Barcelona, Spain) to validate the incubation conditions, and the whole incubation content was filtered through nylon bags (45 µm pore size) that were dried at 60ºC for 48 h to determine organic matter disappearance (OMd). Concentration of methane in the gas produced was determined at successive time intervals along the incubation (0 to 12 h, 12 to 24h, 24 to 36 h and 36 to 48 h), by taking a single sample (5 mL) of the gas produced in each interval. Gas samples for 0 to 12 h and for 12 to 24 h were collected from the third and fourth incubated bottles per treatment, respectively, whereas one of the two bottles used for gas production was used for gas sampling in the 24 to 36 h and 36 to 48 h intervals, the same bottle for both periods. For each substrate and incubation series, methane concentration was extrapolated to average total gas for each time interval to calculate total methane production per unit of incubated substrate. Two milliliters of liquid medium from the bottles incubated for 12 h was also sampled on 0.5 mL solution of 0.5M phosphoric acid with 1 mg of 4-methyl-valeric acid as internal standard, and was stored at -20ºC for the analysis of volatile fatty acid (VFA) concentration.

Analytical procedures
Substrates were analysed for dry matter (DM), organic matter (OM), crude protein and ether extract content according to the AOAC (2005) procedures (methods ref. 934.01, 942.05, 976.05 and 2003.05, respectively). Their concentration in neutral detergent fibre was analysed as described by Mertens (2002) in an Ankom 200 Fibre Analyser (Ankom Technology, New York), using αamylase and sodium sulphite, and results being expressed both including (NDF) or excluding (NDFom) residual ashes. The acid detergent fibre, expressed exclusive of residual ashes (ADFom,ref. 973.18) and lignin determined with sulphuric acid were analysed as described by AOAC (2005) and Robertson & Van Soest (1981), respectively. Incubation residues were also analysed for t heir DM and OM content.
Internal pressure on each bottle was measured with a HD8804 manometer provided with a TP804 pressure gauge (DELTA OHM, Caselle di Selvazzano, Italy). Readings were corrected for the atmospheric pressure and converted to volume (mL) using a pre-established linear regression (n=103, R 2 =0.996) recorded in the same type of bottles and expressed per unit of incubated OM. Results are presented either as accumulated gas volumes (total gas produced after a given period of time) or as a rate of gas production (volume of gas per unit of time produced in a specific interval). Methane concentration in gas samples was measured in an Agilent 6890 apparatus (Agilent Technologies España, Madrid) equipped with a capillary column (HP-FFAP polyethylene glycol TPA, 30 m × 530 µm id), calibrated with a 10% CH 4 standard, with a flux of 2 mL/min at 250 ºC. The frozen samples of the incubation medium were thawed and centrifuged at 13,000 g for 15 minutes at 4 ºC for their analysis of VFA, that were determined by gas chromatography on the same apparatus than for methane analysis.

Calculations and statistical analyses
Results were analysed statistically by ANOVA using the Statistix 10 package (Analytical Software, 2010), considering the substrate (n=6) as factor and the incubation series (n=4) as a block. For total gas production and OMd, the experimental unit was the average of the two bottles per treatment incubated for 48 h in the same run, whereas for methane production (from 12 to 48 h) and VFA pattern at 12 h the value from a single bottle per run was considered. Treatment differences among means with p<0.05 and 0.05<p<0.10 were accepted as representing statistically significant differences and a trend to differences, respectively. When significant, differences were contrasted by the Tukey test.

Results
Cell wall recovery from SH, BP, PK, OH, DA and BS, measured as weight of residue after large scale extraction with neutral detergent, was 586, 478, 669, 774, 534 and 707 g/kg DM of original feeds, respectively. Initial incubation pH (0 h) averaged 6.36 ± 0.20 for the four incubation runs. Along the whole incubation period, pH values were maintained within a narrow range (from 6.38 to 6.66), not existing differences over 0.15 pH units among substrates at any time interval (12, 24 and 48 h) or over 0.25 units among times within each substrate.
The pattern of total accumulated gas production ( Fig.  1) showed substrate differences along the whole incubation period (p<0.01). From 6 h onwards cell walls from BP and SH showed higher volumes of gas (p<0.05), although PK behaved similarly to SH from 6 to 24 h and recorded minimum increases thereafter. Fermentation of the other three substrates did not greatly differ from PK up to 12 h incubation, and at 48 h differences between this and DA and BS again became non-significant. Gas production from OH was the lowest from 24 to 48 h (p<0.05). For a better characterisation of cell wall fermentation, the rate of gas production as volume per time unit is presented in Fig. 2. All substrates reached a maximum gas production at 12 h, except for OH, which rate was relatively constant along the whole incubation, ranging from 2.2 to 4.5 mL/g OM/h. This and the moderate magnitude of increases from the initial 2 h to the maximum for SH (from 7.4 to 11.1 mL/g OM/h) contrast with changes observed with BP, PK and DA (9.4, 6.6 and 6.5 mL/g OM increase) and mainly the 9.7-fold increase (from 1.2 to 12.0 mL/g OM/h) in BS in the same period. Rate of fermentation of BP was the highest from 6 to 12 h and that of OM was the lowest from 10 to 24 h (p<0.05), being the comparison among the other substrates similar to that commented regarding Fig. 1.
Proportion of methane in the gas produced at the different incubation time intervals (Table 2) showed differences only at the end of incubation (36 to 48 h), when methane proportion from fermentation of OH was higher than that with BP, not existing differences among the rest of substrates. When the volume of methane produced was expressed per unit of incubated substrate (Fig. 3), the production with OH was lower than BP, SH and DA from 0 to 12 h, and was the lowest from 24 to 48 h (p<0.05). Higher methane production was observed with SH, BP and PK at 24 h, but differences among PK, DA and BS became non-significant at 36 h. After 48 h of incubation, substrates ranked as follows: SH>BP, PK, DA, BS > OH (p<0.05). The OMd after 48 h incubation (Table 2) was higher with BP and SH, followed by PK and BS and then DA, whereas OH was the lowest (p<0.05).
Total VFA concentration and molar proportions of the major VFAs after 12 h incubation are presented in Table 3. The highest total VFA concentration was recorded with BP, and it was also higher with DA than OH, recording intermediate values with the rest of substrates (p<0.05). No substrate differences were recorded on acetate proportion, whereas that of propionate was higher with BP than PK, OH and DA, and with SH and BS it was higher than PK (p<0.05). Instead, butyrate proportion was higher with PK and OH than BS, and higher in SH than BP and DA. OH recorded higher valerate and branched-chain volatile fatty acids (BCVFA, sum of isobutyrate and isovalerate) proportions than SH and DA (p<0.05).

Discussion
Compared with the analysed NDF proportion of substrates (without discounting ashes) that is shown in Table  1, the proportion of cell wall recovery in the extraction process was 0.99, 1.05, 1.09, 0.99, 1.00 and 0.93, respectively. Extraction efficiency was within ± 0.09 units interval, with extreme values for BS and PK. In the former substrate, the lower recovery could be associated to a partial solubilisation of cell wall when processed at a larger scale, as it was also observed by Barrios Urdaneta et al. (2000). In the case of PK, the difference could be due to a noticeable proportion of cracked stones in this by-product that might bias the NDF proportion in the reference analysis from this substrate because of problems in sampling, aspect that should be minimised when a higher initial In vitro rumen fermentation of cell walls from fibrous sources amount of substrate was used for the large batch extraction respect to chemical determination. In order to validate the process of cell wall isolation in this experiment, it is assumed that extraction accurately reflects the entire cell wall fraction of all sources, and thus allowed for a homogeneous comparison of their fermentation, without interferences from other components of the original feed (Barrios Urdaneta et al., 2000;Zhang et al., 2007). In this regard, it is worth considering that chemical analysis of substrates showed a crude fat content of 88 g/kg DM in PK respect to a range from 3 to 18 g/kg in the other substrates, and a crude protein content of 179, 157 and 133 g/kg DM in SH, PK and DA vs. a range from 48 to 85 g/kg in BP, BS and OH (Table 1).
In other way, some cell wall components such as pectins and to some extent non-lignified xylans are solubilised by neutral detergent (Van Soest, 1994;Jung & Allen, 1995) and thus were not recovered when processed for cell wall extraction. Therefore, fermentation potential of non-processed BP, which contains from 13 to 20% pectins (Miron et al., 2001;FEDNA, 2019) might be underestimated in substrate comparison. In fact, cell wall fermentation measured as gas production after 48 h incubation was 0.86 of that of the whole ingredient of the same batch reported by Ortolani et al. (2020). Other substrates such as SH and DA may also contain noticeable proportions of pectins (around 8 % in both SH and DA; FEDNA, 2012), and thus a reduction in fermentation could be expected when this fraction was removed. However, this might be counterbalanced by the protein washout in the cell wall treatment, as nitrogen content is negatively correlated with gas production (González-Ronquillo et al., 1998;Cone & Van Gelder, 1999). Thus, despite the 179 and 133 g CP/kg DM of SH and DA (Table 1), fermentation of their extracted cell wall fractions resulted in gas volumes after 48 h close to those reported with the original substrate (Ortolani et al., 2020).
Cell wall polysaccharides from BP and SH were rapidly and extensively fermented, at a slower but more constant rate for the latter (Fig. 2), thus making that extent of gas production in the first 24 h interval was 0.74 vs. 0.61 of total gas with BP vs. SH, although both of them reached similar gas volume after 48 h (Fig. 1). Getachew et al. (2004) observed a 0.76 NDF digestibility after 24 h of in vitro incubation. In both cases, such pattern could be related with its high availability of cell wall polysaccharides (Miron et al., 2001;Seo et al., 2009) and low lignin content (0.03 -0.05 of total NDF). Initially, rate of gas production for PK was relatively constant from 2 to  8 h and then increased up to 12 h, but fermentation from 24 h onwards was notably diminished, and only 0.21 of total gas production was produced thereafter. It seems that availability of fermentable polysaccharides was rapidly reduced, and the masking effect of its high lignin content (0.16 of total NDF) over polysaccharides probably was then more apparent. According to results from Hindle et al. (1995), the proportion of rumen undegradable fraction in palm kernel cake may reach up to 0.37 of total NDF. Non-processed OH had a high cell wall content (Table 1), but in the range previously reported (Garleb et al., 1991;Thompson et al., 2000). The low rate and extent of cell wall fermentation of OH, even lower than that from straw, agrees with previous in situ results (Hsu et al., 1987;Thompson et al., 2000), although it may widely vary depending on the oat variety. The low fermentation of OH can be attributed to its high lignin content and the presence of tight lignin/carbohydrate complexes (Garleb et al., 1991). However, for a similar feed such as wheat bran, Miron et al. (2001) suggest that it is cutin (also recovered in the lignin analysis) rather than lignin which restricts microbial fermentation, and Jung & Allen (1995) indicate that determination of lignin as soluble in sulphuric acid tends to underestimate lignin content of grasses. According to Miron et al. (2001), cell walls from dicotyledonous substrates (SH, BP, PK and DA) are higher in cellulose and lower in hemicelluloses compared with monocotyledonous cell walls (OH and BS). In any case, proportion of cellulose and lignin does not apparently affect the rate or extent of cell wall fraction of these substrates. In fact, correlation of both OMd and 48 h gas production with cellulose, considered as the difference between ADF and lignin, or lignin/cutin contents of incubated cell wall substrates rendered low coefficients (-0.015 and -0.012 for cellulose and -0.325 and -0.433 for Lignin (sa), respectively). However, correlations with hemicelluloses, considered as the difference between NDF and ADF, were higher (-0.664 for OMd and -0.685 with 48 h gas production). Miron et al. (2001) recorded different in vitro microbial degradability of cell wall polysaccharides depending on the substrate. Although the potential effect of the lignin proportion is associated with cell wall fermentation pattern (Chesson et al., 1983;Thompson et al., 2000), this is not necessarily a direct relationship (Jung & Allen, 1995), as the type of lignin and the extent of linkages between lignin and heteroxylan side-chains, which difficult hemicelluloses digestion, depend on the nature of the substrate (Jung & Deetz, 1993;Miron et al., 2001).
Methane production can be an index of fermentation efficiency in ruminants, since up to 12% of energy intake is lost as methane (Johnson & Johnson, 1995). Compared to forages, fermentation of some non-forage fibre sources may alter rumen microbiota towards a more amylolytic population, thus leading to a higher propionate production. Propionate may act as hydrogen sink reducing the utilisation of hydrogen in methanogenesis (Wang et al., 2018). Further, the smaller particle size of these feeds promotes a faster passage rate, which leads to a lower methane production (Okine et al., 1989;Beauchemin et al., 2008). This effect is further enhanced as highly digestible fibre sources stay in the rumen for a shorter time, and this may also restrict the time available for fermentation, leading to a lower methane production than less digestible fibre sources. Pardo et al. (2016) estimated a reduction of methane emission from fermentation in dairy goats when increasing proportion of agroindustrial by-products in diet. In general, total methane production ( Fig. 3) agreed with gas production pattern (Fig. 1); however, among the highly fermentable sources, from 24 h onwards methane volume from BP was lower than SH, in response to its numerically lower methane proportion in total gas (p>0.05). The lack of a specific methane pattern, different to that of total gas production, is expectable considering that no differences were recorded in acetate proportion, and substrates differences in that of butyrate were manifested in lower values for BP and DA, that otherwise rendered a higher VFA production at 12 h. Therefore, qualitatively, fermentation of cell wall fractions from fibrous feeds does not allow for substantial differences in methane production, apart to those expected from the quantitative extent of fermentation, that are manifested in total gas and total VFA production. However, substantial differences were observed in propionate proportion, which was higher in BP than in PK, OH and DA. It is worth considering that propionate proportion is inversely related with methane (Moss et al., 2000). Having into account that, stoichiometrically, propionate contribution to the volume of gas produced is lower than that from acetate and butyrate (Beuvink & Spoelstra, 1992;Getachew et al., 1997), certain level of underestimation of cell wall fermentation from sugarbeet pulp can be assumed if measured from the total gas production, as it is reflected by its higher total VFA production. However, it was not the case with OMd, which did not differ between BP and SH. Reasons explaining differences in OM that did not match to those in gas production or VFA concentration are not apparent but respond to previously observed non-extracted substrates comparison (Ortolani et al., 2020).
In summary, fermentation of the cell wall fraction of these fibrous feeds is not directly linked to its chemical composition, at least to their cellulose or lignin proportions, although a correlation with their hemicellulose content has been observed. Cell wall of sugarbeet pulp is highly and rapidly fermentable, producing a high proportion of propionate and rendering a low proportion of methane, and that from soybean hulls behaves similarly. The fermentation rate of palm kernel cake was also high during the first 24 h, so a potential contribution can be assumed if included as ingredient in concentrate compound feeds, considering the low rumen retention time of high concentrate diets. Despite structural and chemical differences of cell wall fractions from dehydrated alfalfa and barley straw, their fermentation was very similar in rate and extent.