| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
LETTER TO THE EDITOR |
Nutrition Laboratory, Department of Clinical Veterinary Medicine, University of Cambridge, 307 Huntingdon Road, Cambridge CB3 0JQ
Kebreab et al. (2002) described a dynamic model of N metabolism in the lactating dairy cow and claimed it as a first step toward a mechanistic approach of nutrient modeling. Although this is a highly aggregated model with the restricted aim of predicting fecal, urinary, and milk N output of the dairy cow, it has a fundamental flaw, several other mistakes, and numerous errors of description. Thus, the model is not an appropriate first-step mechanistic model, and the errors severely limit its appropriateness for even its limited objectives.
Since aggregated parameters of rumen and tissue metabolism were estimated by iteration of the model from data of only diet N input, fecal, urine, and milk N output, the model is empirical, not mechanistic. Although estimates of parameters of a four-pool model were made from independent metabolic studies, these parameters were allowed to change during model simulation with the sole object of providing the best fit of the input and output N data. Consequently, the fitted parameters have no biological meaning. In a mechanistic model of animal metabolism, different dietary inputs of nutrients alter pool sizes (e.g., rumen ammonia, microbial N, blood urea) and consequently flux rates, but do not alter the basic parameters of fractional utilization, affinity constants, or maximum velocity, except when these are in turn modified by some other feedback mechanism (e.g., change in rumen pH, alteration in hormone level). For example, Kebreab et al. (2002) make an initial estimate from the diet composition that 0.43 of diet N was rumen undegradable and 0.57 was degradable. In the final model, new values fitted by optimization are 0.33 and 0.65, respectively. These values are significantly different from the original estimates based on direct experimental determination of protein degradability of the feed types.
As an empirical model, predictions are constrained by the range of conditions used in the derivation of the equations. No information is given on the energy value of the diets. The authors state that "most of the diets ... were isoenergetic." Indeed, dry matter intake (DMI) varied little (SD = 0.86 kg/d, CV = 5.1%) and metabolizable (ME) and fermentable (FME) energy intakes would be similarly restricted. In contrast, the N intake varied much more (CV = 16.7%). In ruminants, as in other animals, energy supply is the prime driving force of metabolism, and protein supply is subsidiary; a N deficiency restricts energy utilization to the point where voluntary intake is reduced and an excess causes additional nonproductive use of energy through the processes of N excretion. This interdependence of energy and protein clearly requires that even a model of N metabolism alone must include energy as a parameter. The model does use an equation [4.3] including FME to limit microbial protein production, but has no energy parameter to drive or limit milk production. The latter is solely driven by the size of the amino acid pool and the efficiency of amino acid utilization. It is no wonder that the model overpredicts milk yields at high diet N inputs. The authors demonstrated that varying the energy concentration of the diet from 7 to 11 MJ FME/kg DM as a sole factor "had major effects on the amount, and more importantly in terms of pollution, the route of N excretion." Nevertheless, the model uses a fixed value of 9 MJ FME/kg DM to constrain microbial protein synthesis and does not automatically vary or take into account this factor from inputs to the model. Consequently, the empirical equations relate solely to a diet with 9 MJ FME/kg DM. Another consequence of the small range of DMI is the lack of any consideration of dry matter or ME intake on rumen turnover rates. The importance of rumen turnover rates is fully accepted in official schemes of nutrient requirements (ARC, 1993; NRC, 2000). The model of Kebreab et al. (2002) has little or no relevance outside the limited range of 15 to 18 kg DMI.
A further indication of the underlying empirical nature is given by the estimates of YMiDi and NEn, which are finalized at the preset minima of 0.05 in each case. If not constrained, the model presumably would have fitted even more nonsensical values. The yield of microbial N from diet N, YMiDi (units are a fraction or g/g, not g N as stated in Table 2 of Kebreab et al., 2002), suggests that 5% of the microbial N is derived directly from dietary N and does not pass through the rumen ammonia pool. Studies of 15N in the rumen have clearly shown that a large part of microbial N does not pass through the rumen ammonia pool, but is derived directly by uptake of peptides or amino acids that must be mainly derived from the dietary N. In sheep fed lucerne chaff, only 0.43 of microbial N was derived from ammonia, leaving 0.57 of microbial N derived from dietary protein (Nolan, 1975). In cattle fed silage diets, the proportion of microbial N from rumen ammonia was 0.55, leaving 0.45 to come from dietary protein. The proportion from dietary protein was enhanced to 0.61 when slowly degraded fishmeal was added to the silage diet (Dawson et al., 1988).
Endogenous fecal N excretion, NEn, has units in g/g and not, as defined in Table 2 of Kebreab et al. (2002), as g N/d. It is defined operationally as a proportion of the nondietary fecal N, but is also an extra proportion of the microbial N that does not enter the body amino acid pool (see below for a further criticism of this estimate of indigestible microbial N). The model makes no explicit allowance for endogenous urinary N loss. The endogenous contribution to fecal N is determined primarily by the amount of energy fermented in the hindgut. This results in fixation of endogenous protein secretions and of urea diffused from the blood into microbial N. Total endogenous N loss in urine and feces determined with intra-gastrically infused nutrients have been estimated at 350 mg N/BW0.75 (AFRC 1993). For a 600-kg cow, this equates to 42.4 g N/d lost in urine and feces, with the partitioning between urine and feces dependent on the energy fermented in the hindgut. In sheep given N-free diets, fecal N accounts for 0.64 (SD = 0.05) of the fecal plus urinary N excreted (Ørskov, 1982). Applying this proportion to the expected total endogenous loss for a 600-kg cow gives an endogenous fecal loss of 27.1 g N/d or 18.5% of the total fecal N loss. If the dietary N is 65% degraded in the rumen and the undegraded amount is 33% with an intestinal true digestibility of between 50 and 90%, then the mean dietary fecal N can be calculated as being between 14.4 and 72.2 g N/d. Subtracting these values from the mean fecal output gives the microbial plus endogenous N. The endogenous fecal N can then be expected to account for between 20 and 36% of the nondietary fecal N, not the minimum constrained 5%.
The mechanistic components initially selected are inadequate. The model assumes that all dietary proteins, regardless of source, are 65% degraded in the rumen with only 33% passing to the duodenum for further digestion. The authors acknowledge that most mechanistic models that estimate microbial protein supply require a detailed knowledge of feed degradation characteristics, but consider these to be too detailed for inclusion in a simple whole-animal model. However, the authors demonstrated that varying the one factor of protein degradability "had a major effect on urine N output." Descriptions of feed protein degradation characteristics are readily available for most feed types (e.g., AFRC 1993; NRC, 2000), so that on-farm estimation of the partitioning of N to milk, urine, or feces is possible from knowledge of the feed components fed or consumed. The value of such models is to facilitate the planning of cost-effective diets to minimize environmental pollution. Even a simple model should take this important factor into account.
Fecal N output is estimated from the sum of three fractions: 1) rumen undegradable and intestinally indigestible dietary N, 2) microbial N that has not contributed to the amino acid pool and 3) endogenous N. Microbial N that does not contribute to the amino acid pool includes 25% nonprotein N (NPN) as well as 11% indigestible true protein. The authors clearly and erroneously define microbial fecal N as 36% of the total microbial N. The NPN is mainly nucleic acids, which are absorbed from the small intestine (purine digestibility 0.83), metabolized, and excreted in the urine. Given that the model estimated the mean microbial N output as 222.1 g N/d, it follows that 31.3 g of nucleic acid N are absorbed. All the pyrimidine N and 0.2 of the purine N are metabolized to ammonia or urea (14.23 g N/d) and may be excreted in the urine or recycled. The remaining 0.8 of absorbed purines is excreted in the urine as allantoin and uric acid (17.1 g N/d). From studies of urinary excretion of 15N urea infused into the blood, 0.65 of blood urea is excreted in the urine of lactating cows (Al-Dehneh et al., 1997). This value contrasts with kUrUn of 0.29 derived in the model. Consequently, microbial nucleic acid N can be expected to contribute 26.36 g N/d to the urine. Thus, fecal N is overestimated by at least 26.4 g N/d (18.0% of the mean recorded excretion) and urine N is underestimated by at least 26.4 (17.3% of the mean recorded excretion). In confirmation of this calculation, the measured purine derivatives in the urine of lactating cows fed grass silage supplemented with 8.1 kg of concentrates was 445 mmol, or 24.9 g N/d (Shingfield and Offer, 1998). Of this, approximately 3.3 g N/d is likely to be from breakdown of body purines, leaving 21.6 g N/d to be derived from microbial nucleic acids. A further problem in the estimation of the split between fecal and urinary N is that the model makes no distinction between microbial N produced in the rumen that is digested in the small intestine and microbial N synthesized in the colon and cecum, which is not further digested and does not contribute to the amino acid pool, but does form a major part of the fecal N. Much of the N utilized to support microbial fermentation of low-N carbohydrate residues in the hindgut comes from diffusion of urea from the blood with consequent repartitioning of N from urine to feces.
The derivation of the parameters for the estimation of microbial N from the urea and ammonia pool is also a cause for concern. In the model description, kUrMi is given an initial value of 30.6 and is defined as a value for the maximal quantity of N assimilated by rumen microbes per unit of FME taken from the literature. The value will not be found in either quoted reference. The units are not clear. According to table 2kij is a fractional rate constant for the i-j transaction, generally per day. Consequently, kUrMi should represent the fraction of the urea-ammonia pool converted to microbial N per unit time, generally per day. From the discussion of sensitivity analysis, it appears that kUrMi is not a rate constant, but a yield value of microbial N/MJ FME x DMI. For the mean rounded DMI of 17 kg, the assigned value of 30.6 corresponds with a yield of 1.8 g microbial N/MJ FME, a value in the middle of the range quoted as derived from Russell et al. (1992) and similar to the value quoted for AFRC (1993) of 1.76 (lactating cows at level of feeding 3x maintenance). Thus, kUrMi has units of g N/MJ FME, but apparently is also multiplied by a constant selected by the authors as representing DMI, estimated here as approximately 17 kg. The value will be inappropriate for any other DMI. The product kUrMi x E in equation [4.3] then represents the maximal microbial N from the daily intake of FME supplied by 17 kg of DM. Since E is also taken as a constant of 9 MJ/kg DM, the maximal microbial N yield from ammonia per day is a constant in the region of 1.8 x 17 x 9 = 275.4 g N/d (or using the fitted final value 31.1 x 9 = 279.9). Clearly this equation does not estimate the potential microbial N synthesis from ammonia in any other situation than a daily supply of 17 x 9 = 153 MJ ME. This potential microbial N from the ammonia pool is constrained by a Michaelis-Menten parameter that gives a rectangular hyperbola relationship between microbial N synthesis and the size of the ammonia pool. Using the minimal and maximal urine N excretions as the extremes of the range, the ammonia pool QUr can be calculated from equations [4.4] and [5.4]: NUn = UUrUn = kUrUn*QUr and the final estimate of kUrUn = 0.29. From equation [4.3], the constrained microbial N production is found to reach a maximum at the maximal QUr, corresponding with the maximal urine N excretion and maximal dietary N input. In contrast, according to AFRC (1993), 17 kg of DMI of 9 MJ FME/kg supplying 153 MJ FME would produce 269.3 g of microbial N in a dairy cow at 3x maintenance. Assuming, as in the model, that the dietary N is 0.65 effective rumen-degradable N (ERDP), then this yield could be met by 414 g of dietary N, a value close to the mean intake and only 65.9% of the maximal N intake in the data set. In other words, microbial N yield is largely constrained by the energy supply in about half of these diets and not by availability of N from the ammonia pool. Consequently, the model appears to underestimate microbial N yield because of an inappropriate constraint.
The consequence of these major errors and omissions is that the model seeks to explain N output as milk, urinary, and fecal N solely from knowledge of dietary N input. The progress of the past 30 yr, from using digestible crude protein to the now widely adopted schemes of metabolizable protein, is completely negated. Indeed, values for digestible crude protein are not used, but the model regresses to using total N. It does not require a dynamic model to come to the authors’ conclusions that urine N output can be decreased by an increase in the energy concentration of the diet by adding more starch, by reducing dietary N, or by reducing dietary protein degradability. These are self-evident from our current knowledge of ruminant metabolism (Ørskov, 1982).
Literature Cited
AFRC. 1993. Energy and Protein Requirements of Ruminants. Agricultural and Food Research Council. CAB International, Wallingford, UK.
Al-Dehneh, A., J. T. Huber, R. Wanderley, C. B. Theurer, M. Pessarakli, and D. DeYoung. 1997. Incorporation of recycled urea N into ruminal bacteria flowing to the small intestine of dairy cows fed a high-grain or high-forage diet. Anim. Feed Sci. Technol. 68:327–338.
Dawson, J. M., C. I. Bruce, P. J. Buttery, M. Gill, and D. E. Beever. 1988. Protein metabolism in the rumen of silage-fed steers: effect of fish meal supplementation. Brit. J. Nutr. 60:339–353.[Medline]
Kebreab, E., J. France, J. A. N. Mills, R. Allison, and J. Dijkstra. 2002. A dynamic model of N metabolism in the lactating cow and an assessment of impact of N excretion on the environment. J. Anim. Sci. 80:248–259.
Nolan, J. V. 1975. Quantitative models of nitrogen metabolism in sheep. Pages 416–431 in Digestion and Metabolism in the Ruminant, I. W. McDonald and A. C. I. Warner, ed. Univ. of New England, Armidale, NSW, Australia.
NRC. 2000. Nutrient Requirements of Dairy Cattle. 7th ed. Natl. Acad. Press, Washington, DC.
Ørskov, E. R. 1982. Protein Nutrition in Ruminants. Academic Press, London.
Shingfield, K. J., and N. W. Offer. 1998. Evaluation of milk allantoin excretion as an index of microbial protein supply in lactating dairy cows. Anim. Sci. 67:371–385.
Russell, J. B., J. D. O’Connor, D. G. Fox, P. J. Van Soest, and C. J. Sniffen. 1992. A net carbohydrate and protein system for evaluating diets: I. Ruminal fermentation. J. Anim. Sci. 70:3551–3561.[Abstract]
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
