Subjects
White-crowned Sparrows (Zonotrichia leucophrys gambelii) were captured in California (Sutter and Colusa counties) between March 2004 and March 2005. For the assessment of sleep a total of 13 birds were recorded, 12 of which were ultimately analyzed for sleep/wake scoring (SLPs). One was analyzed in three seasons (Summer, Fall and Winter), two were analyzed for two seasons (Spring and Summer) and the remaining 9 were analyzed for one season each. 15 DRL birds (DRLs) were assessed for 4 seasons after birds had learned the task, defined as stable performance (4 consecutive sessions with response rate for each session varying by no more than 10% of the mean response rate across all 4 of those sessions), before entering a migratory season. Therefore, each bird was studied for 2 non-migratory seasons and 2 migratory seasons. Migratory birds do not display activity patterns consistent with frank migratory activity for the whole of the spring or fall. However, for the purposes of this study, unless stated otherwise, birds selected from the spring or fall were always in the active migratory state and displaying activity patterns which typify this state, including increased nocturnal activity. Similarly, birds in the summer and winter are always in the non-migratory state. All birds were captured using mist nets under authorization granted by the California Department of Fish and Game and the United States Fish and Wildlife Service.
Sparrows were transported to the University of Wisconsin - Madison where they were individually housed in galvanized wire cages (L: 35 cm × W: 25 cm × H: 32 cm) in environmentally controlled rooms (L: 4.0 m × W: 2.7 m × H: 2.7 m; 22.0 - 24.5°C, 40% relative humidity). Each bird was in visual and auditory contact with other birds in the room. Their daily diet consisted of a seed mixture (Finch Mix, Mounds Pet Food Warehouse, Middleton, WI), grit, romaine lettuce, and one mealworm. The SLPs had access to food and water ad libitum. DRLs had their food restricted for 3 hours per day (see Behavioral Procedures), except on weekends, and access to water was not available during 30-minute test sessions. All procedures were approved by the Institutional Animal Care and Use Committee at the University of Wisconsin - Madison and followed NIH guidelines. The study was conducted in an AAALAC-accredited facility.
Photoperiods
The photoperiods in housing and testing rooms approximated seasonal and geographic changes in day length appropriate for these birds based on their typical location. Dawn and dusk times were changed each Friday evening to reflect the day length at the expected location of Z.l. gambelii on that day of the year. Winter and summer photoperiods were based on US Naval Observatory (http://aa.usno.navy.mil./data) sunrise/sunset tables for Sacramento, CA and Fairbanks, AK, respectively. Spring and fall photoperiods were based on a linear approximation of a direct route between Sacramento and Fairbanks and observed arrival and departure dates [12]. Illuminance during the dark phase was < 0.5 lux. Illuminance during the light phase was 540 - 640 lux measured at the level of the cage floor. Photoperiods for days selected for sleep scoring were as follows: spring 14.75:9.25 LD, summer 21.5:2.5 LD, fall 13.75:10.25 to 16.0:8.0 LD and winter 9.5:14.5 LD. Bird behavior was more variable during the fall, so that recordings for birds in this group were obtained on different days, with correspondingly different lengths of simulated daylight. Operant data were analyzed from birds under the same photoperiodic conditions as above, except in fall when photoperiods ranged from 11.5:12.5LD to 16.0:8LD.
Activity monitoring
Selection of days for sleep scoring from each season was partly based on data from activity monitoring. To measure daily activity, an infrared (IR) photocell sensor (Invisible-Beam Entry Alert, Radio Shack, Fort Worth, TX) was centered behind each bird's home cage, 18 cm from the floor and 17.5 cm from either side of the cage. The sensor was positioned to project the photobeam parallel to and above two perches so that hops and flights from one side to the other disrupted the beam. Beam breaks were tallied every 30 seconds (i.e., interval recording, with 2880 intervals per day) using VitalView (version 4.0) software and transmitter equipment (Mini Mitter Co., Inc., Bend, OR). Although the infrared activity monitoring system may be prone to some inaccuracy because it is insensitive to activity unless the bird crosses the center-line of the cage and overly sensitive to activity when the bird remains near the center of the cage and on the perches, it nevertheless offers a rapid method for assessing gross seasonal changes in behavior. SLPs behavior was also continuously recorded using two infrared-sensitive cameras per bird connected to a digital video storage system (IView PC, Salient Systems Corp., http://www.salientsys.com). Infrared illuminators provided lighting for the cameras during the dark phase. Cameras were positioned on opposite sides of each cage, to afford maximum potential for visual monitoring of behavior. Cameras (commonly available for recording and transmitting over the Internet) were used inside the test chambers to view DRLs during sessions.
Surgery
Only SLPs received surgery. Surgical procedures were performed under isoflurane anesthesia (1.0%-3.5% isoflurane with 500 ml/min O2). The bird's head was stabilized in a stereotaxic device (Kopf Instruments, Tujunga, California, United States), cranial feathers were removed and an incision was made along the midline of the head to expose the cranium. Six small holes were drilled through the cranium to the dura: Two holes were drilled on each hemisphere of the anterior forebrain 2 mm lateral to the midline over the hyperpallium (Wulst). Two additional holes were drilled 2 mm posterior to the anterior holes so that signal was recorded from both the anterior and posterior hyperpallium. Two holes were positioned over the midline of the cerebellum to accommodate the common reference electrode and the ground. Teflon insulated stainless steel electrodes (#791400, A-M Systems Sequim, Washington, United States) were inserted through the holes to the level of the dura and held in place with surgical adhesive (Tissuemend II, Veterinary Products Laboratories, Phoenix, Arizona, United States).
Each electrode was connected to a lightweight, flexible, and electrically shielded recording cable designed for use with small birds [13] (Dragonfly, Inc., Ridgeley, West Virginia, United States). This cable was attached to the skull using dental acrylic (Justi Products, Oxnard, California, United States), and the incision was closed around the acrylic with surgical adhesive (Tissuemend II, Veterinary Products Laboratories, Phoenix, Arizona, United States). Following surgery, each bird was placed in the recording cage and provided with at least 14 days of postoperative recovery and adaptation to the recording cable before experimental observation began.
Electrophysiological recording
The recording cable was attached to a low torque 6-channel mercury commutator and the weight of the recording cable was counterbalanced with a spring; these recording conditions allowed birds to move unimpeded throughout the cage. The EEG signals were referenced to the cerebellar electrode, amplified and band pass filtered (0.3 - 30 Hz and 10 - 90 Hz, respectively) using Grass-telefactor amplifiers (Model 12 Neurodata and 7P511, http://www.grass-telefactor.com) and digitized at 100 Hz (National Instruments PCI 6071E card, http://www.ni.com and Somnologica 2, Flaga hf. Medical Devices, http://www.medcare.com). EEG signals were viewed using Somnologica 3 software (Flaga hf. Medical Devices, http://www.medcare.com).
Sleep-wakefulness scoring
Actograms (e.g., see Figure 1) generated from the IR activity data demonstrated that daytime activity was moderate to high in all seasons. Migratory-specific behavior was best characterized by nighttime IR beam-breaking activity. Blocks consisting of several consecutive days of nocturnal activity were selected for each bird. Video footage was reviewed to ascertain that video quality on the day selected for scoring was adequate. To assess extremes of sleep-wakefulness behavior during non-migratory seasons, days for scoring were chosen during the longest (summer) and shortest (winter) photoperiods. Days were also selected during spring and fall migratory periods when nocturnal activity appeared maximal. We eliminated days for which video recordings were not consistent with IR-measured activity, e.g., a bird slept perched near the middle of the cage and generated many IR beam-breaks with small involuntary movements. The EEG and video recordings were reviewed concurrently to ascertain EEG quality, ensuring that the EEG was not obscured by motion artifact (outside of active wakefulness) and that the quality of the tracing was sufficient to distinguish among vigilance states (wakefulness, drowsiness, SWS, REM sleep). For each of the seasons of winter, spring and summer, we were able to find at least a single date that met these criteria for all the birds being recorded at those times. Fall migratory behavior was more variable within and between birds, both in timing and in day-to-day consistency, than spring migratory behavior, as previously reported [14]; this necessitated the use of data from different days for different birds.
During each season, 4 or 5 birds produced 24-hour records (activity, video, and EEG) that fit our selection criteria, resulting in a total of 16 scored records across four seasons. Unfortunately, data for individual birds could not be collected across all seasons, as EEG signals typically degraded within 3 - 9 months, and, consequently, 10 of the 13 birds were recorded during one season only; 2 birds were recorded for two seasons and 1 was recorded for three seasons. One fall bird originally selected for sleep scoring was ultimately not used in our final data analysis; although IR data indicated that this bird was in the migratory state on the date selected to score his sleep, he was not, on closer inspection of the video, actively engaged in nocturnal migratory activity, and no other migratory days for this bird had EEG data of usable quality. As a result, only 3 fall birds were used in the final data analysis.
Vigilance state was manually scored in 4 s epochs using simultaneous EEG and video recordings. Each epoch was categorized as either wakefulness, drowsiness, SWS or REM sleep based on visual inspection of the EEG from both hemispheres, as well as by analysis of recorded behavior using the standard criteria: Wakefulness was characterized by a high-frequency low-amplitude EEG in both hemispheres. Behavior during wakefulness included hopping and flying around the cage, feeding, drinking, feather preening, and actively scanning the room. During drowsiness, EEG activity was intermediate between that of wakefulness and SWS (i.e., increased amplitude in the low-frequency range relative to wakefulness). Behavioral evidence of drowsiness included birds holding their heads close to their bodies and the position of the eyelids fluctuating between open, partially closed, and completely closed states. During SWS, EEG activity was dominated by slow waves of the highest amplitude. Behaviorally birds were motionless, with closed eyes; the head was either pulled in toward the body and facing forward or resting on the bird's back. REM sleep epochs in birds tend to be brief [15]. In this study, a 4-sec epoch was scored as REM sleep if the EEG amplitude was reduced by at least one-half the amplitude seen in the previous SWS episode and this amplitude reduction lasted for longer than 2 seconds and was accompanied by behavioral signs of REM sleep including muscle hypotonia (feather or head drooping) or, rarely, eye movements. Bouts of sleep were defined as the length of time from a SWS onset to the next bout of wakefulness (W), and waking bout length was likewise defined as the length of time from the onset of W to an epoch of SWS. We also calculated the ratio of sleep bout length to waking bout length as a measure of relative sleep stability for each hour. This measures whether sleep is maintained for longer periods of time than wakefulness. For each 24-hour period, the entire period was scored in 4 s epochs, resulting in 21,600 epochs per day.
Operant Apparatus
Programs for operant behavior testing and data collection were written using MED-PC IV (Med-Associates, St. Albans, VT). Session events were recorded with 10-millisecond resolution. Behavioral testing was conducted in operant chambers (Med-Associates, Model ENV-007). Each chamber contained a translucent pigeon key capable of emitting white light and a food hopper that, when activated, emitted light and provided 5 seconds of access to nyger seed. The key was situated in the center of the panel with the food hopper located to the right of the key. Each chamber was surrounded by a sound-attenuating cubical with a built-in ventilating fan that circulated air into the experimental environment and provided masking noise. A fluorescent light with an 8-watt bulb, which remained on throughout the experiment, was located between the chamber and cubicle to provide additional lighting. Internet cameras were also mounted inside the chamber cubicles.
Operant testing
Food was removed from home cages for 3 hours preceding test sessions. This was done Monday through Friday, although sessions were only conducted Tuesday through Friday every week. Following training, operant testing was conducted for 1 year. Sessions were held once per day and each lasted 30 minutes. Each session began with the darkened key turning white. A differential-reinforcement-of-low-rate (DRL) schedule was in effect throughout the entire session [11]. On the DRL schedule used here, a response was reinforced when it occurred more than 20 seconds after the previous response. Qualifying key pecks immediately opened the reinforcement hopper for 5 seconds; response rates per minute were determined by subtracting out this 5 second reinforcement interval. Each key peck that occurred after 20 seconds or more since the last key peck (DRL 20-sec) was reinforced with food. Qualifying key pecks turned the key dark and simultaneously opened the reinforcement hopper for 5 seconds, followed by the key light turning white again; responses made prior to the required 20 second pause had no effect on the key light but did reset the DRL clock to zero to begin incrementing again. Responses during times when the key was dark had no programmed consequences. The DRL schedule is alternatively referred as reinforcement of long interresponse times (IRTs)[11].
Video was monitored to verify that birds were actually consuming the reinforcer when obtained. Training for the procedure began in the fall for 5 birds, in the spring for 2 birds, and in the summer for 8 birds. For all birds, key pecking behavior came under reliable control of the DRL schedule, as defined above, within 6 months.
Sleep Deprivation
During each season, a 48-hour sleep deprivation was conducted with all DRLs using the following procedure: Members of the research staff entered the housing room once every 3 - 5 minutes or sooner if behavioral signs of sleep, such as eye-closings and inactivity, were prominent (viewed remotely with video cameras). Walking quietly past the cages and occasionally tapping on them provided sufficient stimulation to keep the birds awake; handling was not necessary to induce wakefulness. The sleep deprivation began at lights out on Monday, and continued until lights out on Wednesday evening. DRL sessions were run from Tuesday-Friday each week. Therefore, the first two sessions for the week (Tuesday and Wednesday) occurred under conditions of sleep deprivation; the sessions conducted on Thursday and Friday were considered recovery days. The week prior to the sleep deprivation (Pre Sleep Deprivation) was used as a baseline, and the week subsequent to the sleep deprivation and recovery was used as a return to baseline (Post Recovery). For spring, summer and winter, sleep deprivation was conducted 1 - 2 weeks after the weeks used in the assessment of typical DRL performance for the season. However, during fall, the sleep deprivation probe occurred 1 - 4 weeks after the weeks selected to represent seasonal performance for 10 birds; given; the heterogeneity of the fall migratory behavior, birds used for the sleep deprivation and subsequent DRL testing (post sleep deprivation recovery) in fall were not in the migratory state. The photoperiods (and weeks) during the four sleep deprivation periods were 10.25:13.75 LD for winter (second week of February), 17.25:6.75 LD for spring (second week of May), 20.25:3.75 LD for summer (third week of July), and 13.25:10.75 for fall (fourth week of September). Pre Sleep Deprivation and Post Recovery data were averaged across all 4 sessions of those weeks; Sleep Deprivation and Recovery were averaged across the first 2 sessions and last 2 sessions, respectively, of that week.
Statistics
Based on selection criteria (see Activity monitoring and Sleep-wakefulness scoring), seasons for SLPs were defined as follows: winter 9.5:14.5 LD (first week of January), spring 14.75:9.25 LD (third week of April), summer 21.5:2.5 LD (first week of July), and fall ranging from 13.75:10.25 to 16.0:8.0 LD (fifth week of August to second week of October). These weeks included those with the longest or shortest days (winter and summer), and those with the most nocturnal activity in the home cage (spring and fall). The main effects and interactions of season, lights and migration status on vigilance state were assessed by repeated measures (split plot) ANOVA. All pairwise comparisons were assessed with a familywise α = 0.05 via Tukey HSD following a significant independent ANOVA.
For DRLs, data from these same photoperiods were used to characterize performance with exception of fall; in this case, the week during which nocturnal home-cage activity (see Activity monitoring) best represented migratory restlessness was selected for each bird. Fall weeks for DRLs fell within the range for SLPs specified above.
For analyses of operant behavior, the primary dependent measures were (1) response rate, defined as the ratio of the number of responses to the number of minutes the key light was on, (2) number of food reinforcers obtained, and (3) the behavioral inhibition ratio (sometimes referred to as efficiency), defined as the ratio of the number of reinforcers to the number of responses. For the behavioral inhibition measure, a ratio of 0 represents complete disinhibition; the closer the ratio is to 1.0, the more pronounced the inhibition. On occasion, birds did not respond during the session; for these sessions, the behavioral inhibition ratio, but not response rate or reinforcers obtained, was dropped because the resultant ratio of 1.0 was not representative of performance. Data were assessed across four consecutive seasons for each bird, starting with the first season of reliable responding (see Operant testing). DRL data were averaged across sessions for each season, producing a single value for each photoperiod before it changed.
For analyses of DRL performance by migratory status, data from the migratory seasons were averaged and compared to non-migratory season averages. T-tests were used in these analyses with the Type I error rate set at 0.05; multiple comparisons were not made following these omnibus tests. For analyses of seasonal performance, a repeated measures ANOVA was performed across all four seasons." Behavioral changes within sessions were also assessed by dividing each session into ten 3-minute bins and recording the number of responses and reinforcers during each bin. Analyses of trends were conducted following repeated measures ANOVAs across the ten bins to determine linearity. For each season of sleep deprivation, a repeated measures ANOVA was performed across conditions of Pre Sleep Deprivation (Baseline), Sleep Deprivation, Recovery, and Post Recovery, followed by post-hoc t-test comparisons with the error rate corrected via the FDR procedure.