To study their thermal responses to climatic stress, we implanted seven greater kudu (Tragelaphus strepsiceros) with intra-abdominal, brain, carotid, and subcutaneous temperature data loggers, as well as an activity logger. Each animal was also equipped with a collar holding a miniature black globe thermometer, which we used to assess thermoregulatory behavior. The kudu ranged freely within succulent thicket vegetation of the Eastern Cape Province, South Africa. The kudu spontaneously developed a bacterial pneumonia and consequent fever that lasted between 6 and 10 days. The fever was characterized by a significant increase in mean 24-h abdominal temperature from 38.9 ± 0.2°C to 40.2 ± 0.4°C (means ± SD, t6 = 11.01, P < 0.0001), although the amplitude of body temperature rhythm remained unchanged (t6 = 1.18, P = 0.28). Six of the kudu chose warmer microclimates during the fever than when afebrile (P < 0.0001). Despite the selection of a warmer environment, on the first day of fever, the abdominal-subcutaneous temperature difference was significantly higher than on afebrile days (t5 = 3.06, P = 0.028), indicating vasoconstriction. Some kudu displayed increased frequency of selective brain cooling during the fever, which would have inhibited evaporative heat loss and increased febrile body temperatures, without increasing the metabolic maintenance costs of high body temperatures. Average daily activity during the fever decreased to 60% of afebrile activity (t6 = 3.46, P = 0.014). We therefore have recorded quantitative evidence for autonomic and behavioral fever, as well as sickness behavior, in the form of decreased activity, in a free-living ungulate species.
- selective brain cooling
animals employ a suite of autonomic and behavioral mechanisms to fight an infection. These mechanisms are regulated by proinflammatory cytokines, which act centrally to inhibit the firing rate of warm-sensitive neurons in the hypothalamus, resulting in a rise in the thermoregulatory set point (1, 3, 26). In addition to an increased body temperature during infection, animals consistently exhibit symptoms such as lethargy, reduced engagement in social activities, and reduced food intake (1, 7), collectively referred to as sickness behavior. Sickness behavior originally was thought to be the result of a weakened physiological state but now is viewed as an adaptive reorganization of the host's priorities to facilitate recovery from an infection (10, 18, 23, 31). Indeed, given their ubiquity among both endothermic and ectothermic vertebrates, both fever and sickness behavior are postulated to offer a survival benefit (25, 27, 28, 56).
Although both fever and sickness behavior have been studied extensively in laboratory-housed animals given purified pyrogens experimentally, little is known about how free-living animals respond to opportunistic infection. Unlike naturally occurring fevers, which last several days at least, most experimentally induced fevers last less than 24 h and are not superimposed on the nychthemeral rhythm of body temperature (26, 44). However, in free-living impala, prolonged natural fevers did not disrupt the nychthemeral rhythm of body temperature (24). But neither these antelope nor laboratory-housed goats with chronic fevers were observed to display any obvious signs of sickness behavior (24, 44). It is, therefore, unknown whether sickness behavior is a component of chronic fever in free-living animals.
Laboratory-housed animals, although they may be physically active, often do not have access to the full range of natural behaviors such as thermoregulatory behaviors, social interactions, and reproduction. Changes in behavior in laboratory-housed animals during fever, therefore, may not be similar to behavioral responses of a free-living animal, since wild animals would have to sacrifice parts of their usual behavioral repertoire to employ sickness behavior. For example, both free-living, male, northwestern song sparrows and white-crowned sparrows decreased their territorial aggressive behavior and song when exposed to an exogenous pathogen (48, 49). Such behavioral changes may be costly for wild animals, particularly if they result in an increased susceptibility to predators, decreased reproduction, and parental care, loss of social position, or removal from territories (2, 18, 31, 49, 57). Sickness behavior and immune responses to infection also are metabolically costly (18, 30, 40). Such metabolic costs may be maladaptive, particularly during times of low energy availability. Weight loss following exogenous pyrogen administration in captive male white-crowned sparrows was greater when they had high energy reserves than when they had lower reserves (48). Free-living animals therefore have to trade off any potential benefit of sickness behavior, in supporting metabolic and physiological changes associated with infection, against associated energetic and social costs.
It is difficult to study fever and sickness behavior in wild animals as sickness occurs opportunistically and observations of behavior usually require human presence, which, in turn, may disrupt normal behavior. Serendipitously for us, a herd of free-living antelope (kudu, Tragelaphus strepsiceros), which we had instrumented with temperature and activity data loggers for other purposes, acquired a spontaneous infection, which lasted between 6 and 10 days, 2 to 5 wk after surgery. We obtained remote and continuous measurements of temperatures at various body sites, including the brain, carotid artery, abdomen, and beneath the skin, both before and during the infection. We were able later to extract quantitative data reflecting the kudu's autonomic and behavioral responses to the infection. We show that the free-living kudu exhibited sickness behavior and implemented a suite of coordinated thermoregulatory responses that elevated body temperature during a spontaneous fever.
MATERIALS AND METHODS
Animals and Habitat
The experiment took place between December 2004 and March 2005 (austral summer) at Blaauwkrantz farm (33°32'S 25°23′E), near Port Elizabeth, South Africa. The vegetation in the area has been classified as Sundays Spekboomveld (55), and the habitat is both historical and current habitat for greater kudu (T. strepsiceros) (51).
Seven adult female free-living kudu (body mass 121 ± 8 kg) were captured from a large camp on Blaauwkrantz farm in November 2004. The kudu were immobilized from a helicopter, by a professional capture team under veterinary supervision and were given a long-acting tranquilizer, zuclopenthixol acetate (100 mg im clopixol acuphase; H. Lundbeck, Randburg, South Africa), before being transported to nearby holding pens at River Bend Lodge (∼20-min drive). The kudu were housed in these pens for a 2-wk habituation period to reduce potential perioperative stress.
All experimental procedures were approved by the Animal Ethics Screening Committee of the University of the Witwatersrand (protocol no. 2004/84/4).
For surgery to implant the data loggers, each animal was immobilized with etorphine hydrochloride (6 mg im, M99, Novartis, Johannesburg, South Africa), together with azaperone (100 mg im, Stresnil, Kyron Laboratories, Johannesburg, South Africa) and transported to a veterinary surgery within 200 m of the pens. At the surgery, the animals were placed in sternal recumbency, supported by sandbags, with their heads elevated. Anesthesia was maintained with 1–3% halothane (Fluothane, Zeneca, Johannesburg, South Africa), administered in oxygen via a face mask. Once an adequate plane of anesthesia was established, the effect of the opioid, etorphine, was reversed using diprenorphine hydrochloride (15 mg iv, M5050, Novartis). Respiratory rate, heart rate, arterial oxygen saturation, and rectal temperature were monitored throughout the surgery, lasting ∼2 h.
Under sterile surgical conditions, a suite of miniature data loggers was implanted. All data loggers were covered in an inert wax (Sasol, Johannesburg, South Africa) and dry-sterilized in formaldehyde vapor before implantation. Incision sites were shaved and sterilized with chlorhexidine gluconate (Hibitane; Zeneca, Johannesburg, South Africa). Each animal was fitted with data loggers connected to thermistor sensors for temperature measurement in the carotid artery and brain. A thermistor, inserted in a blind-ended and thin-walled polytetrafluoroethylene (PTFE) tube (OD 1.35 mm, ID 0.97 mm; Straight Aortic Flush 4F Catheter, Cordis, The Netherlands), was advanced 80 mm into the left common carotid artery toward the heart, at a position midway along the length of the neck, and secured in position with a purse-string suture in the artery wall. Outside the artery, the PTFE tube was sealed on a PTFE-coated coaxial cable (150 mm long, OD 3 mm, Belden; Richmond, VA) connecting the thermistor to the temperature-sensitive data logger (see Temperature Measurements). The data logger was positioned subcutaneously, dorsal to the artery. A second data logger, connected to the thermistor positioned in the brain, was positioned subcutaneously caudal to the base of the left ear. Its PTFE-coated cable was advanced subcutaneously over the skull, where it was connected to a head plate and guide tube. The guide tube, constructed from cellulose acetate butyrate tubing (OD 3.2 mm, ID 1.6 mm; World Precision Instruments, Sarasota, FL) sealed at the tip by a steel cap, was 58 mm long and was inserted through a small 2-mm-diameter hole, which was drilled through the cranium, at appropriate coordinates, predetermined from head sections of dead kudu of similar size, so that the probe tip would be positioned near the hypothalamus. The brain guide tube was connected to a subcutaneously implanted plastic head plate (20 × 10 × 5 mm) that was secured to the skull by two bone screws.
A third data logger, with an internal temperature sensor, was inserted, via an incision in the paralumbar fossa, into the abdominal cavity. The muscle layer was sutured closed and a smaller temperature-sensitive data logger (see Temperature Measurements) was inserted subcutaneously before the skin was sutured closed. An additional incision was made on the upper hindlimb, where an activity logger (Actical, Mini-Mitter, Bend, OR) was implanted subcutaneously. The activity logger recorded at 10-min intervals, had dimensions of 40 × 40 × 15 mm, and weighed ∼40 g.
Wounds were treated with a topical antiseptic spray (Necrospray, Centaur Labs, Johannesburg, South Africa) and coated with a tick repellent grease (cypermethrin 0.025% m/m, Bayer Animal Health Pty, Isando, South Africa). Each of the kudu received a long-acting antibiotic (22 ml im, penicillin, Peni La Phenix, Novartis, Virbac Animal Health), a nonsteroidal anti-inflammatory analgesic (5 ml im, phenylbutazone, Phenylarthrite injectable solution, Bayer Animal Health), a long-acting parasiticide (2.5 ml sc, doramectin, Dectomax, Pfizer Laboratories, Sandton, South Africa), Vitamin E and Selenium (5 ml im,vitESe injectable solution, Kyron Laboratories, Johannesburg, South Africa), and perphenazine (100 mg im, Kyron Laboratories).
Before halothane administration was terminated, a neck collar (African Wildlife Tracking, Pretoria, South Africa) was fitted to each kudu. In addition to a tracking radio transmitter, each collar supported a miniature black globe thermometer (“miniglobe”) to allow for the dynamic measurement of the microclimate that the kudu chose to occupy. This technique previously has proven successful on other ungulate species (20). Miniglobe temperature was measured with a thermistor inserted into the center of a matt-black hollow bronze sphere (30-mm diameter, Press Spinning & Stamping, Cape Town, South Africa). The thermistor leads were housed within a spring that was soldered to the miniglobe and filled with flowable silicon and covered in heat-shrink tubing to add flexibility and strength. The base of the spring was attached to the collar, so that the miniglobe stood 20 mm above the collar. The miniglobe temperatures were recorded by a temperature-sensitive data logger (see Temperature Measurements), which was attached to the collar and waterproofed with dental acrylic. A weight on the ventral side of the collar ensured that the miniglobe remained over the dorsum of the neck and could not be shaded by the animal's body.
Following surgery, the kudu were transported back to their pens, where they became ambulatory within ∼10 min. After a three-day recovery period, and following veterinary inspection, they were transported back to Blaauwkrantz farm, where they were released into a 50-ha fenced enclosure with natural forage and water available ad libitum. Two to five weeks after release, the kudu spontaneously developed a lethal pneumonia, and a consequent fever that lasted between 6 and 10 days before the kudu died. Unfortunately, our experimental design, which required that the animals be disturbed by humans as little as possible, and the data logging technology did not allow us to determine that the kudu were febrile until after their death. Where possible we conducted gross-macroscopic and histopathological postmortem examinations. These revealed a severe necrotizing bronchopneumonia and empyema, with extensive neutrophil infiltrates and bacterial colonies. While it is known that wild animals do not inhabit sterile environments (14), the deaths were unexpected; we have performed similar procedures on a variety of mammalian species without ill effects. However, because there are no reports of recovery from surgery in wild kudu, we do not know whether the species is particularly susceptible to delayed postoperative infection or whether they were the victims of an opportunistic infection acquired at the study site.
The miniature thermometric data loggers (StowAway XTI, Onset Computer, Pocasset, MA) used to measure brain, carotid artery, and abdominal temperature, had outside dimensions of ∼50 × 45 × 20 mm and a mass of ∼40 g when covered in wax. These loggers had a resolution of 0.04°C and measurement range from +34 to +46°C. Temperature sensors used to measure brain and carotid blood temperatures were constructed from ruggedized glass-coated bead thermistors with insulated extension leads (bead diameter 0.3 mm; AB0E3-BR11KA103N, Thermometrics, Edison, NJ). The scan interval of the brain and carotid blood loggers was set at 5 min and that of the abdominal logger was set at 20 min. Subcutaneous temperatures were recorded every hour using a smaller thermometric data logger (iButton DS1922T, Maxim, Dallas Semiconductor, Dallas, TX), which had a radius of ∼25 mm, a height of 15 mm, and a weight of ∼10 g when covered with wax. These loggers had a resolution of 0.5°C and a measurement range from 0 to 125°C.
The temperature of each miniglobe was measured with an uncoated bead thermistor (bead diameter 1.5 mm; 27–10K4A801 Onset Computer). The thermistor was connected to a miniature thermometric data logger (Hobo U12–013, Onset Computer), with a temperature range of −20 to +70°C and an intrinsic resolution of 0.35°C. Miniglobe temperatures were recorded instantaneously every 15 min.
All temperature sensors and loggers were calibrated against a high-accuracy thermometer (Quat 100, Heraeus, Hanau, Germany) in an insulated water bath. After calibration, the loggers and their sensors measured blood, brain, and abdominal temperature to an accuracy of better than 0.05°C, subcutaneous temperature to better than 0.5°C and miniglobe temperature to better than 0.4°C.
Meteorological Data Measurements
We collected climatic data from a portable weather station in the enclosure into which the kudu were released (Hobo Weather Station, Onset Computer). We monitored wind speed (m/s), solar radiation (W/m), dry-bulb temperature (°C), relative humidity (%), and standard (150-mm diameter) black globe temperature (°C) for the duration of the study period.
We considered the kudu to be febrile if their mean daily body temperature was at least 0.5°C above normal. Paired Student's t-tests were used to confirm differences in body temperature between febrile and afebrile states. We obtained a complete set of brain temperature and carotid blood temperature data for four febrile kudu for analysis of selective brain cooling. The relationship between brain temperature and carotid blood temperature in each animal was analyzed by sorting all 5-min measurements of carotid blood temperature into 0.1°C classes, and determining the mean, standard deviation, maximum, and minimum brain temperature at each class of carotid blood temperature.
After converting miniglobe temperature to their equivalent standard black globe temperatures (20), we correlated these converted animal miniglobe temperatures against weather station standard black globe temperatures using linear Pearson procedures. To assess whether the animals were conforming to ambient conditions or were selecting microenvironments, we tested whether the slope of the regression equation was significantly different from one (the slope of the line of identity). In addition, we tested whether the slope and elevation of the regression equation were significantly different between febrile and afebrile states using an analysis of covariance (ANCOVA). To compare changes in activity levels during the fever, we calculated the febrile activity levels as a proportion of afebrile activity levels and compared these to a value of one using a paired Student's t-test. Values are expressed as means ± SD.
Over the 6-wk study period during which the kudu were both febrile and afebrile, average dry-bulb temperature, standard black globe temperature, wind speed, and solar radiation varied as a function of time of day (Fig. 1). Average dry-bulb temperature over 24 h was 22.5 ± 2.4°C with a mean daily minimum of 16.8 ± 2.2°C and mean daily maximum of 30.4 ± 5.4°C. Standard black globe temperature was 25.0 ± 3.1°C on average, reaching a mean daily minimum of 16.7 ± 2.3°C and a mean daily maximum of 36.5 ± 6.0°C (Fig. 1A). Solar radiation showed a bell-shaped distribution, with a mean peak of ∼750 W/m2 around solar noon, with variability resulting from a combination of cloudy and clear days. Wind speed increased in the late afternoon, reaching a peak of about 4 m/s around 1600 (Fig. 1B). Rainfall averaged 24 mm/mo in 2004, which was slightly lower than the 50-year rainfall average of 26 mm/mo. However, prevailing weather conditions were not unusual during the study period.
All seven kudu developed a spontaneous fever 2 to 5 wk after surgery. These fevers lasted between 6 and 10 days and were characterized by an upward displacement of the nychthemeral rhythm of body temperature (Fig. 2). The mean 24-h abdominal temperature, for the seven animals, increased significantly from 38.9 ± 0.2°C when the kudu were afebrile to 40.2 ± 0.4°C (t6 = 11.01, P < 0.0001) when they were febrile (Fig. 3). Minimum daily abdominal temperatures increased from 38.0 ± 0.4°C to 39.2 ± 0.5°C (t6 = 5.97, P = 0.001) and maximum daily abdominal temperatures from 39.7 ± 0.3°C to 41.2 ± 0.4°C (t6 = 19.02, P < 0.0001) during the fever. However, the amplitude of nychthemeral rhythm remained unchanged at 1.7 ± 0.5°C when the animals were afebrile and 1.9 ± 0.6°C during fever (t6 = 1.18, P = 0.28).
During febrile days, the upward shift in the nychthemeral rhythm of body temperature appeared to be associated with a change in the thermal environment that the kudu chose to occupy. Figure 4 shows the correlation between the standard black globe temperatures at the sites chosen by a single animal (kudu 5), converted from miniglobe temperatures recorded on the collar, and standard black globe temperatures recorded at a nearby weather station, during both febrile and afebrile states in kudu 5. The slopes for all of the animals were significantly less than one (ANCOVA, P < 0.0001), and the regression lines intersected the line of identity, implying that kudu-selected microclimates cooler than the prevailing environmental conditions at higher environmental heat loads and microclimates warmer than the prevailing environment conditions at lower environmental heat loads. Six of the seven kudu showed no difference in the slope of the regression lines between febrile and afebrile states (ANCOVA, P > 0.17). The remaining animal (kudu 3) showed a significantly increased slope (F1,812 = 4.8, P = 0.03) under the febrile state (0.41 ± 0.04) compared with the afebrile state (0.32 ± 0.01), implying that this kudu selected warmer microclimates when febrile compared with afebrile, particularly at high environmental heat loads. One animal (kudu 7) showed no change in the slope or elevation of the regression line between the febrile and afebrile state. Five of the six kudu, which showed no difference in the slope during afebrile and febrile states, showed an average elevation in points during febrile states, with the y-intercept being significantly (ANCOVA, P < 0.0001) higher in the febrile state (24.0 ± 3.5°C, n = 5) compared with the afebrile state (22.8 ± 4.1°C, n = 5). These results imply that these five kudu were selecting warmer microclimates during their fevers, irrespective of prevailing ambient temperatures.
Despite the selection of warmer environments in the febrile state, there was a larger temperature difference between the abdominal cavity and subcutaneous tissue, indicating a more vasoconstricted periphery, at all times of day (t5 = 2.74, P = 0.04), compared with the afebrile state. The largest temperature differences between the abdominal cavity and subcutaneous tissue occurred on the first day of the fever. The abdominal-subcutaneous temperature differences were significantly higher (t5 = 3.06, P = 0.03) on the first day of fever (1.8 ± 0.9°C) than during the afebrile state (1.3 ± 0.4°C). The average 24-h differences between abdominal and subcutaneous temperatures for the first day of fever, and for an average afebrile day, for all seven kudu, are represented in Figure 5.
In addition to peripheral vasoconstriction and selecting warmer microclimates when they were febrile, all of the kudu displayed sickness behavior in the form of decreased activity. Figure 6 illustrates the average 24-h activity pattern, as detected by movement of the upper hind limb, for a single kudu (kudu 2). Although all the kudu maintained a biphasic activity pattern with crepuscular peaks throughout both febrile and afebrile states, they showed a decrease in activity at all times of day when febrile. On average, daily activity of the seven kudu decreased by 40% during febrile compared with afebrile states (t6 = 3.46, P = 0.01).
Selective Brain Cooling
Selective brain cooling is defined as a brain temperature lower than arterial blood temperature (22). Figure 7 shows the typical pattern of brain and carotid blood temperature recorded for a single kudu (kudu 3) over 4 days before fever was detected and for 4 days of fever. This kudu showed the largest magnitude of selective brain cooling at the highest carotid blood temperatures, which were recorded at the acrophase of the endogenous nychthemeral body temperature rhythm during fever (Fig. 7).
We obtained complete records of brain and carotid blood temperatures for four animals. Figure 8 shows mean, SD, minimum and maximum hypothalamic temperature for each 0.1°C class of carotid blood temperature during both febrile and afebrile states, for those four kudu. The dotted line represents the line of identity; points below this line reflect selective brain cooling. Three of these four kudu displayed selective brain cooling during both the afebrile and the febrile states, and one animal never displayed selective brain cooling (Fig. 8, kudu 1). The threshold for selective brain cooling, defined as the carotid blood temperature at which carotid blood and mean hypothalamic temperatures are equal, was not statistically different (t2 = 1.30, P = 0.32) during afebrile (38.8 ± 0.12°C) and febrile (39.3 ± 0.68°C) states. However, kudu increased the frequency with which they employed selective brain cooling during the fever (t2 = 4.6, P = 0.04), spending 84.4 ± 7.9% of their time with the hypothalamus cooler than carotid blood during the febrile state compared with only 60.5 ± 11.3% during the afebrile state. This higher frequency of selective brain cooling resulted in significantly higher mean selective brain cooling (t4 = 3.2, P = 0.03), as defined by the difference between carotid blood and hypothalamic temperature, during febrile (0.38 ± 0.14°C) compared with afebrile (0.10 ± 0.07°C) states.
Although selective brain cooling was used more often during fever, the mean brain temperature was higher for a given blood temperature when febrile than when afebrile. The slope of selective brain cooling, defined as the difference between carotid blood and hypothalamic temperature as a function of carotid blood temperature (36), was significantly reduced in febrile compared with afebrile states for each of the individuals (P < 0.0001). In addition, the mean slope of lines of regression of brain temperature on carotid blood temperature, above the threshold for selective brain cooling, was lower in afebrile (0.49 ± 0.05) than in febrile (0.63 ± 0.07, t4 = 2.7, P = 0.05) states, indicating that the magnitude of selective brain cooling at any given carotid temperature was reduced during the febrile compared with the afebrile state.
Our results provide the first demonstration, in free-living animals, of certain measurable physiological processes that bring about the elevation of body temperature that is characteristic of fever. We also present the first quantitative evidence for sickness behavior in free-living animals. We could detect these responses because we had instrumented free-living kudu with data loggers capable of producing a more complete continuous record of the thermal status of undisturbed free-living animals than has been achieved previously. The data loggers, which we implanted in the abdomen, revealed that our kudu had an average 1.3°C increase in mean abdominal temperature and reached maximum temperatures ranging between 40.8 and 41.6°C while febrile.
The elevated body temperature was achieved partially through appropriate thermoregulatory behavior, which is less metabolically costly than is implementing autonomic mechanisms. We were able to detect such behavior, without observers present, by using miniature globe thermometers attached to collars, a novel technique we developed (20) for recording microclimate selection by animals. Irrespective of climatic conditions, or time of day, the kudu selected warmer microclimates when they were febrile (Fig. 4) than when they were afebrile. There was a parallel shift in the linear regression lines of the relationship between the selected microclimate and the prevailing ambient conditions, consistent with a change in set point for behavioral thermoregulation. Our technique of measuring selected microclimates also revealed that the kudu used thermoregulatory behavior, in the form of choice of microclimate, when they were both febrile and afebrile; the slope of those regression lines was less than one, showing that the kudu selected appropriate microclimates to reduce the impact of environmental thermal stress to less than one-third of what it would have been without employing thermoregulatory behavior, over the standard black globe temperature range of 10 to 50°C.
If behavioral processes alone are insufficient to increase body temperature to the elevated set point during fever, autonomic processes are invoked. Peripheral vasoconstriction provides the least metabolically costly autonomic response in a cool environment. Using the difference between abdominal and subcutaneous temperatures as an index of peripheral blood flow, we provide evidence that the kudu implemented peripheral vasoconstriction during fever. Irrespective of the time of day, the difference between abdominal and subcutaneous temperature was higher when the kudu were febrile, particularly on the first day of the fever (Fig. 5), than when they were afebrile. The kudu, when febrile, therefore appeared to be more vasoconstricted at all times of day, even though standard black globe temperature reached 35°C, on average, around solar noon (Fig. 1A).
An additional autonomic response that may be invoked to supplement the febrile rise in body temperature is the inhibition of evaporative heat loss (52). Although we had no means of measuring evaporative heat loss continuously in free-living animals, we do have indirect evidence that our kudu, indeed, did inhibit evaporation when febrile. At least three of the animals implemented selective brain cooling more frequently and displayed a greater mean selective brain cooling, when they were febrile (Fig. 8). By lowering brain temperature below carotid blood temperature, selective brain cooling reduces the stimulus from hypothalamic warm receptors for evaporative heat loss. We (13, 42) and others (34, 35, 50) believe that the main thermoregulatory function served by selective brain cooling is inhibition of evaporative heat loss. Even though the magnitude of selective brain cooling at any given carotid blood temperature decreased during fever, kudu spent more time using selective brain cooling. These results indicate that the control of selective brain cooling was altered during fever, in a manner that would suppress evaporative water loss and hence evaporative heat loss.
In response to the increase in set point that occurs at the start of fever, animals increase body temperature by inhibiting evaporative heat loss, instituting peripheral vasoconstriction and selecting warmer microclimates. Once that set point is reached, body temperature is regulated actively around the new set point, and the nychthemeral rhythm of body temperature is preserved during the “plateau” phase of fever (Figs. 2 and 3). In addition to an increased body temperature, our free-living kudu also displayed a decreased activity, conspicuous sickness behavior. When they were febrile, our kudu reduced activity by 40% compared with that when afebrile. Because we recorded activity continuously, by implanted data loggers, we also could show that, just as they preserved the nychthemeral rhythm of body temperature when febrile, the kudu preserved the nychthemeral rhythm of activity, and, specifically, its crepuscular peaks and midday trough (Fig. 6), as if there were a downward shift in the set point of a “regulator of activity”.
Because the presence of human observers distorts the thermoregulation and behavior of animals (42), human presence in the vicinity of the animals was kept to a minimum. Consequently, we do not know what the kudu did to reduce activity when febrile, nor how they set about selecting microclimates; however, kudu have been reported to display clear habitat preference for patches with dense cover (11, 19). In addition, because it was not the original intention of our study to investigate fever patterns, we cannot verify the nature of the pathogen or the cause of death. We also do not have records of other sickness behaviors, for example, feeding, nor did we have any way of monitoring heat production and heat loss, in these free-living animals. We assume that the increased difference between abdominal and subcutaneous tissue temperatures, during fever, reflected peripheral vasoconstriction. But a decrease in environmental temperature would lead to the same result without a change in skin blood flow. However, given that the kudu selected warmer microenvironments during fever, the former explanation is more likely.
Our data provide the first evidence for the proposition that free-living animals elevate body temperature during fever by using the same mechanisms as do human patients and laboratory animals, namely, a coordinated suite of behavioral and autonomic responses. We also have obtained measurements of body temperature from other febrile free-living antelope. The shape and magnitude of the body temperature increases during fever in the kudu are similar to those we saw in free-living impala (24) and springbok (12). The maintenance of the nychthemeral rhythm of body temperature also has been reported during prolonged fevers in goats (44) and humans (39, 46). However, because body temperature during prolonged fever in small laboratory mammals typically is raised much more during the inactive phase than the active phase of their nychthemeral activity patterns, the nychthemeral rhythm of body temperature tends to be reduced or obliterated in small species (16, 41, 43, 47, 53).
Whether or not the body temperature rhythms are maintained, in environments warm enough for evaporative heat loss to be stimulated, body temperature could increase as a result of vasoconstriction or inhibition of evaporative heat loss rather than because of an increase in metabolic heat production. It is likely that evaporative heat loss would be inhibited during the genesis of fever simply because the set point for its initiation would be raised. However, we propose that, in animals capable of selective brain cooling, the implementation of selective brain cooling during fever further inhibits evaporative heat loss. If that were the case, one would expect selective brain cooling to be implemented at the initiation of the fever, which appeared to be the case in our kudu (Fig. 7). However, in oxen (6) and goats (33), selective brain cooling was most conspicuous during the defervescence phase of fever. If sheep are prevented, by upper respiratory bypass, from implementing selective brain cooling during fever, so that brain temperature rises, the elevation in rectal temperature is markedly attenuated, as would be expected if the high brain temperature stimulated evaporative heat loss (37). In species that do not have a carotid rete, and in which there is no selective brain cooling during fever (54), artificial cooling of the brain during fever leads to an exaggerated increase in body temperature (3, 21), without increasing the metabolic cost of fever.
While we could not measure metabolic rate in our kudu, fever is metabolically costly and has been estimated to increase the metabolic heat production of humans by ∼13% with each degree centigrade rise in body temperature (18, 26). This metabolic cost is species specific and dependent on habitat, severity, and duration of infection, nutritional and metabolic status, and the ambient temperature (5, 8, 9, 18). At low ambient temperatures, the increase in body temperature is achieved mainly through heat production, such as shivering; however, at higher ambient temperatures, heat is stored through inhibition of heat loss (for example, cutaneous vasoconstriction and the inhibition of evaporative heat loss), with little change in heat production (5, 17, 21, 52). The selection of warm microclimates by our kudu not only facilitated an increase in body temperature directly but may also have resulted in a preference for the employment of heat conservation mechanisms, such as increased peripheral vasoconstriction and the inhibition of evaporative heat loss.
In addition to the autonomic changes that we have outlined and the selection of warmer microclimates, an additional behavioral mechanism proposed to reduce the metabolic maintenance costs of fever is decreased activity. In theory, reduced activity acts as an energy conservation mechanism, by reducing wasteful wandering activity and convective heat loss associated with moving (18, 31, 56). Decreased activity associated with fever has been reported in various laboratory studies in both rats (15, 41, 45, 53) and mice (32). Conversely, a decreased activity was not observed during sustained experimental fevers in penned goats (44), but unlike rodents, which can be highly active in cages, penned goats have low activity, perhaps too low for a further decrease to be detected during fever. Our study is the first to demonstrate reduced activity in free-living febrile animals. However, inactivity in free-living animals is not without risk, as it could make them more susceptible to predation and reduce their foraging capacity (2, 18, 31, 49, 57).
Perspectives and Significance
As a regulated physiological phenomenon, one would expect both fever and sickness behavior to have survival value (7, 18, 29). Fever has evolved, however, during the so-called host-pathogen evolutionary arms race (10, 38) and could equally benefit the microorganism causing the infection, as it could the host. Banet (4) reported that the survival of rats was directly proportional to the metabolic cost during the rising phase of fever but inversely proportional to the magnitude of fever (4). Our kudu not only appeared to reduce their metabolic costs, through vasoconstriction, warmer microclimate selection, and reduced activity, but also displayed high mean maximum body temperatures of 41.2°C. We do not know whether the fever and sickness behavior displayed by our kudu were beneficial to them; nevertheless, we believe that a proper assessment of the natural role of fever in thermoregulation can be made only by studying fever in free-living animals, just as the roles of other thermoregulatory strategies are revealed only by studies of free-living animals away from human observers (42). By demonstrating, for the first time, coordinated behavioral and autonomic thermoregulatory effectors and sickness behavior in free-living kudu that became infected, we believe that we have made an important first step in that direction.
This study was supported by the National Research Foundation, South Africa, the University of the Witwatersrand Medical Faculty Research Endowment Fund, the Nelson Mandela Metropolitan University and START/ NORAD African Ph.D. fellowship (to R. Hetem).
We thank Arthur and Trinette Rudman for their hospitality and for allowing the study to take place on their farm Blaauwkrantz, Malcolm Rutherford of River Bend Lodge for use of their holding pens and surgical facilities, Dr. Kennedy Erlwanger for assisting with surgical procedures, Andrè Matthee for project assistance and game management expertise, and Sophie and Martin Haupt from African Wildlife Tracking for their help in designing and making the animal collars.
The costs of publication of this article were defrayed in part by the payment of page charges. The article must therefore be hereby marked “advertisement” in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
- Copyright © 2008 the American Physiological Society