Comparison of predictions of digestible supply and measurements of net portal fl uxes of essential amino acids in lactating dairy cows

Estimated digestible amino acids (AA: corrected for endogenous secretions) obtained from the NRC and CNCPS models appear to be good predictors of the AA profi le fl owing into the portal vein. However, for absolute amounts, the NRC model more closely predicts the changes measured in net portal fl uxes compared to the CNCPS. The slopes of the regression “AA net portal fl ux vs AA digested-NRC” are in agreement with AA metabolism across the gut: slope smaller that unity indicating loss for the branched-chain AA (oxidation) and for Thr (endogenous loss); slope equal to unity for Lys, indicating little loss. Slopes greater than unity for His, Met and Phe suggest either an underestimation of the digestible fl ow with the NRC model or an overestimation of the net portal fl uxes for these AA.


INTRODUCTION
Mathematical models have been designed to estimate the requirements and productive responses of lactating dairy cows (e.g., Cornell Net Carbohydrate and Protein System (CNCPS v. 5.0.34) and National Research Council (NRC, 2001).These models rely on empirical relationships between metabolizable protein (i.e.absorbed amino acids) and milk protein output, without taking into account metabolic events occurring in the portal-drained viscera (PDV) and the liver.Hence, our current research seeks for ways to improve the accuracy of whole-animal models by including mechanistic elements to take into account the events to which essential amino acids (EAA) are subjected to during their transfer from duodenum to the mammary gland.
A fi rst step in our modelling approach was to estimate net portal fl uxes of EAA using predictions of digestible EAA fl ow from two models: CNCPS and NRC.Our objective was to compare measurements of net portal fl uxes of EAA in dairy cows against predicted digestible EAA, to determine which of the two models could be used as a starting point for a mechanistic model of EAA utilization in dairy cows.

MATERIAL AND METHODS
Eight treatment means for net portal fl uxes were obtained from the studies conducted by our group (Berthiaume et al., unpublished;Lapierre et al., unpublished;Raggio et al., 2004).Feedstuffs included in the diets in these trials were individually analysed for chemical composition, including the CNCPS fractions and AA.Data from the ingredient analyses (n=33), ration formulations and average intakes from each treatment were entered into the CNCPS (version 5.0.34) and NRC (2001) software packages to obtain estimates of AA fl ows to duodenum.To allow a direct comparison with the net portal fl ux measurements, estimated digestible AA fl ows to duodenum were corrected by subtracting the endogenous secretions as specifi ed for each model.Linear regression analysis of portal fl uxes (y) on the estimated duodenal fl ows (x) served as an indicator of the degree of association between the two variables.Regression analyses were performed using the REG procedure from SAS with the NOINT option.Differences between AA profi les in portal and digestible fl ows were assessed by performing a correspondence analysis (CORRESP procedure) in SAS.

RESULTS
Digestible supply of EAA estimated by CNCPS were numerically higher (5 to 40% higher) than those obtained from the NRC model, with exception of leucine, which was only 92% of the NRC value (Table 1).The estimated digestible EAA fl ows are, on average, 103 and 118% of the measured net portal fl uxes for the NRC and CNCPS, respectively.However, the ratio of digestible fl ow to net portal fl ux for individual EAA ranged from 82 (Met) to 129% (Arg).The amounts estimated by NRC and CNCPS models were numerically lower than the net portal fl ux for Phe (87 and 91%, respectively).
The analysis of correspondence indicated that AA profi les are not different between portal measurement and NRC and CNCPS estimations (P>0.05χ 2 test, 16 df).The fi rst principal axis (54% of the dataset inertia) is a contrast between the results from NRC (higher Leu) and CNCPS models (higher Arg and His); while the second principal axis (46% of inertia) highlights a contrast between the net portal fl uxes (higher Phe and Met) and the two models for estimation of digestible AA (higher Thr).As expected, for both models the slopes of the regression "net portal fl ux vs digestible fl ows" indicate that for all EAA net portal fl ux has a positive linear association with digestible fl ow.The NRC model estimates yielded lower root mean square errors when used to explain the changes in net portal fl ux (Table 2) compared to the CNCPS.For the NRC estimates, His, Met and Phe slopes were signifi cantly (P<0.05)greater than unity.For CNCPS, the same situation arises for Phe (P=0.06).PACHECO D., LAPIERRE H.

DISCUSSION
Although the values originate from totally different approaches (arterio-venous differences for net portal fl uxes vs mathematical models for digestible fl ows), they relate very closely.
For most EAA, the value of the slope of the regression of net portal fl ux vs digestible fl ow from NRC is in agreement with expected metabolism of the EAA across the PDV (Berthiaume et al., 2001;Lobley and Lapierre, 2003).A slope less than unity would indicate loss across the PDV (e.g., oxidation for the branched chain AA, endogenous loss for Thr).A slope equal to unity would indicate very little catabolism for Lys.For His, Met and Phe, the slopes greater than unity indicate marginal increases in net portal fl ux that are greater than the digestible fl ow for these EAA, suggesting underestimation of the digestible fl ow by NRC or overestimation of the net portal fl ux measurements.
Estimated digestible fl ows from CNCPS were higher than NRC, but this model fails to follow the pattern of changes in net portal fl ux, as indicated by its higher RMSE values for all EAA.Also, the slopes for individual EAA do not refl ect the differential metabolism of EAA across the gut as well as those calculated from NRC estimations.

CONCLUSIONS
Digestible fl ows of EAA estimated with NRC better predict the measured changes in net portal fl uxes of EAA in lactating dairy cows.

Table 1 .
Amino acid (AA) amounts (g/d) and profi les (% of essential amino acids) from measured net portal fl uxes, or from estimation of digestible AA by the NRC 2001 and CNCPS 5.0 models

Table 2 .
Regression analysis using the NRC and CNCPS estimated digestible EAA as predictors of net portal fl ux 1 equations are in the form Net portal fl ux, g/d = Slope × digestible fl ow, g/d 2 root mean square error: smaller value indicates better fi t.SE: standard error of the estimate 1