INTRODUCTION

Top Abstract Introduction Methods Results Discussion References

A GENERAL CHARACTERISTIC OF the circadian system is its ability to synchronize with environmental cues, such as light-dark alternation or food restriction. It is known that an endogenous oscillator that produces a rhythm with a free-running period () can be synchronized by an environmental oscillation, named synchronizer or zeitgeber, whose period length (t) differs from . During entrainment, the  progressively changes until it equals t and a stable phase relationship is reached between the endogenous oscillator and its zeitgeber (5, 26).

Self-sustained oscillators can be entrained to periods that deviate from their own natural periods but only within certain limits. These limiting frequencies encompass the "range of entrainment" (1, 4, 5). This range varies depending on the animal species (5), age (11), and the strength of the zeitgeber (5). Beyond this entrainment limit, a circadian rhythm free runs with a period close to that measured in constant conditions (1). However, its frequency may be periodically modulated by the signal of the zeitgeber that it crosses ("relative coordination" phenomenon).

The precise determination of the entrainment limits presents some difficulties. Entrainment often becomes unstable near the limits, and, as a result,  fluctuates periodically (34). Furthermore, in a given organism, various overt rhythms can have different ranges of entrainment (1). Finally, some experimental data suggest that after exposure to long or short zeitgebers, the circadian system may lose its capability to produce self-sustained oscillations, resulting in an extremely wide or even "unlimited" range of entrainment (3).

To date, entrainment limits have been determined using three different methodological approaches: 1) exposure of animals to the same zeitgeber with various periods and measurement of the degree of synchronization of the circadian rhythm (29); 2) exposure to light-dark (LD) cycles of progressively increasing or decreasing periods until the circadian rhythm becomes desynchronized (4); and 3) finally, by deduction based on the period of the pacemaker and the phase-response curve (PRC) of the species (for rat see Refs. 15, 16, 31). Thus the range should extend from  minus the maximum phase advance to  plus the maximum phase delay that can be achieved in the PRC.

The albino rat, which is probably the most widely used laboratory animal, has been the subject of many studies devoted to examining the physiological organization of the circadian timing system. There are, however, relatively few data on the range of entrainment of circadian rhythms in rats. Only the limits of locomotor rhythms have been studied in this rodent (29), and the particular entrainment limits of feeding rhythms remain unknown. Although, to our knowledge, there is no direct evidence that different rhythmic variables in rat show different ranges of entrainment, some observations, such as the appearance of a transient internal desynchronization between temperature and locomotor activity rhythms and the presence of a melatonin circadian rhythm together with the loss of circadian rhythmicity of temperature and locomotor activity in animals exposed to constant light (8), make this an open question. Moreover, the study of feeding behavior and the entrainment of feeding rhythms to cycles other than 24 h can be used as a tool for understanding the relative contributions of circadian and homeostatic processes to the rat's daily feeding pattern. As previously reported (19, 33), the growth rate of rats exposed to ahemeral LD cycles can be modified by shortening or lengthening t within a certain range.

The object of our present investigation was twofold: 1) to determine accurately the range of entrainment of feeding rhythms to symmetrical LD cycles using two experimental approaches and 2) to quantify the interaction of circadian and homeostatic processes that regulate the feeding behavior of rats kept under LD cycles ranging from 22 to 28 h.

	METHODS AND EXPERIMENTAL PROCEDURE

Top Abstract Introduction Methods Results Discussion References

Experiment 1. The purpose of the first experiment was to accurately measure the entrainment limits of adult male and female Wistar rats exposed to progressively increasing or decreasing LD cycles.

Eleven Wistar rats of both sexes (6 males and 5 females), bred and raised in our laboratory, were used in this experiment. Until weaning, the rats were maintained under a 12:12-h LD cycle (light = 250-300 lx on the cages). When the rats were 25 days old, they were separated from their mothers and placed individually in the same light-, temperature-, and humidity-controlled room (light = 250-300 lx on the cages, darkness < 0.5 lx of dim red light continuously illuminated; temperature 22 ± 1°C; relative humidity 60 ± 10%).

Light was provided by fluorescent lamps giving symmetrical LD cycles of increasing or decreasing length. The LD cycle was programmed by means of an electronic timer.

A pelleted diet (Letica, Barcelona, Spain) and tap water were available ad libitum. Cages were cleaned and food and water replaced every 4-5 days, always during the light phase.

Feeding activity was recorded by using a contact eatometer described in detail elsewhere (18, 22). Briefly, the eatometer consisted of a transparent polycarbonate rat-keeping cage (Panlab) covered by a stainless steel grid. Mounted on the inside of the eating area was a hinged grid, which enabled the animal to reach the food when it was pushed up. A microswitch fitted to the swinging grid was activated each time the grid was pushed. This information was then fed into an IBM PC-AT computer.

In the course of this experiment, the animals were kept in a sequence of 4 different light conditions: 1) from weaning to 45 days old, LD cycles of 12:12 h (lights on at 0800 and off at 2000); 2) from 45 to 65 days, LD cycles with t decreasing steadily by 10 min per day from LD 12:12 h to LD 10:10 h; 3) from 65 to 74 days, LD 12:12 h; and 4) finally, from 74 to 117 days, LD cycles with t increasing steadily by 10 min daily from LD 12:12 h to LD 15:15 h.

Experiment 2. In this part, we studied food intake and the temporal organization of feeding behavior in rats exposed to static LD cycles covering a wide range of circadian periodicities.

Seventy male Wistar (Charles River) rats were used. Before the experiment, the animals had been bred and kept with foster mothers under LD cycles of 12:12 h. When the animals were 25 days old, they were separated from the foster mothers and put into individual cages with free access to food and water. The animals were then divided randomly into 7 groups (10 rats per group). Each group was placed in an experimental chamber, in which they were submitted to one LD cycle of 22, 23, 24, 25, 26, 27, or 28 h. To avoid the effects of photoperiod and light intensity, the cycles were symmetrical in all cases (light = 300 lx on the cages, darkness < 0.5 lx of continuous dim red light). This protocol ensured that at the end of the experiment all animals had received the same quantity of light. To ensure that the luminance conditions were in accordance with the experimental protocol, a photoelectric sensor was attached to the top of one rat cage in each chamber and its output was continuously recorded. These records enabled us to control luminance exactly and facilitated subsequent analysis of the data. The experiment was ended by leaving the animals in constant dark (DD) conditions.

The experimental chambers were placed in an isolated room with the same intensity of dim red light as during the dark phase. The temperature and humidity were also controlled and kept constant at 23 ± 0.5°C and 40-60%, respectively.

The feeding behavior of six rats per group was recorded as in expt 1. Every 4 days, from 25 to 85 days of age, the food remaining in the cage and the body weight were recorded for the 10 animals of each group. At the same time, water and previously weighed fresh food were provided.

Data processing and analysis. Food approach events were recorded and stored in 10-min bins. For visual inspection, event records were qualitatively double plotted (48-h time scale) at a resolution of 10 min in a standard actogram format. At the end of the experiment the percentage of the total feeding duration (amount of time spent feeding) during the light and dark phases was calculated to determine the degree of synchronization to the t cycles. We chose feeding duration and not meal duration (the total time from the beginning to the end of a particular meal, including periods with no feeding activity within the meal) because the latter parameter depends largely on the meal separation criterion chosen, whereas feeding duration excludes such inconsistencies because it changes very little between individual animals (17). According to our criterion, desynchronization began when the light feeding duration surpassed 60% of the overall feeding duration during two consecutive t cycles. This criterion was selected because our experimental design prevented the use of a periodogram analysis to establish the uniformity of periods between the rhythm and its zeitgeber and because the low level of events in feeding actograms means that a more objective criterion than visual inspection is necessary. To define the onset of desynchronization, we took into account the normal phase angle of feeding activity in response to LD cycles in the range assayed in this experiment. Visual inspection of the desynchronization dynamics confirms the validity of the criterion selected. Before desynchronization, the percentage of feeding duration during light increased steadily as the phase angle between the circadian rhythm and its zeitgeber increase; however, after desynchronization this parameter became unstable and the light feeding activity reached 50-60%.

Feeding behavior was characterized by calculating the averaged patterns of feeding duration and meal frequency. Feeding activity was recorded as previously described (17). Feeding events with no interruption longer than 20 min were defined as a meal. This time criterion is similar to that used by others to discriminate between pauses within and between meals (17, 21, 27).

To assess the degree of synchronization of the animals to the static t cycles, the data were analyzed by double-plotted graphs, chi-square periodogram (28), and calculation of phase differences between the last LD cycle and the onset and end of  during the first 3 days in DD. The length of  was determined, from well-defined average waveforms, as the period during which the feeding activity profile stabilized above a certain level that represented 50% of all activity counts recorded in the t period. To define the onset of , transient increases of activity (with a duration 60 min) above the 50% level that did not reflect anticipation of the onset of darkness were not taken into account. In those t23 animals that showed two concurrent entrained and free-running components,  was determined from the mean waveform calculated on a 23-h basis.

Because the average shape of the waveforms was not sinusoidal, cosinor analysis could not be used and so the amplitude of circadian rhythms was determined by calculating the differences between mean feeding activity during  and . The average waveform of feeding activity was obtained by averaging 10-min data bins measured throughout the 60 days the rats were exposed to the t cycle. Estimates of  were based on these average waveforms as well as on event records.

A multiple ANOVA was used to determine differences in body weight, food intake, and feeding parameters between groups. Individual comparisons were made by means of the Scheffé test.

	RESULTS

Top Abstract Introduction Methods Results Discussion References

Experiment 1. Entrainment limits to progressive changes in t. The objective of the first experiment was to measure the range of entrainment in response to progressive changes in the t cycle.

Four representative records of feeding activity, taken from two male and two female rats, are shown in Fig. 1. During the first 20 days (t = 24 h), all the rats showed 24-h entrained rhythms; however, as t became shorter  became progressively later, while feeding activity ended precisely at lights on. When t was shortened from 22 h 30 min to 22 h, most rats became desynchronized and the circadian rhythms started to free run with periods longer than 24 h.

View larger version (51K): [in this window] [in a new window]   Fig. 1.   Left: double-plotted event records of feeding activity of 4 representative rats, 2 males (top) and 2 females (bottom), exposed from weaning to light-dark (LD) cycles that decreased steadily by 10 min per day. Scheduled darkness is indicated by solid line. Right: percentage of total feeding duration during dark (D) and light (L) calculated per LD cycle.

A more precise determination of the onset of desynchronization was achieved by calculating feeding duration during light and darkness. Table 1 shows the last t cycle to which animals remained synchronized. All the rats were synchronized between t = 22 h 40 min and t = 26 h. No sex-related differences were detected in the animals.

After a period of resynchronization to LD 12:12, the progressive lengthening of t induced a parallel phase advance in feeding activity (Fig. 2). From approximately t = 26 h, these phase shifts became insufficient to correct the difference between t and . The event records of activity for two rats of each sex are shown in Fig. 2. With minor exceptions, the qualitative features of these records are representative of all the rats in the study. As can be seen, the upper desynchronization limit (between 25 h 50 min and 27 h 30 min) to a much greater extent than the lower. Overall, the mean entrainment range of the rats in our experimental conditions was 4 h 35 min ± 30.3 min.

View larger version (51K): [in this window] [in a new window]   Fig. 2.   Left: double-plotted event records of feeding activity for 4 representative rats, 2 males (top) and 2 females (bottom), exposed from weaning to LD cycles that increased steadily by 10 min per day. Scheduled darkness is indicated by solid line. Right: percentage of total feeding duration during dark and light calculated per LD cycle.

Experiment 2. Entrainment limits for a stable t. Because the limits of entrainment depend on experimental conditions, particularly those referring to gradual changes in t or age, we examined the effect of long-term exposure to stable t cycles. After weaning, each animal was exposed to only one t cycle.

Because none of the rats entrained their circadian rhythms to t = 22 h, in our experimental conditions this period is clearly outside the entrainment range of the rat's circadian system (Fig. 3). The t22 animals exhibited a free-running rhythm little related to the LD cycle. In these conditions,  was longer than in DD. As was to be expected in a state of desynchronization, the starting phase of the free-running rhythm under DD was not related to the previous phase of the LD cycle (Fig. 4).

View larger version (94K): [in this window] [in a new window]   Fig. 3.   Feeding activity records of representative animals maintained under zeitgeber period (t) cycles ranging from 22 h (t22) to t28. Two rats belonging to t23 group are shown because they are representative of animals with 2 (R3/26) and 1 (R3/16) circadian component, respectively. From day 85, the animals were kept in constant dark conditions. For each rat, 3 chi-square periodograms were made corresponding to period marked by arrows.

View larger version (19K): [in this window] [in a new window]   Fig. 4.   Individual values of feeding activity phase of circadian rhythm for first day after removal of time cue. Black points represent individual values of the end of  calculated when circadian rhythms started to free run in constant conditions. Black and white bars show dark and light period, respectively, of the last day of t cycle exposure.

In t23 animals, the circadian rhythms essentially remained synchronized to an average period close to 23 h. However, one-half the animals of this group exhibited a complex pattern consisting of a light-synchronized component of 23 h and a weak free-running component of ~24.5 h. It is difficult to separate true light entrainment from the masking effects induced by light in those animals with two components. However, some evidence suggests that, besides masking, a weak entrainment component may be present: 1) the average waveform of t22 rats only showed two reactive responses to lights on and lights off, whereas the waveform of t23 animals with two components exhibited a more pronounced oscillation in their feeding activity, which cannot be not explained by the masking alone (Fig. 7); 2) the free-running component was more robust in t22 rats than in t23 rats with two components (Fig. 5); and 3) in t23 rats a free-running component appeared gradually. The presence in some animals of two components explained why, when the time cue was removed, the rhythm started to free run from a phase not always determined by the environmental cycle (Fig. 4); however, the weakness of the free-running component meant that it was not possible to establish its exact phase at the beginning of constant conditions. Animals with only the light-entrained component started to free run from a phase that was coincident with the phase of the previous day under LD alternation.

View larger version (132K): [in this window] [in a new window]   Fig. 5.   Double-plotted event records of feeding activity at modulo t for 3 representative animals (A, B, and C) maintained under t22, t23, and t28 cycles. Note instability of phase relationship between feeding rhythm and its zeitgeber under t28.

All the animals exposed to t cycles of 24, 25, and 26 h showed a clear and stable synchronization to LD cycles (Fig. 3), fulfilling all the entrainment criteria: period control, stable phase relationship between circadian rhythm and its zeitgeber, and phase control of the free-running rhythm by the previous environmental cycle.

Animals exposed to 27- and 28-h t cycles seemed to be synchronized to their respective cycles because the period of their circadian rhythms coincided with the period of the environmental cycle and when the time cue was removed the rhythm started to free run from the phase to be expected from previous environmental cycle (Figs. 3 and 4). However, the instability of the phase relationship between the feeding rhythm and its zeitgeber, especially under t = 28 h conditions, suggested that entrainment had not been completely achieved (Fig. 5).

Figure 6 shows the mean of the free-running period of the circadian rhythm of feeding activity when the rats were transferred from t cycles to DD. All the animals showed a free-running period of >24 h, although the previous conditions of entrainment continued to influence  for many subsequent cycles. This relationship was approximately linear for groups t24, t25, t26, and t27, although there was no such relationship in extreme situations (t22, t23, and t28). It is interesting to note that the t23 animals with two components in their circadian rhythms showed higher  values than the t23 rats that were completely synchronized to the LD cycle. The t22 animals, which suffered total desynchronization, showed no clear aftereffects, which suggests that such aftereffects are more severe in the more synchronized animals.

View larger version (14K): [in this window] [in a new window]   Fig. 6.   Free-running period of circadian rhythm of feeding activity when rats were transferred from t cycles to constant darkness. Solid line represents mean value of 6 rats per group.

The averaged waveforms of feeding activity (Fig. 7) clearly illustrate the stability of  despite the considerable changes in the waveforms themselves, which range from unimodal in t23 to trimodal in t > 24 h. The positive masking effect of lights off and the negative masking effect of lights on make it difficult to calculate . In clearly desynchronized animals (t22 rats), this parameter was not determined. The / percentage shows a high level of stability, ranging from 50% in t28 and t24 to 46.4% in t25. Only in one group, t23, was this relationship far from one-half of the circadian cycle (39%), probably because of the negative masking effect induced by lights on and the presence in some animals of separate light-synchronized and free-running components.

View larger version (43K): [in this window] [in a new window]   Fig. 7.   Educed waveform of feeding activity averaged over 60 days of exposure to t cycles. Data are presented as percentage of the mean of the overall circadian period, so that this mean equals 100%. This procedure normalizes ranges of circadian rhythms across subjects. Values are means ± SD of 6 animals per group, except for t23, in which the animals that showed 2 components (2C, n = 3) have been separated from animals with 1 component (1C, n = 3).

Within the range of entrainment, the phase angle difference between the rhythm and the entraining zeitgeber changed systematically, with increasingly earlier phases as periods lengthened and lagging phases with shorter periods (Fig. 8). The onset of activity occurred 6.8 h before lights off when t = 28 h and 3 h after lights off when t = 23 h. A similar pattern was seen at the end of , except in t23 animals, in which  was shortened before its natural occurrence by the masking effect of lights on.

View larger version (19K): [in this window] [in a new window]   Fig. 8.   Phase-angle difference (expressed in degree; 1 full zeitgeber period = 360°) between  onset () or  offset () and midpoint of darkness at various zeitgeber periods for the full range of entrainment. Each value represents mean of 6 animals. Shaded area indicates darkness.

The percentage of nocturnal feeding activity changed with t [F(6,33) = 26.6, P < 0.001]. It was highest in t24 animals (82%) and decreased gradually as t became more distant from 24 h (65% for t = 22 h and 58% for t = 28 h; Fig. 9A). Except in the case of t22 rats, in which the rhythms were not entrained, this effect not only seemed to be the consequence of a variation in the phase angle between the activity and its zeitgeber but also of a real decrease in the amplitude of the feeding circadian rhythm. Considering the mean values of feeding activity as 100%, the amplitudes for t22 to t28 were 9.5, 30.5, 54.1, 47.1, 47.7, 35.8, and 31%, respectively.

View larger version (37K): [in this window] [in a new window]   Fig. 9.   Mean nocturnal feeding activity (A), feeding duration (B), meal frequency (C), and food intake (D) under t cycles ranging from 22 to 28 h. Values are means ± SD of 6 animals (A-C) and 10 animals per group (D). All values are expressed per t cycle and per 24 h. Values for a given parameter that do not share a common letter are significantly different (P < 0.05), as determined by ANOVA followed by Scheffé's test.

Meal frequency per t cycle increased progressively with t [F(6,35) = 2.91, P < 0.05], although only t23 and t28 animals showed statistically significant differences between them (Fig. 9C). This relationship exhibited a linear correlation, in which meal frequency = 0.41 × t + 1.47 (P < 0.001, r = 0.54, n = 41). As shown in Fig. 9C, such a relationship was due to the fact that the number of meals remained practically constant over 24 h. Therefore, ultradian periodicity in the occurrence of meals did not change proportionally with the t span. Similar results were obtained with the feeding duration (the total number of 10-min periods during which contact was made with the eatometer; Fig. 9B). The lengthening of t induced proportional increases in this parameter (feeding duration in min/t = 19.7 × t  74.6, P < 0.001, r = 0.53, n = 41). Furthermore, when these results were presented as feeding duration per 24 h, this parameter remained almost constant [F(6,35) = 0.72, P = 0.63; Fig. 9B].

The total food intake calculated per t cycle is shown in Fig. 9D. This parameter is significantly modified by exposure to t cycles [F(6,63) = 22.9, P < 0.001]. However, rats exposed to t cycles ranging from t24 to t28 showed no statistically significant differences between them. The lowest values corresponded to the animals of the t22 and t23 groups that were statistically different from the values of the other groups (P < 0.05 in all comparisons). It is striking that rats of these groups were either nonsynchronized (t22) or partially synchronized (3 rats t23 with 2 components). In contrast, the highest food intake per t cycle corresponded to the t28 animals. Again, although this group met some entrainment criteria, entrainment did not seem to be complete. In clearly synchronized animals (t24, t25, t26, t27), food intake per 24 h statistically decreased as t increased (Fig. 9D), suggesting that in these groups food intake follows a rhythmic rather than an homeostatic control.

	DISCUSSION

Top Abstract Introduction Methods Results Discussion References

It is widely known that Wistar rats perform nearly 80% of their feeding activity during the dark phase of a 12:12-h LD cycle (21). This basic characteristic can be used to objectively determine the entrainment limits of feeding rhythms to ahemeral LD cycles, because in synchronized rats the shift in the feeding activity phase is not great enough to produce an inversion in the percentages of feeding activity during light or darkness. Moreover, according to our results, when desynchronization occurs, the percentage of feeding duration during light and darkness exhibits rapid and unstable changes with many inversions, which reflects the passage of a free-running rhythm through light and dark phases.

Although it is true that entrainment limits can be predicted indirectly from the PRC, this procedure fails to predict accurately the limits of entrainment to LD cycles. Symmetrical t cycles have "parametric" and "nonparametric" effects on the pacemaker, whereas in the PRC only nonparametric effects are studied.

The entrainment limits observed in rats exposed to progressive changes in t were 22 h 9 min and 26 h 44 min. This is a slightly wider range than that described by Stephan (29) for the motor activity of Wistar rats exposed to t cycles with 2 h of light. Differences in the experimental protocol and in the range of t cycles may explain these discrepancies. It is noteworthy that despite the strong influences of sexual hormones on the circadian system of rats (21), in our experiment, no sex-related differences in the entrainment limits were detected.

The exposure of rats to a wide range of stable t cycles allowed us to examine the circadian rhythm of feeding activity related to each t cycle. The most striking finding related to short t cycles was the appearance in t23 rats of two different components: a light-synchronized and a free-running component. Although the light synchronized component could be explained as the expression of a masking effect of the LD cycle, other contributions must be taken into account to explain the appearance of this circadian component. It is difficult to differentiate the entrainment from the masking effects in the t23 rats possessing two circadian components, although there is evidence to suggest that the masking was not sufficient to explain the presence of a light-synchronized component. 1) One characteristic of masking factors, and of light in particular, is that their effects are noticed irrespective of the LD period (7). If the synchronized component were only due to a masking effect, the feeding rhythms of t22 and t23 two-component rats would be masked in a similar way by LD 11:11 h and LD 11.5:11.5 h, respectively. However, in t23 two-component rats the contribution of "masking" is greater than in t22 rats. 2) The free-running rhythm in t22 is more evident than in t23 two-component rats, suggesting that some circadian component remains synchronized in t23 two-component rats.

In previous papers (20, 32) involving rats exposed to t23 cycles, we found that the period of the light-synchronized component changed every time it crossed the free-running component, indicating the true oscillatory nature of the light-synchronized component.

The appearance in t23 rats of two circadian components, which has also been reported in the motor activity of rats (20), can be explained by the general hypothesis that the circadian timing system of complex organisms consists of multiple circadian oscillators that are coordinated by coupling relationships (9). If we assume that the system is composed of a population of oscillators with slightly different frequencies, it is possible to hypothesize that under constant conditions or under t cycles close to 24 h, the coupling between oscillators is strong enough to keep them running together, generating a clear circadian overt rhythm. However, under t cycles far from the , some oscillators may desynchronize while others remain synchronized to the t cycle. When t was equal to 22 h, all the oscillators seemed to be outside their entrainment limits, so that a free-running overt rhythm appeared. However, in t23 rats, the oscillators can be reorganized into two different groups: the first, composed of those oscillators with a higher frequency and that are more easily entrainable, remained synchronized to t23, whereas the second group, composed of oscillators with a lower frequency and that were poorly entrainable, were responsible for the free-running component.

The most important difference between the dynamic and stable paradigms as regards entrainment limits determination concerned the upper boundaries (26 h 44 min vs. at least 28 h, respectively). Exposure to t27 and t28 induced an apparent synchronization in all rats that was manifested as 1) the coincidence observed between the period of the circadian rhythm of feeding activity and the period of t, 2) the fact that when the LD cycle was removed the rhythm began to free run from a phase determined by the environmental cycle, and 3) the phase angle difference between  of feeding activity and its zeitgeber in t27 and t28 animals, which was as expected for a synchronized situation. On the other hand, other criteria suggest the opposite: phase instability, especially in t28 animals, and the strong positive correlation between  and t in constant conditions when t ranged from 24 to 27 h was lacking in t28 rats, where  was smaller than the expected aftereffects.

The strong ability of rats exposed to long t cycles to synchronize could also reflect the plasticity of the circadian system after the early exposure of rats to the t cycle.

The animal circadian system is not considered to function as a homogeneous unit, and the rat's system in particular (9) is thought to be composed of multiple oscillators, each with its own particular characteristics. For this reason, the entrainment limits cannot be defined according to the classic concept of a circadian rhythm, which may or may not be synchronized to its zeitgeber. The change from an entrained system to one that is not would involve a series of intermediate situations in which certain oscillators slowly lose their entrainment and begin to free run, whereas others, which can be more easily entrained either because their endogenous period is more favorable to entrainment or because connections with the zeitgeber are stronger, remain synchronized.

As regards feeding behavior, our results showed that neither the interval between meals nor total feeding activity could be lengthened or shortened in direct proportion with the LD cycle. The mechanisms responsible for producing the ultradian feeding pattern seemed to be partially independent of circadian organization (e.g., homeostatic processes related to digestion, motility, or metabolic rate can act as an hourglass mechanism). Similar results were obtained by Gerkema et al. (12) in the common vole, in which ultradian periodicity in feeding activity was independent of the circadian period, suggesting the existence of an endogenous ultradian oscillator independent of the circadian pacemaker.

In previous experiments involving rats kept under ahemeral LD cycles varying from 23 to 26 h, we showed that the growth rate of synchronized rats appeared to depend on the number of cycles that the animal had lived through (19, 33). Similar results were also obtained in the present experiment in relation to food consumption per t cycle. All the groups that exhibited light-entrained rhythms (from t = 24 h to t = 28 h) showed the same food consumption per cycle, regardless of cycle length. However, the animals kept under extreme conditions (t22 and t23) had a lower food intake than the animals of other groups. Food intake per t cycle in t22 and in some t23 rats may have been underestimated because their free-running rhythms were longer than 22 and 23 h, respectively.

Our experiments seem to support the idea that in well-synchronized animals, food intake may depend more on the number of cycles that the animal experiences (rhythmic control) than on real time (homeostatic control). Similar findings have been made in studies of the feeding behavior of subjects whose circadian rhythms of sleep-wakefulness free run when exposed to temporal isolation. Total caloric intake and locomotor activity remained constant per cycle irrespective the length of their "subjective days" (6, 14).

Although feeding duration might be considered an indirect estimation of food intake, the differences existing between these two parameters can be explained by the fact that feeding duration depends not only on food intake but also on the rate of ingestion, which, in turn, may depend on the individual, the phase, or the period of the LD cycle.

Although it has been shown that circadian organization and homeostatic control of a physiological process are independently operating factors, our results suggest that circadian and homeostatic processes influence each other so that, to a certain extent, it is possible to modify the rate of certain physiological processes such as food ingestion by altering the LD cycle to which the animals are synchronized.

Perspectives