Screening Approaches for Methane Mitigating Potential of Tannin-containing Plants Under in Vitro Rumen Environment

The aim of the present study was to conduct univariate, bivariate and multivariate (principalcomponent analysis, PCA) approaches in the screening of tannin-containing plants from variouscollection sites for their CH4 mitigating properties. Plant samples were obtained from various collectionsites in different countries, i.e. Indonesia (n = 27 species), Mongolia (n = 14), Switzerland (n = 16) andGermany (n = 3). The plants were incubated in vitro with buffered-rumen fluid at 39oC for 24 h. Totalgas production was recorded as an indicator of feed quality and emission of CH4 was measured. Resultsshowed that, based on bivariate screening, generally, plants possessed low CH4 production had lowquality or low total gas production except Rhus typhina, i.e. 43 ml/200 mg DM. The loading plot of PCAshowed that all phenolic fractions were in the opposite direction with CH4 and total gas production.Plants clustered together in reverse direction to that of CH4 were Bergenia crassifolia root and leaf,Swietenia mahagoni, Clidemia hirta, Peltiphyllum peltatum, Acacia villosa and R. typhina. It wasconluded that, for tannin-containing plants, screenings based on univariate, bivariate and multivariateapproaches in relation to ruminal CH4 emission led to similar results.


INTRODUCTION
Forage quality is a determining factor for the productivity of ruminants. Several important properties related to forage quality are energy content, fiber, protein, anti-nutritional compounds and the digestibility of forage in the digestive tract. On the other hand, the concern on global warming due to accumulation of green-house gases has increased in the last decades, including methane (CH 4 ). Ruminants are among major contributors of atmospheric CH 4 which is generated from microbial fermentation in the rumen (Moss et al., 2000), and may be more prevalent in the tropical area. It has been estimated that ruminants produce 80 million tons of CH 4 annually which accounts for 28% of total anthropogenic CH 4 emissions (Beauchemin et al., 2008). Such emission is not only associated with environmental problem but also reflects the loss of energy from animal and, hence, may reduce the amount of energy for production purposes. Up to 12% of the gross energy from feed consumed by ruminants is lost as CH 4 (McCrabb and Hunter, 1999).
The above-mentioned factors have led to the exploration of forages that are not only good in quality but also are able to reduce CH 4 formation in the rumen. Tannin-containing forages or extracts have been reported to be able to reduce CH 4 emission from ruminants, both in in vitro and in vivo experiments (e.g., Tavendale et al., 2005;Tiemann et al., 2008;Bhatta et al., 2009). However, further research still needs to be conducted to find most promising tannincontaining plants, i.e. effective in mitigating CH 4 emission at simultaneously of high quality. Screenings of plants that possess CH 4 mitigating properties have been done using different approaches, such as using univariate approach (Soliva et al., 2008), bivariate Garcia-Gonzalez et al., 2008) and multivariate (Jayanegara et al., 2011a). However, no studies so far have attempted to address these different approaches simultaneously when screening the plants and to compare the results obtained.
Based on this background, therefore, the aim of the present study is to conduct univariate, bivariate and multivariate approaches in the screening of tannin-containing plants from various collection sites for their CH 4 mitigating properties. A main point of interest is to observe whether these different approaches will lead to a different conclusion when screening the plants. Additionally, as a second objective, this study is aimed at observing the relationships between plant phenolic fractions and CH 4 emission using an extended database of plants collected from various sites, i.e. Indonesia, Mongolia, Switzerland and Germany.

Plant Samples
Plant samples were obtained from various collection sites in different countries, i.e. Indonesia (n = 27 species), Mongolia (n = 14), Switzerland (n = 16) and Germany (n = 3). The data were based on some previous studies, i.e. Jayanegara et al. (2009a) for the plants from Mongolia and Germany, Jayanegara et al. (2011a) for the plants from Indonesia, and Jayanegara et al. (2011b) for the plants from Switzerland. Plant samples from Indonesia consisted of one grass, four herbs, nine shrubs and 13 tree species; all of them are either commonly used as ruminant feeds in rural areas or as traditional veterinary medicinal plants. The plants collected from Mongolia are used locally in the region of production as medicinal plants. Plants from Switzerland were obtained from the region of the Alps at altitudes of 800-2300 m above sea level, consisted of two grasses, eight non-leguminous herbs, three herbaceous legumes and three tree species. The majority of these plants is used for grazing of ruminants during European summer period, particularly for dairy production. The plants from Germany were collected from the Botanical Garden of the University of Hohenheim in Stuttgart. All plant samples were oven dried at 50-60 o C and ground to pass 1 mm sieve. Each plant was encoded by its country of origin, i.e. I, M, S and G for plants obtained from Indonesia, Mongolia, Switzerland and Germany, respectively, and numbered according to alphabetical order within country.

Chemical Composition and In Vitro Procedures
The plant samples were analyzed for their chemical composition. These included crude protein (CP), ether extract (EE; AOAC, 1997), neutral detergent fiber (NDF), acid detergent fiber (ADF; Van Soest et al., 1991), total phenols (TP), non-tannin phenols (NTP), total tannins (TT), condensed tannins (CT; Makkar, 2003a) and hydrolysable tannins (HT; Singh et al., 2005). The plants were incubated in vitro with bufferedrumen fluid by using the Hohenheim gas test method (Menke and Steingass, 1988), incubated at 39 o C for 24 h. In addition, standard hay and concentrate (obtained from the Institute of Animal Nutrition, University of Hohenheim, Stuttgart, Germany) were incubated, serving as control for the successful incubation. After the incubation, Screening Approaches for Methane Mitigating Plants (A. Jayanegara et al.) total gas production was recorded as an indicator of feed quality. Emission of CH 4 was measured by using an infrared methane analyzer (Goel et al., 2008) or a gas chromatograph (Soliva and Hess, 2007). These two different methods result in a similar CH 4 value of hay standard, i.e. between 160-170 ml/l gas.

Statistical Analysis
Data of all plants from various collection sites were tabulated into a database. Minimum and maximum values for each variable were recorded, and mean and standard deviation were calculated. Pearson's correlation analysis was performed to the data, i.e. the chemical composition and fermentation variables (total gas production and CH 4 emission). Screening approaches were based on univariate (CH 4 only), bivariate (total gas and CH 4 ) and multivariate (all variables investigated) using the principal component analysis (PCA). For the univariate approach, the plants considered were those which produced CH 4 ≤ average CH 4 across all plants minus 1 × standard deviation. For the bivariate approach, the screening was based on a twodimensional plot between total gas and CH 4 . For the multivariate approach, a PCA was applied to the data with Kaiser's criterion, i.e. Eigenvalue ≥ 1.0 to extract the principal components (PC) without rotation method. This procedure generated factor loading and factor score. The first two PC were plotted both for the loading and the score. The loading plot is used for describing the relationship among variables and the score plot is used for classifying the plants according to the loading. All statistical analyses were conducted using SPSS version 17.0 (2008) and the figures were generated using SigmaPlot version 11.0 (2008).

Chemical Composition
Summary of chemical composition of the plants investigated is presented in Table 1. Plant samples collected from Indonesia contained the highest CP contents (on average basis) compared to from other collection sites. Very high CP contents were found in the leaves of Carica papaya (386 g/kg DM; code I7) and Manihot esculenta (377 g/kg DM; I17) across all plants. All plants contained low EE with the average less than 35 g/kg DM from the four collection sites. Plants from Mongolia contained higher ADF than the others. Quite a diverse TP contents were found across all collection sites, except those from Switzerland. The average of TP in these plants was 31 g/kg DM with the highest was found in Hedysarum hedysaroides, i.e. 69 g/kg DM (S8). This pattern was similarly observed for TT, CT and HT. Several plants, i.e. Rhus typhina (G2), Acacia villosa (I2) and Clidemia hirta (I8) contained TT more than 200 g/kg DM. Very high CT contents were found in Vaccinium vitis idea (M14) and Swietenia mahagoni (I27) with the values of 175 and 86 g/kg DM, respectively.

Relationships between Variables
Variables CP and EE were unrelated to total gas production and CH 4 ( Table 2). Both NDF and ADF were positively correlated (both at P<0.001) with CH 4 emission, however, only ADF was negatively correlated with total gas (P<0.05). All phenolic fractions were negatively correlated with CH 4 by following the order of magnitude: TP > TT > HT > NTP > CT. The phenolics were also negatively correlated with total gas production.
In relation to screening plants based on the multivariate approach, loading and score plots generated from PCA are shown in Figure 2. The PC1 and PC2 explained 42.3% and 23.3% of the total variation, respectively. It is clear that all phenolic fractions were in the opposite direction with CH 4 and total gas production ( Figure 2a). However, less clear relationships were observed between the other chemical constituents (CP, EE, NDF and ADF) and the in vitro fermentation variables. Some plants were clustered together in reverse direction to that of CH 4 (Figure 2b). These plants were B. crassifolia root (M4), B. crassifolia leaf (M3), S. mahagoni (I27), C. hirta (I8), P. peltatum (G1), A. villosa (I2) and R. typhina (G2). Variable CH 4 and total gas production appeared to be in close direction in the loading plot of PCA.

Plant Phenolics and CH 4 Emission
Negative relationships between all phenolic fractions and CH 4 emission suggest the role of plant phenolics on mitigating the respective green-house gas under anaerobic condition in the  rumen, and this confirms some studies that had been reported previously (Tiemann et al., 2008;Bhatta et al., 2009). Phenolics, particularly tannins reduce CH 4 emission through several ways, i.e. through reduction in fiber digestion which decreases H 2 production, inhibition of the growth of methanogens (Tavendale et al., 2005), and reduction of protozoa population where part of the methanogens are associated symbiotically on the surface or inside the micro-fauna (Vogels et al., 1980). Within the tannin classes, based on the correlation data, both CT and HT contributed to lower CH 4 emission as also shown by Bhatta et al. (2009). It is hypothesized that lower CH 4 due to CT is more through reduction in fiber digestion as compared to the HT, confirmed by much stronger negative correlation of the former with the total gas production. This might be related to the fact that CT is resistant from microbial degradation in the rumen due to its aromatic structure, while HT is degradable due to the presence of carbohydrate in its structure (McSweeney et al., 2001;Makkar, 2003b). Interestingly, the NTP plays also a role in lowering CH 4 with less adverse effect on the total gas production. This is in accordance to our  I1   I2   I3  I4  I5   I6  I7   I8   I9   I10   I11   I12   I13  I14   I15   I16   I17   I18  I19   I20   I21   I22   I23   I24   I25   I26   I27   S1  S2   S3  S4  S5 S6 S7 S8 S9 S10 S11 S12 S13 S14 S15 S16  previous study that the addition of some simple phenolic acids, namely caffeic, p-coumaric, ferulic and cinnamic acids on hay diets at 5 mM concentration decreased in vitro CH 4 emission without impairing the organic matter digestibility (Jayanegara et al., 2009b).

Different Screening Approaches
Screenings based on univariate, bivariate and multivariate approaches led to similar results to certain extent; the promising plants (low CH 4 production) obtained from univariate approach were also presence when screened by bivariate and multivariate approaches. However, the  I1   I2   I3   I4   I5   I6   I7   I8   I9   I10   I11  I12   I13   I14   I15  I16   I17   I18   I19   I20   I21  I22   I23  I24  I25   I26 I27 S1 S2 S3 S4 S5 S6 S7 S8 S9 S10 S11 S12 S13 S14 S15 S16 (b) screening by univariate approach is not as selective as the others. One may obtain plants with low CH 4 production potential, but without considering the other important aspect, i.e. the forage quality. Among the promising plants obtained by the univariate screening, for instance, A. villosa (I2) and E. aquea (I11) resulted in poor quality performance as shown by the very low total gas production (≤ 10 ml/200 mg DM incubated) of both plants. The bivariate approach may overcome this problem by considering the forage quality related variable . Although almost all plants possessed low CH 4 emission had low quality, still, a good quality plant could be obtained, i.e. R. typhina (G2). What is clear from this approach is to avoid the utilization of plants that possessed high CH 4 emission and low total gas production such as Salsola laricifolia (M8), Canna indica (I6) and Hibiscus tiliaceous (I12).
The loading plot of PCA showed the relationships between variables in a twodimensional plot. Variables that are in the same direction are positively correlated whereas variables that are in the opposite direction are negatively correlated (Esbensen, 2004). This apparently provides the same message as the correlation analysis, but not only restricted to pairwise comparisons. Since variable CH 4 and total gas production were in a close direction in the loading plot of PCA, it may imply that there is little probability in obtaining many tannincontaining plants with both desired characteristics, i.e. low CH 4 production potential and high forage quality. Therefore, the screening results of this approach were largely in agreement with the univariate approach. This is probably connected to the role of tannins in mitigating CH 4 emission as discussed above but reducing the forage quality as well at the same time. The adverse effects of tannins on forage digestibility have been well-documented (Makkar, 2003b;Mueller-Harvey, 2006). However, this does not exclude the possibility of finding plants with both desired characteristics, like R. typhina (G2) in the present dataset, since the structure and the activity of tannins in each plants are quite diverse (Mueller-Harvey, 2006).

CONCLUSION
Plant phenolic contents may be used as an indicator for CH 4 mitigating properties of the plants, although they are also reversely related to the forage quality. Different approaches, i.e. univariate, bivariate and multivariate (PCA) for screening of tannin-containing plants in relation to their ruminal CH 4 emission in vitro led to similar results to certain extent. However, screening based on univariate approach has to be conducted cautiously since it does not consider the forage quality, which is also quite an important factor. The bivariate and multivariate approaches may overcome such problem by considering the forage quality related variable. Although almost all tannin-containing plants that possessed low CH 4 emission had low quality in the present study, still, a good quality plant could be obtained, i.e. R. typhina. This is possible since great variation have occurred in the structure and activity of each plant phenolics. The searching of plants with both desired characteristics is subjected to further research.