All procedures were approved by the UCSD Institutional Animal Care and Use Committee. Laboratory bred Syrian hamsters, Mesocricetus auratus (of original Harlan stock; HsdHan: AURA, Harlan, Indianapolis, IN), and Siberian hamsters, Phodopus sungorus (from a colony maintained at UCSD), were housed at 22 ± 2°C with ad libitum access to water and Purina chow (St. Louis, MO). From birth, Syrian and Siberian hamsters were maintained in 14 h light, 10 h dark daily (i.e., LD14:10; lights on 0300 PST) with photophase illumination of 100 – 150 lux and with no scotophase illumination. Beginning with and continuing throughout the experiments, green light-emitting diodes (LEDs) generated a mean scotophase illumination of 0.027 ± 0.007 lux on the floor of the cage. Photophase illumination remained 100 – 150 lux.
Cage changes, which are potent circadian zeitgebers that facilitate rhythm splitting, always occurred during the first 60 min of the daytime scotophase at intervals of 1–2 weeks. These cage changes defined intervals for statistical analysis, which excluded any data collected 24 h after the perturbation.
At 12 weeks of age, sterile radio-telemeters (Mini-Mitter, Bend, OR) were implanted in the abdominal cavity of Syrian hamsters (16 male/2 female) under Nembutal anesthesia, and hamsters were transferred to individual cylindrical polypropylene cages (20.5 cm diam).
Photoperiod and wheel manipulations
At 1000 PST on the tenth day post-surgery, animals were transferred to rectangular polypropylene cages (27 × 20 × 15 cm) equipped with running wheels (17 cm diam) with plastic interleaved through the rungs of these wheels to increase wheel running coordination. Transfer coincided with the onset of the first scotophase of a new LDLD7:5:7:5 cycle (lights off 1000, lights on 1500, lights off 2200, lights on 0300 PST). Animals were divided between two light-tight ventilated chambers with space for 12 cages each.
After 2 weeks in LDLD7:5:7:5, running wheels were permanently immobilized at 0700 PST with plastic zip ties that bound rungs to the top of the cage. After 2 weeks without wheel access in LDLD7:5:7:5, the photoperiod was changed to LDLD9:3:9:3 (lights off 1100, lights on 1400, lights off 2300, lights on 0200 PST) at the beginning of the evening photophase (1500 PST). After four weeks, hamsters were exposed to constant dim illumination (0.027 ± 0.0067 lux) for two additional weeks beginning with the daytime scotophase.
At either 21–24 or 41–44 weeks of age, group housed male Siberian hamsters (n = 22) were transferred at 1000 PST to individual cages coinciding with the beginning of a 5 h scotophase of a new LDLD7:5:7:5. Half of the animals, with equal representation of each age cohort, were housed in cylindrical cages equipped with plastic-wrapped running wheels (17 cm diam). The remaining hamsters were housed in rectangular polycarbonate cages (27 × 20 × 15 cm) without a wheel, but equipped instead with a passive infrared motion detector (PIR; Coral Plus, Visonic, Bloomfield, CT) to detect locomotor activity.
After two weeks the LDLD cycle was changed to LDLD8:4:8:4 by reducing the duration of both scotophases by 30 min at each end. At weekly intervals each scotophase cycle was successively shortened by one hour (to LDLD9:3:9:3, LDLD10:2:10:2 and LDLD11:1:11:1). After one week of the latter photocycle, hamsters were maintained for an additional four weeks in LDLD9:3:9:3. Subsequently, each 9 h photophase was replaced with two 1-h "skeleton pulses" to yield LDLDLDLD1:7:1:3:1:7:1:3 for two weeks.
In both experiments, activity and temperature data were collected with Dataquest III hardware (Mini-mitter, Bend, OR) configured for 6 min bins. Prior to implantation, telemeters were calibrated at 36 and 38.5°C. Wheel-running revolutions triggered mechanical sensors that recorded a single count every half rotation. PIR motion detectors registered activity whenever 3 of 27 zones were crossed. Rhythms were plotted and analyzed in ClockLab (Actimetrics, Evanston, IL) with supplementary analyses with Microsoft Excel. Statistical analyses were performed with Statview 5.0 (SAS Institute, Cary, NC).
Incidence of splitting
Different criteria for splitting were required for the different rhythms measured. For wheel running, rhythms were classified as split in any given interval if there were > 15 wheel revolutions for 3 consecutive 6 min bins in both daily scotophases for at least 3 days. As in past studies [18, 23], there was never any ambiguity about the split versus unsplit status of wheel-running records.
For PIR rhythms and for telemetered GL and Tb data, values were smoothed over 30 min bins and the dataset reduced to 48 values per 24 h period (i.e., every 30 min). For classification as split or unsplit during each analysis interval, a 24 h histogram was produced for each hamster by averaging 5–6 days of this reduced dataset. The rhythm was considered split if each histogram scotophase was associated with elevated activity or Tb levels defined as follows: values exceeding the daily mean by more than one standard deviation for 2 consecutive 30 min bins. In nearly all cases, these objective determinations corresponded with subjective judgments of the rhythms as essentially unimodal (unsplit) or bimodal (split).
Analyses specific to Experiment #1
Tbmean and amplitude
In each analysis interval, mean Tb values were determined by averaging all values over 5–6 days prior to a cage change. Tb amplitude was calculated as the difference between the maximum and minimum values of 5–6 day histogram calculated from 30 min averages as described above. Repeated measures ANOVA was performed for the 12 animals with uninterrupted telemeter function.
Activity-independent component of Tbrhythm
For each animal with complete data in the second 6-day analysis interval of LDLD9:3:9:3 (n = 8 split, n = 4 unsplit hamsters), 30 min Tb values were regressed against 30 min GL. Residual values representing the component of Tb rhythm not accounted for by GL values were retained and averaged at each 30 min time point over 6 days. Alternative models to account for GL yielded very similar results on sample animals. These models included regression of Tb against log GL values using 30 min values; and use of various lag times and integration intervals for the independent variable .
Analyses specific to Experiment #2
Phase angle of entrainment
In each photoperiod, the phase angle of entrainment for PIR data in Experiment #2 was determined from 5–6 day averages of the 30 min smoothed data. For each component of the split activity rhythm, or for the single component of the unsplit rhythm, activity onset was defined as the earliest point near the L/D transition that exceeded the daily mean. Phase angle was expressed in relation to this transition (positive values indicated activity anticipates lights off).
For each scotophase, activity duration (α) was taken as the time difference between the first and last points exceeding the daily activity mean. Both dependent measures (phase angle and activity duration) were regressed against the length of each scotophase for split and unsplit hamsters separately. Because the number and identity of split hamsters changed across the experiment, the dataset was not a proper repeated measures nor were data points fully independent. Recognizing that treating data points as independent would have the effect of reducing statistical power, we opted to do so in linear regression analysis.