University of Groningen Body Temperature Predicts the Direction of Internal Desynchronization in Humans Isolated from Time Cues

University of Groningen

Body temperature predicts the direction of internal desynchronization in humans isolated from time cues Daan, Serge; Honma, Sato; Honma, Ken-ichi

Published in: Journal of Biological Rhythms

DOI: 10.1177/0748730413514357

IMPORTANT NOTE: You are advised to consult the publisher's version (publisher's PDF) if you wish to cite from it. Please check the document version below. Document Version Publisher's PDF, also known as Version of record

Publication date: 2013

Link to publication in University of Groningen/UMCG research database

Citation for published version (APA): Daan, S., Honma, S., & Honma, K. (2013). Body temperature predicts the direction of internal desynchronization in humans isolated from time cues. Journal of Biological Rhythms, 28(6), 403-411. DOI: 10.1177/0748730413514357

Copyright Other than for strictly personal use, it is not permitted to download or to forward/distribute the text or part of it without the consent of the author(s) and/or copyright holder(s), unless the work is under an open content license (like Creative Commons).

Take-down policy If you believe that this document breaches copyright please contact us providing details, and we will remove access to the work immediately and investigate your claim.

Downloaded from the University of Groningen/UMCG research database (Pure): http://www.rug.nl/research/portal. For technical reasons the number of authors shown on this cover page is limited to 10 maximum.

Download date: 11-02-2018 Body Temperature Predicts the Direction of Internal Desynchronization in Humans Isolated from Time Cues

Serge Daan,*,1 Sato Honma, † and Ken-ichi Honma † *Center for Life Sciences, University of Groningen, Groningen, the Netherlands, and †Department of Chronomedicine, Hokkaido University Medical School, Sapporo, Japan

Abstract This publication presents a new analysis of experiments that were carried out in human subjects in isolation from time cues, under supervision of Jürgen Aschoff and Rütger Wever at the Max Planck Institute for Behavioural Physiology (Erling-Andechs, Germany, 1964-1974). Mean rectal temperatures

(t b) were compared between subjects who showed internal desynchronization (ID) and internal synchronization (IS) of the endogenous rhythms of sleep-

wakefulness and of body temperature. The results showed that t b was reduced

in long ID (circadian sleep-wake cycle length [ WSW] > 27 h) and increased in short

ID ( WSW < 22 h) relative to IS. In subjects with both ID and IS sections in the complete record, these differences were also found when comparing only the IS

sections: Low t b during IS predicts the later occurrence of long ID, and high t b predicts the incidence of short ID. While this association is associated with sex

differences in t b, it also occurs within each sex. To the extent that the variation in

tb reflects the variation in heat production (metabolic rate), the results are consistent with the proposition that the spontaneous frequency of the human sleep- wake oscillator is associated with the metabolic rate, as suggested on the basis of the proportionality of meal frequency and sleep-wake frequency. The finding thus has implications for our views on spontaneous sleep timing.

Keywords human sleep-wake rhythm, circadian, internal desynchronization, sex differences, body temperature, metabolic rate

Human subjects in prolonged isolation from first observed by Aschoff (1965) shortly after he external time cues often show internal desynchroni- established the endogenous nature of human circa- zation (ID) (Wever, 1979). This is a state in which the dian rhythms (Aschoff and Wever, 1962). Its occur- alternation of sleep and wakefulness is no longer in rence has been confirmed by several other synchrony with the circadian rhythm of body tem- laboratories (Colin et al., 1968; Jouvet et al., 1974; perature; that is, the 2 rhythms simultaneously Weitzman et al., 1979) and was claimed to occur in exhibit different frequencies. The phenomenon was about one third of all subjects living in isolation from

This article is dedicated to the memory of Jürgen Aschoff and Rütger Wever, pioneers of human circadian rhythms research, on whose data the work is based. 1. To whom all correspondence should be addressed: Serge Daan, Center for Life Sciences, University of Groningen, Hoofdweg 274, 9765 CN Paterswolde, the Netherlands; e-mail: [email protected].

403 Downloaded from jbr.sagepub.com at University of Groningen on December 19, 2013 404 JOURNAL OF BIOLOGICAL RHYTHMS / December 2013 external time cues (Wever, 1979). In ID, the cycle circadian period, suggesting that bowel movement length of the sleep-wake rhythm is highly variable slowed down in long ID. between individuals. It can be either much shorter Aschoff’s (1993) suggestion of a change in the meta- (16-20 h; “short ID”) or much longer (29-68 h; “long bolic rate was further supported by his analysis of the ID”) than the circadian body temperature rhythm, locomotor activity of subjects in isolation, measured which retains periods virtually always between 24 by motion sensors underneath the floor carpets in the and 26 h (Wever, 1979). It is quite unusual for other isolation unit. These data demonstrated that the total circadian rhythms to display such a large variability amount of activity per wake episode was independent in the period as the sleep-wake rhythm in humans. of the duration of wakefulness ( D) and that activity per As the processes leading to sleep and those leading hour was inversely proportional to D (Aschoff, 1993). to wakefulness still elude our scientific understand- Since this activity can account only for a small part of ing, it would be of great interest to better understand the metabolic rate, Aschoff (1993) suggested that basal physiological correlates of the period of the sleep- metabolism must also have been reduced in long ID. If wake rhythm, here indicated by WSW. this were true, one might hypothesize that the sleep- Aschoff and colleagues (1984, 1986) have specu- wake cycle is controlled by the same process that con- lated that there might be an association of the cycle trols heat production and, while normally entrained to length with the overall rate of metabolism. This approximately 24 h by the circadian pacemaker, can hypothesis was based on patterns of meal timing assume widely different period lengths under the during ID rather than on direct measurements of the influence of the overall energy metabolic rate in condi- metabolic rate. Their analysis yielded intriguing tions of isolation. This would provide new insight into results. The authors reported that subjects retained the nature of the buildup and breakdown of sleep need the habit of taking 3 meals per wake episode when during wakefulness and sleep (“Process S”) (Borbély, the latter became greatly compressed or expanded in 1982; Daan et al., 1984). ID. The intervals between breakfast, lunch, and din- No direct measurements of the human metabolic ner were reduced or increased in proportion to the rate during prolonged isolation from time cues exist duration of wakefulness (Aschoff et al., 1986). Hence, to put this hypothesis to the test. Since body tem- subjects eat breakfast, lunch, and dinner once per WSW perature is a potential correlate of metabolic heat of, say, 18 h in “short” ID and once per WSW of, say, 34 production, we have looked for a possible relation- h in “long” ID (these values are the means reported ship between the mean core body temperature and by Wever [1979]). If meals had the same size in the WSW in ID in existing data. The core body temperature different situations of internal synchronization (IS) has nearly always been recorded in human circadian and ID, this would lead to a substantial increase in experiments. Its analysis, however, usually focused overall food intake in short ID and to a large decrease on the timing of peaks and troughs in the tempera- in long ID. Aschoff et al. (1986) deplored that no pre- ture rhythm and virtually never on the mean values. cise quantitative measurement had been taken of We first analyzed the extensive data published by food intake and body weight in their experiments. Wever (1979) in his monumental book. Encouraged The limited data on body weight demonstrated no by these initial results (Fig. 1), we proceeded to the loss or gain following ID (Aschoff, 1989). The authors original data of Aschoff and Wever and performed an suggested that there may have been differences in the analysis of a sizeable sample of their experiments. metabolic rate associated with meal frequency. This speculation gained support in a study on 8 subjects by Green et al. (1987a, 1987b). These authors did ana- DATABASE lyze meal timing, meal size, and body weight and concluded that in long ID, the caloric intake rate The data that we used are all derived from the decreased on average by 21%, without negatively experiments performed by Professors Jürgen Aschoff affecting body weight. In their sample, short ID did and Rütger Wever in the first isolation facility specifi- not occur. Green et al. (1987b) indeed shared the view cally built for the purpose of studying human circa- that the mechanism responsible for the timing of dian rhythms. In this facility, at the Max Planck sleep and waking is also important for the timing of Institut für Verhaltensphysiologie in Erling-Andechs, meals. In an analysis of the timing of defecation, Germany, experiments were performed from July 6, Aschoff (1994) further showed that the interval from 1964 until December 21, 1989. The facility and the the main evening meal (or from wake-up) until the experiments until 1978 have been described in detail first defecation expanded proportionally with the by Wever (1979). In total, 485 experiments were

Downloaded from jbr.sagepub.com at University of Groningen on December 19, 2013 Daan et al. / BODY TEMPERATURE AND INTERNAL DESYNCHRONIZATION 405

preferentially selected who showed both IS and ID (see supplementary online material). We attributed

record sections to IS or ID on the basis of WSW as in

Figure 37 in Wever (1979). A WSW below 22 h was interpreted as short ID and above 27 h as long ID; IS sec-

tions all had a WSW between 22 and 27 h. This corresponded to individual records with the assign- ment by Wever and Aschoff in notes and publica-

tions, and indeed, the WSW values were mostly obtained from them. Since sections always lasted 8 or more days, this means that the body temperature and sleep-wake rhythm shifted at least 1 full cycle relative to each other. Since the original digitization of the data by Wever is no longer available, rectal temperature records (1 value per half-hour) and sleep- wake alternations (each reported onset and end of sleep) were digitized (with half-hour resolution). We

then computed the overall mean tb over the same sec-

tion of the record for which WSW was calculated. Such sections were typically in the order of 10 cycles long Figure 1. Relationship between WSW (h) and the mean t b (°C). The data are taken from Figures 16, 20, 22 to 24, 27, 28, 30, 36, 55 and defined from onset until onset of wakefulness. In to 57, 64, and 78 in Wever (1979). The W is taken from Wever’s SW the case of ID, this natural number of cycles for the (1979) analysis and the mean temperature from the mean of his fitted sine waves. Solid dots = IS; open symbols = short (dia- sleep-wake rhythm means that there is no whole mond) or long (square) ID. Lines connect 2 situations for the number of complete tb cycles. Although this theoreti- same individual. Dashed line = linear regression through all cally causes a bias, we have made computed esti- points: t (°C) = 38.21 – 0.044 * W (h). b SW mates of the error, which is maximally in the order of 0.015 °C, that is, less than the symbol size in our fig- carried out here. Most of these involved a subject ures. Another potential source of error is the tem- staying in complete isolation for 4 weeks. The experi- perature changes with the menstrual cycle in female mental conditions varied considerably over the years. subjects. As this was not recorded in the experiments, The behavior of subjects was studied to assess we simply use long-term overall mean values. responses of free-running rhythms to different con- stant conditions as well as entrainment responses to potential zeitgebers such as light, temperature, elec- RESULTS tromagnetic fields, and others. The paper archive of these experiments and the data collected have recently We first looked at the data extracted from Wever’s been largely restored. 1 It is now hosted by the Institute (1979) book in Figure 1. This suggests an interesting

for Medical Psychology, University of Munich (head: pattern. First, with an increase in WSW, there appears to

Prof. Dr. M.W. Merrow), and made accessible to be an overall decline in the mean tb between subjects.

researchers worldwide. The digitized archive has not The linear regression t b (°C) = 38.21 – 0.044 * WSW (h) is been retrieved so far. a first approximation of this trend. This association is In a pilot approach, we used the data on 14 sub- statistically significant when formally tested as jects reported on by Wever (1979) in his book. This though all data points were independent (Pearson r = analysis used all 14 figures in which the author plot- –0.666 [p = 0.003]; Spearman U = –0.529 [ p = 0.024];

ted the mean circadian sine wave curves fitted to the n = 18). Values of t b during long ID appear to be rectal temperature records of an individual in con- lower, while in the single case of short ID (Fig. 28 in

stant conditions. From this, we derived the overall Wever [1979]), the mean t b is near the upper range of

mean rectal temperature (t b) and compared it in mean tb values observed. This can be no more than a

Figure 1 with the WSW provided in the figure legends preliminary indication for a tendency of t b to be asso-

by Wever (1979). In the original data archive, a larger ciated with WSW in humans. In all 4 of the subjects who subsample (48 subjects) was selected first for the showed both an episode of IS and an episode of ID, absence of any zeitgebers. In particular, subjects were this negative association was also seen within the

Downloaded from jbr.sagepub.com at University of Groningen on December 19, 2013 406 JOURNAL OF BIOLOGICAL RHYTHMS / December 2013

males in the dataset, and the mean core temperature is negatively associated with the period of the sleep- wake cycle. Female and male subjects were differentially distrib- uted over short ID (3 females, 3 males), IS (9 females, 10 males), and long ID (3 females, 20 males). Although this distribution is significantly different from random ( F2 = 6.83; p = 0.033), this is Figure 2. Relationship between WSW (h) and the mean t b (°C). The data are taken from 48 individual subject records in the bunker archive from the Max Planck Institut für Verhaltensphysiologie, Abteilung not reliable evidence, Aschoff. (A) Individual data points. Solid symbols = IS; open symbols = ID. Black = males; red (or as the records ana- shaded) = females. Lines connect 2 points of the same individual when both an IS section and an ID section were present. (B) Mean values per sex and IS status. Each individual contributes 1 value; if sec- lyzed were selected tions with IS and ID are present within an individual, the ID section is chosen. Error bars are ±1 SEM. partly on the basis of the incidence of ID. The deviant distribu- individual, possibly suggesting a causal relationship. tions for the sexes are consistent with results This led us to address the question of the relationship reported earlier by Wirz-Justice et al. (1984) on between WSW and t b in humans more completely. partly the same data. The dataset that we then selected from the archive In 17 subjects, there were 2 sections in the record: comprises 48 subjects (including those in Wever’s 1 with IS and 1 with ID. These allow us to test

[1979] book, which were partly reanalyzed). In these whether the mean t b also changes systematically subjects, there were 65 sections, from each of which within individuals with WSW when the WSW lengthens W 1 mean value for t b and 1 value for SW were deter- or shortens in ID. Figure 3 shows the mean t b values mined (Fig. 2A). Overall, these 2 parameters were for 5 individuals who had both IS and short ID sec- indeed significantly negatively correlated, again tions and for 12 individuals who had both IS and when formally tested and pooling intraindividual long ID sections. The figure shows that there is nei- and interindividual associations (Pearson r = –0.254 ther a systematic increase in individual t b with short [p = 0.041]; Spearman U = –0.402 [ p = 0.001]; n = 65). ID nor a systematic decrease in long ID, as was sug- In Figure 2B, we show interindividual association gested by Figure 1. Intraindividual differences alone by summarizing the data with the overall between the mean t b recorded in IS and ID (groups mean values for each of 6 groups, that is, 3 groups 1 and 3) were tested and found to be not significant for each sex: short ID, IS, and long ID. Here, we by a paired t test: t = 1.418 ( n = 17; p = 0.175). did not include sections with IS when a section Apparently, the differences in core temperature with ID for the same individual was also present to between subjects displaying short or long ID are avoid multiple data from the same individuals. already present during the state of IS that usually

Regression of t b on circadian period and sex (male = 1, precedes ID. During IS (Fig. 3, solid symbols), the W female = 0) yields t b = 37.92 – 0.019 * SW – 0.389 * sex negative association of tb with group (1 = subject also (n = 48; WSW: t = –2.182 [ p = 0.034]; sex: t = –3.235 [ p = has a short ID section; 2 = only IS; 3 = subject also has

0.002]). Correlation coefficients for t b and WSW in this a long ID section) is indeed highly significant: smaller dataset are the following: Pearson r = Pearson r = –0.529 ( p = 0.001; n = 36) and Spearman –0.349 ( p = 0.015) and Spearman U = –0.440 ( p = U = –0.488 ( p = 0.003; n = 36). W 0.002). The partial correlation of t b and SW after The association between t b and WSW in these 36 indi- controlling for sex is r = –0.309 ( p = 0.034). Hence, vidual subjects during IS is analyzed in Figure 4. females have a higher mean core temperature than There is again a negative relationship between core

Downloaded from jbr.sagepub.com at University of Groningen on December 19, 2013 Daan et al. / BODY TEMPERATURE AND INTERNAL DESYNCHRONIZATION 407

W Figure 3. Mean ± SEM values for t b for 3 subgroups of subjects. Figure 4. Relationship between SW (abscissa) and t b (ordinate) Group 1 (diamonds): subjects who had IS during 1 section of the among 36 subjects during IS. Red (or shaded) = females; black = record and short ID during another section (3 females, 2 males). males. Closed symbols = individuals; large symbols = mean

Group 2 (circles): subjects with only IS (8 females, 11 males). WSW and mean t b (both with SEM) for 3 groups per sex. Group 1: Group 3 (squares): subjects with IS and long ID (2 females, 10 subjects who showed short ID in a different section of the record males). Solid symbols = IS; open symbols = ID. (diamonds). Group 2: subjects without ID (circles). Group 3: subjects who showed long ID in a different section of the record (squares). Dashed lines = regression through the means in females (red or shaded) and males (black) separately. temperature and the period of the sleep-wake cycle in the IS situation alone, both in females and in males. DISCUSSION The correlation between WSW and t b during IS, although indicated in both sexes, is not significantly different from zero: Pearson r = –0.307 ( p = 0.068, not signifi- The data of Aschoff and Wever evaluated here cant) and Spearman U = –0.263 ( p = 0.121, not signifi- with respect to one particular aspect unequivocally cant). This holds also for the partial correlation after demonstrate a remarkable association. Human sub- controlling for sex: r = –0.278 ( p = 0.106, not signifi- jects with a relatively high mean core body tempera- cant). Only with 1-tailed criteria, defendable on the ture have a tendency to develop short ID, and those basis of the theoretical prediction, the overall Pearson with a low core temperature often develop long ID. r = –0.307 is interpreted as p = 0.034. We consider the One might presume that the core temperature is significance of the association nonetheless marginal. partly a consequence of ID because ID affects the If we distinguish again between 3 groups of indi- overall proportion of time spent in wakefulness and since wakefulness is associated with a higher core viduals as above, WSW during IS, like t b, also turns out to be associated with group: Pearson r = 0.566 temperature than sleep. This is problematic since (p = 0.000) and Spearman U = 0.559 ( p = 0.000). The long ID is associated with an increase in the proportion of wakefulness (Aschoff et al., 1984; Wirz- mean values and SEMs for both tb and WSW are shown in Table 1. Justice et al., 1984) and short ID with a decrease The data thus show that the subjects developing (Aschoff, 1992). Long ID would thus be expected to

ID did already behave differently during IS. Both be characterized by a higher t b, and short ID by a mean temperature and circadian period during IS, lower t b, whereas the opposite is true. More specific although only marginally correlated with each other speculations concern the calorigenic effect of food, in this set, are reliable predictors of the probability of which should be higher in short ID with a higher desynchronizing by shortening (hot individuals) or meal frequency and less frequent heat-generating lengthening (cool individuals). postural changes in long ID (Wever, 1979). However,

Downloaded from jbr.sagepub.com at University of Groningen on December 19, 2013 408 JOURNAL OF BIOLOGICAL RHYTHMS / December 2013

Table 1. Interindividual mean values for WSW and t b during IS in 3 groups of subjects: 1) women in the preovulatory phase subjects exhibiting short ID in another section of the record, 2) subjects exhibiting under thermoneutral conditions only IS, and 3) subjects exhibiting long ID in another section of the record. (Gagnon and Kenny, 2011). In the Group W , mean ± SEM (h) t , mean ± SEM (°C) n SW b postovulatory phase (Bittel and 1 24.00 ± 0.532 37.50 ± 0.114 5 Henane, 1975), and following exer- 2 24.61 ± 0.157 37.23 ± 0.072 19 cise (Kenny and Jay, 2007; Gagnon 3 25.47 ± 0.198 36.78 ± 0.168 12 and Kenny, 2011), the female mean core temperature is elevated above all of these explanations fail because the analysis in that of men. The differences in thermoregulation are Figure 3 shows that individuals do not change their probably due to lower basal metabolism, more effec- tb when going from the IS state to the ID state. tive insulation, and thermogenic effects of progester- Rather, the pattern is generated by differentiation one (Ogawa, 1996). The increased tendency for between individuals with a different mean core females to enter short ID and for males to enter long temperature. In the state of IS, warmer subjects tend ID has been reported before (Wirz-Justice et al., 1984). to have slightly shorter circadian periods, and they The overall relationship of the tendency towards ID also more often enter the state of short ID. Cooler with temperature cannot solely be attributed to this subjects tend to have slightly longer circadian peri- gender difference since the association also holds for ods in IS and are more prone to enter the state of each of the sexes (Fig. 4). long ID (Table 1). The core temperature is determined by the balance The tendency towards long and short ID in sub- between heat production and heat loss. The associa- jects with a long and short WSW during IS, respectively, tion of a tendency towards ID with high and low core has been described in detail by Wever (1975, 1979 temperatures might be hypothetically attributed to [Fig. 29]). Wever (1979 [Fig. 29]) found that IS subjects either heat production or heat loss or to both. The

who later showed long ID had a mean WSW (and also association with heat production, that is, energy

Wtemp ) of 25.55 h and those who later showed short ID metabolism, is what Aschoff proposed in several had a period of 24.47 h. He explained this significant publications (Aschoff et al., 1986; Aschoff, 1990, distinction by a difference in the endogenous period 1993). His arguments were based on systematic dif- in a sleep-wake oscillator between the 2 subject ferences in the food intake rate (meals per hour), groups. In people with a slow sleep-wake oscillator, confirmed in terms of calories per hour by Green et the temperature oscillator period was lengthened in al. (1987b), and also in locomotor activity per hour as,

IS, and in those with a fast sleep-wake oscillator, it between individuals, WSW became much longer or was slowed down. This explanation is internally con- much shorter than 24 h. Indeed, Aschoff (1990, 1993) sistent with all of Wever’s data and also with our suggested that the activity per cycle remained con- finding that the patterns are generated by differences stant over a wide range of cycle lengths. between individuals rather than attributable to effects The extent of the difference in core body tempera-

of desynchronization. Wever’s (1979) model is of an ture involved in the huge differences between WSW abstract mathematical nature and does not specify values in ID (by a factor of 2-3 when comparing how individuals may physiologically vary to gener- short and long ID) may seem small. The coefficient

ate different directions of ID. of WSW is only –0.019 °C/h, corresponding to a varia-

The pattern is partly related to sex differentiation. tion of only about 0.5 °C over the WSW range of 16 to Sex differences in circadian periods have been estab- 48 h. This, however, is not incompatible with consid- lished before. Females have shorter free-running erable changes in the metabolic rate. Studies on sea- rhythms in IS than males (Wever, 1984). They also sonal adaptation in other large mammals such as the show slightly (6 min) shorter temperature/melatonin Alpine ibex and red deer have recently shown that rhythms under forced desynchrony protocols (Duffy large (>50%) changes in overall metabolism are et al., 2011). Female subjects in our analysis had a sig- strongly positively correlated with core tempera-

nificantly higher mean t b (37.42 °C; n = 15) than males tures varying over less than 1 °C (Signer et al., 2011; (36.90 °C; n = 21) ( t = 3.720; p = 0.001). This is in agree- Turbill et al., 2011). This result points to temperature ment with some other studies under different condi- variability that is indeed induced by the variation in tions (Bittel and Henane, 1975; Cunningham et al., metabolism but buffered by changes in thermal con- 1978; Schellen et al., 2012). A core temperature differ- ductance. Similarly, in humans, the circadian varia- ence is not found when men are compared with tion in heat production has been estimated as 324

Downloaded from jbr.sagepub.com at University of Groningen on December 19, 2013 Daan et al. / BODY TEMPERATURE AND INTERNAL DESYNCHRONIZATION 409

kcal/d or 324/1915 = 17% for a t b range of only 0.44 °C a Q 10 of 3.5. This is high for biochemical processes in (Kräuchi and Wirz-Justice, 1994). Recent studies general and extremely high for circadian oscillations have shown strong interindividual correlations in that are generally buffered against temperature varia- mean human tympanic ( r = 0.82) and rectal ( r = 0.83) tion. If the sleep-wake cycle were so sensitive to core temperatures with the overall daily energy expendi- temperature, it would easily be entrained by and thus ture (Van Marken Lichtenbelt et al., 2001, 2002). retain synchrony with the daily cycle of temperature. Curiously, the hypothesis of metabolic determina- Therefore, Aschoff’s (1993) hypothesis that not tion of WSW is consistent with data on tau mutant the core temperature per se but rather the metabolic hamsters: Oklejewicz et al. (1997, 2000) measured rate has such a dramatic influence on the sleep-wake heat production by indirect calorimetry in homozy- cycle seems preferable. If so, the metabolic rate gote, heterozygote, and wild-type Syrian hamsters in would apparently leave the circadian pacemaker in free-running conditions. They found that basal the SCN unaffected, just as it is resistant against tem- metabolism and the circadian cycle length are perature changes (Buhr et al., 2010), so that body inversely proportional. Although their results may be temperature oscillations remain in the circadian related to associated variations in body mass range, while WSW varies from about one half to twice (Refinetti, 2007), it is possible that the overall meta- the circadian W. The old data that we have evaluated bolic rate may affect the velocity of an oscillation that here were not collected with the aim to test such is controlling the alternation of sleep and wakeful- hypotheses and can at best help generate them. The ness. The latter may be a behavioral relaxation oscil- data also have their limitations that are related to the lator in which wakefulness leads to sleep and sleep time at which they were collected. There was ini- leads to wakefulness, such as the homeostatic sleep tially no high-resolution digital data collection, and drive Process S in the 2-process model of sleep regu- what is left is only the paper records. There was no lation (Borbély, 1982; Daan et al., 1984). It may also be recording of sleep electroencephalograms, so we a more or less localized internal oscillator (Nakao have only self-reported sleep data. The protocols et al., 2002), possibly in the brain, such as the oscilla- allowed subject control over illumination: initially tor involving the orexin neuronal system postulated over ceiling light and later still over bed reading by Phillips et al. (2011). In the first case, lower metab- lights. Even if such self-control can never constitute olism might cause a slower rate of Process S buildup a zeitgeber, this is not how the experiments would and hence a longer WSW, tending towards long ID; in be conducted today. Yet, the long series of experi- the second, it may lead to increased orexin tone with ments performed in the “bunker” in Andechs has the same consequences. The present analyses do not yielded a wealth of pioneer insights into the human allow further speculations on the mechanism. circadian system and can hardly be redone accord- While such roles for energy metabolism are specu- ing to modern standards. We hope this reanalysis lative, the suggestion ties in with a stream of recent may serve as a stimulus towards a new look at the literature pointing to a role of the circadian system in intriguing phenomenon of ID in the new light of the energy metabolism (Bass and Takahashi, 2010; Froy, interconnections between the metabolic rate and 2011; Albrecht, 2012; Stubblefield et al., 2012). This circadian organization. also entails links between circadian anomalies and human health problems that are related to energy metabolism (Rüger and Scheer, 2009; Maury et al., ACKNOWLEDGMENTS 2010; Wyse et al., 2011; Reiter et al., 2012; Roenneberg et al., 2012). Despite intense recent research efforts, S.D. is deeply grateful to the Heiwa Nakajima Foundation there is, however, not a consistent effect of sleep on for a guest professorship, hosted by K.H. and S.H. in energy metabolism across many recent studies Sapporo (2010), where this study was conceived. The (Klingenberg et al., 2012). authors gratefully acknowledge the efforts of Gayline Manalang in digitizing a large part of the temperature data An alternative suggestion might be that the core and Eduardo Mendoza for helping to organize this. In temperature itself rather than heat production would addition, S.D. acknowledges the help of Drs. Anna Wirz- be instrumental in determining the period of the Justice, Reimer Lund, Jürgen Zulley, and Barabara Helm in circadian sleep-wake cycle. This is most unlikely. securing the bunker archive and transferring this on loan to The linear regression through our complete dataset Drs. Till Roenneberg and Martha Merrow to host the is WSW = 140.10 – 3.044 * tb ( p = 0.041; n = 65) (Fig. 2A). archive at the Institute for Medical Psychology of Munich For temperatures around 37 °C, this would translate to University.

Downloaded from jbr.sagepub.com at University of Groningen on December 19, 2013 410 JOURNAL OF BIOLOGICAL RHYTHMS / December 2013

CONFLICT OF INTEREST STATEMENT Colin J, Timbal J, Boutelier C, Houdas Y, and Siffre M (1968) Rhythm of rectal temperature during a 6-month free- running experiment. J Appl Physiol 25:170-176. The author(s) declared no potential conflicts of interest Cunningham D, Stolwijk J, and Wenger C (1978) with respect to the research, authorship, and/or publication Comparative thermoregulatory responses of resting of this article. men and women. J Appl Physiol 45:908-915. Daan S, Beersma DGM, and Borbély AA (1984) Timing of human sleep: recovery process gated by a circadian NOTES pacemaker. Am J Physiol Regul Integr Comp Physiol 246:R161-R178. Supplementary material is available on the journal’s Duffy J, Cain S, Chang A-M, Phillips A, Münch M, Gronfier website at http://jbr.sagepub.com/supplemental. C, Wyatt J, Dijk D-J, Wright K, and Czeisler C (2011) Sex differemnce in the near-24-hour intrinsic period of the 1. This was performed in April 2012 under the auspices of human circadian timing ssystem. Proc Natl Acad Sci U the SRBR Chronohistory Committee by S. Daan, B. Helm, S A 108:15602-15608. R. Lund, A. Wirz-Justice, and J. Zulley. Froy O (2011) The circadian clock and metabolism. Clin Sci 120:65-72. Gagnon D and Kenny G (2011) Sex modulates whole-body sudomotor thermosensitivity during exercise. J Physiol REFERENCES 589:6205-6217. Green J, Pollak CP, and Smith GP (1987a) The effect of Albrecht U (2012) Timing to perfection: the biology of cen- desynchronization on meal patterns of humans living in tral and peripheral circadian clocks. Neuron 74:246-260. time isolation. Physiol Behav 39:203-209. Aschoff J (1965) Circadian rhythms in man: a self-sustained Green J, Pollak CP, and Smith GP (1987b) Meal size and oscillator with an inherent frequency underlies human intermeal interval in human subjects in time isolation. 24-hour periodicity. Science 148:1427-1432. Physiol Behav 41:141-147. Aschoff J (1989) Temporal orientation: circadian clocks in Jouvet M, Mouret J, Chouvet G, and Siffre M (1974) animals and human. Anim Behav 37:881-896. Toward a 48-hour day: experimental bicircadian Aschoff J (1990) Interdependence between locomotor activ- rhythm in man. In The Neurosciences: Third Study ity and duration of wakefulness in humans during iso- Program , Schmitt F and Worden F, eds, pp 491-497. lation. Experientia 46:870-871. Cambridge, MA: MIT Press. Aschoff J (1992) Temporal structures within the human Kenny G and Jay O (2007) Sex differences in postexer- sleep-wake cycle. In Discussions in Neuroscience, Vol. 8: cise esophageal and muscle tissue temperature Circadian Rhythms, Zatz M and Axelrod J, eds, pp 16-20. response. Am J Physiol Regul Integr Comp Physiol Amsterdam: Elsevier. 292:R1632-R1640. Aschoff J (1993) On the relationship between motor activity Klingenberg L, Sjodin A, Holmback U, Astrup A, and and the sleep-wake cycle in humans during temporal Chaput JP (2012) Short sleep duration and its associa- isolation. J Biol Rhythms 8:33-46. tion with energy metabolism. Obes Rev 13:565-577. Aschoff J (1994) The timing of defecation within the sleep- Kräuchi K and Wirz-Justice A (1994) Circadian rhythm of wake cycle of humans during temporal isolation. J Biol heat production, heart rate, and skin and core tempera- Rhythms 9:43-50. ture under unmasking conditions in men. Am J Physiol Aschoff J, Von Goetz C, Wildgruber C, and Wever RA Regul Integr Comp Physiol 267:R819-R829. (1986) Meal timing in humans during isolation without Maury E, Ramsey KM, and Bass J (2010) Circadian rhythms time cues. J Biol Rhythms 1:151-162. and metabolic syndrome: from experimental genetics to Aschoff J and Wever R (1962) Spontanperiodik des human disease. Circ Res 106:447-462. Menschen bei Ausschluss aller Zeitgeber. Nakao M, Yamamoto K, Honma K-I, Hashimoto S, Honma Naturwissenschaften 49:337-342. S, Katayama N, and Yamamoto M (2002) A phase Aschoff J, Wever R, Wildgruber C, and Wirz-Justice A dynamics model of human circadian rhythms. J Biol (1984) Circadian control of meal timing during temporal Rhythms 17:476-489. isolation. Naturwissenschaften 71:354-355. Ogawa T (1996) Thermal physiology of the young and old Bass J and Takahashi JS (2010) Circadian integration of of both sex, part 2: gender differences and effects of age- metabolism and energetics. Science 330:1349-1354. ing [in Japanese]. Human and Living Environment 4:2- Bittel J and Henane R (1975) Comparison of thermal 7. exchanges in men and women under neutral and hot Oklejewicz M, Hut RA, and Daan S (2000) Effect of deute- conditions. J Physiol 250:475-489. rium on the circadian period and metabolism in wild- Borbély AA (1982) A two-process model of sleep regula- type and tau mutant Syrian hamsters. Physiol Behav tion, I: physiological basis and outline. Hum Neurobiol 71:69-74. 1:195-204. Oklejewicz M, Hut RA, Daan S, Loudon ASI, and Stirland Buhr E, Yoo S-H, and Takahashi J (2010) Temperature as a AJ (1997) Metabolic rate changes proportionally to fre- universal resetting cue for mammalian circadian oscilla- quency in tau mutant Syrian hamsters. J Biol Rhythms tors. Science 330:379-385. 12:413-422.

Downloaded from jbr.sagepub.com at University of Groningen on December 19, 2013 Daan et al. / BODY TEMPERATURE AND INTERNAL DESYNCHRONIZATION 411

Phillips A, Czeisler C, and Klerman E (2011) Revisiting Van Marken Lichtenbelt W, Schrauwen P, Van de Kerckhove spontaneous internal desynchrony using a quantitative S, and Westerterp-Plantinga M (2002) Individual varia- model of sleep physiology. J Biol Rhythms 26:441-453. tion in body temperature and energy expenditure in Refinetti R (2007) Absence of circadian and photoperiodic response to mild cold. Am J Physiol Endocrinol Metab conservation of energy expenditure in three rodent spe- 282:E1077-E1083. cies. J Comp Physiol B 177:309-318. Van Marken Lichtenbelt W, Westerterp-Plantenga M, and Reiter R, Tan D-X, Korkmaz A, and Ma S (2012) Obesity van Hoydonck P (2001) Individual varation in the rela- and metabolic syndrome: association with chronodis- tion between body temperature and energy expenditure ruption, sleep deprivation, and melatonin suppression. in response to elevated ambient temperature. Physiol Ann Med 44:564-577. Behav 73:235-242. Roenneberg T, Allebrandt KV, Merrow M, and Vetter C Weitzman E, Czeisler C, and Moore-Ede M (1979) Sleep- (2012) Social jetlag and obesity. Curr Biol 22:939-943. wake, neuroendocrine and body temperature circadian Rüger M and Scheer FAJL (2009) Effects of circadian dis- rhythms under entrained and non-entrained (free-run- ruption on the cardiometabolic system. Rev Endocr ning) conditions in man. In Biological Rhythms and Their Metab Disord 10:245-260. Central Mechanism: A Naito Foundation Symposium , Suda Schellen L, Loomans MGLC, de Wit MH, Olesen BW, and M, Hayaishi O, and Nakagawa H, eds, pp 199-227. van Marken Lichtenbelt WD (2012) The influence of Amsterdam: Elsevier. local effects on thermal sensation under non-uniform Wever R (1975) The circadian multioscillator system of environmental conditions: gender differences in ther- man. Int J Chronobiol 3:19-55. mophysiology, thermal comfort and productivity dur- Wever R (1979) The Circadian System of Man: Results of ing convective and radiant cooling. Physiol Behav Experiments under Temporal Isolation . Berlin: Springer. 107:252-261. Wever R (1984) Sex differences in human circadian rhythms: Signer C, Ruf T, and Arnold W (2011) Hypometabolism and intrinsic periods and sleep fractions. Experientia basking: the strategies of Alpine ibex to endure harsh 40:1226-1234. over-wintering conditions. Funct Ecol 25:537-547. Wirz-Justice A, Wever R, and Aschoff J (1984) Seasonality Stubblefield JJ, Terrien J, and Green CB (2012) Nocturnin: at in freerunning circadian rhythms in man. the crossroads of clocks and metabolism. Trends Naturwissenschaften 71:316-319. Endocrinol Metab 23:326-333. Wyse CA, Selman C, Page MM, Coogan AN, and Hazlerigg Turbill C, Ruf T, Mang T, and Arnold W (2011) Regulation DG (2011) Circadian desynchrony and metabolic dys- of heart rate and rumen temperature in red deer: effects function: did light pollution make us fat? Med of season and food intake. J Exp Biol 214:963-970. Hypotheses 77:1139-1144.

Downloaded from jbr.sagepub.com at University of Groningen on December 19, 2013