Figure S1; Figure S2; Figure S3; Figure S4; Figure S5

Full text

1
Supplementary Materials
Additional detail for statistics
Statistical analyses to compare the effects of experimental periods were performed in R 3.2.5. (R
Core Team (2016). R: A language and environment for statistical computing. R Foundation for
Statistical Computing, Vienna, Austria. URL https://www.R-project.org/)
One-way repeated measures analysis of variance (rmANOVA) was used to examine the
differences across experimental periods given the repeated nature of measurements. rmANOVA
models were specified using the lme function of the ‘nlme’ package (Pinheiro J, Bates D,
DebRoy S, Sarkar D and R Core Team (2016). _nlme: Linear and Nonlinear Mixed Effects
Models_. R package version 3.1-128). Models were specified to examine the effect of
experimental period (the within-subjects factor) on the dependent variables of interest: ethanol
consumption, ethanol preference, transition matrix indices (Pstay), local statistics for locomotor
data (mean, skewness, variance), activity-rest parameters(γ), wavelet band power data (circadian
and relative ultradian) and 2D entropy data. Rats were specified as subjects, and a random
statement where periods were nested in rats was included. For ethanol consumption and ethanol
preference, DEPwk1 and DEPwk2 were not included because bottles contained no alcohol. For
Pstay, DEPwk1 was included for comparison of stability (while DEPwk2 was also calculated, a low
number of accesses/transitions, [e.g. no accesses or only single accesses during the week]
resulted in the index being uninformative- we refrain from making any interpretation). For all
other measures, all experimental periods were analysed. (To confirm that the assumption of
sphericity was not violated, we used the ezANOVA function of the ‘ez’ package).
For all models, whenever significant effects of experimental period were found by the
rmANOVA (significance level of p <0.05), it was necessary to identify where the differences lay.
Although we were generally interested in comparison of BASE to other periods, these
differences were already expected given prior knowledge about the paradigm. As such, post-hoc
multiple comparisons were performed with the Tukey contrasts to search for differences across
all periods. Tukey contrasts were calculated using the glht function of the ‘multcomp’ package
(Torsten Hothorn, Frank Bretz and Peter Westfall (2008). Simultaneous Inference in General
Parametric Models. Biometrical Journal 50(3), 346--363).

a
b
day
a
Wet r
r
u
o
h/
g
k
/
g[ ]
t
EH
O%
5
r
u
o
h
/
g
k
/
g[ ]
t
EH
O%
0
1
r
u
o
h
/
g
k/
g
[]
t
EH
O%
0
2
r
u
o
h
/
g
k/
g[ ]
t
EH
Oo
Ta
tl
r
u
o
h/
g
k
/
g[ ]
A
L
r
u
o
h/
s
t
n
u
o
c
[]
BA S E DE P
wk 1
DE P
wk 2
ER
wk 1
ER
wk 2
ER
wk 3
ER
wk 4
1.0
0.5
0.0
9080706050
1.0
0.5
0.0
1.0
0.5
0.0
1.0
0.5
0.0
2000
1000
0
10
0
H2O H2O H2O H2O
H2OEtOH EtOH
5% 10%
EtOH
20%
H2OEtOH EtOH
5% 10%
EtOH
20%
ER
wk1
ER
wk2
ER
wk3
ER
wk4
DEP
wk1
DEP
wk2
BASE
BASE
Figure S1

Figure S1. Alcohol deprivation effect (ADE) paradigm and sample intensive longitudinal data
(ILD). (a) Schematic of the ADE paradigm: baseline (BASE) followed by two weeks of
deprivation (DEPwk1 and DEPwk2) and four weeks of ethanol (EtOH) reintroduction (ERwk1-4),
and time schedule. (b) Drinking (Water, 5%, 10%, 20%, total EtOH) and locomotor activity
(LA) traces for a sample rat during experimental periods as an example of ILD. NB: During
deprivation (red shaded area) all bottles contained water only; accesses shown then are water and
not EtOH.

BASELINE DEP ER
BASELINE DEP ER
BASELINE DEP ER
1.4
1.2
1.0
0.8
0.6
0.4
0.2
0.0
EtOH 20%
100806040200
Day
1.6
1.2
0.8
0.4
0.0
EtOH 10%
100806040200
Day
BASELINE DEP ER
18
16
14
12
10
8
Preference
100806040200
Day
1.5
1.0
0.5
0.0
-0.5
EtOH 5%
100806040200
Day
a
b
c
d
Figure S2

Figure S2. Average daily consumption of a) 5%, b) 10%, and c) 20% solutions, and d) average
preference during the course of the experiment, demonstrating the alcohol deprivation effect.
Regression lines showing projected trajectories and were fitted using the curves for averages
from day 5-56 (until deprivation). Error bars denote standard deviation. Data from days 1-5 was
lost due to sensor calibration issues.

15
0
10
5
counts2/day
BASE DEPwk1 DEPwk2 ERwk1 ER wk2 E R wk3 E R wk4
Experimental Period
Animal #
2
3
4
5
6
7
9
10
11
13
14
15
16
17
18
19
21
22
23
24
1
cycles/day
10
40
1
Figure S3

Figure S3. Wavelet plots for all animals showing the distribution of power of the locomotor
activity signal across different frequencies and how it changed over time (experimental periods).
Values of moduli of the wavelet coefficients (powers) are colour coded according to their
magnitude (blue indicates a low and red a large value), and the ordinates are represented on a
logarithmic scale. Increased power at low frequencies was observed in DEPwk1, showing
instability in ultradian rhythms and suggestive of the system being near a tipping point.

Animal #
2
3
4
5
6
7
9
10
11
13
14
15
16
17
18
19
21
22
23
24
1
probability
density
0.015
0
0.010
0.005
BASE DEPwk1 DEPwk2 ER wk1 E R wk2 E Rwk3 E R wk4
Experimental Period
BASE DEPwk1 DEPwk2 ER wk1 E R wk2 E Rwk3 ER wk4
x(t)
018
18
x(t-τ)
Figure S4

Figure S4. Limit cycle trajectories (left) and probability density maps (right) of rat locomotor
activity over experimental periods. In BASE, plots show a large clear circle indicating stable
circadian rhythms. In DEPwk1 trajectories become diffuse suggesting instability. From circles
subsequently decrease in size, and once again become clear circles, suggestive of a stabilisation
into a new state (DEPwk2 – ERwk4).

0
10
20
30
40
BASE
DEPwk1
DEPwk2
ER wk1
ER wk2
ER wk3
ER wk4
Radius
2
Figure S5

Figure S5. Moment of inertia about centre of mass (average squared radius) for limit cycles
shown in figure S3 confirms the decrease in the size of the cycle across experimental periods.
Note that the squared radius in DEPwk1 is not significantly smaller than that in BASE, different
from the observation for the circadian power (figure 5d), presumably because this quantity
include increased ultradian power in DEPwk1 (figure 5c). Error bars show S.E.M. *** p < 0.001;
** p < 0.01; * p < 0.05.

Changes in drinking behaviour described by transitions between solutions over experimental
periods.
Period
Average Transitions
H2O
5%
10%
20%
H2O
622.14
15.62
18.10
12.14
BASE
5%
16.86
10.81
2.14
1.48
10%
17.76
1.43
8.48
1.67
20%
13.19
1.95
1.43
3.81
H2O
608.57
26.14
23.52
10.00
DEPwk1
5%
25.14
8.95
10.95
4.05
10%
25.05
3.90
8.38
4.67
20%
11.76
3.38
3.90
2.05
H2O
579.10
6.90
6.95
3.90
DEPwk2
5%
7.43
0.57
0.57
1.43
10%
7.86
0.71
1.33
0.71
20%
4.90
1.33
0.71
0.19*
H2O
474.62
25.90
22.81
27.05
ERwk1
5%
24.62
24.38
7.52
7.90
10%
23.43
8.19
15.57
7.81
20%
26.29
7.81
8.19
17.24
H2O
496.10
13.86
15.10
15.10
ERwk2
5%
14.29
12.81
2.43
1.86
10%
15.33
1.86
9.76
2.00
20%
15.95
2.10
1.86
6.57
H2O
545.38
12.90
12.57
14.81
ERwk3
5%
13.05
4.67
0.95
0.95
10%
13.57
1.05
6.95
0.81
20%
15.48
0.86
1.05
6.14
H2O
422.57
11.48
9.67
17.29
ERwk4
5%
11.71
5.95
1.81
3.10
10%
9.90
3.71
5.24
3.71
20%
16.76
3.33
3.71
9.76

Table S1. Group average number of transitions per period for Water, 5%, 10% and 20% alcohol
solutions. Bold numbers indicate stays with the same alcohol solution. Staying (stable) is
observed in BASE with initial preference for weaker solutions, followed by increased switching
during DEPwk1. In DEPwk2 greatly decreased amounts of accesses were observed. (*During
DEPwk2 two rats switched to the 20% bottle as their main water bottle- these accesses were
interpreted as water accesses). From ERwk1- ERwk4 staying behaviour reappeared accompanied
by a shift in preference, with 20% having the most stays.