Bruin Re,scwrch BuMin,
Vol. 15, pp. 41 l-415,
1985. r Ankho
0361.9230185 $3.00 + .OO
in the U.S.A
Qualitative Analysis of Feeding Behaviour Through Dietary Selection of Nutrients DAVID Research
tS(4) 41 I-415, 1985.~Studies
of brain mechanisms
tlirrury srktion ofnurrirnrs. BRAIN RES BULL and feeding behaviour have often neglected interactions between total food (energy) intake and qualitative aspects of the dietary intake. Generally, experimental animals are presented with a single diet of fixed nutritional composition. Thus, if as a result of a given treatment, an animal has an increased or decreased appetite for a specific nutrient, then this could manifest itself as an increase or decrease in intake from the sole diet offered. As selection of food is a characteristic behaviour of all animals, and their ability to monitor intake of specific nutrients is well known, then, giving experimental animals a choice of dietary constituents could result in a wider understanding of central mechanisms governing food intake. Exploiting the ability of rats to select dietary protein and carbohydrate has suggested that brain 5hydroxytryptamine (5-HT) is involved in the regulation of protein/carbohydrate intake. Evidence from human studies suggests that appetite disturbances which occur in obese and mood-disturbed individuals may be linked to an impaired functioning of the brain 5-HT system. ASHLEY,
D. V. M. @difutitv
V. M. ASHLEY
controlling food intake
RESEARCH into mechanisms governing appetite is one means of investigating feeding disorders in man. It is of major importance to the understanding of obesity, one of the most serious of the nutritional disorders affecting affluent countries, and equally important to the understanding of compulsive eating syndromes, anorexia nervosa and cachexia in patients with tumours. None of the aforementioned disorders are a single entity, and their study can be aided by the development of suitable animal models which share with humans characteristic features of the disease. In classical models, experimental animals are often presented with a single diet of fixed nutritional composition. These studies have invariably concentrated on the voluntary adjustments of food (energy) intake which take place after imposition of various nutritional or environmental constraints or after different metabolic, physiologic or neural manipulations. The result has been a plethora of hypotheses which attempt to explain how food (energy) intake may be controlled. Each theory has had elements which have been relevant and applicable to a given situation. However, none explains all the observed facts. Why, for example, do some people get fat on quite small intakes of food, whilst others apparently with no greater physical activity eat a great deal more without gaining weight? The answer could perhaps lie either in qualitative differences in their food intake, or in the way energy substrates are utilized, or in individual capacities to adapt to differences in the total food intake or to some elements of it. Before attempting to answer these questions, several others must be posed. Are there mechanisms governing qualitative selection of nutritionally adequate diets? How do
voluntary imbalances arise, and what are their metabolic consequences? Can these explain any of the food intake related diseases? Until now, the role of individual nutrients in the mechanisms governing food intake has received little attention. In this paper, some of the factors which may be important when examining nutrient selection are discussed. Data are presented which show how dietary selection of nutrients is being exploited in research on the mechanisms underlying one abnormal feature of feeding in some obese humans: the craving for simple carbohydrate-rich foods. FACTORS
Food choice is the result of a synthesis of information derived from metabolic signals, and from sensory properties of the food. Omnivores first learn about the properties of various foods before using the information to make appropriate decisions on what to eat [15,20]. First choices are often based on palatability (e.g., preferences for sweet over bitter foods). These choices are often metabolically safe but provide no assurance that nutritionally adequate choices are made. Some internal mechanisms help, detecting negative and positive elements of the diet, the result of which will eventually cause rejection of foods having negative metabolic consequences, and acceptance of those producing positive effects . These features of food selection behaviour were recently demonstrated in an experiment aimed at discovering how rats develop an appetite for protein [ 151. Adult rats, offered a choice of 0 and 40% casein diets, taste
FIG. I. Automated feeding apparatus for studying dietary selection. Each cage is equipped with two pellet delivery systems, arranged so that one press on the appropriate lever delivers either a 75 mg pellet of a 6C% casein diet or a similar sized pellet of a protein-free (or/; casein) diet into an adjacent food cup. The lever is then inactivated until the rat triggered a microswitch by pushing open the door of the food cup. The timing and identitv of each lever press and food cup entry are recorded using a HP 9845 desk top computer 1141.
both of the diets on the first day that the choice is offered to them. Thereafter, during several days, most of the rats eat almost all of their food from the casein-free, carbohydraterich diet, perhaps because they prefer its taste or texture. Those that eat inadequate amounts of protein lose weight and reduce total food intake. Soon after, they abruptly change their food choice patterns and begin to eat from the previously rejected high protein food sources. The reason for the abrupt change in food choice is not clear. But, because young rats select adequate protein intakes rapidly (within 4 days), whilst adult rats take a much longer time (-- 10 days) before accepting the high protein food source suggests that body nitrogen depletion and the metabolic need for protein and/or amino acid are playing a role. Other properties of the food, such as texture are also important in determining choice. When offered casein-free and casein-rich powdered diets, some rats fail to eat protein and die  perhaps in consequence of their inability to learn about the metabolic properties of the foods from which they are choosing. However, if diet texture is improved by granulation or taste enhanced by replacing some of the starch with sugar, or by replacing casein with a highly palatable protein such as meat protein, then rats choose adequate protein intake more quickly and survive and grow well.
The young animal has a high metabolic demand for both energy and amino acids and, on a unit body weight basis, eats more than the mature animal. When extreme dietary manipulations are avoided, both young and mature animals usually select more protein than needed to maintain adequate growth. Older rats do not select a lower proportion of energy as protein than younger rats [ 1, 14, 151precluding a precisely metabolic linked control over intake. Metabolic needs during pregnancy and lactation become
so demanding that selection patterns change. Increases in total food intake are accompanied by voluntary increases in selected fat and protein but carbohydrate intake is unaltered [ 191. These changes in food intake may be efficient strategies adopted by the animal to satisfy its increasing needs. Increasing fat consumption, for example, is an energetically efficient way of increasing the energy stored in adipose cells, in preparation for the high, metabolic energy demands during pregnancy and lactation. Increasing protein intake contributes effectively to meeting the amino acid and nitrogen needs of the developing fetus, while conserving maternal stores. Specific increases in calcium and phosphorous by pregnant and lactating rats can also be a protection against compromising maternal stores [ 191.
A clear example that nutritional deficiencies alter dietary food choices was provided by Richter [ 191 who showed that when selecting from a choice of foods in a ‘cafeteria’ paradigm, thiamine deficient rats increase fat intake and decrease carbohydrate intake. Increased fat intake has the advantage to the rat of sparing metabolic utilization of thiamine, thus slowing the development of deficiency symptoms. Similarly, we have shown that both the excess and deficiency of protein in the dietary choice influences food selection . When offered a choice of a high fat and a high carbohydrate diet, both of which contain 32% of the dietary energy as protein, adult rats select on average 60% of their daily intake as fat and 10% as carbohydrate. But when protein provides only 10% of the dietary energy, a quantity insufficient to meet protein requirements, the rats initially prefer the high fat diet and then within 2-3 days switch to consume 60% of their intake from the high carbohydrate diet and 10% from the high fat diet . From the above examples, it is clear that appetite is de-
5-HT AND DIETARY
A. Food intake
El. % protein-enerRy
mm 120 nfn.
‘P-E selected TRP/LNAA
Time (min.) I 1 120 240 . . . . . . Maa
FIG. 2. Effect ofglucose (Glu), tryptophan (Trp), a mixture of amino acids excluding tryptophan (Maa) or NaCI on food intake and percentage protein-energy (95 P-E) selected by rats. The rats were allowed to achieve stable selection patterns from 0 and 60% casein diets for 10 days and then fasted for 22 hr prior to receiving the solution by gavage; 105 minutes later they were again allowed to choose from the 0% and 60% casein diets .
FIG. 3. Plasma tryptophan to neutral amino acid ratios of obese and obese diabetic subjects before and 120 minutes after glucose tolerance test (GTT). The control groups were normal weight of average age 20 (group 1) or 52 years (group II). Obese subjects had either normal GTT (group III), impaired GTT, normal insulin (group IV) hyperinsulinemia (group V) or hypoinsulinemia (group VI) . Significant differences indicated ?re for each time period, expressed relative to young controls @<0.05).
pendent on the supply or the metabolic demand for nutrients, and the consequences of their ingestion. In each case one could speculate as to the brain changes that may be involved and the signals that may ultimately determine the dietary choice. The chemistry of the brain is influenced by food
non-utilizable nutrients in the metabolic pool, thus taxing the mechanisms available for their elimination and storage. Conclusions arising from such experiments are therefore limited, and will not give a clear picture of the treatment’s real effects.
eaten and, consequently, precursors which result
deficiencies of neurotransmitter from imbalances in nutrient intake
may have repercussions on the synthesis and release of neurotransmitters involved in feeding behaviour. Except for the possible involvement of brain 5hydroxytryptamine (S-HT) in the dietary selection of protein and carbohydrate, these aspects of food choice have not been investigated and will provide a fruitful area for future research. However, it should be noted that changes in brain neurotransmitter metabolism induced by drugs (or other treatments, e.g., stress) may produce changes in the overall metabolic profile so that nutritional balance becomes impaired. As maintenance of nutritional equilibrium is one of the first priorities of animal life, the voluntary adjustments in food choice that are then made are often favorable to the restoration of that equilibrium. Under some conditions, voluntary shifts in food choice may not occur which, in the long-term, may have severe consequences and endanger life. Such is the situation when an animal fails to make associations between its profile of nutrient selection and their respective metabolic consequences, these two events being too far removed in time. In the experimental situation where a single diet is fed, the options available to the rat are either to decrease or to increase food intake. Restoration of nutritional homeostasis can only occur with difficulty. A decrease in the animal’s total food intake in response to a given treatment, limits the supply of the specific nutrient(s) against which that behavioural strategy may be directed, but has the disadvantage of limiting the supply of other nutrients which may be necessary for re-establishment of the overall nutritional balance. Likewise an increase in intake creates an overall excess of
DIETARY SELECTION OF PROTEIN AND CARBOHYDRATE Methodologicd
In the paradigm most often employed in studies on protein/carbohydrate selection, the rat is given simultaneous access to a choice of two nutritionally well-defined diets of equal energy density, in which the only difference is in the protein and carbohydrate content. One diet is usually deficient in protein (O-IO%) and the other contains an excess of protein (4&60%). Apart from protein level, the remainder of the nutrients in the diets are adequate to meet the rats’ nutritional requirements. The paradigm has the advantage that the rat may select independently for energy and protein. Its disadvantage lies in the fact that in selecting for protein, the rats are also obliged to select against carbohydrate. This is because protein and carbohydrate contribute the same number of calories to the energy densities of the diets, and the inclusion of one is made at the expense of the other. Thus, in situations where the energy densities of the diets in the choice are maintained constant, a decrease in protein consumption is always associated with an increased carbohydrate consumption. Conversely, if the protein or carbohydrate levels in the dietary choices are made constant (see ) then the energy density of the diets must necessarily be altered and the rats selection for carbohydrate or protein will also be accompanied by a change in energy intake. The interaction of one nutrient with another, for example, fat with protein, has been studied by maintaining constant the nitrogen to caloric level in the diets whilst substitutions are made in dietary fat content . It is sufficient here to
caution against isocaloric substitution of fat for protein or carbohydrate as this may result in diets of completely different nutrient densities. Because consumption for energy dominates other nutrient intake mechanisms , energy and other nutritional needs of the animal may well be satisfied by the more nutrient dense diet alone, particularly if it is the more palatable of the two diets in the choice situation. A three choice situation may be used in which selection is from either fat-rich, carbohydrate-rich or protein-rich food sources. Provided that the diets can be made isocaloric or that their nutrient/energy ratios are kept constant, this method provides a good technique for studying interactions between macronutrients. More complex choice situations have also been described [ 131. For example, rats, when selecting from casein, dextrin, lard, yeast and a salt mixture, choose each dietary constituent and achieve good growth rates. By mixing salt with the yeast, most rats fail to grow and select low casein and yeast intakes. Changing the concentrations of minerals in diets has been shown to alter dietary protein selection by rats [ 161. Use of semisynthetic diets of defined nutritional composition is usually preferable to commercial food preparations, in order to be sure of the available nutrient(s) selection. Interpretation of selection procedures are further complicated by the hyperphagia which variety of diet may produce . Consequently, the best experimental design for dietary selection experiments is one that is relatively simple and in which selection for a specific nutrient can occur independently of energy intake and of adequate intakes of other essential nutrients. As suggested above, the texture of the diets should be consistent; granulated diets being preferable to powdered diets. Before experimental interventions are made, care must also be taken to ensure that individually learnt and stable choice patterns have been well established [ 151. Automated devices for monitoring nutrient selection have advantages over manual measurements in that they are less disruptive and permit monitoring of feeding behaviour patterns. Thus, parameters such as meal frequency, meal-size and meal intervals may be recorded [Ill. An example of such a device is illustrated in Fig. 1. 5-Hydroxytryptamine Sdection
Regulation of ProteinlCarbohydrute
Rats given dietary choices select constant intakes of protein and carbohydrate as a proportion of total food intake both on a daily and long-term basis [5, 15, 171. Whilst various mechanisms have been shown to affect the amounts of protein and carbohydrate selected and which are not related to the central nervous system [3,15] it has become evident that central nervous system control of nutrient specific appetites do exist. Since the description of an inverse relationship between protein intake and plasma tryptophan (TRP) to large neutral amino acid (LNAA) ratio in self-selecting rats  a number of experiments have confirmed that when synthesis or release of the neurotransmitter 5-HT is grossly disturbed, then changes in the proportions of the dietary intake selected as protein and carbohydrate ensue. Consequently, drugs that deplete brain 5-HT, such as para-chlorophenylalanine (PCPA) or 5,7_dihydroxytryptamine (5,7-DHT) and lesions of serotoninergic cell bodies in the midbrain Rapht nuclei cause decreases in protein intake and concomitant increases in carbohydrate intake . Similar findings have been made
in free-feeding and starved rats given the anorectic drug fenfluramine [3, 4, 231. Though the time course of feeding responses differ in free-feeding and fasted rats (perhaps a consequence of starvation which itself alters the metabolism of brain 5-HT and energy status), the data are consistent with an effect of enhanced 5-HT neurotransmission on protein/carbohydrate intake. Other experimental situations which are known to increase synthesis of brain 5-HT such as giving TRP or glucose by gavage to hungry rats , induce preferences for protein, whilst lowering brain 5-HT by giving TRP-free mixtures of neutral amino acids produces preferences for carbohydrate (see Fig. 2). The possible significance of these findings to human food selection was suggested by Wurtman and his colleagues [22,24], upon noting that some obese people exhibit abnormal cravings for carbohydrate-rich foods. Indeed, normal humans would not be expected to have demonstrable involvement of brain 5-HT in their food choices. As in rats, one should only expect clear demonstration of the neurotransmitter’s involvement when neurotransmission is grossly altered. AS obesity is often associated with abnormalities in the metabolism of insulin, and therefore brain TRP uptake, we carried out a study in which plasma amino acid profiles of four groups of overnight fasted obese and obese-diabetic persons were examined, before and 120 minutes after a glucose tolerance test (Fig. 3). All of the obese subjects were more than 20% above their ideal body weight and had been obese for 8-30 years. They were classified according to the duration of their obesity, their glucose tolerance and plasma insulin response. One group (group III), obese for 8.6 years, had normal plasma glucose and insulin responses to oral glucose tests. Another group (group IV) obese for 17 years, had impaired glucose tolerance but normal plasma insulin levels. In the other two groups (groups V and VI) obesity was associated with overt diabetes. the cause of which was either insulin resistance (hyperinsulinemia) or an impaired secretion of insulin (hypoinsulinemia). The subjects of these latter two groups had been obese for a mean of 28 and 32 years, respectively. Control subjects consisted of two subgroups of normal weight individuals. One was of average age 20 years (group I), the other (group II) 52 years, age-matched to the older obese diabetic group (group IV). Fasting plasma TRP/LNAA ratios are shown in Fig. 3. In obese non-diabetics (group III), plasma TRP/LNAA ratios were 18% lower than in young controls (group I) and similar to values of older normal weight individuals (group II). When obesity was associated with impaired glucose tolerance or insulin secretion (groups III-VI) plasma TRP/LNAA ratios were further reduced. Extrapolation from animal data [3,4] suggests that this decrement may be enough to reduce brain 5-HT. An increased preference for carbohydrate-rich foods which may characterize the feeding behaviour of these latter groups could therefore be due to a deficient synthesis of brain 5-HT. It is possible that foods ‘craved’ by these persons, be not only palliative [ lg] but the result of a metabolic drive to normalise brain 5-HT deficits, which by these means may be impossible. Giving glucose to the obese patients of our study does not normalise plasma TRP/LNAA ratios (Fig. 3). Recent data from Wurtman’s laboratory showing that obese persons with low fasting plasma TRP/LNAA ratios, when fed high carbohydrate lunches did not normahse plasma TRPiLNAA ratios, during 6 hr following the meal even when TRP in doses of up to 1 g were given (R. J. Wurtman. personal communication). Thus, these per-
5-HT AND DIETARY
sons may be caught in a cycle of events, whereby eating of the foods which eventually make them more obese, is in attempt to normalise a fundamentally incorrigible neurotransmitter deficit. There is circumstantial evidence for a link between defective S-HT neurotransmission and carbohydrate cravings in obesity. Thus TRP, when given to obese subjects, has been reported to reduce carbohydrate cravings [22,24]. In patients given TRP therapy for depression, it is possible that .5-HT receptor hypersensitivity may ensue and induce cravings for protein-rich foods (von Leber-Good, personal communication). The anorectic drug fenfluramine reportediy reduces the number of intermeal carbohydrate snacks taken by compulsive carbohydrate-craving obese subjects 1241. Amytriptyline and other tricyclic antidepressants have been reported to promote weight gain in humans [8, 12, 181 which
may be due not only to a voracious appetite for sweet, carbohydrate-rich foods  but also to alterations in basal metabolic rate and energy substrate utilization induced by the drugs [lo]. CONCLUSION
Allowing a rat to choose its dietary intake of nutrients seems to offer advantages in the study of relationships between nutrition, brain neurotransmitter metabolism, and behaviour. Indeed, dietary selection has provided a hypothesis for how the choice of protein and carbohydrate may be regulated by brain 5-HT neurotransmission. Studies in humans have indicated that some obese may be “calmed” by a craving for carbohydrate-rich foods. Factors contributing to the development of this abnormal food practice (and to the disturbance in 5-HT neurotransmission) are as yet unknown.
REFERENCES I. Anderson, G. H. Control of protein and energy intake: role of plasma amino acids and brain neurotransmitters. Ccln J Physiol Phurmucol 2.
3. 4. 5.
13. L&t, J. Seff-selection of dietary components. In: Hun&o& of Ph,vsidogy. Vd I. Alimcntarv Canul, edited by C. F. Code.
DC: Amer. Physiol. Sot., 1967. pp. 367-386. P. D. and L. Arimanana. Circadian rhythms of food intake and protein selection in young and old rats. Annu RPV
Arimanana, L., D. V. Ashley, D. Furniss and P. D. Leathwood. Protein/carbohydrate selection in rats following administration
of tryptophan, glucose or a mixture of amino acids. In: Yr+rc~s.s in Tr~p~~~p~un and ~~r~~f~Jnin Research, edited by H. B. Schlossberger, W. Kochen, B. Linzen and H. Steinhart. Berlin: de Gruyter, 1984, pp. 549-552. Ashley, D. V. Factors affecting the selection of protein and carbohydrate from a dietary choice. Nutr Ret, 5: 555-571, 1985. Ashley, D. V. Dietary control of brain neurotransmitter synthesis: implications in the etiology of obesity. Inr J Vitamin NW Res Suppl 29 (in press). Ashley, D. V. and G. H. Anderson. Food intake regulation in the weanling rat: effects of the most limiting essential amino acids of gluten, casein and zein on the self selection of protein and energy. .I Nltrr 105: 1405-1411, 1975. Ashley, D. V. and G. H. Anderson. Correlation between the plasma tryptophan to neutral amino acid ratio and protein intake in the self-selecting weanfing rat. J N14rr 105: 141%1421, 1975. Ashley, D. V., D. V. Coscina and G. H. Anderson. Selective decrease in protein intake following brain serotonin depletion.
Chr[~~~)philrrna~(~l 1: 255-258, 1984. 15, Leathwood, P. D. and D. V. Ashley. Strategies in protein choice by weanling and adult rats. Apdctire 4: 97-112, 1983. 16. Lenrohon. C. E.. D. V. Ashiev. T. L. Woodeer and G. H.
Liff, Sci 24: 973-984,
8. Berken, G., D. Weinstein and W. Stern. Weight gain. A side effect of tricyclic antidepressants. JAffkt Dis 7: 133-138, 1984. 9. Bise, A., P. D. Leathwood and D. V. Ashley. Influence of taste and texture on protein selection in the rat. In/ J Vitamin Nutr Rf.s 53: 232, i983. 10. Fernstrom, M. H., L. H. Epstein, D. Spiker and D. Kupfer. Resting metabolic rate is reduced in patients treated with antidepressants. Bid Psyc~hirrfr~. in press. Il. Kissileff, H. R. Free feeding in normal and “recovered lateraf” rats monitored by a pellet-detecting eatometer. Physiol Bc>hrrv 5: 163-173.
12. Kupfer, D. J.. P. A. Cable and D. Rubenstein. Change in weight during treatment for depression. P.s~~~~~~f~rn.Md 41: 535-543, 1979.
Anderson. Effect’of mineral m&ure in diet on”protein intake regulation in the weanling rat. J Nurr 109: 827-831, 1979. 17. Musten, B., D. Peace and G. H. Anderson. Food intake regulation in the weanling rat: self-selection of protein and energy. J Nutr 104: X3-572. 1974. 18. Paykel, K. S., P. S. Mueller and P. D. de la Vergne. Amitriptyline, weight gain and carbohydrate craving: a side effect. Br J P.yvchiatry 123: 501-507, 1973. 19. Richter. C. P. Total self-regulatory functions in animals and human beings. Hur\w Lec,f Ser 38: 63-103, 1943. 20. Rozin, P. The selechon of foods by rats. humans and other animals. Adv Studs Rehuv 6: 21-76, 1976. 21. Scaflani. A. Appe&e and hunger in experimental obesity syndromes. In: Hungv: Basic ~~~~h~t~~srn.s and Clinicml Imp&u.I tions, edited by D. Novin, W. Wyrwicka and G. Bray. New
York: Raven Press, 1976, pp. 281-295. 22. Wurtman, J. J., R. J. Wurtman, J. H. Growden, P. Henry, A. Libscoule and S. Zeisel. Carbohydrate craving in obese people: suppression by treatments affecting serotoninergic neurotransmission. Int J &ting Dis 1: 2-11, 1981. 23. Wu~man, J. J. and R. J. Wu~man. Fenfluramine and tluoxetine spares protein consumption while suppressing caloric intake by rats. Scirnce 185: 183-184, 1977. 24. Wurtman, J. J. and R. J. Wurtman. Impaired control of appetite for carbohydrates in some patients with eating disorders: Treatment with pharmacological agents. In: The Psychohiulogy ofAnore.uio Narwso, edited by K. M. Pirke and 13. Ploog. Berlin: Springer Verlag, 1984, pp. 12-21.