I. PROLOGUE: A NEW FRAMEWORK FOR THE CONTROL OF FEEDING

Top Previous Next References

Curiosity, interest, and questions about the acquisition, preparation, and presentation of food, the cultural importance of food, appropriate diet and menu selection to promote optimal nutrition, states of hunger and satiety, regulation of body fat, body weight, and energy balance naturally arise in most of us from our daily experiences and contact with food. Among scientists with interests and training in physiology, behavior, neuroscience, metabolism, physiological psychology, and/or behavioral neuroscience, many have given at least passing thought and perhaps even a little professional attention and effort to this topic. However, a few become so captivated by this complex, multifactorial regulatory system at the interface between "internal" physiology and the "external" world that they then devote their working scientific lives to grappling with these issues. These scientists from many different disciplines continually try to tease out and identify the fundamental facts, the organs, tissue, cells and molecules, the physiological mechanisms, the levels and patterns of organization, and the decision algorithms and rules involved. Their ultimate goal is to provide a deep and complete mechanistic understanding of the regulation of feeding behavior.

The subject of this review is our perspective on the current status and, what we believe are, promising future directions of this search for a more complete understanding of the regulation of feeding behavior in laboratory rats and humans. We will not be comprehensive in the sense of enumeration of all the ideas or theoretical and experimental approaches that our former and current colleagues have productively or unproductively applied to the problem. Instead, we attempt to provide a new framework for understanding the control of feeding behavior, with special emphasis on the evolution of hunger and the initiation of feeding, including theoretical and experimental components. We will begin by providing, in section II, a historical perspective on the role of blood glucose in the control of feeding. In section IIIA, theoretical approaches that have been applied to the control of feeding and had a strong influence on experimental feeding research are summarized. We then provide a statement and overview of our current theory that has emerged from our studies of the role of blood glucose dynamics or "patterns" in the control of meal initiation in section IIIB. Then, we review in detail the experimental studies on meal initiation that have been conducted, first in rats (sect. IV) and then in humans (sect. V). Sections A and B of the rat and human data presentations provide the reader with the basic data set upon which our theory is based. The remaining sections may be too specialized for the general reader. We conclude with a discussion of the future directions of the work, limitations, and the implications for the understanding of the control of feeding behavior and the regulation of energy balance in section VI. Section VIA presents a future research agenda specific to meal initiation that may also be too detailed for the general reader. However, section VI, B and C, is intended for both specialist and the general reader and forms the summary of the review. Here a discussion of limitations of previous studies of feeding, obstacles to linking eating to physiological processes, the limitations of our "patterns as signals" concept, and the implications for our theory for the control of food intake and body energy balance are presented. Along the way, we present a mixture of philosophy, theoretical arguments, experimental design issues, as well as formal presentation and discussion of experimental results. We hope that this journey into the complexities of contemporary feeding research will prove interesting and useful to the reader.

At the outset, we want to acknowledge and explicitly state that the ideas, concepts, and implications of the experimental results discussed here have been strongly influenced by the experimental approach, integrative perspective, and brilliant insights into the regulation of feeding of Jacques Le Magnen and his students and colleagues. Professor Le Magnen, who died in May 2002, was a major intellectual force in the field of the regulation of feeding throughout his career. In addition, Donald Novin, Stephen Woods, and G. P. Smith and their students and colleagues as well as the work of many others have also had important influences on our research and thinking. Of course, our research in the area of meal initiation and the physiology of hunger descends from the seminal work of Jean Mayer and his students and colleagues. We have extended his view by adding and emphasizing a dynamic perspective that proposes patterns of blood glucose, rather than glucose utilization, as endogenous signals for meal initiation. Jean Mayer wrote the following about his glucostatic theory that is based on glucose utilization:

As for the glucostatic theory, ... its reception has illustrated to a certain extent the fate which, William James warned, awaited all new concepts: First people say it is not true; then they say it is of no general significance; and finally, they say that anyway it had been know for a long time.... At any rate, what lasting value, if any, it will achieve will be inexorably decided by the test of time.

Jean Mayer (77)

We are truly "standing on the shoulders of those that have gone before" as we take this look, first, backward and, then, forward in time at the complexity of feeding behavior and its regulation.

	II. HISTORICAL PERSPECTIVE ON THE ROLE OF BLOOD GLUCOSE IN THE CONTROL OF FEEDING

Top Previous Next References

Among the major issues in the control of food intake, the identification of the biochemical basis for hunger and meal initiation, and the signals controlling these neurobiological processes have been the subject of extensive research and debate for many decades. These studies led to the classical aminostatic, thermostatic, and lipostatic theories of food intake control as well as more recent hypotheses that are discussed in section III (6, 9, 37, 46-50, 70-72, 93, 122) . However, all of these theoretical ideas were preceded by the notion that glucose uptake and utilization played a central and metabolically privileged role in the control of hunger, satiety, and the regulation of body energy balance. Although these notions were first discussed and suggested by Carlson in a classic text published in 1916 (37), they were formalized by Jean Mayer into the classical glucostatic theory in the mid 1950s (76, 77).

The observation that the sensation of hunger was often associated with discomfort referred to the abdominal region was the basis of early attempts to explain the regulation of food intake. Cannon and Washburn (36) reported that "hunger pangs" were associated with contractions of an empty stomach. Washburn swallowed a flexible tube to which a balloon was attached at the end. Stomach contractions could be monitored because each contraction would increase the pressure in the partially inflated balloon. In general, Washburn's reports of hunger pangs coincided with measured gastric contractions. This reported relationship between hunger pangs and gastric contractions, together with the observation that injection of glucose reduced gastric contractions, led Carlson (37) to propose that food intake was regulated by feedback from the stomach.

Mayer proposed that decreased glucose utilization, which was detected by the brain at glucosensitive sites, in unspecified locations, represented the stimulus for meal initiation. The glucostatic theory of hunger postulated that reduced glucose utilization in these critical brain regions leads to perception and expression of hunger. He also argued that increased glucose utilization in these same glucosensitive sites leads to decreased hunger and the cessation of eating. Mayer proposed that decreased glucose utilization or "metabolic hypoglycemia," the point at which the peripheral arteriovenous difference in blood glucose (A-V delta glucose) becomes negligible and glucose is no longer entering "metabolizing cells," was the signal for meal initiation. He viewed his signal, metabolic hypoglycemia, as reflecting the point at which the energy substrate flux was at a minimum or turning in the direction of increasing fatty acid utilization; in other words, the point of the beginning of carbohydrate depletion. In his 1955 paper, Mayer (77) explicitly argues that the glucostatic theory would account for the short-term regulation of hunger and food intake, while he invoked a lipostatic mechanism to account for the long-term regulation of body weight and energy balance. Mayer observed that the temporal changes in blood glucose concentration were correlated and consistent with the observed time domain of changes in hunger and food intake in rats and humans. From this observation, he argued that changes in blood glucose concentration and/or arteriovenous differences in glucose concentration reflected or mirrored the postulated glucose uptake and utilization in glucosensitive brain areas and could be used as surrogates of these unobservable parameters. He also argued that the changes in body weight would be consistent and controlled by a slower lipostatic mechanism in which increases in fat content will be followed by increased fat utilization, with the resulting sparing effect on carbohydrates (77).

After formulation of the glucostatic hypothesis, there was a blossoming of research interest in the biological basis of hunger and satiety in both humans and laboratory rats. Mayer's specific hypothesis and its straightforward predictions relating behavior and metabolic processes were disseminated widely, attracted a lot of attention, and quickly became objects of a flurry of research directed at validation and testing of the hypothesis. The initial results of several studies, including studies in humans conducted in Mayer's laboratory, demonstrated that arteriovenous differences in blood glucose concentration were correlated with hunger ratings and food intake under some circumstances (125, 128). However, other research comparing the arteriovenous differences in blood glucose and hunger ratings or food intake failed to observe a correlation under other situations (2, 129). In other research, hunger ratings and food intake were measured following exogenous infusions of glucose. Again, the results of these investigations were either consistent (7, 125-127) or inconsistent (2, 55, 133) with a role for glucose in the onset of hunger. A review of this research area by one of Mayer's early collaborators, Dr. Ted VanItallie, has provided an important historical perspective (127). Mayer acknowledges the very important contribution of Dr. T. VanItallie to the broadening of his initial theory and its application to different metabolic situations including the hyperphagia of diabetes (76).

Numerous experimental studies emphasize the role of decreased glucose utilization or decreased intracellular glucose concentrations rather than the absolute level of blood glucose as the stimulus for meal initiation. The observed induction of feeding by administration of pharmacological doses of insulin (8, 45, 68, 75, 83) or of nonmetabolizable glucose analogs (73, 117, 123), the satiating effects of small glucose infusions or gastric loads (91, 92), and effects of central injections of glucose and 2-deoxyglucose (4, 99, 101, 123) all strongly suggest a role for decreased glucose uptake and utilization, possibly modulated by insulin, at a target site or sites in the control of meal initiation. However, other experimental results appeared inconsistent with the glucostatic theory. When intravenous glucose infusions (using peripheral veins) with or without insulin were administered before meals, no delay in meal initiation or reduction in meal size was observed (2, 55, 86). Furthermore, the observations that large, prolonged decreases in blood glucose were required to induce feeding following insulin administration and that the onset of feeding often occurred when the blood glucose had returned to baseline have also been used as evidence against the glucostatic theory (44, 119, 120).

By the mid 1970s, the weight of the experimental evidence and its contemporaneous interpretation had cast serious doubt on the glucostatic hypothesis. In the absence of a strong advocate (Mayer had left the field and active research by that time) and in the face of strong and vocal attacks, interest in research motivated by the glucostatic hypothesis based on glucose utilization waned. Mayer himself was quite fatalistic about the role of hypoglycemia, or any theory based on hypoglycemia, ever being consistent with the well-established hyperphagia commonly observed in diabetes:

As the possible role of hypoglycemia. ... , its study was pursued no further, especially in the view of the lack of correlation encountered between hunger and absolute levels of blood glucose and then apparently insurmountable difficulty of diabetic hyperphagia.

Jean Mayer (77)

Both Mayer and VanItallie have argued that the ability of the glucostatic theory to adequately explain the hyperphagia commonly observed in uncontrolled diabetes demonstrated the strength of the theory based on glucose utilization rather than glucose concentration (76, 77, 127). The reasoning was that an uncontrolled, insulin-deficient diabetic would experience hunger because the blood glucose, although high, was not being properly utilized and hyperphagia would follow.

Two other parallel developments in the field also contributed to the abandonment of the glucostatic hypothesis based on glucose utilization. First, the prominence and primacy of the hypothalamus in the control of feeding behavior and regulation of energy balance strongly argued as a critical component of the dual-center hypothesis published by Stellar in 1954 (121) was also being forcefully challenged by scientists advocating the importance of peripheral mechanisms involved in the control of feeding. Second, the demonstration that the gastrointestinal hormone cholecystokinin (CCK) rapidly and potently reduced meal size in rats through a peripheral mechanism (51) caused a major shift in research and conceptual focus from meal initiation and hunger to meal termination and satiety. The combination of these trends led to a paradigm shift in feeding research away from the glucostatic hypothesis and central mechanisms toward meal termination, satiety, and peripheral mechanisms.

More recently, a signal for the initiation of freely taken meals in rats with continuous access to familiar food has been identified: a brief fall and rise in blood glucose concentration before ingestion of food (Fig. 1). The identification of this signal was a direct result of a series of technological innovations leading to computer-based continuous monitoring of blood glucose in freely moving rats (14, 74, 84, 120).

View larger version (12K): [in this window] [in a new window]   Fig. 1. Composite time course of the functional coupling between transient decline in blood glucose and meal initiation. Average time course of blood glucose concentration in the prevented access condition was expressed as percent change from baseline concentrations. Data are mean values. The latest estimates of the onset and offset times for the meal initiation signal to be above threshold are indicated by downward and upward arrows, respectively. For details, see text. [From Campfield and Smith (18), with permission from Elsevier Science.]

The ability to monitor blood glucose continuously in freely behaving animals, in contrast to discrete blood sampling at fixed 10-, 15-, or 30-min intervals, led to renewed consideration of the role of blood glucose in meal initiation. Transient declines in blood glucose were first described by Louis-Sylvestre and Le Magnen (74), who showed that a fall in blood glucose was correlated with meal initiation in the rat. They observed that blood glucose concentration declined 6-8% at 5.0 ± 0.3 min before meal onset in both the dark and light phases of the light-dark cycle.

We have confirmed and extended these initial findings. We have provided experimental evidence supporting the hypothesis that spontaneous, self-resolving transient declines in blood glucose precede and signal meal initiation in free-feeding rats. This evidence was obtained using on-line, computer-based technology for continuous monitoring of blood glucose concentration in freely behaving rats. In nondeprived free-feeding rats, this signal precedes food-seeking behavior and the initiation of a meal but does not predict the size of the meal or the timing of meal termination.

	III. RESEARCH STRATEGIES AND STATEMENT OF THE THEORY

Top Previous Next References

Research on the control of feeding behavior over many decades has been as varied, multidisciplinary, and complex as the phenomena of feeding behavior and its regulation. As in any field of scientific investigation, most of the research has been focused on a variety of major and minor issues over the years. Individual scientists or groups of scientists in one or more laboratories have become interested in an issue and have experimentally addressed that issue by applying the conceptual framework and technology that was available or preferred by them. The list of questions in Table 1 has been among the major issues in the field.

Each of these questions has been further subdivided into classes of factors (e.g., sensory, gastrointestinal, metabolites, hormonal, neural), site(s) of origin or action, and antecedent or prevailing metabolic states. A set of theoretical models has been employed and has led to several different research strategies being utilized in the field (5, 9, 10, 67, 70-72, 93, 122). These are briefly described below.

The dominant theoretical approach and research strategy to each of these questions has been the so-called "depletion/repletion" model. In this construct, it is postulated that a critical variable, usually associated with the magnitude of a specific "store," will cause ingestive behavior when it is reduced below an implicit threshold. Feeding behavior will cease when the level of this variable has returned or been "repleted" to the "defended" level. Among the earliest suggestions for such a critical variable was the volume of stomach contents. In this model, depletion was represented by gastric contractions generated by an "empty" stomach and repletion was represented by the meal-related expansion of the volume of the stomach (37). The depletion/repletion model for the control of feeding behavior was derived from a direct translation of the negative-feedback control systems common in engineering to the physiological problem of the regulation of energy balance. Thus changes in carbohydrate, fat, and/or protein body stores were postulated to lead to the generation of appropriate "error signals" and appropriate changes in food intake to return the regulated store or stores to the desired level. When the theory was reduced to practice, however, it was appreciated that the body fat store was fairly easy to estimate and manipulate, while the carbohydrate and protein stores were not. Measurable surrogates such as liver and, in some cases, muscle glycogen and lean body or fat-free mass were used to represent carbohydrate and protein stores, respectively. Despite these practical difficulties and limitations, this theoretical model has dominated the field and a large experimental literature has accumulated based on this theoretical approach (5, 9, 10, 67, 70-72, 93, 122). Indeed, in spite of widely appreciated deficiencies, most discussions of the control of feeding behavior begin, and all too often end, with the consideration of only the depletion/repletion model.

The research strategy most often suggested by this model was to experimentally manipulate one of the critical stores, most often body weight, and to measure the resulting change in food intake. This model also stimulated a search for the identification of the "critical or error" signals that lead to the observed change in food intake.

The search for a representation of body energy stores in the plasma led to another major theoretical construct in the field: the regulation of blood levels of metabolic substrates or "fuels." The most important example was the glucostatic theory of Mayer (76, 77). Other examples of this type of model were the aminostatic, thermostatic, and lipostatic theories of food intake control (5, 9, 10, 70-72, 93, 122). The glucostatic theory and its experimental evaluation were discussed in detail above in section II. Briefly, this theory held that the rate of glucose utilization in a privileged brain region controlled food intake. Blood glucose concentration and/or its arteriovenous difference were used to infer the rates of glucose utilization. Thus hypoglycemia and its associated decreased glucose utilization led to initiation of feeding, whereas postprandial hyperglycemia and increased glucose utilization resulted in cessation of feeding and a period of satiety. However, as the theory spread throughout the field, much of the experimental focus and debate became centered on blood glucose concentrations rather than on rates of glucose utilization in an unknown brain region.

The lipostatic hypothesis, originally postulated by G. Kennedy (65) and J. Mayer (77), also proposed that metabolic fuels or signals generated from adipose tissue, circulated in the blood, and acted on the brain to match or balance energy intake and energy expenditure and body fat mass over the long term (days or weeks). The candidate signal molecules most well studied were the fatty acids and glycerol released from adipose tissue by lipolysis. Thus increased adipose tissue-derived fatty acids and/or glycerol led to feeding, whereas postprandial decrease in lipolysis resulted in cessation of feeding and a period of satiety (70-72).

Another candidate has emerged as a result of the cloning of the obese (ob) gene and the identification of adipose tissue as the primary source of its gene product, leptin (also known as OB protein) (141). The existence of a circulating signal from adipose tissue that acts on the brain to reduce food intake was established by the studies of Hervey and Coleman and their colleagues (39-42, 57-59). When cross-circulation (or parabiosis) experiments were performed between lean and obese rats by Hervey and colleagues at the University of Leeds (58, 59), the lean partner reduced its food intake and lost weight while the obese partner continued to gain weight. These studies suggested that the obese partner was producing increased amounts of a factor that was proportional to body fat and the factor crossed into the circulation of the lean partner where it acted on the brain to reduce food intake and body weight. When cross-circulation (or parabiosis) experiments were performed in ob/ob and db/db mice by Coleman and colleagues at the Jackson Lab (39-42, 57), they also observed that the ob/ob partner reduced its food intake and lost weight, while the db/db partner maintained both its food intake and body weight. These studies led Coleman to conclude that ob/ob mice fail to produce a circulating factor (or factors), perhaps from adipose tissue, which normally acts on their brain to reduce food intake. However, ob/ob mice still retain the ability to respond to this factor. He also concluded that db/db mice produce this circulating factor, but their brain cannot respond to it (39, 41). Experimental attempts to purify a bioactive factor from the adipose tissue of overfed, obese rats independently conducted by Harris and Martin (56) and Hulsey and Martin (61) were not successful. Initial, impure extracts did have biological activity, but this activity was lost after further purification. The insightful hypotheses of Hervey and Coleman have been proven by the emerging biology of the leptin pathway since 1995. Evidence suggests that leptin appears to play a major role in the long-term regulation of adipose tissue mass through coordinated regulation of feeding behavior, metabolism, autonomic nervous system, and body energy balance in rodents, primates, and humans (13, 25-27).

The aminostatic hypothesis, originally postulated by Mellinkoff et al. (81), proposed that amino acids were candidate signals generated from the breakdown of protein stores in muscle, circulated in the blood, and acted on the brain to match or balance energy intake and energy expenditure and body fat mass over the long term (days or weeks). Thus increased muscle catabolism and elevation of amino acids led to feeding, while postprandial uptake of amino acids from the plasma into muscle resulted in cessation of feeding and a period of satiety. These ideas have been pursued, and the current status of the field has been reviewed by Gietzen (52).

The thermostatic hypothesis, originally postulated by Brobeck (11), also proposed that changes in skin, visceral, and core temperature could induce behaviors that would appropriately alter energy balance and match energy intake and energy expenditure and body fat mass over the long term (days or weeks). Thus hypothermia and decreased thermogenesis and its associated metabolic effects led to feeding, while postprandial hyperthermia resulted in cessation of feeding and a period of satiety. In addition to the very large literature on the role of thermogenesis in energy balance, this concept has been studied and reviewed by Woods and Strubbe (138) and Himms-Hagen (60).

The research strategy most often suggested by this model was to experimentally manipulate one of the putative circulating metabolic substrates or fuels, such as blood glucose, and to measure the resulting changes in food intake. These experiments were most often performed in animals and humans in specific metabolic states with the goal of modifying the anticipated behavioral response to one inappropriate for the prevailing metabolic state (99). For example, food-deprived or "hungry" subjects should eat less than expected following the experimental increase of blood glucose concentration. Although this model was originally conceived to be compatible with several blood-borne metabolic substrates modulating food intake under different circumstances, it unfortunately led to narrowly focused and mutually exclusive, competing searches for the identification of the "critical" or "major" blood-borne metabolite that globally controlled food intake. Although such a single blood-borne signal was not required or implied by this conceptual model, it was often invoked to justify the primacy of one factor over another and to push to the margins of the discussion and debate interesting multifactorial models that required integration of these multiple signals over different temporal domains and metabolic states.

The shift in focus away from glucose utilization led many in the field to consider the glucose molecule itself to be the signal for the control of food intake. This was consistent with the trend in the field to consider molecules that circulate in the blood (e.g., insulin, free fatty acids, glycerol, glucagon, CCK, gastrin, catecholamines, adipsin, enterostatin, leptin) to be candidate control signals for the regulation of food intake (5, 9, 13, 25, 70-72, 93, 122). This concept emerged from endocrinology and metabolic physiology and gave the field a much longer list of candidate signals. Many of these signals could be accounted for or associated with the glucostatic, aminostatic, and lipostatic theories of food intake control. However, other signals emerged from the renewed interest in gastric and intestinal mechanisms in the digestion and absorption of nutrients as well as the control of food intake.

The discovery and elucidation of brain/gut peptides that had powerful and often dramatic effects on food intake, and the striking advances achieved in receptor biochemistry and molecular biology also contributed to the "molecule as signal" paradigm. In recent years, the effects of brain and/or gut peptides, peptide fragments, mixtures of peptides, and peptide receptor antagonists on food intake have been studied (118). The success of molecular biology and its penetration into the field of ingestive behavior have made gene expression studies and the use of antisense nucleotide probes to disrupt expression of neuropeptides, so-called knock-out animals, possible. These techniques have identified new brain neuropeptides and receptors and provided new understanding of well-known neuropeptides, involved in the regulation of food intake such as neuropeptide Y (NPY), proopiomelanocortin (POMC), agouti-related peptide, melanocortin receptors and the interaction of these neuropeptides with leptin (26). Also, studies of neuro- or regulatory peptides have awakened interest in the classical neurotransmitters. Although common to earlier research, specific written descriptions of linkages between molecular signals and body stores are now often sketchy or missing. Thus we study molecules per se and construct theories of feeding control based on molecules, for example, the insulin, CCK, or serotonin hypotheses (63, 104, 122, 135).

One of the most compelling theoretical constructs in the field of feeding and body energy balance is the brain insulin hypothesis developed by Stephen Woods and colleagues at the University of Washington in Seattle, and now at the University of Cincinnati (63, 104, 135). This model accounts for the central integration of a circulating signal related to adipose tissue mass, plasma insulin concentration, and the resultant modulation of daily food intake and, thus, body weight regulation. The current formulation of this evolving theory and the experimental evidence supporting it have been reviewed (63, 104, 105, 135). In the brain insulin hypothesis, the 24-h and day-to-day fluctuation or pattern of plasma insulin concentration, when integrated by brain insulin transport and subsequent insulin-dependent signal transduction, is a critical variable of the control of body weight. Although many other factors are involved in the regulation of body weight, the brain insulin hypothesis has provided important insights.

The recent identification by positional cloning of the genes, and predicted gene products, responsible for single gene mutant models of obesity in rats and mice (e.g., ob, db, tub, fat, and agouti) has increased the focus on alterations of feeding behavior mediated by a single factor to the exclusion of others (e.g., leptin only, melanocortin receptor only), rather than the integration of multiple factors (13, 25).

The research strategy most often suggested by this modification of the metabolic substrate or fuel model was very similar to that inspired by the original model as discussed above.

Increasing attention has been focused recently on behavioral sequences or patterns. Well-established orosensory motor patterns have been described by Grill and Norgren (54). Also, stereotypic satiety behavioral sequences have been described by Smith and co-workers (1, 118). Thus complex motor programs underlying feeding are thought to reside in the central nervous system (CNS), and signals related to body stores or specific molecules may "initiate" or "trigger" them under appropriate circumstances.

The consideration of sequences in which a complex temporal and/or spatial pattern fulfills the role of initiator of these motor programs has emerged as an alternative theoretical construct (22, 23). Thus the temporal pattern of a specific molecule (e.g., transient decline in blood glucose), the pattern of several molecules (e.g., nutrient flux across the intestine), the temporal pattern of a specific molecule in a specific context of other patterns (e.g., insulin before, during, or after a meal; oral signals; or gastric distension) or the spatial pattern caused by the passage of a specific molecule through body compartments (e.g., glucose interacting with multiple glucose receptive neural elements, insulin in the brain and cerebrospinal fluid, leptin from adipocytes to the brain) may act as control signals that the CNS uses to organize feeding behavior. Within this conceptual framework, it is not the store or molecule (neither its concentration or absolute amount) but rather the dynamic "pattern" of the molecule that conveys critical "information" to the CNS.

Another key element of the pattern as signal concept is the notion of representation of peripheral events and states within the CNS. The CNS may contain "representations" of metabolic and behavioral states: absorptive and postabsorptive states, hunger, and satiety. Although often thought of in terms of spatial maps, transformations, or homunculi, representation in this context is considered to be a dynamic pattern of activity in a set of neurons that function as a central analog of sensory somatic or visceral events. Thus we seek these representations of peripheral body stores, metabolic state, and the external world and how these representations interact to control feeding behavior.

What form would these representations take? Based on our knowledge of the nervous system and by analogy to other brain systems involved in sensorimotor integration and chemosensory detection and processing, these representations would probably be patterns of electrical and/or neurochemical activity of one or more neural networks controlling feeding. Thus, in this theoretical construct, patterns in the periphery carry information that is detected and recognized by central neural networks, which results in modified patterns of activity of these neural networks. Specific patterns of activity of these networks are postulated to correspond to the behavioral states associated with meal initiation and meal termination. Transition from the representation corresponding to satiety (a period of no food intake) during intermeal intervals to the representation appropriate for, and corresponding to, meal initiation within the CNS will cause the onset of feeding. Meal termination will occur when the representation corresponding to satiety is restored by reversal of the meal initiation pattern activity by the complementary pattern corresponding to satiation (the active process of meal termination, see Refs. 70-72). If we could capture a visual image of the patterns of activity of the feeding network before, during, and after a meal, we would see a transition from the pattern characteristic of the intermeal interval to that of meal initiation and food ingestion followed by a transition back to the pattern corresponding to satiety. Thus activity patterns corresponding to two active processes, meal initiation and satiation, will be superimposed on the hypothesized passive or low-activity steady state corresponding to satiety (a period of no food intake) each time an animal eats a meal. In this theoretical construct, the components of a behavioral sequence become distinct patterns of activity in discrete neuronal networks and are linked to the underlying physiological and biochemical dynamics of these networks. Mechanistic decomposition of the behavioral sequences of feeding will require the identification and characterization of the major components of the key elements of that sequence: initiation, maintenance, and termination. Perturbation or shaping of feeding behavior would be then translated into modulation of these complex activity patterns and their integration by the brain.

Since formulating this conceptual framework for feeding behavior in a review published in 1990 (23), we have come to believe that its utility both in terms of explanatory power and an ability to tightly link feeding behavior and physiology has been demonstrated. We remain optimistic about its potential to refocus our field on the message rather than the messenger. This point of view also has the potential to synthesize much of our field because, rather than debating the merits of transient declines in blood glucose or hypothalamic norepinephrine or hindbrain CCK or circulating leptin, we can ask how all of these components or elements of the pattern are integrated to elicit specific behaviors underlying feeding. This hypothesis is also focused on the detection and recognition of these multiple patterns by the widely distributed neuronal networks controlling feeding. Thus the goal of the search is not the pattern but rather the patterns that together form the central representations of meal initiation, maintenance, and termination. The integrative focus of this conceptualization has much in common with and is a descendent of earlier integrative approaches of the balance of lipogenesis and lipolysis rates (70-72), combining energy flux in the plasma (5, 6) and neurons that integrate specific molecular signals (84). However, this concept has an even broader focus on the totality of central representations of peripheral events related to feeding. An analogy may be that the brain is responding, conditioned on its current state, to the entire "symphony" rather than the "voice of individual sections" of the orchestra. Finally, the ability of the "pattern as signal" construct to integrate the dynamics of blood glucose before, during, and after meals into one or more important messages related to feeding behavior suggests that a similar focus on the patterns of insulin, leptin, and CCK rather than a focus on the location or concentration of these hormones may also yield important insights. The application of this conceptual framework and its related research strategy applied to the problem of meal initiation will be the focus and subject of this review.

The experimental studies presented and reviewed in our previous publications (14, 18, 22-24, 29, 34, 35, 111, 112, 114) and to be reviewed below have led to the formulation of a signal detection and recognition theory of meal initiation. The major assertions of this theory are that: 1) transient declines in blood glucose represent "endogenous metabolic patterns," 2) transient declines in blood glucose are signals in the form of "patterns" that are detected and recognized by the central neural network that controls feeding behavior, and 3) these patterns are "mapped into" meal initiation under free-feeding conditions. The phrase mapped into meal initiation is meant to describe the process of transformation, in the mathematical sense, or establishing a unique correspondence between the recognition of the transient decline in blood glucose and the activation of the motor program for feeding within the CNS.

The distinguishing feature of this assertion is that it is the temporal pattern, shape, or waveform of blood glucose dynamics rather than the glucose molecule, or the absolute decrease in blood glucose, or blood glucose concentration or glucose utilization that is detected and contains "critical information" that is extracted by the central nervous system to control meal initiation.

We propose that the processes of detection, recognition of a transient decline in blood glucose, and its mapping into meal initiation behavior are accomplished by the set of spatially and temporally distributed processes shown in Figure 2. Meal initiation will occur only if a "timing or probe signal" was generated recently and a transient decline in blood glucose is detected in the brain and recognized. Figure 2 depicts the information and signal flow through the set of sequential processes without regard to the anatomical localization of the process. The "timing or probe signal" represents an "inquiry" or "interrogation" or "probing" of the periphery by the CNS regarding its ability and capacity to maintain glucose homeostasis in upcoming time interval. This signal is generated by a signal generator (in the CNS) that controls its frequency or timing. The signal acts on the biochemical subsystem that regulates blood glucose concentration and, depending on the peripheral metabolic state and its ability and capacity to "maintain glucose homeostasis" over the next time interval, a transient decline in blood glucose (TDBG) will be generated or not. An average transient decline in blood glucose is shown in Figure 1.

View larger version (21K): [in this window] [in a new window]   Fig. 2. Proposed information and signal flow in the pattern detection and recognition model. The information flow chart is shown without regard for the anatomic location of each process. The process of meal initiation begins with the generation of a "probe or timing signal" within the brain, continues as this signal "passes through" the peripheral metabolic system acting on the blood glucose regulatory system, and ends with an affirmative decision to activate the "meal initiation program." The blood glucose regulatory system generates an insulin spike and a transient decline in blood glucose (represented by the sketch; middle). See text for details.

Any changes in blood glucose concentration from baseline values are detected by peripheral and central glucose responsive neural elements which form the "TDBG detector" process in Figure 2. The recognition of the presence of a transient decline in blood glucose that meets all the necessary criteria from among the many changes in blood glucose concentration occurring in the peripheral blood is reported by the TDBG recognition process. This process is located in both the CNS and the periphery, and its output acts on the "decision algorithm" process. The decision algorithm unit computes the conditional probability of activating the stored meal initiation program given the state of the three inputs it receives: 1) output from the TDBG recognition process, the recognition of TDBG; 2) peripheral metabolic state; and 3) presence of a recent "probe or timing signal." If a transient decline in blood glucose is recognized in the presence of a favorable peripheral metabolic state and the recognition of recent generation of the "probe of timing" signal, then the output of the decision algorithm will be yes and the meal initiation program will be activated. By a "favorable metabolic state" we mean the later part of the intermeal interval which would allow a transient decline in blood glucose in response to the probe or timing signal. The transient decline in blood glucose will indicate that additional energy intake from a meal may be required to maintain blood glucose over the coming time interval (see below). An "unfavorable" metabolic state would be just after a meal. Unless other events interfere with the meal initiation program or the act of feeding, the meal initiation program, when activated, generates all of the motor acts required for expressing meal initiation behavior.

Our current working hypothesis is shown in Figure 3 and is as follows. In free-feeding rats, we have shown that there are brief, spontaneous plasma insulin peaks antecedent to each decline in blood glucose (19, 29). We postulate that a "probe or timing signal" generated by the CNS results in a brief change in the firing rate of the parasympathetic (vagal) and/or sympathetic nervous system efferents which innervate pancreatic -cells, liver, adipose tissue, and the gastrointestinal tract. The change in autonomic firing rate causes a brief insulin spike from pancreatic -cells, along with other responses, which, in the presence of an appropriate peripheral metabolic state, induces a transient decline in blood glucose by decreasing hepatic glucose production and/or increasing peripheral glucose disposal. In the presence of hyperglycemia, the brief insulin spike, along with other responses, may have to be larger in magnitude to induce a transient decline in blood glucose (see sect. IVB2). The peripheral metabolic state is postulated to "condition," in a probabilistic sense, or "gate," in a signal flow sense, the likelihood of a feeding response to the activation of vagal and/or sympathetic efferents. Thus a central/peripheral interaction is proposed in which a centrally generated meal initiation signal must "pass through" the peripheral metabolic system, and it will be mapped into feeding behavior only if a transient decline in blood glucose of the correct shape or pattern occurs in response to a recent probe signal. If the output of the decision algorithm remains in the "no" state, either because of the absence of a transient decline in blood glucose or the failure of a decline to meet all the criteria for meal initiation or an unfavorable peripheral metabolic state (e.g., recent meal), the TDBG recognition unit will be reset to the "no decline" state. If the peripheral metabolic state cannot maintain glucose homeostasis over the next time interval, without additional energy intake from a meal, a transient decline in blood glucose in response to a probe signal will occur. On the other hand, if glucose homeostasis can be maintained over the next time interval without additional energy intake (e.g., liver glycogen breakdown and utilization), a transient decline in blood glucose will not occur. However, if a transient decline in blood glucose occurs but is ignored and no food is eaten, the only consequence may be the occurrence of the next transient decline in blood glucose sooner than expected. In this case, blood glucose may be maintained through novel involvement or activity of the liver or other organs.

View larger version (31K): [in this window] [in a new window]   Fig. 3. Block diagram representation of the information and signal flow in our conceptual model of the pattern detection and recognition theory of meal initiation. A "probe or timing signal" is generated within the brain (top left) and causes a brief change in firing rate of the parasympathetic (PNS) and/or sympathetic (SNS) afferents which innervate pancreatic -cells, glucose regulatory system, and gastrointestinal (GI) tract (middle and bottom left). If additional energy (a meal) is needed to maintain glucose homeostasis over the next time interval, the probe signal (e.g., insulin spike) will "pass through" the peripheral metabolic system acting on the blood glucose regulatory system and generate a transient decline in blood glucose (represented by the sketch; middle). Changes in blood glucose concentration will be detected by peripheral and central glucose responsive neural elements (middle and top right) and then the transient decline in blood glucose is "recognized" in the brain (top right). If the criteria of the decision algorithm (top) are satisfied (transient decline recognized, favorable peripheral metabolic state, recent probe signal), the "meal initiation program" will be activated and meal initiation behavior will be observed (top left). If the decision criteria are not satisfied, the "recognition" unit will be reset and meal initiation will not be observed until the next transient decline in blood glucose occurs. TDBG, transient decline in blood glucose; GPR, glucose production rate; GDR, glucose disposal rate.

The model assumes that some probe or timing signals will not result in transient declines in blood glucose, and, therefore, meal initiation will not be observed. Therefore, a prediction of our model is that the number of probe or timing signals should be greater than the number of meals. The acetylcholine analog bethanechol, which provokes a brief spike in insulin concentration, was used to induce transient declines in blood glucose and meal initiation in rats. In these studies, transient declines in blood glucose were observed in 63% of the trials (10 of 16 trials), and meal initiation was observed in 56%. In contrast, 38% of injections of bethanechol (6 of 16 trials) were not followed by a transient decline in blood glucose and meal initiation, thus providing support for this hypothesis (111). In these same experiments, the same dose of bethanechol that induced transient declines in blood glucose in some circumstances (e.g., time during of light-dark cycle), failed to do so in others (e.g., light phase).

The processes of detection and recognition are considered distinct in our formulation because we interpret our experimental results to mean that the minute-by-minute blood glucose concentration is detected by glucose responsive peripheral afferents and central neurons. Thus blood glucose is continuously represented and always available to the CNS. The process of detection begins with a representation of the relatively steady intermeal concentration of blood glucose ("baseline") and "reports" the detection of all departures in blood glucose concentration from that baseline to the TDBG recognition unit. In contrast, the unique process of recognition that an ongoing pattern of blood glucose concentration "matches" or "fits" the criteria for a transient decline in blood glucose is the function of the TDBG recognition unit. Thus the recognition of a transient decline in blood glucose is the result of a two-step process: 1) detection that the blood glucose concentration is changing and 2) the recognition that the pattern of the blood glucose change meets the criteria for transient declines in blood glucose. Another way of looking at the function of these two processes is that the TDBG recognition unit must recognize the shape or pattern of transient declines, but only transient declines, out of the set of all other changes in blood glucose concentration detected throughout the day.

The output of the decision algorithm was considered to also be yes throughout the duration of the transient decline in blood glucose, but the output is assumed to switch to no after the blood glucose concentration returned to baseline for 6 min, or more. Therefore, the time interval between the output of the decision algorithm switching from the no state to the yes state, following the initial recognition of the transient decline in blood glucose, to the output switching back from yes to no states was of short duration (~12 min in rats) and persisted <6 min after the end of a blood glucose decline.

The necessary conditions for a fall and rise in blood glucose concentration in rats to be recognized as a transient decline in blood glucose have also been calculated from our experimental data set. The following necessary conditions must be met sequentially by all transient declines in blood glucose that signal meal initiation in rats: 1) the slope of the falling phase must be within 0.4 and 1.5 mg/dl · min; 2) the nadir of the decline must be at least 6% below baseline and must occur between 40 and 60% of the total duration of the decline, and the total duration of the transient decline must be longer than 6 min; and 3) the slope of the rising phase must be within 0.5 and 1.5 mg/dl · min.

These necessary conditions, which specify an approximately symmetrical fall and rise in glucose concentration, shown in Figure 1, provide a first-order approximation of the criteria for recognition of a transient decline in blood glucose by the TDBG recognition unit in our model.

The conditions required for a yes output of the decision algorithm and activation of the meal initiation program are proposed to be as follows: the recognition of a transient decline in blood glucose that was generated by a receptive or favorable peripheral metabolic state in response to the "recent" generation of a probe or timing signal by the CNS.

As discussed above, if the output of the decision algorithm remains in the no state, because these criteria were not met, the TDBG recognition unit will be reset to the no decline state.

Although the conceptual basis and a basic structure of our working hypothesis has been represented in our model shown in Figure 3, many aspects of the model remain to be determined in quantitative terms. However, we feel that this model, and the hypotheses on which it is based, provide a new framework for the understanding of feeding behavior, with special emphasis on the evolution of hunger, the initiation of feeding and its dependence on patterns of blood glucose.

	IV. PHYSIOLOGICAL AND BEHAVIORAL STUDIES OF TRANSIENT DECLINES IN BLOOD GLUCOSE AS A SIGNAL FOR MEAL INITIATION: STUDIES IN RATS

Top Previous Next References

A.  Basic Studies Under Free-Feeding Conditions in Rats

1.  Transient declines in blood glucose as a signal for meal initiation

Blood glucose concentration and meal pattern were continuously monitored in rats feeding freely using methods described previously in detail (14). Since all of the studies reviewed have utilized these same techniques and the evidence supporting a role for glucose in meal initiation is critically dependent on these technological innovations, it is important that the reader has a clear understanding of these methods. Thus these methods will be briefly described.

Adult rats were housed in individual cages with free access to food and tap water (Fig. 4). Powdered rat food was placed in a food cup fitted with strain gauge weighing apparatus. Animals were kept in a temperature-controlled room with 12:12-h light-dark schedule. Rats were implanted with chronic cannulas in the right atria of heart. After a seven-day recovery period, characterized by resumption of consistent gain in body weight and normal meal pattern, experimental studies were conducted in freely moving rats in their home cages. Heparin sodium (200 U) was injected intravenously, and 45 min later, blood withdrawal (25 µl/min) for continuous blood glucose monitoring was initiated. Venous blood was withdrawn from freely behaving rats through polyethylene tubing attached to the cardiac cannula and injected into the sample chamber of a glucose analyzer (YSI model 23A or 27). The analog outputs from both the food cup and the glucose analyzer were sampled 8-10 times a minute, amplified, digitized, and interfaced to a microcomputer. Blood glucose monitoring was continued for up to 3 h. Experiments have been conducted throughout the light-dark cycle (14).

View larger version (29K): [in this window] [in a new window]   Fig. 4. Experimental set-up for studies of continuous recording of blood glucose concentration and meal initiation. Note that the animal remains in its home cage during blood glucose recording. [From Campfield et al. (14), with permission from Elsevier Science.]

When the blood glucose concentration and meal pattern were monitored continuously in free-feeding rats, a transient fall and rise in blood glucose was observed before each meal independent of the light-dark cycle. The average time course of blood glucose in nine early experiments is shown in Figure 5. The blood glucose concentration was expressed as percent change from baseline; the time 0 reference point was chosen as the minimum blood glucose, and the data points were averaged at 1-min increments or decrements before and after the nadir. During an average decline, blood glucose concentration fell gradually to 11.6 ± 1.2% below baseline. Blood glucose concentration began to decline 12.1 ± 1.7 min before the onset of food intake and continued to decline to a minimum at 5.4 ± 1.5 min before meal onset. Note that these average times were measured from the beginning of the meal as opposed to the minimum glucose concentration. Correlations between meal size and percent maximum decline in blood glucose concentration and the total duration of the decline were both nonsignificant. Thus transient declines in blood glucose predict meal initiation but not meal size.

View larger version (19K): [in this window] [in a new window]   Fig. 5. Average time course of the transient decline in blood glucose when access to food was free or prevented. Blood glucose concentrations are expressed as percent change from the intermeal baseline concentrations in this figure. The minimum glucose concentration has been taken as the time 0 reference. Data were selected each minute from the reference point and averaged in each of nine experiments. Data are means ± SE. The onset of food-seeking behavior is indicated by the vertical arrows in free access () and prevented access () groups. Mean baseline glucose concentrations were 104 ± 5 and 96 ± 3 mg/dl, respectively. Note the similar time course of blood glucose and onset of food-seeking behavior when food access was free and prevented. [From Campfield and Smith (18), with permission from Elsevier Science.]

Random fluctuations in blood glucose occurred in these experiments that were not related to feeding. Analysis of the lower limits of transient declines in blood glucose that were associated with meal initiation indicated that blood glucose decreases with magnitudes >6%, and durations of >6 min invariably preceded food intake. When these criteria are applied to our large data set, we conclude that no such transient decline in blood glucose has been observed in the absence of food intake nor has feeding been observed in the absence of such a transient decline in blood glucose concentration.

These findings are consistent with and confirm the original observations of a premeal decline in blood glucose by Louis-Sylvestre and LeMagnen (74). They also are consistent with the decrease in metabolisme de fond before meal initiation reported by Nicolaidis and Even (85). The striking similarity to the time course of blood glucose and metabolisme de fond before meals suggests that blood glucose could be a major contributor to this measure of the resting metabolism of the rat.

These results are also consistent with the premeal patterns in both liver and skin temperature reported by Woods and Strubbe (138). They have observed a rise in liver and skin temperature from the intermeal minimum temperature preceding spontaneous meal initiation in rats. Although meals of different size began at different temperatures, all meals terminated at constant liver and skin temperatures. Our results also permit the reevaluation of the reports that meal initiation after insulin administration occurred when blood glucose was near or at baseline levels (44, 119, 120). Although these results have been used to argue against the glucostatic theory of Mayer (76, 77), the temporal relation between blood glucose and meal initiation is remarkably similar following both intravenous insulin and spontaneous transient declines in blood glucose.

Having documented the temporal relation between transient declines in blood glucose and meal initiation in free-feeding conditions, the next set of questions addressed the nature of this systemic signal for meal initiation. Furthermore, the magnitude or strength and the time course of the functional coupling between blood glucose dynamics and feeding behavior were also investigated. These questions were addressed by preventing access to food before and during declines in blood glucose, restoring access to food at various times following the beginning of the decline in blood glucose, and measuring the latency to feeding and the next decline in blood glucose (18). In experiments conducted across the light-dark transition, the food cup was covered before and during the transient fall of blood glucose and was uncovered only after blood glucose had returned to the baseline concentration. The time course of blood glucose concentration before an expected meal, expressed as percent change from baseline, in these experiments is compared with similar experiments in which rats had free access to food in Figure 5. The time course of blood glucose concentration in both groups was very similar. Food seeking behavior (i.e., orienting and/or moving to the food cup, sniffing, trying to remove the cover) occurred with a latency comparable to that seen in free-feeding animals. Food-seeking behavior was transient and ended as the glucose concentration increased toward baseline.

In similar experiments, the cage and food cup were thoroughly cleaned before the experiment, and no food was placed in the food cup (n = 6-9). The time course of blood glucose and the latency to food-seeking behavior observed were not different in the total absence of food. Thus neither preventing access to nor the absence of food affected the time course of the transient fall and rise of blood glucose before an expected meal and the latency to food-seeking behavior.

In these experiments, the food cup was uncovered 6-8 min after blood glucose had returned to the baseline concentration and the latency to meal initiation was measured (n = 6-9). In all cases, food-seeking behavior ceased before uncovering the food cup. Feeding was not observed immediately after the uncovering of the food cup but rather after a normal intermeal interval and a second transient fall in blood glucose and rise toward baseline. Meal initiation following the second decline in blood glucose occurred an average of 84 ± 8 min after the beginning of the first decline in blood glucose. This compares with an average intermeal interval for these rats at this phase of the light-dark cycle of 98 ± 7 min. The mean coefficient of variation of blood glucose concentration during the interdecline interval was 2.1 ± 0.5%. The size of the meal was not different from meals eaten at this time by these rats in the free access condition. In other experiments, the food cup was uncovered 4-9 min before blood glucose returned to baseline. In these experiments, rats moved to the food cup and feeding began within 2 min after removal of the food cup cover. These experiments indicate that when access to food was restored as the blood glucose was rising toward baseline, meal initiation occurred as expected.

Combination of these studies and our studies in the free-feeding condition allowed calculation of the approximate temporal evolution of the functional coupling between blood glucose and meal initiation. The resulting composite time course is shown in Figure 1. The latency to meal initiation in the free-feeding condition (12.1 ± 1.7 min from the beginning of the decline in glucose) was taken to be the latest time that the glucose-dependent signal for meal initiation exceeded its threshold and coupled blood glucose to feeding behavior (downward arrow). Because meals were initiated with a short latency when access to food was restored while the glucose was rising toward baseline, the meal initiation signal was considered to be above threshold throughout this period. The termination of the functional coupling was then taken as 6 min after blood glucose concentration returned to baseline because when food access was restored at this time, or later, a normal intermeal interval and a second transient decline in blood glucose preceded meal initiation (upward arrow). Therefore, the functional coupling between blood glucose and feeding behavior was of short duration (~12 min) and persisted <6 min after the end of a blood glucose decline. The transient nature of this coupling required a second decline in blood glucose to initiate feeding when the rat was unable to eat within this narrow temporal window.

In other experiments, a novel food (orange slice, potato chip, or chocolate chip cookie) was presented 30 min after the food cup was uncovered. Access to powdered food had been restored 8 min after the end of the transient decline in blood glucose. The novel food was eaten in seven experiments with an average latency of 2.5 ± 0.9 min after presentation but without a prior decline in blood glucose. These studies demonstrate that novel foods with strong sensory qualities can be eaten without any prior changes in blood glucose concentration (18).

In addition to transient declines in blood glucose, other physiological and metabolic changes can occur in the premeal period. When rats are presented food on fixed time schedules for several days, gastric contractions increase just before anticipated periods of food availability (43). Increased gastric contractions have also been observed following hypoglycemia (83, 125, 126). Because transient declines in blood glucose also occur prior to food availability in conditioning studies in rats (see below), it is probable that gastric contractions occur in association with blood glucose dynamics and meal initiation. Additional metabolic changes are discussed in section IVD.

In summary, the nature of the signal controlling meal initiation that we are proposing is very different from that proposed by Mayer. In his glucostatic theory, Mayer proposed that decreased glucose utilization or metabolic hypoglycemia, the point at which the peripheral arteriovenous difference in blood glucose (A-V delta glucose) becomes negligible and glucose is no longer entering metabolizing cells, was the signal for meal initiation. In contrast, we propose that it is a specific pattern of blood glucose concentration, the transient decline in blood glucose, that signals and controls meal initiation under most conditions. Mayer viewed his signal, metabolic hypoglycemia, as reflecting the point at which the energy substrate flux was at a minimum or turning in the direction of increasing fatty acid utilization, in other words, the point of the beginning of carbohydrate depletion. However, we view transient declines in blood glucose as signals that "interrogate" the peripheral metabolic state and signal meal initiation if additional energy intake will be needed to maintain glucose homeostasis for the coming time interval. Thus transient declines in blood glucose are not the result of a change in direction or magnitude of metabolic flux indicating depletion, but rather are metabolic patterns that signal energy intake to avoid depletion of metabolic fuels. Another important difference is that, theoretically at least, dynamic patterns of blood glucose could explain meal initiation and hunger in the context of a variety of diet patterns and metabolic conditions, while the Mayer glucose utilization, which dependent on changes in glucose utilization and metabolism, lacks this property. Finally, rather than a difference in glucose utilization (a metabolic process) in central glucoreceptors, as represented by peripheral glucose concentration difference across metabolically active tissues, the transient decline of blood glucose is a dynamic pattern that is detected and recognized by central and peripheral glucose-sensitive neural elements and mapped into meal initiation.

Based on their initial studies, Louis-Sylvestre and LeMagnen (74) concluded that blood glucose concentration was among the feedback signals in the regulation of feeding and body energy storage, and the observed decline in blood glucose concentration before meal onset was either the signal for meal initiation or a consequence of the true signal. In either case, a causal relation between the decline in blood glucose concentration and feeding was proposed (70). This proposal has proven to be quite controversial. Several investigators argued that the premeal decline in blood glucose concentration was correlated with rather than causally related to meal initiation (see commentaries in Ref. 70). Motivated by the controversy over the causality of the observed premeal decline in blood glucose, we conducted a series of experiments designed to determine whether transient declines in blood glucose induce meal initiation.

Our first approach was to infuse glucose intravenously to partially block the premeal decline in blood glucose and assess the effect on subsequent feeding behavior. In some experiments, 10% glucose (up to 0.2 ml) was infused over a 5-min period beginning as soon as a fall in blood glucose prior to an expected meal was "recognized." Glucose infusions were administered each time a decline in glucose was recognized. Isotonic saline also was administered in separate experiments to control for the nonspecific effects of infusions (14). The transient declines in blood glucose observed in these experiments are summarized in Figure 6. Comparison of the results obtained indicates that although the initial rate of decline, the timing, and the magnitude of the glucose nadir were not significantly affected by the glucose infusions, the duration of the decline was decreased and the latency to the anticipated meal was increased markedly (318 ± 94 compared with 12.1 ± 1.7 min). It is important to note that 0.2 ml of 10% glucose contains at most 9% of the calories consumed in the smallest meal. The very long latency to the anticipated meal in glucose-infused rats may have been overestimated. The sensitivity of the chart recorder used to monitor the meal pattern was lower (minimum meal size approximately 0.2 g) in these early studies than the computer-based recording system used in later studies. Thus earlier small meals may have failed to be detected leading to an overestimate in the latency. Despite this potential overestimate, these studies indicate that glucose infusions that change the shape of the transient declines in blood glucose delay meal initiation (14).

View larger version (10K): [in this window] [in a new window]   Fig. 6. Effects of glucose clamp on transient decline in blood glucose and latency to the anticipated meal. Comparison of the timing and magnitude of the maximum percent decline and time of the return to baseline in experiments without infusions (solid line), with saline infusions (dotted line), and with glucose infusion (heavy dotted line). The arrows represent the time of onset of the next meal in each group. Note decreased decline duration and increased latency to meal initiation with glucose infusion. [From Campfield et al. (14), with permission from Elsevier Science.]

These studies suggest that it is the shape or pattern of the transient declines in blood glucose that affects meal initiation. Further support for this hypothesis was provided by the results of additional experiments in which glucose was infused during the rising phase at the end of the transient decline. In these experiments, the anticipated meal occurred with normal latency despite the glucose infusion. These observations suggest that it is not glucose itself that uncouples the transient declines in blood glucose from meal initiation but rather a modification of the shape of the transient decline in blood glucose. Taken together, these results offer an explanation for the previously reported failure of intravenous glucose infusions to delay the onset of feeding or reduce meal size (44, 119, 120). Our studies demonstrate that to block meal initiation, a glucose infusion must occur during the early phase of the transient decline in blood glucose. Thus meal initiation would not have been blocked if glucose had been infused either before the decline or toward its end (14).

Additional experiments were performed in which glycine, -hydroxybutyrate, or fructose was infused intravenously instead of glucose. The effects of these infusions on the parameters of the transient decline and the latency to meal initiation were minimal; only glucose infusions blocked meal initiation. Although it would have been interesting, glucosamine and glucagon were not studied. This result suggests that changes in blood glucose rather than any other nutrient tested provides, or generates, the signal for meal initiation.

Additional supportive evidence was obtained from experiments in which fructose was infused intravenously in free-feeding rats to produce transient declines in blood glucose observed before meal initiation (114). Random sequences of fructose (doses range 0.05-0.2 ml of 10%), separated by at least 30 min, were infused intravenously over 2 min during intermeal intervals. During the early dark phase, fructose was followed by slight decreases or increases in blood glucose (4 to +10% at 6 min). In the light phase, however, three types of dose-dependent declines in blood glucose were observed. The first pattern was a fall to a suppressed level (6%) that was maintained for at least 30 min. The second pattern was a transient fall (10% at 8 min) and return to baseline at 28 min. However, the third pattern mimicked the transient decline in blood glucose observed before meal initiation: a transient fall (9% at 8 min) and return to baseline at 17 min, during which meal initiation occurred with a latency within the normal range for spontaneous meals following transient declines in blood glucose. No feeding was observed after the other two blood glucose response patterns induced by fructose infusion. The observation of feeding during a period of low probability for spontaneous meals only following a fall and rise in blood glucose induced by fructose that mimicked the transient decline in blood glucose provides further evidence that patterns in blood glucose dynamics are a causal