Biology_of_Turtles

sumption rates and increases in plasma lactate—begins at about 22 mmHg (Milton, 1994). Recent studies have shown a wide variation in the capacity of neurons to tolerate hypoxia, reflective of the function and the degree of hypoxia normally encountered (Milton & Prentice, in press). In the mammalian brain, the hippocampus, corpus striatum, and Purkinje cells are the most vulnerable to hypoxic or ischemic damage; within the hippocampus, CA1 neurons are more sensitive than cells of the CA3 region (Pulsinelli, 1985). For comparison, let us look first at the sequence of events in the mammalian brain that occur when oxygen is lacking—for example, as might happen in the event of a cardiac arrest, or more locally in the event of a stroke.

12.2The Brain in Crisis

As the end electron acceptor that makes mitochondrial oxidative phosphorylation possible, oxygen is critical to nearly all forms of life on earth. Severe hypoxia or cerebral ischemia forces tissues to rely on the ATP supplied by anaerobic glycolysis, which supplies only 2 moles of ATP per mole of glucose rather than the far greater amounts of ATP generated from oxidative phosphorylation. In theory, each mole of glucose supplies 36 moles of ATP by oxidative phosphorylation, though empirical measurements indicate that true ATP production is closer to 25 moles/mole glucose (Hochachka & Somero, 2002). When forced to rely on less than 1/10th of its normal energy supply, cellular ATP demand rapidly outstrips ATP supply and ATP-driven processes begin to fail. In the neurons, failure of ATP-dependent ion transporters leads to the breakdown of membrane potential; K+ leakage out of the cells triggers membrane depolarization and the release of excitatory amino acids (EAA) such as aspartate and glutamate (Lutz et al., 2003). Elevated levels of glutamate, due to both increased release and decreased reuptake, are generally regarded as a significant factor leading to hypoxic or ischemic cell death. High extracellular glutamate stimulates NMDA receptors and permits an influx of calcium from the extracellular space (Harukuni & Bhardwaj, 2006). Increased intracellular Ca2+ is thought to lead to neurodegenerative events including the generation of excess levels of reactive oxygen species (ROS), lipid peroxidation, and the activation of executioner caspases that trigger apoptosis (programmed cell death). The activation of glutamate receptors (Dugan et al., 1995; Sharp et al., 2005; Kahlert et al., 2005) and elevated intracellular calcium (Sharikabad et al., 2004; Nagy et al., 2004) are associated with free-radical formation, and the direct activation of NMDA receptors results in a massive release of hydroxyl radicals (OH) (Lancelot et al., 1998; Laplanche et al., 2003). High intracellular calcium also activates proteases, lipases, and endonucleases, which in turn destroy cellular integrity (Lutz et al., 2003).

In addition to glutamate related damage, hypoxic or ischemic events cause the release of the monoamine dopamine into the extracellular space; however, in contrast to other excitotoxins, increases are seen before cellular depolarization, even under conditions of mild hypoxia (Globus et al., 1988; Huang et al., 1994) and may be a major cause of hypoxic/ischemic damage (Mitsuyo et al., 2003). Dopamine is thought to contribute to neuronal damage by increasing the release of excitatory neurotransmitters, through the production of oxygen free radicals, by inhibiting Na+/K+ ATPase and by uncoupling glucose metabolism from cerebral blood flow (Lutz et al., 2003). Following hypoxia/ ischemia, additional neuronal damage occurs during reperfusion, thought to be caused by the postischemic release of oxygen radicals, the synthesis of nitric oxide, inflammation, and an imbalance between the excitatory and inhibitory neurotransmitter systems (Berger et al., 2002); this damage can continue for days after the initial hypoxic/ischemic insult.

12.3The Anatomy and Physiology of Anoxic Survival

By contrast, certain species of freshwater turtle can survive extended periods of anoxia and fully recover (Ultsch, 2006), as can the Crucian carp (Carassius carassius), a northern European fish that likewise winters in iced-over, anoxic, ponds (Lutz et al., 2003). Extended anoxic survival involves the downregulation of energy-demanding processes and those that are potentially damaging to

cellular survival, and a concurrent upregulation of a variety of protective mechanisms that decrease energy demand and promote cell survival. For example, factors that are downregulated include the enzymes of both glycolysis and oxidative phosphorylation (Brooks & Storey, 1988; Brooks & Storey, 1989), an overall decrease in protein synthesis (Fraser et al., 2001), the release of excitotoxic neurotransmitters (Nilsson & Lutz, 1991; Milton & Lutz, 1998; Milton et al., 2002; Milton & Lutz, 2005), and the activity of ion channels, all of which decrease electrical activity (Fernandes et al., 1997) and thus lower both energy demand and overall metabolic rate.

Conversely, a number of pro-survival mechanisms are either already constitutively elevated or are clearly upregulated, including increasing glycolytic energy supply through high constitutive tissue levels of glycogen (McDougal et al., 1968; Clark & Miller, 1973), increases in plasma glucose (Milton, 1994), increased brain blood flow (Hylland et al., 1994), the release of inhibitory neuromodulators such as adenosine (Nilsson & Lutz, 1992) and GABA (Nilsson & Lutz, 1991), as well as the maintenance of (Lutz & Manuel, 1999) or increase in (Lutz & Leone-Kabler, 1995) neurotransmitter receptor function. Changes at the molecular level also occur, including increased expression or activation of heat shock proteins (Prentice et al., 2004), members of the mitogen activated kinase families (Greenway & Storey, 2000; Milton, unpublished data), and the recently discovered neuroglobin (Milton et al., 2006).

12.3.1 Energy Supply and the Enzymes of Anaerobic Glycolysis

The upregulation of energy conserving processes, coupled to the downregulation of energy expenditures, leaves the turtle in a state of profound, reversible, metabolic depression. The key to survival is to lower energy demand to meet the reduced energy supplied by anaerobic glycolysis; thus, the ability to carry out widespread glycolysis is critical. A study by Suarez et al. (1989) comparing glucose use in rats and turtles showed that after accounting for temperature differences, there was at least a sixfold difference in metabolic rates between rats and turtles, suggesting that lower rates of ATP use in turtles are inherently protective in anoxia as they are more easily supported by anaerobic glycolysis. However, lower metabolic rates and temperature differences alone are clearly not sufficient for anoxia survival, as the brains of other poikilotherms respond to anoxia in a manner similar to mammals. Remarkably, at their respective biological temperatures, the glycolytic capacity of rats and turtles is similar, as determined by the activities of hexokinase and lactate dehydrogenase (Suarez et al., 1989). This indicates not only a greater reliance on glycolysis by turtles even under normoxic conditions but a ready capacity for anaerobic glycolysis in anoxia. Key enzymes that have been studied in relation to brain metabolism include both those critical to oxidative phosphorylation, primarily cytochrome c oxidase, and those of the glycolytic pathway including hexokinase, lactate dehydrogenase, and glycogen phosphorylase. As is also seen in mammals, there is a detectable general pattern of differences in metabolism between gray and white matter in the turtle brain, such as gray/white matter differences in glucose metabolism indicated by 2-deoxyglucose use and citrate synthase (Sokoloff et al., 1977; Suarez et al., 1989). These differences result in a repeatedly observed rostral-to-caudal gradation of metabolic indicators, including the enzymes of energy production and antioxidants, that in general declines in parallel with the increase in white matter content of the more caudal brain regions (Oke et al., 1987), perhaps because the metabolic rate of the soma (and greater numbers of energy consuming dendrites and synapses) is greater than that of axons and a greater proportion of gray matter is somata.

As the terminal oxidase of the electron transport chain, the activity of cytochrome oxidase (CO) is directly linked to oxygen consumption. As neural activity is constrained, at least in part, by the availability of cellular ATP, and the activity of CO in aerobic metabolism determines the level of ATP available to the cell, then CO activity in turn can serve as a marker of metabolic capacity (Sakata et al., 2005). Hence, CO histochemistry has been used extensively to follow changes in neural metabolism after experimental manipulations; for example, significant decreases in CO activity have been shown to follow experimentally induced decreases in excitatory afferent input

dominant. However, an alternative explanation is that it is more critical to maintain function in some regions of the brain during anoxia than in others. For example, in the striatum the release and reuptake of neurotransmitters continues during long-term anoxia (Milton & Lutz, 1998; Milton et al., 2002) despite energetic costs of reuptake transport mechanisms of as high as 1.5 ATP/molecule (Swanson & Duan, 1999). The continued release and re-uptake of these compounds implies a likely function for the neurotransmitters during anoxia, perhaps to maintain neuronal networks or prepare for the eventual upregulation of metabolism during recovery (Milton et al., 2002), but the energetic demands of transport also means a continuous energy supply is required. Thus, the high activity levels of both GP and CO would allow aerobic metabolism under basal conditions, with a change to anaerobiosis during anoxia at levels sufficient to maintain neurotransmitter homeostasis, albeit at a reduced level. Along these lines, Partata et al. (1999) note the remarkable overlap in distribution of GP and CO in the medulla, where HK activity was also found to be high (Xia et al., 1992). The authors also suggested that the CO and HK overlap is critical as CO may regulate the onset of glycolysis; lactate production is known to begin before CO is fully reduced in the anoxic turtle brain (Lutz et al., 1984; Lutz et al., 1985). By contrast, the dorsal ventricular ridge and most layers of the optic tectum expressed significant GP levels but not CO activity, indicating some continued function for the eye during anoxia (Rosenberg & Ariel, 1990; Fan et al., 1995, 1997). However, the BON showed detectable CO but not GP (Partata et al., 1999), though continued normal electrical activity has been shown to occur in the accessory optic system for several days in vitro (Ariel & Fan, 1993; Johnson et al., 1998). The sum of these metabolic enzymes studies then implies that the continued function of the brainstem in anoxia is critical to anoxic survival or recovery.

Alternatively, these studies did not measure enzyme activity levels during anoxia, and it has been known for many years that enzymes of the glycolytic pathway as well as those of aerobic metabolism are strongly downregulated in anoxia both in the brain and in other organs (Brooks & Storey, 1988, 1989). In turtles, as in other good facultative anaerobes, glycolytic activation is attenuated after the first hour or two of anoxia; as the rate of energy consumption gets more and more depressed, so does the rate of energy production (Kelly & Storey, 1988). Despite the widespread presence of glycogen phosphorylase reported by Partata et al. (1999), GP activity in the brain is reduced by 70% over the initial hours of anoxia (Brooks & Storey, 1988). Phosphofructokinase (PFK) shows increased inhibition as well, reflecting its role as the primary locus of glycolytic rate depression (Brooks & Storey, 1988). In most animals, hypoxia induces a sharp increase in anaerobic energy production (the Pasteur Effect) as metabolic supply attempts to keep pace with metabolic demand. The Pasteur Effect results primarily from the release of metabolic control at the PFK locus (Storey, 1988); in the turtle, covalent modification of enzymes results rather in a depression of metabolic rate (Brooks & Storey, 1988, 1989).

Energy production then continues in the anoxic turtle brain, albeit at greatly reduced rates that suffice to meet the reduced energy demands of the hypometabolic state, with some sections of the brain metabolically equipped for extended glycolysis. The distribution of glycolytic enzymes within the brain is likely to indicate critical areas that maintain function even in extended anoxia; the hypometabolism of anoxia is not simply a “shutting down” of the brain but is a selective process presumed to maximize energy savings while allowing critical processes to continue. Studies have revealed that whereas many processes are minimized, others continue in anoxia, including the continued release and reuptake of neurotransmitters, the upregulation of protective molecular pathways, and selective electrical activity (Lutz & Milton, 2004).

12.3.2 Ion Channels

As with all cells of the body, neurons maintain their ion balance across the cell membrane with a higher concentration of K+ ions inside the cell compared to the outside, whereas Na+ ion gradients are the reverse. These ion differences are used by the neuron to generate action potentials, and the restoration of ion gradients after an action potential—plus counteracting continuous leak channels

autoradiography of bound saxitoxin. As with the voltage-gated calcium channels, Na+ channels in the rat generally followed established metabolic gradients, with a very heterogeneous distribution. Saxitoxin binding in the rat brain was highest in the cortex, hippocampus, amygdala, and cerebellum and low in the brainstem and spinal cord (Xia & Haddad, 1993). Whereas saxitoxin in the turtle had similar binding properties as in the rat, overall binding density was much lower, especially in the cerebellum and rostral areas such as the cortex. Very few areas, including the medullary raphe, had similar or higher channel density in the turtle compared to the rat (Xia & Haddad, 1993). However, the differences in receptor density are unlikely to explain the approximately 100-fold difference in anoxia sensitivity between the turtle and mammalian brains, a conclusion also reached by Edwards et al. (1989). That study found that turtle synaptosomes had a voltage-gated Na+ channel density approximately a third that of rat synaptosomes—but again, this alone is of little significance when compared to the enormous differences in anoxia tolerance (Edwards et al., 1989).

Yet another difference in ion channels between turtles and rats lies in the density of ATPsensitive potassium channels (K-ATP). K-ATP channels were first discovered in cardiac myocytes (Noma, 1983) and have subsequently been found in many other excitable cell types (Ashcroft, 1988), including neurons (Mourre et al., 1989; Krnjevic, 1993). These channels are normally inhibited by physiological levels of ATP but open when ATP levels decrease (Ben Ari et al., 1990; Cameron & Baghdady, 1994). As the open probability of K-ATP channels is thus directly linked to intracellular energy (ATP/ADP) levels, the channels link the metabolic status of a cell to its electrical activity (Liss & Roeper, 2001). Sulfonylurea binding studies indicate the K-ATP are widely expressed throughout different regions of the brain, including pyramidal and striatal neurons, various nuclei of the hypothalamus, and GABAergic and dopaminergic neurons of the substantia nigra (Liss & Roeper, 2001).

K-ATP channels are thought to play a critical protective role in the early stages of brain ischemia in mammals (Ben Ari et al., 1990) through hyperpolarization of the membrane, by increasing brain blood flow (and thus glucose delivery) (Tomiyama et al., 1999), and decreasing the release of excitatory neurotransmitters (Tanaka et al., 1996; Wind et al., 1997). Blockade of K-ATP in the anoxic rat brain increases neuronal depolarization, thus suggesting that open K-ATP can dampen the anoxia-induced depolarization of CNS neurons (Jiang & Haddad, 1991). By contrast, K-ATP activation does not have any significant effect on K+ flux in newborn rats or adult turtles, suggesting that these channels are not a major route of K+ loss in more hypoxia-tolerant animals (Jiang & Haddad, 1991). Using the binding of glibenclamide, a specific inhibitor that binds to the sulfonylurea subunit of K-ATP channels, Xia and Haddad (1991) compared the binding properties and distribution of K-ATP in the rat and turtle brains with autoradiography. Highand low-affinity binding channels were revealed in both, but the rat had far higher binding densities and a more heterogeneous distribution. In the rat, there was a more than a ten-fold difference between the highest and lowest densities, with highest K-ATP levels in the substantia nigra, hippocampus, cerebellum, and a few thalamic nuclei. Intermediate levels were found in the basal ganglia, septum, thalamus, and hypoglossal nuclei, whereas densities were lowest in the cerebellum, brainstem, and spinal cord (Xia & Haddad, 1991). The cortex itself is highly variable; K-ATP are present at high densities in the hippocampal CA3 and CA4 regions and the dentate gyrus but are only intermediate in the CA1 and CA2 regions. Interestingly, the CA3 region has proven to be less sensitive to hypoxic-ischemic brain damage than the CA1 area (Pulsinelli, 1985); perhaps this is related to the protection afforded by higher densities of these ATP-sensitive channels.

In the rat brain, we again see a general rostral-to-caudal gradient, though there is less specific correlation between K-ATP channels and the enzymes of metabolism, indicating perhaps a greater association of K-ATP with specific pathways and nuclei rather than with generalized brain metabolic activity. For example, high densities in the substantia nigra could be related to the theoretical role of K-ATP channels in depressing the excessive release of excitotoxins such as dopamine in the face of an energy deficit; the substantia nigra has a high density of dopaminergic neurons that terminate in the striatum (Parent, 1979). As an aside, stimulation of these channels to increase

hypoxic/ischemic neuronal survival in the mammalian brain has been an interest for nearly two decades with little result.

However, in contrast to high K-ATP channel density in the adult rat, binding densities were very low in the newborn rat and turtle brains, both more hypoxia-tolerant than the adult rat. Average densities of the K-ATP channels were 25 and 36 fmol/mg protein in the turtle and newborn rat, respectively, versus a mean 219 fmol/mg in the adult rat (Xia & Haddad, 1991). Also in sharp contrast to the adult rat was the overall homogeneity of binding density in turtles: densities in most brain regions ranged from 7 to 25 fmol/mg protein in turtles, whereas in the adult rat values ranged from 46 fmol/mg in the spinal cord to around 309 fmol/mg in the substantia nigra. Whereas densities were lower and more homogeneous in the newborn rat than the adult, this trend was even greater in the turtle brain, especially in the more caudal areas. Xia and Haddad (1991) reported densities too low to measure in the spinal cord, intermediate binding in the rostral and cerebellar regions, and values as high as 40 fmol/ng in the striatum. K-ATP densities in the turtle brain are thus about 1/10th the values of the rat brain, in keeping with their lower overall metabolic intensity; the authors conclude that K-ATP channel density then could not be the critical difference that allows for extended anoxia in turtles (Xia & Haddad, 1991).

However, more recent work has shown that even if K-ATP density is not high in the turtle brain, K-ATP channel function is of critical importance in anoxia tolerance, performing many of the same protective functions as have been reported in mammalian neurons (although apparently with more success). In mammalian cells, the opening of K-ATP channels in response to low intracellular ATP is in part protective because the K+ outflow temporarily hyperpolarizes the cell. It is critical to realize that if ATP levels do not recover, the increased conductance will result in a more rapid efflux of K+ and shorten the time to depolarization (Lutz et al., 2003); indeed, the K-ATP channels are thought to be a major route of K+ efflux during anoxic depolarization in mammals (Xia & Haddad, 1991), though they appear of less importance in the depolarizing turtle brain (Pek-Scott & Lutz, 1998). In turtles, there is much evidence that membrane potential is maintained during anoxia (Lutz et al., 2003), and anoxia is associated with a decrease in K+ efflux that can be initially blocked by K- ATP blockers (Pek-Scott & Lutz, 1998). Whereas the addition of glibenclamide blocks this anoxiainduced reduction in K+ efflux during the first hour, the effect is diminished by 2 hours anoxia and completely disappears by 4 hours anoxia; these differences are thought to be associated with a temporary fall and subsequent recovery in cellular energy stores (Pek-Scott & Lutz, 1998). The temporary opening of K-ATP channels is protective in the turtle brain during a period of temporary energy crisis: open K-ATP are associated with the decreased release of both dopamine (Milton & Lutz, 2005) and glutamate (Milton et al., 2002) during a period when the turtle brain has perhaps less energy to spare on reuptake transport mechanisms.

However, whereas the distribution and densities of ion channels may not reveal differences between anoxia-tolerant and hypoxia-sensitive animals, or determine which parts of the brain are more or less active in anoxia, changes in ion channel activity are more revealing. Modulating the conductance of ion channels can result in significant alterations to metabolism. For example, increasing Na+ channel activity with veratridine increases energy consumption by 20% in rat and turtle synaptosomes (Edwards et al., 1989). Conversely, blocking voltage-dependent Na+ channels with tetrodotoxin causes a 22% decrease in the energy consumption of turtle synaptosomes (Edwards et al., 1989) and a 20% decrease in the in situ perfused rat brain (Xie et al., 1994). Thus, the ability to alter ion conductance could be a powerful mechanism to decrease metabolic costs by reducing the costs of ion pumping. Extensive evidence shows that ion flux through membrane channels indeed decreases in the anoxic turtle brain (“channel arrest”), including decreased K+ flux (Pek-Scott & Lutz, 1998) and a reduction in the density of voltage-gated Na+ channels (Perez-Pinzon et al., 1992). In the isolated turtle cerebellum, 4 hours anoxia produced a 42% decrease in the density of voltagegated Na+ channels (Perez-Pinzon et al., 1992).

Anoxia-induced decreases in Ca2+ influx through the glutamate-responsive N-methyl-D- aspartate (NMDA) receptors of the turtle cerebrocortex have also been well described (Bickler &