The elusive cephalic phase insulin response: triggers, mechanisms, and functions

Abstract

The cephalic phase insulin response (CPIR) is classically defined as a head receptor-induced early release of insulin during eating that precedes a postabsorptive rise in blood glucose. Here we discuss, first, the various stimuli that elicit the CPIR and the sensory signaling pathways (sensory limb) involved; second, the efferent pathways that control the various endocrine events associated with eating (motor limb); and third, what is known about the central integrative processes linking the sensory and motor limbs. Fourth, in doing so, we identify open questions and problems with respect to the CPIR in general. Specifically, we consider test conditions that allow, or may not allow, the stimulus to reach the potentially relevant taste receptors and to trigger a CPIR. The possible significance of sweetness and palatability as crucial stimulus features and whether conditioning plays a role in the CPIR are also discussed. Moreover, we ponder the utility of the strict classical CPIR definition based on what is known about the effects of vagal motor neuron activation and thereby acetylcholine on the β-cells, together with the difficulties of the accurate assessment of insulin release. Finally, we weigh the evidence of the physiological and clinical relevance of the cephalic contribution to the release of insulin that occurs during and after a meal. These points are critical for the interpretation of the existing data, and they support a sharper focus on the role of head receptors in the overall insulin response to eating rather than relying solely on the classical CPIR definition.

CLINICAL HIGHLIGHTS.

The cephalic phase insulin response (CPIR) is physiologically and clinically important because it prepares the body for the optimal assimilation of ingested food, particularly glucose-containing carbohydrates, and, indeed, facilitates the process. A disrupted or absent CPIR is implicated in the development of insulin resistance and type II diabetes mellitus. The absence of oropharyngeal contact with food during enteral or parenteral feeding regimens may be complicit in the disruptions of glucose homeostasis seen in some patients. This review discusses the various stimuli that trigger a CPIR, the sensory and neural mechanisms involved, and its physiological functions. In this context, it is critical to emphasize that cephalic stimulation of insulin secretion continues throughout the meal, which is neglected by the strict classical definition requiring insulin release prior to any increase in circulating glucose. Accordingly, a focus on the role of cephalic stimulation in the overall insulin response to eating, rather than a sole emphasis on the beginning of the meal, more fully captures its physiological and clinical impact.

1. INTRODUCTION

1.1. Preamble

Eating is accompanied by complex physiological processes. Naturally, these processes comprise the stimulation of digestive secretions (saliva, gastric, and pancreatic exocrine secretions) to efficiently accomplish the breakdown of the ingested food and changes in motility that ensure proper gastrointestinal transit. In addition, even before nutrients are absorbed, eating triggers the release of several gastrointestinal (see Refs. 1, 2) and metabolic (Refs. 3, 4) hormones, activates the autonomic nervous system (e.g., Refs. 5–7), and induces metabolic changes, including an increase in thermogenesis (e.g., Refs. 8, 9). While it is obvious that eating must trigger the release of digestive secretions and initiate changes in gastrointestinal motility, it is not immediately clear why eating should trigger the release of metabolic hormones and induce metabolic changes before nutrient absorption. Yet, all these responses combine to accomplish an efficient processing of the ingested nutrients from digestion and absorption to postabsorptive metabolism and storage. Many of the underlying physiological mechanisms are already activated early in the meal, or even before the onset of eating. This early stimulation results from the activation of head (cephalic) receptors. These phenomena are therefore collectively termed cephalic phase responses. Evidence suggests that meal expectations, visual and olfactory cues, and, importantly, oropharyngeal chemosensory stimuli are all capable, to a greater or lesser degree, of triggering some of these responses (e.g., Refs. 10–15).

1.2. Historical Background

Cephalic phase responses were first identified ∼120 years ago by Pavlov (see Ref. 13). He received the 1904 Nobel Prize in Physiology or Medicine for his work on the physiology of digestion. In a series of experiments in dogs, he showed that the taste of food in the oral cavity not only stimulated the secretion of saliva but also of gastric and pancreatic juices. These secretory responses were independent of gastric distension or nutrient digestion and absorption because the dogs were equipped with esophageal and gastric fistulas to prevent any effect that the nutrients might otherwise have in the stomach or small intestine (13). In addition, Pavlov showed that these responses could be conditioned. After repetitive pairing of a bell or the sight of food, referred to as the conditioned stimulus, with food ingestion, referred to as the unconditioned stimulus, the presentation of the conditioned stimulus alone came to trigger digestive secretions, referred to as the conditioned response(s) (13, 16).

Interestingly, Pavlov also found that a subdiaphragmatic vagotomy blocked the gastric and pancreatic secretory responses, thus providing what is probably the first evidence of a major role of the vagus nerve in these secretions (13). It was actually Pavlov who coined the term “cephalic phase responses,” with the critical criteria that they are neurally mediated, anticipatory responses that occur before nutrient absorption (13). Only many years later did the advent of reliable radioimmunoassays to measure the circulating concentrations of hormones herald the search for cephalic phase endocrine responses, and in this context, attention rapidly focused on insulin, i.e., on the so-called cephalic phase insulin response (CPIR).

Given the regular, recurring nature of eating, it is also not surprising that many, if not all cephalic phase responses, can be conditioned. In a remarkable review, published in 1991 (7), Woods writes “The act of eating, although necessary for the provision of energy, is a particularly disruptive event in a homeostatic sense. Just as humans learn responses to help them tolerate the administration of dangerous drugs, so do they learn to make anticipatory responses that help minimize the impact of meals on the body, to limit the amount of food consumed within any individual meal, to recruit several parts of the protective stress-response system while meals are being processed, and to limit postprandial behaviors so as to minimize the possibility of disrupting homeostatic systems even more.”

1.3. Cephalic Phase Insulin Response: What Is It?

While there are multiple cephalic phase responses, and even several involving hormones, traditionally, the emphasis has been on the CPIR, which is the focus here as well. Many authors somewhat narrowly define the CPIR as a head receptor-triggered, neurally mediated insulin release that occurs before nutrient absorption and is transient (e.g., Refs. 17–22). The latter feature of this standard definition ignores the fact that the stimulation of cephalic receptors continues throughout the meal and, hence, is not limited to the early phase of ingestion (see FIGURE 1). Interestingly, with respect to digestive exocrine secretions, this had already been beautifully described by Pavlov (13), who emphasized the importance of the continuation of oral stimulation for normal gastric secretion during eating. With respect to the CPIR, this was demonstrated by D’Alessio et al. (23) and then further discussed in relation to its possible effects on eating by Smeets and colleagues (24). In fact, much earlier Louis-Sylvestre and Le Magnen (25) had already alluded to the fact that the cephalic stimulation continues throughout the meal by referring to the initial phase as an “on response triggered by the first oral contact with the food.” The preabsorptive increase in insulin is therefore mainly an interpretive demand that allows for the dissociation of the site of stimulation under many experimental conditions. That is, if blood glucose increases, it is difficult to disambiguate whether the rise in insulin represents a CPIR or is due to direct stimulation of pancreatic β-cells. The preabsorptive increase in circulating insulin is usually small compared to its overall prandial and postprandial release, but because the oral stimulation continues throughout the meal, the cephalic phase contribution to the total insulin secretion in response to eating is much greater than commonly assumed. The continuation of cephalic stimulation throughout the meal is also reflected by the fact that the cephalic stimulation of insulin release has been shown to have a substantial effect on postprandial glucose metabolism (e.g., Refs. 26–29). In addition to the direct measurement of the dynamics of circulating insulin and glucose, investigators therefore often infer a CPIR indirectly from changes in glucose tolerance (see Ref. 14).

FIGURE 1.

A hypothetical timeline of carbohydrate (CHO) processing (i), and the dynamics (not shown to scale) of oropharyngeal-gastrointestinal (Oro-GI) glucose (ii), blood glucose (iii), and blood insulin (iv) after food is either ingested via the mouth (A) or given directly into the stomach (B). Note that in this diagram the cephalic stimulation of insulin release is not a transient event only seen at the beginning of a meal but instead it lasts throughout.

In any case, the CPIR should be considered in the context of the whole insulin response to a meal rather than as an isolated event. For example, it has been known for quite some time that glucose given orally results in a greater insulin response than the same amount of glucose given intravenously (30). Of course, such observations do not allow for the localization of the origin or the identification of the mechanism of this phenomenon. In fact, the CPIR represents only the first of three stages of eating in which different stimuli act to influence insulin secretion (FIGURE 1). The second stage is characterized by the incretins glucose-dependent insulinotropic peptide (GIP; formerly gastric inhibitory polypeptide) and glucagon-like peptide-1 (GLP-1) (see Ref. 31). A vago-vagal reflex from the stomach triggered by antral distension may also contribute (32). The third and last stage is represented by the direct effects of glucose and some amino acids (33, 34) on the β-cells. As far as incretins are concerned, it is well known that their potential to stimulate insulin release by themselves is limited. Their major physiological effect is an enhancement of glucose-induced insulin release (31). With respect to a direct effect on pancreatic β-cells under physiological conditions, GIP is probably the more important incretin, whereas endogenous GLP-1 triggers insulin in large part through a vago-vagal reflex. However, GLP-1-based substances, in particular the GLP-1 receptor agonists, are the more useful compounds for the treatment of type II diabetes (31). That incretins do not massively stimulate insulin release per se, but only when glucose is present, makes their receptor agonists therapeutically useful because this feature reduces the risk of producing potentially dangerous hypoglycemia. The conventional CPIR definition neglects the fact that there is a similar interaction between vagal efferent activation and glucose with respect to the stimulation of insulin secretion. That is, the direct stimulatory effect of acetylcholine on insulin secretion is glucose dependent (35). In turn, in isolated pancreatic islets, cholinergic agonists, which have been considered a proxy of vagal efferent activation, sensitize the β-cells for glucose in a sustained way (for ∼45 min) even after the drugs are removed (36), an effect that may well represent the major contribution of vagal stimulation to the overall meal-induced insulin release (23, 29, 36).

1.4. Cephalic Phase Insulin Response: What Is Its Function?

1.4.1. General considerations.

It is often assumed that there are some eating-related physiological mechanisms that evolved to protect us from overeating and, hence, gaining weight. Perhaps the most prominent example of this misguided view is the idea that dietary-induced thermogenesis (DIT) serves as a buffer to compensate for overconsumption (37). This assumption ignores the fact that evolution is driven by natural selection, which favors traits that translate into reproductive success. Given the regular famines that occurred in the human past, DIT would have been eliminated long ago if its only purpose were to waste energy (see Ref. 38 for a thorough discussion). Along these lines of reasoning, and given the broad range of cephalic phase responses, it seems obvious that they serve mainly to prepare the animal for the arrival of food, i.e., for the optimal digestion, absorption, and utilization of ingested nutrients and hence to ensure maximum ingestion capacity with minimum disturbance of homeostasis. This certainly holds also for the CPIR (see below and Refs. 7, 39, 40). For our ancestors, it was certainly advantageous to be able to eat large amounts of food when available because it enhanced chances of survival during the next period of food shortages or starvation that lurked around the next corner. Of course, this function ultimately also requires stopping ingestion when the negative consequences of continued eating outweigh its benefits. A contribution of the CPIR to satiation (meal termination) is therefore feasible. In other words, while some evidence suggests that cephalic responses in general and the CPIR, in particular, are positively correlated to meal size (see Refs. 7, 24, 25, 39, 40), the continuing cephalic stimulation during a meal may eventually also activate short-term satiation signals that limit meal size (see Refs. 7, 24). Although arguably, from an evolutionary point of view, the ability to ingest large meals is an advantage, in the modern food environment large portion sizes and the resulting large meals may promote overeating (e.g., Ref. 41).

1.4.2. What happens without CPIR?

Again, Pavlov was probably the first to demonstrate the substantial effects of eliminating cephalic stimulation on gastrointestinal motility, food digestion, and nutrient absorption (13, 16). For instance, he showed that placing food directly into the stomach produced much less gastric juices than when the same food was normally ingested or even sham fed. Unfortunately, because bypassing the oral cavity has several effects on gastrointestinal (GI) functions that by themselves can affect insulin secretion (13, 16, 42–44), it is difficult to attribute any change in insulin release to the absence of a direct effect of cephalic stimulation on β-cells or indirect effects of altered GI mechanisms. Because of its small size and supposedly short duration, the physiological relevance of the CPIR has often been questioned. Nevertheless, evidence for the physiological relevance of the CPIR is derived from observations indicating that its elimination compromises postprandial glucose handling, i.e., enhances the postprandial increase in blood glucose (26–28, 45), an effect that appears to be primarily due to a reduced suppression of hepatic glucose production (46). Based on sham-feeding studies in animals and modified sham-feeding (MSF) studies in humans (see sect. 1.5 for explanations), the CPIR has traditionally been estimated to be very small (∼1% of the overall insulin response to a meal) (see Ref. 14). However, this considers only the preabsorptive part of the cephalic stimulation, and the picture is completely different when the whole cephalic stimulation of the β-cells during eating is considered (23).

1.4.3. Metabolic effects of the CPIR.

Using a double isotope technique to trace rates of total glucose appearance, total glucose disappearance, intestinal glucose absorption, hepatic glucose production, and the metabolic clearance rate of glucose, Proietto and colleagues (47) showed that oral glucose stimulation improved particularly the efficiency of glucose disposal in the early phase after a glucose load. Moreover, consistent with this outcome, insulin concentrations measured in the jugular vein were lower with oral glucose delivery than when the same amount of glucose was infused into the stomach. Insulin’s major function in the liver is to activate glucokinase and thereby indirectly stimulate glucose uptake (48). The enhanced glucose disposal with oral glucose stimulation is therefore consistent with the assumption that in this situation the early increase in insulin specifically stimulates hepatic glucose uptake (49). This makes perfect physiological sense because it promotes the repletion of hepatic glycogen reserves and efficiently dampens the postprandial increase in systemic blood glucose levels. At the same time, normal chow meals in rats have been shown to be accompanied by brief, transient glycogenolysis and an increase in glucose output (50), similar to what has also been observed after oral stimulation with glucose (51). Sympathetic stimulation as well as a prandial release of glucagon may contribute to this phenomenon (51, 52). This is interesting because one important effect of early phase insulin secretion is to counteract the hyperglycemic effect of glucagon (53). While these studies were performed in rats and dogs, later studies in humans (54) convincingly supported the important role of the early phase insulin response in inhibiting endogenous hepatic glucose production. This is clinically relevant because many studies indicate that the rapid postprandial rise in circulating glucose is a major contributor to the pathogenesis of diabetic complications (see Ref. 55 for a detailed discussion).

A prandial stimulation of hepatic glucose output at a time when insulin supposedly starts to enhance hepatic glucose uptake appears counterintuitive but may simply be an example of the general principle that physiological regulation is usually a shift in equilibria and that it is therefore not unusual, but in fact rather common, for two opposing actions to occur at the same time. Multiple factors may then determine in which direction the balance shifts. Applied to the previous example, this means that the early rise in insulin presumably shifts the balance of hepatic glucose release and uptake toward the latter.

The relevance of the CPIR for postprandial glucose handling is further supported by findings in rats indicating that the insulin increase was delayed when food was injected into the stomach, while the postprandial increase in glucose was enhanced (28). Similarly, in a study in humans (56), ingestion of a mixed meal (sandwich, sugar pie, and soft drink, 755 kcal, duration 10 min) produced a CPIR, whereas infusion of the same amount of calories into the stomach did not and caused postprandial disturbances in carbohydrate metabolism (56). Accordingly, years later, other studies in humans reported that direct gastric administration of glucose in combination with sham feeding significantly lowered postprandial blood glucose levels that were otherwise observed with either intragastric (29) or intravenous (57) glucose administration. In line with this function, a loss of cephalic, or early phase, insulin release is a common feature of patients with impaired glucose tolerance (e.g., Refs. 55, 58, 59). It develops early in the disease and may contribute to the pathogenesis of type II diabetes mellitus (55, 60).

Further evidence for an important physiological function of CPIR is derived from a study in rats showing that 30-min feeding as well as 4-h sham feeding of carbohydrate-rich solid or liquid diets, respectively, increased circulating insulin and lipoprotein lipase activity in adipose tissue, resulting in lower circulating triglyceride and free fatty acid levels compared to the fasted state (61). The increase in insulin occurred before (in the feeding experiment) or without any (in the sham-feeding experiment) increase in blood glucose (61).

All these changes combine to efficiently handle the nutrients derived from digestion and absorption, and in particular the glucose derived from ingested carbohydrates, i.e., to minimize the postprandial increase in circulating glucose by indirectly enhancing its uptake into the liver and by rapidly reducing the secretion of glucagon, which stimulates hepatic fatty acid oxidation, glycogenolysis, and gluconeogenesis. At the same time, insulin reduces adipose tissue lipolysis, thus interrupting the flow of fuel for hepatic fatty acid oxidation and gluconeogenesis. Collectively, these findings indicate that oral sensory stimulation and the resulting CPIR are important for postprandial glucose metabolism and that the mechanisms triggered by this cephalic stimulation are involved in normal glucose homeostasis.

1.5. Cephalic Phase Insulin Response: How Is It Detected?

Because of its short duration and small size, the preabsorptive insulin release is often challenging to measure directly and it is best achieved with sophisticated blood sampling technology. The common baseline fluctuations in insulin that are due to the pulsatility of its release that is particularly evident in the hepatic portal vein (62–65) pose another problem for detection. These methodological limitations may account for some disparities in the literature. For instance, whereas a CPIR is usually observed in animals (mainly rodents), in which a small insulin peak occurs within the first few minutes of oral sensory stimulation and is apparently independent of any nutrient absorption (e.g., Refs. 23, 28, 66–70), the CPIR is more “elusive” in humans. As Pullicin and colleagues (19) noted, this may partly also be due to the lack of consistency in the literature with respect to what constitutes a cephalic phase response and to the fact that often times arbitrary operational definitions have been applied.

Because of the limited methodological options in humans compared to animals, the question of whether a CPIR can be shown in humans has been often addressed using a variety of different techniques: The test subjects either first, expect a meal but are only allowed to see and smell the foods, a procedure occasionally called a “tease meal” or “tease feeding” (71); or second, they are exposed to taste solutions that they have to spit out after a short time period or they may chew and taste real foods before they expectorate them. This technique is also called modified sham feeding (MSF), in analogy to the sham-feeding technique in animals, in which ingested solutions drain through a fistula from the esophagus or stomach without activating intestinal or postabsorptive mechanisms. The MSF technique was first used in relation to studies of gastric acid secretion (72) or thermogenesis (9) and introduced into human CPIR research by Teff and colleagues (73). Third and finally, subjects may be allowed to see, smell, taste, chew, and swallow the food, i.e., to ingest it. In general, because of the technical limitations in human experiments, other parameters are often measured in addition to the recording of circulating insulin concentrations. Thus, glucose tolerance may be assessed as an indirect measure of a CPIR, and plasma concentrations of pancreatic polypeptide (PP) or of C-peptide, as an indicator of vagal efferent activation and of insulin release, respectively, may also be measured (see sects. 5.8 and 5.9, respectively).

1.6. Consistencies, Disparities, and Potential Resolutions—the Purpose of This Review

As mentioned above, cephalic phase responses include GI motility changes as well as exocrine and endocrine secretions, and other physiological phenomena. To deal with all these responses including their special features, their physiological functions, and their ambiguities would be beyond the scope of this review. We therefore focus on the CPIR and refer to other hormones, particularly to C-peptide and PP, only in relation to the CPIR. Clearly, the CPIR is the most investigated cephalic phase response because it may be practically the most relevant to glucose homeostasis. However, the rigor with which the CPIR has been validated is uneven, especially with respect to the sensory triggers and the magnitude of the response. Upon the development of radioimmunoassays for peptide hormones, in the 1960s, a host of data from several laboratories started to emerge, employing different techniques to investigate the CPIR in various species. Most of the original animal studies were performed in rats (e.g., Refs. 66–68, 74, 75), but other mammals, such as dogs (69, 76, 77), lambs and rabbits (78), sheep (79), rhesus macaques (23), and mice (70, 80) have been tested as well. In animals, a CPIR appears to be more consistently observed than in humans, but often a given stimulus might trigger a CPIR in one study but not another. Thus, the phenomenon is rather elusive as the title of this article reflects.

In sum, while there is ample evidence that the CPIR does exist, there are open questions about its scope and properties, as well as its underlying mechanisms, and afferent and efferent neural pathways. The phenomenon is interesting from a basic science perspective as well as from a translational point of view, and there are several recent reviews from different groups that focus on various specific aspects of the CPIR (14, 15, 17, 19, 81–83), but we have chosen to take a more integrative approach to the topic.

We review the current literature on the CPIR, the triggers and conditions under which it takes place, and its neural and physiological mechanisms. Specifically, we discuss 1) the various stimuli that elicit the CPIR and the sensory signaling pathways involved (sensory limb); 2) the efferent pathways that control the various endocrine events associated with eating (motor limb); and 3) what is known about the central integrative processes linking the sensory and motor limbs. When evaluating the pertinent animal and human¹ studies with respect to the reliability and/or magnitude of the CPIR, we discuss problems and open questions with respect to test conditions that allow, or may not allow, the stimulus to reach the potentially relevant taste receptors and to trigger a CPIR. We also touch upon the possible roles of sweetness and palatability as crucial stimulus features, and whether conditioning plays a major role for the CPIR, as it does for other cephalic phase responses. Importantly, as the CPIR, according to its classical definition, reflects an early release of insulin before any nutrient absorption, we question the usefulness of this criterion based on what is known about the effects of vagal motor neuron activation and thereby acetylcholine on the β-cells together with the general difficulties of the accurate assessment of insulin release and the limited reliability of its substitute measures. Finally, we weigh the evidence of the physiological and clinical relevance of the cephalic or neural contribution to the insulin release in response to eating based on the pathophysiological consequences of its absence or elimination. In this context, we consider that 1) cephalic stimulation can have substantial metabolic effects even if it is short and transient (29), and 2) it normally continues throughout the meal (e.g., Ref. 23). These points critically underlie the interpretation of the existing data, and they suggest that a focus on the cephalic contribution to the overall insulin response to a meal in general may be more instructive than a strict reliance on the classical CPIR definition, which is mainly dictated by the quest to isolate cephalic, (i.e., preabsorptive), from postabsorptive sites and mechanisms of stimulation.

2. THE SENSORY LIMB

2.1. Preamble

The concept of the adequate stimulus has been a fundamental principle of sensory psychology and physiology since the days of Johannes Müller (84). It contends that every sensory receptor has a form of energy to which it responds best. Thus, photoreceptors respond best to photons of a certain wavelength; auditory hair cells, depending on their location on the basilar membrane respond best to certain frequencies of sound pressure; taste receptors respond best to compounds with certain chemical properties; and so on. Decades of modern psychophysical and neurophysiological findings have verified this concept. Moreover, it is equally clear that a given receptor can respond to other stimuli for which it is not optimally tuned but less well. In this context, a starting point in understanding the CPIR is to define what the adequate stimulus is and what the relevant head receptors are.

While the evidence supporting glucose as the adequate stimulus is reasonably strong, the ability of other sugars, nonnutritive sweeteners, and complex stimuli to elicit a CPIR is mixed, as is the role of the taste system as a critical sensory trigger.

2.2. The Adequate Stimulus

2.2.1. d-Glucose.

Theoretically speaking, the stimulation of taste receptors by d-glucose, should represent the primary stage of the sensory process by which insulin is preabsorptively released. Arguably, glucose should be the cardinal stimulus because insulin, among its functions, serves to maintain a euglycemic state and facilitates the utilization of this specific monosaccharide as an energy source or for fuel storage. Because glucose has low vapor pressure, it would be poorly effective at stimulating olfactory receptors, and, except in certain experimental conditions, during ingestion this stimulus must contact the sensory receptors of the oral cavity, in which a subset of taste bud cells respond well to the sugar.

The most compelling evidence for the primacy of glucose as the stimulus triggering a CPIR comes from experiments with rodent models. More than four decades ago, using adult female rats fitted with a cardiac catheter, Louis-Sylvestre (67) allowed the animals to drink 1 mL of 50% glucose solution over 1 min and clearly identified two early peaks in plasma insulin (>3 times increase from baseline): one at 2.5 min and the other at 5.0 min after ingestion, with both preceding any rise in blood glucose. The intragastric infusion of the same amount of 50% glucose solution produced only one early insulin peak of approximately the same magnitude but somewhat later (around 3.5 min after infusion onset) than after oral ingestion. Subdiaphragmatic vagotomy eliminated both insulin peaks confirming that they were neurally mediated and not due to direct stimulation of the β-cells by glucose. The two peaks were interpreted as representing an oral and enteric origin of stimulation respectively. Siegel et al. (85) presented glucose (40% wt/vol, 1 g/kg body weight) to control rats and rats with pancreas transplants and then sampled blood from the jugular vein at various time points after the start of drinking. In contrast to the rats with transplants, the control animals displayed an early rise in plasma insulin at 2 min that preceded any significant rise in blood glucose. Together, these findings provided a nice demonstration that glucose was an effective stimulus. However, was it a specific one?

The issue of stimulus specificity was thoroughly addressed by Grill et al. (74) who infused small volumes of various sugar solutions, sugar alcohols, and sodium saccharin into the oral cavity of rats. Although glucose was not the most preferred stimulus (e.g., Ref. 86), they found that it was the only one that caused a significant rise in plasma insulin levels (measured from the vena cava or jugular vein) within the first minute of oral delivery without any increase in plasma glucose and did so even when the orally ingested fluid drained from an open gastric cannula. This result was consistent with the results in an earlier study in rats showing that oral glucose was superior to sucrose or a glucose/fructose mixture in generating a rise in plasma insulin despite that all these sugars led to similar levels of plasma glucose (87). Other investigators also reported a CPIR in response to high-glucose concentrations. In one of these studies (88), maltose and fructose were also effective stimuli, and in three others (51, 68, 89), saccharin was. The issue of stimulus specificity is explored further below.

Several decades later, Glendinning and his colleagues (70) trained mice to lick a small volume of a stimulus, adjusted for body weight, from a drinking spout within 3 min; on different occasions, the same volume was delivered by gavage bypassing the oral cavity. Oral ingestion of high concentrations of glucose caused a significant rise in insulin within 5 min whereas gastrically delivered glucose did not. Oral delivery of isomolar sucrose, but not fructose, was also effective. Of course, sucrose is a disaccharide of both a glucose and fructose moiety. Follow-up experiments (80) showed that only glucose and saccharides that had glucose moieties, including sucrose, maltose, and maltodextrin, were effective at eliciting a CPIR; fructose, α-methyl-d-glucopyranoside (α-MDG; a nonmetabolizable form of glucose that binds to SGLT1; more on this below), and several nonnutritive sweeteners were all ineffective. Glucose was the most effective, then sucrose, then maltose, then maltodextrin. When the latter three stimuli were mixed with acarbose, an inhibitor of the enzymatic activity of amylases (released from exocrine pancreas and salivary glands) and α-glucosidases (found in intestinal lumen and apical membranes of taste receptor cells; see below), it significantly suppressed the CPIR. The outcome of the acarbose manipulation does not necessarily prove that the trigger of the insulin release is oral, but it does show that free glucose is necessary. One interpretive limitation to the findings above is that blood glucose was above baseline when the plasma insulin measurements were taken, making it difficult to ascertain how much of the response was due to oral versus pancreatic stimulation by glucose. Nevertheless, the outcome is consistent with other studies using glucose as an oral stimulus (51, 67, 68, 74, 87, 89).

The rodent model results demonstrating that glucose is an effective stimulus are complemented by findings in the dog. In fact, Fischer et al. (69, 76, 77, 90) were probably the first to report a CPIR in animals. In a series of experiments in adult male and female dogs, a highly concentrated glucose solution infused into the oral cavity caused an early rise in insulin levels measured in the hepatic portal vein or in the posthepatic venous circulation before a detectable increase in blood glucose levels (69, 76). Consistent with what Louis-Sylvestre described for rats (67), the early increase in insulin featured two peaks and was independent of the amount of orally administered glucose. The experimenters could block the first and attenuate the second peak by oral mucosal anesthesia, indicating that they resulted from stimulation of local nerve endings (76). Moreover, in follow-up studies, the insulin rise was abolished after bilateral truncal vagotomy (90) and diminished after complete (parasympathetic and sympathetic) pancreatic denervation (91). The application of sodium cyclamate, a nonnutritive sweetener, had no effect on insulin release, conferring some degree of chemospecificity to the effect (76). The early insulin peaks occurred also in dogs with esophageal or gastric fistulas, i.e., when the orally delivered glucose did not reach the stomach or the intestine, and when the glucose was infused into the esophageal fistula, the first insulin peak did not occur (77). Together, these findings in essence exclude any pre- or postabsorptive stimulation of insulin release from the gastrointestinal tract and indicate that there is a CPIR in dogs that is triggered by oral sensory stimulation and mediated by the vagus.

In humans, Yamazaki and Sakaguchi (92) placed a gauze saturated with different concentrations of d-glucose ranging from 277 to 2,220 mM and containing a mixture 36% α- and 64% β-anomers on the tip of the tongue for 1 min and found a concentration-dependent rise in plasma insulin levels at 3 min poststimulation. In a subset of tested participants, this was blocked by atropine treatment (10 µg/kg) administered 45 min before blood sampling. When these researchers tested the relatively pure α- and β-anomers separately at 1.1 M, the β-anomer was less effective than the α or the equilibrated mutarotated mixture, which themselves did not differ, despite that the participants did not report a difference in sweetness in the glucose anomers. The fact that glucose stimulation of the anterior tongue, which accounts for roughly less than a fifth of the total taste buds in the oral cavity, was effective is noteworthy. With this in mind, it is important to also note that the 277-mM concentration was not an effective trigger of insulin release. Which taste bud fields are stimulated, and the concentration of pure sugar stimuli used, are critical parameters to consider when evaluating outcomes across studies on CPIRs. We are aware of only two explicit negative findings with glucose as the stimulus (93, 94), in which either a 2-min oral rinse (0.9 M) with glucose solution (94) or its ingestion (300 mL, 25% wt/vol glucose) (93) failed to elicit a significant CPIR in humans. In the latter case, however, both plasma insulin and blood glucose did not appear to rise until ∼25 min after the start of the ingestion of the concentrated glucose solution, which is somewhat uncharacteristic of glucose tolerance tests. Moreover, in this study (93), the 10 min blood sampling intervals may have prevented the detection of a CPIR. In contrast, Mandel and Breslin (95) presented participants with 50 mL of a 10% glucose solution consumed over 20 min and found a significant increase in plasma insulin at 9 min before an increase in blood glucose. Finally, Smeets and colleagues (96) reported that in healthy normal-weight men drinking 300-mL solutions of 75 g glucose, 880 mg aspartame, or 78.9 g maltodextrin or water, only glucose caused an early (at 5 min after the onset of drinking) increase in circulating insulin before an increase in blood glucose. As will be shown in this article, of all the stimuli tested, and despite some negative findings in humans (see above), pure glucose is the one that leads to the most consistent positive outcomes.

2.2.2. Other mono- and disaccharides.

Some of the work described above supporting the primacy of glucose in eliciting a CPIR showed that other mono- and disaccharides generated weaker, if any, effects and only if they were glucose-containing. That said, others have found effects in response to sugars other than pure glucose. Berridge et al. (88) reported that a 1.0-min intraoral infusion (1.0 mL) of 15% maltose and 15% fructose could generate a CPIR in rats. Tonosaki et al. (97) found an increase in plasma insulin levels in rats 3 min after a 45-s infusion of 1.0 M sucrose that preceded a significant rise in blood glucose and was eliminated by bilateral transection of the chorda tympani nerve (CT) innervating the taste buds of the anterior tongue. These data are somewhat difficult to reconcile with the inability of similar intraoral infusions of 0.8 M fructose and maltose to elicit a reliable cephalic phase response in rats (74) and the inability of 1.0 M fructose licked by mice to have an effect (70, 80). Tonosaki et al. (97) also tested starch and found it to be an ineffective stimulus. Thus, if any hydrolysis was occurring in the oral cavity to produce glucose or maltose as a function of the actions of amylase and disaccharidases (see Refs. 80, 98), it was at an insufficient level to trigger a reliable response.

In humans, Just et al. (99) found that the application of taste strips saturated with sucrose on the tongue followed by closing of the mouth led to a significant CPIR. On the one hand, it is possible that disaccharidases on the apical membrane of taste receptor cells hydrolyzed the sucrose to produce a source of glucose (see sect. 2.3.3), but, on the other hand, saccharin was equally effective at eliciting the response. Duskova et al. (100) reported a small but significant rise in insulin after a mouth rinse with 5% sucrose. Other groups did not observe a CPIR to sucrose (20, 101, 102). Dhillon et al. (17) separated participants with overweight or obesity into cephalic phase insulin responders and nonresponders and then tested them with either 8.8% sucrose in a beverage (59.2 mL) or 30.3% sucrose in a gelatin cube (16.4 mL) that was presented orally for 14 min in a MSF protocol with a crossover design. The responders displayed a greater insulin response to the sucrose presented in solid relative to the liquid form, but both forms were effective. The authors apparently tried to match the amount of sucrose delivered to the oral cavity (∼5 g), but it is difficult to disambiguate the effect of sucrose concentration from the phase of the matter in which the stimulus was presented, the latter engenders a different, but overlapping, set of oromotor actions.

2.2.3. Nonnutritive sweeteners.

Nonnutritive sweeteners have the experimental appeal of being sweet to humans and being treated by rodents as bearing strong perceptual and affective similarity with sucrose (e.g., Refs. 103–106) but have no calories and do not raise blood glucose. Of course, to the extent that such compounds are absorbed, it remains possible that they could stimulate the pancreas directly given that the T1R2 and T1R3 subunits of the cardinal taste receptor that binds to sweeteners have been found to be expressed in pancreatic β-cells (107–111). An additional interpretive complication is that, although these compounds bind with the T1R2 + T1R3 heterodimeric taste receptor, they also have sideband tastes and, at least some, are known to bind to certain members of the T2R family of taste receptors that are activated by compounds reported as “bitter” by humans (112). Thus, the perceptual experience they evoke is not identical to pure sugars. Implicit in the strategy to use such stimuli to investigate CPIR is that the perception of “sweetness”² is key, but this may not be the case (more on this below).

Early studies in rat (113, 114) and rabbit (115) demonstrated that saccharin consumption exacerbated the deleterious physiological effects of high-dose insulin administration suggesting that this nonnutritive sweetener could cause hypoglycemia potentially via pancreatic insulin release. As noted above, Berthoud et al. (68) and Powley and Berthoud (116) demonstrated that the oral delivery of a small amount of sodium saccharin (0.15%) was similarly effective as 50% glucose at eliciting an early insulin response in adult male rats. Likewise, Ionescu et al. (51) were able to demonstrate a significant rise of insulin in rats 1–2 min after ingestion of a small volume (1 mL) of saccharin (0.15%) or glucose (70%). Ninomiya et al. (89) reported that, similar to 50% glucose, sodium saccharin also triggered an insulin response and did so in a concentration-dependent manner (0.0125–0.125%); the same group later replicated the CPIR to 0.125% saccharin (117). Tonosaki et al. (97) also reported that, in addition to sucrose, saccharin was an effective stimulus for a CPIR. These reports stand out from the animal literature reviewed above because they challenge the necessity of a glucose moiety in the oral stimulus, but not all studies have found that nonnutritive sweeteners are capable of orally triggering an insulin release (74, 76, 80, 118). It remains unclear whether species or strain differences account for some of the disparities in the effectiveness of nonnutritive sweeteners to trigger CPIR, as Berthoud and Powley (118) have speculated, or whether other unidentified factors (see also Ref. 80) contribute. We discuss the potential involvement of learning on CPIR in sect. 5.7.

In humans, some studies have demonstrated that nonnutritive sweeteners are effective stimuli (99, 119), but the extensive opportunity in daily life to associate the “sweet” taste of these stimuli with postabsorptive rises in blood glucose somewhat mitigates the significance of such findings. Accordingly, it is perhaps more noteworthy that not all studies do find a CPIR to nonnutritive sweeteners (20, 101, 120–122). Dhillon et al. (17), who, as described above, separated participants with overweight or obesity into cephalic phase insulin responders and nonresponders, found that the responders not only displayed insulin release to orally restricted presentations of sucrose, but also of sucralose, in both solid and liquid form. Bruce et al. (71) did not find a significant rise in serum insulin levels to chewing gum adulterated with the nonnutritive sweetener aspartame; however, they did find that the aspartame was effective when combined with the sight and smell of a palatable breakfast meal. In this regard, it is noteworthy that an aspartame- and fat-containing strawberry-flavored gelatin mousse (200 kcal) that was swallowed by men with normal weight or obesity, led to a significant rise in plasma insulin levels peaking at 4–5 min postingestion without any observable rise in blood glucose (18, 26).

In summary, it seems possible that the capricious nature of a nonnutritive sweetener-stimulated CPIR may depend, in part, on some features of the experimental design that may facilitate associative learning, at least in some cases (see sect. 5.7 for a more thorough discussion of this issue).

2.2.4. Complex oral stimuli.

In pioneering experiments by Steffens (28) and Strubbe and colleagues (66, 123), complex diets rich in carbohydrates were capable of eliciting an early rise in plasma insulin levels before any increase in blood glucose upon initiation of a meal in male rats equipped with cardiac catheters. Still, there was certainly a significant amount of glucose and oligosaccharides contained in the oral stimulus. They also observed the early increase in insulin when they offered carbohydrate-free food or even “food” without any caloric value. However, because these latter diets were tested at the end of the series of other experimental manipulations, it remains possible that the outcome was influenced by this prior experience. In a different study, an evaporated milk liquid diet containing 40% carbohydrate and adulterated with sodium saccharin (0.06%) was sufficient to promote a substantial cephalic phase insulin response in rats that were sham fed (124). Later, Berthoud and Powley (118) reported a clear and reliable CPIR also in rats ingesting regular chow; this response was modulated by selective transections of vagal branches (see sect. 3.3.5). At approximately the same time as the early studies in rats were done, Fischer and colleagues (90, 91, 125) showed in dogs that not only an oral glucose load, but also real or sham ingestion of mashed meat caused a biphasic early insulin release that occurred before any increase in circulating amino acid concentrations. This was abolished by bilateral truncal vagotomy, and diminished by pancreatic denervation. In sheep, a CPIR was observed as early as 1–2 min after the start of the presentation of complex high fiber feed (Lucerne pellets) without a rise in blood glucose, which was attenuated by vagotomy (79) (also see sect. 3.3.5). In rhesus macaques, a mixed meal (fruit and monkey chow) triggered a CPIR as early as 2 min after meal onset whereas plasma glucose did not increase before 10 min (23). Recently, Wiedemann et al. (126) reported a CPIR in mice immediately after a single bite of a chow pellet with no concomitant rise in blood glucose. This CPIR was shown to be mediated in large part by IL-1β release from microglia (see also sect. 4.4). Because the chow was familiar, it is unclear whether this represents an unconditioned or conditioned response to the oral sensory properties of the food. Nevertheless, collectively, all these studies show that, in animal models, oral contact with complex stimuli can trigger a preabsorptive insulin release.

In humans, the profile of results is more heterogeneous. Bellisle et al. (127, 128) detected a significant rise in insulin beginning within a few minutes after the start of ingestion of small cocktail sandwiches of various foods that preceded any detectable increase in blood glucose concentration. The latter is strong support for a preabsorptive trigger, but it is difficult to distinguish between an oral, gastric, and enteric site. It is also noteworthy that Bellisle and colleagues (127) reported large variability among and within subjects, in agreement with the view that there are possibly subsets of individuals who are more or less responsive.

Perhaps the most comprehensive series of studies on CPIR in humans, especially with regard to complex oral stimuli, was conducted by Teff and her colleagues. In one study (73), the MSF technique was employed in men to isolate oral sensory stimulation. After an overnight fast, participants took bites of a peanut butter sandwich, chewed without swallowing as long as it felt comfortable, and then expectorated the food. This was repeated for a total of 2 min. In addition to the MSF condition, on other occasions they were allowed to eat the sandwich. Very similar early (0–10 min, with a peak around 4 min) increases in insulin and C-peptide levels occurred in both the MSF and fed conditions. Glucose, norepinephrine, epinephrine, and free fatty acids did not change from baseline levels during this time. Teff et al. (101) also found that, in contrast to when 13.7% sucrose or a variety of nonnutritive sweeteners were tested, MSF of apple pie produced a reliable increase in venous insulin levels within 3 min after oral stimulation in men.

In another study, Teff and colleagues (26) found that individuals with obesity who ingested an aspartame-sweetened, strawberry-flavored gelatin mousse enriched with dairy fat exhibited greater CPIRs than normal-weight subjects when the increases from baseline were expressed in absolute (difference) as opposed to relative (percent) terms. This suggests that the more pronounced CPIRs in the obese state that have occasionally been reported under various experimental conditions (11, 12, 51) may simply reflect the usually higher baseline insulin levels under this condition (26). They also observed that the CPIR and the postprandial increase in circulating insulin were positively correlated in normal-weight but not in obese subjects, suggesting that in normal-weight healthy people the neurally mediated cephalic phase and the postprandial pancreatic stimulation are consistent. Other groups reported a CPIR in healthy men ingesting a high glycemic index meal of chicken cream soup and biscuit (supplemented with 50 g sucrose in a 200-mL aqueous solution) (129) or a sugar pie (325 kcal, 51.7 g carbohydrates) (130).

In a series of studies, LeBlanc and his group showed that high-protein meals (250 g beef steak or 195 g beef steak compared with 250 g cod fillet) with negligible amounts of carbohydrates (130–132) or pure whipped cream (325 kcal, 35.6 g fat) (130), triggered a CPIR, indicating that proteins and even fats can serve as a stimulus for CPIR. In an earlier study by the same group (56), ingestion of a mixed meal (sandwich, sugar pie, and soft drink, 755 kcal, duration 10 min) produced a significant early (2 min after meal onset) increase in circulating insulin before the increase in blood glucose, which in fact decreased for the first 4 minutes. When the same amount of calories as ingested in the meal was tube fed into the stomach, no early increase in insulin was observed, and later (between 15 and 90 min) insulin increased higher and remained elevated longer than after eating the meal (56). Moreover, throughout this time, the resting metabolic rate and the respiratory exchange ratio were higher after meal ingestion than after stomach tube feeding (56), indicating that the oral stimulation had a substantial and lasting effect on postprandial carbohydrate metabolism. A CPIR was also observed after ingestion of a chocolate drink sweetened with sucrose or aspartame (119). Other authors reported a CPIR in response to a high-fat breakfast administered in capsules and followed immediately by a 2-h MSF protocol with frequent 10-s exposures to low- and high-fat cream cheese, with the latter one producing a much bigger insulin response (133). Another study (57) in men receiving intravenous glucose infusions did not find a CPIR or any effect on the area under the curve (AUC) of insulin or C-peptide after MSF of a meat and sauce stimulus. The design of the experiment, however, may have obscured a CPIR because blood glucose level was clamped at 8 mmol/L, and it is possible that the direct stimulation of the pancreatic β-cells by the infused glucose overpowered and, hence, masked any neurally mediated early rise in insulin. The authors did observe that the sham feeding under these conditions improved the glucose tolerance and speculated that this was due to a reduced hepatic glucose output because their insulin and C-peptide data supposedly did not permit any other conclusion.

Other studies in the literature are at odds with one another, even some conducted in the same laboratory. When men sampled foods from a palatable assortment of breakfast items in a MSF preparation, there was no evidence of a CPIR (nor for glucagon, GLP-1, GIP, or ghrelin) despite that glucose moieties were constituents in these complex stimuli. Oral sampling, however, led to an increase in venous levels of PP, which is released by the gamma cells of the pancreatic islets, suggesting cephalic activation of vagal motor fibers but without causing insulin release (134). This is in stark contrast to an earlier study from the same group in which 10 min after women ingested a breakfast meal, insulin levels significantly increased in plasma, whereas glucose levels did not until 15 min postingestion (21). In this study, the CPIR was partially blocked by concomitant treatment with trimethaphan, an autonomic preganglionic synaptic antagonist, or atropine, a parasympathetic cholinergic antagonist, implicating parasympathetic and perhaps sympathetic mechanisms as being involved in the motor limb of the reflex (see sect. 3.3).

Bello et al. (135) were unable to see any increase in insulin secretion triggered by oral sampling of chocolate half-and-half or a nonfat equivalent. In another study that used foods of different macronutrient composition in men, only the carbohydrate-rich stimulus (frosted cereal) caused a small but significant early insulin response at the end of the 3 min MSF session (136), suggesting that the CPIR under these conditions is specific for carbohydrates. Eliasson and colleagues (137) found an early rise in insulin occurring around 3–5 min after the onset of eating a vanilla-caramel apple muffin, whereas blood glucose levels started to increase later. However, such a simple distinction in the effectiveness of an oral stimulus to trigger a CPIR based on its macronutrient intake is belied by other findings in the literature (138–140). For example, Crystal and Teff (138) sampled blood every 2 min and found no CPIR after the onset of MSF (3 min) of either a high-fat or high-carbohydrate cake in normal-weight women classified either as nonrestrained or restrained eaters. Using a MSF protocol, Lasschuijt et al. (141) were also unable to demonstrate a CPIR to a complex gel diet that varied in sucrose concentration. They did, however, find some evidence of cephalic phase insulin responders and nonresponders to some of the stimuli.

Interestingly, in another study, patients with kidney-pancreas transplantation did not display a CPIR to a palatable pizza meal, but controls did; a finding that implies sufficient innervation of the islets is necessary to observe the response (142) (see sect. 3.3). Robertson et al. (143) also used a pizza meal coupled with a full-fat milk and cream drink and reported that either MSF or ingestion produced a poststimulation rise in plasma insulin. Strangely enough, the MSF condition, but not the consumption, appeared to trigger an early rise in plasma insulin, but this time period was not explicitly analyzed in the study.

Broberg and Bernstein (144) compared blood glucose and plasma insulin levels in response to first, the presentation, and then 10 min later, the consumption, of a cinnamon roll in bulimic patients and healthy but dieting women. The authors saw an increase in plasma insulin before an increase in blood glucose when the participants ingested the cinnamon roll but not in response to the mere presence of it (144). They observed no difference between bulimic and healthy women, indicating that bulimic women do not have abnormal patterns of insulin release. Using a similar design, the same authors (145) found a CPIR in women with anorexia nervosa but not in normal-weight healthy controls, despite the fact that the former ate much less of this cinnamon roll than the latter.

Morey et al. (146) tested tomato soup composed primarily of carbohydrate that was either ingested or intragastrically infused in men. This resulted in different dynamics of the increases in insulin, whereas blood glucose levels were not different between the two conditions and started to rise after the plasma insulin level. In this study, the characteristic early peak in insulin around 3–5 min that was detected in other complex meal studies (e.g., Refs. 18, 21, 73, 127, 128, 137) was not apparent. Rather, the insulin AUC for the first 15 min was greater after oral ingestion of the soup than after intragastric infusion (146). Nevertheless, the very consistent early and transient increase in circulating insulin among studies that find a CPIR emphasizes the importance of the timing of blood draws. Given the inter- and even intraindividual differences, sampling blood at 5 min or more after stimulus exposure increases the risk of missing the early insulin peak, which may explain some of the negative findings in studies of CPIR (147–149). On the other hand, this cannot easily explain some other failures to observe a CPIR to complex stimuli that have met this methodological criterion (e.g., Refs. 134, 135, 138, 150).

2.2.5. Effectiveness of nonoral stimulation.

There is evidence in humans that the mere presentation of food without oral sampling can lead to a CPIR suggesting that visual and olfactory stimuli have some degree of effectiveness. The presumably first documentation of an increase in circulating insulin in response to only a visual and olfactory meal stimulus was published in 1971 (45). The authors studied the effects of such a “tease meal” on plasma insulin, free fatty acids, and glucose concentrations in six nondiabetic adolescents of both sexes with obesity. They observed an increase in plasma insulin and a decrease in plasma free fatty acids without a change in blood glucose levels. However, the peak insulin concentrations occurred 15–20 min after the stimulus (i.e., much later than might be expected for a CPIR) and the increases also lasted longer. It is therefore unclear whether the observed changes were really a CPIR or whether some other mechanism was involved. In an even earlier study, an increase in plasma insulin was observed in three out of seven subjects in relation to hunger or an imaginary food consumption in hypnosis (151), also not a real CPIR, but suggesting that insulin release can be stimulated in some individuals by neural signals independent of real ingestion.

Sjostrom and colleagues (12) reported an increase in plasma insulin calculated as the AUC from 1–10 or 1–20 min after exposing women with obesity and normal-weight controls fasted for 16 hours to the sight and smell of food. The response was more pronounced in the former compared with the latter group and in general variable. Interestingly, atropine blocked the insulin response completely, implicating the involvement of the vagus (12). Both features appear to be consistent with what has been reported for the CPIR in rats (67, 118, 124), dogs (90), and sheep (79). In another study (11), the visual and olfactory presentation of a meal also produced very similar results, i.e., a small and variable increase in insulin between 3 and 9 min after meal presentation with no changes in blood glucose, and a more pronounced effect in individuals with obesity.

Other investigators (10) measured insulin, glucose, glucagon, and salivation in response to a tease meal in 14 women and men fasted for 17 h and reported an increase in insulin as well as increased salivation. The former was relatively small, peaked (∼20% above baseline) around 6 min after stimulus presentation, and occurred in 11 out of the 14 subjects. Plasma levels of glucose and glucagon did not change throughout the measurement period.

One tease meal study specifically addressed possible changes in the magnitude of the CPIR after weight loss in 20 patients with obesity (152). The authors found substantial interindividual differences, which did not differ and correlated (r = 0.61) before and after weight loss. They categorized the subjects into positive responders (n = 8), intermediate responders (n = 5), and nonresponders (n = 7). It is possible that this was simply parceling out a uniform distribution with a central tendency close to zero change, particularly because even in the positive responders the increases from baseline were less than 10%. That said, Teff et al. (18) found that when participants were tested three times over a 5-day period, the correlation between the level of insulin release elicited by an aspartame-sweetened fat-containing strawberry mousse was highly reliable (r = 0.83). Collectively these data suggest that individual differences in the extent of a CPIR are consistent across sessions, lending credence to the view that some humans are, for whatever reason, inherently more responsive than others.

Some other tease meal studies performed in normal-weight and/or individuals with obesity also found evidence for a CPIR (153, 154). Although animal models and several human studies reported more pronounced insulin responses as a function of obesity (e.g., Refs. 11, 12, 51, but see Ref. 26 and sect. 2.2.4), one study observed the opposite, i.e., whereas normal-weight controls showed a CPIR in response to visual and olfactory stimulation, participants with obesity did not (154). While the outcome of this latter study stands out from the others and may simply reflect the volatile nature of the CPIR in humans, it may also be related to the experimental design, in which measurements stopped 4 min after stimulus presentation, such that part of the response might have been missed. As the increases in normal-weight subjects were relatively small, although statistically significant, they may simply reflect the pulsatile nature of the insulin release (20). Interestingly, however, in yet another tease meal study (71), the combined visual and olfactory stimuli failed to trigger insulin release, but the addition of a sweet taste (aspartame) stimulus, which alone was also ineffective, did lead to a CPIR.

2.2.6. Coda: the adequate stimulus.

Taken together, the evidence supporting the effectiveness of complex food stimuli, as well as the ability of all stimuli other than glucose, whether tasted or not, to trigger a CPIR is mixed. One does get the sense that under the correct conditions, insulin release can be observed to nonglucose stimuli. There is also the possibility that in some cases such release might represent a conditioned response depending on the experimental design (for more discussion, see sect. 5.7). In humans, most, but not all, studies use the MSF preparation to obviate the interpretive complication of postabsorptive increases in blood glucose that could stimulate the pancreatic β-cells directly. In these studies, it is questionable how much the posterior tongue taste receptors are stimulated. Thus, a portion of the signal arising from the peripheral gustatory system might be weak or absent. In addition, the lack of the characteristic sensory stimulation involved in the act of swallowing, arising from the muscles and somatosensory afferents associated with the pharynx and esophagus, might blunt responsiveness; this could even be more of a factor for conditioned responses because without swallowing, the predictive value of restricted stimulation of the anterior oral cavity to signal an imminent increase in blood glucose could be compromised. Accordingly, in studies involving humans, it appears that one factor underlying the presence or absence of a CPIR to a stimulus is whether it is swallowed. The critical oral stimulation features that lead to a CPIR have yet to be identified, with one exception: glucose consistently works. TABLE 1 provides a summary of the stimulus and test conditions that have been effective at producing a CPIR in various species.

Table 1.

Studies that reported a cephalic phase insulin response to the listed stimulus

Authors (Ref. No.)	Year	Species/Strain	Simple vs. Complex Stimulus	Stimulus	Stimulus Concentration	Stimulus Duration	Swallowed?
Fischer et al. (69)	1972	Dog	Simple	Glucose	∼48% (wt/vol)	2 min	Yes
Fischer et al. (76)	1972	Dog	Simple	Glucose	∼48% (wt/vol)	2 min	Yes
Fisher et al. (125)	1976	Dog	Simple	Glucose	∼48% (wt/vol)	2 min	Yes
Hommel et al. (77)	1972	Dog	Simple	Glucose	∼44% (wt/vol)	2 min	Yes
Fisher et al. (90)	1976	Dog	Simple	Glucose	∼48% (wt/vol)	2.6 min	Yes
Louis-Sylvestre (67)	1976	Rat	Simple	Glucose	50% (wt/vol)	1 min	Yes
Berthoud et al. (68)	1980	Rat	Simple	Glucose	50% (wt/vol)	∼1 min	Yes
Siegel et al. (85)	1980	Rat	Simple	Glucose	40% 9wt/vol)	URc	Yes
Berridge et al. (88)	1981	Rat	Simple	Glucose	50% (wt/vol)	1 min	Yes
Hara and Saito (87)	1981	Rat	Simple	Glucose	33% (wt/vol)	3 min	Yes
Fischer et al. (91)	1982	Dog	Simple	Glucose	∼48% (wt/vol)	2 min	Yes
Grill et al. (74)	1984	Rat	Simple	Glucose	0.8 M	1 min	Yes
Grill et al. (74)	1984	Rat	Simple	Glucose	0.8 M	1 min	Yesa
Yamazaki and Sakaguchi (92)	1986	Humans	Simple	Glucose	≥555 mM	1 min	No
Ionescu et al. (51)	1988	Rat	Simple	Glucose	70% (wt/vol)	0.5–1 min	Yes
Ninomiya et al. (89)	1988	Rat	Simple	Glucose	70% (wt/vol)	1 min	Yes
Ninomiya et al. (117)	1989	Rat	Simple	Glucose	50% (wt/vol)	1 min	Yes
Smeets et al. (96)	2005	Humans	Simple	Glucose	25% (wt/vol)	2 min	Yes
Mandel and Breslin (95)	2012	Humans	Simple	Glucose	10% (wt/vol)	20 min	Yes
Glendinning et al. (70)	2015	B6 mice	Simple	Glucose	2.8 M	≤3 min	Yes
Glendinning et al. (70)	2015	B6 mice	Simple	Glucose	1.0 M	≤3 min	Yes
Glendinning et al. (70)	2015	T1R3 KO mice	Simple	Glucose	1.0 M	≤3 min	Yes
Glendinning et al. (80)	2017	B6 mice	Simple	Glucose	1.0 M	≤3 min	Yes
Glendinning et al. (80)	2017	B6 mice	Simple	Glucose	2.8 M	≤3 min	Yes
Glendinning et al. (80)	2017	CALHM1 KO mice	Simple	Glucose	2.8 M	≤3 min	Yes
Glendinning et al. (80)	2017	P2X2/X3 KO mice	Simple	Glucose	2.8 M	≤3 min	Yes
Glendinning et al. (80)	2017	SGLT1 KO mice	Simple	Glucose	1.0 M	≤3 min	Yes
Berridge et al. (88)	1981	Rat	Simple	Fructose	15% (wt/vol)	1 min	Yes
Berridge et al. (88)	1981	Rat	Simple	Maltose	15% (wt/vol)	1 min	Yes
Glendinning et al. (80)	2017	B6 mice	Simple	Maltose	0.6 M	≤3 min	Yes
Glendinning et al. (80)	2017	B6 mice	Simple	Polycose	30% (wt/vol)	≤3 min	Yes
Berthoud et al. (68)	1980	Rat	Simple	Saccharin	0.15%	∼1 min	Yes
Powley and Berthoud (116)	1985	Rat	Simple	Saccharin	0.15% (wt/vol)	∼2 min	Yes
Powley and Berthoud (116)	1985	Rat	Simple	Saccharin	0.15% (wt/vol)	∼10 min	Yesa
Ionescu et al. (51)	1988	Rat	Simple	Saccharin	0.15% (wt/vol)	0.5–1 min	Yes
Ninomiya et al. (89)	1988	Rat	Simple	Saccharin	0.125% (wt/vol)	1 min	Yes
Ninomiya et al. (117)	1989	Rat	Simple	Saccharin	0.125% (wt/vol)	1 min	Yes
Tonosaki et al. (97)	2007	Rat	Simple	Saccharin	0.01 M	0.75 min	Yes
Just et al. (99)	2008	Humans	Simple	Saccharin	0.01 M	0.75 min	No
Dhillon et al. (17)	2017	Humans	Simple	Sucralose	0.013% (wt/vol)	14 min	No
Dhillon et al. (17)	2017	Humans	Simple	Sucralose	0.066% (wt/vol)	14 min	No
Tonosaki et al. (97)	2007	Rat	Simple	Sucrose	1.0 M	0.75 min	Yes
Just et al. (99)	2008	Humans	Simple	Sucrose	1.0 M	0.75 min	No
Duskova et al. (100)	2013	Humans	Simple	Sucrose	5% (wt/vol)	URc	No
Glendinning et al. (70)	2015	B6 mice	Simple	Sucrose	1.0 M	≤3 min	Yes
Glendinning et al. (70)	2015	T1R3 KO mice	Simple	Sucrose	1.0 M	≤3 min	Yes
Glendinning et al. (80)	2017	B6 mice	Simple	Sucrose	1.0 M	≤3 min	Yes
Dhillon et al. (17)	2017	Humans	Simple	Sucrose (beverage)	8.8% (wt/vol)	14 min	No
Dhillon et al. (17)	2017	Humans	Simple	Sucrose (gelatin cube)	30.3% (wt/vol)	14 min	No
Berthoud et al. (68)	1980	Rat	Simple	Tap water	50% (wt/vol)	1 min	Yes
Teff et al. (101)	1995	Humans	Complex	Apple pie	N/A	1 and 3 min	No
Teff et al. (18)	1991	Humans	Complex	Aspartame and strawberry-flavored gelatin with dairy fat	N/A	2 min	Yes
Teff et al. (26)	1993	Humans	Complex	Aspartame and strawberry-flavored gelatin with dairy fat	N/A	2 min	Yes
LeBlanc et al. (132)	1998	Humans	Complex	Beef steak (250 g)	N/A	URc	Yes
Soucy and LeBlanc (131)	1999	Humans	Complex	Beef steak or cod fillet	N/A	URc	Yes
LeBlanc et al. (130)	1996	Humans	Complex	Beef steak or cod fillet or sugar pie or whipped cream, all 325 kcal	N/A	URc	Yes
Strubbe and Steffens (66)	1975	Rat	Complex	Carb-free diet	N/A	∼3–5 min	Yes
Steffens (28)	1976	Rat	Complex	Carb-rich chow	N/A	URc	Yes
Picard et al. (61)	1999	Rat	Complex	Carb-rich solid diet	N/A	30 min	Yes
Picard et al. (61)	1999	Rat	Complex	Carb-rich liquid diet	N/A	4 ha	Yes
Sakurai et al. (129)	2012	Humans	Complex	Chicken cream soup, biscuit added sucrose	N/A	URc	Yes
Melchior et al. (119)	1991	Humans	Complex	Chocolate drink with sucrose or aspartame	N/A	≤ 5 min	Yes
Wiedemann et al. (126)	2022	Mice	Complex	Chow	N/A	65 s	Yes
Broberg and Bernstein (anorexic) (145)	1989	Humansb	Complex	Cinnamon rolls	N/A	URc	Yes
Bellisle et al. (127)	1983	Humans	Complex	Cocktail sandwiches	N/A	∼8–12 min	Yes
Bellisle et al. (128)	1985	Humans	Complex	Cocktail sandwiches	N/A	≥16 min	Yes
Broberg and Bernstein (144)	1989	Humans	Complex	Cinnamon roll	N/A	URc	Yes
Mandel and Breslin (95)	2012	Humanse	Complex	Corn starch	10% (wt/vol)	20 min	Yes
Chavez-Jauregui et al. (133)	2010	Humans	Complex	Cream cheese, low and high fat	N/A	Multiple 10 s	Yes
D’Alesssio et al. (23)	2001	Rhesus macaques	Complex	Fruit and monkey chow	N/A	∼24 min	Yes
Berthoud and Powley (118)	1990	Rat	Complex	Lab chow	N/A	10 min	Yes
Herath et al. (79)	1999	Sheep	Complex	Lucerne pellets	N/A	URc	Yes
Fischer et al. (88)	1976	Dog	Complex	Mashed beef	N/A	5.5 min	Yes
Fischer et al. (91)	1982	Dog	Complex	Mashed beef	N/A	URc	Yes
Fischer et al. (125)	1976	Dog	Complex	Mashed beef, rice, and bread	N/A	3.6 min	Yesa
Fisher et al. (125)	1976	Dog	Complex	Mashed beef, rice, and bread	N/A	2.2 min	Yes
Strubbe and van Wachem (123)	1980	Rat	Complex	Moistened chow	NA	∼3 min	Yes
Eliasson et al. (137)	2017	Humans	Complex	Muffins	N/A	5 min	Yes
Strubbe and Steffens (66)	1975	Rat	Complex	Noncaloric diet	N/A	∼3–5 min	Yes
Lucas et al. (155)	1987	Humans	Complex	Onion tart, tuna tart, and quiche	N/A	1–13 min	Yes
Teff et al. (73)	1993	Humans	Complex	Peanut butter sandwich	N/A	2 min	No
Teff and Engelman (156)	1996	Humans	Complexd	Peanut Butter sandwich	N/A	5 min	No
Secchi et al. (142)	1995	Humans	Complex	Pizza	N/A	5–10 min	Yes
Robertson et al. (143)	2001	Humans	Complex	Pizza, milk, and cream drink	N/A	10–15 min	No
Berthoud and Jeanrenaud (124)	1982	Rat	Complex	Saccharin-flavored milk diet	N/A	∼11 min	Yesa
LeBlanc et al. (56)	1984	Humans	Complex	Sandwich, sugar pie, and soft drink	N/A	10 min	Yes
Ahren and Holst (21)	2001	Humans	Complex	Standard breakfast	N/A	5 min	Yes
Strubbe and Steffens (66)	1975	Rat	Complex	Standard laboratory diet	N/A	∼3–5 min	Yes
Zhu et al. (136)	2014	Humans	Complex	Sugared cereals	N/A	3 min	Yes
Steffens (28)	1976	Rat	Complex	Vivonex (85% carb)	N/A	∼8 min	Yes

CALHM1, calcium homeostasis modulator 1; KO, knockout; N/A, not available; P2X2/X3R, purinergic receptor 2, subtypes 2 and 3; T1R3, taste type 1 receptor 3; SGLT, sodium-glucose cotransporter. ^aSham feeding; ingested contents drained from gastric cannula. ^bOnly found in women with anorexia not in controls. ^cUR: unreported. ^dModified sham feeding after intragastric glucose load with insulin and blood glucose measured. ^eOnly individuals with high amylase activity.

2.3. The Relevant Cephalic Receptors: the Peripheral Gustatory System

2.3.1. Taste is multidimensional.

Given their position at the entrance of the alimentary tract, the oral chemoreceptors are well suited to signal an impending glucose load. For one to attain a complete understanding of potential sources of variability in the outcomes of experiments in the literature, it is necessary to first be mindful of the anatomical and functional organization of the gustatory system.

If form follows function, then the best place to start the discussion of how the organization of the gustatory system relates to mechanisms of CPIR is to review what functions taste serves. First, it is important to recognize that what most people refer to as “taste” is not strictly so, or, in some cases, not even taste at all. They are referring to “flavor,” which is the perceptual combination of the sight, smell, texture, temperature, and taste of foods and fluids. If there is any doubt, hold your nose the next time you drink your favorite glass of wine; it will “taste” entirely different because olfaction, regardless of whether it is retronasal (inside-out) or orthonasal (outside-in), requires airflow through the nasal turbinates, and the absence of smell changes the perceptual experience quite substantially. Thus, taste simply defined, at least for our purposes here, refers to the physiological and behavioral sequalae initiated by the activation of taste receptor cells. Second, perception cannot be measured directly and can only be inferred from behavior. Third, perception, as inferred from behavior, may be unnecessary for cephalic phase reflexes. This point, which will be discussed in more detail below, is a corollary of the view that taste function is multidimensional and can be divided into at least three domains (157) (FIGURE 2): the “sensory-discriminative domain” refers to processes that underlie the discrimination and identification of taste compounds. Stimulus intensity and quality, prothetic and metathetic continua, respectively, in the parlance of sensory psychology, are associated with this function. The “motivational/affective domain” refers to processes that encourage or dissuade ingestion; whether a stimulus tastes “bad” or “good” is an informal way of understanding the nature of this function (see Ref. 158). The “physiological domain,” which is the most relevant to the topic of this paper, refers to physiological reflexes, such as CPIR, that are triggered by an adequate stimulus binding with its cognate taste receptor.

FIGURE 2.

Depiction of a simplified hypothetical framework for how signals arising from the interaction of a ligand (glucose, red hexagon) with two different receptors, T1R2 + T1R3 and an unidentified specific glucosensor (perhaps a GLUT or SGLT, see FIGURE 3), expressed in taste receptor cells might be differentially channeled into brain circuits subserving different taste functions. In the example, a second ligand (fructose, green pentagon) can only bind one (T1R2+T1R3) of the two receptor types, whereas glucose can activate both. Thus, glucose and fructose (and other potential ligands binding with the T1R2 + T1R3) activate similar neural circuits leading to identical outcomes in terms of perception and motivation and affect (with the exception that differences in binding affinity would lead to differences in relative intensity at isomolar concentrations). However, only glucose can stimulate the second chemospecific receptor, the output of which is channeled into circuits that give rise to a cephalic phase insulin response. Some support for this model derives from the facts that glucose and fructose: 1) appear to lead to identical qualitative taste perceptions, and 2) have a similar unconditioned positive taste valence that is virtually eliminated by genetic deletion of one or both subunits of the T1R2 + T1R3 receptor without affecting the ability of glucose to elicit a CPIR. The colored lines within the Integrative Circuits box represent the potential interconnections between the different neural circuits that underlie these taste functions. Importantly, this model makes no assumptions about the brain site(s) of these processes. Such interconnectivity would allow for conditioning to play a role in modulating the CPIR to a normally ineffective ligand and could also support the development of a hedonic discrimination between glucose and other T1R2 + T1R3 ligands such as fructose. T1R2, taste type 1 receptor 2; T1R3, taste type 1 receptor 3; SGLT, sodium-glucose cotransporter. The generic pink-shaded protein with a ? in the potential glucosensors signifies the possible contribution of some other receptor type that remains to be identified. See text for further discussion.

A given chemical stimulus can engage one or more of these domains or none, although in the case of the latter, it would likely not meet the definition of a taste stimulus. Importantly, while these functional domains are not necessarily mutually exclusive, they are dissociable. Different taste stimuli in isolation or in a mixture can be equally preferred or aversive yet clearly qualitatively distinguishable. Likewise, a stimulus like NaCl at low and high concentrations can be equivalently identified as eliciting the prototypical salt taste quality, despite that its hedonic valence substantially changes across the concentration range. Thus, potentially, an adequate stimulus that triggers a CPIR (physiological function) may not be perceived (sensory-discriminative function) and may not elicit appetitive and consummatory responses (motivational/affective function).

To understand how an adequate stimulus initiates the chain of events in the gustatory system ending with an insulin release, one must consider what the potential taste receptors are, where they are located, and how they transmit their signals to the brain. The following paragraphs give a brief overview of the anatomical, physiological, and molecular organization of the peripheral gustatory system. This provides the foundation upon which to understand the possible mechanisms by which some degree of chemospecificity can be conferred among potential triggers and why in some experiments particular stimuli may be less effective than expected.

2.3.2. The anatomy, physiology, and molecular organization of the peripheral gustatory system.

The fundamental sensory organs of the gustatory system are the taste buds, which are distributed in distinct fields in the oral, pharyngeal, and laryngeal epithelia. Taste buds are aggregates of ∼50–100 specialized elongated epithelial cells collectively having a tulip-shaped appearance. The apical membranes of a subset of these cells protrude through a hole in the stratified squamous cell layer called the taste pore through which dissolved chemical compounds gain access to the taste receptors. Within the taste bud, at least three cell types³ have been identified, type 1, type 2, and type 3, that can be dissociated by the specific receptors, signal intermediaries, and other markers that they express (159, 160). The type 1 cell does not express any currently identified taste receptors and is thought to play a supporting role and express the glial marker glutamate-aspartate transporter 1 (GLAST1), but there is some evidence that it might be involved in sodium taste transduction (161, 162). The type 2 cell possesses the G-protein coupled receptors, the ligands of which give rise to the sweet, bitter, and the so-called umami qualitative taste perceptions. These include the taste type 1 receptor (T1R) family members that form heterodimers that bind with amino acids (T1R1 + T1R3) and sugars and sweeteners (T1R2+ T1R3)⁴ as well as the taste type 2 receptor (T2R) family of receptors that bind with bitter tasting ligands. Although still debated, there is substantial evidence supporting that the expression of these receptor classes is segregated among type 2 taste receptor cells (163–165). For example, a cell that expresses the T1R2 + T1R3 receptor would not express a T2R or a T1R1 + T1R3 receptor. The type 3 cells were recently discovered to possess the otopetrin-1 (OTOP-1) proton channel, which is the primary receptor associated with a sour taste (166). The taste bud cell types that possess the epithelial sodium channel that has been shown to be critical for sodium taste in rodent models (167–174) remain to be clearly identified (see Refs. 159, 160). Indeed, this abbreviated summary of taste bud cell types and their receptors belies the complexity of the potential coexpression of receptors for various ligands, both conventional (sugars, salts, acids, alkaloids) and unconventional (fats, maltodextrins, starches) and their transduction machinery, but a detailed discussion is beyond the scope of the question at hand here.

In many mammals, including rats and humans, taste buds are housed in specialized structures called papillae (see Refs. 159, 175–178). The fungiform papillae, which appear as nipple-like structures, are found on the surface of the anterior two-thirds of the tongue. These taste buds are innervated by the chorda tympani (CT) branch of the facial nerve with the pseudo-unipolar sensory neuron cell bodies residing in the geniculate ganglion. Another branch of the facial nerve, the greater superficial petrosal (GSP; also with neuron cell bodies in the geniculate ganglion), innervates the taste buds of the soft palate that do not reside in any specialized papillae; in rodents, the GSP also innervates taste buds along a strip of tissue at the border between the soft and hard palate called the “Geschmacksstreifen” as well as taste buds in the incisive papilla surrounding the openings of the nasoincisor ducts. The taste buds of the posterior tongue are found in the circumvallate and foliate papillae. In the latter, they line the walls of the trenches, which appear as a set of slits on the lateral margins of the tongue. In humans, the circumvallate papillae lie in two diagonal rows on either side of the midline spanning the posterior lingual surface just behind the intermolar eminence. Rodents have only a single circumvallate papilla positioned on the midline. From a top view, these papillae appear as moats surrounding islands of lingual tissue; the taste buds line the walls of the moat. The posterior tongue taste buds are innervated by the lingual branch of the glossopharyngeal nerve (GL) arising from the petrosal ganglion, where the neuron cell bodies reside (see Refs. 175, 177, 179). Finally, there are some taste buds (5–10% of the total) that are distributed in and around the epiglottis and laryngeal epithelium, which are thought to be involved in the protection of the airways (180–184). These taste buds are innervated by the superior laryngeal branch of the vagus (SLN; nodose ganglion).

The afferent fibers of the gustatory nerves penetrate the bottom of the taste bud through the lamina propria where they make conventional and unconventional synapses with taste receptor cells along their long axis (185–187). Recent results from elegant taste bud reconstructions based on serial blockface electron microscopy (187) have revealed that, for the most part, at least in the mouse circumvallate papillae, a given nerve fiber only synapses with one cell type (i.e., type II or type III). Moreover, while a given taste receptor cell can be innervated by more than one nerve fiber, a substantial proportion of nerve fibers, especially those destined for type II cells, innervate a single taste receptor cell. Thus, notwithstanding the caveat that some branching can potentially occur pregeminally, the anatomy of the taste bud is capable of supporting both a very specific and a less specific signal to the brain.

In type II cells, the principal mechanism of neurotransmission is not vesicular, but rather is through the release of ATP through the calcium homeostasis modulator channel 1 and 3 (CALHM1/3) where it binds with P2X2/X3 ionotropic receptors positioned on the postsynaptic membrane of gustatory nerve fibers (188–190). Based on the fact that the CT and GL of P2X2/X3 knockout mice do not respond to citric acid (or any other prototypical taste stimulus) (188, 191–193), it appears that type III cells also use ATP as a neurotransmitter but the conduit subserving its release appears to be independent of CALHM1 (190) and remains to be identified. Type III cells additionally make more classic synaptic connections with gustatory afferents and have a vesicular release of serotonin where it binds with 5-hydroxytryptamine subtype 3 A receptors (5-HT_3A) expressed on the nerve fiber membranes (187, 194–198). In support of a functional role for such synapses, CT responses to acids and other taste stimuli are blunted, but not abolished, by genetic deletion or pharmacological antagonism of the 5-HT_3A receptor (197).

In sum, the evidence strongly implicates ATP and P2X2/X3 as an essential neurotransmitter and postsynaptic receptor, respectively, for peripheral taste signals to reach the brain. As would be expected, taste-related behavioral and gustatory nerve responses to all prototypical taste compounds are virtually eliminated in P2X2/X3 knockout (KO) mice, with the exception that these animals have relatively normal avoidance of citric acid (191, 192). The latter, however, appears to be attributable to the activation of TRPV1 receptors on oral somatosensory afferent fibers (192, 199). Nevertheless, the presence of serotonin in type III cells and the expression of 5-HT_3A receptors on taste afferents along with the findings that functional silencing of the latter attenuates CT nerve responses suggest that there are potential routes of neural transmission, other than through the P2X2/X3 receptor, that taste signals can use to reach the nucleus of the solitary tract (NTS) even when generated from the same receptor cell, albeit such pathways are not sufficient to maintain taste sensibility. Given all the findings discussed above, it is vexing that P2X2/X3 KO (and CALHM1 KO) mice appear to display normal CPIRs to oral glucose (Ref. 80, but see sect. 2.3.3). One critical caveat, however, is that none of the studies of taste nerve responsiveness in P2X2/X3 receptor KO mice that we are aware of used glucose as a stimulus. The CT response to 300 mM glucose was significantly suppressed by a little more than one-half in CALHM1 KO mice, but it did not appear to be eliminated (190).

From the standpoint of reflexive neural circuitry especially with respect to the CPIR, it is noteworthy that all these taste nerves project to the rostral part (r) of the NTS and that the terminal fields are distributed in a rough orotopic fashion with significant overlap. The taste afferent fibers of the facial nerve (CT and GSP) project most rostrally in the NTS and those of the GL terminate slightly more caudally followed by those of the SLN of the vagus (200–203) (see sect. 4.2).

2.3.3. Potential glucose-specific sensing mechanisms in taste receptor cells.

If indeed glucose is the optimal unconditioned stimulus relative to other sugars, nonnutritive sweeteners, and diets with complex macronutrient composition to trigger a CPIR, then the fact that the T1R2 + T1R3 heterodimeric taste receptor binds with all sugars and nonnutritive sweeteners and disaccharide products of starch hydrolysis (i.e., maltose) creates a conundrum for understanding the basis of such a chemospecific effect (164, 204, 205). The latter suggests that there must be some transduction pathway that is T1R2 + T1R3-independent and can signal specifically the oral presence of glucose. Over the last decade, there has been growing evidence in rodents that taste receptor cells express various forms of glucose transporters including GLUT1-5, GLUT8, GLUT9, and SGLT1 (206–211). In addition, SUR1 and Kir6.1, which are both components of the K_ATP channel, are also expressed (206, 210, 211). Moreover, glucokinase is expressed in type III taste bud cells and its lingual modulation has been shown to affect taste-related behavioral responses to glucose (212). The expression of GLUT1 and 2, glucokinase, and components of the K_ATP channel is of particular significance because these are critical parts of the glucose-sensing mechanism in pancreatic β-cells. Likewise, the expression of SGLT1 is also of relevance because it is one of the main throughways for glucose absorption in the small bowel brush border membrane (213–215) and is pivotal for glucose sensing in enteroendocrine and intestinal epithelial cells (215–218). Importantly, fructose uses GLUT5 but can also use GLUT2. The presence of SGLT1 and some GLUT members, notwithstanding GLUT5 and perhaps GLUT2, provides the basis for a potential glucose-specific pathway into taste receptor cells. However, in these cells, there is a tendency for some of these transporters (GLUT1, GLUT4, and SGLT1), as well as SUR1, to be coexpressed with T1R3 (208, 210, 211). Thus, given that fructose and glucose both bind with the T1R2 + T1R3 receptor, any discrimination between fructose and glucose that was dependent on SGLT1 or GLUTs in the same cell would require a differential output signal (e.g., neurotransmitter), an arrangement that has yet to be resolved.

Notwithstanding that caveat, there is accumulating functional evidence that some of these proteins are involved with glucose sensing in the gustatory system and, importantly, are necessary for CPIRs. Recently, the potential involvement of SGLT1 in gustatory glucose detection has been supported by the finding that small amounts of sodium added to a solution of glucose or sucrose, but not of a nonnutritive sweetener (SC45647), enhanced the responses of the CT and GL and was blocked by the addition of 1 mM phlorizin (an SGLT inhibitor) (219). The CT response to a nonnutritive sweetener was also unaffected by the addition of phlorizin. The CT responses to glucose and sucrose were observed in T1R3 KO mice, albeit much reduced, but responsiveness to these sugars was still enhanced by the addition of 10 mM NaCl and suppressed by the addition of 1.0 mM phlorizin. The effectiveness of NaCl and phlorizin to influence the response of the CT nerve to sucrose or glucose but not to nonnutritive sweeteners even in the KO mice testifies to the presence of a T1R3-independent SGLT pathway ultimately activating some taste receptor cells. This conclusion is further buttressed by some studies that have shown that the CT in T1R3 KO mice displays significant concentration-dependent responses to sugar solutions including glucose and fructose, but not surprisingly, such activity is markedly reduced relative to that seen in the wild-type (WT) controls (70, 205, 220). Collectively, these data are important because T1R3 KO mice can still display a CPIR to glucose (although see caveats below). On the other hand, fructose is an ineffective stimulus for a CPIR, even in WT mice, and so it remains a conundrum how the two monosaccharides are distinguished neurally beyond the taste receptor cell. Interestingly, one study found that CT responses to glucose were not different between WT and T1R3 KO mice although sucrose and fructose responses were markedly suppressed in the latter strain, but this seems to be the exception rather than the rule in the literature (221). It is also important to consider that the weak concentration-dependent responsiveness to sugars displayed by the CT in T1R3 KO mice, was eliminated in animals that had both the T1R2 and T1R3 deleted (205). Because there are some cells that express T1R3 alone, Zhao et al. (205) hypothesized that the T1R3 might form a homomer to maintain some degree, albeit compromised, of sugar signaling capacity (also see Ref. 222). Nevertheless, the fact that sugars, including fructose and glucose, all appear to bind with the T1R2 + T1R3 as well as perhaps T1R3 alone (homomer), mitigates the role of T1Rs in glucose-specific sensing.

It is noteworthy that Yasumatsu et al. (219) found that the CT and GL responses to sucrose in WT and T1R3-KO mice were just as affected by NaCl and phlorizin as those to glucose. On the surface, this would seem to challenge the potential specificity of the response and the involvement of SGLT1. However, there are various saccharide-hydrolyzing enzymes including sucrase-isomaltase, maltase-glucoamylase, lactase, trehalase, and α-glucosidase C, that are expressed in the apical portion of some taste receptor cells (98). Thus, these results support the argument that the action of sucrase could hydrolyze sucrose and provide a source of glucose to activate the cell. Indeed, the presence of α-glucosidases where the taste receptor proteins reside coupled with the action of salivary amylase leaves open the possibility that complex foods as well as certain sugars could be effective stimuli at generating a cephalic phase response by virtue of their glucose content. Yasumatsu et al. (219) speculated that the sodium cotransported into the taste receptor cell may contribute to the depolarization of the cell as it is thought to do in the L-cells of the intestine (223–225). The presence of the K_ATP channel offers another or additional means of depolarization but would require the metabolism of the glucose in the cell. The low NaCl concentrations used in a mixture with glucose should be well within the range of values found in saliva especially considering that sodium content increases proportionally with flow rate, which should typically be higher during ingestion (226–230). The proposed glucose-sensing mechanisms in taste receptor cells are depicted in FIGURE 3.

FIGURE 3.

A: an idealized type II taste receptor cell that expresses the T1R2 + T1R3 heterodimer, which binds with sugars and other sweeteners, is depicted. The intracellular signaling cascades leading to the release of ATP through a CALHM1/3 channel are also illustrated. The ATP binds with P2X2/X3 receptors found on nearby afferent fibers. This form of neurotransmission involves an unconventional synapse. B: the various T1R-independent proteins in the apical membrane of some taste receptor cells that have been proposed to serve in the potential specific sensing of glucose in the oral cavity. The generic pink-shaded protein with a ? in the apical membrane signifies the possible contribution of some other receptor type that remains to be identified. The corresponding possible transduction pathways leading to putative neurotransmitter release are also depicted. α-Amylase in saliva and α-glucosidases expressed by some taste receptor cells and found in the taste pore are thought to hydrolyze various glucose polymers, and di-, and tri-saccharides providing a source of glucose moieties to be transported either through GLUTs or SGLT1 into the cell. There is evidence that glucokinase expressed in some type III cells is involved. The ATP produced by glucose catabolism is thought to close a K_ATP channel causing depolarization of the cell membrane. Additionally, the sodium cations co-transported with glucose through the SGLT1 are thought to contribute to membrane depolarization. The neurotransmitter and the mechanism of its release are unclear because there is evidence that P2X2/X3 knockout mice still display a cephalic phase insulin release to glucose. As discussed in the text, there are caveats for many of these proposed glucose-sensing mechanisms that remain to be resolved. CALHM1, calcium homeostasis modulator 1; DAG, diacylgycerol, GLUTs, glucose transporter(s); IP3, inositol triphosphate; Kir6.1, inward rectifying potassium channel subtype 6.1; P2X2/X3R, purinergic receptor 2, subtype 2, and subtype 3; PIP2, phosphatidylinositol 4,5-bisphosphate; PLCβ2, phospholipase C-β2; SGLT1, sodium-glucose cotransporter-1; SUR1, sulfonylurea receptor subtype-1; T1R2, taste type 1 receptor 2, T1R3, taste type 1 receptor 3; TRPM4, transient receptor potential melastatin type 4; TRPM5, transient receptor potential melastatin type 5; ↑ Vm, change in membrane potential. See text for more discussion.

These findings in the rodent model recently received support from a human psychophysical study demonstrating glucose detection thresholds measured with a two-alternative forced-choice staircase procedure were modestly, but significantly, lowered (= increased sensitivity) by adding 20 mM NaCl (which was in the comparison solution) and were more substantially raised (= decreased sensitivity) by the addition of 0.2 mM phlorizin (231). The same effects were observed when α-MDG, a nonmetabolizable compound that can use SGLT1, was the stimulus. This suggests that intracellular metabolism of the sodium cotransported stimulus via SGLT1 is not required to see an effect on threshold. Critically, the addition of either NaCl or phlorizin has no effect on fructose threshold, a result consistent with the inability of fructose to access the cell through SGLT1.

As discussed above, Glendinning et al. (70) trained WT and T1R3 KO mice to lick a small amount of 2.8 M (50%) glucose and measured the insulin and blood glucose levels at baseline, and 5, 15, 30, and 60 min following the initiation of licking (maximum duration 3 min). Other WT and T1R3 KO mice were treated the same way but had similar amounts of the glucose solution gastrically gavaged (<1 s). At 5 min there was a significant and similar increase in plasma insulin in both the WT and the T1R3 KO mice that had the glucose delivered orally but not in the mice that had it delivered gastrically. Although there appeared to be an increase in blood glucose in both modes of stimulus delivery at 5 min, the fact that only the oral glucose caused a rise in plasma insulin at that time point was reasonably constituted as evidence of a cephalic phase response. Moreover, blood glucose levels over the 60 min were significantly lower for the oral compared with the intragastric delivery, showing much better glucose tolerance. These results in conjunction with the fact that T1R3 KO mice are unable to detect glucose in a psychophysical task, nor display concentration-dependent licking of the sugar in a brief access test (70, 232, 233), highlight the point that while a cephalic phase response to a taste stimulus (e.g., glucose) requires reception, it does not necessarily require perception (or affective responses) (FIGURE 2).

Glendinning et al. (70) went on to show that, while 1.0-M solutions of glucose and sucrose were both capable of raising plasma insulin at 5 min after licking initiation in both WT and T1R3 KO mice, fructose was not. However, blood glucose levels were not reported and so whether the disparities in the effectiveness of these sugars to trigger the insulin release is of oral or pancreatic origin remains unclear. In a subsequent study (80), they measured plasma insulin and blood glucose every minute for the first 5 min after licking of 1.0 M glucose started and found that the first detectable significant rise in plasma insulin occurred at minute 3, whereas it did not occur for blood glucose until minute 4. By minute 5, the rise in blood glucose was quite clear and so taking a blood draw at that time or any time after minute 3 makes it difficult to disambiguate whether the stimulus is critically activating an oral or pancreatic site. With this caveat in mind, at 5 min after the initiation of licking, the order of effectiveness of various saccharide stimuli to generate a significant plasma insulin rise was glucose ≥ sucrose ≥ maltose = Polycose (a maltodextrin). Fructose and α-MDG and the nonnutritive sweeteners saccharin, Ace K, sucralose, and SC45647 were all ineffective. Of course, none of the ineffective stimuli, including α-MDG, which can be transported by SGLT1, contain metabolizable glucose. The disaccharides and the maltodextrin could potentially be hydrolyzed by the α-glucosidases noted above to produce glucose. In fact, when 5 mM acarbose (an inhibitor of both α-glucosidases and α-amylase) was mixed with the stimuli, the ability of sucrose, maltose, and Polycose to trigger a 5 min rise in insulin was abolished. But where is the site of action of the glucose? Is it at the apical membrane of the taste receptor cell or is enzymatic hydrolysis happening in the small bowel to cause a postabsorptive direct stimulation of the β-cell? Because blood glucose was reportedly elevated at min 5 when there was no acarbose but was markedly reduced or abolished when the enzymatic inhibitor was present, it is difficult to unequivocally distinguish between an oral and pancreatic site of action. In support of an oral site, Mandel and Breslin (95) found that humans with high salivary amylase activity displayed a CPIR at 9 min after the start of ingestion of 50 g of corn starch (10% solution) over a 20-min period whereas those with low salivary amylase activity did not, despite that blood glucose levels did not increase above baseline at that time point.

Glendinning et al. (80) also found an increase in plasma insulin at 5 min after the start of licking of 2.8 M glucose in CALHM1, P2X2/X3, and SGLT1 KO mice. In the CALHM1 KO mice, there was a significant increase in blood glucose levels at 5 min making the insulin result more equivocal, but for the P2X2/X3 and SGLT1 KO mice there was not. However, the result in the latter KO strain must be viewed with some caution because the plasma insulin levels, while significantly above baseline in both WT and KO mice, were exceptionally low, possibly due to the very low carbohydrate diets used to maintain the mice of both genotypes. When SUR1 KO mice were tested with orally presented 2.8 M glucose, they did not display a significant rise in plasma insulin levels at 5 min despite displaying a significant increase in blood glucose. Expectedly, the SUR1 KO also showed a very weak insulin response, at best, throughout the 60-min period, very modestly but significantly rising above baseline only at 15 min. In support of the necessity of SUR1 in the response, when K_ATP channels were closed by the addition of 0.15 mM glyburide, which works through SUR1, to the 0.5-M glucose oral stimulus, the plasma insulin response at 5 min significantly increased. Likewise, when K_ATP channels were opened by the addition of 0.25 mM diazoxide to the 1.0-M glucose oral stimulus, the plasma insulin response at 5 min significantly decreased. Of course, all these manipulations provide compelling support for the necessity of both the SUR1 subunit and the action of K_ATP channels in the glucose-triggered rise in plasma insulin at 5-min after the initiation of licking, but given the entire profile of results, it is equivocal whether the site of glucose action is in the taste receptor cells or in the pancreatic β-cells or both.

To the extent that the critical stimulating events are indeed in the taste receptor cells, the results from the KO mice suggest that the SUR1-dependent closure of K_ATP channels as a function of the metabolism of intracellular glucose does not rely on SGLT1 for extracellular glucose to cross the cell membrane. Perhaps one of the GLUTs is contributing. It should also be noted that while the competence of SGLT1 KO mice demonstrated that the cotransporter appears to be unnecessary, caveats notwithstanding, this result does not prove that SGLT1 is not sufficient and thus it still could be contributing. In this regard, it would be instructive to test the effects of the addition of phlorizin to the glucose stimulus on CPIR. Moreover, a different mode of synaptic transmission in the K_ATP-expressing taste receptor cells, other than the release of ATP through the CAHLM1/3 channel and binding with P2X2/X3 receptors on afferent nerve fibers, must be at play. The fact that P2X2/X3 does not appear to be necessary for the glucose-specific stimulation of early insulin release abates the interpretive tension that, like SGLT1 and GLUTs, SUR1 appears to be coexpressed with T1R3 in taste receptor cells. The T1R2 + T1R3 heterodimer is stimulated by both glucose and fructose, but only the former is effective at triggering the insulin response. In other words, there must be two transduction pathways for glucose in the same subset of taste receptor cells, one that is T1R3-dependent and the other that is SUR1-dependent, which lead to release of different neurotransmitters.

2.3.4. The neural code for glucose.

Regardless of the expression and coexpression patterns of transporters, receptors, and transduction intermediaries in taste receptor cells, what ultimately counts, is the nature of the neural signal transmitted to the brain through the gustatory nerves described above. Although it remains controversial, there is no escaping the fact that in the peripheral gustatory system, as elegantly demonstrated decades ago by the pioneering rodent single fiber recordings of Marion Frank and her colleagues (234, 235), there is a subset of fibers and ganglion cells that respond selectively to qualitative classes of taste compounds (specialists) and other peripheral neurons that respond more generally (generalists) (see Refs. 179, 236, 237).⁵ This presence of specialist units in the peripheral gustatory system has been taken as support for a labeled-line coding mechanism by which activity in a single class of neurons gives rise to a specific qualitative taste sensation. This contrasts with another popular neural coding model, the across-neuron pattern or ensemble code, which posits that taste quality is represented by the pattern of activity in a population of neurons at a given level of the gustatory neuraxis (see Ref. 179). As some have argued, the fact that the breadth of tuning in some geniculate ganglion cells increases as the concentration of test stimuli is raised undermines the labeled-line model (238). In considering the merit of this view, it is important to note that even at midrange concentrations tuning was still quite narrow, with half the neurons responding to only one of the prototypical taste stimuli representing a given taste quality and with the vast majority of the remaining neurons still responding quite well to their best stimulus as indicated by entropy values less than 0.5 (see Ref. 179 for a discussion on the use of the entropy value to represent the breadth of tuning). It is true that when the concentrations in the test panel were even raised further there was far less selectivity in the response profiles, there were nonetheless still some cells that responded only to a single compound from the test array, and most maintained their best stimulus category. More to the point, absolute response specificity is not a requirement of the labeled-line coding model. The only strict theoretical obligation is that whatever stimulates the neuron, regardless of whether it is the adequate stimulus, the activity is interpreted by the brain as a representation of its label (e.g., a given taste quality). In this sense, it is a microcosm of the Law of Specific Nerve Energies (84). As noted above, sensory receptors often respond to stimuli that are outside the range of their adequate stimulus provided that the intensity is high enough. However, with respect to the theoretical framework of the multidimensional nature of taste function, it is quite likely that more than just taste quality is represented across the population of afferents projecting to the brain. In other words, quality may be just one label (see Refs. 157, 176, 179). There could be other “labels” such as hedonic valence, or relevance to specific physiological processes (salivation, hormone release, etc.). The taste literature is dominated by attention to the neural coding of taste quality without much consideration of other features of a taste stimulus, except for perhaps intensity, that can be represented (see Refs. 157, 179). In the context of our scrutiny of CPIRs, the relative absence of glucose in the test stimulus panels of many neurophysiology studies, with some notable exceptions, has created a relative data scotoma in the literature. That said, Yasumatsu et al. (219) identified a class of single fibers in the CT of WT and T1R3 KO mice for which glucose responses were enhanced by phlorizin but responses to the nonnutritive sweetener SC45647 were not, suggesting that an SGLT-dependent taste transduction pathway might activate peripheral units relatively narrowly tuned to glucose.

In nonhuman primate work on the gustatory regions of the insular, opercular, and orbitofrontal cortices, glucose has been the sweetener of choice for the examination of the response profiles of neurons. However, because other pure sweetener compounds (especially sugars) were not included in the test panels (e.g., fructose, sucrose), the question of specificity cannot be addressed. Nevertheless, some rodents studies of central gustatory structures have included glucose in their stimulus test panels, and while there is at times some separation of glucose from other sugars in multidimensional scale spaces (e.g., Ref. 239), neural responses to glucose are generally correlated with those to other sugars and there is no striking evidence of glucose-specific units (240–243). For example, a few studies of the mouse NTS (242, 243) or the parabrachial nucleus (PB) (240) found units that were responsive to glucose but not in a fashion that was ostensibly distinguishable from other sweeteners. Moreover, Kalyanasunder et al. (242) adapted the tongue to artificial saliva containing 15 mM NaCl (see Ref. 244), which also served as the solvent for the taste stimuli; based on the findings discussed above, the 15 mM concentration should have been sufficient to amplify neural responsiveness to glucose presumably through the SGLT1. In fact, in the absence of a functional T1R2 + T1R3 taste receptor, the response to glucose was substantially attenuated but not entirely eliminated (242, 243). The same was true for fructose. The fact that some significant responsiveness to glucose and fructose remained, albeit severely blunted, supports the presence of a T1R2 + T1R3-independent signaling pathway for these monosaccharides. This conclusion is further buttressed by some studies that have shown that the CT in T1R3 KO mice displays significant concentration-dependent responding to sugar solutions including glucose and fructose, but not surprisingly, such activity is markedly reduced relative to that seen in the WT controls (70, 205, 219, 220). Collectively, these data are important because T1R3 KO mice can still display a CPIR to glucose. On the other hand, fructose is ineffective, even in WT mice, and so it remains a conundrum how the two monosaccharides are distinguished neurally beyond the taste receptor cell.

In evaluating the significance of the neurophysiological literature especially regarding the nature of neuronal responses generated by oral application of glucose relative to other sugars, one must consider that there is a methodological constraint, in at least anesthetized preparations, that biases the results toward neurons with anterior tongue and palatal receptive fields. Namely, it is sometimes difficult to achieve adequate stimulus perfusion into the trenches of the foliate and circumvallate papilla. Frank (234) commented on this and found that it was necessary to insert a pipette into the trench(es) of the rat circumvallate and foliate papillae to record optimal responses from glossopharyngeal single fibers. This is not to say that neurons with posterior tongue receptive fields cannot be identified, as the detailed oral mapping studies of Travers and her colleagues (e.g., Refs. 245–248) have demonstrated, but perhaps there are some units that are going undetected because the stimulus is not reaching the complete repertoire of taste receptor cells in those fields.

2.3.5. Discrimination among sugars: behavioral evidence.

Although not required, if there was evidence that glucose could be discriminated from other sugars, particularly fructose, based on taste quality and not intensity, then this would provide some qualified support for the presence of a T1R2 + T1R3-independent transduction pathway and provide a potential means by which glucose could have privileged status at generating a taste-evoked insulin response. As will be shown, however, the composite findings from the literature on this issue, not surprisingly, paint a complex picture.

Adopting a strategy like that used in the metameric color-matching experiments conducted by vision scientists to study photopic vision (see Ref. 249), Breslin et al. (250) tested whether humans could discriminate among sugars based on quality. One sugar, at a fixed suprathreshold concentration, was chosen as the standard stimulus, and another sugar, serving as the comparison stimulus, was presented at different concentrations across trials. A two-alternative forced-choice procedure was used in which three cups were presented two of which had the standard and the third had the comparison, or two cups had the comparison (at the single chosen concentration for that trial) and the third had the standard. The participant had to choose the odd one. When this was done, there was always a concentration of the comparison that could not be discriminated from the standard for most sugar comparisons. As the concentration of the comparison sugar was raised or lowered from the undistinguishable one, discrimination performance improved, apparently because intensity cues came into play. The one exception was when concentrations were greater than 100 mM, maltose was discriminable from fructose, but in general, fructose, sucrose, and glucose all appeared to be indiscriminable when intensity was controlled. Breslin et al. (250, 251) coined the term “monogeusia” to refer to the fact that the sugars generate an identical qualitative taste perception. Importantly regarding the discussion here, glucose was indiscriminable from fructose and presaged the discovery of the T1R2 + T1R3 sugar taste receptor at the beginning of the next millennium. Of course, to the extent that SGLT1 is part of a critical glucose-specific taste sensor, perhaps if a small amount of sodium had been placed in the stimuli, glucose would have been discriminable from fructose. On the other hand, it is important to remember that the saliva that bathes the oral epithelium contains sodium. In any event, based on these findings, it appears that humans cannot discriminate glucose from fructose and thus if glucose works and fructose does not at eliciting a CPIR in humans, then it follows that perceiving a difference between the two sugars is unnecessary.

Early rodent work found that a taste avoidance conditioned to a given sugar solution did not entirely generalize to other sugar stimuli, but some of the generalization decrement could have been along the intensity dimension as opposed to quality (see Ref. 252). In other words, just like humans, animals can make discriminations between strong and weak tastes of the same quality; not every ligand for a given receptor would be expected to have the same binding affinity. Spector and colleagues (252) controlled for intensity and demonstrated that while a taste avoidance conditioned to either sucrose or maltose cross generalized, the two disaccharides also did not appear to be completely identical. Later explicit psychophysical discrimination experiments confirmed that rats could distinguish the taste quality of sucrose and maltose even when intensity cues are minimized by varying concentration (253). Transection of the CT or GL alone had no effect on the discrimination, but combined transection of the CT and GSP, depriving the brain of taste signals from cranial nerve VII, substantially impaired the discrimination showing that it was based largely on oral signals (see Ref. 254 for more discussion). As a relevant aside, a variety of nerve transection studies of this ilk confirmed that in the rat the input of the facial nerve is essential for the maintenance of normal sensory-discriminative taste function; the input of the GL appears to be unnecessary (178, 255–264). This is important because, at a minimum, it shows that the respective inputs arising from the different gustatory nerves of the rat do not contribute equally to all taste functions (see Refs. 178, 265 for more discussion).

Like rats, mice trained in a psychophysical taste discrimination task, were able to distinguish sucrose from maltose, but only poorly (266). However, they were entirely unable to discriminate sucrose from glucose. Perhaps the extra glucose moiety produced upon maltose hydrolysis made that stimulus easier to discriminate from sucrose, which upon hydrolysis would only produce a single glucose moiety (and a fructose moiety). These mice were also unable to discriminate sucrose from fructose. Although hydrolyzed sucrose would engage both the T1R2 + T1R3 receptor and a glucose-specific sensor, fructose would only engage the former and thus the two sugars would not stimulate an identical set of receptors and should be discriminable, but perhaps, as was possibly the case with the sucrose versus glucose discrimination, the strength of the glucose-specific signal was not sufficient in the background of T1R activation. Unfortunately, a direct test of whether glucose was discriminable from fructose was not conducted in that study. Moreover, perhaps if the taste stimuli would have been dissolved in a weak NaCl solution, better discrimination among some of these sugars would have been revealed as a function of the contribution of SGLT1. Nevertheless, Dotson and Spector (266) concluded that, with the possible partial exception of maltose, sugars produce a unitary taste quality in the rodent gustatory system, a conclusion that is in register with the findings of Breslin et al. (250, 251) in humans.

Sclafani and coworkers (267–273) have shown that glucose is a superior unconditioned stimulus for flavor-nutrient learning and that fructose is relatively ineffective. Schier and her colleagues (274–278) have demonstrated that the difference between these two monosaccharides in the potency of their postoral consequences to support flavor-nutrient conditioning also appears to foster an experience-dependent ability for rodents to rapidly discriminate between them based on orosensory signals. If rats and mice are given individual experience ingesting glucose alone on some occasions alternated with fructose alone on others, the animals will come to display enhanced licking to glucose over fructose across a range of isomolar concentrations when tested together in a brief access test. By virtue of the small stimulus volumes and immediate responses on a given trial in this test paradigm, the behavior is thought to be guided by events of orosensory origin. What is surprising about these findings, however, is that there must be an orosensory feature that allows the animal to distinguish between these sugars in the brief access test. Furthermore, the enhanced hedonic responding to glucose after sufficient single-bottle experience with glucose and fructose does not appear to depend on the T1R2 + T1R3 receptor (275, 279) nor on the TRPM5 signaling intermediary for G-protein-coupled taste receptors in type II cells (278). With respect to the latter, an enhanced avidity for glucose over fructose does appear to be more difficult to train in TRPM5 KO mice (278, 279), but with sufficient ingestive experience, these mice can express differential responses to the two sugars when pitted together in a brief access test (278). In addition, α-MDG or galactose is not interchangeable with glucose in this phenomenon, suggesting that SGLT1 is not the origin of the T1R2 + T1R3-independent pathway underlying this particular taste discrimination (270, 272, 275, 278, 279). As noted above (sect. 2.3.3), however, there is evidence that the enzyme glucokinase is expressed in a subset of mouse taste bud cells, generally type III cells, and that it plays a role in taste-related behavioral responsiveness to glucose. More to the point, its selective lingual knockdown impairs the behavioral expression of the learned hedonic glucose versus fructose discrimination in mice (212).

Some have argued that these sugars may have an olfactory signature that serves as the basis of the discrimination learning because, at times, but not always, there is a greater latency to approach the drinking spout on fructose trials during testing (275, 278, 280), and experimental damage to either the olfactory bulbs or mucosa abolishes either the learning or expression of the exposure-induced glucose versus fructose hedonic discrimination (280). However, while disrupting, or even eliminating, competent performance in this hedonic sugar discrimination paradigm by impairing olfactory function provides evidence that this sensory system must remain intact for rodents to distinguish glucose from fructose in a normal fashion, it does not demonstrate that such signals are sufficient. In other words, these results do not rule out a significant contribution from taste signals (see Refs. 275, 278). Parsimony would justify the expectation that the brain would make use of, and depend on, a variety of sensory signals, including those from both the gustatory and olfactory systems, to support such a learned hedonic discrimination based on the postoral consequences of the ingesta.

There are at least two ways the trained hedonic discrimination between glucose and fructose in rodents described above is relevant to the issue of CPIR. First, it is clear that rodents, at least, have the capacity to differentially respond to glucose and fructose based on input from head receptors. The duration of the trials in the brief access test, as its name implies, is very short (10 s), and it is highly unlikely, based on the volumes consumed, that animals would be able to use postoral receptor systems to influence the ongoing behavior within a single trial. In this case, the differential response is in the degree of appetitive and consummatory behavior elicited by the two sugars respectively, but potentially it could be in the form of the presence or absence of a CPIR. Second, naïve rodents treat both fructose and glucose similarly at isomolar concentrations in the brief access test. It is not until animals have sufficient experience with each of the sugars individually that they display enhanced licking to glucose and not to fructose. Accordingly, experience may be required to allow potential discriminable cues that are present between sugars to exert control on the animal’s taste-related behavior or perhaps physiological processes such as CPIR.

2.3.6. Coda: Taste is multidimensional.

It is quite possible that animals and humans cannot perceptually distinguish glucose from other sugars. Indeed, the experience-dependent glucose versus fructose discrimination work in mice just discussed above involved measures in the hedonic domain. That is, the tasks were designed to determine which sugar evoked greater responses or preference in a licking/drinking test. Obviously, there would have to be some discriminable signal acting on affective/reward circuits in the brain to guide the behavior, but this does not necessarily mean that the respective taste perceptions generated are different (see Ref. 277). On the other hand, the psychophysical work discussed above, which assessed function in the sensory-discriminative domain, suggests that at least some sugars might in fact be discriminable (e.g., sucrose versus maltose). Regardless, it is important to distinguish between the reported qualitative taste perception that an oral stimulus evokes (e.g., sweetness) and the potential for that compound to influence neural circuits that might be independent of perception. For example, decerebrate rats, in which the entire forebrain is neurally isolated from the rest of the nervous system, will increase their ingestive oromotor behaviors, often used as an indicator of hedonic responsiveness, as a function of increasing concentration of sucrose directly infused into the oral cavity. However, it is difficult to argue that such rats are experiencing a perception of sweetness or even positive affect. Of relevance to the point at hand, Flynn et al. (281) demonstrated that such rats will display a CPIR to intraorally delivered glucose. Such a view is buttressed by the findings of Yamazaki and Sakaguchi (92) demonstrating that, in humans, a 1.1-M application of the α-anomer of glucose is more effective than the β-anomer at stimulating a CPIR even though the same participants judge the sweetness of the two anomers to be similar. Such functional compartmentalization of sensory brain circuits is a concept difficult for some to grasp. A salient example from the visual system is the phenomenon of “blindsight” (see Refs. 282–285). Humans and monkeys with sufficient damage to the primary visual cortex are unable to report the presence of many features of a visual stimulus and thus for all intents and purposes can be considered blind. Yet, such individuals can perform above chance in locating certain types of stimuli placed in their visual field or in tracking moving stimuli. From the standpoint of reflexes, their pupils constrict and dilate in response to the intensity of light shined on their photoreceptors. Clearly, the remaining undamaged parts of the visual system are able to maintain some degree of competent responses to photons in the visual spectrum.

Thus, while the adequate stimulus must be transduced at the sensory organ (reception), there is no theoretical prerequisite that the signal is perceived (perception) to produce a cephalic phase response (FIGURE 2).

3. THE MOTOR LIMB

3.1. Preamble

This section will discuss what is currently known about how the brain uses exteroceptive modalities to drive CPIR. The fundamental organizing principle of the CPIR requires that the brain process signals arising from exteroceptive modalities in a way that alters endocrine secretion via neural effector projections to endocrine cells. The most important of these modalities is gustation, but also to a lesser degree, olfaction and vision (see sect. 2.2.5). For pancreatic islets, the activated motor outputs from the brain are modified parasympathetic and sympathetic drives.

We begin this section with an overview of the organization of the sensorimotor components responsible for CPIR. The structural organization of basal insulin secretory mechanisms is then described, followed by the neural motor inputs to pancreatic islets and β-cells. These are the main components engaged by the brain to control the insulin secretion seen with preabsorptive stimulation. Finally, we consider the possible brain integrative mechanisms that link primarily gustatory processes (see sect. 2.3) with motor control mechanisms in the dorsal medulla. These integrative mechanisms are the least well-understood aspects of CPIR control.

3.2. Overview: from Sensory Transduction to Motor Execution

The reflex is the simplest command chain for all motor actions whether the effector is skeletal muscle, smooth muscle, exocrine cells, or endocrine cells [see Gallistel (286)]. By definition, a reflex enables sensory information to directly control motor actions without intervening integrative mechanisms (287). For the parasympathetic drive to pancreatic islets that involves postabsorptive glucose (i.e., glucose as it enters the hepatic portal vein from the gut), reflex control involves vago-vagal or spino-vagal reflexes, the sensory limbs of which exist from the duodenum and the walls of the hepatic portal and superior mesenteric veins to the dorsal motor nucleus of the vagus (DMX) (288–290). Hypothetically, CPIR could be initiated very rapidly by a reflex that begins with the oral glucosensory information that is transmitted by gustatory afferent fibers to rNTS neurons, which then project directly to the DMX, the outputs of which then influence intrapancreatic parasympathetic ganglia and then β-cells (FIGURE 4). However, as we discuss in sect. 4, the nature of any such connections is currently unclear.

FIGURE 4.

A hypothetical schematic of the neural circuitry that can influence insulin release from β-cells (red text and arrows) in response to gustatory stimulation. The primary neural drive for cephalic phase insulin release (CPIR) consists of preganglionic parasympathetic neurons [vagal motor neurons (VMN)] in the dorsal motor nucleus of the vagus (DMX) that project to the pancreas to increase insulin release (brown text and arrows). Please see text and FIGURES 5–7 for more detailed representations of the intrapancreatic processes. How the neural control of these DMX neurons is organized for CPIR is still unclear, but this schema (1: green text and arrows) is a reasonable starting point based on the literature. The light tan box highlights this circuitry, which is also represented in FIGURE 5. Glucose in the oropharynx interacts with receptors expressed in taste receptor cells (see FIGURE 3) and generates sensory signals in the taste buds that are conveyed to the rostral nucleus of the solitary tract (rNTS) by the chorda tympani nerve (CT), the glossopharyngeal nerve (GL), the greater superficial petrosal nerve (GSP), and the superior laryngeal nerve (SLN) (purple text and arrows; see text for more details). Output from the rNTS is directed in two ways: first, either directly or indirectly to the DMX by as yet undefined projections (green dashed arrow); and second, to integrative regions in the caudal brainstem, primarily in the reticular formation and, in rodents, the parabrachial nucleus. In turn, these integrative regions project to the DMX (green arrows), but they also have bidirectional projections with various forebrain regions, including the hypothalamus, that may influence CPIR by direct and indirect projections to the DMX. DMX control of CPIR may also be influenced by glucagon-like peptide-1 (GLP-1) released from L-cells (2: blue text and arrows) in response to increasing glucose in the gut (see FIGURE 5). GLP-1 is detected by receptors (small blue circles) expressed by vagal sensory nerves that are located close to L-cells and in the hepatic portal vein wall (dark gray box). These vagal sensory nerves (VSN) then project to the medial part of the NTS (mNTS). Finally, glucose in the posthepatic circulation acts directly on β-cells to stimulate insulin secretion (3: black text and arrows).

What is known about the fundamental organization of how motivated behaviors are controlled by the brain (287) offers a template for a general understanding of the sensorimotor integration required for CPIR. Because motivated behaviors are obviously far more complex than cephalic phase responses and involve very different motor effectors, using such a template may at first glance seem unhelpful. However, there are some fundamental similarities. For example, the minimum circuitry for orally provoked autonomic reflexes including CPIR and salivation is, like that for food intake (291), still functional in decerebrate rats (281, 292, 293). However, full expression of eating behaviors requires complex sets of projections from the hypothalamus and other forebrain regions (FIGURE 4) (287). Some of these regions, the paraventricular hypothalamic nucleus (PVH) and the parasubthalamic nucleus (PTSN) in the lateral hypothalamic area (LHA) for example, provide very robust projections to the DMX (294, 295), but, again, the significance of these descending forebrain projections for CPIR is unknown. The fact that cephalic phase salivation and GI motility can be conditioned (13, 16), like insulin secretion in some circumstances (e.g., Refs. 25, 296), also points to the existence of a control network that is more complex and extensive than a simple reflex. We discuss these possibilities in greater detail in sect. 4.

3.3. Mechanisms That Control Insulin Secretion

3.3.1. General features.

Insulin secretion rates and patterns are controlled by a multilayered and complex set of neural, paracrine, and humoral mechanisms (297–300). These include 1) direct neural innervation of β-cells; 2) innervation of islets by intrapancreatic ganglia, particularly postganglionic parasympathetic terminals; 3) paracrine signals from β and other islet cells, particularly alpha cells; and 4) a variety of humoral agents, the most important of which is glucose, but also amino acids and fatty acids (e.g., Refs. 301–303), the incretins GIP and GLP-1 (31, 304–306), and leptin (307–309). FIGURE 5 summarizes parts of this multilayered control, which provides great adaptive flexibility for β-cells to secrete insulin in response to the many and varied signals that arise from ingestion, mastication, digestion, and postabsorption (also see FIGURE 1).

FIGURE 5.

Three mechanisms are activated by glucose as food from a meal passes from the oropharynx to the small intestine. 1: Cephalic phase (brown text and arrows) is the first of three consecutively engaged mechanisms that glucose uses to increase the release of insulin from β-cells in pancreatic islets (red text and arrows). When a meal begins, the action of glucose on taste receptors in the oropharynx leads via central integrative mechanisms (green dashed line and arrow) to the activation of vagal parasympathetic preganglionic neurons in the dorsal motor nucleus of the vagus (DMX). These vagal motor neurons (VMN) innervate intra-pancreatic ganglia (IPG) whose postganglionic neurons then innervate pancreatic islets and release the acetylcholine that potentiates the actions of ambient glucose to increase insulin release from β-cells (brown text and arrows). The light tan box indicates the brain circuitry shown in more detail in FIGURE 4. 2: The actions of incretins constitute the second mechanism (blue text and arrows) that is initiated as food transits within the small intestine where the increasing amounts of glucose that result from digestion act on enteroendocrine K-cells to release glucose-dependent insulinotropic peptide (GIP) and on L-cells to release glucagon-like peptide-1 (GLP-1). GIP enters the hepatic portal vein and then the posthepatic circulation to act as an incretin on β-cells. GLP-1 mainly acts on GLP-1 receptors (solid blue circles) on vagal sensory nerve (VSN) endings found in close proximity to L-cells and in the wall of the hepatic portal vein wall (dark gray box). These VSNs project to the medulla. Most of the GLP-1 released into the hepatic portal vein during normal meals is degraded by the liver and is presumably not the major stimulus for β-cell insulin secretion. 3: The final mechanism (black text and arrows) is the direct action on β-cells of the increased blood glucose concentrations that result from digestion and absorption.

3.3.2. Effects of circulating glucose.

The most important level of β-cell control is the direct action of glucose. Outside of the brain, the homeostatic level for ambient extracellular glucose concentrations is ∼5 mM, which is actively maintained, in part, by basal insulin secretion from β-cells (310, 311). Insulin secretion increases once ambient glucose concentrations rise above this value, and it decreases significantly as blood glucose falls below ∼4.5 mM (312). The relationship between glucose concentration and insulin secretion rate is nonlinear in these circumstances (310). The ability of β-cells to operate as glucosensors to enable this control is very well documented (for reviews, see Refs. 310, 313). They are excitable, and the interaction of glucose with the mechanisms that control their electrical activity intimately couples local glucose concentrations with insulin secretion (314). Paracrine factors released within islets, particularly from alpha cells (315), as well as interactions between β-cells (316) also contribute.

3.3.3. The control of basal insulin secretion.

The foundation of basal insulin control during euglycemia is the network of β-cells within islets. In these circumstances insulin, like many other hormones, is released in a series of ultradian pulses. These are evident in the hepatic portal vein (62–65). During euglycemia in humans, fasted 3-month-old rats, and dogs, each pulse has a period of ∼4–5 min and a peak amplitude of 400–600 pmol/L (∼1400–2,000 pg/mL). This amplitude is at least fivefold greater than in the posthepatic vasculature (63, 65). Increased pulse mass [the total quantity of insulin secreted during an individual pulse (64)] and amplitude occur during hyperglycemia in humans (65) and as rats age (317).

The release of insulin in discrete measurable pulses requires local coordination between β-cells and between islets (62, 318–320). Although incompletely understood, inter-β-cell communication is achieved by a complex network of β-cell gap junctions, paracrine signaling by a host of locally released chemical signals and by electrical coupling (314, 320). The identification of specialist pacemaker β-cells (“hubs”) has provided a component that may coordinate the insulin secretion from individual islets (321), but this remains to be comprehensively tested (322, 323). Also, the way that the interislet coordination (required to enable large discrete insulin pulses) is achieved remains unknown. One theoretical model involves liver-islet interactions that control islet electrical coupling mechanisms (324); coordination via intrapancreatic parasympathetic ganglia is another model (318, 319).

The fact that a multitude of extra- and intrapancreatically derived chemical signals can influence the activity of these ganglia (297, 300) makes it difficult to resolve whether preganglionic parasympathetic and postganglionic sympathetic inputs play any direct role in insulin pulse generation. What is clear, however, is that preganglionic parasympathetic inputs must interact in some way with the intrinsic pulse generation machinery in islets to produce the greater secretion rates seen with CPIR. However, as we discuss later, exactly how this interaction occurs is unknown.

3.3.4. Autonomic innervation of the pancreas and its effects on insulin secretion.

The endocrine pancreas is innervated by parasympathetic and sympathetic motor nerves (300, 325), and by vagal and spinal sensory nerves (VSNs and SSNs, respectively) (300, 325, 326). The SSNs are implicated in nociception and inflammatory processes (326) but apparently play little role in modulating basal insulin secretion (326). VSNs innervate islets and can interact with beta cells (327). The organization of neural inputs to islets and β-cells is complex and species differences are also evident.

To mediate insulin and glucagon secretion, the endocrine pancreas is innervated by the gastric and hepatic branches of the vagus nerve but not its celiac branch (118, 328). These preganglionic axons originate from a large mechanistically heterogeneous pool of neurons in the medial longitudinal columns of the left and right DMX (328–333). Electrical vagal activation has long been known to stimulate insulin and in some species also glucagon secretion (see Ref. 334).

Sympathetic innervation originates in the thoracic intermediolateral column of the spinal cord, with preganglionic axons projecting to the celiac and superior mesenteric ganglia. Postganglionic sympathetic axons then enter the pancreas via the splanchnic nerve.

There are three sites where the brain can influence insulin secretion: intrapancreatic ganglia; islets; and β-cells (FIGURES 5 AND 6). Each location has distinct innervation patterns and sensitivity to a variety of chemical signals (297, 298) thereby enabling an expansive potential for the integrative control of insulin secretion.

FIGURE 6.

A schematic to illustrate how parasympathetic neural drive, incretins, and glucose can engage intracellular signaling pathways in β-cells to increase insulin release. When a meal begins food is digested as it passes from the oropharynx to the stomach and then into the small intestine. Insulin release is stimulated by β-cell signaling pathways associated with three distinct but interacting mechanisms that originate in the oropharynx and small intestine. 1: Cephalic phase insulin release (CIPR) is stimulated by the interaction of glucose with taste receptors in the oropharynx (also see FIGURE 4). These mechanisms then stimulate the release of acetylcholine from vagal preganglionic nerves (i), and then in pancreatic islets from postganglionic parasympathetic motor fibers (brown text and arrows). Some cholinergic terminals are found distal to β-cells (ii), while in some species these terminals are found more proximal and in close apposition to β-cells (iii). Acetylcholine then binds to M3 muscarinic G-protein-coupled receptors (GPCRs) in the β-cell plasma membrane (brown square), which in turn activate membrane-bound phospholipase C (PLC) to catalyze the conversion of phosphatidylinositol 4,5-bisphosphate (PIP2) to diacylglycerol (DAG) and inositol trisphosphate (IP3). DAG-dependent mechanisms then activate protein kinase C (PKC) to increase the phosphorylation of proteins involved with insulin packaging, trafficking, and release (red text and arrows). IP3 increases the release of calcium (Ca²⁺) from intracellular stores thereby increasing insulin release into the hepatic portal vein. Note that acetylcholine-dependent mechanisms in β-cells can potentiate the actions of glucose on insulin secretion without concomitant increases in glucose (also see FIGURE 7). 2: As food is digested in the small intestine increased secretion of incretins from enteroendocrine cells is stimulated by elevated free glucose concentrations (see FIGURES 1 AND 5). During a meal glucose-dependent insulinotropic peptide (GIP) is the principal incretin (blue text and arrows) that binds to GPCRs in the β-cell plasma membrane (blue square) to activate adenyl cyclase (AC) and, via cAMP, protein kinase A (PKA). Downstream signaling from PKA can, like PKC, increase the phosphorylation of proteins involved with insulin packaging, trafficking, and release. This emphasizes the interactions between acetylcholine and incretins to potentiate the actions of glucose on insulin release (also see FIGURE 7). 3: Glucose absorption from the small intestine increases glucose in the blood and pancreatic islets. Glucose (black text and arrows) enters the β-cell via glucose transporters (GLUT) in its plasma membrane; primarily GLUT 1 in humans, GLUT 2 in mice. Glucose, together with certain amino acids and nonesterified fatty acids, is oxidized to release the energy that is used to synthesize ATP. The increasing ATP/ADP ratio then closes KATP channels in the β-cell plasma membrane, which helps to depolarize the β-cell and opens Ca²⁺ channels, thereby rapidly increasing intracellular Ca²⁺ concentrations (small gray box with red border) and insulin release.

3.3.4.1 INTRAPANCREATIC GANGLIA.

After entering the pancreas, sympathetic and parasympathetic axons terminate on neurons in a limited number of intrapancreatic ganglia that are distributed throughout the pancreas (300, 329, 335, 336). In turn, these ganglia innervate a limited number of islets, with both elements being in close proximity (<100 µm) in the mouse and human pancreas (335). Parasympathetic and sympathetic drive can control the release of a diverse set of chemical signals from intrapancreatic ganglionic neurons. These include acetylcholine, norepinephrine, vasoactive intestinal peptide, gastrin-releasing peptide, neuropeptide Y, galanin, pituitary adenylate cyclase-activating polypeptide, nitric oxide, and substance P (297, 299, 300).

3.3.4.2. INNERVATION OF PANCREATIC ISLETS.

In a recent and comprehensive whole organ three-dimensional image analysis of the innervation of the pancreas in mice and humans, Alvarsson et al. (335) found that only a minority of islets are innervated by sympathetic (27%) and parasympathetic nerves (35%) (FIGURE 6). However, these innervated islets are significantly larger than those lacking a nerve supply meaning that a majority of the total pancreatic islet volume receives axons from sympathetic and parasympathetic postganglionic neurons, at least in mice. The duodenal lobes of both species have a greater innervation density than other parts of the pancreas (335).

3.3.4.3. β-CELL INNERVATION.

The number of axon terminals directly apposing β-cells (FIGURE 6) appears to be sparse in mice (335) and sparse to nonexistent in humans and rats (329, 336). This means that direct parasympathetic influences on β-cell function must be largely mediated by the more distal innervation within individual islets (FIGURE 6) in a manner that is less synaptic and more akin to the volume transmission seen for a variety of neuroactive agents in the brain (337, 338). However, the presence of extensive electrical coupling mechanisms between β-cells (314) leaves open the possibility that the effects of innervating only a few β-cells (or even none at all, as is apparent in rats) could be propagated by electrical coupling more widely within the β-cell population of an islet.

At first glance, it appears peculiar that acetylcholine should selectively trigger insulin release from β-cells given the fact that its receptors are also expressed on other islet cells. In fact, cholinergic agonists and activation of vagal motor neurons also trigger the release of PP (339) and in several species (mouse, rat, and dog) even glucagon (e.g., Ref. 340). However, some facts need to be considered in this context. First, human islets consist of ∼60–70% β-cells and 20–30% alpha cells (341, 342), whereas the remaining 10% are delta (somatostatin-producing), gamma (PP-producing), and epsilon (ghrelin-producing) cells, which means that there are many more β-cells than any other cell type. Second, human alpha cells also release acetylcholine (343) and are closely associated with β-cells (342), an arrangement that predisposes the cells to paracrine interactions. Third, several other substances influence the effect of acetylcholine on the β-cells, for instance circulating glucose, incretins (GIP and GLP-1), or peptides that are coreleased from vagal motor nerve terminals together with acetylcholine. Fourth, the effects of glucose differ substantially among the various islet cells, i.e., while glucose triggers the release of insulin from β-cells, an effect that is potentiated by acetylcholine (FIGURE 6) (35), it inhibits glucagon and PP release from alpha and gamma cells, respectively (339). Last, but not least, sympathetic activation or sympathomimetic substances generally inhibit insulin and trigger glucagon release (e.g., Refs. 344, 345), whereas they have scarcely any effect on PP release (339). Thus, acetylcholine has the potential to initiate the release of insulin, PP, and glucagon from pancreatic islets. However, because of the complex morphology, and because glucose, sympathomimetic agents, and various other substances have differential effects on the secretion of these three hormones, the “cocktail” of hormones released differs depending on the specific neuroendocrine and metabolic milieu.

Sympathetic islet innervation appears to be closely associated with controlling islet blood flow in humans (336). However, sympathetic axons in human and mouse islets do appear to be in sufficiently close proximity to alpha and β-cells to directly control glucagon secretion and possibly reduce insulin secretion in a way that can increase glycemia (345, 346).

Taken together, these studies show that the signaling environment within the islet means that vagal motor fibers and, hence, acetylcholine, mainly sensitize the β-cells to glucose (FIGURES 6 AND 7). Acetylcholine released within the islet as a consequence of cephalic phase stimulation, combined with the actions of incretins, prepares the β-cell for the enhanced stimulation generated by postabsorptive glucose (FIGURE 7). Acetylcholine thus facilitates rather than triggers insulin release in these circumstances.

FIGURE 7.

There are three mechanisms that act in pancreatic islets and β-cells to control insulin release before and during a meal. This schematic illustrates the interactions between these mechanisms to control insulin release (red text and arrows), and how they change during a meal. They are as follows: 1: the parasympathetic motor system and the intra-pancreatic ganglia (IPG) (brown text and arrows); 2: incretins (blue text and arrows), primarily glucose-dependent insulinotropic peptide (GIP); and 3: the direct action of glucose on β-cells (black text and arrows). Mechanisms 2 and 3 are activated postabsorptively. The different text and arrow sizes denote relative concentrations and activation levels. Note, this schematic omits the details of signaling mechanisms, together with their molecular components, found in islets and β-cells. See FIGURE 6 and the text for more details of these mechanisms. A: in the absence of significant challenges to blood glucose homeostasis, insulin is released from β-cells primarily in response to ambient glucose to help maintain euglycemia. Influences from the parasympathetic motor system and incretins are minimal. B: as food enters the oral cavity it is chewed and digestion begins. Free glucose is detected by taste receptors in the oropharynx (purple circle) whose cognate primary sensory neurons project to the rostral part of the nucleus of the solitary tract (see FIGURE 4). This process involves integrative circuits, mainly in medulla (green arrow), that rapidly stimulates the parasympathetic motor system to release acetylcholine into pancreatic islets. Acetylcholine then potentiates the ability of glucose to stimulate insulin release, which occurs before the glucose produced from digestion enters the circulation. C: as the concentration of digested glucose and other nutrients increase in the gut, stimulate the release of GIP and other incretins, and move into the hepatic portal vein. The actions of increasing glucose concentrations on β-cells are potentiated by GIP. Because CPIR may well be maintained throughout the meal, the continuing potentiating actions of acetylcholine on β-cells continue to help maintain insulin secretion.

3.3.5. Which of these pathways are involved with CPIR?

Three techniques have been used to investigate how the brain communicates with the pancreas to initiate CPIR: 1) pancreas or islet transplantation, 2) selective nerve stimulation or nerve cuts, and 3) experimental manipulation of the receptors responsible for neurotransmitter/neuromodulator signaling.

3.3.5.1. ISLET OR PANCREAS TRANSPLANTATION.

Pancreas and islets survive transplantation into various sites to support a degree of glycemic control in pancreatectomized or diabetic subjects. Ostensibly, they offer a direct way to examine islet and β-cell function in the absence of neural inputs. This type of experiment has been performed in rats and humans to examine whether pancreas/islet transplants still support CPIR (68, 85, 123, 142). In all cases, controlled oral glucose delivery resulted in a CPIR (preabsorptive) under intact conditions that was abolished by pancreatectomy with transplants. In streptozotocin-induced diabetic rats that had received intrahepatic islet isografts, oral saccharin and water did not produce any insulin response, and the response to intravenous glucose was reduced by ∼30% (68). Similarly, Strubbe and van Wachem (123) reported that the CPIR observed in normal control rats was abolished in streptozotocin-induced diabetic rats with neonatal pancreas transplants under the kidney capsule.

In a follow-up study (124), Berthoud and colleagues showed that continuous sham feeding of a liquid, carbohydrate diet (40% carbohydrate, 29% lipid, 29% protein) containing saccharin (0.06% wt/vol) triggered a substantial CPIR that lasted ∼7 min without any change in blood glucose. Interestingly, prior intravenous administration of atropine methyl nitrate (2 mg/kg) blocked the response and was potentiated by intravenous phentolamine (1.0 or 2.5 mg/kg). These results are supportive of a CPIR that was mediated by cholinergic vagal preganglionic fibers and counterbalanced by a concomitant alpha-adrenergic sympathetic activation. In addition, the authors observed that sham feeding did not increase circulating glucagon, whereas real feeding did. The authors therefore concluded that insulin released as a consequence of the actions of acetylcholine (i.e., CPIR) is not immediately countered by glucagon release or by an increase in hepatic glucose production that is driven directly by adrenergic actions (124). A further study from the same laboratory confirmed and extended these findings by suggesting that CPIR originates from rhrombicbrain/midbrain processing of sensory signals that is modulated by diencephalic inputs (347).

Results from a study that compared kidney-pancreas transplanted patients with control subjects with respect to the occurrence of a CPIR in relation to a palatable pizza meal (142) also support the critical role of autonomic nervous system efferent innervation in the initiation of the CPIR. Controls in this study showed a substantial increase in plasma insulin immediately after meal onset in the absence of any change in blood glucose levels, whereas the kidney-pancreas-transplanted patients did not show any early change in plasma insulin in response to the pizza meal, indicating that an intact, presumably efferent, innervation of the pancreas is necessary for this response. Plasma levels of the incretin GIP also did not change in response to the meal. In addition to insulin, the authors measured the levels of C-peptide and found that it also increased only in controls but to a lesser extent than insulin. Others (e.g., Refs. 73, 142, 348) also observed such a discrepancy between increases in C-peptide and insulin in the peripheral circulation. This is not surprising because the baseline molar plasma level of C-peptide is manyfold higher than that of insulin, which is due to the different tissue distribution and plasma elimination kinetics of both hormones (349) (see also sect. 5.8). Upon stimulation, insulin and C-peptide are released in a 1:1 molar ratio, which can cause a relatively smaller increase in C-peptide despite that insulin is partly extracted by the liver (see Ref. 349).

Considered together these results support the idea that orosensory signals trigger neurally mediated insulin release. However, possible disruptions to the complex intrapancreatic/islet synaptic and paracrine signaling in the transplants, together with some degree of reinnervation, particularly sympathetic fibers, over extended survival periods (350–352) constrain how far this type of preparation can be used to rule out humoral factor contributions or to support unequivocally an exclusive neural drive for CPIR.

3.3.5.2. SELECTIVE NERVE STIMULATION OR NERVE CUTS.

The fact that the vagus nerve can influence insulin secretion has been known for at least 50 years (353, 354). Berthoud et al. (118, 328) later showed in an important set of experiments in rats that selective stimulation of the gastric and hepatic branches but not the celiac branches of the vagus increased insulin secretion into the hepatic portal vein. These findings considered together make the motor portion of the vagus the primary target for investigating how the brain drives CPIR. Since that time, support has grown from a variety of experimental approaches that parasympathetic drive in the vagus nerve is the principal brain output for this function.

The first reports to use nerve cuts found that bilateral subdiaphragmatic vagotomy in rats (67) and bilateral truncal vagotomy or complete pancreatic deafferentation (i.e., parasympathetic and sympathetic) in dogs abolished the CPIR to an oral glucose challenge (90, 91). Selective section of the abomasal (which supplies the ruminant fourth stomach), pyloric, and duodenal branches of the vagus in sheep also suppressed the CPIR to food intake that was evident in sham-sectioned animals without a concomitant rise in jugular vein glucose concentrations (79).

More specific vagal nerve cuts in rats showed that sectioning the gastric and hepatic but not the celiac vagal branches abolished the rapid CPIR to chow ingestion that was evident before any rise in superior vena cava blood glucose could be measured (118). Taken together with the nerve stimulation results from the same group (328), these vagal nerve sections make a strong case that the gastric and hepatic vagal branches carry the critical information from the brain to drive CPIR in rats.

One caveat that should be noted when considering how vagal nerve sections are interpreted has yet to be examined experimentally. Complete section of the vagus or any of its branches also disrupts the flow of sensory information along the VSN from the GI tract and the hepatic portal vein wall into the dorsal medulla. This means that it is not possible to exclude contributions from vago-vagal reflex loops that originate from intestinal incretin release to help potentiate the parasympathetic drive to CPIR (FIGURES 4 AND 5).

3.3.5.3. EXPERIMENTAL MANIPULATION OF AUTONOMIC SIGNALING SYSTEMS.

The early recognition that parasympathetic drive via the vagus nerve was the primary neural effector for insulin release in general, and CPIR in particular, meant that the majority of studies since then have focused on cholinergic mechanisms (reviewed in Refs. 355, 356). Thus, the first use of pharmacology to investigate the neural signaling systems involved with CPIR-targeted parasympathetic drive mechanisms (124). This study showed that intravenous atropine (a muscarinic receptor antagonist) blocked the CPIR in sham-fed rats, whereas it was potentiated by intravenous phentolamine (a nonspecific alpha-adrenergic antagonist), an observation that can be regarded as consistent with the idea that suppressing the sympathetic drive to islets is an important component for controlling the magnitude of CPIR driven primarily by parasympathetic innervation.

Intravenous atropine also attenuated the CPIR by ∼90% and 20% respectively in rhesus macaques (23) and humans (21). However, both studies reported a greater suppression after treatment with trimethaphan, which blocks sympathetic and parasympathetic neurotransmission in autonomic ganglia. These results were interpreted as showing that autonomic nerves provide the major drive for CPIR in humans, but there were significant contributions from noncholinergic signaling mechanisms, e.g., the peptides coreleased by parasympathetic nerves (21). Little progress has apparently been made since then to clarify the role of coreleased peptides in controlling insulin secretion via vagal stimulation. Berthoud and colleagues (328) used hexamethonium (a nicotinic receptor antagonist) to block insulin and glucagon secretory responses to electrical stimulation of the two gastric and hepatic branches but not the celiac branches of the vagus nerve, thereby substantiating a role for vagal preganglionic neurons in controlling insulin release.

When critically evaluating the studies that have examined the effects of manipulating receptor function on CPIR, it is worth noting that the majority have administered agents systemically rather than using specific targeting of the individual components in the motor chain. This general approach is understandable given the technology that was available at the time and has provided useful information about the broad involvement of particular signaling systems. However, it has limited ability to determine exactly how these receptor systems operate within the complex signaling environments that exist preganglionically, or in intrapancreatic ganglia, islets, or β-cells (298, 299).

Two recent studies have addressed this problem with more nuanced approaches. The first used intraocular transplantation of islets in mice to examine the effects that parasympathetic innervation has on islet cell function (357). Because the intraocular parasympathetic islet innervation controls islet function in a light intensity-dependent manner, this novel study concluded the autonomic nervous system acts as a modulator of islet function rather than as a driver. A second study used optogenetic methods to stimulate cholinergic axons in the left cervical vagus or cholinergic terminals in the pancreas of mice expressing ChannelRhodopsin2 in cholinergic neurons (358). Both manipulations significantly increased plasma insulin concentrations, while pancreatic stimulation also rapidly decreased blood glucose, an effect that had not been seen previously using electrical stimulation.

4. CENTRAL INTEGRATION

4.1. Preamble

Despite the large numbers of studies that have characterized the expression of CPIR, how the underlying neural circuitry is organized to enable this outcome remains largely unknown. Beyond the fact that the primary stimulus is clearly gustatory and that this information must somehow engage parasympathetic and possibly sympathetic drive to the islets and β-cells, the intervening steps of this chain within the brain are still obscure. While it is possible to identify likely routes, details are still elusive.

4.2. Cranial Nerves and Their Terminations in the Brain

The routes from taste buds into the brain are very well understood and have been described in detail elsewhere (see sect. 2.3). In summary, FIGURE 4 shows that taste information is conveyed by the CT, GSP, GL, and the SLN, all of which terminate in a rough orotopic manner in the rostral central part (RC) of the rostral (r) NTS⁶ (200–203, 359, 360). The rNTS contains three other subdivisions: the medial (M), ventral (V), and rostrolateral (RL) parts (see Ref. 360). The principal ascending projections from the rNTS to the gustatory part of the PB originate in its RC part, while its V, RL, and M subdivisions send projections, respectively, to the reticular formation, salivary preganglionic parasympathetic neurons, and caudal NTS. These caudally projecting neurons from the M part of the rNTS are the ones that are of greatest significance for CPIR. They are discussed further in sect. 4.4.

4.3. Brain Motor Control Networks for Pancreatic Function

The pancreas was an early target when retrogradely transported neurotropic pseudorabies viruses (PRV) were first used in the 1990s to identify the brain components that control peripheral organ function (361–364). Retrograde transmission reveals the various levels of motor control neurons in the brain for the targeted organ. Using selective pancreatic parasympathetic and sympathetic deafferentation, these studies were consistently able to identify the principal regions in the brain-wide pancreatic control networks.

Of particular interest in terms of CPIR initiation, the studies from Loewy et al. (362–364) remain the most useful because their results are expressed as whole brain maps and tables. They identified the following as key regions retrogradely infected from the sympathectomized pancreas: DMX, PVH, dorsomedial hypothalamic nucleus, parts of the LHA (possibly including the PTSN), the arcuate nucleus, central area of the amygdala (CEA), catecholaminergic neurons in the rostral and caudal ventrolateral medulla and the dorsomedial medulla, and the gustatory and perirhinal parts of the insular cortex. For the most part, these regions coincide with those closely involved with controlling eating behaviors and energy balance (see Ref. 287). However, none of these viral tracing studies found infected neurons in any part of the PB. One other aspect of the study of Buijs and colleagues (361) is worth noting. The NTS and that part of the rostral ventrolateral medulla that contains noradrenergic neurons (the A5 region) form an interconnected network that can influence both sympathetic and parasympathetic drive to the pancreas. If the initiation and magnitude of CPIR are the result of modulating the counterbalanced control of sympathetic and parasympathetic drives, then this modulation may be enabled by interactions between these two regions.

Because the injections in all these experiments were spatially targeted, their major interpretational caveat was that none were able to specify whether the motor control networks identified in the brain influenced the endocrine or parts of the exocrine pancreas, or their circulatory elements. This problem has been addressed in two more recent studies that used a PRV whose replication and retrograde transmission were controlled by β-cell-specific Cre drivers.

The first study (365) used a β-cell Cre driver (MIP-Cre^ERT) that was controlled by tamoxifen exposure. Results were consistent with the earlier PRV tracing studies from Bujis, Loewy, and their colleagues (361–364). However, there was no discrimination in this study between sympathetic and parasympathetic trajectories, and the locations and the degrees of infections of neurons identified in the brain were only reported at low-resolution regional levels. These PRV-tracing experiments were accompanied by functional studies showing that directly manipulating the glucosensing capability in many of the regions identified using Cre-dependent PRV replication could directly influence glycemic control.

A second comprehensive study identified a direct sympathetic drive to Cre-Ins 1-expressing β-cells via the intermediolateral column in the spinal cord that was controlled by oxytocin (OXY)-expressing neurons in the PVH (PVH^OXY) (346). Optogenetically driving these PVH^OXY neurons quickly reduced insulin secretion and increased blood glucose, while silencing them was followed by hypoglycemia. From the perspective of CPIR initiation, however, no PRV-infected neurons were seen in the DMX of these animals, meaning that if CPIR primarily results from parasympathetic drive, then this PVH^OXY projection system has little to no role in its mediation. However, it is not known whether reducing this PVH^OXY-mediated inhibition of the sympathetic drive to β-cells contributes to the magnitude of CPIR by counterbalancing the parasympathetic drive in the manner we just discussed. While investigating the role of vagally mediated parasympathetic drive to insulin secretion has taken precedence for many years, it seems reasonable to also explore a role for the inhibition of sympathetic drive during CPIR, which by its nature, would be a more elusive target for study.

Considered together, all of these viral tracing studies highlight the significant technical difficulties associated with identifying islet-specific parasympathetic control neurons. This problem clouds a detailed dissection of their role in the wider aspects of parasympathetic stimulation of insulin secretion, let alone those that contribute specifically to CPIR. In particular, identifying the specific brain neurons that are directly involved with parasympathetic control of β-cells as opposed to the specific brain regions that more broadly innervate the pancreas may prove to be a formidable technical challenge using neurotropic viruses. This is because only ∼35% of islets are reportedly innervated by the parasympathetic system (335). Notably, this islet innervation only very rarely appears to terminate directly on rodent β-cells (329, 335) in a manner that would facilitate efficient and accurate viral transfer. The majority of β-cells may therefore receive little to no synapse-like parasympathetic innervation, an observation that may help explain the lack of parasympathetic PRV labeling in the experiments of Papazoglou and colleagues (346) using the β-cell-specific Cre-driver Ins-1 to drive PRV replication.

4.4. Integration of Gustatory Information with the Motor Control of Insulin Secretion

An obligatory requirement for the ability of oral stimuli to initiate CPIR is that gustatory information from the oral cavity must, at some locus in the brain, have the ability to influence the neural drive to β-cells. Although this requirement is fundamental to CPIR initiation, the location and nature of this interaction are still unknown.

A strong body of evidence discussed earlier (sect. 3.3.4) points to the DMX as the final common pathway through which gustatory signals eventually trigger CPIR (FIGURE 4). If this is the case, then how is it directed toward these parasympathetic preganglionic neurons in the DMX in a way that rapidly stimulates insulin secretion? Is there also a role for simultaneously inhibiting sympathetic drive to β-cells to potentiate insulin release? Because gustatory afferent fibers terminate in the rNTS (see sect. 4.2), the outputs of these first-order rNTS neurons must be able to engage autonomic premotor control neurons either directly or indirectly to drive β-cells (FIGURE 4).

The fact that decerebrate rats can still show CPIR to orally delivered glucose (281) means that the minimum required circuitry is located somewhere in the midbrain and rhombicbrain (347). However, what this minimum circuitry is and whether it is sufficient for the CPIR seen in intact animals in more complex situations (e.g., with conditioning) is unknown. Decerebration disconnects the forebrain, including the hypothalamus and all regions in the endbrain, from more caudal brain regions. This means that the forebrain is not necessary for the basic expression of CPIR. There are currently at least three unknowns in this model. Is the minimum circuitry that drives CPIR simply a direct projection from the rostral NTS to the DMX (i.e., a reflex)? Does such a projection exist? Is intervening processing elsewhere in the brain required, meaning that the command chain is no longer a reflex?

4.4.1. Integration in the NTS.

For the most part, the various regions identified by the PRV tracing studies from the pancreas that we described in sect. 4.3 do not provide an immediately obvious answer to the existence of direct or indirect rNTS to DMX projections. In two studies where PRV injections were made into the sympathectomized pancreas (362, 363), infected neurons were apparent in what appears to be the terminal zones of the gustatory nerves in the rNTS (Refs. 200, 202, 203, 366, and see Ref. 360) but any functional relationship between these PRV infected neurons and those second-order rNTS neurons that receive taste input is unknown. Evidence for a direct rNTS to DMX projection is certainly not strong (see Refs. 367, 368), meaning that the existence of a gustatory-driven reflex for parasympathetic-mediated insulin secretion is currently unresolved.

Some insight into how rNTS neurons and their projections might be organized to integrate oral sensory information to control autonomic output can be gained from the work of Bradley and colleagues on the secretomotor output of the salivary glands (369, 370). In this model, rNTS neurons, as the first recipient of gustatory signals from the oropharynx, integrate this information before conveying it to preganglionic salivatory neurons (369, 370). It is tempting to speculate that the rNTS may perform a similar function with respect to CPIR, although details of the effector projections remain unclear.

The most detailed descriptions of how the NTS controls DMX function, including insulin secretion, is by way of a heterogenous array of local connections from the adjacent NTS. These include neurons that use a variety of chemical signals at their synapses with DMX neurons, including GABA, glutamate, nitric oxide, GLP-1, or catecholamines (332, 333, 371–378).

Given that activating GABAergic inputs from the NTS to the DMX increases both glucagon secretion and blood glucose in the absence of a change in insulin secretion (373), it seems possible that activating glutamatergic inputs and/or inhibiting GABAergic inputs to the DMX from the NTS are significant factors in mediating the parasympathetic drive for CPIR. With this in mind, perhaps the most parsimonious circuitry for how gustatory information reaches DMX neurons from the rNTS is that transmission is mediated by way of extended intra-NTS projections that control the activity of those NTS neurons that modulate DMX output (FIGURE 4), but if this is the case, a detailed picture of any intra-NTS connections that can enable this means of communication is currently lacking. It should be noted, however, that there are reports alluding to connections from the medial part of the rNTS to more caudal regions of the NTS (360, 368). These may be capable of modulating the activity of those NTS neurons that are known to control DMX function. Although the medial part of the rNTS from where these projections originate does not receive direct inputs from gustatory nerves, the microcircuitry of this region is structured in a way that could enable the transfer of gustatory signals into the medial part of the rNTS (379). Given the potential importance of these intra-NTS projections for CPIR, a more detailed examination of the rostral to caudal NTS projections (or potential rNTS projections to the DMX) would be very informative.

4.4.2. Integration in the forebrain.

Given the fundamentals of how autonomic control is organized by the brain (380, 381), the role of particular forebrain regions as modulators of motor control networks in the hindbrain and medulla seems a reasonable model to use for CPIR. We have already discussed how a basic circuit for controlling CPIR may exist in the NTS and DMX (FIGURE 4), so the question now arises about how this circuitry is modulated by projections from the forebrain and where in the forebrain are the opportunities for interactions between regions that process gustatory information and those involved with pancreatic control.

Once gustatory information moves rostrally from the rNTS, the opportunities for the types of interactions with regions that can directly influence parasympathetic drive become less obvious. In rodents, gustatory information passes from the first-order neurons in the rNTS to the gustatory parts of the PB. From there, it is conveyed in two directions. First, to the thalamus and then the gustatory cortex. However, as noted in sect. 4.3, based on evidence from pancreatic injections of retrogradely transported PRV in rodents, neither the PB nor the gustatory thalamus are regions that are closely associated with the brain’s motor control network for the pancreas. Second, subcortically to parts of the bed nuclei of the terminal stria, the amygdala, and the hypothalamus. In the primate brain, gustatory information bypasses the gustatory PB completely and is conveyed directly to the thalamus and then gustatory cortex, which means opportunities for integration along these routes are possibly even more sparse than those in rodents (360, 382, 383).

Two facts relegate any discussion of potential loci of integrative mechanisms in the forebrain to conjecture. First, there are few prominent points of overlap between known sites of gustatory processing within the brain and regions clearly identified by PRV infections from the pancreas. Second, there have only been very few traditional forebrain lesion studies that reported the effects on CPIR or conditioned insulin release (40, 68, 296, 384, 385), and none that have used gene-targeted methods. However, by virtue of their position in the motor control network for the pancreas (361, 362), three regions seem to be reasonable starting points in this regard: the insular cortex, the CEA, and the PTSN in the LHA. All three are to some extent interconnected (386, 387) as well as having projections to the NTS and/or DMX (295, 387, 388). Whether or how these regions, or indeed others like the ventromedial hypothalamus (40, 68), are involved with the expression or modulation of CPIR awaits rigorous experimental dissection of these circuits. Likewise, more studies are warranted to follow up on recent findings in mice suggesting that IL-1β released from hypothalamic microglia contributes to the mediation of the CPIR (126).

5. CRITICAL POINTS AND OPEN QUESTIONS

5.1. Preamble

Up to this point, we have reviewed what is known about the relevant features of the sensory and motor limbs as well as the potential central links involved in mediating the CPIR. In this section, we discuss critical test conditions that are required to ensure that the stimulus can reach and activate the relevant taste receptors in the oral cavity to initiate a neural signal that triggers a CPIR (FIGURE 8). We also discuss the possible roles of sweetness and palatability as crucial stimulus features because several studies suggest that they may be important. Because it has been known since Pavlov’s work that conditioning plays a major role in several different cephalic phase responses, it seems logical that this is also true for insulin release. We therefore weigh the evidence for and against the role of conditioning in the CPIR. This is an important question because it also relates to its physiological relevance. Moreover, as the CPIR, according to its traditional definition, reflects an early release of insulin, we touch upon the general difficulties to accurately assess insulin release, i.e., we consider factors and conditions that may prevent the detection of a CPIR as traditionally defined, and we evaluate the reliability and usefulness of various approaches to overcome these measurement problems. Finally, we weigh the evidence of the physiological relevance of the CPIR in light of the pathophysiological consequences of its absence or elimination. All these points are relevant with respect to the interpretation of available data, and their thorough consideration reveals several open questions and problems, which may guide the way for future research.

FIGURE 8.

Section 5 discusses nine critical points and open questions about cephalic phase insulin release (CPIR). FIGURE 8 shows how these nine topics (sects 5.2 to 5.10) relate to the processes that enable the information flow from the oropharyngeal sensory systems (purple) that begins with increased amounts of glucose in the oropharynx during a meal (black), and ends with the initiation of insulin release from the β-cell (red) in response to cholinergic stimulation from parasympathetic innervation (brown). The color scheme used here is the same as that shown in FIGURE 4 and others in this review.

5.2. Glucose as the Adequate Stimulus

The stimulus conditions under which a CPIR is and is not observed remain elusive as the title of this review implies. For the most part, however, in both animal models and in humans, glucose appears to consistently work (TABLE 1). However, in most studies, the concentrations used are quite high approaching 1.0 M or even double that. As noted above, Yamazaki and Sakaguchi (92) found that, in humans, while glucose was an effective stimulus, the CPIR was concentration-dependent and quarter molar concentrations of the sugar were insufficient to generate the response. Accordingly, even with what appears to be perhaps the optimal chemical stimulus, although more studies using glucose would help confirm such a claim, there are limitations to its effectiveness to orally trigger the response. It follows, then, whether disaccharides, oligosaccharides, and more complex foods that have a source of glucose potentially liberated upon enzymatic hydrolysis are effective, might depend on the amount of free glucose produced in any given stimulation protocol. Indeed, Glendinning et al. (80) found that in mice, glucose-containing saccharides such as sucrose, maltose, and maltodextrin could all produce a CPIR when low volumes were orally ingested, but the response was eliminated by adulterating the stimulus solutions with acarbose, which inhibits the activity of α-amylase and α-glucosidases. Unfortunately, because the acarbose blocked digestion of sugars postabsorptively as indicated by an elimination of any rise in blood glucose after ingestion, it remains unclear whether the enzyme-inhibiting effect represents an oral or postoral site of action. In humans, salivary amylase activity was found to be associated with preabsorptive insulin release during ingestion of 500 mL of 10% cornstarch solution, while the CPIR produced by 10% glucose did not differ between participants with high and low salivary amylase activity (95). This is worthy of further investigation using a sham-feeding preparation in animal models as well as the MSF in humans in which glucose-containing stimuli are presented with and without the addition of acarbose. To the extent that it is an oral event, the relative amount of glucose produced in the initial stages of hydrolysis in the mouth can potentially make a difference and may be dependent on the specific features of the stimulus delivery procedure.

If glucose has a privileged role in stimulating a CPIR, there are several conundrums that remain to be understood with respect to how such specificity is neurally conferred. Although there is growing support for the existence of some potential glucose-sensing mechanisms in taste receptor cells, these tend to be expressed with the T1R3 protein and so it is unclear how the synaptic output of the cell differentiates between glucose and other sugars. Likewise, while there is some evidence of T1R3-independent neural transmission of input arising from stimulation with sugar solutions, there are still questions about how the brain distinguishes glucose from fructose and other sweeteners.

5.3. Adequate Stimulation of Oral Taste Receptor Fields

Nevertheless, even if an adequate stimulus is present, it must reach the critical receptors. It remains unclear whether all taste buds, and their oral fields, which are innervated by different gustatory nerve branches, have equipotentiality with respect to their contribution to the CPIR. There is strong evidence, at least in the rat, that the neural input transmitted by the different gustatory nerves does not contribute equally to all taste functions [see Spector (178); Spector and Glendinning (265)]. Therefore, if the stimulus is not sufficiently reaching the relevant taste receptors, then a CPIR might be weak or absent. In this regard, the MSF procedure, which has the distinct interpretive benefit of preventing stimulus contact with postoral sites, may also equally prevent stimulus contact with posterior tongue taste receptors. The circumvallate and foliate papillae are positioned quite posterior in the tongue and the degree to which they are stimulated when the act of swallowing does not occur is most likely less than optimal and even minimal. That would mean that the field that contains the most taste buds (in the rat ∼60% of total) and their associated signals in the glossopharyngeal nerve are inadvertently taken out of the equation so to speak. For example, if light were only shined on the fovea, the functional properties of vision would be quite different from what would be inferred from the same light restricted to the more peripheral parts of the retina. Even in cases where restricted stimulation of the anterior tongue or MSF produces a CPIR to a given stimulus, it could perhaps be more (or less) effective if the stimulus has access to the posterior tongue. For instance, Yamazaki and Sakaguchi (92) placed gauze saturated with different concentrations of glucose ranging from 0.28 M to 2.22 M on the tip of the tongue for 1 min and reported that the higher concentrations triggered a CPIR, but the lower one did not. It seems possible that the lowest concentration would have been effective had it stimulated a broader range of taste bud fields. Of course, this remains speculation, but the relative lack of accessibility of the stimulus to the posterior tongue in many stimulation protocols could be at the root of some of the disparities in outcomes in the literature.

5.4. Swallowing

Minding the caveat that negative outcomes often go unreported, we were struck by the fact that virtually every published study with animal models included positive results with at least some orally sampled stimulus, whereas the results associated with human studies were more equivocal. There are several reasons why this may be so, including the potential for prior experience with a multitude of foods to affect the outcomes of studies with human participants (see sect. 5.7). However, one of the most salient differences between studies with humans and those with animal models is that in the latter case the stimulus is always swallowed, whereas in the former case, the stimulus is often expectorated (i.e., MSF) to circumvent the interpretive confound of direct postabsorptive stimulation of the β-cells. With respect to the latter issue, most (but not all) animal model studies, focused on the early insulin response to the oral stimulus before any rise in blood glucose was detected and a couple actually included a sham-feeding preparation in which the ingestants dripped out of an open gastric cannula (see TABLE 1). Not only would this likely increase the range of stimulation of taste bud fields to include the posterior tongue, but it would also cause stimulation of the somatosensory afferents of the oral, pharyngeal, and esophageal epithelium and include the reafference signals of the swallowing musculature. Arguably, there is not a more reliable predictor of an impending postingestive load, almost always including glucose under natural eating conditions, than that. Obviously, some MSF studies were able to generate a CPIR despite that no swallowing occurred and so this final ingestive act is not a strict necessity. That said, the act of swallowing could potentially be a factor that determines the breadth and robustness of a CPIR. This is a possibility that is worthy of further investigation.

5.5. Sweetness as a Critical Stimulus Feature?

There is a predilection in the literature for authors to assume that sweetness is the critical sensory attribute that triggers a CPIR. The uncontested fact that one of the primary functions of insulin is to promote a euglycemic state and that humans report glucose as possessing a “sweet” taste quality likely provides the basis for this line of thinking. Of course, as the findings discussed in the preceding pages reveal, the evidence supporting such a view is mixed. In some cases, fructose and nonnutritive sweeteners are effective stimuli in studies with either animal models or humans, and, in other cases, they are not. At least some complex stimuli (e.g., cocktail sandwiches, peanut butter, pizza) for which it would be difficult to claim sweetness as the dominant qualitative taste perception (although admittedly this is a point that would have to come under strict psychophysical investigation) can generate a CPIR. Within the context of the theoretical framework regarding the multidimensionality of taste function discussed above, it is not axiomatic that an effective stimulus for a CPIR even be perceived.

5.6. Palatability as a Critical Stimulus Feature

While a consistent theme in the literature is the possibility that the palatability of the stimulus is an important feature that determines the presence or extent of a CPIR (e.g., Refs. 25, 40, 128), the results in support of this view are equivocal. Lucas et al. (155) found that a meal of palatable foods, as rated by the participants, resulted in a detectable rise in plasma insulin levels before any increase in blood glucose in at least some participants; meals with less palatable food were much less effective, if at all. In contrast, Teff and Engelman (29) found that both palatable and unpalatable foods were equally capable of eliciting a CPIR to MSF in women. Whereas Taylor and Feldman (149) reported that an “appetizing” steak dinner that was sham fed to normal-weight participants was ineffective at stimulating an insulin release even though PP levels increased. Eight patients with duodenal ulcers (389) did not display an increase in plasma insulin (15–120 min) in response to MSF of an “appetizing” meal of steak and potatoes for various durations (4–30 min), but the potential complication of the condition of the small bowel cannot entirely be dismissed. Given these disparities in the literature, it is difficult at this time to draw any firm conclusions about the role of palatability in modulating CPIR.

5.7. The Potential for Conditioning to Influence CPIR

Pavlov clearly showed that digestive reflexes can come under conditioned control (13, 16). Accordingly, we would expect the CPIR to be subject to associative learning. Given the diversity of foods consumed over a lifetime in humans coupled with the fact that virtually every meal causes a rise in blood glucose, except for some specific beverages and snacks, it is surprising that the mere sight, smell, taste, and texture of food does not always produce a CPIR regardless of whether swallowing occurs. Rodent models, which are normally maintained on a single chow diet, might be more amenable for tracking down the unconditioned stimulus attributes that lead to an unconditioned CPIR. Moreover, given their standard and unvarying dietary maintenance conditions, rodents would be excellent subjects for studying the ability of novel neutral stimuli to lead to conditioned insulin release. Unfortunately, there have been only a few explicit attempts to study the conditionability of the CPIR, but there are some studies that provide support. It has been shown in rats that even the anticipation of a meal at a specific time of the day can lead to an insulin release (25, 390) that is blocked by atropine suggesting that the hormonal secretion is a vagally mediated conditioned response (390). There are examples of conditioned hypoglycemia to not only saccharin but to non-“sweet” stimuli as well as even odors (113, 391); the mechanism, while not explicitly investigated in these studies, presumably involves insulin release. When a neutral conditioned stimulus is paired with an insulin injection, but the resulting hypoglycemia is prevented by intraperitoneal glucose administration, the conditioned stimulus is still effective at reducing blood glucose after training (392). Indeed, a complex light/tone/odor stimulus that was paired with food presentation was capable of eliciting a significant increase in plasma insulin after training sessions, and this response was eliminated by electrolytic lesions bilaterally placed in the ventromedial hypothalamus or by bilateral subdiaphragmatic vagotomy (296). Although the neural manipulations were relatively imprecise by today’s standards, these findings nonetheless support that the effect is dependent on the brain.

In an early study by Berthoud et al. (68) in which 0.15% saccharin was just as effective a stimulus for a CPIR as was 50% glucose, the rats were first adapted to the oral presentation of the stimuli over a 1-wk period, which means they were experiencing 50% glucose on some occasions and 0.15% saccharin on others potentially setting the stage for some conditioning to occur. That said, Berthoud and Powley (118) found saccharin to be an ineffective oral stimulus for insulin release despite that the animals had ample prior exposure to sweetened condensed milk diet. Ninomiya et al. (89) reported that the amplitude of a CPIR to 0.125% saccharin decreased across repeated tests; the authors interpreted this as potential evidence of extinction, but it is not entirely clear how the response was conditioned in the first place. Tonosaki et al. (97) found that along with sucrose, saccharin was an effective stimulus for a CPIR. However, the order of stimulus testing was unclear, and if sucrose were tested first, then it seems possible that the saccharin-induced insulin release was a conditioned response.

Berthoud et al. (347) presented three rats with three sessions of a glucose + saccharin mixture which led to a clear CPIR. They then presented the saccharin alone for 10 sessions and found that the CPIR to this nonnutritive sweetener was stable. This outcome could be interpreted as suggesting that the saccharin was either an unconditioned stimulus or a conditioned stimulus resistant to extinction. The fact that other animal studies were unable to demonstrate a CPIR to saccharin (74, 80, 118) or to sodium cyclamate (76) would favor the latter interpretation.

In humans, Bellisle et al. (128) observed spontaneous cyclic oscillations of basal plasma insulin around lunchtime when no food was presented emulating what was found in animal model studies (25, 390) Another study with humans found that when an odorant was paired with intravenous insulin administration on four separate trials, the presentation of the odorant alone was capable of generating a decrease in blood glucose (393, 394), an outcome consistent with the conditioned hypoglycemia seen in the animal experiments (113, 390, 395). However, in a similar design, pairings of the same odorant with intravenous glucose (or insulin) injections were insufficient to lead to a CPIR. Nevertheless, the fact that insulin can, under certain circumstances in at least animal studies, be conditionally released by a neutral stimulus that has been paired with food as well as the fact that glycemia can come under Pavlovian conditioned control leaves open the possibility that some of the inter- and intrastudy variability in CPIR may be due to differential prior experience with the cephalic stimulus. On the other hand, Dhillon et al. (17) reported that a 2-wk training period in which participants were explicitly exposed to either sucrose or sucralose in daily sessions did not influence the response to the other sweetener after the 2-wk exposure. In other words, there was no evidence of extinction of the CPIR to sucrose and no amplification of the response to sucralose.

Notwithstanding the results of Dhillon et al. (17), it would seem that human studies, in which dietary history cannot be explicitly controlled, are most vulnerable to the potential for conditioning to confer potency to certain stimuli, such as nonnutritive sweeteners, to elicit an insulin release. At issue is, whether some orally delivered compounds, such as glucose, are unconditionally capable of triggering a CPIR, while other cephalic stimuli are only effective when sufficiently paired with increases in blood glucose or insulin. To the extent to which a CPIR to a given stimulus is associatively conditioned, the possibility of extinction episodes must also be considered. Accordingly, the potential for learning and extinction to influence a CPIR to at least certain stimuli poses challenges to finding unifying outcomes across studies.

Regarding the conditionability of the insulin response to taste stimulation, Berridge et al. (88) demonstrated that a preabsorptive insulin release to a 50% glucose (or to 15% fructose or 15% maltose) solution in rats could be eliminated by pairing oral infusion of the sugar stimulus with LiCl administration, which causes nausea in humans and resulted in a conditioned aversion to the taste stimulus. Ninomiya et al. (117) replicated this outcome with 50% glucose and showed that familiarity with 0.125% sodium saccharin before conditioning did not dampen the aversion or the CPIR to glucose. One caveat regarding the insulin response is that blood glucose levels were not measured and so it is difficult to differentiate between pre- and postabsorptive factors in triggering the response. On the other hand, saccharin (0.125%), which should not ostensibly increase blood glucose levels, unconditionally triggered a CPIR, which was mitigated when a taste aversion was conditioned to this nonnutritive sweet stimulus. Interestingly, a conditioned taste aversion to saccharin did not eliminate a subsequently tested CPIR to glucose and vice versa suggesting that glucose and saccharin are both capable of eliciting a CPIR, but not necessarily through a shared perceptual feature (i.e., “sweetness”; see sect. 5.5). In sum, not only might it be possible for neutral stimuli to be conditioned to elicit an orally triggered secretion of insulin, but it also appears that an unconditionally adequate taste stimulus like glucose can be conditioned to be ineffective. This is worthy of further investigation because it is an example of how learned associations can override seemingly hardwired physiological reflexes. For example, through learning, could the secretion of saliva elicited by the acid taste of a lemon be abolished?

5.8. The Difficulties Assessing Insulin Release

The classic definition of the CPIR implies that the characteristic peak in circulating insulin within the first minutes of eating reflects an early release of the hormone. While this cannot be taken for granted (see below), it is also difficult to accurately assess insulin release for several reasons. First, insulin is released into the hepatic portal vein, and a substantial and varying amount (in general, 40–90%, up to 80% in humans) is extracted by the liver upon first pass by a receptor-mediated process (e.g., Refs. 396–398). Second, the pattern of insulin delivery into the hepatic portal vein, i.e., the secretory pulses (see sect. 3.2.3), is a major determinant of hepatic insulin extraction (398). An important function of insulin in the liver is the activation of glucokinase (48). Phosphorylation traps the glucose inside the cells and is the first step of glycogen synthesis or glycolysis; it also keeps the intracellular concentration of free glucose low, thus promoting passive hepatocellular uptake of glucose through GLUT2. As might be expected given this function, hepatic insulin extraction and action, increase in response to enteral or parenteral glucose administration (see Refs. 396, 398). Glucose-induced stimulation of insulin release is characterized by increased pulse mass (65, 317). As it is the insulin pulse mass that mainly determines the hepatic extraction (398), a major part of the early insulin release presumably ends up in the liver. This means, however, that even a substantial early increase in insulin release may not be fully reflected in the peripheral circulation. Therefore, peripheral measurements of plasma insulin, the only way to routinely measure circulating insulin concentrations in humans, generally provide only a rough idea of how much insulin is released into the hepatic portal vein and may easily miss a CPIR.

Experimental insulin measurements in the hepatic portal vein, which are possible in rodents and other animals, avoid the dilemma of extraction by the liver, but are subject to changes in blood flow and are still hampered by the pulsatility of the hormone’s release. Thus, in rats and humans the blood flow in the hepatic portal vein, as estimated by various methods, increases during eating by ∼100% (e.g., Refs. 399–401). This means that the secretion of insulin must already increase substantially to just maintain the basal insulin concentration in the hepatic portal vein. Moreover, and as mentioned earlier (see sect. 3.2.3) insulin is released in a pulsatile (period: 4–5 min) fashion (62–65), and the pulses in the hepatic portal vein are at least 5-fold greater than in the posthepatic circulation (63, 65). The problem of the pulsatility of release and its consequences for the measurement requirements and the interpretation of results was in fact emphasized in at least two CPIR studies in humans (e.g., Refs. 20, 128). Both studies observed oscillations of basal peripheral plasma insulin concentrations of similar periods (12–20 and 12 min, respectively) and amplitudes in the range of 1–10 µU/mL. One of these studies (128), in which insulin levels were measured continuously, observed a CPIR, while the other one (20) did not. Of course, the substantial amplitude of the basal plasma insulin concentration means that any CPIR must be greater than the baseline pulsatility to be reliably detected. Interestingly, Abdallah and colleagues (20), who did not report a CPIR, still observed postexposure decreases in plasma insulin and blood glucose levels. One possible explanation is that there was an increase in insulin that was undetectable peripherally because of its inherent pulsatility and because it was filtered out by the liver, where the insulin caused an inhibition of glycogenolysis (49, 53, 54) and, hence, contributed to the postexposure decrease in blood glucose. In any case, the pulsatility of release and the variable hepatic extraction together with the changes in blood flow make it very difficult to accurately measure short-term releases of insulin because they add substantial variability to any measurements, which can easily prevent the detection of a CPIR.

C-peptide is cosecreted with insulin in a 1:1 molar relation (402) and, unlike insulin, is not extracted by the liver (see Ref. 349). Particularly in humans, in whom hepatic portal vein insulin measurements for experimental reasons are virtually impossible, peripheral C-peptide levels are often considered to be a good indicator of pancreatic insulin release (e.g., Refs. 29, 73, 403). Unfortunately, however, the tissue distribution and elimination kinetics of both hormones differ substantially (349), which complicates the estimate of insulin release from the C-peptide levels. For instance, as C-peptide has a much longer biological half-life than insulin and accumulates in the circulation (404), it is difficult to detect secretory pulses. Frequent sampling, very accurate measurements of C-peptide, and sophisticated mathematical methods (stochastic nonparametric deconvolution) are therefore necessary to accurately estimate insulin release from C-peptide measurements (62, 348, 398). Nevertheless, peripheral C-peptide levels have routinely been measured to that end in the context of CPIR: For example, in one study Teff and colleagues (73) used a peanut butter sandwich as a complex oral stimulus (see sect. 2.2.3) and observed similar early (0–10 min, with a peak around 4 min) increases in insulin and C-peptide in an MSF condition or when the participants were allowed to eat the sandwich. In another study that addressed the effects of oral stimulation together with intragastric glucose infusions on glucose tolerance, again using the peanut butter sandwich as an oral stimulus, Teff and Engelman (29) reported similar increases and AUCs for C-peptide and insulin during the first half of the 195 min sampling period, when the intragastric glucose infusion was combined with the oral stimulus, compared to the intragastric glucose infusion alone. Reflecting the positive effect of oral stimulation on glucose tolerance, plasma glucose levels were lower, and the AUC was smaller with the oral stimulation. The comparison of absolute changes in peripheral plasma concentrations usually reveals cephalic phase C-peptide responses that are greater than the CPIR (see Ref. 403). If the relative changes of both hormones to the corresponding baseline levels are compared, the cephalic phase C-peptide response is mostly smaller than the CPIR (e.g., Refs. 71, 142, 348). At first glance, this appears to be strange given that some of the released insulin is extracted by the liver whereas C-peptide is not. However, the baseline molar plasma level of C-peptide is substantially higher than that of insulin (405) because of the different tissue distribution and plasma elimination kinetics of both hormones. Despite the first pass hepatic insulin extraction, an equimolar release of insulin and C-peptide therefore often increases insulin in peripheral plasma more than C-peptide relative to their respective baselines. Finally, two studies employing an MSF design did not report any cephalic phase response for either insulin or C-peptide (57, 134). Interestingly, however, in one of these studies (57), the authors observed that the MSF improved glucose tolerance, which was supposedly due to a reduced hepatic glucose output. Given all the limitations mentioned above and the fact that a shift from hepatic glucose output to glucose uptake is exactly what should be expected from insulin (e.g., Refs. 49, 53, 54), it cannot be excluded that there was an early release of insulin in this study that was simply missed in the peripheral blood samples. In sum, although C-peptide levels can provide some measure of insulin release (403), they are apparently less reliable than often assumed.

5.9. PP as an Indicator of Vagal Efferent Stimulation

As mentioned above, methodological and practical difficulties often hamper the assessment of the CPIR, as traditionally defined, in humans. As vagal efferent stimulation supposedly is a critical component of the CPIR, investigators have often turned to measuring PP in addition to insulin. PP inhibits pancreatic exocrine secretion as well as gall bladder emptying (406) and has long been implicated in the control of eating and regulation of body weight (see Ref. 407). Because vagal efferent activation is a major stimulus of PP release, an increase in circulating PP is often considered a reliable marker of vagal efferent activity that should also trigger the early insulin release during a meal, i.e., the CPIR. As the increase in plasma PP in response to vagal efferent stimulation is usually larger than the increase in circulating insulin (14), PP measurements may even allow for the detection of graded responses that might not be discernable based on changes in insulin. The downside of this, however, is that a vagal efferent activation that triggers the release of PP may simply not be potent enough to also trigger the release of insulin. At first glance, this appears somehow strange because in humans the β-cells account for 60–70% of all islet cells. The α-cells account for 20–30%, and Δ-, γ-, and ε-cells (which produce somatostatin, PP, or ghrelin, respectively) account for the remaining 10% of the pancreatic islet cells (341, 342). Therefore, other factors than the number of cells or their surface area must account for this differential effect. As mentioned before (see sect. 3.3.4), acetylcholine sensitizes the β-cells to glucose (36), and insulin release in response to vagal efferent activation depends heavily on the presence of glucose (35), whereas PP release is inhibited by glucose (339). At the beginning of a meal, before any increase in circulating glucose, an activation of vagal motor fibers may therefore primarily stimulate PP release. This fact presumably also contributes to the observed failure of antral distension to trigger the release of insulin, despite a massive stimulation of PP secretion (408). Moreover, unlike the release of insulin (409), the release of PP is scarcely inhibited by sympathetic activation (339). The immediate sympathetic and parasympathetic activation by eating (5) may therefore also contribute to the differential dynamics of prandial insulin and PP increases. Finally, there is no hepatic extraction of PP (410), whereas hepatic insulin extraction in humans reaches 80% (398). All these factors combined in essence must increase the relative amount of PP compared to insulin in the peripheral, posthepatic circulation. Perhaps not surprisingly given these facts and complex interactions, in several studies increases in circulating PP levels were not associated with a CPIR (121, 138–140, 147, 149, 411).

In addition, the correlation between vagal efferent activation and increases in plasma PP is not as clear as generally assumed. A CPIR can apparently happen without any concomitant increase in peripheral plasma PP levels (71). Moreover, Veedfald and colleagues (412) measured plasma PP levels in truncal-vagotomized patients and controls in response to an oral glucose tolerance test (OGTT) and found that oral glucose-stimulated PP secretion similarly in both groups. This clearly indicates that ingested glucose can stimulate PP release independent of an intact vagal innervation. Although the experiment did not specifically look at a CPIR because the earliest blood sample was taken 15 min after the OGTT, the data still show that a potent stimulation of PP release is possible without vagal efferent activation.

In sum, although increases in circulating PP presumably reflect mostly an activation of vagal motor fibers, increases in insulin and PP may be dissociated for several reasons, which must be kept in mind for the interpretation of PP data.

5.10. CPIR, a Peculiarity or Physiological Necessity?

As discussed earlier, during a meal, hepatic portal vein blood flow increases dramatically, sympathetic stimulation occurs together with the activation of vagal motor fibers, and blood glucose levels are still low. Given all these factors that counteract a detectable increase in circulating insulin during the early phase of a meal, it is almost surprising that a CPIR, as originally defined, is observed occasionally, in fact even more often than occasionally, in humans and apparently most of the time in rodents. Of course, the latter statement requires some caution; the picture may be distorted because it is usually very difficult to publish negative results from animal experiments. In any case, and as mentioned in the beginning, the preabsorptive increase in insulin before any changes in blood glucose levels is mainly an interpretive demand that allows for the dissociation of the site of stimulation under many experimental conditions, but it does not reflect physiological relevance. In other words, the absence of a detectable CPIR that fits the standard definition should not be misinterpreted as indicating that there is no cephalic phase of insulin release or that it is not physiologically relevant.

In fact, strong evidence indicates that the cephalic stimulation of vagal motor fibers targeting the pancreatic islets is physiologically important for prandial and postprandial glucose homeostasis (e.g., Refs. 21, 23, 29, 56, 130). The preabsorptive increase in circulating insulin is usually small compared to its overall prandial and postprandial release, but the oral stimulation continues throughout the meal. D’Alessio and colleagues (23) showed that blockade of the autonomic nervous system control of the β-cells by infusion of either the muscarinic blocker atropine or of the ganglionic blocker trimethaphan reduced the overall insulin response to a mixed nutrient meal in rhesus macaques by ∼75%, despite that the concomitant increases in blood glucose and plasma incretin levels were unaffected (23) (FIGURE 9). This provides impressive evidence for the crucial role of the continuous autonomic nervous system input to the pancreatic β-cells for the meal-induced prandial and postprandial rise in circulating insulin. Although it cannot be excluded that these pharmacologic treatments also eliminated some of the incretin effect on insulin release based on the potential activation of vago-vagal reflexes (413), a significant part of their effect was presumably due to the elimination of cephalic phase autonomic input.

FIGURE 9.

Plasma concentrations of glucose (black), insulin (red), and glucose-dependent insulinotropic peptide (GIP; blue) before, during, and after test meals in six adult rhesus macaques infused with saline (top) or atropine (bottom) from 20 min prior to 60 min after the onset of a test meal consisting of one banana, half of an apple cut into four pieces, and three monkey biscuits. During saline infusion, plasma levels of insulin, glucose, and GIP increased from baseline levels continuously throughout the meal and remained elevated thereafter. As indicated by the bigger symbols and asterisks, the increase in insulin was already significant at 2 min, whereas the increases in glucose and GIP did not reach significance before 10 min into the meal. During atropine infusion, the prandial increases in glucose and GIP were similar as during saline infusion (both increases over baseline significant after 15 min), but the plasma level of insulin remained virtually unchanged and did not increase significantly over baseline during the 60 min of the infusion, indicating that a substantial part of the overall insulin response to the meal was mediated by cephalic stimulation. See text for further details. Data are from Ref. 23.

Moreover, even a brief and transient cephalic stimulation can have delayed effects on metabolism. Teff and Engelman (29) showed that in men who received a 75-g glucose load intragastrically delivered over 5 min, MSF of a peanut butter sandwich during that period significantly lowered the area under the blood glucose curve in the following 195 min by more than one-third (FIGURE 10). The oral stimulation condition led to significantly higher plasma insulin levels (about 30%) during the first 75 min after load administration compared with controls. In this strategy, rather than attempting to measure an early and relatively small rise in insulin as is done in most studies of CPIR, Teff and Engelman (29) assessed the contribution of oral stimulation to glucose tolerance. Together, these effects document that the cephalic phase contribution to the total insulin secretion in response to eating is much greater than commonly assumed, and they presumably account for the substantial influence of the CPIR on postprandial glucose metabolism (e.g., Refs. 26–28).

FIGURE 10.

Effect of oral sensory stimulation on plasma levels (mean ± SE) of glucose (black), insulin (red), and C-peptide (green) of eight normal-weight men after administration of 75 g of glucose intragastrically, with and without oral sensory stimulation. Time 0 indicates onset of glucose administration (beige bar) and sensory stimulation. Dashed lines represent control condition (no sensory stimulation), and solid lines represent sensory stimulation condition. Inset: areas under the curve (AUC) for conditions 1 and 2 (hatched bar, control condition; solid bar, sensory stimulation condition). *P < 0.05, **P < 0.03, significant differences between conditions. The oral stimulation led to significantly higher plasma insulin levels during the first 75 min after glucose administration load administration and subsequently to significantly lower glucose concentrations, indicating better glucose tolerance compared with controls. Figure is modified from Ref. 29, with permission from the American Physiological Society.

Vagal efferent activation and, hence, acetylcholine mainly sensitizes the β-cells to glucose (FIGURES 6 AND 7). Acetylcholine thereby prepares the β-cell and thus facilitates rather than triggers insulin release. This sensitizing action is 1) consistent with the similar effect of incretins (FIGURES 6 AND 7) and 2) is presumably important because it reduces the risk of an uncontrolled release of insulin through such a neural mechanism, which could cause potentially dangerous hypoglycemia if it occurred independent of glucose. Moreover, the lack of cephalic stimulation of the β-cells may partly explain why glucose homeostasis in critically ill patients receiving parenteral or enteral nutrition is often disturbed (414, 415). Likewise, a loss of cephalic phase stimulation of insulin release compromises postprandial glucose homeostasis and is often observed in patients with impaired glucose tolerance (55, 58). It is an early symptom (60) and supposedly contributes to the pathogenesis of type II diabetes mellitus (see Ref. 55 for a detailed discussion).

Given the complexity of the mechanisms that lead to the CPIR, several factors that influence these processes (e.g., the macronutrient composition and amount of the food eaten, the energy status of the individual, the time of the circadian cycle, etc.) may affect the CPIR and/or the critical balance between hepatic glucose uptake and output in response to eating. As to the macronutrient content of the meal, it is clear that circulating amino acid concentrations rapidly and substantially increase after mixed nutrient meals, and several amino acids potently stimulate insulin release by acting directly on the pancreatic β-cells (301–303). They can also enhance glucose-stimulated insulin release (303) (see FIGURE 6). Dietary fat is considered noninsulinogenic. Nevertheless, nonesterified fatty acids (NEFA) can stimulate insulin release from the β-cells by at least three different mechanisms (303), and fatty acids have also been shown to enhance glucose-stimulated insulin release (303). However, because of the lymphatic transport after absorption, circulating NEFA levels increase with a substantial delay (∼90–120 min) after eating (416). Therefore, diet-derived NEFAs presumably do not intimately interact with the other mechanisms that govern the CPIR or the prandial postabsorptive release of insulin, although such interactions are feasible for adipose tissue-derived NEFA in fasted individuals or during the inactive phase of the light-dark cycle that is characterized by a prevalence of lipolysis.

There is no doubt that proteins and fats, and/or certain amino acids and fatty acids, provide oro-sensory stimuli for the initiation of various cephalic responses (e.g., Refs. 13, 16, 42, 417–420). Several studies indicate that proteins/amino acids (90, 91, 125, 130–132) and even fats/fatty acids (130, 133) also serve as oral stimuli that can initiate a CPIR and may therefore somehow contribute to the pertinent effects of mixed nutrient meals. Crystal and Teff (138) reported no CPIR in response to high-fat oral stimulation, despite that the stimulus caused a massive release of PP. The reasons for this discrepancy, i.e., for the absence of a CPIR despite the strong PP response suggesting vagal motor activation, are unknown. Perhaps the pronounced decrease in blood glucose promoted PP and counteracted insulin release, or any sensory quality of fat provoked the substantial PP secretion. Interestingly, however, in another study with a complex oral stimulus, a strawberry-flavored mousse that consisted mainly of fat (19.9 g) and contained only small amounts of protein (2.9 g) and carbohydrates (2.4 g), authors from the same group observed a very clear and reproducible CPIR (18). However, the mousse in this experiment was sweetened with aspartame; hence, rather than the carbohydrates or proteins, the added aspartame may have accounted for the CPIR (see sect. 2.2.4). In an MSF study with foods with different macronutrient compositions (high fat, high carbohydrate, and high protein) in men, Zhu et al. (136) observed that only the high-carbohydrate food induced a CPIR. In sum, proteins or amino acids and perhaps even dietary fats or fatty acids appear to provide potent stimuli for a CPIR at least under certain conditions and in some species. Regarding fat, it is interesting that the CPIR has been shown to activate lipoprotein lipase in adipose tissue and thus contribute to postabsorptive fat metabolism and that fatty acids stimulate insulin release by acting directly on the pancreatic β-cells.

Another open question is whether vagal motor activation and/or acetylcholine sensitize the β-cells also for NEFA and thus have an important function in fat metabolism, as they do for glucose and some amino acids (421). The initial intracellular stimulatory mechanisms of insulin release in response to glucose, amino acids, and NEFA are different (303), but in the end, insulin release always depends on an increase in intracellular calcium, which, depending on the stimulus, is produced by varying combinations of ionotropic calcium channels, and acetylcholine via M3 muscarinic receptors, G_q- and IP₃-mediated release of calcium from the endoplasmic reticulum (FIGURE 6). Therefore, an early activation of vagal motor fibers, as part of the cephalic stimulation, and the resulting action of acetylcholine should, at least in theory, also sensitize the β-cells for amino acids or NEFA, but at least to our knowledge, this has not yet been specifically tested.

6. SYNOPSIS

We started this review by revisiting the available literature on the CPIR and evaluating the triggers and conditions under which it takes place and its neural and physiological mechanisms. From this, a picture emerged that the CPIR in humans occurs less consistently and may depend on more complex stimulation than in animals. For instance, in humans a CPIR is usually observed when the experiments are designed to have the participants ingest a real meal. In this case, the full range of visual, olfactory, gustatory, and tactile stimulation including the reafference signals associated with chewing and swallowing are present. In contrast, the mere stimulation of taste receptors in the frontal part of the oral cavity or even MSF or other procedures that provide only part of the complex sensory stimulation of a real meal is often insufficient to trigger a CPIR. Clearly, however, in humans as in animals, vagal motor neuron signaling appears to be crucial.

Notwithstanding the complexity of stimulation that appears to be required to initiate a CPIR, particularly in humans, strong evidence indicates that glucose plays a special role as an oral stimulus in all mammals investigated so far. This makes perfect sense because a major function of the CPIR is to limit the prandial and postprandial rise in blood glucose levels, thereby contributing to glucose homeostasis. In this context, it is important to emphasize that activation of the orally expressed T1R2/T1R3 heterodimer that is commonly called the “sweet” receptor is not sufficient to trigger a CPIR. While interesting alternative possibilities have been proposed for how oral stimulation with glucose may cause a neurally mediated insulin release, the exact mechanism still needs to be identified. In general, the thorough analysis of the available data and the consideration of the physiological mechanisms involved raise questions about the somewhat strict classical definition of the CPIR as a head receptor-triggered, neurally mediated, and transient insulin release that occurs before nutrient absorption. We therefore end this review in the hope that we have managed to raise awareness of the limitations of this classical definition. In particular, as discussed earlier, the traditional definition demands that the insulin release be preabsorptive and thus transient, constraints that are unnecessarily restrictive for at least two reasons. First, acetylcholine, the major chemical effector of vagal motor neuron activation, primarily sensitizes the β-cells for glucose and some amino acids, but, by itself, does not initiate a large release of insulin (FIGURES 6 AND 7). Therefore, the activation of vagal motor neurons can hardly be expected to cause a substantial insulin release before the absorption of glucose from ingested carbohydrates. Second, the cephalic activation of vagal motor fibers and, hence, stimulation of insulin release, continues throughout the meal (e.g., Ref. 23; FIGURES 7 AND 9) and is not limited to its beginning. Moreover, even a brief cephalic stimulation can have substantial effects on glucose homeostasis (29). The contribution of the cephalic phase or neural stimulation of insulin release to the overall insulin response to a meal is therefore much greater than commonly assumed (23, 29).

In fact, because of the widespread focus on the early phase of the neurally mediated insulin release during eating as proof of a CPIR and, hence, as an interpretative demand, investigators are trying to measure a small insulin response that likely requires optimal conditions to be detected. This is because the neural mechanisms initiated by oropharyngeal glucose are designed to prime the β-cells for a response to glucose (in a manner similar to the incretin effect) rather than to serve as a direct trigger for insulin release. The limited usefulness of the strict classical CPIR definition, coupled with the experimental need to distinguish between cephalically and postabsorptively stimulated insulin release, should therefore prompt us to look for other measures of the cephalic or neural insulin release that better reflect its physiological and clinical relevance as a critical component of the overall insulin response to a meal.

GRANTS

The work on this review was supported by National Institute of Diabetes and Digestive and Kidney Diseases (NIDDK) Grant 5R01DK118910-04 “Neural Bases of Cephalic Phase Endocrine Responses” (to W.L., A.G.W., and A.C.S.); Swiss National Science Foundation, International Short Visit Grant IZK0Z3_158027; “Brain Targets of Sensory Information From the Hepatic Portal Vein” (to W.L. and A.G.W.); and NIDDK R01DK121531 “A Brain-Wide Neural Network for Glucosensory-Motor Integration During Hypoglycemia” (to A.G.W.).

DISCLOSURES

A. C. Spector is a member of the Scientific Advisory Board of Gila Therapeutics.

AUTHOR CONTRIBUTIONS

W.L., A.G.W., and A.C.S. prepared figures; W.L., A.G.W., and A.C.S. drafted manuscript; W.L., A.G.W., and A.C.S. edited and revised manuscript; W.L., A.G.W., and A.C.S. approved final version of manuscript.