Contents - Previous - Next

The metabolic basis of amino acid requirements


* Nutrition Research Unit, London School of Hygiene and Tropical Medicine, St. Pancras Hospital, 4 Pancras Way, London NW1 OPE, U.K.

1. Introduction: The nature of the problem
2. Nutrient requirement models
3. The Millward & Rivers requirement model: Qualitative aspects
4. The variable extrinsic component of the maintenance requirement
5. The anabolic drive
6. Hormonal components of the anabolic drive
7. Protein requirements: Formal statement
8. The issue of protein quality
9. Stable isotope studies
10. Practical experience of biological values of dietary protein
11. Urea salvage
12. Indispensable amino acid requirements for the anabolic drive
13. Conclusions


The Millward & Rivers model for protein requirements accounts for both functional intrinsic needs for amino acids (growth and obligatory amino acid catabolism), and for extrinsic responses to habitual protein intake in terms of oxidative losses of amino acids. These losses are part of homeostatic control, under acute and chronic regulation by dietary protein feeding, and reflect the potential toxicity of many indispensable amino acids (IAA). Because postabsorptive amino acid losses increase with increasing habitual protein intake, the diet must include sufficient protein to replace losses occurring in the previous postabsorptive state, i.e., the more you eat, the more you need for overall balance. Judgement of the value of any particular intake requires consideration of the regulatory influence of dietary amino acids on the organism: the anabolic drive. The optimal protein requirement comprises the sum of intrinsic needs for protein accretion, obligatory amino acid consumption, and sufficient amounts for an appropriate regulatory response or anabolic drive. Such a response might be an agreed rate of height growth or index of immunocompetence. While the issue of protein quality is relevant to each of these components, accretion accounts for so little in humans that no relationship should be expected between protein quality for growing animals and human needs. Transient accretion during diurnal cycling could involve some recycling of lysine and threonine, reducing their dietary needs.

The IAA requirement to cover obligatory amino acid needs cannot be predicted from first principles, but judging from pig studies is dominated by S-amino acids and threonine, with little lysine need. The IAA requirement for the anabolic drive is unknown. In any case, the emerging evidence that changes in the dietary amino acid composition can be achieved during urea salvage and bacterial amino acid synthesis in the lower gut means that the relationship between dietary need for protein and IAA and dietary protein quality becomes even more difficult to predict.

1. Introduction: The nature of the problem

Figure 1. N-balance studies in preschool children.

Discussion of the effect of amino acid composition of protein on protein-energy interactions requires both relevant data and an agreed model of the nutritional requirements for protein and energy, according to which the significance of the data can be evaluated. In fact, data of any sort are scarce, so that the potential for useful discussion is extremely limited. More importantly, this limited existing information is very difficult to interpret, in part because the criteria by which we have judged the impact of changes in energy intake on protein metabolism have been inadequate. This can be illustrated with data from two N-balance studies in preschool children (Figure 1) fed traditional diets, corn and beans in Guatemala (TORUN and VITERI, 1981), and rice in Hyderabad, India (IYENGAR et al., 1981). Digestible protein intakes were close to the 1985 requirement values (FAO/WHO/UNU, 1985) in each case. Digestible energy intakes were reduced, from levels which were slightly below the most recent UK Dietary Reference Values (UK Department of Health, 1991), to values which we would judge to be grossly inadequate. Both diets maintained a large and positive N balance (more than twice the value judged necessary for normal growth) at all levels of intake. With the rice diet, the N balance fell with reduced energy intake, but at the lowest energy intake N balance was still adequate, as judged against the level necessary for normal growth.

We can interpret these findings in two ways. One is that, in children, there is no reason to worry about the impact of any reduction in energy intake on protein utilisation, since they can maintain adequate N balance with traditional vegetarian diets supplying current protein requirements, even when energy intakes are markedly reduced. The other interpretation is that the measurement of N balance alone does not give us sufficient information to judge the adequacy of the dietary intakes, or to account for the metabolic responses occurring in these children. There does appear to be an emerging consensus favouring the latter view which, in any case, is not new (HEGSTED, 1976). For this reason, in the rest of this paper, I want to explore just how much we do understand of the metabolic basis of the requirements for protein and amino acids, and to argue for a more rational model for determining amino acid requirements.

2. Nutrient requirement models

For most nutrients, it is useful to consider requirements according to the scheme in Table 1.

In this scheme, there are two categories of requirements, functional and regulatory. Functional requirements are mainly intrinsic, representing demands of the organism. Regulatory requirements are extrinsic being responses to nutrition and other variable environmental influences. For energy, functional requirements include growth, physical activity and cellular maintenance, which involve reasonable fixed stoichiometries. Evidence for a significant regulatory component of energy expenditure (e.g., DIT) is relatively poor, albeit unresolved (NORGAN, 1990; SHETTY, 1990). Thus, acceptable models for energy requirements include estimations of only functional needs.

The situation is different for protein. Figure 2 shows the protein requirements for all ages. I have calculated the growth requirements on the basis of the unadjusted estimated average increment rates at each age (FAO/WHO/UNU, 1985). I have then included the obligatory losses based on estimates in children and in adults. These losses do not change much with age. Provision for growth and replacements of obligatory losses represent the functional needs, and it is quite clear that, at all ages, the total of these two components is never much more than 50% of the total requirement, even less in the case of children.

The key problem, then, is to account for the nature of the rest, identified as the oxidative losses in Figure 2. This was the aim of our amino acid requirement model (MILLWARD and RIVERS, 1988).

Table 1. Nutritional requirement models: functional and regulatory components which must be balanced by the diet


Functional needs (intrinsic)

Regulatory losses (extrinsic)







(small components)



oxidative losses


(large component)

NST = non-shivering thermogenesis
DIT = diet-induced thermogenesis
ONL = obligatory nitrogen loss

Figure 2. Current values of the protein requirements (FAO/WHO/UNU, 1985).

Figure 3. The Millward & Rivers requirement model (MILLWARD and RIVERS, 1988).

The protein requirement will equal the sum of needs for protein gain (Gn) and needs to balance losses (the obligatory losses (Lo), and regulatory oxidative losses which vary with intake (Lr)). Thus, R = Gn + Lo + Lr.

3. The Millward & Rivers requirement model: Qualitative aspects

One way of representing the model qualitatively is shown in Figure 3. In its simplest form, it includes needs for new tissue and maintenance. We view maintenance as a variable component, comprising both fixed functional demands and variable regulatory extrinsic components. It is defined operationally as the dietary intake needed to balance all losses, including those which occur on feeding protein.

There are three components which are functional demands or intrinsic. The first one is the fixed component relating to the net requirement, i.e., the need to deposit new protein. Conceptually this is straightforward. The other two are part of the maintenance needs. These are the intrinsic requirements for obligatory amino acid consumption in non-protein pathways, conceptually straightforward even if poorly understood, and the anabolic drive.

4. The variable extrinsic component of the maintenance requirement

Our understanding of this is based on two principles which need to be considered separately.

4.1. Indispensable amino acids as toxic metabolites
4.2. Diurnal cycling

4.1. Indispensable amino acids as toxic metabolites

The branched-chain amino acids (BCAAs), aromatic amino acids and sulphur amino acids all have tightly regulated oxidative pathways, which, in general have low KM for the rate-limiting step, and consequently exist in small tissue pools (KREBS, 1972; WATERLOW and FERN, 1981). For example, the KM for tryptophan pyrrolase is 0.15 mM, and for the branched-chain s-oxoacid dehydrogenase the KM is also at sub mM levels. As reviewed previously (MILLWARD and RIVERS, 1988), the regulatory mechanisms of these catabolic pathways include activation by feeding and an induction of their activity by the protein content of the diet. As a result of this, increases in the concentrations of the BCAAs, aromatic and sulphur amino acids on feeding are minimised. Furthermore, we know that this is a feature of these specific amino acids, and not amino acids in general. Some dispensable amino acids, such as glutamine, can be tolerated at very much larger concentrations (>20 mM; MILLWARD et al., 1982). From the perspective of these aspects of metabolic regulation, we would conclude that the organism treats many IAAs as toxic metabolites. We know that they are toxic when their concentrations rise as the result of an inability to remove them because the enzymes are lacking as in maple syrup urine disease and phenylketonuria.

Figure 4 (a). Amino acids in human skeletal muscle.

4a: Changes in IAA in human muscle at 3 hours.

Figure 4 (b). Amino acids in human skeletal muscle.

4b: Changes in IAA in human muscle at 7 hours, after a meal of 50 g of albumin. Values recalculated from BERGSTROM et al. (1990) to show the changes per kg body weight assuming that muscle water accounts for 500 mL per kg body weight.

This is not the case for ail IAAs. Lysine and threonine are different with higher KM (18 and 52 mM, respectively), larger pool sizes, (about 1 mM) and there is less evidence of the fine control of their catabolic pathways.

The difference in the handling of lysine and leucine may well be important. From this metabolic perspective we would predict that, after feeding, the branched-chain, aromatic and sulphur amino acids not incorporated into protein would be removed very rapidly, more so than lysine or threonine. Figure 4 shows data on the changes in human muscle (obtained by muscle biopsies) at 3(4a) and 7(4b) hours after a meal of 50 g of albumin (BERGSTROM, FURST and VINNARS, 1990). The increases have been calculated per kg body weight and expressed in relation to the amino acid content of the meal. For leucine and l although the intakes are the same, and although removal of these two amino acids into protein will be at the same rate (since their concentrations in protein are similar), the increase in the concentration of lysine is twice that of leucine. The same is true for threonine in comparison with valine. By seven hours, the concentration of all these IAAs had fallen below the baseline with the exception of threonine and lysine, where there was still an excess of amino acid over the baseline value.

Clearly these changes in the free pool size are a consequence of several processes which deliver them to and remove them from the tissue, and changes in oxidation with feeding are only part of those processes, but these data do confirm that the organism is not prepared to let any of the pool sizes of these IAAs expand for very long, with the exception of lysine and threonine. The implications of these differences between lysine and threonine on the one hand, and the other IAAs on the other, will be considered further below.

The concept of IAAs as toxic metabolites involves the need not only to be able to activate oxidation by feeding, but also to induce the capacity for oxidation. We know from animal experiments that the branched-chain s-oxodehydrogenase is induced by dietary protein (HARRIS et al., 1986; PATSTON et al., 1986), SO we would expect to see the human organism act in the same way.

In fact we have investigated this with studies looking at the extent to which the amount of leucine oxidation depends on the previous protein intake level. We fed diets of increasing protein from 0.35 to 2 g per kg for two weeks, and measured leucine oxidation and deposition as protein. The results are shown in Figure 5.

The increasing protein intake induced increasing leucine oxidation, as well as increasing deposition as protein. However, when we switched from the high-protein diet to a moderate intake, and measured the response after two days, the high rate of leucine oxidation was still occurring, so that all of the dietary and some of the body protein was oxidised. The previous intake had conditioned the body to expect a high-protein diet and had induced an oxidative capacity which was now too much for the lower intake. By seven days of the lower intake, the oxidative capacity had readjusted back to a lower, more appropriate level.

This demonstrates the first principle of our model, that the amino acid oxidative capacity of the organism is determined by the habitual protein intake.

Figure 5. Leucine oxidation and deposition into protein in relation to the dietary protein intake level.

Measurements involve 13C leucine infusions in normal adults fed the diets for 10-14 days except where indicated (MILLWARD), D.J., PACY, P.J.H., PRICE, G.M., QUEVADO, R.M., HALLIDAY, D.. Unpublished results).

Figure 6. Magnitude of diurnal cycling of body protein in relation to habitual protein intake.

Values calculated from 12-hour nitrogen balances corrected for changes in the body urea pool measured over a 48-hour period in adults habituated to the intakes for at least 14 days (MILLWARD, D.J., PACY, P.J.H., PRICE, G.M., QUEVADO, R.M., HALLIDAY, D.. Unpublished results).

Figure 7. Schematic representation of the diurnal changes in body protein stores in relation to a set-point regulation of the upper limit of body protein.

The nature of the regulation of the set-point is not understood, but is a function of height.

4.2. Diurnal cycling

The second principle which explains the variable maintenance requirement is the diurnal pattern of food intake exhibited by our species. As demonstrated many years ago in animals (GARLICK et al., 1973; MILLWARD et al., 1974), and man (CLUGSTON and GARLICK, 1982), during the substantial postabsorptive phase of the diurnal cycle, body protein will be lost and will need to be replaced during feeding. However, the novel feature of our model is the suggestion that the amount of these postabsorptive losses will increase as intake increases, because of the induction of the oxidative enzymes by the dietary protein (MILLWARD et al., 1991a; b). The result of this would be an increasing amplitude of diurnal cycling, with increasing protein intake. If this occurs, then the protein requirements needed to balance postabsorptive losses will increase with intake.

Evidence that this does occur is shown in Figure 6. It shows measurements of the magnitude of diurnal cycling based on 12-hour N balances, measured over a 48-hour period after two weeks on the diet, corrected for changes in the body urea pool. Measurement with 13C leucine and 2H phenylalanine confirms that these are protein gains and losses (MILLWARD et al., 1991a; b).

Figure 7 provides another perspective on the same idea. Body protein stores are regulated by a mechanism which sets the upper limit. The nature of this mechanism is currently unknown, but it appears to be some function of height, and a possible mechanism for muscle has been described (MILLWARD, 1989). Although the 'body stores full' level may be achievable over a wide range of intakes, the increasing postabsorptive losses induced by increasing intake necessitate increasing fed-state gains for balance to be achieved. If the diet is changed to a lower level, then the acute response, i.e., before adaption has occurred, will involve both excessive postabsorptive loss and a feeding response which, because of the lower intake and excessive fed-state oxidation, would result in insufficient protein deposition to replace the losses.

To summarise, maintenance protein requirements for overall balance can be defined as a dietary intake to allow sufficient daytime gain to replace the prevailing night-time losses.

5. The anabolic drive

This is the third intrinsic component of the protein requirement, and represents the need for dietary amino acids to exert an appropriate regulatory influence on the organism, a transient function exerted prior to oxidation. The magnitude of this is currently unknown, and the targets for it, height growth and immunocompetence, for example, are also somewhat speculative at the moment.

As discussed elsewhere (MILLWARD et al., 1990), the anabolic drive can be conceived in terms of exerting homeostatic and homeorhetic influences as well as exerting some functional regulation. Dietary protein is well known to regulate the growth of skeletal muscle (JEPSON, BATES and MILLWARD, 1988), and, in children, growth in height may be another indicator of homeorhetic control by protein, at least according to the evidence assembled by GOLDEN (1985). Our own evidence from rat studies (YAHYA et al., 1989; MILLWARD, 1990) clearly shows the effect of dietary protein on height growth. In adults we might assess body composition and organ function. We know that muscle function is nutritionally sensitive (LOPES et al., 1982). There is also evidence that kidney function might be adversely influenced by excessive protein intake (BRENNER, MEYER and HOSTETTER, 1982), and this might define the upper safe limit in intake. Our main point is that, without these indicators, we cannot judge the efficacy of any level of protein intake.

6. Hormonal components of the anabolic drive

As to the mechanisms involved, in the rat we have clear evidence of an interaction of dietary amino acids with the major growth-mediating hormones (MILLWARD, 1989; 1990). In Figure 8 the anabolic drive is represented as a combination of direct regulatory influences of dietary protein on tissues, together with indirect hormonal influences of insulin and growth hormone, IGF-1 and T3 (see MILLWARD, 1989). The sensitivity of insulin secretion to amino acid intake, most marked in the rat, allows it to mediate its homeostatic role acting with amino acids. However, it is the indirect influence of insulin, mediated through its involvement in the peripheral metabolism of growth hormone and thyroid hormone metabolism and action, which constitutes its homeorhetic role.

Figure 8. Hormonal components of the anabolic drive.

Selected aspects of the peripheral changes in hormone levels in response to dietary protein are shown with the central role of insulin identified (see MILLWARD, 1989; 1990).

A potential scheme for the homeostatic and homeorhetic components of the anabolic drive on muscle and bone is shown in Figure 9 (MILLWARD, 1990). In this scheme, insulin mediates homeostatic regulation by controlling the reversible deposition of protein (shown here for skeletal muscle but also including liver). However, homeorhetic regulation of growth involves events in the myofibre, satellite cells and fibroblasts of muscle, and in the proliferating chondrocytes of the growth plate of bone. This requires a sustained nutrient input in which insulin levels are elevated for long enough to increase levels of IGF-1, T3 and the other growth factors, which together mediate such responses. Clearly, the scheme is not intended to be comprehensive. IFG-1's role in muscle connective tissue growth has yet to be confirmed, but seems likely, especially since satellite cell proliferation is most sensitive to IGFs rather than insulin (DODSON, ALLEN and HOSSNER, 1985). The interactions between insulin and IGF-1 action via modulation of the IGF-1BP1 are still somewhat tentative, with the role of T3 in satellite cells and muscle connective tissue cells not yet confirmed, but based on no more than its similar known role in chondrocyte maturation (BURCH and VAN WIJK, 1987). However the important feature of the scheme is the concept that growth hormone, IGF-1 and T3 levels represent an integrated response to food intake with sufficient elasticity in their regulation that several meals need to be missed before malnutrition is sensed and growth is shut down.

Figure 9. Homeostatic and homeorhetic components of the anabolic drive on muscle and bone (see MILLWARD, 1990).

As discussed elsewhere (MILLWARD, 1990), the extent to which this scheme, developed for the rat, obtains in the human, has yet to be confirmed. The role of insulin may be less obvious, since insulin secretion is much less sensitive to dietary protein. What is important, however, is the considerable evidence for a regulatory role of dietary amino acids on the organism.

7. Protein requirements: Formal statement

From this we can construct a protein requirement statement, as in Table 2. In our original paper we said that the requirement will be equal to G (growth) + Lo (ONL) + Lr (the regulatory oxidative loss) and argued that, since the magnitude of Lr varies with the intake, we could only define an operative value for Lr and the requirement. Since balance can be achieved over a range of intakes, we need to be able to make judgements about the value to the organism of any particular level of oxidative loss. This is the point at which the importance of the anabolic drive can be recognised, since it allows definition of an optimal requirement level (Ropt), as distinct from the minimum level (Rmin) which, in practice, has been the previous objective.

Table 2. Components of the minimum and optimum protein requirement

Functional needs (intrinsic)

Regulatory losses (extrinsic)

growth (G)

ONL (Lo)

anabolic drive (Lr optimum)

oxidative losses (Lr operative)

minimum requirement:
minimum intake for N balance after adaptation
Rmin = G + Lo + Lr min

optimum requirement:
intake for N balance after adaptation and desired functional response
Ropt = G + Lo + Lr opt

8. The issue of protein quality

In the context of this model, protein quality and IAA needs can be considered in relation to the three intrinsic components of Figure 3, namely protein accretion, minimum obligatory needs and the anabolic drive.

8.1. Accretion: Both net and transient
8.2. Minimum obligatory needs: Theoretical predictions

8.1. Accretion: Both net and transient

The IAA needs for tissue accretion are uncontroversial; they correspond to the amino acid content of tissue protein gain. According to current estimates of protein requirements, however, accretion accounts for < 20% at 3 months, falling to only 6% for the 10-year-old. This means that there may be little relationship between estimates of protein quality studied in growing animals and human needs. Indeed FAO/WHO has now formally abandoned the use of rat growth assays in assessing protein quality for humans (FAO/WHO, 1991).

However, one interesting but complex problem which has not previously been considered relates to the IAA needs for accretion during diurnal cycling, i.e., the transient accretion in subjects in overall balance. In one sense, IAA needs for transient accretion are only different from needs for growth to the extent that the composition of the proteins gained and lost during diurnal cycling differ from the rest of body tissues. Since, as far as we know, proteins of the visceral tissues and skeletal muscle are involved in diurnal cycling, we would not expect any major difference in terms of average composition.

However, one reason why IAA needs for diurnal cycling may differ from growth needs is the possibility that some recycling might occur, i.e., the recycling of amino acids liberated during the catabolic postabsorptive phase into net deposition during feeding (Figure 10). For this to happen, the free pool size of the IAAs must be large enough to allow such recycling to occur. Clearly, on the basis of the above discussion, this is only possible for Lysine and threonine. Some indication of how much of these free pools might be available to supplement dietary protein, which was inadequate in lysine and threonine, was given by the results of the muscle biopsy studies of Bergstrom, already referred to in Figure 4. In these studies, the authors reported the changes in muscle lysine and threonine following a protein-free meal after which protein deposition would have been entirely dependent on the free amino acid pools. The fall in the free lysine and threonine pools in muscle were sufficient to have enabled about 250 mg protein deposition per kg body weight. Given that for an adult, consuming the current DRV of 0.75 g/kg/d, our studies indicate that about 50% of this is deposited during feeding (i.e., 380 mg), then the free pool of lysine and threonine could, in theory, contribute two thirds of their needs, if the dietary source was inadequate. If this is the case, then the organism in overall balance may be less sensitive to poor dietary quality in terms of lysine and threonine needs for transient protein deposition, than during growth with net requirements.

Figure 10. Recycling of amino acids liberated during the catabolic postabsorptive phase into net deposition during feeding.

These arguments, however, assume that net protein deposition during feeding is only limited by amino acid supply, and it is not known whether this is the case. In the context of the model implicit in Figure 7, the maximum body protein stores are limited by some 'body-full' signal which controls the upper limit of the protein stores in the lean body mass. In this model, fed state deposition is only possible if 'space' is created by the postabsorptive losses. We don't know what these factors are that could limit protein deposition once the 'space' was 'filled', but they may well be unrelated to IAA supply.

Another possibility which is novel, is that dietary IAA needs for the transient deposition during diurnal cycling may be limited by an amino acid not usually considered to be limiting on the basis of our experience in animal growth assays. In our model, the fate of each dietary amino acid is determined by the competing processes of oxidation (Lr) and gain. When deposition is growth, as in children recovering from malnutrition, the evidence is that this is very efficient, with little oxidation as judged by nitrogen-balance studies (WATERLOW and WILLS, 1960). It must be assumed that, notwithstanding any activation of oxidation on feeding, the greater activation of protein deposition removes amino acids into protein at such a rate as to maintain very low tissue concentrations and low oxidation rates. In the circumstances of transient gain, if the stimulation of protein deposition is not as powerful as in a growing organism, then activation of oxidation on feeding may become a more important regulator of utilisation. On the basis of our results (MILLWARD et al., 1991b), the efficiency of protein deposition during feeding of milk protein in subjects in near balance, i.e., the slope of the fed state gain-intake plot, is low at about 50%. This would allow for the idea that the increase in oxidation (Lr) has become a more important factor in limiting amino acid supply for protein deposition. In this case, since the activation of oxidation of leucine is particularly marked on feeding, the leucine content of dietary protein could be the limiting factor.

Clearly, the resolution of these possibilities requires more information than we have at present.

Contents - Previous - Next