Downloaded from http://rspb.royalsocietypublishing.org/ on June 14, 2017
Social forces can impact the circadian
clocks of cohabiting hamsters
Matthew J. Paul†, Premananda Indic and William J. Schwartz
rspb.royalsocietypublishing.org
Research
Cite this article: Paul MJ, Indic P, Schwartz
WJ. 2014 Social forces can impact the circadian
clocks of cohabiting hamsters. Proc. R. Soc. B
281: 20132535.
http://dx.doi.org/10.1098/rspb.2013.2535
Received: 26 September 2013
Accepted: 13 January 2014
Subject Areas:
behaviour, neuroscience
Keywords:
social interactions, circadian rhythms,
body temperature, wavelet analysis
Author for correspondence:
Matthew J. Paul
e-mail: [email protected]
†
Present address: Neuroscience Institute,
Georgia State University, PO Box 5030,
Atlanta, GA 30302-5030, USA.
Department of Neurology, University of Massachusetts Medical School, Worcester, MA 01655, USA
A number of field and laboratory studies have shown that the social environment influences daily rhythms in numerous species. However, underlying
mechanisms, including the circadian system’s role, are not known. Obstacles
to this research have been the inability to track and objectively analyse
rhythms of individual animals housed together. Here, we employed temperature dataloggers to track individual body temperature rhythms of pairs
of cohabiting male Syrian hamsters (Mesocricetus auratus) in constant darkness and applied a continuous wavelet transform to determine the phase
of rhythm onset before, during, and after cohabitation. Cohabitation altered
the predicted trajectory of rhythm onsets in 34% of individuals, representing
58% of pairs, compared to 12% of hamsters single-housed as ‘virtual pair’
controls. Deviation from the predicted trajectory was by a change in circadian period (t), which tended to be asymmetric—affecting one individual
of the pair in nine of 11 affected pairs—with hints that dominance might
play a role. These data implicate a change in the speed of the circadian
clock as one mechanism whereby social factors can alter daily rhythms.
Miniature dataloggers coupled with wavelet analyses should provide powerful tools for future studies investigating the principles and mechanisms
mediating social influences on daily timing.
1. Introduction
Organisms have evolved endogenous, approximately 24 h (circadian) clocks to
anticipate and prepare for predictable changes in the day– night environment.
However, the period (t) of these internal clocks does not precisely match the
Earth’s daily cycle and must be adjusted by environmental factors (e.g. light,
food availability and temperature) to optimally time the organism’s daily
rhythms in physiology and behaviour. Field and laboratory studies have
shown that interactions between individuals can also influence the timing of
daily rhythms and thus may be critical for the adaptive function of the clock,
but elucidating the circadian role(s) and mechanism(s) of social cues has been
problematic (for reviews, see [1–4]).
Some of the difficulties with this research are illustrated by previous
work investigating social effects on the circadian rhythms of Syrian hamsters
(Mesocricetus auratus). Among widely available laboratory species, hamsters
would seem to be well suited as subjects for such studies: they respond predictably to a range of non-photic phase-shifting cues [5] and have been used for
dissecting the neurobiological and hormonal substrates that underlie social
interactions, especially agonistic (conflict) behaviours [6,7]. Nevertheless, the
effects of social cues on hamster rhythms have been inconsistent. Closely
housed or co-housed hamsters not in physical contact do not affect each
other’s rhythms [8,9], although more rapid re-entrainment to a phase-advanced
light –dark (LD) cycle has been reported when males are paired with oestrous
females [10]. Other studies have restricted the interval of social contact to a few
hours per day. This approach has led to conflicting results [11 –14], perhaps due
to differences in experimental design (blind versus intact animals, maintenance
in continuous darkness versus light, interaction via neutral cage versus resident –intruder paradigm). Refinetti et al. [13] co-housed two hamsters in
physical contact with each other but with markedly different t’s; one was
entrained to a short 23.3 h LD cycle while the other was blinded and free ran
& 2014 The Author(s) Published by the Royal Society. All rights reserved.
Downloaded from http://rspb.royalsocietypublishing.org/ on June 14, 2017
Figure 1 illustrates the experimental timeline. As detailed
fully in Material and methods, the Tb rhythm of each hamster
was recorded by an implanted datalogger for a total of
85 days in DD, representing 13–14 days before randomized
pairing, 62 –63 days of cohabitation with a single running
wheel and 9 –10 days following separation. Figure 2 illustrates data analysis for one of the hamster pairs. Figure 2a
shows double-plotted actograms of the Tb rhythms in DD
of a pair of hamsters before, during (within the boxes) and
after cohabitation. They are displayed in green (no. 99) and
red (no. 94), with both rhythms merged in the adjacent
enlarged actogram; simultaneous activity appears yellow.
As described in Material and methods, rhythm onsets were
determined by transforming the raw data using a continuous
wavelet function (Mexican Hat). Onsets were defined as the
zero crossing along the upward slope of the wavelet-derived
circadian component of the Tb data. This method accurately
tracks Tb rhythm onset occurring, on average, 94 min
before the onset of wheel-running activity [16]. Figure 2b
(upper figure) shows the time series obtained for three days
during cohabitation (when hamster no. 99 phase-leads
eg
in
9 weeks
DD
n
se
pa
ra
tio
9 weeks
en
d
m
ea
su
re
s
24 weeks
6 L : 18 D
Tb
r
pa ecor
iri di
ng ng
s sb
surgeries
Figure 1. Timeline of experimental procedures. 6 L : 18 D, photoperiod with
6 h of light and 18 h of dark per day; DD, constant darkness.
hamster no. 94 by several hours), with wavelet-determined
onsets indicated as black marks; in the lower figure, onsets
are overlaid on an excerpt of hamster no. 94’s single-plotted
actogram around the time of initial pairing. In Figure 2c,
onset phase is double-plotted over time (denoted by the
green (no. 99) or red (no. 94) dots) with the days before
and after cohabitation denoted by shading. The blue line represents the predicted trajectory of onset phases over the
duration of the experiment, as extrapolated from the linear
regression of onset phases from the 13 to 14 days before cohabitation; dashed blue lines enclose the 95% prediction interval
(sp., the area in which 95% of all data points would be
expected to fall based on the regression line; this is not to
be confused with the 95% confidence interval, which is the
area with a 95% chance of containing the true regression
line). Plotting each hamster’s predicted and actual onsets in
this manner provides an objective method to determine
whether, and at which point, t of a hamster’s circadian Tb
rhythm is stably altered during cohabitation.
(a) Cohabitation alters Tb rhythm t in a subset of
hamsters
For hamster no. 94, figure 2c shows that the onset phases of
its Tb rhythm during cohabitation deviated markedly from
that predicted, beginning around day 17 (4 days after pairing)
and remaining outside the prediction interval for the duration of the experiment; on the other hand, onset phases of
hamster no. 99 remained within the 95% prediction interval.
For all the cohabiting pairs (n ¼ 19), 13/38 (34%) individuals
exhibited a trajectory of onsets outside their 95% prediction
interval; among the virtual pairs (controls; n ¼ 13), only
3/26 (12%) individuals deviated from their 95% prediction
interval. These proportions were significantly different by
x 2 analysis ( p , 0.05). Deviation from the prediction interval
in affected animals was by a change of t. The mean absolute
change of t (difference between t before cohabitation and t
for the 10 days prior to separation) in affected animals was
larger than in non-affected and virtually paired animals,
and the latter did not differ from each other (mean + s.e. ¼
0.13 + 0.02, 0.05 + 0.01, 0.07 + 0.01 for affected, non-affected
and virtually paired hamsters, respectively, p , 0.0001, oneway ANOVA; affected versus non-affected and virtually
paired hamsters, p , 0.001 for both comparisons, Fisher’s
PLSD). When assessed after separation, affected hamsters
continued to exhibit a greater change of t than non-affected
and virtually paired hamsters (mean + s.e. absolute t
change (before minus after cohabitation) ¼ 0.13 + 0.02,
Proc. R. Soc. B 281: 20132535
2. Results
2
rspb.royalsocietypublishing.org
with a t approximately 24 h. There was no synchronization or
relative coordination of the activity of the blind hamster to
that of the entrained one. The conclusion from all these
studies has been that social factors provide only weak
inputs to the hamster circadian clock.
If social interactions do in fact represent a weak coupling
force between individuals, then oscillator theory would
suggest that strong effects might be revealed only when cohabitants are in direct, unrestrained physical contact for a
relatively long period of time and their free-running t’s are
relatively close to one another [15]. Only recently has it
become feasible to test this prediction. Rather than relying
on the inspection of group activity data displayed on actograms, it is now possible to continuously record rhythms
from individual, cohabiting animals over months and to analyse the data using rigorous methods that do not demand
invariance of period and amplitude over time. To our knowledge, no reported experiment in hamsters (or any other
animal) has incorporated such a design, which should be
more sensitive for inducing and detecting social influences
on circadian rhythmicity.
Here, we report the results of recording individual body
temperature (Tb) rhythms of pairs of male Syrian hamsters,
co-housed in a cage with a single running wheel for two
months in constant darkness (DD). We determined the
phase of rhythm onset before, during, and after cohabitation
using a continuous wavelet transform that we have applied
and validated for this purpose [16]. Data from paired hamsters were compared to control hamsters that were singly
housed without cagemates; pairs and controls were treated
identically, with the timing of daily care, behavioural observations and recording of experimental measures of each
control hamster tethered (in time) to another control hamster
to form a series of ‘virtual’ pairs. We demonstrate that the
social environment can exert a significant effect on the circadian clocks of cohabiting hamsters, manifested as a change of
t that, if it occurs, usually involves only one hamster of an
affected pair.
Downloaded from http://rspb.royalsocietypublishing.org/ on June 14, 2017
(a)
(b)
3
no. 99
06.00
12.00
18.00
0
06.00
12.00
18.00
0.8
0
06.00
12.00
18.00
0
06.00
12.00
18.00
0
amplitude
0
0.4
0
–0.4
–0.8
47
48
days
49
rspb.royalsocietypublishing.org
0
no. 94
06.00
12.00
18.00
0
06.00
12.00
18.00
Proc. R. Soc. B 281: 20132535
0
0
24
phase difference in hours
phase difference in hours
(c) 24
12
0
–12
–24
0
20
40
days
60
80
12
0
–12
–24
0
20
40
days
60
80
Figure 2. Cohabitation alters t in a subset of hamsters. Representative body temperature (Tb) records of a cohoused hamster pair before, during and after cohabitation, in which t was altered by cohabitation in one of the cohabitants. Tb rhythms are plotted in (a) as individual actograms (left) and as a combined actogram
(enlarged actogram, right). Black-lined boxes overlaid on the actograms indicate the duration of cohabitation. Elevated Tb of hamster nos. 99 and 94 are depicted in
green and red, respectively, with overlapping intervals of elevated Tb in the combined actogram indicated in yellow. Phase onsets of Tb rhythms were extrapolated
using wavelet analysis and plotted in (b) as black marks overlaying the wavelet-derived Tb time series for both cohabitants (upper plot; green and red line ¼
extrapolated Tb rhythms of hamster nos. 99 and 94, respectively) and an excerpt of hamster no. 94’s Tb actogram before and after pairing (lower plot). Phase onsets
are double-plotted across days in (c) for hamster no. 99 (left, green dots) and hamster no. 94 (right, red dots) along with their predicted trajectory of phase onsets
(blue line) and their prediction interval (dashed blue lines). Intervals before and after cohabitation are denoted by shading.
0.07 + 0.01 and 0.06 + 0.01 for affected, non-affected and
virtually paired hamsters, respectively; p , 0.003, one-way
ANOVA; affected versus non-affected and virtually paired
hamsters, p , 0.004 for both comparisons, Fisher’s PLSD).
When the entire sample of actually paired hamsters was
grouped together, the mean absolute change of t did not
differ from virtually paired hamsters. Six of the affected
hamsters exhibited a shortening of t, whereas the remaining
seven exhibited a lengthening of t. Of note, nine of the
affected hamsters were co-housed with a non-affected cagemate; the remaining four affected animals represented two
pairs. Thus, a total of 11 of 19 actual pairs were affected by
cohabitation, and for the majority of these affected pairs,
the change of t was asymmetric, occurring in only one
hamster of the pair. Unlike the change of t, there were no
differences in the change of amplitude between affected,
non-affected and virtually paired hamsters (mean + s.e.
absolute change of amplitude ¼ 0.09 + 0.02, 0.08 + 0.02,
0.09 + 0.02 for affected, non-affected and virtually paired
hamsters, respectively; p . 0.95, one-way ANOVA).
(b) Attributes of hamsters and hamster pairs
We found no significant differences in pre-cohabitation circadian attributes or body mass between hamsters that exhibited
a change of t and those that did not, whether grouped as
‘affected’ (n ¼ 13) versus ‘non-affected’ (n ¼ 25) hamsters or
as hamsters from ‘affected’ (n ¼ 22) versus ‘non-affected’
Downloaded from http://rspb.royalsocietypublishing.org/ on June 14, 2017
Table 1. Mean + s.e. of circadian and physical attributes of hamsters in actual pairs. BM, body mass.
4
non-affected hamsters
affected pairs
non-affected pairs
24.04 + 0.03
0.72 + 0.02
24.00 + 0.02
0.72 + 0.02
24.02 + 0.02
0.71 + 0.02
24.01 + 0.03
0.74 + 0.02
10.60 + 0.13
0.69 + 0.07
10.41 + 0.12
0.72 + 0.06
10.60 + 0.13
0.73 + 0.06
10.31 + 0.13
0.68 + 0.06
circadian
0.04 + 0.003
0.04 + 0.003
0.04 + 0.003
0.04 + 0.004
150.4 + 4.4
154.0 + 3.3
154.3 + 3.7
149.1 + 3.5
155.6 + 5.2
158.7 + 3.7
159.4 + 4.1
155.1 + 4.4
(n ¼ 16) pairs (table 1). In addition, we compared average
intra-pair differences for each of these measures in ‘affected’
(n ¼ 11) versus ‘non-affected’ (n ¼ 8) pairs and again found
no significant differences (table 2). The phase relationship
between Tb rhythms at the onset of pairing did not predict
a t change during cohabitation. In addition, the number of
days of cohabitation until the phase trajectory deviated
from the prediction interval as well as the phase relationship
at deviation were variable; and whether the change of t in the
affected hamster was biased towards or away from the unaffected hamster was indeterminate in some cases and variable
in others. Estimated testis volume (ETV) after, but not before,
cohabitation was negatively correlated with the absolute
change of t ( pre-cohabitation t minus post-cohabitation t;
R 2 ¼ 0.12, p , 0.05, regression; figure 3a).
We examined circadian, physical and behavioural attributes of hamsters in the nine pairs that exhibited the
asymmetrical change of t (i.e. pairs with only one affected
hamster; table 3). Of the circadian attributes, only amplitude
precision was predictive, with a higher proportion of affected
hamsters exhibiting higher precision (six of eight pairs (in
one pair precision was equal); p , 0.05, x 2 analysis). Overall,
however, cycle-to-cycle variability in this measure was extremely low (mean + s.e. for all animals ¼ 0.04 + 0.003). There
were some indications that dominance status might be an
important variable. In seven of the nine pairs, the affected
hamster weighed less than the non-affected hamster at pairing
( p , 0.05, x 2 analysis); differential body mass often predicts
dominance status with the heavier hamster being dominant
(reviewed in [17]). By the end of the experiment, however,
the proportion of affected hamsters with a lower body
mass fell short of significance (six of the nine pairs; p ¼ 0.16,
x 2 analysis). Affected hamsters in six of the nine pairs had a
lower ETV at week 29 (nine weeks before pairing), but this proportion was not significant ( p ¼ 0.16, x 2 analysis). Notably, by
the end of the experiment, the affected hamsters in all nine
pairs had significantly lower ETVs ( p , 0.05, x 2 analysis), due
to a decrease in ETV of the affected hamster in each pair (from
week 29 to the end of the experiment: change in ETV + s.e.
of affected versus non-affected hamsters was 2911.0 +
326.7 mm3 versus 214.2 + 242.8 mm3, p , 0.05, Student’s
t-test). Similar to that seen for all paired hamsters, ETV measures at the end of the experiment for hamsters from asymmetric
pairs were negatively correlated with the absolute change
of t (pre-cohabitation t minus post-cohabitation t; R 2 ¼ 0.3,
p , 0.02, regression; figure 3b). Interestingly, at the time of
Table 2. Mean + s.e. intra-pair difference in circadian and physical
attributes of hamsters in actual pairs.
affected
pairs
non-affected
pairs
0.13 + 0.02
0.11 + 0.02
0.17 + 0.02
0.09 + 0.02
0.58 + 0.10
0.47 + 0.12
0.31 + 0.09
0.02 + 0.003
0.34 + 0.06
0.02 + 0.006
21.6 + 5.4
13.0 + 5.1
21.2 + 7.7
19.4 + 6.6
circadian
t (h)
amplitude (8)
a (h)
t precision (h)
amplitude precision (8)
physical
BM at pairing (g)
BM at end (g)
the t change during cohabitation, the affected hamster was the
second of each pair in the wheel (seven of seven pairs
(the order of individuals in the remaining two pairs was indeterminate, as their rhythms were approx. 1808 out of phase);
p , 0.05, x 2 analysis). Nonetheless, we were able to unequivocally assign dominance in only three of the nine pairs owing to
an absence of agonistic behaviours during behavioural observations; in these pairs, the affected hamster was subordinate in
two and dominant in one.
3. Discussion
Here, we show that the social environment can impact circadian rhythmicity of Syrian hamsters by altering the period (t)
of the circadian clock. Previous studies in this species have
been inconsistent and limited to acute phase shifts or to
modifications in light entrainment (acceleration in the rate
of re-entrainment to a shift in the LD cycle) [8–14]. Because
it was not possible in these studies to track individual activity
rhythms when animals were housed together, they restricted
social interactions to brief encounters, limited the physical
contact between individuals or used blinded hamsters; each
of these may have diminished the impact of social cues.
Here, we implanted temperature dataloggers that allowed
us to record rhythms of individual animals pair-housed
in complete physical contact for an extended duration
Proc. R. Soc. B 281: 20132535
t (h)
amplitude (8)
a (h)
t precision (h)
amplitude precision (8)
physical
BM at pairing (g)
BM at end (g)
rspb.royalsocietypublishing.org
affected hamsters
Downloaded from http://rspb.royalsocietypublishing.org/ on June 14, 2017
(a)
5
|t change| (h)
rspb.royalsocietypublishing.org
0.3
0.2
0.1
0
Proc. R. Soc. B 281: 20132535
(b)
|t change| (h)
0.3
0.2
0.1
0
2000
3000
ETV
4000
(mm2)
5000
6000
before cohabitation
7000
2000
3000
4000
ETV
5000
(mm2)
6000
7000
at end
Figure 3. t change of cohabitants is correlated with testis size after cohabitation. Regression analyses of absolute t change of individuals ( pre-cohabitation t minus
post-cohabitation t) with their ETV before (left) and after (right) cohabitation as assessed (a) in all hamster pairs or (b) in asymmetric pairs only.
Table 3. Circadian, physical and behavioural attributes of hamsters in
asymmetric pairs.
no. affected hamsters/
asym. pairs
circadian
t (shorter)
amplitude (lower)
a (shorter)
t precision (higher)
amplitude precision (higher)
physical
BM at pairing (lighter)
BM at end (lighter)
ETV before pairing (lower)
ETV at end (lower)
behavioural
wheel occupation (second)
dominance (subordinate)
4/9
3/9
6/9
4/9
6/8*
7/9*
6/9
6/9
9/9*
7/7*
2/3
*p , 0.05.
(approx. two months). We suspect that such unrestrained
interaction for a prolonged period of time is likely to be an
important factor for eliciting cohabitation effects on the
circadian clock.
The other advance we made was to employ wavelet transforms to analyse the body temperature rhythms. Traditionally,
t has been calculated using methods that assume sinusoidal
waveforms and rhythms with period and amplitude that are
constant over time or using subjectively determined phase
points (e.g. visual inspection of actograms); such methods
are inappropriate for the present application, in which the
oscillators (animals) are combined and are not operating
(living) independently from one another. Wavelet-determined
rhythm onset provided an objective phase marker that permitted the analyses exemplified in figure 2c, in which the
trajectory of onsets over time could be precisely mapped.
Various degrees of social entrainment have been reported
for a number of species indicating that social influences on t
are not restricted to hamsters (for reviews, see [1–4]). Similar
to reports in other species [18], we found that the effects of
the social environment occurred in a subset of hamsters, supporting the notion that there are individual differences in
the sensitivities of the circadian system to social factors.
In this experiment, 13/38 (34%) of individuals, representing
11 of 19 (58%) pairs, deviated from the prediction interval.
We investigated possible factors responsible for these individual differences by evaluating several circadian and physical
attributes of the cohabitants. However, no differences were
detected between affected and non-affected hamsters or
pairs in the variables considered. Oscillator theory would
suggest that the effects of cohabitation on circadian rhythmicity might be more likely in pairs of hamsters with more
similar t’s; however, we failed to detect a statistically significant difference in the intra-pair t difference between affected
and non-affected pairs ( p ¼ 0.27, Student’s t-test), even
though the means were in the predicted direction (smaller
in affected pairs). Perhaps a larger sample size containing
a wider range of intra-pair t differences is needed to reach
statistical significance.
Downloaded from http://rspb.royalsocietypublishing.org/ on June 14, 2017
4. Material and methods
Sixty-eight male Syrian hamsters (M. auratus) housed from birth
on a 14 L : 10 D cycle were purchased from Harlan Laboratories
(Haslett, MI, USA) at five to nine weeks of age. Two pairs of
6
Proc. R. Soc. B 281: 20132535
socially induced changes in wheel running, either by social
facilitation or by social exclusion, could contribute to the t
changes observed in this study. Given that we could not
determine the wheel-running activity of individual hamsters,
we could not assess whether hamsters that shortened t
increased their wheel-running activity during cohabitation
and vice versa. Future studies providing co-housed hamsters
with two, one or no running wheels are required to parse out
the effect of the running wheel, if any, in cohabitationinduced t changes. Another strategy would be to provide a
large wheel that allows both hamsters to run simultaneously
(assuming hamsters would do so and coordinate their
running speed).
At present, it is not clear which neural substrates mediate
social influences on the circadian system. Social modulation
of t might act through serotonin inputs to the suprachiasmatic
nucleus, which are thought to mediate the effects of arousal and
wheel running on circadian activity (reviewed in [31,32]).
Another possibility is that social forces might act through the
intergeniculate leaflet of the thalamus, a pathway known to
mediate the influence of other non-photic factors [33].
These findings raise the question as to whether social
influences on t play a functional role in the life of the animal.
In the wild, animals inhabit complex ecosystems involving
interactions with both conspecifics and heterospecifics, and
the ability to adjust one’s daily activity in response to these
interactions could be beneficial. For example, family groups
and communities might coordinate their efforts to achieve
common goals [34,35]. Conversely, when the social context
involves competition, territoriality, dominance hierarchies or
predator avoidance, one would predict segregation rather
than synchronization of activity [21,36–42]. A change in t
may alter the timing of activity onset under entrained conditions (i.e. the phase angle of entrainment), which could
bring the activity of individuals either closer together or further
apart even while entrained to the ambient LD cycle [2].
There is a growing literature describing numerous examples
of inter-organismal influences on daily activity. Because of technical difficulties in tracking individual activity rhythms of
group-housed animals, the mechanisms underlying most of
these examples are not known. Here, we describe the use of
two relatively recent advances, temperature-sensitive dataloggers and wavelet analyses, which enabled us to overcome
this limitation. In doing so, we uncovered evidence in support of one potential mechanism—alterations in the speed
of the circadian clock. Many questions remain. Do other
species use this same mechanism to adjust their daily
activity? Can temporal synchronization and/or temporal segregation be explained by social influences on t interacting
with light entrainment? What factors are responsible for individual differences in, and what environmental conditions
promote, social influences on circadian activity? The use of
temperature-sensitive dataloggers coupled with wavelet
analyses, as described here, should provide powerful tools
for future investigations asking these and other questions
pertaining to social chronobiology.
rspb.royalsocietypublishing.org
These data confirm and extend our findings from a preliminary study in which we monitored wheel-running activity in
hamsters (without implanted temperature dataloggers) cohoused with a single running wheel in DD [19]. In that
report, we found that long-term cohabitation altered the
wheel-running activity of the hamster pairs during cohabitation and that a subset of individuals emerged from the
cohabitation interval with an altered t. In both that report
and this one, the cohabitation effect on t tended to be
asymmetric, i.e. affecting only one individual of the pair. Dominance hierarchies could explain this asymmetry if social
factors preferentially impact the dominant or subordinate
individual. Deer mice (Peromyscus maniculatus) co-housed in
constant darkness exhibited synchronous group activity
rhythms with a period similar to the endogenous rhythm of
the dominant mouse [20]. There is also anecdotal evidence
that dominance could lead to temporal segregation. Bovet
[21] reported one example in which three subordinate
long-tailed field mice (Apodemus sylvaticus) avoided the
time of peak out-of-burrow activity of a dominant mouse
when housed together in a laboratory cage. While we did
not observe sufficient aggression in our hamsters to determine the intra-pair hierarchy, we found circumstantial
evidence suggesting dominance might play a role: affected
hamsters weighed less than their non-affected cagemates
and experienced partial testicular suppression (decreases in
testis size). Notably, testis size at the end of the experiment
was negatively correlated with the absolute change of t
(figure 3). One interesting hypothesis is that low levels
of testosterone, here due to social subjugation [22], render
the circadian system of hamsters more susceptible to environmental influences, which could be social or not. Castration
decreases and testosterone replacement restores t precision
in male Syrian hamsters [23], and in degus (Octodon degus)
testosterone decreases the responsiveness of the circadian
system to social cues that facilitate re-entrainment to shifts in
the LD cycle [24].
Of note, we observed more dramatic t changes in our preliminary report [19]—including cases of complete temporal
segregation—than in this study. While we did not measure
testis size in our preliminary report, hamsters in that report
were entrained to a long (14 L : 10 D) photoperiod before
release into DD. Hamsters are a photoperiodic species (for
review, see [25]) and interpret DD as a short winter day
length, so those hamsters almost certainly exhibited much
more extensive testicular regression and decreased circulating
testosterone than that seen in the affected hamsters of this
report. This greater testicular suppression may explain the
larger t changes in that experiment. Alternatively or in
addition, DD may have triggered photoperiodic behavioural
changes that amplify the effects of the social environment on
the circadian system (e.g. increased aggression [26,27]).
In this study, we removed this potentially confounding
variable by first maintaining the hamsters in a short day
length (6 L : 18 D photoperiod) for 24 weeks, a length of
time known to render hamsters unresponsive to the seasonal
effects of short day lengths ( photorefractory).
We housed hamster pairs in cages with a single running
wheel, a manipulation intended to induce competition for a
limited resource that could be resolved through temporal segregation (wheel running is rewarding to hamsters [28]).
However, wheel running also provides feedback to the circadian clock with high levels shortening t [29,30]. Thus,
Downloaded from http://rspb.royalsocietypublishing.org/ on June 14, 2017
Figure 1 illustrates our experimental protocol. Hamsters were
maintained in a short day length (6 L : 18 D cycle) for 24 weeks
before they were released into DD, a length of time known to
render hamsters unresponsive to the seasonal effects of short
day lengths ( photorefractory); ETV was recorded every four
weeks to track reproductive state and the development of photorefractoriness. For ETV measures, hamsters were anaesthetized
with isoflurane vapours, and the length and width of the testis
were measured externally using callipers. ETV was calculated
as the product of the length width2 [43]. For half of the hamsters, lights went off at 10 : 00 local time; for the remaining
half, lights went off at 22 : 00. After 24 weeks, all hamsters
were transferred to DD for nine weeks before being paired—
these procedures ensured a spread of activity onsets and thus a
variety of phase relations between individuals at the time of pairing. When placed in DD, each hamster was provided with a
running wheel (diameter ¼ 7 inches). Temperature iButtons
were surgically implanted after three to five weeks in DD and
programmed to begin recording Tb after a two to four week
recovery period. The Tb rhythm of each hamster was recorded
for a total of 85 days in DD, representing 13 – 14 days before pairing, 62 – 63 days of cohabitation with a single running wheel and
9 – 10 days following separation. Hamsters were then sacrificed
and iButtons removed and interrogated.
(b) Pairing and behavioural observations
At pairing, hamsters were weighed, and pairs placed together in
a fresh cage. Pairs continued to have access to a single running
wheel, which was previously occupied by one hamster of the
pair; prior wheel ‘ownership’ was noted but did not influence
the results of the experiment. During the first 30 min of cohabitation, hamsters were observed under dim red light for any
aggressive or dominance behaviours. At the end of the experiment (9– 10 days after separation), body mass and ETV were
recorded, and hamster pairs were reunited under dim red light
for 30 min for a second behavioural observation test. To discriminate the behaviour of the individuals within each pair, one
hamster was given an ear tag on each ear. For each hamster, a
dominance score was calculated by summing the number of
dominance behaviours (attacks and chases) and subtracting
the number of subordinate behaviours (flees and on-back postures). Dominance was assigned only in pairs for which one
hamster had a positive score (dominant hamster) and the other
a negative score (subordinate hamster) for at least one of the
observation tests. While dominance was sometimes indeterminate in one of the two tests, dominance status was never
reversed in the second observation test. During cohabitation,
hamsters were checked regularly (initially once per day for the
first week, then approximately three times per week thereafter)
for behavioural interactions and signs of physical discomfort.
Single-housed controls were treated identically to paired hamsters, except that they were not provided with a cagemate.
Nonetheless, timing of daily care, behavioural observations and
recording of experimental measures of each single-housed
(c) Implantation of iButtons
Prior to surgery, iButton temperature dataloggers (Dallas
Semiconductor DS1922, Maxim Integrated Products, Inc.,
Sunnyvale, CA, USA; height, radius and weight ¼ 5.89 mm,
17.35 mm and 2.95 g, respectively) were coated in paraffin/elvax
wax (Mini-Mitter Co., Inc., Sunriver, OR, USA) and sterilized by
sequential soak in povidine iodine solution (overnight) followed
by a sterile saline rinse. On the day of surgery, hamsters were
anaesthetized with ketamine (100– 150 mg kg21, IP) and xylazine
(10 mg kg21, IP). A small patch of fur was shaved on the lower
abdominal surface and the area wiped clean with povidoneiodine solution. Incisions in the skin and abdominal wall, just
large enough to allow passage of the iButton, were made within
this shaved area, and the iButton was placed within the peritoneal
cavity. The abdominal wall and skin were closed using sterile
sutures and wound clips, respectively, and the wound treated
with antibiotic ointment (Neomycin and Polymyxin B Sulfates
and Bacitracin Zinc Ointment USP).
(d) Data collection and analysis
Body temperature was recorded every 15 min by a digital thermometer encased within the iButton. Temperature records for
each hamster were converted to .awd files and plotted as temperature actograms using ACTIVIEW software (Mini-Mitter Co.,
Inc.). These actograms were thresholded using each hamster’s
mean body temperature value for the entire 85-day record.
Continuous wavelet transforms were used to determine cycleto-cycle peak circadian amplitudes (Morlet function) and phases
of rhythm onset and offset (Mexican Hat function), as we have
described previously [16]. For each hamster, we calculated the
cycle-to-cycle phase difference between rhythm onset and a
reference marker (04.00 local time) and plotted the evolution of
this phase difference to obtain the phase trajectory. The actual
phase trajectory was compared to the predicted phase trajectory
with 95% prediction interval, as extrapolated from the linear
regression of onset phases from the 13 to 14 days before cohabitation. Change of t was determined by subtracting the slope of
the pre-cohabitation regression line from that of the lines calculated during the 10 days immediately before or after separation.
Precision of t and amplitude were calculated from the cycleto-cycle variation in Mexican Hat-determined onsets and
Morlet-determined circadian peaks. Alpha was determined for
each cycle as the difference between rhythm onset and offset.
All comparisons of actual pairs versus virtual pairs, affected
hamsters versus non-affected hamsters, and affected pairs
versus non-affected pairs were conducted using the Student’s
t-test (for means) or x 2 analysis (for proportions). One-way
ANOVA was used to compare means of affected hamsters, nonaffected hamsters and virtually paired hamsters. All statistical
tests were conducted using STATVIEW 5.0.1 (SAS Institute Inc.,
Cary, NC, USA).
All procedures were in accordance with the NIH Guide for the Care
and Use of Laboratory Animals and approved by the Institutional
Animal Care and Use Committee of the University of Massachusetts
Medical School.
Acknowledgements. We would like to thank T.R. Virgil for thoughtful
discussions. The contents of this report are solely the responsibility
of the authors and do not necessarily represent the official views of
the NIH.
Data accessibility. Raw Tb data are available from the Dryad Digital
Repository at (http://doi.org/10.5061/dryad.6c45t).
Funding statement. Supported by NINDS F32 NS058135 (M.J.P.) and
NIGMS R01 GM094109 (W.J.S.).
7
Proc. R. Soc. B 281: 20132535
(a) Experimental procedures
hamster were tethered (in time) to another single-housed
hamster to form a ‘virtual’ pair.
rspb.royalsocietypublishing.org
cohoused hamsters were excluded from the data analyses. In one
pair, a hamster displayed signs of sickness during the final weeks
of the study (weight loss, hunched posture, decreased body
temperature and low levels of locomotor activity). In the other,
several weeks of body temperature data of one hamster were
lost due to an iButton failure. Thus, the total number of hamsters
analysed was 64. Upon arrival, hamsters were housed singly in
plastic cages (1900 10.500 800 ) contained within well-ventilated,
light-proof environmental compartments with food and water
available ad libitum.
Downloaded from http://rspb.royalsocietypublishing.org/ on June 14, 2017
References
2.
3.
5.
6.
7.
8.
9.
10.
11.
12.
13.
14.
15. Winfree AT. 2001 The geometry of biological time.
New York, NY: Springer.
16. Leise TL, Indic P, Paul MJ, Schwartz WJ. 2013
Wavelet meets actogram. J. Biol. Rhythms 28,
62 –68. (doi:10.1177/0748730412468693)
17. Siegel HI. 1985 Aggressive behavior. In The hamster:
reproduction and behavior. New York, NY:
Plenum Press.
18. Menaker M, Eskin A. 1966 Entrainment of circadian
rhythms by sound in Passer domesticus. Science
154, 1579–1581. (doi:10.1126/science.154.3756.
1579)
19. Paul MJ, Schwartz WJ. 2007 On the chronobiology
of cohabitation. Cold Spring Harb. Symp. Quant.
Biol. 72, 615 –621. (doi:10.1101/sqb.2007.72.042)
20. Crowley M, Bovet J. 1980 Social synchronization of
circadian rhythms in deer mice (Peromyscus
maniculatus). Behav. Ecol. Sociobiol. 7, 99 – 105.
(doi:10.1007/BF00299514)
21. Bovet J. 1972 On the social behavior in a stable
group of long-tailed field mice (Apodemus
sylvaticus). II. Its relations with distribution of daily
activity. Behaviour 41, 55 –67. (doi:10.1163/
156853972X00194)
22. Huhman KL, Moore TO, Ferris CF, Mougey EH,
Meyerhoff JL. 1991 Acute and repeated exposure to
social conflict in male golden hamsters: increases in
plasma POMC-peptides and cortisol and decreases
in plasma testosterone. Horm. Behav. 25, 206– 216.
(doi:10.1016/0018-506X(91)90051-I)
23. Morin LP, Cummings LA. 1981 Effect of surgical or
photoperiodic castration, testosterone replacement
or pinealectomy on male hamster running
rhythmicity. Physiol. Behav. 26, 825–838. (doi:10.
1016/0031-9384(81)90106-2)
24. Jechura TJ, Walsh JM, Lee TM. 2003 Testosterone
suppresses circadian responsiveness to social cues in
the diurnal rodent Octodon degus. J. Biol. Rhythms
18, 43 –50. (doi:10.1177/0748730402239675)
25. Paul MJ, Zucker I, Schwartz WJ. 2008 Tracking the
seasons: the internal calendars of vertebrates. Phil.
Trans. R. Soc. B 363, 341–361. (doi:10.1098/rstb.
2007.2143)
26. Garrett JW, Campbell CS. 1980 Changes in social
behavior of the male golden hamster accompanying
photoperiodic changes in reproduction. Horm.
Behav. 14, 303 –318. (doi:10.1016/0018506X(80)90020-3)
27. Jasnow AM, Huhman KL, Bartness TJ, Demas GE.
2002 Short days and exogenous melatonin increase
aggression of male Syrian hamsters (Mesocricetus
auratus). Horm. Behav. 42, 13 –20. (doi:10.1006/
hbeh.2002.1797)
28. Sherwin CM. 1998 Voluntary wheel running: a
review and novel interpretation. Anim. Behav. 56,
11 –27. (doi:10.1006/anbe.1998.0836)
29. Pratt BL, Goldman BD. 1986 Environmental
influences upon circadian periodicity of Syrian
hamsters. Physiol. Behav. 36, 91– 95. (doi:10.1016/
0031-9384(86)90079-X)
30. Mrosovsky N. 1999 Further experiments on the
relationship between the period of circadian
rhythms and locomotor activity levels in hamsters.
Physiol. Behav. 66, 797–801. (doi:10.1016/S00319384(99)00022-0)
31. Morin LP. 1999 Serotonin and the regulation of
mammalian circadian rhythmicity. Ann. Med. 31,
12– 33. (doi:10.3109/07853899909019259)
32. Yannielli P, Harrington ME. 2004 Let there be ‘more’
light: enhancement of light actions on the circadian
system through non-photic pathways. Prog.
Neurobiol. 74, 59 –76. (doi:10.1016/j.pneurobio.
2004.06.001)
33. Harrington ME. 1997 The ventral lateral geniculate
nucleus and the intergeniculate leaflet: interrelated
structures in the visual and circadian systems.
Neurosci. Biobehav. Rev. 21, 705– 727. (doi:10.
1016/S0149-7634(96)00019-X)
34. Bovet J, Oertli EF. 1974 Free-running circadian
activity rhythms in free-living beaver (Castor
canadensis). J. Comp. Physiol. 92, 1– 10. (doi:10.
1007/BF00696522)
35. Frisch B, Koeniger N. 1994 Social synchronization of
the activity rhythms of honeybees within a colony.
Behav. Ecol. Sociobiol. 35, 91– 98. (doi:10.1007/
BF00171498)
36. Shkolnik A. 1971 Diurnal activity in a small desert
rodent. Int. J. Biometeorol. 15, 115–120. (doi:10.
1007/BF01803884)
37. Glass GE, Slade NA. 1980 The effect of Sigmodon
Hispidus on spatial and temporal activity of Microtus
ochrogaster: evidence for competition. Ecology 61,
358–370. (doi:10.2307/1935194)
38. Ziv Y, Abramsky Z, Kotler BP, Subach A. 1993
Interference competition and temporal and habitat
partitioning in two gerbil species. Oikos 66,
237–246. (doi:10.2307/3544810)
39. Harrington LA, Harrington AL, Yamaguchi N, Thom
MD, Ferreras P, Windham TR, MacDonald DW. 2009
The impact of native competitors on an alien
invasive: temporal niche shifts to avoid interspecific
aggression? Ecology 90, 1207– 1216. (doi:10.1890/
08-0302.1)
40. Fenn MGP, MacDonald DW. 1995 Use of middens by
red foxes: risk reverses rhythms of rats. J. Mammal.
76, 130 –136. (doi:10.2307/1382321)
41. Fraser DF, Gilliam JF, Akkara JT, Albanese BW, Snider
SB. 2004 Night feeding by guppies under predator
release: effects on growth and daytime courtship.
Ecology 85, 312–319. (doi:10.1890/03-3023)
42. Swarts HM, Crooks KR, Willits N, Woodroffe R. 2009
Possible contemporary evolution in an endangered
species, the Santa Cruz Island fox. Anim. Conserv.
12, 120 –127. (doi:10.1111/j.1469-1795.2008.
00229.x)
43. Gorman MR. 1995 Seasonal adaptations of Siberian
hamsters. I. Accelerated gonadal and somatic
development in increasing versus static long day
lengths. Biol. Reprod. 53, 110–115. (doi:10.1095/
biolreprod53.1.110)
Proc. R. Soc. B 281: 20132535
4.
Davidson AJ, Menaker M. 2003 Birds of a feather
clock together—sometimes: social synchronization
of circadian rhythms. Curr. Opin. Neurobiol. 13,
765–769. (doi:10.1016/j.conb.2003.10.011)
Mistlberger RE, Skene DJ. 2004 Social influences on
mammalian circadian rhythms: animal and human
studies. Biol. Rev. 79, 533–556. (doi:10.1017/
S1464793103006353)
Favreau A, Richard-Yris M-A, Bertin A, Houdelier C,
Lumineau S. 2009 Social influences on circadian
behavioural rhythms in vertebrates. Anim. Behav.
77, 983–989. (doi:10.1016/j.anbehav.2009.01.004)
Castillo-Ruiz A, Paul MJ, Schwartz WJ. 2012 In
search of a temporal niche: social interactions. Prog.
Brain Res. 199, 267–280. (doi:10.1016/B978-0444-59427-3.00016-2)
Mrosovsky N. 1996 Locomotor activity and nonphotic influences on circadian clocks. Biol. Rev. 71,
343 – 372. (doi:10.1111/j.1469-185X.1996.
tb01278.x)
Albers HE, Huhman KL, Meisel RL. 2002 Hormonal
basis of social conflict and communication. In
Hormones, brain, and behavior (eds DW Pfaff,
AP Arnold, AM Etgen, SE Fahrbach, RT Rubin),
pp. 393 –433. San Diego, CA: Academic Press.
Huhman KL. 2006 Social conflict models: can they
inform us about human psychopathology? Horm.
Behav. 50, 640 – 646. (doi:10.1016/j.yhbeh.2006.
06.022)
Davis FC, Stice S, Menaker M. 1987 Activity and
reproductive state in the hamster: independent
control by social stimuli and a circadian pacemaker.
Physiol. Behav. 40, 583– 590. (doi:10.1016/00319384(87)90101-6)
Gattermann R, Weinandy R. 1997 Lack of social
entrainment of circadian activity rhythms in the
solitary golden hamster and in the highly social
Mongolian gerbil. Biol. Rhythm Res. 28, 85 – 93.
(doi:10.1076/brhm.28.3.5.85.13122)
Honrado GI, Mrosovsky N. 1989 Arousal by sexual
stimuli accelerates the re-entrainment of hamsters
to phase advanced light-dark cycles. Behav.
Ecol. Sociobiol. 25, 57 – 63. (doi:10.1007/
BF00299711)
Aschoff J, Von Goetz C. 1988 Masking of circadian
activity rhythms in male golden hamsters by the
presence of females. Behav. Ecol. Sociobiol. 22,
409–412. (doi:10.1007/BF00294978)
Mrosovsky N. 1988 Phase response curves for social
entrainment. J. Comp. Physiol. A 162, 35 –46.
(doi:10.1007/BF01342701)
Refinetti R, Nelson DE, Menaker M. 1992 Social
stimuli fail to act as entraining agents of circadian
rhythms in the golden hamster. J. Comp. Physiol. A
170, 181–187. (doi:10.1007/BF00196900)
Mistlberger RE, Antle MC, Webb IC, Jones M,
Weinberg J, Pollock MS. 2003 Circadian clock
resetting by arousal in Syrian hamsters: the role of
stress and activity. Am. J. Physiol. Regul. Integr.
Comp. Physiol. 285, R917 –R925.
rspb.royalsocietypublishing.org
1.
8