Using in vitro methods to estimate metabolizable energy content of five forage legumes harvested under different defoliation systems / Einsatz von in vitro Methoden zur Schätzung der umsetzbaren Energie in fünf Futterleguminosen aus unterschiedlichen Nutzungssystemen

A total of 431 samples were derived from two field trials in Noer, Germany (16 m ASL; 8.7°C MAT; 774 mm AP) and Gumpenstein, Austria (710 m ASL; 6.8°C MAT; 1010 mm AP), and these were harvested for two years. The experimental designs at both sites were carried out as randomized block designs with three replicates each. Up to five legume species (white clover (cv. Klondyke), WC; red clover (cv. Pirat), RC; lucerne (cv. Ameristand), LC; birdsfoot trefoil (cv. Rocco), BT; kura clover (cv. Endura), KC) were grown in binary mixtures with perennial ryegrass (Lolium perenne L.). Three different defoliation systems were applied: (a) a 3-cut system (cut at every 50±5 days, first cutting after ear emergence of grass), established at both study sites; (b) a 5-cut system (cut at every 30±3 days, first cut when the first node of grass is detectable); (c) rotational grazing (RG; nine animals were allowed to graze on plots of size 1500 m² for 3–5 days at intervals of 30±3 days). The management systems (b) and (c) were only established at the study site in Germany. Fresh plants were harvested keeping a cutting height of 5 cm and the growth was separated into grass and legume fractions. Samples were oven-dried at a temperature of 58°C and ground using a Cyclotech mill (Tecator, Haverhill, MA, USA) to a particle size of 1 mm.

All available legume samples were scanned twice using near infrared spectrometry (NIRS) Systems 5000 scanning monochromator (Perstrop Analytical Inc., Silver Spring, MD, USA) and software (ISI-version) for data collection and manipulation was supplied by Infrasoft International® (ISI, Port Matilda, PA, USA). Samples with H-values exceeding 3.0 were excluded from the calibration procedure. The calibration subsets that were selected (H-value 0.6) represented the whole sample spectrum, while the validation subsets were randomly selected after ranking the spectral data according to their H distance. Calibrations were developed by regressing laboratory determined values against the NIR spectral data (Shenk and Westerhaus, 1991). The minimum F statistics for terms included in the equation was 8.0. Subsets of samples were chosen for wet chemical analysis.

N was analysed by a rapid combustion (850°C), conversion of all N products to N₂, and subsequent measurement by thermoconductivity cell (elementar-analysator Vario MAX CN, Fa. Elementar Analysensysteme, Hanau, Germany). The results were expressed as crude protein, i.e. N × 6.25.

The pepsin-cellulase method (CM) was carried out according to De Boever et al. (1988) in Germany, following the guidelines of VDLUFA Standard Methodology (VDLUFA, 1993). Briefly, the method involved a preliminary incubation for 24 h with pepsin/HCl at 40°C, followed by heating for 45 min at 80°C and a second incubation with a commercial cellulase Onozuka R-10 from Trichoderma viride (Merck, Darmstadt, Germany).

The ME content by the CM method (ME_CM) was computed by applying the estimating equation for legumes derived by Weissbach et al. (1996) on the values obtained with the in vitro method:

MECM(MJ/kg DM)=13.98−0.0147×CA−0.0137×IOM+0.00234×CP,$${\rm{ME}}_{{\rm{CM}}} \left( {{\rm{MJ/kg}}\,{\rm{DM}}} \right) = 13.98 - 0.0147 \times {\rm{CA}} - 0.0137 \times {\rm{IOM}} + 0.00234 \times {\rm{CP}},$$(1)

where CA is crude ash content (g/kg DM), IOM is enzymatically insoluble organic matter (g/kg DM) and CP is crude protein content (g/kg DM).

Four standard samples with known in vivo digestibility values were included for in vitro analyses. Single standard samples were randomly included per run in quadruplicate. Results of the standard samples were calculated as cellulase organic matter digestibility (CM-OMD; % DM), based on the calculated loss upon ashing (L; g), initial weight of the sample (W; g), dry matter content (DM; %) and ash content (CA; %) (VDLUFA 1993).

(2)CM − OMD (% DM)=100−(L×1 000 000)/(W × DM × (100−CA))$${\rm{CM - OMD}}\,(\% \,{\rm{DM}}) = 100 - ({\rm{L}} \times 1\,000\,000)/({\rm{W}}\, \times \,{\rm{DM}}\, \times \,(100 - {\rm{CA}}))$$

The two-stage method by Tilley and Terry (1963) (TT) was carried out with modifications in Austria. Rumen fluid was obtained prior to the morning feeding from two rumen-fistulated steers fed with a diet of seasonal green forage from mixed swards and supplemented with concentrates. The buffer solution was dispensed according to McDougall (1948). Prior to incubation, rumen fluid and buffer solution were mixed in the proportion 1:4 (v/v). The pepsin solution was prepared by dissolving 20 g of 1:10,000 pepsin (Sigma-Aldrich, Germany) in 1000 ml distilled water. Where required, steps were carried out under anaerobic conditions, flushing buffers and solutions with gaseous CO₂.

Dried forage samples (0.5 g) were weighed in triplicate into 100 ml Erlenmeyer flasks and 50 ml of the rumen liquor-buffer solution were added. Remaining air was then expelled with CO₂ and flasks sealed with perforated Parafilm, followed by incubation of samples and blanks at 38.5°C for 48 h in the dark. At the end of the first incubation period, pH value was adjusted to 1.5 units by using 2.2N HCl, 5 ml of pepsin solution was added and then the flasks were incubated again for 48 h at 38.5°C. At the end of the second incubation period, samples were filtered (Macherey Nagel MN 640w, Germany), dried at 104°C for 4 h and weighed before ashing at 450°C.

The TT-OMD was calculated as difference between the OM of the sample before incubation and the residual OM. Residues after incubation measured in blanks were deducted. Outliers (if the deviation of a replicate exceeded 3 % of the mean) were excluded from further calculations.

The same four standard samples with known in vivo digestibility values, similar to the CM method, were included for the in vitro TT analyses, and all standard samples were analyzed in six replicates within each run.

The TT-OMD values of the standard samples were compared within each run to their corresponding in vivo values by linear regression. A correction equation for in vitro values of the samples was generated in terms of variability due to rumen fluid quality between the runs, and to express the results of the samples as estimated in vivo OMD. ME content for the samples in the present experiment was estimated by regression analysis of the analyzed TT-OMD (g/kg DM) with benchmark values obtained from the Deutsche Landwirtschaftsgesellschaft (DLG) tables (DLG, 1997). The benchmark values published in the tables originate from in vivo trials with lactating cows. Regression analyses were performed separately for the first cut, and then for subsequent regrowths on data of grassland with high proportion of legumes, giving the following equations:

1st cut: METT(MJ/kg DM) = 0.0174 × TT-OMD − 1.2677; R2=0.99; SEM = 0.07$${\rm{1}}^{{\rm{st}}} {\rm{ cut:}}\,\;\;\;\;\;\;\;\;\;\;\;\;{\rm{ME}}_{{\rm{TT}}} \left( {{\rm{MJ/kg DM}}} \right)\;\; = \;\;0.0{\rm{174}}\;\; \times \;\;{\rm{TT - OMD}}\;\; - \;\;{\rm{1}}.{\rm{2677}};{\rm{ R^2}} = 0.{\rm{99}};{\rm{ }}\;SEM\;\; = \;\;0.0{\rm{7}}$$(3)

regrowths: METT (MJ/kg DM) = 0.0172 × TT-OMD − 1.0306; R2=0.99; SEM = 0.08$${\rm{regrowths:}}\quad \quad {\rm{ME}}_{{\rm{TT}}} \;\left( {{\rm{MJ/kg DM}}} \right)\;\; = \;\;0.0{\rm{172}}\;\; \times \;\;{\rm{TT - OMD}}\;\; - \;\;{\rm{1}}.0{\rm{3}}0{\rm{6}};{\rm{ R^2}} = 0.{\rm{99}};{\rm{ }}\;SEM{\rm{ }} = {\rm{ }}0.0{\rm{8}}$$(4)

A mixed model analysis was calculated for energy content data (ME) of each defoliation system and for the 3-cut system of both sites using PROC MIXED by considering cut, species and method as fixed factors (SAS Institute Inc., 2004). Years and site (data was available for the 3-cut system in both sites) were included in the data set, not as classificatory factors in the statistical model, but considered as replicates. Cuts (or grazing cycles) were treated as a repeated measurement assuming a symmetric covariance structure. In case of significant interactions (P < 0.05), linear contrasts were calculated using the SLICE procedure of SAS. Comparison of least squares means were performed by t-test. To avoid random significances, probabilities were adjusted by the Bonferroni–Holm procedure at P < 0.05 (Holm, 1979).

The relationship between the NIRS-estimated ME contents based on the two different in vitro methods was determined by linear regression analysis for different factor combinations (forage legume, defoliation system and harvest season). Parameters of the determined regression equations (slope, intercept) were tested to evaluate whether they differed from unity or zero, respectively.

As shown by the mixed model and regression analyses of all data, the determination of ME of forage legumes clearly differed depending on the in vitro method used, with systematic higher values predicted using the CM method.

The general limitations given within the methods and the respective equations are possible reasons for this study. Whereas OMD values obtained with TT are dependent on the variation obtained for the standard samples measured simultaneously in each run, the CM method does not need simultaneous measurement of standard samples, as the equation is based on several in vivo digestibility trials. Although a low number of samples originated from in vivo trials, the weak repeatability in TT could be due to methodological aspects of the in vivo digestibility method, or due to the variations accounted for, for example, the harvest time or botanical composition of the standard samples.

For the CM method, a regression equation was developed for ME estimation separately for legumes by Weissbach et al. (1996). This equation is based on 20 in vivo digestibility trials, including LC and RC, preserved as silage, hay and oven-dried forage, in all covering a range of 29 to 77% OMD. As found in other studies, the in vitro CM-OMD values generally agreed with in vivo values, regardless of the feed type analyzed (De Boever et al., 1988). The larger variation of values obtained from the TT procedure was reflected in the larger variation of results of the standard sample analysis (Table 2). Although the difference to the in vivo OMD were lower, the ME estimates based on TT gave systematically lower ME values when considering cuts, legume species and management types compared to the CM method in the present study.

The defoliation frequency caused deviations of the intercepts, as the 5-cut system and RG were larger than zero (Figure 2). Compared to the 3-cut system, the plants in the 5-cut system and RG were harvested at an earlier development stage and thus showed high ME values (Table 4), as confirmed in previous experiments including grazing management (Kleen et al., 2011). The differences were consistent between sites and years.

Enzyme-based predictions of in vivo digestibility and ME content can vary with forage species and harvest season (Barber et al., 1990; Givens et al., 1995). Differences between harvest seasons were observed in in vivo digestibility equations for grasses using either enzyme or rumen inoculum based methods (Jones and Theodorou, 2000). Season influence on ME content is quite often observed, but less often for forage legumes. In the present study, the lowest content of ME was observed for the CM method in the third cut for the 5-cut system and RG (Table 4). In the autumn harvest, higher ME with the CM method was estimated to be comparable to the summer harvest, whereas with the TT method only a slight increase could be observed. For animal nutrition, the first harvest is quantitatively and qualitatively more important. From the second cut onwards, similar values for the ME content among cuts are mostly assumed, irrespective of management. However, the results suggested that the TT method may also underestimate the ME content in cuts late in the season (e.g., in autumn), especially at a higher defoliation frequency (Table 4), and undergrazing (Figure 2). Although a higher ME content in autumn harvests was observed, the way to use the grassland in this season is disputable, as the higher ME content is masked by low DM yield, higher forage legume proportion and poor protein quality (Kleen et al., 2011; Krawutschke et al., 2013), which may result in higher N losses in the production system, especially under grazing.

Forage legumes may have leaves with higher digestibility than stems, compared to grasses. In this case, management systems resulting in higher leaf:stem ratio may substantially improve OMD in legumes with erect, crown-forming growing habit (Annicchiarico, 2007). For legumes like LC, RC or BT, higher leaf:stem proportions are observed for higher cutting frequency (Gierus et al., 2012). Thus, these legumes have a stronger influence on higher digestibility and ME content due to an altered leaf:stem ratio with low maturity at harvest in the 5-cut system in the present study. Ranking legumes based on the leaf:stem proportion as used in the present study (Table 3 and 4) confirms the observations by others (Nordkvist and Åman, 1986). LC samples confirmed having a higher ME content in the 5-cut system and RG system in comparison to the 3-cut system. Achieving higher leaf:stem proportion would be a management option to improve forage quality for certain forage legume species with erect, crown-forming growth habit. However, the prediction of ME of several legumes submitted to different management systems shows that the largest ME values were observed for white clover, while comparing the 3-cut system at both sites (Table 3) or among defoliation systems in Germany (Table 4). Both CM and TT methods were able to measure the higher ME content for white clover consistently in Austria and Germany.

The comparison of ME_TT and ME_CM for individual data within legume species reveals that the higher ME values were equally well predicted compared to the lower ME contents for different legume species. Although higher in ME content, white clover did not show better prediction using either TT or CM in comparison to other species. Using legume forage samples with known in vivo OMD revealed that the CM method may overestimate the ME contents of the samples, whereas the TT method may underestimate the values. However, white clover (at both sites), together with KC (in Germany only) were the species with the highest ME content, independent of season (cut number), defoliation system, or site, which was confirmed using both methods.

Compared to other forage legumes, the enzyme polyphenol oxidase is very active in RC (Eickler et al., 2011). Using substrates like caffeic acid, phaselic acid and clovamide present in the plant, the enzyme catalyses the reaction and produces quinones and in this way may cause a complexation of proteins (Jones et al., 1995). The complexation is comparable to that observed for condensed tannins, which are present in BT, and the extent of their presence is dependent on their concentration in forage and diet. One may suggest that condensed tannins or quinones cause a stronger impact on the rumen microbes when the TT method is the method of choice, resembling the determination of in vivo OMD with non-fitted ruminal microflora of animals to tannins. The effect may be ignored in the CM method due to an unsuitable pH value in the pepsin step, and seems to be comparable to the OMD estimation in animals already adapted to secondary compounds. However, these differences are not always observed in the literature. Reviewed by Aufrère and Guérin (1996), prediction of OMD by enzymatic methods is, in general, accurate. For forages containing tannins, enzymatic methods overestimate OMD compared to in vivo digestibility, because of lower digestion in vitro due to the presence of tannins. Similar effects might occur in RC containing polyphenol oxidase activity, as the produced quinones may act in some cases similarly as condensed tannins.

The effect of polyphenols, either condensed tannins in BT or quinones in RC, on an annual basis was not apparent in the present study for the ME estimate. The ME estimates among legume species from TT and CM methods were lowest for BT and comparable for LC (Table 3 and 4), whereas the ME estimates of RC were only lower than white clover. The secondary plant components, quinones and condensed tannins, may have varied in their contents during the growing season, which was supported by the observed species × cut interactions over methods in the present study, and was also observed in other studies (Eickler et al., 2011). However, the variation was small but enough to show the influence of species (RC and BT) on a lower ME estimation as average over methods. This may be related to the content of condensed tannins or quinones formed within these species.

While in vitro methods are suitable for the prediction of ME content of forages, in vivo are used to validate the predictions. Variation may arise due to limitations of in vitro analysis to include deviations of OMD due to feed intake level, diet selection of species and plant organs or animal species. In addition, in vitro methods may accumulate fermentation end products and may not consider the effects of passage rate, neither solid nor liquid phase (Dijkstra et al., 2005). However, feeding trials performed to quantify the level of inclusion of forage legumes on animal performance response showed controversial results (Steinshamn, 2010). For the present study, the large number of samples to test variations due to site, harvest date and defoliation system made the in vivo validation considerably more difficult, especially under grazing conditions. However, the present results supported the assumption that in vitro estimation methods reflect the ME content in forage legumes in the following order of decreasing relevance: legume species, defoliation system and harvest date.