Hyperglycemia associated with diabetes mellitus will exacerbate neurologic injury after global brain ischemia. Studies in a rat model of forebrain ischemia (bilateral carotid occlusion plus hypotension for 10 min) discovered that acute restoration of normoglycemia in diabetics, using an insulin infusion, resulted in a neurologic outcome that was similar to normoglycemic rats without diabetes. The current study evaluated cerebral glucose, glycogen, lactate, and high-energy phosphate concentrations to identify metabolic correlates that might account for an alteration in postischemic outcome.
Fifty-four pentobarbital-anesthetized Sprague-Dawley rats were assigned to three groups: chronically hyperglycemic diabetic rats (D; N = 18); insulin-treated, acutely normoglycemic diabetic rats (ID; N = 18); and nondiabetic rats (ND; N = 18). These groups were further divided into groups of six rats each that received either no ischemia, forebrain ischemia of 10 min duration without reperfusion, or ischemia plus 15 min of reperfusion. Brains were excised after in situ freezing, and metabolites were measured using enzymatic fluorometric techniques.
Before ischemia, D rats had greater concentrations of brain glucose (12.18 +/- 2.67 micromol/g) than did either ID (5.10 +/- 1.33) or ND (3.20 +/- 0.27) rats (P < 0.05). Preischemic brain glycogen was similar in all groups. At the completion of ischemia, brain lactate concentrations in D were 86% greater than in ID and 61% greater than in ND (P < 0.05), reflecting a higher intraischemic consumption of glucose plus glycogen in D (P < 0.05). High-energy phosphate concentrations, as assessed by the energy charge of the adenylate pool, were better preserved in D (energy charge = 0.60 +/- 0.28) than in either ID (0.29 +/- 0.09) or ND (0.36 +/- 0.07; P < 0.05) rats. After 15 min of reperfusion, the energy charge returned to preischemic values (i.e., 0.91-0.92) in all groups.
These studies demonstrated greater intraischemic carbohydrate consumption and lactate production in D than in ID or ND rats. Under these conditions, intraischemic-but not postischemic-energy status was better in D rats. Acute insulin therapy in ID rats resulted in a metabolic profile that was similar to that of ND rats. These results suggest that, in this model, primary energy failure during ischemia is not the origin of greater injury in hyperglycemic diabetics, nor is energy enhancement the origin of improved outcome after acute insulin treatment.
Key words: Brain: cerebral metabolism; global cerebral ischemia. Diabetes mellitus: hyperglycemia. Hormones: insulin. Metabolism: adenosine triphosphate.
THERE is a greater incidence of cerebral ischemic insults in persons with diabetes mellitus than those who do not have the disease. [1-4]Additionally, as a result of their systemic vascular disease, persons with diabetes may be more likely to require cardiac surgery [5-7]and related therapies such as invasive angiography, in which there is a therapy-associated risk of cerebral ischemia. [7-9]When coupled with the recent observation that the prevalence of diabetes mellitus is dramatically increasing in the United States (i.e., from 11 million in 1983 to 16 million in 1995), anesthesiologists and intensivists can expect to care for more diabetic patients in whom there is concern for cerebral ischemia.
In addition to a high incidence of ischemia, once diabetics experience a cerebral ischemic insult, they have a worse outcome than nondiabetics. [2,3]Although a variety of systemic [1,3,4,10]and cerebral vascular [2,4,11,12]factors may contribute to the worse morbidity and mortality, another testable hypothesis is that diabetes mellitus-induced changes in brain carbohydrate content adversely affect periischemic brain metabolism, including brain energy metabolism. The latter issue was examined by our study.
Increases in brain glucose concentrations at the onset of ischemia (as occurs with untreated diabetes mellitus [13,14]) will worsen outcome after ischemia. The presumed operant mechanism is that the anaerobic metabolism of glucose results in an intracellular acidosis [2,16]that is toxic to the cell. [17-19]Furthermore, glucose originating as either glycogen or free brain glucose should contribute to the enhanced cellular acidosis. [2,20,21]Hydrogen ions accumulating from this process may poison the cell by either direct toxicity or by initiating secondary adverse effects. [17,19,22,23].
Some authors have theorized that the secondary adverse effects of lactic acidosis involve a detrimental influence on brain energy metabolism; however, as recently reviewed, the nature of that relationship is controversial.
Of further concern, chronic hyperglycemia might have an influence on cerebral glucose physiology, [13,14,21,24]resulting in dissimilar effects in persons with and without diabetes. Additionally, during attempts to restore normoglycemia in previously hyperglycemic diabetics, a possible insulin-associated hypermetabolic state might adversely affect high-energy phosphate concentrations.
The purpose of the current study in diabetic rats, nondiabetic rats, and insulin-treated diabetic rats, was to quantify brain glucose and glycogen concentrations in nonischemic brain. The study also determined the relationships among brain glucose and glycogen consumption, lactate accumulation, and concentrations of high-energy phosphates before, during, and after a 10-min period of near-complete forebrain ischemia. Finally, the study evaluated whether acute insulin-treatment of diabetes mellitus would restore brain metabolite concentrations to the nondiabetic state.
Materials and Methods
This protocol, conducted in 54 Sprague-Dawley rats, was reviewed and approved by the Institutional Animal Care and Use Committee of Mayo Foundation. One week before the study, diabetes mellitus was induced in 36 rats with intraperitoneal streptozocin 65 mg/kg. [14,21,25]All rats were fasted for 10-12 h before the study but had free access to water. Rats were weighed and then anesthetized in an induction box with 3% halothane in oxygen. The trachea was intubated via a tracheostomy, and the lungs were mechanically ventilated. Intramuscular pancuronium (0.5 mg) was given hourly to provide muscle paralysis, and anesthesia was maintained with 1.3% inspired halothane in nitrogen and oxygen for the remainder of the preparatory period.
Temperature was measured using both a rectal thermistor and a needle thermistor inserted beneath the temporalis muscle (73A, Yellow Springs Instruments, Yellow Springs, OH). Temperatures were maintained at 37 degrees C using a heating lamp and pad. A femoral artery and vein were cannulated with polyethylene catheters (PE-50, Becton Dickinson, Parsippany, NJ). Through a neck incision, the carotid arteries were identified and, using sutures, isolated. Thereafter, a Silastic cannula (ID = 1.2 mm) was inserted via the jugular vein into the superior vena cava, and the rats were given 50 units intravenous heparin. The superior vena caval catheter and the isolated carotid arteries would be used subsequently in the production of cerebral ischemia. Arterial blood gases were determined using electrodes at 37 degrees C (Instrumentation Laboratories, Lexington, MA). Blood glucose, plasma glucose, and plasma lactate were measured using a glucose and lactate analyzer (23A, Yellow Springs). The former device has a glucose detection range of 0-28 micro mol/ml (0-500 mg/dl) and a sensitivity of 0.05 micro mol/dl. .
The skin of the scalp was reflected, exposing the calvarium, and a funnel was secured to the exposed bone. The bottom of the funnel was filled with a sheet of Parafilm (American National Can, Greenwich, CT) to prevent heat and moisture loss. The continuous electroencephalogram was monitored from biparietal electrodes glued to the calvarium and was amplified and recorded using a polygraph and strip recorder (78B, Grass Instruments, Quincy, MA). The animal was then placed in the supine position, and the head was secured in a stereotactic head frame.
In earlier studies done in our laboratory, in which we evaluated carbohydrate metabolism in diabetic rats, [13,14]we discovered that, during periods of minimal surgical stimulation, many halothane-anesthetized rats were unable to maintain blood pressure within protocol criteria. This problem was subsequently avoided by using a background of pentobarbital anesthesia. Thus, on completing the surgical preparation in the current study, halothane was discontinued, and the rats were administered 6 mg pentobarbital intravenously. Supplemental pentobarbital was given in 2-mg increments to maintain a synchronized, continuous electroencephalogram pattern consistent with a surgical plane of anesthesia. [21,26]Once a stable electroencephalogram pattern was achieved, 10-min stabilization periods were allowed, after each of which ventilation and oxygenation were adjusted until the changes resulted in data within predetermined protocol criteria: arterial carbon dioxide tension 38+/-2 mmHg (mean+/-range), arterial oxygen tension 150+/-25 mmHg, mean arterial blood pressure > 60 mmHg, and temperature 37.0+/-0.5 degrees C. Inclusion into the study was based on the following criteria: Diabetic rats were required to have a blood glucose concentration > 200 mg/dl. Nondiabetic rats were required to have a blood glucose concentration of > 60 but < 120 mg/dl before any intervention. All study animals were allowed at least 20 min for stabilization after the completion of the surgical preparation.
According to the presence of diabetes mellitus-induced hyperglycemia and the type of fluid infusion, rats were assigned to three groups: (1) diabetic rats treated with saline placebo (D; N = 18), (2) diabetic rats treated with insulin (ID; N = 18), and (3) nondiabetic rats treated with saline placebo (ND; N = 18). The sequence of study was randomized.
Studies of baseline metabolism were performed in 6 D rats, 6 ID rats, and 6 ND rats that received fluid infusions, but were not exposed to an ischemic insult. After obtaining control measurements, ND and D rats received a continuous infusion of 0.9% saline solution at a rate of 2 ml/h. The ID rats received the same infusion of saline to which regular porcine/bovine insulin (Eli Lilly, Indianapolis, IN) had been added to produce a final insulin delivery of 0.75 units/h. In ID rats, the goal was to reduce blood glucose to values similar to those in ND, during a 90-150-min period. Attempts were made to achieve similar durations of fluid infusions in all groups. Once the desired study conditions were achieved, the stereotactic device was rotated so that the rats were prone. Thereafter, the brains were rapidly frozen in situ by removing the insulation covering the exposed calvarium, and filling the funnel with liquid nitrogen. [13,27]Concomitant with brain freezing, another arterial blood sample was obtained for analysis. Liquid nitrogen was used to bathe the brains during removal and transportation to temporary storage in a -76 degrees C freezer. Later, the brains were removed to a -25 degrees C environment, the venous sinuses and meninges were dissected away, the hemispheres were separated from each other, and the cortex was dissected from the remainder of the cerebrum. Cortical tissue samples were subsequently subjected to chemical analyses.
Studies of intraischemic and postischemic metabolism were performed in 36 rats. These rats were treated similarly to the initial 18 D, ID, and ND rats, except that they were subjected to forebrain ischemia of 10 min duration. In all rats, fluid infusions (including the insulin infusion) were discontinued at the onset of ischemia. Ischemia was induced using a modification of the technique described previously by Smith et al. Briefly, rats received 1.0 mg/100 g body weight trimethaphan intravenously, and, concomitantly, blood flow through both carotid arteries was interrupted with removable surgical aneurysm clips. Blood was withdrawn or infused via the jugular venous catheter into a 10-ml heparinized syringe to maintain mean arterial blood pressure at 45 +/-5 mmHg. Duration of ischemia was calculated from the onset of the isoelectric electroencephalogram. At the completion of 10 min ischemia, and without reperfusion, the brains of 6 D, 6 ID, and 6 ND rats were frozen in situ, and harvested for subsequent brain analysis. In the remaining 18 rats, cerebral circulation was restored after 10 min of ischemia. This was accomplished by reinfusing the previously withdrawn blood and removing the carotid occlusion clamps. Additionally, rats were treated with sodium bicarbonate 0.3 mEq on reperfusion. Rats whose mean arterial blood pressure was < 90 mmHg at 2 min postischemia were also given 4 micro gram phenylephrine intravenously as needed to maintain mean arterial blood pressure > 90 mmHg. After exactly 15 min of reperfusion, physiologic variables were measured, and blood and brain tissues were harvested for chemical analysis, as in previous groups.
Brain concentrations of glucose, lactate, and high-energy phosphates (adenosine triphosphate [ATP], adenosine diphosphate, adenosine monophosphate, and phosphocreatine) were measured using previously described enzymatic fluorometric techniques. [13,14,21,29,30]The energy charge of the adenylate pool was calculated using the formula: energy charge = [ATP] + 0.5 [adenosine diphosphate]/[ATP] + [adenosine diphosphate ] + [adenosine monophosphate]. Brain glycogen content was estimated from a 0.1 -g aliquot of tissue using a modification of methods described by Passonneau and Lauderdale and Swanson et al. Brain tissue was homogenized in 0.03 N HCl and heated for 10 min at 100 degrees C. A portion of the homogenate was subjected to glucose analysis, and the remaining portion was incubated with amyloglucosidase (Sigma Chemical Co., St. Louis, MO), which hydrolyzed the glycogen into glycosyl units. Glycogen content, expressed as glycosyl units, [21,33]was estimated as the difference between the glucose level of hydrolyzed and nonhydrolyzed homogenates. This method was validated in our laboratory by analyzing samples that contained known quantities of oyster glycogen (Sigma). .
From the above data, several calculations were made. Total brain carbohydrate (TBC) was defined by the formula:
TBC = brain glucose concentration + brain glycogen concentration.
Additionally, in the 18 rats subjected to ischemia without reperfusion, intraischemic changes in glucose, glycogen, TBC, and lactate concentrations were calculated using the formula:
Intraischemic change = individual metabolite value at the completion of 10 min of ischemia - mean preischemic baseline value.
Using this convention, a negative value denoted metabolite consumption (e.g., glucose, glycogen, TBC), and a positive value denoted production (e.g., lactate).
Data within treatment groups (i.e., D, ID, or ND) were compared to the preischemic subgroups, and data after each intervention (e.g., fluid infusion; fluid infusion plus ischemia) were compared among treatment groups (D, ID, and ND), using a two-way analysis of variance that considered only the main effects of group and intervention. A one-way analysis of variance was used to compare intraischemic metabolite production or consumption data. After all analysis of variance calculations, post hoc comparisons were made with F-tests. A probability of < 0.05 was considered statistically significant. All data are reported as mean+/-SD.
Groups tended to be well matched for control physiologic data both before fluid infusions (untabulated data) and at the time of brain harvesting (Table 1). Notable exceptions are as follows: At the time of brain freezing, mean arterial blood pressure in the nonischemic groups was always within protocol criteria (Table 1), although significantly less in diabetic D and ID rats than ND rats. This perhaps reflected a degree of diabetes-induced dehydration. Further, at all study intervals, there were small deviations from the target PaO2and PaCO2values that were observed only at the time of brain freezing (i.e., they were not observed during the preischemic control period [untabulated data] or during spot checks of blood gases before brain harvesting). These deviations were likely due to turning the rats from supine to prone, to facilitate in situ brain freezing. Thus, pulmonary mechanics and hemodynamics may have been altered shortly before brain harvesting and obtaining final blood samples. New-onset deviations in mean arterial blood pressure in rats receiving 10 min of ischemia without reperfusion did, however, achieve statistical significance among groups, although all remained within the protocol range.
The study weight of diabetic rats (262+/-41; mean +/-SD) was less than that of nondiabetic rats (330+/-34 g; P < 0.001 by unpaired t tests). Diabetic rats lost 40+/-26 g weight in the week after streptozocin injection, despite ample access to rat chow and water. Similar weight loss in diabetic rats has been observed in previous studies from our laboratory. [14,21].
The duration of fluid infusion did not significantly differ among the various study groups for a given study interval. Infusion duration was 159+/-58 min for the nonischemic groups, 114 +/-43 min for the 10 min ischemia groups, and 108+/-34 min for the ischemia plus 15 min reperfusion groups.
Plasma Glucose Concentrations
At the time of brain harvesting, plasma glucose concentrations in D were 2.6- to 4.0-fold greater than that in ID and ND; however, plasma glucose concentrations did not differ between ID and ND (Table 1).
Brain Glucose and Glycogen
Before ischemia, brain glucose concentrations in D were 2.4-3.8-fold greater than in either ID or ND (Table 2; P < 0.05). At the completion of ischemia, brain glucose concentrations were reduced in all groups; however, residual brain glucose concentrations in D were greater than in either ID or ND (Figure 1and Table 2and Table 3; P < 0.05). Once circulation was restored, brain glucose concentrations increased to 88%, 106%, and 213% of baseline in D, ID, and ND, respectively (Figure 1and Table 2).
Before ischemia, brain glycogen did not differ among groups. At the completion of ischemia, brain glycogen concentrations were reduced in all groups (Table 3; P < 0.05). Once circulation was restored, glycogen concentrations--unlike glucose concentrations--continued to decline in all groups (Table 2).
In Table 3, the patterns of intraischemic TBC consumption are given. During ischemia, TBC consumption was D > ID > ND (P < 0.05 for D vs. ND). In D rats, only 27% of consumed TBC originated as glycogen. In contrast, in ID and ND rats, 52% and 63%, respectively, of consumed TBC originated as glycogen.
Before ischemia, brain lactate concentrations were similar among groups (Table 2). As expected, postischemic lactate concentrations were greater in D (i.e., the group with the greatest intraischemic carbohydrate consumption) than in either ID or ND rats (Table 2and Table 3). With reperfusion, lactate concentrations decreased in all groups to approximately one-third to one-half the peak values (Table 2).
Brain High-energy Phosphates
Brain concentrations of high-energy phosphates and energy charges are presented in Table 2and Figure 2. Preischemic data were similar among groups. Although high-energy phosphates were severely depleted during ischemia in all groups, ATP concentrations at the completion of ischemia were 2.6-4.2-fold greater in D than in ID or ND, and the energy charge of the adenylate pool was 1.7-2.1-fold greater (P < 0.05).
After 15 min of recirculation, phosphocreatine and ATP concentrations had almost returned to baseline values (Table 2). Although the sum of the adenyl nucleotides was slightly reduced from baseline, energy charge had returned to normal (i.e., 0.91-0.92) in all groups.
Although it is well established that diabetics have a worse outcome after cerebral ischemia than nondiabetics, [2,3]and that worse outcome is related, in part, to hyperglycemia, [2,3]the effects of insulin treatment on this process are largely unknown. Palumbo et al. retrospectively discovered that, in humans with adult-on-set diabetes mellitus, good or excellent diabetes control was associated with fewer transient ischemic attacks and strokes than poor diabetes control. However, the authors did not state whether control resulted from more stringent insulin therapy, noninsulin therapies, or varying severity of disease in the two study populations. In a recent multicenter study of stringent versus traditional glucose control in a population of 1,441 humans, with a mean duration of 6.5 yr of observation, there was insufficient statistical power to reach conclusions on differences in outcome after stroke. In contrast, using a rat model of drug-induced diabetes mellitus (streptozocin-treatment 5-7 days before study) and forebrain ischemia similar to that in the current study, Warner et al. demonstrated that, compared to nondiabetic rats, there was a worse neurologic outcome in hyperglycemic diabetics. However, when acutely treated with insulin, diabetic rats had an outcome similar to that of normoglycemic nondiabetic rats. The metabolic basis of this benefit was not elucidated.
Based on our study results, several conclusions can be reached regarding energy metabolism in the rat model of forebrain ischemia: (1) acute insulin treatment in previously hyperglycemic diabetics restores brain metabolism to a pattern similar to that in normoglycemic nondiabetics, and (2) a worse outcome in hyperglycemic diabetics does not correlate with a greater depletion of high-energy phosphates, nor can an improved outcome with insulin therapy be attributed to better preservation of high-energy phosphates.
In all groups, TBC consumption during ischemia, and production of lactate, were proportional to preischemic TBC concentrations. Furthermore, during ischemia, the mobilization of glycogen was an important source of carbohydrate, yet the relative contribution of glycogen differed among groups. In D rats, glycogen contributed 27% of the total glucose and glycogen consumption, as opposed to approximately twice this fraction in ID and ND rats (Table 3). The observation of similar carbohydrate metabolism in ID and ND rats suggests that the pattern and extent of intraischemic TBC consumption were determined largely by brain glucose concentration. Thus, by acutely reducing plasma and brain glucose concentrations with insulin, it was possible to shift the brain from a pattern of diabetic metabolism to nondiabetic metabolism. Taken together, these data suggest that hyperglycemia of 1-week duration in group D did not fundamentally affect the kinetics and metabolism of glucose.
Our studies also quantified lactate production during ischemia. In models of complete or near-complete cerebral ischemia (i.e., models in which lactate concentrations should be a marker of brain pH [see discussion later]), it has been estimated that brain lactate concentrations of 16-20 micro mol/g are required to produce morphologic injury. [2,21]This conclusion is consistent with observations in the current model. Hyperglycemic diabetics are reported to experience severe functional and histologic injury with this model ; we observed brain lactate concentrations of 29 micro mol/g (i.e., values in excess of the toxic threshold). In contrast, normoglycemic nondiabetic rats and insulin-treated diabetic rats, 10 min of ischemia is reported to produce mild to moderate functional and histologic injury ; we observed brain lactate concentrations of 15-18 micro mol/g at the completion of ischemia (i.e., values barely within the threshold range).
The intraischemic accumulation of lactic acid is theorized to poison the cell by interfering with a variety of metabolic processes including protein synthesis and, in some studies, energy production. [17,19,21,22]The latter issue is both controversial and complex because, as glucose and glycogen are hydrolyzed to produce lactate, equimolar quantities of lactate and ATP are produced. [22,36]However, as the ATP is consumed, the process yields hydrogen ions. Thus, the hydrogen ions associated with lactic acidosis do not originate from glucose glycolysis to lactate, per se, but instead may result from the hydrolysis of ATP. [22,36]Because most ATP produced by anaerobic glycolysis (as well as a portion of the basal ATP concentration) is hydrolyzed (Table 2), it follows that there will be a comparable intraischemic accumulation of lactate and hydrogen ions. In this setting, at any instant, the relationship among energy reserves, lactate, and hydrogen ions will depend on both: (1) the production of energy, and (2) the consumption of energy. This balance may be influenced by a potentially toxic effect of lactic acid on the cell, perhaps by initiating futile, [18,23]energy-consuming metabolic pathways.
As reviewed recently, the effect of increases in brain glucose on periischemic brain high-energy phosphate concentrations has been evaluated in several studies in nondiabetic animals with conflicting results. Studies in diabetic subjects are limited. Despite these limitations, results from the current study are consistent with those of our previous study in rats exposed to complete cerebral ischemia. In both, when exposed to an ischemic insult that is clinically survivable in normoglycemic subjects, intraischemic high-energy phosphate concentrations were greater in hyperglycemic subjects than normoglycemic subjects. This appears true regardless of whether hyperglycemia results from diabetes mellitus or glucose infusion, or whether normoglycemia is associated with nondiabetic subjects or acute insulin treatment of diabetes (Figure 2and Table 2). At the completion of ischemia, high-energy phosphate concentrations were restored to near-baseline concentrations, and, thus, probably were of no importance in further modulating ischemic injury.
Another concern is that, as reported by Pelligrino et al., the acute, insulin-induced restoration of normoglycemia in diabetic rats may induce a state of cerebral hypermetabolism. If a generalized metabolic activation was induced in our study, an increased lactate production and worsened energy status might be expected; however, no such pattern was observed. At the completion of ischemia, lactate and high-energy phosphate concentrations were similar in ID and ND (Figure 2and Table 2). Thus, the insulin effect observed by Pelligrino et al. in nonischemic nitrous oxide-sedated rats apparently did not influence intraischemic and postischemic metabolism in our pentobarbital-anesthetized rats.
When compared to nondiabetic rats, it is not intuitively obvious why diabetes mellitus is associated with a greater concentration of intraischemic cerebral high-energy phosphate concentrations (Table 2), yet worse--postischemic injury. [2,3,35]Further, it is not obvious why acute insulin treatment in diabetic subjects, which reduces the intraischemic high-energy phosphate concentrations to those of nondiabetic subjects (Table 2), also lessens postischemic injury. .
Before extrapolating from the present data obtained in rats to the diabetic state in humans, we urge the reader to consider the following. Unlike the rats we studied, humans typically: (A) have longer periods of diabetes mellitus before the ischemic insult, (B) have coexisting diseases (e.g., hypertension, vasculopathies) that may influence outcome, and (C) are not anesthetized at the time of ischemia. [1-4,10,11,34]It was beyond the scope of the current study to determine the relative influence of these factors in modulating postischemic outcome. Thus, further study is warranted.
With the above limitations acknowledged, the current data can be summarized as follows: Nondiabetic rats and rats with drug-induced diabetes mellitus demonstrated that diabetes affected brain glucose concentrations, but not brain glycogen concentrations. Despite a greater intraischemic accumulation of lactate in hyperglycemic diabetic rats, reflecting a greater lactic acidosis, concentrations of high-energy brain metabolites were better preserved in hyperglycemic diabetics than in normoglycemic, nondiabetic rats. Acute insulin therapy in chronically hyperglycemic rats converted brain metabolism from a diabetic to a nondiabetic pattern. After 15 min of recirculation, brain concentrations of adenosine nucleotides and phosphocreatine were either normal or near normal in all study groups. Taken together, these results suggest that, when exposed to an ischemic insult that is consistent with postischemic survival with residual brain injury, the enhanced cerebral injury associated with lactic acidosis in diabetics is probably mediated, in large part, by factors other than energy failure.
The authors thank Richard Koenig and Rebecca Wilson for technical support; and Michael E. Johnson, MD, Ph.D., for critiquing the manuscript.