Homeostatic regulation of protein intake: in search of a mechanism

Abstract

Free-living organisms must procure adequate nutrition by negotiating an environment in which both the quality and quantity of food vary markedly. Recent decades have seen marked progress in our understanding of neural regulation of feeding behavior. However, this progress has occurred largely in the context of energy intake, despite the fact that food intake is influenced by more than just the energy content of the diet. A large number of behavioral studies indicate that both the quantity and quality of dietary protein can markedly influence food intake. High-protein diets tend to reduce intake, low-protein diets tend to increase intake, and rodent models seem to self-select between diets in order to meet protein requirements and avoid diets that are imbalanced in amino acids. Recent work suggests that the amino acid leucine regulates food intake by altering mTOR and AMPK signaling in the hypothalamus, while activation of GCN2 within the anterior piriform cortex contributes to the detection and avoidance of amino acid-imbalanced diets. This review focuses on the role that these and other signaling systems may play in mediating the homeostatic regulation of protein balance, and in doing so, highlights our lack of knowledge regarding the physiological and neurobiological mechanisms that might underpin such a regulatory phenomenon.

the regulation of food intake is one of the most essential phenomena in biology. Survival is dependent on the organism procuring a variety of macro- and micronutrients within an environment in which food availability and quality can be unreliable. The regulation of energy intake exists at the forefront of research on ingestive behavior, perhaps rightly so, considering that obesity is at its core a disease of energy excess. An abundant literature indicates that lean animals actively balance energy intake and energy expenditure (energy homeostasis), and in recent decades, an intricate neuroendocrine network has been characterized (10). Yet these advances in our understanding of the central nervous system regulation of energy balance have not translated into a marked impact on the obesity epidemic. Some have even questioned the power of this regulatory system, considering the relative ease with which most individuals gain weight.

When considered in the context of a natural environment, it seems likely that food intake is driven by more than simply a need for energy. Considering the necessity of protein consumption and the fact that there is no storage site for body protein that is analogous to adipose tissue, it seems likely that physiological mechanisms exist to ensure an adequate supply of protein. This review will address the overarching question of whether protein intake is physiologically regulated, focusing initially on the historical behavioral evidence, moving to potential mechanisms underlying the detection of dietary protein content, and then finally discussing potential mechanisms that might contribute to the detection of physiological protein balance and subsequent regulation of protein selection. It should become apparent from this discussion that ample behavioral evidence supports a regulation of protein intake but that the cellular and molecular mechanisms that might underlie such a regulation are largely undescribed.

Basic Phenomena and Historical Perspective

The effect of variations in dietary protein quality and quantity has a long, rich historical background. Experimental manipulation of dietary protein in laboratory and agricultural models dates back over 100 years. Collectively, these data suggest that protein can have a profound impact on food intake, via two similar, but potentially separate, mechanisms. First, variations in absolute protein quantity lead to changes in feeding behavior, such that high-protein diets suppress food intake, moderately low-protein diets increase food intake, and very low protein diets (protein insufficient) suppress food intake and stunt growth. Second, variations in protein quality (the balance of amino acids) also influence food intake, as many species show a keen ability to detect and avoid imbalanced diets. As such, this literature provides clear evidence that altering dietary protein or amino acid content can alter food intake. The central question of this review is whether these changes in food intake reflect a homeostatic regulation of protein intake. Is protein intake actively matched to physiological demand, or alternatively, do the changes in intake reflect some other endpoint, such as the varying sensory properties of the food or a more generalized avoidance of unhealthy diets?

Effects of High-Protein Diets on Food Intake

There is ample evidence in both human and animal models that high-protein diets suppress food intake over the short term (1, 142). Protein is widely regarded as the most satiating macronutrient per calorie, with high-protein preloads reducing both perceived hunger and food intake to a greater extent than low-protein preloads (5, 9, 59, 99, 128, 140). The longer-term effects of high protein on food intake and body weight are somewhat more complicated. In humans, a substantial number of clinical studies indicate that high-protein diets promote weight and adiposity loss (37, 38, 60, 141). This effect seems to be due to reductions in food intake, maintenance of fat-free mass, and increases in energy expenditure (142). Similar evidence exists in rodents, with rats rapidly reducing intake in response to a high-protein diet (9, 53, 65). However, other studies describe an initial suppression of food intake followed by a recovery toward normal, likely due to adaptive increases in amino acid degradation and metabolism (15, 58, 95). With adequate adaptation, rats have been shown to consume and grow rather well on high-protein diets, with high protein tending to reduce adiposity but maintain lean mass (58, 65). Yet animals will also readily abandon the high-protein source in favor of a control diet when given the choice (96). In summary, these data suggest that animals will generally avoid excessive protein intake; however, it is unclear whether this response reflects a specific detection of protein (positive protein balance).

Effects of low-protein diets on food intake.

In contrast to the rather abundant literature assessing the effects of high-protein diets on food intake and body weight, there have been fewer experiments focusing on the response to a low-protein diet. This limitation seems to partly stem from the fact that the feeding response to a low-protein diet is highly variable, depending on both the severity of protein restriction and the physiological state of the animal. However, the data most consistently suggest that rats respond to moderately low-protein diets with an initial increase in food intake, theoretically in an effort to increase protein intake (36, 83, 144–146). In addition, it was demonstrated that the hyperphagia occurred regardless of whether the isocaloric diet was balanced by increasing fat or carbohydrate (144). However, if protein levels decrease to a point where hyperphagia cannot overcome the protein deficit, rats will abandon this approach and spontaneously reduce food intake (36, 95). As such, low-protein diets have a biphasic effect on food intake, with moderate reduction tending to increase intake, but severe reductions reducing intake. The hyperphagic effect can be readily observed in very young, growing rats, but is more difficult to detect in mature rats (145), likely because protein requirement drops with age. It should be pointed out that this hyperphagia in response to moderately low-protein diets is not detected in all studies, even in young rats (95, 146). In summary, these data suggest that modest reductions in dietary protein increase food intake, resulting in an increase in energy intake and body fat mass but decreases in lean mass (36). This effect also has the hallmark of “protein leveraging”, which will be discussed later in relation to the geometric modeling of intake and selection.

Amino acid imbalance.

It has been known for nearly 100 years that diets that are devoid of certain amino acids fail to support growth, and as such, a particular set of amino acids are essential for life and indispensable in the diet (110). More interesting is the observation that rats appear to rapidly detect and readily avoid these imbalanced diets (53, 70, 114). Considering that this phenomenon has been demonstrated following depletion of multiple amino acids, it seems likely that rats are capable either of detecting individual amino acids or the metabolic consequence of the depletion of any single amino acid. Building upon the work of A. E. Harpers and others (53, 70), Dorothy Gietzen and colleagues (61) provided a specific mechanism for this behavior by demonstrating that the avoidance of imbalanced diets represented a learned aversion, which could be conditioned by both the diet and cues associated with the diet (46). This learned aversion is mediated by critical molecular events within the anterior piriform cortex (APC), in particular, the accumulation of uncharged tRNAs and the resultant activation of the GCN2 kinase (52, 112). Replacement of the missing amino acid locally within the APC, or deletion of GCN2, is sufficient to attenuate this learned aversion (52, 74, 113). These observations provide clear support for a detection of protein quality (amino acid profile). A central question, however, is whether this rather elegant mechanism for the sensing of amino acid imbalance is in any way related to the more general changes in food intake on low- and high-protein diets. Does the suppression of food intake induced by a high-protein diet depend on GCN2 signaling within the APC? To date, this question has not been adequately addressed, but lesions of the APC which block hypophagia on an imbalanced diet do not appear to block hypophagia of a high-protein diet (69).

Summary and caveats.

The preceding discussion is just the tip of a large literature, complete coverage of which is well beyond the scope of this review. The primary point is that variations in dietary protein quantity and quality can have significant effects on food intake. The question is whether these changes actually reflect a specific, homeostatic regulation of protein intake. While on the surface, one may say that the obvious answer is yes, in reality, this interpretation is fraught with caveats. For instance, nearly all of the studies reviewed to this point have focused on single diets, and suffer from the fact that variations in protein are offset by variations in another macronutrient. Isocaloric, low-protein diets are by definition simultaneously high in carbohydrate and/or fat. Thus, when an animal increases their intake on a low-protein diet, is this increase due to the reduction in protein or the increase in carbohydrate? At least one experiment suggests that low-protein hyperphagia is evident regardless of whether the diet is counterbalanced with fat or carbohydrate (144). It is also possible that a selection for a particular diet might be driven by characteristics secondary to the macronutrients. For instance, perhaps the selection for or against protein is due to the sensory properties (taste, smell, and texture) of that particular protein source, and as such, the observations should be replicated with multiple, distinct macronutrient sources (133). Alternatively, the alterations in food intake in response to a high-protein diet may stem from an avoidance of toxic levels of amino acids in the blood, and thus, the reduction in food intake is not a specific regulation of protein but instead a more general avoidance of an unhealthy diet (53). A similar perspective could be hypothesized for the avoidance of imbalanced diets.

Lastly, a distinction should be drawn between the evidence supporting a specific appetite for protein and the evidence supporting the regulation of protein intake. The phrase “specific appetite” is rooted in work on salt and water intake, in which the drive to consume salt or water appears to be specific and innate (43, 127, 134). Whether a similar analogy can be drawn with protein intake is currently unclear. Umami flavor notwithstanding, it is unclear whether an innate “taste” for protein is hard-wired. Certainly, protein may have distinct sensory properties (taste or smell) that can be used to discriminate between high and low protein, but it seems likely that protein consumption is also influenced by an association between these sensory properties and postingestive consequences of the diet (33). By analogy, animals clearly have a taste for sweet, but few would argue that animals regulate energy balance by using sweet taste to measure the energy content of the diet. Energy balance is achieved when internal signals of energy status interact with neurobiological pathways controlling food intake, and it seems likely that protein balance would be achieved through similar mechanisms.

Physiological Regulation of Protein Intake

Protein selection.

The literature reviewed previously focuses on feeding responses to single-choice diets that are high, low, or imbalanced in protein. Although these studies are experimentally tractable, they do not reflect the practical decisions that animals make in a free-living environment. Omnivores, such as humans and rodents, are rarely exposed to a single diet, but instead to a broad range of food sources, each with a unique composition of macronutrients and micronutrients. Individuals must, therefore, navigate and select between this diversity of food sources. The simple question, therefore, is whether regulatory systems exist to ensure a supply of protein that is adequate for growth and physiology.

There is a large body of data indicating that a wide range of species will self-select between diets that are high and low in protein, and in so doing meet protein requirements (2, 40, 41, 62, 66, 79, 84, 96, 132, 146). There is, however, some debate as to whether this self-selection produces a precise regulation of protein intake, or a more general maintenance of protein between certain minimal and maximal levels (66, 96). Rats that are adapted to and growing well on a very high-protein diet will shift intake to a control diet if provided with the opportunity (132) and will similarly overeat protein following protein depletion (33, 146). Lastly, if exposed to diets that are imbalanced but complementary in their amino acids profile, rats will self-select between these diets and balance their amino acid requirements (41). Taken together, these observations strongly suggest that protein intake is balanced, and that animals self-select in order to meet protein requirements.

Yet there is also evidence to the contrary. Work by Galef and colleagues provide several scenarios in which rats fail to select an adequate amount of protein, even when that choice is available (8, 42). This seems to particularly occur when an adequate protein source is provided within the context of a large number of inadequate choices, and/or when the protein source is made relatively unpalatable. These settings obviously make it difficult for the animal to identify the source of adequate protein, but one would expect that rats would eventually solve this issue if protein were strongly regulated (105). Another concern is that many studies of protein selection involve two diets in which selection of adequate protein is indistinguishable from a random selection between the diets, such that the balanced intake is not definitive evidence of protein regulation. Lastly, Reed et al. (105) demonstrated that rats with prior experience with carbohydrate or fat sources would preferentially consume these macronutrients over protein when subsequently given a choice, ultimately leading to protein deprivation. Thus, while animals seem to regulate protein intake in some settings, these observations indicate that protein intake can be dysregulated by changes in availability, palatability, and prior experience.

Protein selection in geometric models of feeding.

Perhaps the strongest data supporting the ability of an animal to navigate through “nutrient space” derive from the Geometric Framework, developed by Steven Simpson and David Raubenheimer (20, 122, 123). As discussed previously, studies of macronutrient selection are impaired by the fact that an increase of one macronutrient must be offset by a decrease in a different macronutrient. The geometric approach addresses this and other concerns by using a geometric state space model to quantify the intake of individual nutrients across a range of diets and choices. The reader is referred to reviews by Simpson and Raubenheimer, in which they explain the approach in more detail (122, 123), but, in essence, the approach involves plotting the consumption of an individual nutrient relative to other components in the diet (e.g., protein vs. carbohydrate, or protein vs. energy). This analysis allows one to compare the consumption of protein relative to carbohydrate or energy, even when the individual is self-selecting from a variety of food sources that themselves differ in macronutrient content. The result is a quantitative assessment of macronutrient consumption in a free-feeding situation. Initially applied to insects, it was determined that multiple insect species will select among food sources in order to attain a target amount of protein and carbohydrate, while fat intake is unregulated. These and other authors have more recently applied this approach to other species, and the data suggest that species as diverse as fish, insects, rodents, and pigs seek to consume a fixed amount of both protein and carbohydrate, and thus regulate intake around a specific protein:carbohydrate target (63, 103, 122, 125). These observations suggest that protein intake is regulated more specifically than just the maintenance within a general upper and lower limit.

The geometric approach also provides a second, more interesting observation that relates to the balance or competition between macronutrients. When faced with diets that do not allow an individual to simultaneously reach its protein and carbohydrate targets, the individual must balance overeating one nutrient against undereating the other, and evidence in insects and rodents indicates that protein intake is prioritized over carbohydrate. This effect has been termed “protein leveraging”, as small changes in protein content can induce profound changes in energy intake (122, 123, 125). From this perspective, the hyperphagia detected on a low-protein diet is due to the leveraging effect of protein, with energy being overconsumed in an effort to consume a target amount of protein. Some organisms seem to exert a very profound protein leverage, with protein precisely regulated at the expense of carbohydrate, while in others, the leveraging effect is less pronounced and results in a compromise between the underconsumption of protein and the overconsumption of energy (122, 125). Interestingly, this balance is also shifted in carnivorous vs. omnivorous species. While omnivorous species seem to prioritize protein over carbohydrate, carnivorous species (such as the cat) seem to primarily defend against excess carbohydrate consumption, while also eating to a protein target (56). Metabolic adaptations in the cat result in additional nutritional requirements that are not present in omnivores, while other metabolic changes impair the cat's ability to metabolize carbohydrates. Thus, a preference for protein and strong avoidance of carbohydrates fits well with an animal whose diet consists primarily of animal tissue. Work in a variety of other species supports the concept of regulated macronutrient selection that is influenced by the specific physiological and environmental characteristics of the individual (75, 102, 115). Lastly, the concept of protein leveraging has recently been extended to humans, with reductions in dietary protein leading to increases in energy intake, at least over the short term (47, 121). Taken together, these data provide a very compelling case for both the regulation of protein intake and a potential role for protein in the regulation of energy intake and obesity (17, 123).

Protein intake in response to changes in physiological protein demand.

If protein intake is being matched to protein requirements, then we should also expect that changes in this requirement, (i.e., physiological changes in protein demand) should also produce an alteration in protein intake. The experiments described in prior sections involve behavioral changes in response to altered protein supply. If we are to hypothesize the existence of a protein balance, then protein intake and selection should also be responsive to changes in the physiological need or demand for protein.

Just as energy restriction induces an adaptive increase in energy intake once food again becomes available, protein-restricted animals exhibit a specific increase in protein intake and motivation for protein when subsequently given the opportunity (15, 30–33, 45, 62, 146). Such a result provides strong evidence that negative protein balance is detected and promotes a selection for protein. This behavioral response is also detected in other settings. Young growing animals have a relatively high-protein requirement, and multiple experiments in several species indicate that young animals select a higher-percent protein compared with older individuals (16, 40, 57, 145). Animal lines that are bred for high muscle growth tend to choose diets higher in protein compared to slower-growing lines (40). Lastly, chronic injections of growth hormone increase muscle mass, reduce body fat, and produce an increased selection for dietary protein (97, 107). Taken together, these data indicate that regulatory mechanisms monitor the body's physiological need for protein, and this regulation occurs independently from energy balance (84). Importantly, this mechanism is likely independent of any manipulation or characteristic of the diet, instead reflecting a direct detection of protein stores or metabolism, analogous to the role of leptin in regulating energy intake.

What Is Dietary Protein, and How Are Low- and High-Protein Diets Discriminated?

It is relatively straightforward to quantify the amount of calories consumed in a diet, and to break down caloric consumption by the individual macronutrients. Quantifying micronutrient intake is also straightforward. But when an animal chooses between low or high protein or increases protein intake in response to changes in physiological protein demand, what are they identifying and selecting for in the diet? Metabolically, there is no such thing as protein per se, because protein represents nitrogen, essential and nonessential amino acids, and in some sense, digested dipeptides and tripeptides. We can express protein as nitrogen content (crude protein) or as caloric content, but neither metric truly captures the balance between protein quality (amino acid balance) and quantity. Therefore, if an animal chooses protein, it is actually rather difficult to define what dietary component is being selected. Evidence indicates that protein is more than nitrogen content, as animals select a complete protein source over an incomplete protein source, even if both diets provide similar levels of nitrogen (crude protein). Yet beyond these facts, the specifics of dietary protein intake are unclear. Do animals eat for a limiting amino acid, do they actively seek a balance of all amino acids, or conversely, do they only avoid marked imbalances? Additionally, are they actually selecting for these individual components (individual amino acids), or instead more generally pairing sensory cues and dietary characteristics with the postingestive consequences of the diet?

Taste.

Taste plays an obvious role in ingestive behavior, but to what extent does taste directly drive protein intake? If taste does contribute to protein intake, there must be a specific protein or amino acid taste, this taste must be interpreted by the brain, and effector systems must integrate a response to these stimuli in the context of absorbed and metabolized protein products, as well as the organism's current homeostatic balance.

The major taste receptors are G protein-coupled receptors and ion channels that are expressed by specialized taste sensory cells. The G protein gustducin is specifically expressed in taste cells and is critical for the perception of sweet, bitter, and the savory (umami) flavors (76). Gustducin is expressed in ∼25–30% of taste cells (76, 148) and interacts with receptors belonging to the T1R family to produce tastes for both sweet and umami (89). The savory taste, termed umami (monosodium glutamate), is mediated by the T1R1-T1R3 heterodimer. (89). Activation of taste receptors leads to downstream signaling cascades that include cAMP, inositol triphosphate, phospholipase C, and calcium influx, activation of the taste cell, and ultimately the activation of afferent sensory neurons. In addition to expression on the tongue, gustducin and various taste receptors are also expressed on enteroendocrine cells throughout the gastrointestinal (GI) tract (106). These cells produce a variety of hormones, including peptide YY (PYY) and glucagon-like peptide-1 (GLP-1), which are known to regulate food intake (26, 34, 35), and thus these post-oral taste receptors may communicate information to the brain regarding the nutrient content in the GI tract (111).

It is clear that the umami taste is associated with consumption of savory, high-protein foods. Yet is umami really a “protein taste”, such that the protein content of a food can be quantified by the umami content or intensity? While monosodium glutamate (MSG) is the compound most closely associated with umami, the umami flavor is also represented by a rather diverse set of compounds (29). In addition, the T1R1/T1R3 heterodimer appears to be capable of responding to amino acids other than MSG (86), and umami is also detected in the absence of T1R3 (27). These data collectively suggest that there is likely more than one receptor mediating the taste termed umami and that individual amino acids besides MSG represent unique tastes (18, 73, 150). Indeed, some amino acids provide a sweet taste, while others taste bitter. The umami flavor, therefore, is derived from a diversity of amino acids, peptides, and nucleotides, potentially contributing to the perception of dietary protein content. It remains very unclear whether and to what extent protein selection is driven by umami. Similarly, diets also have other unique sensory properties beyond just taste (e.g., smell and texture), which could also contribute to the detection of protein in the diet. Much remains to be done in terms of amino acid taste receptor signaling and its integration into the regulation of protein intake.

Post-oral sensing and learned associations.

It is well described that, much like the tongue, the intestine can “taste” nutrients, and, as such, individual nutrients can be directly detected in the post-oral GI system. GI nutrient sensing is a large and active field, and a variety of mechanisms have been described through which dietary protein might be detected in the gut (104, 136). For instance, glutamate represents one of the most abundant amino acids in dietary protein, and luminal glutamate provides an important energy source to intestinal enterocytes and influences the firing of vagal afferent fibers (11, 139). Sweet, bitter, and umami taste receptors are all expressed within the post-oral GI system, and recent work has also highlighted calcium-sensing receptors as detectors for a broad range of amino acids (23, 44). Collectively, these sensing mechanisms orchestrate and coordinate GI function by regulating both vagal afferents and gut hormone secretion, and in so doing, they also have ample opportunity to impact the brain. Animals clearly detect and react positively to the presence of protein in the GI system, as the infusion of protein and amino acids directly into the gut is sufficient to condition a learned preference (91, 92, 137). Interestingly, this conditioning effect is independent of changes in circulating nutrient levels or vagal activity, at least for glucose (118), suggesting that a currently unknown signal is communicating the presence of nutrients to the brain. Yet carbohydrate and fat can also condition a learned preference, and it is currently unclear whether nutrients represent a general or macronutrient-specific reinforcer.

These data collectively suggest that learning may play a key role in the regulation of protein selection. While amino acids may have a discrete taste, it is problematic to hypothesize that the preference for a low- or high-protein diet is driven solely by a taste for specific amino acids. More likely is the association of postingestive consequences of specific foods with specific cues within the diet (79), including taste, sight, smell, and texture. For instance, protein-deprived hamsters exhibit a strong preference for high protein, but the identification of the high-protein diet is driven primarily by flavors associated with that diet (33). These data suggest that protein intake is not a specific, hard-wired appetite for protein, but instead a learned association between dietary cues and postingestive consequences.

Physiological Signals of Protein

Direct effects of amino acids on the brain.

The most straightforward mechanism for protein detection would be a direct effect of amino acids on brain areas regulating food intake (78), as both dietary and circulating amino acids clearly have access to the brain (21, 55). Considering that tryptophan and tyrosine serve as precursors for key neurotransmitters (serotonin and dopamine, respectively), an attractive early hypothesis was that these two amino acids might provide a unique signal related to dietary protein intake (53). While variations in dietary protein do seem to influence brain amino acid levels, particularly the ratio of tryptophan to large neutral amino acids (3, 95), subsequent interventional studies demonstrated that tryptophan administration fails to alter protein selection, despite markedly altering brain tryptophan and serotonin levels (81, 94). Thus, although tryptophan and serotonin alter food intake in general, there seems to be limited support for either tryptophan or serotonin as specific regulators of protein selection.

More recent experiments have focused on the branched-chain amino acids (BCAAs), particularly leucine, as signals of dietary protein. Leucine suppresses food intake when injected into the brain (12, 25, 83, 109), and it appears to induce this effect by acting locally within the mediobasal hypothalamus (12), via activation of mammalian target of rapamycin (mTOR) and/or inhibition of AMPK signaling (25, 83, 109), or via activation of BCAA metabolism (12). While most plasma amino acids are maintained within relatively constant levels despite dietary protein content, plasma BCAAs increase in direct proportion to dietary protein content (95). These observations suggest that BCAAs may provide a unique circulating signal of dietary protein content. Recent work indicates that increasing dietary leucine or BCAAs can reproduce the anorectic effects of high-protein diets, while also influencing specific signaling events within peripheral tissues and the brain (87, 109). However, this anorectic effect of leucine is primarily observed when it is added in excess to the diet, or injected into the animal (intracerebroventricularly), resulting in pharmacological, not physiological, increases. It remains unclear whether low levels of brain amino acids signal protein deficiency, or whether the high levels of brain amino acids induced by injection result in shifted protein preference. Anderson et al. (2) allowed rats to self-select between low- and high-protein diets, but supplemented the low-protein diet with BCAAs to determine whether the added BCAAs would alter the perception of the low-protein diet. Supplementation with BCAAs had little effect on protein selection, indicating that dietary or circulating BCAAs are not associated with alterations in protein intake or selection. Leucine can clearly suppress food intake by activating signaling systems within the hypothalamus, yet it is currently unclear whether these effects are primarily of pharmacological importance, or if they are also of relevance to the homeostatic regulation of protein intake.

The strongest evidence supporting a direct action of amino acids on the brain comes from work in the field of amino acid imbalance. As described above, diets that are severely imbalanced in their amino acid composition reduce food intake and induce a learned aversion. Subsequent experiments have demonstrated that this process involves a direct detection of this amino acid imbalance within the brain, specifically the anterior piriform cortex (APC). Interestingly, in this case, the detection is not of amino acid excess or deficiency, but the imbalance of the amino acid profile, which results in an accumulation of uncharged tRNAs for the missing amino acid (52). This accumulation of uncharged tRNA activates the kinase GCN2 locally within the APC, and replacement of the missing amino acid locally within the APC or deletion of GCN2 is sufficient to attenuate this learned aversion (52, 74, 113). As such, these observations clearly demonstrate a physiologically relevant effect of amino acids acting directly in the brain.

Detection in the gut.

It is well described that the presence, digestion, and/or absorption of nutrients by the GI tract leads to changes in the secretion of a wide variety of gut hormones, including CCK, GLP-1, PYY, and ghrelin (10, 130), with these and other hormones contributing to the general satiating effect of protein. However, are any of these hormones providing information specific to protein, such that they specifically regulate protein intake and selection? To varying degrees, each of these of hormones is also altered in response to carbohydrate and lipid. While a great body of literature indicates that these hormones regulate food intake, their effects on macronutrients are less clear. As such, there is no compelling evidence to support any of these gut hormones as key signals of dietary protein content. The one exception is PYY, as it was shown that PYY knockout mice fail to reduce food intake when placed on high-protein diets (6). Whether these mice also exhibit an altered selection for protein is unclear. In addition to the secretion of various gut-derived hormones, the presence of protein or amino acids within the gut could also influence the brain through vagally mediated signals, as protein activates the vagus nerve and induces c-Fos within the brain stem (135). Yet again, it is unclear whether vagal signals contribute to the specific detection of protein, as an intact vagus nerve is not required for the anorectic effects of a high-protein diet (64, 71). In summary, both gut hormones and vagal signals are activated in response to protein intake, but there is no convincing evidence that either endocrine or neural signals derived from the gut provide information specific to the regulation of protein intake.

Detection of physiological protein need.

While oral or gut-derived signals are activated by protein and may contribute to the effects of high and low protein diets, it seems much less likely that these signals contribute to the detection of physiological protein need. As discussed above, increased physiological protein demand leads to a specific selection for protein, and it is not clear how gut-derived signals would mediate this effect. What then could be the signal of negative protein balance? One possibility is an alteration in the circulating concentrations of amino acids. As noted above, there is ample evidence that exogenous amino acids (leucine) suppress intake, but to date, there is no compelling data to suggest that alterations in endogenous circulating or brain amino acids produce an altered selection for protein. An additional possibility for protein regulation is the production of signals linked to liver or muscle protein metabolism. The liver is an important site for whole body nitrogen and amino acid metabolism, catabolizing amino acids in order to avoid their toxic buildup. This regulatory mechanism is highly sensitive to amino acid concentrations, with degradation capacity increasing in the face of excess amino acids but decreasing to spare amino acids when protein is restricted. Similar to observations in the brain, signaling systems such as mTOR, AMPK, and GCN2 are all altered within the liver in response to changes in dietary protein (22, 50). Perhaps signals related to hepatic amino acid metabolism communicate with the brain to regulate protein intake, though, to date, no such signal is described.

Skeletal muscle also actively shifts between catabolic and anabolic processes based on amino acid availability and protein need, and muscle represents the primary storage site for amino acids in the body. Periods of growth or exercise promote the uptake of amino acids and protein synthesis, whereas fasting results in muscle breakdown and the release of free amino acids, principally for hepatic gluconeogenesis. While a discussion of the endocrine and molecular regulation of skeletal muscle protein synthesis and degradation is beyond the scope of this review, it seems possible that variations in skeletal muscle amino acid metabolism may be detected by the brain. For instance, there is strong evidence that members of the transforming growth factor-β family, such as activin, follistatin, and myostatin (67, 77), regulate muscle mass, and these molecules are found in the circulation. Yet to date, there is little evidence suggesting that these muscle-regulatory factors regulate food intake. Muscle also produces a wide variety of cytokines, many of which have the capacity to act within the brain. Muscle secretion of IL-6 is particularly increased in response to exercise, and thus, it is possible that IL-6 or some other muscle-derived myokine might act in the brain via a mechanism analogous to adipokines, such as leptin (7, 90, 108). Yet similar to the above, there has been little if any work focusing on myokines as specific regulators of protein intake and selection.

Although it is clear that the brain detects and responds to variations in protein need, the mechanism through which this detection is mediated is unclear. While available data support the possibility that amino acids may directly act within the brain, this approach seems on the surface to be too simplistic to explain the full behavioral phenomena. By analogy, it is well accepted that glucose can act within the brain to regulate food intake, and yet glucose is certainly not the only, and perhaps not the most important, signal of energy status. Just as hormones, such as leptin, convey important information regarding both acute and long-term energy stores to the brain, muscle- or liver-derived signals might also provide information related to protein metabolism and storage. To date, there are no likely candidates for a myokine that acts in the brain as signal of lean body mass.

Neurobiology of Protein Selection

Despite an abundant literature focusing on the manipulation of dietary protein, the APC is the only brain area with an established role in the regulation of protein intake and selection. As discussed previously, the APC is essential for the detection and avoidance of diets that are devoid of essential amino acids, and this response is mediated by the buildup of uncharged tRNAs and the activation of GCN2 kinase (46, 74, 112). With this work, Gietzen and colleagues (46, 112) have described the clearest model currently available for the detection of dietary amino acid imbalance, extending their observations from the behavioral to the molecular level. Yet despite the clarity of this model, to date, there have been no data connecting it to the larger issues of homeostatic regulation of protein selection. Although lesions of the APC block the aversion to a imbalanced diet, APC lesions apparently do not alter the response to a high-protein diet (69), and currently, no evidence indicates that the APC contributes to the more general regulation of food intake. These data suggest that the detection of amino acid imbalance may be mediated by a separate mechanism from the detection and regulation of protein balance. If not the APC, then what other brain area may contribute to the homeostatic regulation of protein balance?

The hypothalamus is classically associated with the homeostatic regulation of energy balance, and it would be logical to hypothesize that the hypothalamus contributes to the homeostatic regulation of protein balance. It is clear that leucine is sufficient to suppress food intake when administered directly into the brain, including directly into the hypothalamus, and available evidence supports two likely mechanisms. The first is the regulation of the classic energy-sensing molecules mTOR and AMPK. It is well known that leucine stimulates mTOR activity in peripheral tissues such as muscle and liver, and not surprisingly that leucine activates mTOR signaling within the brain. The downstream effects of mTOR are mediated by at least two signaling molecules, pS6K1 and 4E-BP1, both of which are linked to the regulation of protein translation at the ribosome. Although the data supporting 4E-BP1 as a regulator of food intake are limited, there are several observations indicating that mTOR-S6K1 signaling, particularly within the hypothalamus, influences feeding behavior (13, 24, 25, 49, 83, 138, 147). Even more so, it seems that mTOR signaling may be a node for more than just protein, as feeding relevant hormones, such as insulin and leptin, also activate mTOR signaling within the brain (25). AMPK is another molecule classically associated with the detection of cellular energy status, and a large literature implicates AMPK in the regulation of feeding behavior (80, 149). Both dietary and intracerebroventricularly administered leucine was shown to inhibit AMPK (109). Given evidence for cross-talk between mTOR and AMPK signaling, it seems likely that at least a portion of leucine signaling is mediated via AMPK. In addition to the activation of classic signaling pathways, there is also evidence, suggesting that the metabolism of amino acids, particularly BCAAs, may provide signals relevant to food intake. Blouet et al. (12) demonstrated that local injection of downstream products of BCAA metabolism (α-ketoisocaproic acid or α-ketoisovaleric acid), or pharmacologic manipulation of BCAA metabolism, also altered food intake. Genetic deletion of components of BCAA metabolism also leads to alterations in energy expenditure, food intake, and food selection (101, 119).

Hypothalamic neuropeptide Y (NPY)/Agouti-related peptide (AgRP) and POMC neurons are classically associated with the regulation of food intake in response to circulating signals, such as leptin and insulin (10, 82), and these neurons are also implicated in the response to leucine and dietary protein. Npy and Agrp mRNA expression tend to be increased in animals consuming a low-protein diet (83, 144) but decreased in animals on a high-protein diet (109), and leucine was shown to regulate both AgRP and POMC expression via an mTOR-dependent mechanism (83, 109). Leucine also appears to stimulate Pomc expression and influence both mTOR and AMPK activity within these cell populations (12, 25, 109). Thus, there is ample evidence that these neuronal populations classically associated with energy sensing are also responsive to amino acids. Beyond the hypothalamus, there is also evidence supporting a role for brain stem neurons in mediating the effect of protein, as both the nucleus of the solitary tract (NTS) appears to contribute to the effects of intracerebroventricular leucine, and also to respond to dietary protein (12, 28, 39, 117).

Although these data collectively provide a cellular and neural mechanism for leucine-dependent decreases in food intake, the role of these signaling systems in the specific regulation of protein intake is much less clear. The arcuate nucleus and NTS are primarily associated with energy homeostasis, and there is no evidence that leucine-dependent regulation of mTOR and AMPK in any of these brains areas causes a specific alteration in protein intake or selection. Considering the ample evidence that leptin, insulin, glucose, fatty acids, and a variety of other circulating signals impact both mTOR and AMPK, it is much more likely that these two molecules contribute to a general regulation of food intake, as opposed to a specific effect on protein intake and selection (13, 14, 24, 25, 80, 138, 147). As such, these data, together with the role of the APC in detecting amino acid imbalance, provide two disparate models. The APC is clearly involved in a protein-specific behavior, namely the avoidance of imbalanced diets, but there is no data connecting this mechanism to larger issues of protein homeostasis. Contrastingly, there are strong data linking leucine signaling within the hypothalamus to the general regulation of food intake, but none of these data provide any support for a specific effect of hypothalamic amino acid signaling to protein intake. Interestingly, there are few data implicating GCN2 within the hypothalamus as regulating protein intake, and mTOR is apparently unnecessary for the detection of amino acid imbalance (51). Additional experiments are clearly necessarily to connect these two disparate data sets and determine whether either contributes to the homeostatic regulation of protein balance.

The neural regulation of protein selection is imbedded within a larger literature focusing on the neurobiological substrates driving macronutrient selection. Over the last several decades, a variety of studies have implicated specific molecules or brain areas in the regulation of macronutrient selection. NPY, AGRP, 5HT, galanin, opioids, leptin, insulin, ghrelin, enterostatin, endocannabinoids, GLP-1, and glucagon have all been implicated in macronutrient selection studies (4, 19, 68, 85, 93, 100, 120, 124, 126, 131, 143, 151). Yet these studies represent largely isolated and at times conflicting observations, and a consensus view of macronutrient selection has failed to emerge (133). As a result, there is no compelling evidence that any brain area, neuronal population, neuropeptide, neurotransmitter, or intracellular signaling molecule specifically influences the selection of macronutrients, particularly protein.

As an alternative, it seems possible that protein selection is embedded within neurobiological systems that are already established for the regulation of food intake and other motivated behaviors (Fig. 1). For instance, it is well established that the motivation to procure and select between foods (or any other reward) is driven by cortico-limbic reward circuitry. To put it simply, the animal selects that which is most rewarding at the time. It also seems likely that this reward value (incentive salience) is influenced by need state, such that protein is particularly rewarding in the protein-deprived state. The motivation for protein interfaces with prior experience with food (food memories), resulting in selection of high-protein foods and food cues in the protein-deprived state. Thus, at its most basic, this system requires brain regions associated with the detection of need state, areas associated with reward-based decision making, and areas associated with memories of food and exteroceptive cues associated with those foods. Although the role of reward, learning, and memory in food intake and selection is well established, the glaring weakness of this model is the identity of protein-specific need state detectors. Where and how protein depletion is detected is currently unknown, although the APC likely contributes to this detection of single amino acid deprivation and/or amino acid imbalance.

Fig. 1.

Regulation of protein intake and selection. Foods containing dietary protein generate exteroceptive cues (e.g., taste and smell), resulting in learned associations between the sensory properties of the food and the postingestive consequences of its consumption. Interoceptive cues related to dietary protein content (gastrointestinal tract and liver) and protein stores/amino acid metabolism (muscle and liver) converge on the central nervous system, although the specific nature of these signals remains unclear. Cues of protein status are likely integrated within areas of the anterior piriform cortex (APC), hypothalamus, and/or brain stem, collectively serving to detect protein deficiency or imbalance. Cortico-limbic reward circuitry interacts with prior food memories to influence selection based on need state. In other words, protein depletion generates a need state signal, such that high-protein foods and the memories associated with those foods become increasingly salient and thus preferred. OFC, orbitofrontal cortex; Hipp, hippocampus; Amy, amygdala; VTA, ventral tegmental area; NAcb, nucleus accumbens.

Perspectives and Significance

Throughout this review, the question has been asked: Is protein intake homeostatically regulated? On the basis of the literature, the conclusion is a tentative yes. Variations in dietary protein can induce significant changes in food intake, and animals seem to adaptively alter food intake and protein selection in settings of protein deficiency. Work using the geometric framework strongly suggests that animals eat to a protein intake target and, if necessary, compromise energy intake to reach this target. These data collectively suggest that a mechanism does exist to detect protein balance and alter food intake in response to protein deficiency or excess. This conclusion is tempered by two concerns: First, there are settings in which animals will fail to consume adequate protein in a self-selection model. Second, the underlying neuroendocrine mechanism governing this putative regulation is virtually undescribed. In the absence of a mechanistic underpinning, any experimentally observed phenomena should be viewed with caution, as the observed phenomena could be driven by some unknown experimental artifact. Until a likely mechanism for protein intake is identified and manipulated, these behavioral outcomes will remain only interesting observations.

Regardless of whether protein is homeostatically regulated, the data clearly indicate that protein affects feeding behavior and body weight and that amino acid intake and metabolism have important implications beyond food intake. Work in aging models suggests that protein restriction and/or amino acid imbalance can increase longevity (48, 88, 129), with restriction of essential amino acids, particularly methionine, reducing body adiposity, and increasing insulin sensitivity (54, 98). Leucine and other BCAAs appear to interact with insulin signaling in several tissues (72, 87), and the leucine-dependent stimulation of hypothalamic mTOR has revealed new ways of considering how both fuels and hormones act in the brain to regulate feeding (13, 14, 24, 116). Lastly, protein intake and selection may have important implications for obesity, if the observations of protein leveraging are extendable to humans (47, 121). The conclusion that animals prioritize protein intake over energy intake indicates that modest changes in dietary protein content may lead to marked changes in energy intake (123). Recent decades have seen an explosion in the availability of foods composed of fat and processed carbohydrates. These foods offer a large amount of calories for a low cost when compared to either protein-rich foods or fruits and vegetables. Thus, both economic and social forces drive consumers, particularly those with limited incomes, toward a diet that is rich in energy but poor in protein and essential micronutrients (17). It is tempting to speculate that the overconsumption of these fat- and sugar-rich diets is driven not only by the hedonic aspects of the food, but also by the fact that they are relatively low in protein. If protein intake is truly prioritized over energy intake, then defining the mechanism of this regulation may open new doors for the clinical and pharmacological regulation of food intake. Future studies are necessary to fully make this connection, but it remains possible that a fuller understanding of the neurobiology of protein intake may provide new opportunities for manipulation of food intake.

GRANTS

This work was supported by a grant from the National Institutes of Health (R-01-DK-081563).

DISCLOSURES

No conflicts of interest, financial or otherwise, are declared by the authors.

AUTHOR CONTRIBUTIONS

CDM conceived and wrote the article. SDR and TMH contributed to the writing and editing of the article, as well as discussion of content.