THE COMPARATIVE PHYSIOLOGY OF FOOD DEPRIVATION: From Feast to Famine

Key Words feeding, fasting, starvation, metabolism, atrophy, digestion, gastrointestinal organs, specic dynamic action, phenotypic plasticity Abstract The ability of animals to survive food deprivation is clearly of considerable survival value. Unsurprisingly, therefore, all animals exhibit adaptive biochemical and physiological responses to the lack of food. Many animals inhabit environments in which food availability uctuates or encounters with appropriate food items are rare and unpredictable; these species offer interesting opportunities to study physiological adaptations to fasting and starvation. When deprived of food, animals employ various behavioral, physiological, and structural responses to reduce metabolism, which prolongs the period in which energy reserves can cover metabolism. Such behavioral responses can include a reduction in spontaneous activity and a lowering in body temperature, although in later stages of food deprivation in which starvation commences, activity may increase as food-searching is activated. In most animals, the gastrointestinal tract undergoes marked atrophy when digestive processes are curtailed; this structural response and others seem particularly pronounced in species that normally feed at intermittent intervals. Such animals, however, must be able to restore digestive functions soon after feeding, and these transitions appear to occur at low metabolic costs.

INTRODUCTION
All animals supply the energy required for basal metabolism, physical activity, growth, and reproduction from their food. When food is not available, animals must use internal energy stores to fuel these activities. Starvation resistance reects an animals ability to store energy and control its allocation during extreme resource limitation. Many animals live in environments in which food abundance and quality vary drastically over time, and periods of starvation are common. Bacteria can lead a feast-or-famine existence (1, 2), many planktonic species live in a resource-limited world (3), and many sh overwinter with little or no food. Sommer (4) suggested organisms can be grouped according to their responses to variations in food supplies. Velocity specialists are those species with high,

maximum population-growth rates that respond to sudden increases in resource abundance with rapid increases in population density; many bacteria fall into this category. Afnity specialists can maintain growth at low-food-intake levels; this may describe many tropical sh. Lastly, storage specialists such as hibernating mammals create large internal stores that increase the chances of survival when resource abundance is extremely low. Many of the animals that inhabit environments with uctuating food availability are adapted to consume very large meals when prey is available. Classic examples are sit-and-wait predators, such as snakes; such species feed only once or a few times a year. Very large meals followed by low rates of energy expenditure, typical of crocodiles as well as a number of other reptiles and various sh, allow for days or weeks between feeding. When the lack of food is due to seasonal changes in temperature or water availability, animals may enter into dormancy, during which digestive processes are curtailed. With the exception of birds and some large mammals, which migrate to more desirable areas, dormancy is common in vertebrates living in temperate or arctic environments that hibernate during cold periods of the year. In tropical areas, many vertebrates, with the notable exception of birds, enter into a dormant estivating state during dry, and typically warm, periods. Fasting may also occur when animals engage in activities that compete with or even preclude feeding or the search for food. Such activities include migration, moulting, or the care of eggs or young (e.g., 57). Examples of voluntary anorexia exist in all major groups of vertebrates and may be very prolonged. Some species of penguins, for example, do not eat for several months when tending to eggs (8, 9); Pacic salmon do not feed during their upstream migration, which can be more than 1000 km in length (10); and eels do not feed during their migration across the Atlantic to spawn. The ability of animals to survive food deprivation is clearly of considerable survival value. Unsurprisingly, therefore, all animals exhibit adaptive biochemical and physiological responses to the lack of food. These responses prolong survival when food is not available. Equally important, however, these responses also help animals to (a) preserve physiological functions so that behaviors, such as physical activity to avoid predators or to seek food, can be maintained and (b) ensure that the animals can resume digestive and metabolic processes when food becomes available again.

OVERALL FASTING TOLERANCE IS DETERMINED BY ENERGY STORES RELATIVE TO USAGE

When animals experience food deprivation, they must derive the energetic costs for basal metabolism, physical activity, growth, and reproduction from the internal energy available at the onset of fasting. Large energy stores at the onset of fasting, therefore, obviously aid in prolonging starvation tolerance; animals commonly gain weight before dormancy. A reduction in metabolism also prolongs starvation tolerance, and various biochemical and physiological responses to food

deprivation contribute to the efcient use of resources. Animals reduce metabolism in diverse ways. Many animals inhibit reproduction and reduce both activity and body temperature. In fact, many animals breed only when food supplies are readily available, for example, in the springtime in temperate regions, and they do not breed during the winter, when food supplies are limited. The response to starvation is integrated at all levels of organization and is directed toward the survival of the species. Starvation has been more extensively studied in birds and mammals than in invertebrates and other vertebrates. Birds and mammals are special among vertebrates because they normally must eat at regular intervals owing to their high metabolic rates relative to their body stores. Most denitions and ideas of fasting come from the human literature. The extent to which our knowledge of fasting and starvation in birds and mammals can be transferred to other animals is not clear, as we discuss below.

The Phases of Fasting and Starvation in Mammals and Birds

The responses to absolute food deprivation in birds and mammals proceed in stages, culminating in death. The initial period involves fasting, and the later stages starvation. The demarcation between these two states is rarely appreciated, perhaps owing to lack of denition. In humans, fasting often refers to abstinence from food, whereas starvation is used for a state of extreme hunger resulting from a prolonged lack of essential nutrients. In other words, starving is a state in which an animal, having depleted energy stores, normally would feed to continue normal physiological processes. The metabolic transitions of food deprivation in birds and mammals have been divided into three phases. Most investigators probably would agree that the transition from fasting to starvation occurs by the end of phase II or the start of phase III. Alternatively, and not in conict with this view, others have argued that the transition between fasting and starvation may occur whenever animals opt to abort voluntary anorexia (sensu Reference 5). In our view, fasting should denote voluntary anorexia. Thus, salmon moving upstream and eels crossing the Atlantic to spawn are fasting rather than starving. The stages of starvation, if they exist in sh, amphibians, and reptiles, span a much more extended time frame, and the distinction between fasting and starvation becomes somewhat esoteric. In any event, the severity of the food deprivation and time required before an animal enters actual starvation vary among species, and the transition to starvation will depend on individual responses and nutritional status as well as a number of environmental conditions. Thus, whereas a small endothermic animal may be starving within a day of lacking food, it would take much longer for large ectothermic animals to undergo the same transition. Future studies should address questions such as whether large predatory ectotherms, such as sharks, that have not fed for a few days, are starving or fasting while waiting for the next meal. Similarly, is a hibernating ground squirrel starving or fasting? The different phases of response to food deprivation, which are dened according to the progressive metabolic changes that occur, were initially characterized

for humans and other mammals (e.g., 1114). These stages have been successfully applied to birds (8, 9, 15), and although not thoroughly investigated in ectothermic vertebrates, the overall progression in metabolic adaptation appears similar for most vertebrates. As a major difference, however, resting and maximal metabolism are much lower in ectothermic vertebrates than in endotherms (e.g., 16). For a given body composition and amount of energy stores, ectothermic vertebrates therefore can maintain normal metabolic functions for much longer than can endothermic animals, thereby deferring the detrimental consequences of food deprivation. Thus many ectothermic vertebrates can tolerate more lengthy starvation than can endotherms. Eels, for example, can migrate many thousands of kilometers over almost a year without feeding and may survive lack of food for many years (17, 18), whereas a similar-sized mammal would die from starvation within a few days or a week. Likewise, small animals are much more susceptible to food deprivation than are larger animals. In mammals, the three metabolic phases during food deprivation are characterized as follows on the bases of the primary fuel available for use and the associated changes in overall body mass: Phase I. The postabsorptive phase is the initial phase of fasting immediately after the last meal has been absorbed from the gastrointestinal tract. During this period, which normally lasts for hours, metabolism is largely fueled by glycogenolysis, or glycogen depletion of liver stores, which maintain constant blood sugar levels. In addition, fatty acids are liberated from adipose depots, and the availability of plasma fatty acids allows for some tissues, such as skeletal muscle, to spare the overall use of glucose. Phase II. When liver glycogen stores are depleted, gluconeogenesis becomes necessary to supply the requirements of glucose-requiring organs such as the brain. In humans, the initial fuel for gluconeogenesis is amino acids from proteolysis of muscle protein, but this contribution falls markedly as increased amounts of glycerol, another substrate for gluconeogenesis, is liberated from adipose tissues. Increased oxidation of fatty acids leads to an elevated production of ketone bodies, which can be used as an oxidative fuel in many tissues including the brain. As phase II progresses, protein degradation is rather slow, and degradation of adipose tissue fuels most bodily metabolism. In owls, lipid contributes more than 90% of the energy consumption in phase II, and approximately 2.5% of the energy consumption is derived from protein (19). In humans, this state can be maintained for several weeks and has often been referred to as a period of adapted starvation. Because of the high energy content of lipid, weight loss is rather slow during this state. Phase III. If starvation continues until the adipose stores are depleted, muscle is rapidly degraded for gluconeogenesis. The rapid loss in muscle mass cannot be sustained for long and eventually kills the animal. Figure 1 presents these three phases in rats. It shows how, as rats enter phase III, the rate of body mass loss along with nitrogenous waste production and its excretion increase as a result of protein degradation. Although similar metabolic changes and the transitions between fasting and starvation remain to be studied in detail for ectothermic vertebrates, numerous

Figure 1 Changes in body mass, daily loss of body mass, and excretion of nitrogenous waste in rats during the three phases of starvation (modied from Reference 48).

studies have reported on the gradual but slow decrease in body weight and somatic indices as food is withheld. Most species studied utilize fat before protein is degraded (e.g., 10, 2024). Other reports suggest that in ectotherms, as in mammals, glycogen is utilized even before lipid or protein (25, 26). Table 1 shows expression proles of the main energy-generating pathways that are related to the three phases of the mammalian response to starvation. Mammals have to utilize energy reserves much earlier than sh in response to starvation. Genes that encode protein products in lipolysis and protein turnover were induced after 24 and 48 h of starvation in mice and rats. Figure 2 shows that starvation did not trigger signicant changes in gene expression in carp until after at least 16 days of food deprivation. Lipolysis genes, such as -oxidation, remained unchanged in carp liver throughout the six weeks of food deprivation, reecting the fact that, unlike in mice and rats, hepatic lipid utilization was not enhanced in carp. Carp hepatic ubiquitin-proteasome genes were upregulated by approximately 1.3-fold after 28 days of starvation but did not trigger a signicant decline in total hepatic protein content. In fact, hepatic protein appears to be well conserved during starvation in carp (2730). Of course, before mobilizing hepatic reserves, carp use other lipid sources such as visceral lipid, as do rainbow trout (31). Carp contained a large amount of visceral lipids, although they were not quantied in the experiment. Carp hepatic glycogen, on the other hand, was mobilized during the rst four days of starvation and then declined again after six weeks, which coincided with an increase in glycolytic gene expression. Hepatic glycogen was not exhausted completely in carp after 100 days of starvation (32). Early mobilization of hepatic glycogen in carp may be related to glucagon release during the initial phase of starvation; this has been observed in teleosts and may be a response to stress rather than starvation (24, 33). Migrating salmon utilize lipid and spare protein until later in the migration phase, when lipid stores are almost completely depleted (10).

REDUCTIONS IN ENERGY EXPENDITURE DURINGFOOD DEPRIVATION

Responses to starvation occur at the behavioral, physiological, biochemical, and molecular levels. In general, the time to reach starvation-induced death increases with body mass (3437), reecting larger animals greater abilities to lower specic metabolic rate and increase stores of energy. Reductions in energy expenditure can also occur via a reduction in body temperature, which reduces metabolic rate (the Q10 or Arrhenius effect). Animals also can decrease energy expenditure during starvation by reducing locomotor activity as well as other behavioral and physiological functions such as reproduction and care for young. Many of these activities, although interconnected, are often studied in isolation. Protein synthesis is decreased, and expression of many metabolic genes is also reduced. Whether gene expression is reduced in response to decreased energy expenditure or vice versa is not clear.

Gene expression of mammals and carp liver in response to starvation. Upregulation of genes involved in lipolysis and proteolysis is observed only in starved mammals after 1 to 2 days of starvation, but it is not observed in common carp even after weeks of food deprivation. For details of gene lists, please refer to original publications.
TABLE 1

SSH technique identies only induced butnot suppressed genes.

Expression proles of statistically signicant genes of carp liver indicate that gene expression does not change until after 16 days of starvation. Genes that have similar expression proles are clustered together.
Figure 2

Basal and Resting Metabolic Rates

Humans and other mammals decrease resting metabolism during fasting and starvation (38). A very pronounced example is that of the golden spiny mouse, Acomys russatus, which inhabits dry deserts in the Middle East. Within one day, this small rodent apparently reduces oxygen uptake to half of the normal value and maintains this low metabolism when kept on restricted food availability for two weeks (39). This reduction takes places without changes in body temperature and may represent sympathetic control of energy-requiring processes (39). However, a recent study on the same species showed that food restriction elicits a reduction in both body temperature and metabolism that resembles torpor in other mammalian species (40). Nevertheless, although most animals do seem to lower body temperature when food is limited, substantial reductions in basal metabolic rate may occur. In salmon, for example, oxygen uptake decreased gradually over approximately two months when food was withheld (22). A reduction in basal metabolic rate, with no attendant decline in body temperature, requires that some energy-requiring processes be reduced at the cellular level. The changes in cellular metabolism responsible for a reduction in basal metabolism in fasting or food-restricted animals have not been studied, but the response may involve some of the same mechanisms as those occurring during the metabolic reduction observed during hypoxia. Many vertebrates respond to lack of oxygen by lowering protein synthesis, and a lower membrane permeability decreases the demand for active ion transport. Cell cycle may arrest, and cell proliferation may decrease, leading to the observed reduction in growth (41). Certainly food restriction and therefore the lower rate of intestinal nutrient uptake should also decrease protein synthesis, cell proliferation, and growth. It is difcult to envision, however, that changes in membrane properties would occur without an associated loss of function. Measuring basal metabolic rate while controlling for changes in spontaneous activity or alertness and sleep is difcult. Also, as described in more detail below, digestion also involves specic dynamic action and a rise in metabolism of animals seemingly at basal conditions. Fasting or starvation may therefore be associated with an apparent decline in basal metabolism that actually should be ascribed to a gradual cessation of digestive processes. The mammalian digestive tract displays great morphological and functional changes in response to starvation. Epithelial cell renewal and cell migration from crypt to villi tips were both reduced in starved mice (42) and rats (43). Total intestinal and jejunal mucosal mass was decreased by a factor of two in fasting rats, accompanied by reductions in jejunal crypt size and villi size and numbers as well as increases in villi tip cell apoptosis (44). Another recent study on rats, however, reports only minor changes in intestinal villus apoptosis (45). In any event, these studies concur that during phase III, apoptosis decreases and cell proliferation and migration increase (44, 45). Altogether, this upregulation of cellular events and presumably of absorptive capacity may reect preparation for future feeding (45). The restoration of the intestine during phase III

of starvation may be related to a refeeding signal, which has been described in penguins (46), and/or to behavioral changes in food hunting that mammals display during this phase (47). Intestinal restoration may also be related to increased expression of genes encoding orexigenic, hypothalamic peptides such as neuropeptide Y, agouti-related protein, and pro-opiomelanocortin (48). Rapid restoration of intestinal structures was observed during refeeding in mammals (49). This restoration took place as quickly as 30 minutes after refeeding following phase II starvation (43). By contrast, starvation had no signicant effect on the intestinal tract of the common carp. After 42 days of starvation, intestinal mucosal thickness of carp was not affected, and cellular events (e.g., apoptosis and cell proliferation) of the gut remained active. The expression of many digestive genesincluding those for chemotrypsin A and B, elastases, trypsins, carboxypeptidase A and B, proproteinase E, and amylase 3were suppressed greatly during starvation in carp, with most of the downregulation occurring after 16 days of starvation. The acquisition and processing of food are expensive processes; their cessation is manifested as a reduction in basal metabolic rate. The large reduction in gut mucosal surface area during starvation probably results from a large reduction in energy expenditure in maintaining the gut. This reduction in gut energy expenditure constitutes a part of the overall bodily reduction in energy expenditure during starvation. Several studies reporting on metabolic depression of fasting animals did not report body temperature, and some of the decline in metabolism may stem from hypothermia. Thus, part of the alleged reduction in standard metabolic rate during food deprivation may be ascribed to factors that lead to a reduction in cellular metabolism.

Body Temperature
In both birds and mammals, fasting and the associated depletion of energy reserves are important physiological cues to initiate torpor, which is a reduction in body temperature during inactive parts of the diurnal cycle (50). Thus, in some species, torpor occurs only when energy stores have reached a certain threshold and can be prevented by articial administration of nutrients such as glucose (e.g., 36, 40, 51 53). The gradual depletion of energy stores may also explain why the hypothermic response is enhanced as fasting is prolonged (e.g., 54). Torpor reduces energy usage by the direct effect of temperature on metabolism and because the metabolism of activity is negligible. Hypothermia is more pronounced in small as compared to large mammals and birds (e.g., 5557). Some hummingbirds, for example, may decrease body temperature by as much as 30 C. Small animals may benet from this body-size effect because of their higher mass-specic metabolism and greater ease for heat transfer due to their large surface area relative to body mass. However, body mass alone does not explain the occurrence and patterns of torpor. Many small birds, such as passerines, rarely reduce body temperature by more than 35 C, whereas some larger birds can undergo much larger changes (e.g., 5759). Torpor appears to be more pronounced in animals that inhabit areas in which large

uctuations in temperature and food availability are common, such as deserts. However, torpor also occurs in laboratory rats (e.g., 60) and may therefore be a rather common and widespread response. An entrance into torpor in response to food deprivation has been described in various animals. Nocturnal reductions in peripheral temperatures, associated with lower heart rate and presumably reduced metabolism, occur in large mammals such as red deer and reindeer (e.g., 61, 62). Torpor also occurs in primates; for example, torpor in response to food deprivation has recently been documented in the gray mouse lemur, Microcebus murinu (63). Also, barnacle geese and Puerto Rican todies undergoing long-term migration without feeding reduce body temperature by several degrees (64, 65). However, as Schleucher (57) points out, neither food supply nor energetic stress per se appear to be the ultimate factor determining the hypothermic response in these species, as the response is more pronounced in fatter premigratory birds. Torpor, therefore, may be a strategy to reduce energy expenditure during accumulation of fat stores. Thus, in addition to fasting and energy status of the individual, diverse ecological, morphological, and physiological variables, breeding, or migration periods, as well as physical parameters such as weather and annual cycles, are likely to inuence the extent to which different endotherms utilize torpor (e.g., 57). Ectothermic animals rely on external heat sources and appropriate behavior to regulate body temperature, and when provided with these opportunities, they maintain remarkably constant and well-regulated body temperatures. The effects of food deprivation have been studied in a few species freely selecting body temperature in laboratory settings. Several of these studies on sh and lizards have shown reductions in the preferred body temperature by a few degrees, which develops progressively as food is withheld (e.g., 66, 67).

Physical Activity
Decreasing physical activity and allowing body temperature to decline are likely to contribute more to energy sparing, and thereby to tolerance of starvation, than do reductions in basal metabolic rate, which are comparatively small. When food is not available, however, animals may search more actively for food, at the expense of increased energy usage, or decrease activity so as to reduce energy expenditure. An animals use of these alternatives depends on its foraging mode, the causes of food deprivation, and many other aspects of the animals natural history. In general, sitand-wait predators are likely to reduce activity when food is not available, whereas active hunters and grazers are more likely to increase activity as they search for food. Furthermore, although many animals reduce physical activity during the initial phases of fasting, many other animals exhibit a marked stimulation of activity during the later and more critical phases of starvation. In captive rats, food deprivation leads to reduced physical activity during the initial phases of food deprivation (68), followed by a marked hyperactivity when the animals enter phase III of starvation (47; see also Reference 69). Similar events

ensue in captive emperor penguins, in which the transition from phase II to III and the associated depletion of fat stores coincide with increased activity and escape behaviors. Teleonomically, these responses appear benecial, as the transition to phase III of starvation signies that existing resources are limited and that need for food is acute. Some [but not all (23, 66)] fasting sh and amphibians reduce activity. Mendez & Wieser (21) proposed that the behavioral response of sh to starvation consists of three phases, which has some resemblance to the biochemical changes outlined above. The rst phase is short lasting (approximately 24 h) and involves the increased activity of food searching. A transition phase, in which the sh gradually reduce swimming activity and thereby lower energy expenditure, then follows. The third and nal phase, adaptation, is characterized by low activity and metabolism, which persist until the sh are presented with the possibility of food. Van Dijk et al. (66) did not observe the stress phase in fasting roach (Rutilus rutilus), and it is quite likely that the specic responses will vary among sh with different behaviors and with the experimental setting.

Reproduction
Reproduction is energy expensive and requires either large energy stores in the mother or a ready source of food for both parents and offspring. Reproduction in many animals coincides with a high probability of food. Starvation is an inhibitor of reproduction in vertebrates; for example, most anorexic human females do not menstruate and cannot conceive (38). Female hamsters generally fail to ovulate and show little interest in sex if deprived of food for one or two estrous cycles. Ovulatory failure in these animals is related to an absence of an ovulatory gonadotropin surge and a set of immature follicles (70). Starvation for three days suppresses sexual receptivity in female rats, and this is associated with a reduction in the estrogenic response in the ventromedial nuclei of the hyopothalamus, critical for some reproductive behaviors (71). Sexually mature zebrash spawn daily, but when they are starved, the number of eggs they produce per day drops rapidly (Figure 3) (26). The prompt decline in egg production is associated with decreased expression of CYP19a, an enzyme that converts testosterone into estrogen in female zebrash (Figure 4). Not all animals exhibit inhibition of reproduction during starvation. Tessier et al. (72) suggested that reproduction during starvation may be advantageous for short-lived species. Some rotifer species maintain or even increase egg production during starvation (73), whereas other rotifer species reduce reproduction and survive starvation for longer than the reproducing rotifer species. Those rotifers that increase reproduction when energy supplies are limited invest resources in their offspring, which presumably have a better chance of surviving starvation. This is at the expense of the parents survival, possibly because of the accompanying benets of predator avoidance, reduced energy requirements of the young, and/or increased chance of moving to a resource-rich environment.

Figure 3 Daily egg production of zebrash when fed continuously for 10 days and after 11 days of food deprivation (fasting). A parallel experiment on continuously fed sh, which served as a control group, is shown on the right side of the gure. Each group represents results from 26 pairs of zebrash, sex ratio 1:1, and presented as mean S.E. Food-deprived sh produced signicantly lower numbers of eggs produced than did fed sh (modied from Reference 26).

Some vertebrates starve during reproduction; the Pacic salmon and eel are classic examples. Yellow eels feed and grow in freshwater, but they stop feeding when they become silver and start their migration across the Atlantic to spawn in the Sargasso Sea. Eels that have been starved in both seawater (74) and freshwater (75) for up to three to four years have survived. During this time, these eels lost

Figure 4 RT-PCR analysis of CYP19a mRNA expression in female gonad of control and zebrash starved for 11 days.

between 70 and 80% of their body weight. Interestingly, starvation was associated with an increase in growth hormone (GH) brain-cell hypertrophy and plasma GH levels (75). GH binds predominately to liver cell membrane receptors, but the increase in plasma GH during starvation is not associated with an increase in liver growth, as these liver GH receptors are downregulated. Atlantic salmon and trout species spawn repeatedly, whereas Pacic salmon die after their rst spawning. The lifestyles of Pacic salmon vary between species. Sockeye salmon, Oncorhynchus nerka, are born in freshwater and enter the ocean as one- or two-year olds, weighing between 4 and 15 g. They return to their natal stream to spawn after two or three years at sea, weighing between 1.6 and 3 kg (76). These animals do not feed once they have entered the river. The upstream migration may be more than 1000 km, depending on the river. Not only do these animals swim such large distances but they also produce numerous eggs. The gonad of the female reaches 14% of the pre-spawning total body mass. The starving migrating Pacic salmon use fat to fuel both their upstream migration and egg production; at death, both sexes will have expended more than 95% of their fat reserves (10). The fat reserves when the sh enters the river are, in general, proportional to the distance to be traveled. Female pink salmon spend less energy on migration than do males but more on gonad production such that, in the end, total energy expenditure is approximately the same for both sexes. When fat reserves start to exhaust, rst protein (from white muscle, but not heart or red muscle) and then carbohydrates are utilized (77). When the sh reaches the spawning ground, the caloric content of the sh is reduced to less than half of that which existed when the sh entered the river (10). There is increased interrenal activity and cortisol production, presumably directing some of the metabolic changes. The sh spawn and are usually in a moribund condition associated with energy depletion. All of the sh die, but they are not all moribund, and the cause of death is not clear, although the depletion of energy reserves must be an important component. Other salmonids, such as trout and Atlantic salmon, are repeat spawners. Immediately after reproduction, survival of the parent rather than the offspring is usually favored; this is particularly true for vertebrates. In penguins that fast on the ice during incubation of their eggs, the transition to phase III also leads to increased activity, and parents will abandon their eggs to secure their own survival at the expense of successful reproduction (46). There are, however, exceptions. Some species of octopus continually ventilate their developing embryos; feeding behavior is inhibited during this time, and these octopi can starve to death in the process. Adult cuttlesh, Sepia ofcinalis, migrate toward coastal waters to spawn and then die a month later. They also stop feeding and age rapidly during this period, with a marked deterioration in long-term memory. The cessation of feeding is associated with defects in visuomotor coordination as a result of degenerative changes in the central nervous system (78). There may be selection for genes that cause the rapid death of the postreproductive, or even just older and largersized, individuals within the population. This may be the case in spawning salmon and lampreys. Nothing is known of such death genes, but if they exist, they may

be functionally similar to genes observed in Sepia (78), in which degenerative changes in the central nervous system lead to a loss in prey capture ability and, as a consequence, death by starvation.

PHENOTYPIC PLASTICITY OF THE VISCERAL ORGANS IN RESPONSE TO DIGESTIVE STATE

The size and functional capacity of most visceral organs and muscle change in response to the physiological demands that are placed upon them (e.g., 79, 80). This phenotypic plasticity is pronounced for the gastrointestinal organs, which undergo a marked structural and functional reduction during fasting. The gastrointestinal organs are very metabolically active and have been estimated to account for as much as 40% of basal metabolic rate (e.g., 81). Thus, a reduction in organ size during fasting may confer a signicant energetic savings, which may contribute to a marked reduction in the basal metabolic rate of fasting animals. Within nonmammalian vertebrates, the effects of food deprivation on gastrointestinal organs have predominantly been studied in snakes. This group of reptiles has attracted particular interest because they tolerate very prolonged fasting periodsin some cases up to several yearsand because they can ingest very large meals. Thus, under natural conditions, some snakes may eat only a few times a year, but when they do eat, they can consume prey items of 50% of their own body mass or more (e.g., 82, 83). When snakes such as pythons or rattlesnakes eat these large meals after fasting for a few weeks or longer, the mass of the small intestine increases drastically within the rst 1224 h after ingestion (8486). The gut wall of reptiles, like that of other vertebrates, consists of an outer muscular coat and an inner mucosal layer with an epithelial lining toward the gut lumen (80, 87, 88). The mucosa in particular increases in mass upon feeding (Figure 5) (85, 89), and is attended by a many-fold increase in the transport capacities for various amino acids and glucose (85). This rise in nutrient transport capacity likely reects that the length of the intestinal microvilli increases almost vefold within 24 h (Figure 6) (90). In ectothermic vertebrates, whether the nutrient transport proteins are being synthesized de novo as the enterocytes expand and microvilli lengthen, or whether the increased nutrient transport capacity merely reects the increased surface area and that more transport proteins are exposed to the lumen, is unknown. When expressed relative to the mass of the intestine, nutrient transport capacity actually increases in mammals during hibernation (9193). Thus, although the mucosa atrophies, the intestinal transport proteins are well preserved, and the mRNA levels of the transporter protein SGLT1 does not change during hibernation in ground squirrels (94). This is also the case for the activity and mRNA levels of Na+ , K+ ATPase (94). The membrane potential of the enterocytes actually increases slightly during hibernation, which enhances the Na+ gradient that drives many of the intestinal nutrient transporters (92). It is not clear how this hyperpolarization

Figure 5 Effects of feeding on the mass of the small intestine of the snake Python molurus before and after ingestion of a meal equalling 25% of the snakes body mass. Intestinal mass is shown for fasting snakes (time 0). Each bar represents the total mass of the intestine and is divided into three parts representing the three parts of the intestine: proximal (black), middle (gray), and distal (white). The data are modied from Reference 90.

of the enterocytes affects cellular metabolism, and it certainly would be of interest to perform similar studies in ectothermic vertebrates in which cellular functions could be compared at similar temperatures in fasting and digesting animals. In contrast to that of the mucosa, the thickness of the intestinal muscle layer appears unchanged (85, 89). Furthermore, digestive status does not affect the total number of gut neurons in the intestinal muscle layer, spontaneous activity of the muscle layer in vitro, or the motility responses of isolated intestinal preparations when exposed to various excitatory neuropeptides (87). The mass and nutrient transport capacity of the gastrointestinal system undergo progressive reductions during subsequent food deprivation. Although other species of nonmammalian vertebrates have received much less attention than have snakes, phenotypic plasticity, both in terms of mass of the organs and their functional correlates, is seemingly universal, albeit less pronounced, in species with a more continuous feeding pattern, where prolonged periods of food deprivation are less common (e.g., 95). Thus, a progressive reduction in the intestinal epithelium during fasting, which is rapidly reversed after feeding, occurs in all other major groups of ectothermic vertebrates (e.g., 9598, 150, 151).

Figure 6 Electron micrographs of proximal intestinal microvilli of Python molurus during fasting (a) and at 0.25 (b), 0.5 (c), 1 (d ), 3 (e), 6 ( f ), and 14 (g) days after ingestion of a meal equalling 25% of the snakes body mass. Note the immediate lengthening of the microvilli and the subsequent regression (bar = 1 m). Modied from Reference 90.

Organ growth can occur by increased cell size (hypertrophy) or cell proliferation (hyperplasia). A two- or threefold increase in mucosal mass through hyperplasia would require extremely high rates of cell division and mitotic activities and presumably would be energetically expensive. Several recent studies on snakes and other ectothermic vertebrates have shown that the feeding-induced rise in intestinal mass is due predominantly to increased size of the individual enterocytes (84, 89, 90, 99101), suggesting that hypertrophy is the major mechanism. Thus, although cell proliferation may start early in the digestive phase, cell division reaches its maximal rate rather late in the digestive process. Therefore, the cells that have been worn down during digestion may be replaced. In this manner, the fully functional gut can be rapidly restored when food becomes available again (89). In all ectothermic vertebrates, the epitheliumwhich in fasting animals is pseudostratied, with folded cell membranes of neighboring cellsmay be rapidly unfolded after feeding and converted to a single layer of cells with stretched membranes as the enterocytes expand (reviewed in Reference 80). Each enterocyte appears to swell owing to a very rapid incorporation of lipid droplets (89, 90, 100), and there is also a small increase in uid content as relative wet mass of the intestine increases (87). Although evidence is still inconclusive, the lipid droplets likely come from the ingested food; alternatively, some of the lipids may stem from fat bodies in the body of the predator. Although incorporation of the lipid droplets certainly must account for a major part of enterocytic expansion, Starck & Beese (89) also have suggested that increased lymphatic pressure contributes. The increased water content of the enterocytes, however, cannot be caused by lymph pressure per se but would require movements of osmolytes such that osmotically obliged water is dragged along. Clearly, this aspect needs further experimental clarication. The structure and function of the intestines of birds and mammals are also exible. However, mammals and birds normally feed on a much more regular basis than do ectothermic vertebrates, and the former generally have a constant renewal of the gut epithelium. Structural and functional changes occur rapidly

after food deprivation in small mammals, whose high metabolism places extra premium on energy-saving mechanisms. For frequent feeders that are not adjusted to long periods of fasting, prolonged food deprivation or actual starvation may be more destructive to the gut, as compared to animals that normally experience long periods of fasting. The reduction in intestinal mass is due to atrophy, and the restoration of the gut upon subsequent feeding is accomplished by hyperplasia, although hypertrophy also contributes (e.g., 49, 102105). This is also the case in hibernating mammals, although lower body temperature and metabolism greatly extend starvation tolerance (9193, 106). The evolution of endothermy, which occurred independently in birds and mammals and has required much higher rates of nutrient uptake across the gut because of the high metabolism requirements (107), seemingly has led to a structure for which gastrointestinal and digestive plasticity is energetically more expensive. The signals that elicit the growth of the gastrointestinal organs are not well understood in nonmammalian vertebrates. In mammals, gastrointestinal growth can be elicited through luminal, hormonal, neural, and secretory pathways (e.g., 108). Although these regulatory pathways appear phylogenetically old and conserved (e.g., 88), few studies have experimentally investigated the respective roles of the individual mechanisms. Secor et al. (109) performed systematic infusions of nutrients into the intestine of fasting animals. Infusion of amino acids or protein directly into the intestine increased intestinal mass and transport capacity, whereas infusion of glucose, lipid, or bile had no effect (109). However, only infusion of homogenized rats caused a structural and functional response equivalent to that elicited by a normal meal (109). Cephalic responses, investigated by allowing the snake to constrict a prey item, followed by its removal, did not affect the intestine (109). Luminal signals predominantly from protein, therefore, appear sufcient for intestinal expansion and rise in transport capacity during digestion. However, the mucosal mass and transport capacity of surgically isolated portions of the intestine (Thiry-Vella loops) increase in voluntarily eating snakes (101), so hormonal and/or neural pathways also seem to contribute to gut expansion. Circulating levels of a number of regulatory peptides, some released from various gastrointestinal organs, increase dramatically during digestion (110); some of these peptides may serve as trophic factors.

THE METABOLIC RESPONSE TO DIGESTION:SPECIFIC DYNAMIC ACTION

Lavoisier rst showed that metabolism increases in response to digestion, and this metabolic response, documented in all animals investigated, now represents a general phenomenon. The postprandial rise in metabolism, normally referred to as the specic dynamic action of food (SDA), includes the energetic costs associated with the ingestion, digestion, absorption, and assimilation of the food. Thus, the physiological mechanisms that underlie the SDA response may vary among

different animals depending on feeding habits, food composition, temperature, and other factors. In the past ten years, SDA in carnivorous reptiles has received much attention, owing both to its magnitude and its potential to elucidate the large structural and functional changes in the gut. Also, in the animals that exhibit a large SDA response, the costs of digestion may account for a large proportion of the total energy budget, and the metabolic response to feeding becomes ecologically relevant (e.g., 111, 112). The SDA response is normally characterized as the factorial rise in oxygen uptake and by its duration. Another useful manner of expressing the response is via the SDA coefcient, which is the integrated metabolic response, calculated in caloric equivalents, relative to the energy consumed. Although some have criticized this parameter (113), it provides information on the energetic costs of digestion and allows therefore for a quantitative comparison between digestive responses under different environmental parameters or among different types or amounts of food ingested. There does not appear to be an anaerobic contribution to the SDA response, and the entire response therefore is reected in the rate of oxygen uptake (114117). The causes and determinants of the SDA response in vertebrates is beyond the scope of the present review, and this area has been summarized elsewhere (95, 118120). Also, the respiratory and cardiovascular correlates of the high metabolic rate during digestion have been reviewed recently (121, 122). The effects of fasting duration on the SDA response can be a particularly insightful example of the phenotypic plasticity of the gastrointestinal organs. Thus, Overgaard et al. (123) studied the effects of the previous fasting duration on SDA response. Upon feeding, animals exhibit elevated intestinal mass and function for many days, suggesting that if the expansion of the gut is energetically expensive, then a second meal, ingested while intestinal function is still elevated, should elicit a SDA response smaller than the rst response. Overgaard et al. (123) showed that the SDA coefcient does not change with a fasting duration between 360 days (Figure 7) and that intestinal growth does not constitute a major contributor to SDA response. Fasting duration does not affect the SDA coefcient in skinks or rattlesnakes either (124, 125). A small contribution of intestinal growth was also suggested for turtles (126). Collectively, these ndings are consistent with the proposal that intestinal expansion is structurally simple and energetically cheap (89). Secor (127) subsequently estimated that gastrointestinal upregulation contributes only 5% of the SDA response in pythons. A recent study of frogs, nevertheless, shows that the rate of digestion of the rst meal following three months of estivation is slower than for subsequent meals and that reconstitution of the gut may account for this delay (98). The efciencies of accumulation of various nutrients, however, were not affected (98). A recent study on snakes has implied that the stomach and the secretion of acid and digestive enzymes are the main contributors to the SDA response (97). In this study, the SDA response to a meal of 25% of the snakes body weight was reduced by more than half when digesting a liquid meal. The study further showed that the response was a third of its normal value when the liquid meal was infused

Figure 7 The SDA coefcient in the snake Python molurus following fasting periods of various duration. The SDA coefcient does not change with the duration of the previous fasting duration, indicating that structural and functional upregulation of the intestine occurs at a low energetic cost (modied from Reference 123).

directly into the small intestine. It was estimated therefore that gastric functions contribute 55% of the SDA response and that the stomach operates on a paybefore-pumping principle, in which the snakes must spend endogenous energy sources to initiate acid production and other digestive processes before ingested nutrients can be absorbed and used for metabolic pathways. To investigate this possibility further, we recently used another strategy: tying off of the pylorus, which is the anatomical connection between the stomach and the intestine, so that the chyme was unable to enter the intestine from the stomach. In the thusoperated animals, the SDA response was completely abolished, whereas shamoperated animals had a normal response (Figure 8). Visual inspection of the prey items clearly indicated that gastric functions had started digestion, and these data therefore suggest that secretion of acid and digestive enzymes can proceed at a relatively low energetic cost. That gastric acid secretion has a low energetic cost is further supported by the observation that treatment with omeprazole, a specic inhibitor of the H+ , K+ ATPase that drives gastric acid secretion, does not affect the SDA response in another snake species, Boa constrictor (128).

Figure 8 The metabolic response to digestion in snakes (Python molurus) in which the pylorus has been ligated to prevent chime from entering the intestine from the stomach (M. Andersen, H. Cueto, & T. Wang, unpublished data).

In most animals studied, the SDA response elicited by a given food type increases proportionally with meal size, and both the maximal oxygen uptake during digestion and the duration of the response increase as meal size increases (e.g., 129133). In most cases, the SDA coefcient is unaffected by meal size, indicating that the costs of digestion are proportional to the amount of food ingested. Although these data are often interpreted to reect that it is merely the caloric content of the food that determines the SDA, numerous studies have documented that protein-rich meals elicit larger metabolic changes than do diets composed of fat or carbohydrates. Thus, force-feeding reptiles with fat or carbohydrates elicits almost no metabolic response (e.g., 134138). This would indicate that stimulation of protein synthesis in response to high circulating levels of amino acids (139) is a major contributor to the SDA response (140). The role of protein metabolism in the SDA response (140, 141) is pivotal in fasting catsh, toads, alligators, and pythons in which either systemic infusion of amino acids, or infusion of protein or amino acids directly into the stomach, leads to a rise in metabolism that is comparable to that observed during normal feeding (138, 141144). In catsh and pythons, inhibition of protein synthesis with cyclohexamide completely abolishes the SDA response (138, 142, 143). If increased protein synthesis is indeed the

major contributor to the SDA response, metabolism of all organs should increase during the postprandial period, a reasonable suggestion in light of the very high growth efciency of snakes in which some 4060% of ingested food is directed to growth (123, 145, 146). Obviously, the resulting rates of growth must require protein synthesis in all organs.

SUMMARY AND FUTURE DIRECTIONS

Digestive status affects virtually all physiological and behavioral responses, and selective pressure to enhance feeding strategies and digestive processes must be signicant. The ectothermic vertebrates, with their lower metabolic rates, can endure prolonged periods of fasting, and many of these species exhibit much more pronounced changes in gastrointestinal organs than are normal in healthy mammals (see also Reference 147). The extreme structural and functional changes in their dynamic guts make ectothermic vertebrates useful models to explore largely unresolved issues regarding the interaction and prioritization of physiological functions among organ systems. These issues are of basic physiological importance. Such studies may contribute to our understanding of the mechanisms that enable organs to adapt to physiological demands. They also may help us to understand the factors that in humans can promote intestinal repair following either intestinal resections or diseases such as colitis and Crohns disease in which there is inammatory destruction. ACKNOWLEDGMENTS The authors are supported by the Danish Research Council as well as the Research Grants Council of Hong Kong Special Administrative Region, Peoples Republic of China (project number: CityU RGC1224/02M).