### INTRODUCTION

*Clitoria ternatea*L. and by many common names, such as Cunhã, Clitoria, Butterfly pea, Blue pea, Conchitas, and Kordofan pea, belongs to the

*Fabaceae*family,

*Faboidae*subfamily,

*Phaseoleae*tribe, and

*Clitoriinae*subtribe. The origin of Cunhã is obscure (Hall, 1985; Gomez and Kalamani, 2003; Cook et al., 2005). Some authors attribute the origin of this legume to tropical America (Upadhyaya and Pachauri, 1983), but it is more likely that its origin is the Ternate Island in the Molluca archipelago, Indonesia (Gupta et al., 2010). In Brazil, the beginning of Cunhã cultivation is also obscure, but this legume may be considered a well adapted perennial, vigorous, twining, climbing, and summer growing tropical plant with pinnate leaves, bearing 5 to 7 elliptical and 3 to 5 cm long leaflets. The flowers are solitary or paired and the fruits are linear flat and sparsely pubescent pods containing 8 to 10 dark seeds at its maturity (Staples, 1992; Gomez and Kalamani, 2003; Cook et al., 2005). Cunhã has been cultivated as pure stands to be grazed for short periods (protein bank), fed as a fresh fodder, and harvested for haymaking. It can also be sown with other forage plants forming both cultivated and natural permanent pastures (Staples, 1992; Araújo Filho et al., 1996; Avalos et al., 2004).

### MATERIAL AND METHODS

*C. ternatea*L.) was sown in early spring (October) of 2008 in a 900 m

^{2}area containing Xanthic and Eutrophic Oxisols with 10% declivity located in Northern Rio de Janeiro State (21°42′33″ S and 41°20′23″ W; 12 m of altitude), Brazil. The area is located in a region where an Aw climate (Köppen standards) predominates with an annual rainfall of 800 mm. The seeds were manually scarified over a concrete-made sidewalk for approximately one minute prior to being inoculated with Rhizobium of the cowpea group, and 4 to 6 seeds were sown manually at every 0.2 m along 0.8±0.3 m spaced lines.

_{2}O, and P

_{2}O

_{5}equivalent to 1,500, 100, and 50 kg/(ha·yr), respectively, in addition to a single dose of goat manure (approx. 20 ton/(ha·yr)). A uniformity harvest was performed one month before the end of the winter (June 21st to September 23rd) of 2010 during which the minimum temperature and minimum relative humidity had risen (Figure 1a and 1b). The harvesting ages (treatments) were defined according to a pilot trial conducted to gather an idea of the plant developmental stages. Therefore, four treatments were randomly assigned to the experimental units after the day of the uniformity harvest as follows: five plots were harvested after 35 d (early vegetative growth), four plots were harvested after 50 d (flowering), four plots were harvested after 70 d (early seed pod), and five plots were harvested after 90 d (ripe seed pod). The area was irrigated according to the rainfall occurrence. Since the uniformity harvest, the rainfall accumulated 3 mm for 30 d, 43 mm for 60 d, and 269 mm for the entire 90 d period of the field trial. Therefore, the irrigation was performed once to twice a week in the first 60 d; however, the irrigation was stopped during the last 30 d of the experimental period because of the rise in the rainfall and in the relative humidity (Figure 1).

*in vitro*gas production.

_{2}SO

_{4}and approximately 1.5 g of a 56:1 mixture of Na

_{2}SO

_{4}and Cu

_{2}SO

_{4}.5H

_{2}O in 100 mL tubes using aluminum digestion blocks according to the guidelines outlined in method 984.13 and 2,001.11, including N recovery assays with certified NH

_{4}H

_{2}PO

_{4}, and lysine-HCl (AOAC, 1998; Thiex et al., 2002). The insoluble fiber content (aNDFom) was assayed with sodium sulfite and two additions of a standardized solution of heat-stable amylase, and with ash excluded according to method 2002.04 (Mertens, 2002). The non-fibrous carbohydrates were estimated with the following equation:

_{2}SO

_{4}→ KMnO

_{4}→ ash (Van Soest, 1994, p. 147).

*in vitro*incubations were performed in a 39°C water bath using 100 mL serum amber bottles sealed with butyl rubber stoppers and aluminum crimp seals. Individual samples of air-dried leaves, stems, and pods (approx. 0.5 g; nearest to 0.1 mg) were transferred into the flasks. The forage samples were incubated with 40 mL of a reduced culture medium with 10 mL of rumen inoculum as previously described by Goering and Van Soest (1970). The culture medium, reducing solution, and inoculum were prepared as a single batch (Hall and Mertens, 2008). The inoculum was obtained from a six-year-old healthy Holstein-Zebu steer that weighed 550 kg. The steer was maintained in a pasture of palisade grass (

*Urochloa brizantha*(Hochst. ex A. Rich) R. Webster) during the winter (dry period), and the steer was supplemented

*ad libitum*with chopped 1.5-year-old sugar cane (

*Saccharum*spp.) and 1 kg/d of a concentrate containing 290 g/kg of soybean meal, 680 g/kg of ground corn, and 30 g/kg of a commercial mineral salt. The liquid and fibrous mat of the rumen contents were abundant and presented characteristic appearance, color, and odor. The rumen fluid and fibrous mat were collected separately to completely fill respective thermal bottles (2,000 mL each). Approximately 250 g of the fibrous mat was then blended for 60 s with 500 mL of the rumen fluid under continuous CO

_{2}gassing, and the mixture was then filtered through four layers of cheesecloth. The filtered inoculum was added to the reduced culture medium in a 4:1 ratio, and the mixture was maintained at 39°C with CO

_{2}gassing until the mixture was transferred to the flasks.

*V*

*(Eqs. (1) to (4)) is the cumulative gas production over time (t, h);*

_{t}*V*

*(Eqs. (1) to (2)) is the asymptotic gas volume reached for a single pool substrate;*

_{f}*k*(h

^{−1}) is the fractional rate constant of cumulative gas production inferable as the digestion rate of a single pool substrate (Eqs. (1) to (2)); and

*L*(Eq. (2)) is the discrete lag time (h). Eq. (3) is a dual-pool model designed to estimate the asymptotic gas production of fast (

*V*

_{f}_{1}) and slow (

*V*

_{f}_{2}) digesting substrates (pools) with their respective

*k*

*and*

_{1}*k*

*degradation rates, which are both expressed as h*

_{2}^{−1}. In Eq. (3), the fast digesting pool is fermented as a first-order process without lag, and the second pool follows a logistic pattern with a lag time (

*L*; h). Eq. (4) was designed to fit sigmoid-shaped patterns in which fast and slow digesting pools yield asymptotic gas volumes (

*V*

_{f}_{1}and

*V*

_{f}_{2}) at

*k*

_{1}and

*k*

_{2}rates (h

^{−1}) after a common lag time (L; h) for both pools. The term

*e*is the base of the natural logarithms (Eqs. (2) to (4)), and ɛ is the random error term (Eqs. (1) to (4)). The mean digestion time (MDT, h) for each equation was calculated as follows:

*MDT*= 1/

*k*for Eq. (1);

*MDT*= 1/

*k*+

*L*for Eq. (2); and

*MDT*= 1/

*k*

_{1}+1/

*k*

_{2}+L for Eq. (3) and for Eq. (4).

*Y*

*is the observation measured in the*

_{ij}*j-*th plot at the

*i-*th harvesting age (α

_{i}) after the uniformity harvest. The fixed effects in Eq. (5) are the mean (μ) and α

_{i}, and the random effect is the usual error term (e

_{ij}) that estimates the error among experimental plots. Eq. (5) was fitted using the PROC MIXED procedure of SAS (version 9) with maximum likelihood as the estimation method. The repeated command was used with plots as subjects. The variance-covariance matrix was modeled as variance components, and as the unrestricted variance-covariance structure with treatment grouping to account for heterogeneous variances (Littell et al., 2006). The likelihood of the different variance-covariance structures was assessed by computing Akaike information criteria as suggested by Vieira et al. (2012). Null hypotheses regarding treatments, and their linear, quadratic, and cubic effects were rejected when p<0.05. Tendencies regarding treatment effects were considered when 0.05<p<0.10.

*ŷ*

*± (*

_{x}*U*

*− L*

_{r}_{r})/2; where

*ŷ*

*is the predicted dependent variable for a given harvesting age,*

_{x}*x*; and

*U*

*and*

_{r}*L*

*are the upper and lower limits, respectively, of the 95% CI. For absent treatment effects, the 95% CI for the least squares means of the dependent variable at each harvesting age (*

_{r}*ȳ*

*) was provided as follows:*

_{x}*ȳ*

*± (*

_{x}*U*

*−*

_{r}*L*

*)/2. The approximate 99% confidence interval (99% CI) was estimated for x*

_{r}_{m}as the abscissa coordinate of the maximum or minimum of the fitted quadratic polynomial according to Neter and Wasserman (1974).

### RESULTS

*k*

_{1}of leaves; and the

*k*

_{1},

*k*

_{2}, and

*MDT*of stems. The alternative model used in these exceptions was the unstructured variance-covariance matrix with treatment grouping to account for heterogeneous variances.

*F*-ratio test was applied to demonstrate the mass accumulation of pods between the last two harvesting ages with respective p-values as shown in Table 1. The leaf:whole plant DM mass ratio (LR) presented a sigmoid decreasing pattern after visual inspection. This dimensionless proportion was transformed as

_{m}= 73±3 d after the uniformity harvest, and this maximum DMY for leaves was 1,541±238 kg/ha. Confidence intervals for the variables and parameters studied are displayed in Table 2 and 3.

_{m}= 72±3 d. Ash contents in the leaves and pods were unaffected by the harvesting age (Table 1).

_{m}was approx. 55±3 d. The

*k*

_{1}and

*k*

_{2}of the leaves and

*k*of the pods were unaffected by the harvesting age (Table 1). For stems, the

*k*

_{1}peaked at 0.119±0.010/h close to 68±3 d, and the

*MDT*reached a minimum at 60±3.6 h close to 59±2 d.

### DISCUSSION

*in vitro*gas production may reveal some intrinsic properties of the feedstuff studied. The lag time, fractional rates and asymptotic gas production are affected by inhibitory substances, such as lignin, tannins, and other phenolic and plant defensive compounds that naturally occur in the substrate. Tropical legumes are an example of feedstuffs rich in these inhibitory substances (Van Soest, 1994; Longland et al., 1995; Mertens, 2005). In the present study, these chemicals were not quantified in the Cunhã parts, but Juma et al. (2006) reported an average tannin content of 17.1 g/kg DM.

*Trifolium pratense*L.) and alfalfa (

*Medicago sativa*L.) with asymptotic gas productions of 21.0 and 20.2 mL/0.1 g of DM, respectively. Moreover, they also reported values of 0.64 and 0.54 as the

*V*

_{f}_{1}/(

*V*

_{f}_{1}+

*V*

_{f}_{2}) ratio for clover and alfalfa, respectively, and they reported that the fractional rates of the fast (non-fibrous) and slow (fibrous) digesting carbohydrate pools are 0.153 and 0.028/h, respectively, for clover, and 0.110 and 0.030/h, respectively, for alfalfa. The calculated

*MDT*from the kinetic values reported by Schofield and Pell (1995) were 42.8 and 43.5 h for clover and alfalfa, respectively. In the present study, the fast digesting pool at 70 d presented

*V*

_{f}_{1}/(

*V*

_{f}_{1}+

*V*

_{f}_{2}) ratios of 0.53 and 0.40 for leaves and stems, respectively, and the gas production from pods was more likely to have originated from a uniform digesting pool. At 50 d, i.e., close to the ideal harvesting age of 56 d (Araújo Filho et al., 1994), the same ratios were 0.55 and 0.45 for leaves and stems, respectively. The basic difference between the

*MDT*calculated from Schofield and Pell (1995) and the

*MDT*calculated in the present study is based on the magnitude of the faster lag times presented by Schofield and Pell (1995).

*V*

_{f}_{2}of both leaves and stems, we inferred that the extent of digestion of fibrous carbohydrates might had also been affected. The extent of digestion of the fibrous carbohydrates probably peaked between 50 to 60 d, i.e., at the same age the maximum

*V*

_{f}_{2}for leaves and stems had occurred. The lignin content affects the extent of digestion of fibrous carbohydrates (Smith et al., 1972). Lignification completely prevents digestion of xylem and tracheary tissues of legumes. Nonetheless, other factors prevent microbial digestion too. The accessibility to inner surfaces of the cell walls, the release of phenolic-carbohydrate complexes in the vicinity of the sites where adherent microorganisms are digesting the forage particles, and the release of other inhibitory secondary compounds of the forage plant that may alter the fractional rate of fiber digestion (Wilson and Mertens, 1995; Mertens, 2005). Nonetheless, in our study

*k*

_{2}was unaffected by the harvesting age on both leaves and stems of Cunhã within the 35 to 90 d range.

### IMPLICATIONS

*C. ternatea*L.), has a good potential to be cultivated under irrigation because it yields good quality fodder, especially if this legume is offered fresh and with a high content of leaves. The leaves of this legume are the least affected part by the maturation process. Cunhã yields forage with a potential nutritive value comparable to the traditionally cultivated forage legume crops (e.g., alfalfa or clover) despite the possible positive linear effect of maturity on aNDFom and lignin (sa) contents in its leaves. Here we provided evidences for defining an ideal management strategy for this forage crop: an absent harvesting age effect over the contents of nutritional entities, over the fractional rate, and over the asymptotic gas production of the fast digesting fraction of the leaves. There is also supporting evidence that the gas production from the slow digesting fraction of leaves peaks within the 50 to 70 d range for the harvesting age of Cunhã.