Content uploaded by Tatiana S. Leite
Author content
All content in this area was uploaded by Tatiana S. Leite on Mar 26, 2021
Content may be subject to copyright.
Content uploaded by Sylvia Medeiros
Author content
All content in this area was uploaded by Sylvia Medeiros on Mar 26, 2021
Content may be subject to copyright.
iScience
Article
Cyclic alternation of quiet and active sleep states in
the octopus
Sylvia Lima de
Souza Medeiros,
Mizziara Marlen
Matias de Paiva,
Paulo Henrique
Lopes, ..., Sandro
de Souza, Tatiana
Silva Leite, Sidarta
Ribeiro
sidartaribeiro@neuro.ufrn.br
HIGHLIGHTS
Octopus has ‘Quiet’ and
‘Active sleep’, with
different episode duration
and periodicity
States differ on arousal
thresholds, skin color and
texture, and eye and
mantle movement
The results suggest that
octopus has a sleep cycle
analogous to that of
amniotes
Medeiros et al., iScience --,
102223
--,2021ª2021 The
Authors.
https://doi.org/10.1016/
j.isci.2021.102223
ll
OPEN ACCESS
iScience
Article
Cyclic alternation of quiet
and active sleep states in the octopus
Sylvia Lima de Souza Medeiros,
1,2,3
Mizziara Marlen Matias de Paiva,
1,3
Paulo Henrique Lopes,
4,7
Wilfredo Blanco,
4,5,7
Franc¸oise Dantas de Lima,
6
Jaime Bruno Cirne de Oliveira,
1
Ina
´cio Gomes Medeiros,
1,7
Eduardo Bouth Sequerra,
1
Sandro de Souza,
1,5,7
Tatiana Silva Leite,
6
and Sidarta Ribeiro
1,2,3,8,
*
SUMMARY
Previous observations suggest the existence of ‘Active sleep’ in cephalopods. To
investigate in detail the behavioral structure of cephalopod sleep, we video-re-
corded four adult specimens of Octopus insularis and quantified their distinct
states and transitions. Changes in skin color and texture and movements of eyes
and mantle were assessed using automated image processing tools, and arousal
threshold was measured using sensory stimulation. Two distinct states unrespon-
sive to stimulation occurred in tandem. The first was a ‘Quiet sleep’ state with uni-
formly pale skin, closed pupils, and long episode durations (median 415.2 s). The
second was an ‘Active sleep’ state with dynamic skin patterns of color and texture,
rapid eye movements, and short episode durations (median 40.8 s). ‘Active sleep’
was periodic (60% of recurrences between 26 and 39 min) and occurred mostly af-
ter ‘Quiet sleep’ (82% of transitions). These results suggest that cephalopods have
an ultradian sleep cycle analogous to that of amniotes.
INTRODUCTION
Sleep is a well-studied behavior in amniotes such as mammals (Vanderwolf, 1969;Timo-Iaria et al., 1970;
Delorme et al., 1964), birds (Low et al., 2008;Ayala-Guerrero et al., 1988), and some reptiles (Shein-Idelson
et al., 2016;Libourel et al., 2018;Norimoto et al., 2020). Electrophysiological recordings in amniotes show
distinct spectral profiles that comprise two major alternating sleep states, one quiet and another active
(Gervasoni et al., 2004;Noda et al., 1969;Steriade et al., 1993). Much less is known about neurobiological
rhythms in invertebrates because electrophysiologi cal recordings remain very challenging in these animals,
due to technical difficulties caused by a soft body, a rigid carapace, or life in the aquatic environment (Hen-
dricks et al., 2000).
Despite these limitations, the study of invertebrate sleep has advanced using behavioral criteria originally
developed to investigate mammalian sleep. These criteria comprise stereotyped or species-specific pos-
tures, maintenance of behavioral quiescence, elevated arousal threshold, state reversibility by sensory
stimulation, responsiveness, and homeostatic regulation able to cause sleep rebound after deprivation
(Hendricks et al., 2000;Campbell and Toblew, 1984;Greenspan et al., 2001;Meisel et al., 2011).
Among cephalopods, Octopus vulgaris (Cuvier, 1797) (Octopodidae: Cephalopod) meets all the criteria
to define sleep (Meisel et al., 2011). For example, Octopus vulgaris specimens choose a preferred resting
place and assume a typical posture of head lowered, arms curled around the body, motionless body
except for sporadic shrinking and rapid, random movement of the suckers, pale body color, narrowed
or completely closed eye pupils, and reduced ventilation rate (Meisel et al., 2011). Quiescent animals
exposed to vibratory stimulation showed an elevated arousal threshold and state reversibility after
intense stimulation. Furthermore, deprivation of the quiescence state led to a rest rebound (Meisel
et al., 2011).
The sleep criteria mentioned earlier, except for the increase in the arousal threshold during quiescence,
have also been met by Sepia officinalis (Linnaeus, 1758;Frank et al., 2012;Iglesias et al., 2019). Behavioral
observations suggest the existence of a sleep state in which the animal is partially buried in the substrate
with pupils closed and another sleep state with rapid chromatophore changes accompanied by
1
Brain Institute, Federal
University of Rio Grande do
Norte, Natal, Rio Grande do
Norte, Brazil
2
Graduate Program in
Psychobiology, Federal
University of Rio Grande do
Norte, Natal, Rio Grande do
Norte, Brazil
3
Graduate Program in
Neuroscience, Federal
University of Rio Grande do
Norte, Natal, Rio Grande do
Norte, Brazil
4
Computer Science
Department, State University
of Rio Grande do Norte,
Natal, Rio Grande do Norte,
Brazil
5
Bioinformatics
Multidisciplinary
Environment, Federal
University of Rio Grande do
Norte, Natal, Rio Grande do
Norte, Brazil
6
Department of Ecology and
Zoology, Federal University
of Santa Catarina,
Floriano
´polis, Santa Catarina,
Brazil
7
Graduate Program in
Bioinformatics, Federal
University of Rio Grande do
Norte, Natal, Rio Grande do
Norte, Brazil
8
Lead contact
*Correspondence:
sidartaribeiro@neuro.ufrn.br
https://doi.org/10.1016/j.isci.
2021.102223
iScience --, 102223, --, 2021 ª2021 The Authors.
This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/). 1
ll
OPEN ACCESS
Please cite this article in press as: Medeiros et al., Cyclic alternation of quiet and active sleep states in the octopus, iScience (2021), https://
doi.org/10.1016/j.isci.2021.102223
skin-texture changes, rapid eye movements, and arm twitching. Altogether, the behavioral evidence points
to the existence of an ultradian rhythm (Frank et al., 2012;Iglesias et al., 2019).
Despite the mounting evidence that cephalopods display a Quiet/Active sleep cycleakin to amniotes, one
can still argue that perhaps the specimens were not asleep but rather in a state of quiet alertness. This is a
possibility because (1) responsiveness to stimulation was not tested in Sepia officinalis and (2) the studies of
Octopus vulgaris investigated the quiescence state with pale body pattern, but not the ‘Active sleep’ state.
Here we set out to address these gaps through a comprehensive behavioral quantification of all the sleep
and waking states observed in Octopus insularis (Leite and Haimovici, 2008). Based on previous research
(Meisel et al., 2011;Frank et al., 2012;Iglesias et al., 2019), we hypothesized that this species displays at
least two consecutive quiescence states unresponsive to stimulation, comprising ‘Quiet sleep’ and ‘Active
sleep’ states. To test this hypothesis, animals were video-recorded and systematically exposed to sensory
stimuli during each state of interest to assess potential differences in the arousal threshold across the wake-
sleep cycle (Figure 1).
RESULTS
Ethogram analysis reveals different quiescence states
The behaviors scored during the day were ‘Active’ state, ‘Alert’ state and five variations of the quiescence
state, categorized as Quiet with open pupils (‘QOP’) (Figure 2A; Video S1), Quiet with closed pupil (‘Quiet
sleep’) (Figure 2B; Video S2), Quiet with the ‘half and half’ skin pattern (‘QHH’) (Figure 2C; Video S3), Quiet
with dynamic body pattern and movement of the eyes (‘Active sleep’) (Figure 2D; Video S4), and Quiet with
only one eye movement (‘QOEM’) (Figure 2E; Video S5)(Table 1). For the behaviors recorded during the
night, we observed the following types of quiescence states: ‘QOP’ (Video S6), ‘Quiet sleep’ (Video S7),
‘Active sleep’ (Video S8), and ‘QOEM’ (Video S9).
Among the quiescent states, ‘QHH’, ‘Quiet sleep’, and ‘Active sleep’ resembled sleep behavior. To
further investigate this resemblance we performed the arousal threshold test during these states and
during the ‘Alert’ state, which allowed for a comparison of differences in responsiveness to sensory
stimulation.
Figure 1. Experimental scheme
Animals were first subjected to acclimatization, followed by behavioral video recordings to assess the wake-sleep cycle
and then a protracted period, during which the arousal threshold was measured. Visual stimulation was performed in
three animals (octopuses 2, 3, and 4), whereas the vibratory stimulation test was performed in only one animal (octopus 3),
which initially did not respond to the visual stimulation.
ll
OPEN ACCESS
2iScience --, 102223, --, 2021
Please cite this article in press as: Medeiros et al., Cyclic alternation of quiet and active sleep states in the octopus, iScience (2021), https://
doi.org/10.1016/j.isci.2021.102223
iScienc
e
Article
Figure 2. Behaviors observed during the different quiescence states
A total of five states were observed during quiescence: (A) Quiet state with open pupils; (B) ‘Quiet sleep’ with closed pupils; (C) Quiet with half and half skin
pattern; (D) ‘Active sleep’ (quiet with dynamic pattern and eye movement); (E) Quiet with the movement of only one eye. See descriptions of each state in Table 1.
ll
OPEN ACCESS
iScience --, 102223, --, 2021 3
Please cite this article in press as: Medeiros et al., Cyclic alternation of quiet and active sleep states in the octopus, iScience (2021), https://
doi.org/10.1016/j.isci.2021.102223
iScienc
e
Article
Measuring arousal threshold
Visual stimulation test
The arousal response elicited by visual stimulation was analyzed in only three animals, because a few days
after we finished the wake-sleep recordings of octopus 1, this animal began to show signs of decrease in
health condition, such as less interest for food, so we excluded it from the arousal threshold tests.
These tests were performed when the octopuses were in the states ‘Alert’, Half and half (‘QHH’), ‘Quiet
sleep’ and ‘Active sleep’. The behavioral states showed significant differences in time spent to react to
the visual stimulation (Kruskal-Wallis p = 1.30 310
12
). The highest latency was observed in the ‘Active
sleep’ state (median 32.0 s, first quartile 22.5 s, third quartile 38.0 s), followed by the ‘Quiet sleep’ (median
6.0 s, first quartile 4.0 s, third quartile 9.0 s), ‘QHH’ (median 6.0 s, first quartile 4.0 s, third quartile 8.0 s), and
finally the ‘Alert’ state with the shortest latency (median 4.0 s, first quartile 2.0 s, third quartile 6.0 s) (Fig-
ure 3A, upper panel).
The ‘Active sleep’ had the lowest number of stimulation trials with any reaction (highest number of stimuli
classified as ‘‘absence of reaction’’), followed by ‘Quiet sleep’, ‘QHH’ and the ‘Alert’ state (Figure 3A,
bottom panel). Besides, the chi-square pairwise comparison showed significant differences
regarding the absence of reaction between the states: ‘Quiet sleep’ and ‘Alert’ state (chi-squared test,
p = 6.11 310
10
), ‘Quiet sleep’ and ‘QHH’ (chi-squared test, p = 3.51 310
4
), ‘Alert state’ and ‘QHH’
(chi-squared test, p = 1.39 310
2
), ‘Alert state’ and ‘Active sleep’ (chi-squared test, p = 2.20 310
16
),
‘Active sleep’ and ‘QHH’ (chi-squared test, p = 1.02 310
8
). The only states that did not differ significantly
from each other were ‘Active sleep’ and ‘Quiet sleep’ (chi-squared test, p = 0.06).
Vibratory stimulation test
The vibratory stimulation test was performed with one octopus (n = 1), and the results showed (Figure 3B)
that in the ‘Alert’ and ‘QHH’ states, most responses were recorded at level 1. The ‘Quiet sleep’ state
showed slower reactions, with a predominance of levels 2 and 4. The slowest reactions occurred during
the ‘Active sleep’ state: during most trials, the animal responded only at level 5 or showed complete
absence of reaction. The Pearson’s chi-squared test showed that there were significant differences in
response time across the behaviors (p = 1.93 310
4
). There were more responses at levels 4 and 5 or
‘‘absence of reaction’’ for ‘Quiet sleep’ and ‘Active sleep’ states than for ‘Alert’ and ‘QHH’ states. Although
the chi-squared pairwise-comparison showed that the ‘Active sleep’ responses differed significantly from
those sampled during the ‘Alert’ state (p = 5.52 310
5
)orQHH(p=1.70310
5
), responses during
‘Quiet sleep’ were not significantly different from those sampled within the ‘Alert’ state (p = 0.11) or
‘QHH’ (p = 0.08).
Table 1. Description of the different quiescence states found in Octopus insularis
States Description
Quiet with dynamic body pattern and eye
movements (ACTIVE SLEEP)
The animal dynamically changes the skin color and texture and
moves both eyes while contracting the suckers and the body,
with muscular twitches.
Quiet with closed pupil (QUIET SLEEP) Quiescence with the pupils of the eyes narrowed to a slit and
pale body color. The animal can also sporadically display
random movements of the suckers and arm tips without touching
the environmental surfaces. However, these movements are much
softer and slower than during the ‘Active sleep’ detailed above.
Quiet with open pupil (QOP) Quiescence with generally pale body color, head lowered and
motionless body, except for respiratory movements, and sporadic
shrinking.
Quiet with the half and half body
pattern (QHH)
The animal is head lowered and motionless and suddenly changes
the skin pattern, which is generally pale during this quiescence
state, to the ‘half and half’ skin pattern.
Quiet with only one eye movement (QOEM) A cyclic one eye movement behavior, appearing pupil contraction
and dilation together with exophthalmos.
ll
OPEN ACCESS
4iScience --, 102223, --, 2021
Please cite this article in press as: Medeiros et al., Cyclic alternation of quiet and active sleep states in the octopus, iScience (2021), https://
doi.org/10.1016/j.isci.2021.102223
iScienc
e
Article
Characterization of spontaneous behavioral alternation
The day-time behavioral states analyzed showed distinct distributions of episode durations: The ‘Active
sleep’ (median 40.8 s, first quartile 25.8 s, third quartile 52.2 s) and the ‘QHH’ (median 10.2 s, first quartile
7.2 s, third quartile 13.8 s) states displayed much shorter durations than the ‘Alert’ state (median 2.83 min,
first quartile 1.32 min, third quartile 6.9 min) and ‘Quiet sleep’ (median 6.92 min, first quartile 2.63 min, third
quartile 15.04 min) states (Figure 4A). The ‘Quiet sleep’ state was the one with the most variation in dura-
tion. Kruskal-Wallis tests followed by Wilcoxon pairwise comparisons were conducted to determine
whether there was any pattern of the duration of this state in relation to the states preceding or following
it. These analyses also aimed to investigate the relationship between ‘Active sleep’ and ‘Quiet sleep’ ep-
isodes. The states that immediately preceded the ‘Quiet sleep’ state were ‘Alert’, ‘QOP’, ‘QHH’, and
‘Active sleep’, and the durations of ‘Quiet sleep’ episodes preceded by ‘Active sleep’ episodes were signif-
icantly shorter (Kruskal-Wallis p = 4.91 310
7
)(Figure 5A).
The states immediately following the ‘Quiet sleep’ state were ‘Alert’, ‘QHH’, and ‘Active sleep’, and only
the duration of the ‘Quiet sleep’ episodes immediately followed by ‘Active sleep’ episodes differed
from the others by being significantly longer (Kruskal-Wallis p = 1.18 310
14
)(Figure 5B).
A
B
Percentage of the
absence of reaction
Visual test (n = 3 octopuses)
Vibration test (n = 1 octopus)
Alert Active Sleep
Quiet Sleep
Half and Half
Alert Active Sleep
Quiet Sleep
Half and Half
Latency of reactions (s)
Absolute frequency
Before
Impact
Low
Impact
Medium
Impact
Strong
Impact
Maximum
Impact
Maximum Impact with
absence of reaction
Figure 3. Reaction times after visual or vibratory stimulation
(A) The top panel shows the latency in seconds (y axis) of the animals’ reaction on visual test when they were in the
behaviors: ‘Alert’ state, ‘Quiet with half and half’ (QHH), ‘Quiet sleep’, and ‘Active sleep’ (x axis). The highest latency was
during ‘Active sleep’ and the lowest was during the ‘Alert state.’ The bottom panel shows the percentage of absence of
reaction for the same states. ‘Active sleep’ was the behavior with more absence of reaction, whereas the ‘Alert’ state had
less absence of reaction.
(B) Frequency of animal’s reaction on vibratory test (20 trials for each behavior). The lines represent the animals’ behaviors:
‘Alert’ (orange), ‘QHH’ (light blue), ‘Quiet sleep’ (dark blue), and ‘Active sleep’ (green). On the x axis, the category ‘‘Before
impact’’ comprises trials when the animal reacted even before the hammer’s impact, i.e. in response to the hammer’s
movement. The inclination angles were 10(low), 20(medium), 30(strong), and 40(maximum impact). When the animal
did not react even with the maximum impact, the trial was scored as ‘‘maximum Impact with absence of reaction.’’ The
‘Alert’ state and ‘QHH’ showed the highest frequency of responses ‘‘before impact’’ (10 times), and this value decreased
as the impact increased. The reaction frequency for these behaviors was below 2 times when the impact was maximum.
‘Quiet sleep’ showed little variation in frequency across different hammer impacts. Furthermore, it was more difficult to
evoke a reaction when the animal was in ‘Active sleep’, so that the reaction frequency value increased as the hammer’s
angle increased.
ll
OPEN ACCESS
iScience --, 102223, --, 2021 5
Please cite this article in press as: Medeiros et al., Cyclic alternation of quiet and active sleep states in the octopus, iScience (2021), https://
doi.org/10.1016/j.isci.2021.102223
iScienc
e
Article
The large variation in the duration of ‘Quiet sleep’ episodes prompted us to analyze more deeply the rela-
tionship between the durations of ‘Quiet sleep’ episodes and the likelihood of observing a neighboring
‘Active sleep’ episode. First, we looked for differences in the duration of ‘Quiet sleep’ episodes when
they immediately preceded or did not immediately precede ‘Active sleep’ episodes: using Kernel Density
Estimation, we plotted the durations distribution of the ‘Quiet sleep’ episodes immediately preceding
‘Active sleep’ episodes and of the ‘Quiet sleep’ episodes not immediately preceding ‘Active sleep’ epi-
sodes. These distributions show that the ‘Quiet sleep’ episodes that precede ‘Active sleep’ episodes
are generally longer than those that do not precede ‘Active sleep’ episodes. The Kolmogorov-Smirnov
test showed that these distributions were significantly different (p = 1.36 310
5
). Figure S1A suggests
that it is more likely that an octopus initiates an ‘Active sleep’ episode when they have been in a ‘Quiet
sleep’ episode for a long time period. For this reason, we further sorted the ‘Quiet sleep’ episodes in
two distinct categories: long and short. To classify these behaviors more accurately, we used a non-arbi-
trary parameter to split them into long and short ‘Quiet sleep’ episodes. Namely, we calculated the median
time between (1) the mode of the distribution of ‘Quiet sleep’ episodes that occurred immediately before
‘Active sleep’ episodes and (2) the mode of other ‘Quiet sleep’ episodes that did not occur immediately
before an ‘Active sleep’ episode (Figure S1A). In this way, short ‘Quiet sleep’ episodes had durations %
6.51 min, whereas long ‘Quiet sleep’ episodes lasted >6.51 min.
Next, Kernel Density Estimation and the Kolmogorov-Smirnov test were used to compare the distribution
of durations for ‘Quiet sleep’ episodes occurring immediately after ‘Active sleep’ episodes versus ‘Quiet
sleep’ episodes occurring not immediately after ‘Active sleep’ episodes (p = 0.22). The distributions of
the ‘Active sleep’ episodes immediately preceding ‘Quiet sleep’ episodes and not immediately preceding
‘Quiet sleep’ episodes were also compared (p = 0.98). Likewise, we further compared the ‘Active sleep’
episodes immediately succeeding ‘Quiet sleep’ episodes and not immediately succeeding ‘Quiet sleep’
episodes (p = 0.27). None of these comparisons showed significant differences (Figures S1B–S1D). These
results indicate that there is a relationship between the duration of ‘Quiet sleep’ episodes and the
B
A
‘Quiet sleep’ ‘Quiet sleep’
Short Active
Long
sleep
Active
Active
sleep
Long ‘Quiet sleep’
Short ‘Quiet sleep’
Quiet with
Half and half
Duration of episodes (min)
Quiet with
Quiet with
One eye movement
Open pupil
and Half
One Eye
Movement
Alert Quiet with
open pupil
Half
Figure 4. Characterization of the behaviors observed in Octopus insularis
Each color represents one behavior analyzed during the experiment.
(A) Boxplot and histogram of all animals’ behavior observed through the video recordings during the 12 h of the light
period. Data from the behaviors with shorter durations were zoomed in (top right inset).
(B) Pie chart showing the proportion of the total duration of each behavior for all animals.
ll
OPEN ACCESS
6iScience --, 102223, --, 2021
Please cite this article in press as: Medeiros et al., Cyclic alternation of quiet and active sleep states in the octopus, iScience (2021), https://
doi.org/10.1016/j.isci.2021.102223
iScienc
e
Article
subsequent occurrence of ‘Active sleep’ episodes, whereas the duration of ‘Active sleep’ episodes has no
relationship with the occurrence of subsequent ‘Quiet sleep’ episodes. Importantly, short ‘Quiet sleep’ ep-
isodes are rarely followed by ‘Active sleep’ episodes.
The behaviors with the highest total time during the diurnal video recordings were the ‘Alert’ state
(37.85%), ‘QOP’ (28.42%), and long ‘Quiet sleep’ (22.63%), whereas the ones with the lowest total time
were ‘Active’ (6.42%), short ‘Quiet sleep’ (3.5%), ‘QHH’ body pattern (0.57%), ‘Active sleep’ (0.52%), and
‘QOEM’ (0.09%) (Figure 4B).
The behaviors with the highest frequencies per day were ‘Alert’ (35.56%) and ‘QOP’ (31.31%), whereas the
behaviors ‘QOEM’ (0.50%) and ‘Active sleep’ (7.81%) had the lowest frequencies (Table S1).
There was a significant difference between the duration of almost all states observed. All quiescence be-
haviors differed except ‘QOEM.’ In addition, the ‘Active’, ‘Alert’, and ‘QOP’ states in which the octopus is
possibly awake differed from all other quiescence states (‘Quiet sleep’, ‘Active sleep’, and ‘QHH’) except
‘QOEM’, as shown in Table S2.
Ultradian cyclicity of ‘active sleep’ and long ‘quiet sleep’ episodes
To analyze the cyclic pattern comprising the long ‘Quiet sleep’ and the ‘Active sleep,’ we excluded the
feeding period because it interfered in the sleep cycle. From the total of 74 long ‘Quiet sleep’ intervals
A
B
Duration of Quiet sleep episodes as a function of its precedent state
Active sleep
Duration of Quiet sleep
episodes (min)
Alert Quiet with
open pupil
Quiet with
half and half
Duration of Quiet sleep episodes as a function of its subsequent state
Duration of Quiet sleep
episodes (min)
Active sleep
Alert Quiet with
half and half
Figure 5. Duration of ‘Quiet sleep’ episodes as a function of precedent or subsequent states
States that occurred less than 10 times before or after ‘Quiet sleep’ were discarded from this analysis.
(A) Boxplot and histogram showing the duration in minutes (y axis) of ‘Quiet sleep’ episodes that occurred after the
behaviors ‘Alert’, Quiet with open pupil (‘QOP’), Quiet with half and half (‘QHH’), and ‘Active sleep’ (x axis). The behaviors
‘Alert’,‘QOP’,and‘QHH’occurredbeforelong‘Quietsleep’episodes, with medians 10.73, 7.41, and 5.84, respectively.
‘Active sleep’ episodes preceded the shortest ‘Quiet sleep’ episodes.
(B) Boxplot and histogram showing the duration in min (y axis) of ‘Quiet sleep’ episodes that occurred prior to the
behaviors ‘Alert’, ‘QHH’, and ‘Active sleep’ (x axis). Long ‘Quiet sleep’ episodes (median value of 13.33 min) often led to
‘Active sleep’ episodes. ‘Active sleep’ was the only behavior that occurred after ‘Quiet sleep’ episodes with durations
above 30 min. ‘QHH’ episodes were preceded by ‘Quiet sleep’ episodes of short duration (median 3.9 min), whereas
‘Alert’ episodes were preceded by ‘Quiet sleep’ episodes of intermediate duration (median 4.6 min).
ll
OPEN ACCESS
iScience --, 102223, --, 2021 7
Please cite this article in press as: Medeiros et al., Cyclic alternation of quiet and active sleep states in the octopus, iScience (2021), https://
doi.org/10.1016/j.isci.2021.102223
iScienc
e
Article
(time spent between the end of one episode and the beginning of the next episode of the same behavior),
60% had durations between 6.77 min and 25.48 min, i.e., with a wide range of durations (Figure 6A; Table S3
and Figure S2). From the total of 79 ‘Active sleep’ intervals, 60% had duration between 29.58 min and
32.98 min, i.e. with a narrow range of durations (Figure 6B; Table S4).
A hypnogram was elaborated from the behavioral observation of one representative animal (octopus 3) for
200 continuous minutes in order to exemplify the ultradian cyclicity. Concomitantly with the hypnogram, we
analyzed skin changes in color and texture, represented respectively by general skin color and localized
skin patterning dynamics (Figure 7). Tracking of eyes and mantle movements was also performed for the
same period (Figure S3). These parameters varied substantially across behaviors. For instance, during
‘Quiet sleep’ there was less variation of these parameters an ddur ing ‘Active sleep’ there was a more accen-
tuated variation of the measures, whereas during the ‘Alert’ state we observed a constant and not extreme
oscillation of skin color and texture, with mantle and eye movements.
General skin color dynamics across behavioral states
The pixel analysis of mean brightness values showedcharacteristi c profiles of chromatophore color changes
for each behavioral state (Figure 8A; Figure S4;Figure 7 second panel). When animals were in the ‘Alert’
A
B
Quiet sleep
Intervals between Long quiet sleep episodes
Intervals between REM-like episodes
Amount of
Active sleep
Intervals between Active sleep episodes Amount of
Intervals
111 1
111129 6 32 12
16 8 36 13
Duration of Long
Quiet sleep
interval (min)
Duration of Long
Quiet sleep
interval (min)
Duration of
Active sleep
interval (min)
Duration of
Active sleep
interval (min)
Octopus 2 Octopus 3 Octopus 4
Octopus 1 Octopus 2 Octopus 3 Octopus 4
Octopus 1
Intervals
Figure 6. Ultradian Cyclicity of ‘Active sleep’ and long ‘Quiet sleep’ episodes
Each background color represents one individual octopus
(A) ‘Quiet sleep’ characterization: (Top panel) plots of intervals between long ‘Quiet sleep’ with histogram on the right;
(Bottom panel) boxplot and histogram for the durations of intervals between long ‘Quiet sleep’ episodes for each octopus.
(B) ‘Active sleep’ characterization: (Top panel) plots of intervals between ‘Active sleep’ with its histogram distributions on
the right; (Bottom panel) boxplot and histogram for the ‘Active sleep’ interval durations of each octopus. For better
visualization of the distribution, we excluded from the plot the data points 227.78 min and 150.45 min of the ‘Active sleep’
intervals and data point 173.47 min of the long ’Quiet sleep’ intervals.
ll
OPEN ACCESS
8iScience --, 102223, --, 2021
Please cite this article in press as: Medeiros et al., Cyclic alternation of quiet and active sleep states in the octopus, iScience (2021), https://
doi.org/10.1016/j.isci.2021.102223
iScienc
e
Article
state, brightness intensities recordedfrom the skin kept oscillating (mantle—median 0.56, first quartile 0.22,
third quartile 1.25; head—median 0.75, first quartile 0.34, third quartile 1.59), whereas during the ‘Quiet
sleep’ state there was very little variation in themeasurements (mantle—median 0.21, first quartile 0.09, third
quartile 0.47; head—median 0.25, first quartile 0.11, third quartile 0.61). However, during ‘Active sleep’ ep-
isodes there was a conspicuous and abrupt decrease in the measures, followed by intense brightness vari-
ation throughout the duration of the episode (mantle—median 2.09, first quartile 0.93, third quartile 3.94;
head—median 2.8, first quartile 1.35, third quartile 5.14). This difference in color variation between these
three states was confirmed by the statistical analyses for head and mantle (Kruskal-Wallis, head
p = 2.20 310
16
;mantlep=2.20310
16
). The pairwise comparisons (Wilcoxon) with Bonferroni adjustment
showed that in the head and mantlethere were more variations in skin color during ‘Active sleep’ than during
‘Quiet sleep’ (head p = 2.00 310
16
;mantlep=2.00310
16
) and ‘Alert’ state (head p = 2.00 310
16
;mantle
p = 2.00 310
16
). There was also more color variation duringthe ‘Alert’ state than during ‘Quiet sleep’ (head
p = 2.00 310
16
;mantlep=5.30310
12
). These results are shown by the raw data and Gaussian filter of the
number of white pixels over time (Figure 8A) and by the median of the variations shown by each octopus
(Figure 8B). The other octopuses presented similar profiles, as shown in the Figure S4.
Localized skin patterning dynamics across behavioral states
The localized skin dynamics is caused by texture changes due to extension and retraction of the papillae
and changes in color patterning within specific body regions.Thesedynamicchangesinskinwereinves-
tigated using the Canny algorithm, which captured edges in the image frames. The edges were visualized
through white pixels that indicate the skin papillae’s texture and thecolor patterning complexity (Figure 9A;
Figure 7, third panel; Figure S5). There was a significant difference in variation of skin texture and
complexity of color patterns across ‘Quiet sleep’ (median 1.00, first quartile 0.00, third quartile 3.00), ‘Active
sleep’ (median 8.00, first quartile 3.00, third quartile 4. 16.00), and the ‘Alert’ state (median 2.00, first quartile
1.00, third quartile 5.00) for all the octopuses (Kruskal-Wallis p = 2.20 310
16
).Thepairwisecomparisons
20
80
140
200
Hypnogram
Long ‘Quiet sleep’
Short ‘Quiet sleep’
Active sleep
Half and half
600
1050
1500
Amount of white pixels ROI'S mean brightness
Localized skin patterning dynamics
General skin color dynamics
Head
Mantle
0 20 40 60 80 100 120 140 160 081002
Time (min)
150
Figure 7. Hypnogram with corresponding measurements of general skin color dynamics and localized skin
patterning dynamics
The figure depicts the behaviors of one representative animal (octopus 3) along 200 min of diurnal recordings, beginning
at 06:36:50 and finishing at 09:56:50. The plot of general skin color dynamics depicts the mean color from two regions of
interest delimited over the head and dorsal mantle (detailed in Figure 7). The plot of localized skin patterning dynamics
shows changes in skin texture represented by the variation of the number of white pixels captured by the Canny algorithm
along time. For both skin dynamic analyses, a marked variation of color and texture concomitant with the occurrence of
‘Active sleep’ can be observed.
ll
OPEN ACCESS
iScience --, 102223, --, 2021 9
Please cite this article in press as: Medeiros et al., Cyclic alternation of quiet and active sleep states in the octopus, iScience (2021), https://
doi.org/10.1016/j.isci.2021.102223
iScienc
e
Article
(Wilcoxon) with Bonferroni adjustment showed that there was significantly more variation in skin texture
during ‘Active sleep’ than during ‘Quiet sleep’ (p = 2.00 310
16
)and‘Alert’state(p=2.00310
16
). There
was also significantly more variation during the ‘Alert’ state than during ‘Quiet sleep’ (p = 3.40 310
16
), as
shown by the raw data and Gaussian filter of the number of white pixels over time (Figure 9A) and by the
B
Quiet
Sleep
Active
Sleep
Alert Quiet
Sleep
Active
Sleep
Alert
Quiet
Sleep
Active
Sleep
Alert Quiet
Sleep
Active
Sleep
Alert
Quiet
Sleep
Active
Sleep
Alert Quiet
Sleep
Active
Sleep
Aler
t
Time (s)
0
6
0
3
0
7
0
7
0
5
0
5
ROI’s mean brightness
Median of ROi’s
brightness variation
Median of ROi’s
brightness variation
Octopus 4
Octopus 2
Octopus 1 Octopus 3
Octopus 4
Octopus 3
Minimal Minimal Maximal Maximal
Head
Mantle
Head
Mantle
Octopus 3
A
Figure 8. General skin color dynamics across ‘Quiet sleep’, ‘Active sleep’, and the ‘Alert’ state
(A) The left panel shows changes in skin color represented by the variation of the ROI’s mean brightness (y axis) along 174 s
(x axis) for five transitions across ‘Quiet sleep’ (blue zone), ‘Active sleep’ (green zone), and the ‘Alert’ state (orange zone)
from octopus 3. The intervals analyzed for ‘Quiet sleep’ and the ‘Alert state’ had the same duration of the ‘Active sleep’
episode that occurred between them. The examples are organized from the shortest ‘Active sleep’ episode, on the top, to
the longest, on the bottom. The green line indicates the head’s color and the blue line indicates the mantle’s color. Both
signals are well followed by the Gaussian filter (red line). All the behavioral transitions show a similar profile, with little
value variation during ‘Quiet sleep’ and the ‘Alert’ state, and with a marked variation during ‘Active sleep’, so that both
extreme values (highlighted by the blue and red dots) occurred during ‘Active sleep’. The right panel shows a picture of
the octopus with the ROI’s selected for head (green) and mantle (blue) when the animal color achieved its maximum value
(most pale) and minimum value (most dark).
(B) Scatterplots showing the medians of delta (500 ms) Gaussian values for the color analysis of each behavior per octopus
(n = 4). The top panel shows results for the head, and the bottom panel shows results for the mantle. Note that the mantle
was properly visible in two animals only, so octopuses 1 and 2 were not included in the mantle color analyses. The color
variation in all animals followed a general profile in which the highest color variation occurred during ‘Active sleep’.
ll
OPEN ACCESS
10 iScience --, 102223, --, 2021
Please cite this article in press as: Medeiros et al., Cyclic alternation of quiet and active sleep states in the octopus, iScience (2021), https://
doi.org/10.1016/j.isci.2021.102223
iScienc
e
Article
median of the variations shown by each octopus (Figure 9B). The octopuses 1, 2, and 4 presented similar
profiles, as shown in the Figure S5.
Eye movements across behavioral states
For each octopus we analyzed eye movements for five periods comprising a natural sequence of ‘Quiet
sleep’, ‘Active sleep’, and the ‘Alert’ state. The octopuses showed significant differences in eye movements
across states (Kruskal-Wallis p = 2.20 310
16
). Besides, pairwise comparisons (Wilcoxon) with Bonferroni
adjustment show that the increase rate during ‘Active sleep’ (median 0.56, first quartile 0.32, third quartile
1.32) was significant in comparison with ‘Quiet sleep’ (median 0.36, first quartile 0.21, third quartile 0.79)
(p = 2.00 310
16
) and ‘Alert’ (median 0.44, first quartile 0.25, third quartile 1.02) (p = 5.20 310
15
). There
was also a significant increase in eye movements during the ‘Alert’ state in comparison with ‘Quiet sleep’
(p = 8.60 310
7
)(Figure 10).
Mantle ventilation movements across behavioral states
The octopuses 3 and 4 showed significant differences in mantle movement across ‘Quiet sleep’ (median 0.70,
first quartile 0.29, third quartile 4.67), ‘Active sleep’ (median 1.12, first quartile 0.41, third quartile 4.60), and the
‘Alert’ state (median 0.67, first quartile 0.28, third quartile 3.23) (Kruskal-Wallis p = 1.43 310
8
). The pairwise
Octopus 4
Octopus 2
Octopus 1 Octopus 3
Octopus 3
Amount of white pixels
Time (s)
A
B
Quiet
Sleep
Active
Sleep
Alert Alert
Active
Sleep
Quiet
Sleep
Alert
Active
Sleep
Quiet
Sleep
Quiet
Sleep
Active
Sleep
Alert
Minimal white pixels Maximal white pixels
Median of white
pixels variation
Figure 9. Localized skin patterning dynamics across ‘Quiet sleep’, ‘Active sleep’, and the ‘Alert’ state
(A) The left panel shows changes in skin texture represented by the variation of the number of white pixels (y axis) along
174 s (x axis) for five transitions across ‘Quiet sleep’ (blue zone), ‘Active sleep’ (green zone), and the ‘Alert’ state (orange
zone) from octopus 3. The intervals analyzed for ‘Quiet sleep’ and Alert had the same duration of the ‘Active sleep’ that
occurred between them. The examples are organized from the shortest ‘Active sleep’ episode, on the top, to the longest,
on the bottom. The red line is a Gaussian filter and follows closely the real data, represented by the black line. All the
behavioral transitions show a similar profile, with little variation during ‘Quiet sleep’ and the ‘Alert’ state, and with a
marked variation during ‘Active sleep’, so that both extreme values occurred during ‘Active sleep’ and all the highest
values occurred after the lowest values. The top value among the five sampled transitions is on the last line, having 1,367
white pixels and the lowest value is on the first line, with 281 white pixels. The right panel shows the moment during ‘Active
sleep’ when the animal displayed minimum body pattern complexity, represented by the minimal number of white pixels
captured by the Canny edges algorithm (blue dots on the graphs and minimal white pixels column); the maximal count of
white pixels are represented by the red dots on the graphs and maximum white pixels (red) column.
(B) Scatterplots showing the medians of delta (500 ms) Gaussian values for the behavioral analysis of each octopus (n = 4).
The median values of ‘Active sleep’ are mostly higher than during other behaviors (except for one transition in octopus 1,
when the value did not change across behaviors).
ll
OPEN ACCESS
iScience --, 102223, --, 2021 11
Please cite this article in press as: Medeiros et al., Cyclic alternation of quiet and active sleep states in the octopus, iScience (2021), https://
doi.org/10.1016/j.isci.2021.102223
iScienc
e
Article
Octopus 3
Time (s)
Variation in position (pixel)
Quiet
Sleep
Active
Sleep
Alert Alert
Active
Sleep
Quiet
Sleep
Alert
Active
Sleep
Quiet
Sleep
Quiet
Sleep
Active
Sleep
Aler
t
Octopus 4
Octopus 2
Octopus 1 Octopus 3
position variation (pixel)
Median of the
A
B
Figure 10. Eye movements across ‘Quiet sleep’, ‘Active sleep’, and the ‘Alert’ state
(A) The left panel shows the sum of the position variation for the left and right eyes (y axis) over time (x axis) for five
transitions between ‘Quiet sleep’ (blue zone), ‘Active sleep’ (green zone), and the ‘Alert’ state (orange zone) for octopus3.
The intervals analyzed for ‘Quiet sleep’ and the ‘Alert’ state had the same duration of the ‘Active sleep’ that occurred
between them. The examples are organized from the shortest ‘Active sleep’, on top, to the longest, on the bottom. The
red line on the graph is a Gaussian filter and follows closely the real data, represented by the black line. The right panels
display an example of the region where marks were placed on each eye to perform the tracking, with zoomed-in images
on the right.
(B) Scatterplots show the medians of delta (300 ms) Gaussian values for the behavioral analysis of each octopus (n = 4).
ll
OPEN ACCESS
12 iScience --, 102223, --, 2021
Please cite this article in press as: Medeiros et al., Cyclic alternation of quiet and active sleep states in the octopus, iScience (2021), https://
doi.org/10.1016/j.isci.2021.102223
iScienc
e
Article
comparison (Wilcoxon) with Bonferroni adjustment shows that the increased rate during ‘Active sleep’ was sig-
nificant in comparison with ‘Quiet sleep’ (p = 2.90 310
5
) and the ‘Alert’ state (p = 6.50 310
12
). Besides, the
increase in mantle movements during ‘Quiet sleep’ in comparison withthe ‘Alert’ state was also significant (p =
0.04). Figure 11 shows in detail for octopus 3 the variation of mantle position over time for the behavioral
sequences comprising ‘Quiet sleep’, ‘Active sleep’, and the ‘Alert’ state (Figures 11A) and the scatterplot
with the medians of these transitions for both octopuses (Figures 11B).
Behavioral transitions show a characteristic pattern
To quantify comprehensively how specimens of Octopus insularis transit between specific behavioral
states, the entire dataset recorded during daytime was subjected to graph analysis. The graphs of behav-
ioral transitions show that some of the states were more correlated with each other and that there was a
pattern in the sequence of their transitions (Figure 12). ‘Active sleep’ occurred after ‘Quiet sleep’ 82% of
the times, with 18% coming from short ‘Quiet sleep’ episodes and 64% from long ‘Quiet sleep’ episodes.
The behaviors with the highest probability to occur after long ‘Quiet sleep’ were ‘Active sleep’ (57%), the
‘Alert’ state (24%), and ‘QHH’ (16%). Furthermore, ‘QHH’ was more strongly correlated with quiet
behaviors, with 50% of the inputs coming from ‘QOP’ and 45% of all short ‘Quiet sleep’ outputs leading
to ‘QHH’. The ‘Quiet sleep’ episodes were mostly preceded by ‘QHH’, the ‘Alert’ state, and ‘QOP’, but
almost never by ‘Active sleep’, with significant differences between their frequencies (chi-squared test,
p = 2.20 310
16
). When this occurred the ‘Quiet sleep’ episodes were almost invariably short (chi-squared
test, p = 4.27 310
3
).
DISCUSSION
In the present work, we showed that Octopus insularis displays two different quiescence states that fulfill
the behavioral criteria for sleep, namely ‘Quiet sleep’ and ‘Active sleep’. We characterized these states with
regard to general body skin color dynamics, local skin patterning dynamics, eye and mantle movements,
episode duration, episode periodicity, transition probabilities, and arousal thresholds. Altogether, the re-
sults point to a cyclic-state dynamics in which the behavioral sequence ‘Quiet sleep’ to ‘Active sleep’ to
‘Alert’ state is prevalent.
‘Quiet sleep’ has already been observed in Octopus vulgaris (Meisel et al., 2011), but ‘Active sleep’ has not
yet been described in the literature for the octopus. However, this ‘Active sleep’ is similar to the behavior
named REM-like sleep in Sepia officinalis (Iglesias et al., 2019), by analogy with rapid-eye movement (REM)
sleep in mammals. In both species there were eye movements, dynamic chromatophore patterning, and a
sudden simultaneous darkening of the mantle and head chromatophores at state onset (Frank et al., 2012;
Iglesias et al., 2019). Another similarity between the ‘Active sleep’ here described and the REM-like state
described in Sepia officinalis (Iglesias et al., 2019) is the presence of an ultradian rhythm for both behaviors,
with a characteristic cycle. The periodicity of REM-like sleep in Sepia officinalis was 36.34 G1.46 min,
whereas in Octopus insularis 60% of the intervals between ‘Active sleep’ episodes had durations between
29.58 min and 32.98 min. Importantly, an ultradian sleep rhythm is also observed in mammals (Trachsel
et al., 1991), birds (Walker and Berger, 1972), and in the reptile bearded dragon Pogona vitticeps (Ahl,
1926;Norimoto et al., 2020).
Regarding this periodicity, octopus 4 differed a bit from the others with regard to durations of the intervals
between consecutive ‘Active sleep’ episodes. Reptiles, such as the bearded dragon, can change the period
of neural oscillations according to the environment’s temperature, so that an increase in temperature leads
toadecreaseinperiodicity(Shein-Idelson et al., 2016). However, considering that during this study thewa-
ter temperature was kept identical for all animals and that octopus 4 was the female with the highest body
weight and probably the most mature animal assessed (Lima et al., 2014), it is possible that Octopus insu-
laris specimens undergo changes in their sleep cycles as they age.
During quiescence animals also presented the behavior ‘QHH’, which was also observed in O. vulgaris dur-
ing resting periods (Meisel et al., 2011). For this reason, we included this state in the arousal threshold
experiment. The visual latency test performed with Octopus insularis showed a significant and gradual in-
crease in arousal threshold from the ‘Alert’ state to QHH, ‘Quiet sleep’, and ‘Active sleep’. These results
strongly suggest the existence of different sleep states in octopuses.
ll
OPEN ACCESS
iScience --, 102223, --, 2021 13
Please cite this article in press as: Medeiros et al., Cyclic alternation of quiet and active sleep states in the octopus, iScience (2021), https://
doi.org/10.1016/j.isci.2021.102223
iScienc
e
Article
Octopus 3
position variation (pixel)
Median of the
A
BOctopus 3
Time (s)
Variation in position (pixel)
Quiet
Sleep
Active
Sleep
Alert Quiet
Sleep
Active
Sleep
Alert
Octopus 4
Figure 11. Mantle ventilation movements across ‘Quiet sleep’, ‘Active sleep’, and the ‘Alert’ state
(A) The left panel is showing sum of the position variation of seven different spots chosen in the mantle for the tracking
(y axis) along the time (x axis) for five transitions between ‘Quiet sleep’ (blue zone), ‘Active sleep’ (green zone), and the
‘Alert’ state (orange zone) from octopus 3. The intervals analyzed for ‘Quiet sleep’ and the ‘Alert’ state had the same
duration of the ‘Active sleep’ that occurred between them. The five examples are organized from the shortest ‘Active
sleep’ episode, on top, to the longest episode, on the bottom. The red line on the graph is a Gaussian filter and follows
closely the real data, represented by the black line. The right panels display an example of the region where seven marks
(red circles) were placed on the mantle to perform the tracking, with zoomed-in images on the right.
(B) Scatterplots showing the medians of delta (500 ms) Gaussian values for the behavioral analysis of each octopus (n = 2).
ll
OPEN ACCESS
14 iScience --, 102223, --, 2021
Please cite this article in press as: Medeiros et al., Cyclic alternation of quiet and active sleep states in the octopus, iScience (2021), https://
doi.org/10.1016/j.isci.2021.102223
iScienc
e
Article
A
B
Half
Half
and
One Eye
Movement
Movement
One Eye
Half
Half
and
Open
Pupil
Open
Pupil
Active
Sleep
Active
Sleep
Sleep
Quiet
Long
Active
Active
Long
Quiet
Sleep
Quiet
Sleep
Short
Short
Quiet
Sleep
Inputs
Outputs
Quiet
Quiet
with
with
Figure 12. Proportions of transitions between behaviors
Graphs show the proportions of behavioral transitions that occurred along the 180 h and 49 min of video recordings (n = 4
octopuses). Each node of the graph represents one behavior described on the ethogram and the edges indicate
transitions.
(A) Percentage of the total inputs for each behavior.
(B) Percentage of total outputs for each behavior.
ll
OPEN ACCESS
iScience --, 102223, --, 2021 15
Please cite this article in press as: Medeiros et al., Cyclic alternation of quiet and active sleep states in the octopus, iScience (2021), https://
doi.org/10.1016/j.isci.2021.102223
iScienc
e
Article
We also hypothesized that the ‘‘half and half’’ body pattern is related to a rest behavior, something also
reported for Octopus vulgaris, including an alternation of the body half that gets dark a few minutes later
(Figure 2C) (Meisel et al., 2011). However, other authors have reported that this behavior is related to intra-
specific interactions with other cephalopods, such as in the squid Sepioteuthis sepioidea (Blainville, 1823;
Mather, 2016). In our experiment, it is possible that the octopuses were seeing their own reflection in the
aquarium wall and the ‘QHH’ is a response to environmental stimuli rather than a resting behavior. Never-
theless, in agreement with our hypothesis, the arousal threshold during ‘QHH’ was significantly higher in
comparison with the ‘Alert’ state, and the peaks occurred during periods of quiescence. Thus, this state
may be related to a lighter sleep state preceding or interspersing the ‘Quiet sleep’.
The ‘Quiet state with only one eye movement’ (‘QOEM’), which has never been reported for any cepha-
lopod species, also seems to be a rest behavior. However, further studies must investigate whether this un-
usual type of eye movement can also occur with both eyes simultaneously, whether it can also be observed
in the natural environment, and what are the physiological processes underlying it. The fact that this
behavior also occurred during the night, with lights off, indicates that it is not triggered as a response of
the reflexive glass tank (Video S9).
Our study demonstrates that quiescent states shown by specimens of Octopus insularis fit most of the
behavioral criteria for sleep. The existence in cephalopods of at least two different sleep states within an
ultradian wake-sleep cycle contrasts with the apparent existence of a single sleep state in other mollusks,
such as in Aplysia californica (Cooper, 1863;Vorster and Born, 2017)andLymnaea stagnalis (Linnaeus, 1758;
Stephenson and Lewis, 2011). If extended to multiple species of non-cephalopod mollusks, this difference
suggests an independent evolution of ‘Active sleep’ in cephalopods and amniotes.
One of the main interests in the field is to establish parallels between the ‘Quiet sleep’ and ‘Active sleep’
states presented here and the different physiological responses found in the mammalian non-REM (NREM)
and REM sleep states. Given that the ‘Active sleep’ state has not been described for octopuses before, it is
necessary to investigate whether the similarity of this state with REM sleep goes beyond the behavioralsim-
ilarities observed in this work, such as elevated arousal threshold, eye movements, and body twitches
(Aserinsky and Kleitman, 1955;Dement, 1958;Dillon and Webb, 1965). The similarities between the ‘Quiet
sleep’ of octopuses and NREM sleep in vertebrates, such as behavi oral quiescence, with only brief and min-
imal movements (Muzet et al., 1972), and increased arousal threshold (Neckelmann and Ursin, 1993),
prompt the need for further investigation on whether octopuses and mammas undergo similar physiolog-
ical processes. Recordings of brain activity during ‘Quiet sleep’ and ‘Active sleep’ can be used to search for
the occurrence of typical neural oscillations found in mammals (Steriade et al., 1993,Hobson and Mccarley,
1971), but there is of course a major limitation in this comparison, because the vertebrate and cephalopod
brains are not homologous and differ substantially in tissue organization. Furthermore, given that it is very
challenging to successfully place electrodes in the octopus brain, (Brown et al., 2006) remains the only pub-
lished electrophysiological study of octopus’ sleep. However, although it shows interesting findings in four
specimens of O. vulgaris, such as increase in neural activity during quiescence, the video recordings were
sampled using time lapse. For this reason, we can speculate that ‘Active sleep’, which has a brief episode
duration, could have been severely undersampled. Furthermore, it is possible that during such increase in
brain activity the animal was actually undergoing a short ‘Active sleep’ episode as described here, which
may have been missed in the video record. Therefore, we recommend further studies in this field to assess
brain electrophysiology in continuous behavioral recordings.
As proposed by (Reiter et al., 2018), it is reasonable that chromatophore expansion could serve as a proxy
for motor neuron activity, because each chromatophore is controlled by a small number of motor neurons,
and each motor neuron controls a small number of chromatophores (Reed, 1995). Therefore, the skin
patterning computational analyses can be useful to understand neural activity of these animals. The anal-
ysis of skin color and texture dynamics revealed differences in chromatophore pattern and papillae expo-
sure across episodes of the ‘Alert state’, ‘Quiet sleep’, and ‘Active sleep’. During the ‘Al ert’ state there were
oscillations, whereas during ‘Quiet sleep’ they remained relatively constant and during the ‘Active sleep’
they change abruptly, in accordance with the dynamic body pattern. Furthermore, the onset of ‘Active
sleep’ episodes was characterized by a sudden and simultaneous darkening of the mantle and head, as
described for what has been called REM-like state in Sepia officinalis (Iglesias et al., 2019). In addition,
we also found a moderate increase in eye movements and ventilation rate during ‘Active sleep’.
ll
OPEN ACCESS
16 iScience --, 102223, --, 2021
Please cite this article in press as: Medeiros et al., Cyclic alternation of quiet and active sleep states in the octopus, iScience (2021), https://
doi.org/10.1016/j.isci.2021.102223
iScienc
e
Article
Conclusions
Despite the great evolutionary distance between vertebrates and invertebrates, the sleep behavior of Octopus
insularis shares many features with the sleep of amniotes, including the ultradian cyclic pattern observedfor the
‘Active sleep’ state. The recent evidence of ‘Active Sleep’ in Drosophila (Tainton-Heap et al., 2021) suggests
strong selection pressure across evolution for an alternation between ‘Quiet’ and ‘Active’ sleep states. The
occurrence of the non-REM/REM alternation in mammals, birds, and in some reptiles, such as in the bearded
dragon and in the argentine tegu Salvator merianae (Dume
´ril and Bibron, 1839;Libourel et al., 2018;Shein-Idel-
son et al., 2016), points to a common origin of the wake-sleep cycle in these groups of animals, which share a
common ancestor (Libourel et al., 2018). However, considering that cephalopods split from vertebrates more
than 500 million years ago (Vitti, 2013;Shu et al., 2001), it is likely that the sleep behaviors observed here, despite
their similarity to those found in amniotes, are analogous rather than homologous to these states.
By the same token, the fact that analogous nervous systems such as the cephalopod brain and the vertebrate
brain evolved similarbehavioral sequences across the wake-sleep cycle is strongly suggestive of convergent evo-
lution. Cephalopods have evolved de novo neural structures termed lobes, including the vertical lobe that is
involved in long-term memory and shares some functional features with the mammalian hippocampus (Gutnick
et al., 2016;Nixon and Young, 2003). Indeed, cephalopod evolution seems to have converged with vertebrates
with regard to the neural mechanisms underlying learning (Gutnick et al., 2016;Shomrat et al., 2015;Hochner
et al., 2003). It remains to be investigated whether the physiological functions of sleep, in this far-evolving taxon,
also resemble the functions performed in amniotes, such as metabolic detoxification (Xie et al., 2013;Hablitz
et al., 2019)andcognitiveprocessing(Boyce et al., 2016;Blanco et al., 2015).
Limitations of the study
A more accurate analysis of the behavior ‘QHH’ would involve estimating the arousal threshold using the visual
stimulus to selectively affect only the dark side or the pale side separately. We tried to do it, but it proved very
difficult to have the animal stay long enough in the same position, toensure that it was really seeing the monitor
with only one eye. Furthermore, even if we could establish beyond doubt that just one eye was able to see the
monitor, the other eye still could be stimulated by reflections on the aquarium’s glass.
If ‘QHH’ is in fact a sleep state, it may be akin to the unihemispheric sleep observed in bird and aquatic
mammals (Rattenborg et al., 1999,2000,2016;Ridgway et al., 2006;Lyamin et al., 2002;Mascetti, 2016). Uni-
hemispheric sleep occurs during slow wave sleep (SWS), and the eye corresponding to the sleeping hemi-
sphere remains closed. However, another limitation of this study was that we were not able to compare the
pupil contractions recorded from the pale and dark sides during the ‘QHH’. Pupil size was difficult to mea-
sure because (1) ‘QHH’ is a behavior with reduced frequency and very short duration; (2) we were often able
to observe only one of the eyes because of the camera position; (3) commonly the eye from the dark side
also gets dark, making it difficult to distinguish the pupil from the rest of the eye; and (4) given the asym-
metries in skin color, we could not assume that both eyes had their pupils equally open or closed.
For a deeper understanding of‘Active sleep’ it will be important to quantify more precisely the occurrenceof
body twitches. Because the twitches are distributed all over the octopus’ body, variations of focus and body
position made these twitches difficult to quantify here. Anothe r caveat of the present study is that we did not
assess sleep rebound. This hallmark of sleep has been documented in Octopus vulgaris (Brown et al., 2006;
Meisel et al., 2011), so it is quite likely that Octopus insularis also has a sleep homeostatic regulation.
Resource availability
Lead contact
Further information and requests for resources and reagents should be directed to and will be fulfilled by
the Lead Contact, Sidarta Ribeiro (sidartaribeiro@neuro.ufrn.br).
Materials availability
Our study did not generate new unique reagents.
Data and code availability
Original materials, data and code have been deposited in OSF at https://osf.io/f6jyu/?
view_only=c8341a1535ad472d908f7a5b629e99a3.
ll
OPEN ACCESS
iScience --, 102223, --, 2021 17
Please cite this article in press as: Medeiros et al., Cyclic alternation of quiet and active sleep states in the octopus, iScience (2021), https://
doi.org/10.1016/j.isci.2021.102223
iScienc
e
Article
METHODS
Allmethodscanbefoundintheaccompanyingtransparent methods supplemental file.
SUPPLEMENTAL INFORMATION
Supplemental information can be found online at https://doi.org/10.1016/j.isci.2021.102223.
ACKNOWLEDGMENTS
Funding was obtained from the Coordenac¸a
˜odeAperfeic¸oamento de Pessoal de Nı
´vel Superior (CAPES),
Project - Cie
ˆncias do Mar II-23038.004807/2014- 01, Undergraduate scientific research scholarship from
CNPq, from the State University of Rio Grande do Norte (UERN), grants #308775/2015-5 and #408145/
2016-1 from Conselho Nacional de Desenvolvimento Cientı
´fico e Tecnolo
´gico (CNPq), and grant #2013/
07699-0 from the Sa
˜o Paulo Research Foundation (FAPESP) Center for Neuromathematics. We thank the
Mind the Graph platform (www.mindthegraph.com) for facilitating figure preparation. We thank Annie
da Costa Souza and Renato Junqueira de Souza Dantas for helping with animal collection in the field, Al-
berto Medeiros for photography edition, Claudio Queiroz for granting access to software, and the staff of
the Brain Institute and BioME (UFRN) for their indispensable technical and logistical assistance, especially
Ana Elvira Oliveira and Eronildo Lira de Santana.
AUTHOR CONTRIBUTIONS
SLSM, SR, and TSL conceived and designed the experiments and literature search; SLSM and SS secured
laboratory space and funds for the research; SLSM and SR wrote the paper and PHL contributed with the
writing; WB and EBS contributed with study design and data interpretation; SLSM and MMMP performed
the experiments and collected and analyzed the data; SLSM, MMMP, and FDL performed specimen collec-
tion in the field; FDL, SLSM, PHL, and IGM performed the statistical analysis; SLSM, MMMP, PHL, JBCO,
IGM, and SR prepared figures and/or tables; SR, TSL, FDL, WB, EBS, and SS reviewed the paper and
approved the final version.
DECLARATION OF INTERESTS
The authors declare no competing interests.
Received: June 19, 2020
Revised: December 22, 2020
Accepted: February 18, 2021
Published: March 25, 2021
REFERENCES
Aserinsky, E., and Kleitman, N. (1955). A motility
cycle in sleeping infants as manifested by ocular
and gross bodily activity. J. Appl. Physiol. 8,
11–18.
Ahl, E. (1926). Neue Eidechsen und Amphibien.
Zoologischer Anzeiger 67, 186–192.
Ayala-Guerrero, F., Perez, M.C., and Calderon, A.
(1988). Sleep patterns in the bird Aratinga
canicularis. Physiol. Behav. 43, 585–589.
Blanco, W., Pereira, C.M., Cota, V.R., Souza, A.C.,
Renno-Costa, C., Santos, S., Dias, G., Guerreiro,
A.M., Tort, A.B., Neto, A.D., and Ribeiro, S. (2015).
Synaptic Homeostasis and restructuring across
the sleep-wake cycle. PLoS Comput. Biol. 11,
e1004241.
Blainville, H.M.D. (1823). de., Me
´moire sur les
espe
`ces du genre calmar (Loligo Lamarck).
Journal de Physique, de Chimie, d’Histoire
naturelle et des Arts 96, 116–135, Available online
at. https://www.biodiversitylibrary.org/page/
6176264.
Boyce, R., Glasgow, S.D., Williams, S., and
Adamantidis, A. (2016). Causal evidence for the
role of REM sleep theta rhythm in contextual
memory consolidation. Science 352, 812–816.
Brown, E.R., Piscopo, S., De Stefano, R., and
Giuditta, A. (2006). Brain and behavioural
evidence for rest-activity cycles in Octopus
vulgaris. Behav. Brain Res. 172, 355–359.
Campbell, S.S., and Toblew, I. (1984). Animal
Sleep: a review of sleep duration across
phylogeny. Neurosci. Biobehav. Rev. 8,
269–300.
Cooper, J.G. (1863). On new or rare Mollusca
inhabiting the coast of California.— No. II.
Proceedings of the California Academy of
Sciences. ser. 1, 56–60.
Cuvier, G. (1797). Tableau e
´le
´mentaire de
l’Histoire Nature
´lle des animaux. Baudouin,
Imprimeur du Corps le
´gislatif et de l’Institut
national, place du Carrousel, N. 662, an 6, Paris:
i-xvi, 1–710, Pl. 1-14.
Delorme, F., Vimont, P., and Jouvet, D. (1964).
Statistical study of the wakefulness-sleep cycle in
the cat. C R. Seances Soc. Biol. Fil. 158, 2128–
2130.
Dement, W. (1958). The occurrence of low
voltage, fast, electroencephalogram patterns
during behavioral sleep in the cat.
Electroencephalogr. Clin. Neurophysiol. 10,
291–296.
Dillon, R.F., and Webb, W.B. (1965). Threshold of
arousal from "activated" sleep in the rat.
J. Comp. Physiol. Psychol. 59, 446–447.
Dume
´ril, A.M., and Bibron, G. (1839). ‘‘Le
sauvegarde de Me
´rian, Salvator Merianæ,
Nobis’’. Erpe
´tologie ge
´ne
´rale ou Histoire
naturelle comple
`te des reptiles. 5. (Paris: Roret),
pp. 85–90.
Frank, M.G., Waldrop, R.H., Dumoulin, M., Aton,
S., and Boal, J.G. (2012). A preliminary analysis of
sleep-like states in the cuttlefish Sepia officinalis.
PLoS One 7, e38125.
ll
OPEN ACCESS
18 iScience --, 102223, --, 2021
Please cite this article in press as: Medeiros et al., Cyclic alternation of quiet and active sleep states in the octopus, iScience (2021), https://
doi.org/10.1016/j.isci.2021.102223
iScienc
e
Article
Gervasoni, D., Lin, S.C., Ribeiro, S., Soares, E.S.,
Pantoja, J., and Nicolelis, M.A. (2004). Global
forebrain dynamics predict rat behavioral states
and their transitions. J. Neurosci. 24, 11137–
11147.
Greenspan, R.J., Tononi, G., Cirelli, C., and Shaw,
P.J. (2001). Sleep and the fruit fly. Trends
Neurosci. 24, 142–145.
Gutnick, T., Shomrat, T., Mather, J.A., and Kuba,
M.J. (2016). The cephalopod brain: Motion
control, learning, and cognition. In Physiology of
Molluscs: A Collection of Selected Reviews, S.
Saleuddin and S. Mukai, eds. (Apple Academic
Press), 139–177.
Hablitz, L.M., Vinitsky, H.S., Sun, Q., Staeger, F.F.,
Sigurdsson, B., Mortensen, K.N., Lilius, T.O., and
Nedergaard, M. (2019). Increased glymphatic
influx is correlated with high EEG delta power and
low heart rate in mice under anesthesia. Sci. Adv.
5, eaav5447.
Hendricks, J.C., Finn, S.M., Panckeri, K.A.,
Chavkin, J., Williams, J.A., Sehgal, A., and Pack,
A.L. (2000). Rest in Drosophila is a sleeplike state.
Neuron 25, 129–138.
Hobson, J.A., and Mccarley, R.W. (1971). Cortical
unit activity in sleep and waking.
Electroencephalogr. Clin. Neurophysiol. 30,
97–112.
Hochner, B., Brown, E.R., Langella, M., Shomrat,
T., and Fiorito, G. (2003). A learning and memory
area in the octopus brain manifests a vertebrate-
like long-term potentiation. J. Neurophysiol. 90,
3547–3554.
Iglesias, T.L., Boal, J.G., Frank, M.G., Zeil, J., and
Hanlon, R.T. (2019). Cyclic nature of the REM-like
state in the cuttlefish Sepia officinalis. J. Exp. Biol.
222, 174862.
Leite and Haimovici, 2008, https://doi.org/10.
1093/mollus/eym050.
Libourel, P.A., Barrillot, B., Arthaud, S., Massot,
B., Morel, A.L., Beuf, O., Herrel, A., and Luppi,
P.H. (2018). Partial homologies between sleep
states in lizards, mammals, and birds suggest a
complex evolution of sleep states in amniotes.
PLoS Biol. 16, e2005982.
Lima, F.D., Leite, T.S., Haimovici, M., and Oliveira,
J.E.L. (2014). Gonadal development and
reproductive strategies of the tropical octopus
(Octopus insularis) in northeast Brazil.
Hydrobiologia 725, 7–21.
Linnaeus, C. (1758). Systema Naturae per regna
tria naturae, secundum classes, ordines, genera,
species, cum characteribus, differentiis,
synonymis, locis. Editio decima, reformata [10th
revised edition], vol. 1: 824 pp. (Holmiae:
Laurentius Salvius), p. 658, Available online at.
https://biodiversitylibrary.org/page/726886.
Low, P.S., Shank, S.S., Sejnowski, T.J., and
Margoliash, D. (2008). Mammalian-like features of
sleep structure in zebra finches. Proc. Natl. Acad.
Sci. U S A 105, 9081–9086.
Lyamin, O.I., Mukhametov, L.M., Siegel, J.M.,
Nazarenko, E.A., Polyakova, I.G., and Shpak, O.V.
(2002). Unihemispheric slow wave sleep and the
state of the eyes in a white whale. Behav. Brain
Res. 129, 125–129.
Mascetti, G.G. (2016). Unihemispheric sleep and
asymmetrical sleep: behavioral,
neurophysiological, and functional perspectives.
Nat. Sci. Sleep 8, 221–238.
Mather, J. (2016). Mating games squid play:
reproductive behavior and sexual skin displays in
Caribbean reef squid Sepioteuthis sepioidea.
Mar. Freshw. Behav. Physiol. 49, 359–373.
Meisel, D.V., Byrne, R.A., Mather, J.A., and Kuba,
M. (2011). Behavioral sleep in Octopus vulgaris.
Vie et Milieu 61, 185–190.
Muzet, A., Naitoh, P., Townsend, R.E., and
Johnson, L.C. (1972). Body movements during
sleep as a predictor of stage change. Psychon.
Sci. 29, 7–10.
Neckelmann, D., and Ursin, R. (1993). Sleep
stages and EEG power spectrum in relation to
acoustical stimulus arousal threshold in the rat.
Sleep 16, 467–477.
Nixon, M., and Young, J.Z. (2003). The Brains and
Lives of Cephalopods (Oxford University Press).
Noda, H., Manohar, S., and Adey, W.R. (1969).
Spontaneous activity of cat hippocampal neurons
in sleep and wakefulness. Exp. Neurol. 24,
217–231.
Norimoto, H., Fenk, L.A., Li, H., Tosches, M.A.,
Gallego-Flores, T., Hain, D., Reiter, S., Kobayashi,
R., Macias, A., Arends, A., et al. (2020). A
Claustrum in reptiles and its role in slow-wave
sleep. Nature 578, 413–418.
Rattenborg, N.C., Amlaner, C.J., and Lima, S.L.
(2000). Behavioral, neurophysiological and
evolutionary perspectives on unihemispheric
sleep. Neurosci. Biobehav. Rev. 24, 817–842.
Rattenborg, N.C., Lima, S.L., and Amlaner, C.J.
(1999). Half-awake to the risk of predation. Nature
397, 397–398.
Rattenborg, N.C., Voirin, B., Cruz, S.M., Tisdale,
R., Dell’omo, G., Lipp, H.P., Wikelski, M., and
Vyssotski, A.L. (2016). Evidence that birds sleep in
mid-flight. Nat. Commun. 7, 12468.
Reed, C.M. (1995). The ultrastructure and
innervation of muscles controlling
chromatophore expansion in the squid, Loligo
vulgaris. Cell Tissue Res. 282, 503–512.
Reiter, S., Hu
¨lsdunk, P., Woo, T., Lauterbach, M.A.,
Eberle, J.S., Akay, L.A., Longo, A., Meier-Credo,
J., Kretschmer, F., Langer, J.D., et al. (2018).
Elucidating the control and development of skin
patterning in cuttlefish. Nature 562, 361–366.
Ridgway, S., Carder, D., Finneran, J., Keogh, M.,
Kamolnick, T., Todd, M., and Goldblatt, A. (2006).
Dolphin continuous auditory vigilance for five
days. J. Exp. Biol. 209, 3621–3628.
Shein-Idelson, M., Ondracek, J.M., Liaw, H.P.,
Reiter, S., and Laurent, G. (2016). Slow waves,
sharp waves, ripples, and REM in sleeping
dragons. Science 352, 590–595.
Shomrat, T., Turchetti-Maia, A.L., Stern-Mentch,
N., Basil, J.A., and Hochner, B. (2015). The vertical
lobe of cephalopods: an attractive brain structure
for understanding the evolution of advanced
learning and memory systems. J. Comp. Physiol.
A201, 947–956.
Shu, D.G., Chen, L., Han, J., and Zhang, X.L.
(2001). An early cambrian tunicate from China.
Nature 411, 472–473.
Stephenson, R., and Lewis, V. (2011). Behavioural
evidence for a sleep-like quiescent state in a
pulmonate mollusc, Lymnaea stagnalis
(Linnaeus). J. Exp. Biol. 214, 747–756.
Steriade, M., Mccormick, D.A., and Sejnowski,
T.J. (1993). Thalamocortical oscillations in the
sleeping and aroused brain. Science 262,
679–685.
Tainton-Heap, L.A.L., Kirszenblat, L.C., Notaras,
E.T., Feng, K., Shaw, P.J., and Van Swinderen, B.
(2021). A paradoxical kind of sleep in Drosophila
melanogaster. Curr. Biol. 31, 578–590.
Timo-Iaria, C., Negrao, N., Schmidek, W.R.,
Hoshino, K., Lobato De Menezes, C.E., and Leme
Da Rocha, T. (1970). Phases and states of sleep in
the rat. Physiol. Behav. 5, 1057–1062.
Trachsel, L., Tobler, I., Achermann, P., and
Borbe
´ly, A.A. (1991). Sleep continuity and the
REM-nonREM cycle in the rat under baseline
conditions and after sleep deprivation. Physiol.
Behav. 49, 575–580.
Vanderwolf, C.H. (1969). Hippocampal electrical
activity and voluntary movement in the rat.
Electroencephalogr. Clin. Neurophysiol. 26,
407–418.
Vitti, J.J. (2013). Cephalopod cognition in an
evolutionary context: implications for ethology.
Biosemiotics 6, 393.
Vorster, A.P.A., and Born, J. (2017). Sleep
supports inhibitory operant conditioning memory
in Aplysia. Learn. Mem. 24, 252–256.
Walker, J.M., and Berger, R.J. (1972). Sleep in the
domestic pigeon (Columba livia). Behav. Biol. 7,
195–203.
Xie, L., Kang, H., Xu, Q., Chen, M.J., Liao, Y.,
Thiyagarajan, M., O’donnell, J., Christensen, D.J.,
Nicholson, C., Iliff, J.J., et al. (2013). Sleep drives
metabolite clearance from the adult brain.
Science 342, 373–377.
ll
OPEN ACCESS
iScience --, 102223, --, 2021 19
Please cite this article in press as: Medeiros et al., Cyclic alternation of quiet and active sleep states in the octopus, iScience (2021), https://
doi.org/10.1016/j.isci.2021.102223
iScienc
e
Article
iScience, Volume --
Supplemental information
Cyclic alternation of quiet
and active sleep states in the octopus
Sylvia Lima de Souza Medeiros, Mizziara Marlen Matias de Paiva, Paulo Henrique
Lopes, Wilfredo Blanco, Françoise Dantas de Lima, Jaime Bruno Cirne de
Oliveira, Inácio Gomes Medeiros, Eduardo Bouth Sequerra, Sandro de Souza, Tatiana
Silva Leite, and Sidarta Ribeiro
2
Supplementary Figure S1. Duration distributions of ‘Quiet sleep’ and ‘Active
sleep’ episodes, according to the states that either preceded or succeeded
them, Related to Figure 5.
3
Kernel Density Estimation was used to compare the distributions of durations
of: (A) ‘Quiet sleep’ immediately and not immediately preceding ‘Active sleep’
(black vertical line indicates the average of modes from two distributions). (B)
‘Quiet sleep’ immediately succeeding and not immediately succeeding ‘Active
sleep’. (C) ‘Active sleep’ immediately and not immediately preceding ‘Quiet
sleep’. (D) ‘Active sleep’ immediately and not immediately succeeding ‘Quiet
sleep’. The x-axis displays durations in min, and the y-axis displays the
probability of a given duration to occur.
4
Supplementary Figure S2. Episode duration and intervals between episodes for
‘Quiet sleep’ (long and short) and ‘Active sleep’, Related to figure 6.
5
Each background color represents one individual octopus. (A) The top panel shows
the durations of the ‘Quiet sleep’ episodes, with a histogram on the right. The bottom
6
panel shows the bloxpot and histogram of these durations for each octopus. (B) The
top panel shows the duration of the intervals between all the ‘Quiet sleep’ episodes,
with a histogram on the right. The bottom panel shows the bloxpot and histogram of
these intervals durations for each octopus. (C) Top panel represents the durations of
each ‘Active sleep’ episode, with histograms on the right. The bottom panel shows the
boxplot and histogram of ‘Active sleep’ episodes durations for each octopus.
Supplementary Figure S3. Hypnogram with measurements of general skin color
dynamics, localized skin patterning dynamics, eye and mantle movements,
Related to figure 7.
7
8
The figure depicts the behaviors of one representative animal (octopus 3) along 200
min of diurnal recordings, beginning at 06:36:50 and finishing at 09:56:50. The plot of
general skin color dynamics depicts the mean color from two regions of interest
delimited over the head and dorsal mantle (detailed in Figure 8). The plot of localized
skin patterning dynamics shows changes in skin texture represented by the variation
of the number of white pixels captured by the Canny algorithm along time. For both
skin dynamic analyses, a marked variation of color and texture concomitant with the
occurrence of ‘Active sleep’ can be observed. The eye position plots show the
movements tracked for the right and left eyes. The mantle position plot shows mantle
movements represented by the sum of seven different regions of the mantle (detailed
in Figure 11).
9
Supplementary Figure S4. General skin color dynamics across ‘Quiet sleep’,
‘Active sleep’ and the ‘Alert’ state, Related to Figure 8.
(A), (D) and (G) is showing changes in skin color for octopuses 1, 2 and 4
10
respectively, represented by the variation of the ROI’s mean color (y-axis) along time
(x-axis) of five transitions among ‘Quiet sleep’ (blue zone), ‘Active sleep’ (green zone)
and ‘Alert’ state (orange zone). The intervals analyzed for ‘Quiet sleep’ and the ‘Alert’
state had the same duration of the ‘Active sleep’ that occurred between them.
Besides, the 5 examples are organized from the shortest ‘Active sleep’ episode, on
top, to the longest episode, on the bottom. The green line represents the head’s color
and the blue line the mantle’s color. The lines are well followed by the Gaussian filter
(red line). All the behavior transitions have a similar profile, with little value variation
during ‘Quiet sleep’ and the ‘Alert’ state, and with a marked variation during ‘Active
sleep’, which comprised both extreme values (highlighted by the blue and red dots).
(B), (E) and (H) are showing the picture of the octopuses with the ROI’s selected for
head (green) and mantle (blue) when the animal color achieved its maximum value
(more pale) and minimum value (more dark). (C), (F) and (I) The scatterplots show
the medians of delta (500 ms) Gaussian values for the color analysis of each behavior
for the region of head (octopuses 1, 2 and 4) and mantle (octopus 4).
11
Supplementary Figure S5. Local skin dynamics across the states ‘Quiet sleep’,
‘Active sleep’ and the ‘Alert’ state for octopuses 1, 2 and 4, Related to figure 9.
(A), (D) and (G) is showing changes in skin texture and color pattern complexity for
12
octopuses 1, 2 and 4 respectively, which is represented by the variation of the
number of white pixels (y-axis) along time (x-axis) of five transitions among ‘Quiet
sleep’ (blue zone), ‘Active sleep’ (green zone) and the ‘Alert’ state (orange zone). The
intervals analyzed for ‘Quiet sleep’ and ‘Alert’ had the same duration of the ‘Active
sleep’ that occurred between them. Besides, the 5 examples are organized from the
shortest ‘Active sleep’, on top, to the longest, on the bottom. The red line on the graph
is a Gaussian filter and follows closely the real data, represented by the black line. All
the behavior transitions have a similar graph profile, with little values variation during
‘Quiet sleep’ and ‘Alert’ state and with a marked variation on ‘Active sleep’, so that
both extreme values (blue and red dots) occurred during ‘Active sleep’ and all the
highest values occurring after the lowest values. (B), (E) and (H) show the rectangular
area cropped from the analyzed frames from octopuses 1, 2 and 4, respectively,
including: (1) the mask demonstrated by the white regions which delimited the animal
body, excluding arms and background; (2) the moment on ‘Active sleep’ when the
animal is with lowest and highest body pattern complexity (texture and color),
represented by the minimal and maximum number of white pixels captured by the
Canny borders algorithm (blue and red dots on the graphs). (C), (F) and (I) represents
the scatterplot graphs of octopuses 1, 2 and 4 respectively, showing the medians of
delta (500 ms) Gaussian values for skin dynamics analysis. The median values of
‘Active sleep’ are mostly higher than other behaviors (excepted one transition on
octopus 1 where the value does not change across the behavior).
13
Supplemental tables
Supplemental Table S1. Average of the absolute frequency per day and median
duration of the behavioral states, Related to Figure 4.
States
Average frequency/day
Median Duration (min)
Active
10.00
2.32
Alert
35.56
2.83
Quiet with Open Pupil
31.31
3.61
Quiet sleep
17.00
6.92
Quiet with Half and
Half
27.44
0.17
Active sleep
7.81
0.68
Quiet with Only One
Eye Movement
0.50
0.92
Supplemental Table S2. P values for comparisons between states of the
duration of the states observed. The values in bold indicate significant
differences (Kruskal-Wallis test), Related to figure 4.
Active
Alert
QOP
QHH
Quiet sleep
Active sleep
Alert
0.28
-
-
-
-
-
Quiet with Open
Pupil
0.09
1
-
-
-
-
Quiet with Half and
Half
0.00
0
0
-
-
-
Quiet sleep
0
0
0
0
-
-
Active sleep
0
0
0
0
0
-
Quiet with Only
One Eye Movement
0.33
0.12
0.16
0.01
0.01
1
14
Supplemental Table S3. Cyclicity of long ‘Quiet sleep’ episodes, for a total of 142
episodes analyzed, Related to Figure 6.
Number of
Intervals
Percent
Episodes
74
100%
6.77, 7.42, 7.43, 7.58, 7.67, 8.0, 8.13, 8.2, 8.35,
9.08, 9.12, 9.2, 9.23, 9.25, 9.25, 9.88, 10.13, 10.22,
10.45, 10.5, 11.48, 12.12, 12.25, 13.42, 13.85,
14.17, 14.17, 14.25, 14.4, 15.15, 15.47, 15.93,
16.35, 16.4, 17.85, 17.93, 18.3, 20.6, 21.28, 21.57,
22.45, 23.83, 24.58, 25.22, 25.48, 27.05, 27.07,
28.07, 28.37, 28.55, 28.58, 28.85, 29.27, 32.02,
32.63, 34.15, 35.63, 37.67, 38.22, 38.9, 41.2, 42.15,
42.85, 46.73, 49.07, 49.32, 49.52, 57.2, 59.55,
59.62, 59.77, 66.3, 108.62, 173.47.
45
60%
6.77, 7.42, 7.43, 7.58, 7.67, 8.0, 8.13, 8.2, 8.35,
9.08, 9.12, 9.2, 9.23, 9.25, 9.25, 9.88, 10.13, 10.22,
10.45, 10.5, 11.48, 12.12, 12.25, 13.42, 13.85,
14.17, 14.17, 14.25, 14.4, 15.15, 15.47, 15.93,
16.35, 16.4, 17.85, 17.93, 18.3, 20.6, 21.28, 21.57,
22.45, 23.83, 24.58, 25.22, 25.48.
15
Supplemental Table S4. Cyclicity of ‘Active sleep’ episodes, for a total of 79
intervals analyzed, Related to Figure 6.
Number
of
Intervals
Percent
Episodes
79
100%
20.83, 22.42, 27.83, 28.45, 29.58, 29.7, 30.17,
30.27, 31.97, 32.23, 32.55, 32.62, 32.67, 32.75,
32.92, 32.93, 32.98, 33.17, 33.23, 33.23, 33.38,
33.77, 34.05, 34.1, 34.15, 34.38, 34.6, 34.75, 35.45,
35.48, 35.48, 35.52, 35.85, 36.0, 36.05, 36.18,
36.32, 36.4, 36.45, 36.92, 36.97, 37.12, 37.95,
38.32, 38.32, 38.62, 39.02, 39.05, 39.15, 40.67,
41.25, 42.57, 44.13, 44.95, 46.27, 46.28, 50.4, 51.9,
54.22, 55.25, 56.05, 58.35, 60.77, 61.48, 61.95,
62.15, 68.0, 70.38, 75.53, 75.7, 78.15, 81.32, 81.52,
87.55, 96.68, 99.28, 103.03, 227.79, 150.45.
47
61%
29.58, 33.38, 34.15, 34.75, 34.38, 27.83, 38.32,
32.67, 32.75, 28.45, 32.23, 33.17, 32.62, 31.97,
38.62, 34.1, 32.93, 35.48, 34.05, 33.23, 30.17, 29.7,
35.52, 35.48, 35.45, 36.05, 34.6, 35.85, 33.77,
39.15, 36.18, 36.45, 37.12, 32.55, 36.4, 33.23,
32.92, 39.05, 36.32, 30.27, 39.02, 36.97, 37.95,
38.32, 36, 36.92, 32.98.
16
Transparent Methods
Animals
The general guidelines for care and husbandry of the octopuses during this
study were in accordance with the Directive 2010/63/E.U. Brazil’s animal welfare
regulation does not include cephalopods, so this study did not require a protocol or
approval number. The capture of specimens was approved by the Instituto Chico
Mendes de Conservação da Biodiversidade (ICMBIO) through the Permanent License
for the collection of zoological material number 33754-1.
Four specimens of Octopus insularis, a tropical species from Western Atlantic
(Leite et al., 2008, Lima et al., 2017) one male (155g) and three females (300g, 450g,
535g) were collected by S.L.S.M. and M.M.M.P. at the Buzios Beach (Latitude
6°0'14.29"S, Longitude 35° 6'21.01"W) between March 2018 and February 2019. The
animals were brought to the laboratory at the Federal University of Rio Grande do
Norte, where they were acclimatized for 10 days before the experiments began. The
octopuses were kept individually in a 1.0 x 0.7 x 0.6 meters glass tank, which was part
of a closed circulation system including a sump comprising protein skimmers, passive
filters, and a de-nitrification tank to filter the water.
The animals were provided with a semi-natural enriched environment, including
sand, stones, and PVC pipes, with a water temperature at 25 ºC. In natural
environments, these animals are commonly found in the temperature range from 24 ºC
to 29 ºC (Longo et al., 2015, Leite et al., 2009). S.L.S.M. and M.M.M.P. fed each
octopus twice a day (9:00 and 17:00) with defrosted crabs or shrimps. A 12h/12h cycle
light/dark schedule was established using a timer, with lights on at 5:50 and off at
17:50. At the end of the studies, the animals were anesthetized with MgCl2 3.5% in
seawater and euthanized by de-cerebration.
17
According to their weight, we numbered the octopuses from the lightest
(octopus 1 - male, 155g) to the heaviest (octopus 4 - female, 535g).
Video Recording and Ethogram
Two HD cameras (Sony-Handycam HDR-CX550 and Intelbras IC3) were
placed at the aquarium’s opposite sides to record behaviors in colored videos. Each
animal was video recorded during the 12 h of light period (with few interruptions due to
camera adjustments), for four days. Light-time recordings were chosen because the
Octopus vulgaris from Bermuda is mostly inactive during the day in its natural
environment (Mather, 1988), and recent studies suggest that the common octopus
found in this region is actually the O. insularis (Lima et al., 2017). Furthermore,
according to field research developed in Brazil, O. insularis has crepuscular habits,
being less active during the day (Leite, personal communication). From the visual
inspection of the videos, an ethogram was created for the behaviors displayed by each
animal. From a total of 192h of recordings, 11h and 11min were discarded due to
memory card limitations and feeding periods, when the dynamics of spontaneous
behavioral transitions was disturbed. The ethogram with the sequence of behaviors
and each behavior’s duration was elaborated from the visual inspection of 180h and
49min of video recordings (n = 4 specimens, 48 h from octopus 1, 46h and 46min
from octopus 2, 47h and 32min from octopus 3, 38h and 30min from octopus 4).
The behaviors were initially classified within the categories Active, Alert, and
Quiet, as described in (Brown et al., 2006). We then investigated behavioral changes
during the Quiescence state that could be similar to ‘Quiet with closed pupil’ or to
‘Quiet with the half and half skin pattern’ (‘QHH’) observed in Octopus vulgaris (Brown
et al., 2006, Meisel et al., 2011), as well as the ‘Active sleep’ observed in Sepia
18
officinalis (Frank et al., 2012, Iglesias et al., 2019), and searched for any other
quiescence state not yet described in cephalopods. The duration and frequency of
these behaviors were assessed. We also asked whether there was a sequential
pattern in their occurrence. For comparison, we tried to record all the animals during
the night. However, the camera’s infrared light has a short-range through the water
column, and hours of recording were lost when animals moved away from the focus.
Despite this difficulty, we recorded 114 h from one animal (octopus 3), which tended to
stay quiet in the aquarium corner, maintaining its position relative to the camera placed
at a close distance.
Measurements of arousal threshold
Octopuses 2, 3, and 4 were visually stimulated to quantify the arousal threshold
across distinct behavioral states. The visual stimulation (detailed below) was
performed when the animal was in the ‘Alert’ state and during the different states
initially categorized within the Quiescence state (Table 1).
Octopus 3 initially failed to respond to the visual stimulus even when awake
and alert. Thus, in the case of this animal, we decided to apply a vibratory stimulus
(details below) to confirm the lack of sensory responses across the same behavioral
states investigated by using the visual stimulus. Two weeks after the vibratory
stimulation test, we attempted the visual stimulation again with the same octopus and
were then able to elicit arousal responses. Inspections of the video recordings suggest
that octopus 3 was still in the process of habituation when first subjected to the visual
stimulation test, which led to a freezing behavior during the trials. We kept this animal
in the study because it was in good health, as indicated by key physiological
19
parameters (Fiorito et al., 2015). Octopuses commonly display such intraspecific
behavioral variation (Mather and Anderson, 1993).
The behaviors considered as reactions to sensory stimulation were changes in
body pattern (color and texture) and sudden alterations in posture. In the visual
stimulation case, reactions also included orienting the head to look at the screen and
attempts to attack the crabs moving on the screen (see below). The visual and
vibratory stimuli were mild, near the perceptual threshold, to properly discriminate the
different states’ levels of responsiveness.
During these sensory stimulation tests, each octopus was kept alone in the
experimental room. Simultaneously, the behavior was recorded through a camera that
transmitted video signals in real-time to a monitor in the next room, where a
researcher (either S.L.S.M. or M.M.M.P.) supervised the experiment. Both tests
occurred between 8:00 and 16:30.
Visual stimulation test to quantify arousal threshold
A computer screen (25.8 x 14.5 cm, pixel resolution 1366 x 768, maximum
illuminance of 71 lux, average illuminance of 32 lux) was placed in contact with the
wall of the aquarium as close as possible to the octopus, so that the animal could
easily visualize it (Figure 1, Visual stimulation). When the animal was in one of the
behavioral states of interest (Table 1), a 50 s video showing live crabs moving was
initiated remotely. The time taken by the octopus to react to the stimulus was
assessed by video inspection. When the animals failed to show any reaction
throughout the 50 s, the trial was scored as ‘absence of reaction’.
20
Vibratory stimulation test to quantify arousal threshold
A rubber hammer was attached to the aquarium’s upper corner and tied to a
nylon wire passed to the experimenter's room through a window. This wire had four
marks so that pulling it up to one of these marks would set the hammer at the angles
of 10°, 20°, 30°, and 40° respectively, to hit the aquarium wall with four different,
increasing intensities (Figure 1, Vibratory stimulation). The hammer was triggered
when the animal was in each of the behaviors of interest mentioned above. The
hammer was only triggered to strike with a more vigorous intensity when the animal
failed to respond to the previous (weaker) stimulus. Whenever the animal elicited a
behavioral response before the hammer hit with the smallest intensity (presumably
due to visual perception of the hammer’s movement), the arousal threshold was
recorded as level 1. If the behavioral response was elicited after the hit with the angle
of 10°, the arousal threshold was recorded as level 2, and so on for the other angles.
When the animal failed to react to all intensities (up to the 40° angle), the trial was
considered as ‘Maximum Impact with an absence of reaction’. This procedure was
video recorded for subsequent analysis.
Behavioral transition graphs
Behavioral transition graphs were elaborated from the ethogram using Python
version 3.6.9 (Rossum and Drake Jr., 2011) to quantify the sequential order of the
behaviors observed in the video recordings (n=4, a total of 180h and 49min of
recordings during the light phase, over four days). The directed graphs were used to
characterize all the behavioral transitions observed and to determine whether there
are typical behavioral sequences across the wake-sleep cycle.
21
Quantification of general skin color dynamics
We used pixel color intensity values to compare chromatophore changes across
behavioral states and assess whether each behavior has specific characteristics and
different variation rates. The original videos were recorded at the rate of 29.97
frames/s in the RGB (Red, Green and Blue) color system. Therefore, to measure pixel
color intensities we used brightness (i.e., amount of light), which is represented by
range of integer values (grayscale), going from 0 (black color, absence of brightness)
to 255 (white color, highest brightness). This range of values came from the
conversion of the three RGB values, present in the original image (video frame), to
grayscale. For the conversion from RGB to grayscale, we used the function cvtColor()
from the OpenCv library (Bradski and Kaehler, 2008). We delimited two square
regions of interest (ROIs; 5x5, a total of 25 pixels) located on the octopus' body, over
the dorsal mantle and the head. These regions were chosen because 1) they contain
small and more numerous chromatophores with multiple innervations in comparison
with other areas, thus providing a reliable sample of the highly complex changes in
whole-body color (Leite and Mather, 2008) and 2) they remain almost in a static
position when the octopus displays the sleep-like behaviors of interest. The choice of
dimension was based on a previous study of Sepia officinalis (Iglesias et al., 2019).
The mean brightness of the region of interest, ranging from 0 (black) to 255 (white),
was calculated in each video frame for each behavior of interest.
Quantification of local skin patterning dynamics across behavioral states
Localized skin dynamics, caused by texture changes due to extension and
retraction of the papillae and changes in color patterning within specific body regions,
were investigated using the Canny algorithm for edge detection (Green, 2002). To
22
execute the Canny algorithm more precisely and faster within each ROI, we carried
out two steps: (1) every video frame was cropped to obtain a smaller rectangular
region delimitating exactly where the animal was located; (2) a customized mask was
applied to restrict the ROI further to include the arms but not the arm tips and suckers,
which were prone to movement artifacts (Figure S5). Edge changes across frames of
interest were calculated as the difference in the number of white pixels generated by
the Canny algorithm. We smoothed this profile using a Gaussian filter (sigma = 2), with
a frame-to-frame interval of 500 ms.
Quantification of eye and mantle movements
Eye and mantle movements were tracked using DeepLabCut (Mathis et al.,
2018), with two marks (one for each eye) and seven marks for the mantle. Then, the
sum of the position variation of the left and right eye and the mean values for the
seven marks on the mantle were used for the analysis. A 50 layers deep convolutional
neural network (ResNet-50) was employed for automatic marking, using 30 frames per
video for manual marking (to produce train and test datasets) and a training fraction of
70%. To remove artifacts, coordinates > or < 2 SD from the mean were replaced by
the average value of the segment considered. Eye and mantle movements across
frames of interest were calculated using a Gaussian-smoothed time series (sigma = 2)
of the Euclidean distances between X-Y coordinate pairs, with a frame-to-frame
interval of 300 ms. After performing a parametrical analysis, this interval was chosen
to search for the time interval that maximized differences across states (data not
shown). Eye movements were quantified in all animals, but mantle movements were
quantified in only two animals (octopuses 3 and 4) because the other two animals
23
preferred using the PVC pipe as a shelter, so the mantle was not visible during their
quiescence behaviors.
Statistical Analysis
The assumptions of normality of the distribution and homogeneity of the
variance were examined with the Kolmogorov–Smirnov and Levene tests (Zar, 2007).
Since none of the datasets met the parametric assumptions, non-parametric tests for
repeated measures were conducted. To verify whether there were significant
differences in episode duration, texture and color skin dynamics, eye and mantle
movement across behavioral states, the Kruskal-Wallis test was used. Pairwise
comparisons between states of interest were performed using the Wilcoxon signed-
ranks test, with Bonferroni correction for multiple comparisons. The same statistical
tests were carried out to analyze response latencies after visual stimulation across
behavioral states and a chi-squared test to compare the states’ absence of reaction. In
the case of the vibratory stimulation, which was performed only in the animal that was
initially unresponsive to visual stimulation, chi-square tests were performed to verify
differences in the number of observations at each intensity level across behavioral
states. All statistical analyses were conducted using the software R version 3.3.2 or
Python version 3.6.9.
24
Supplemental References
BRADSKI, G. & KAEHLER, A. 2008. Learning OpenCV: Computer Vision with the OpenCV
Library, Sebastopol, O’Reilly Media.
BROWN, E. R., PISCOPO, S., DE STEFANO, R. & GIUDITTA, A. 2006. Brain and
behavioural evidence for rest-activity cycles in Octopus vulgaris. Behav Brain Res,
172, 355-9.
FIORITO, G., AFFUSO, A., BASIL, J., A., C., DE GIROLAMO, P., D'ANGELO, L., DICKEL, L.,
GESTAL, C., GRASSO, F., KUBA, M., MARK, F., MELILLO, D., OSORIO, D.,
PERKINS, K., PONTE, G., SHASHAR, N., SMITH, D., SMITH, J. & ANDREWS, P. L.
2015. Guidelines for the Care and Welfare of Cephalopods in Research -A consensus
based on an initiative by CephRes, FELASA and the Boyd Group. Lab Anim, 49, 1-90.
FRANK, M. G., WALDROP, R. H., DUMOULIN, M., ATON, S. & BOAL, J. G. 2012. A
Preliminary Analysis of Sleep-Like States in the Cuttlefish Sepia officinalis. PLoS ONE,
7, e38125.
GREEN, B. 2002. Canny edge detection tutorial [Online]. Available:
https://web.archive.org/web/20160324173252/http://dasl.mem.drexel.edu/alumni/bGree
n/www.pages.drexel.edu/_weg22/can_tut.html [Accessed].
IGLESIAS, T. L., BOAL, J. G., FRANK, M. G., ZEIL, J. & HANLON, R. T. 2019. Cyclic nature
of the REM-like state in the cuttlefish Sepia officinalis. Journal of Experimental Biology,
222, 174862.
LEITE, T. S., HAIMOVICI, M., MATHER, J. & LINS OLIVEIRA, J. E. 2009. Habitat, distribution,
and abundance of the commercial octopus (Octopus insularis) in a tropical oceanic
island, Brazil: Information for management of an artisanal fishery inside a marine
protected area. Fisheries Research, 98, 85-91.
LEITE, T. S., HAIMOVICI, M., MOLINA, W. & WARNKE, K. 2008. Morphological and genetic
description of Octopus insularis, a new cryptic species in the Octopus vulgaris complex
(Cephalopoda: Octopodidae) from the tropical southwestern Atlantic. J Molluscan Stud,
74, 63–74.
LEITE, T. S. & MATHER, J. 2008. A new approach to octopuses’ body pattern analysis: a
framework for taxonomy and behavioral studies. Amer Malac Bull, 24, 31–41.
LIMA, F. D., BERBEL-FILHO, W. M., LEITE, T. S., ROSAS, C. & LIMA, S. M. Q. 2017.
Occurrence of Octopus insularis Leite and Haimovici, 2008 in the Tropical
Northwestern Atlantic and implications of species misidentification to octopus fisheries
management. Mar Biodiv, 47, 723–734.
LONGO, G. O., MORAIS, R. A., MARTINS, C. D., MENDES, T. C., AUED, A. W., CÂNDIDO,
D. V., DE OLIVEIRA, J. C., NUNES, L. T., FONTOURA, L., SISSINI, M. N.,
TESCHIMA, M. M., SILVA, M., RAMLOV, F., GOUVEA, L. P., FERREIRA, C. E.,
SEGAL, B., HORTA, P. A. & FLOETER, S. R. 2015. Between-Habitat Variation of
Benthic Cover, Reef Fish Assemblage and Feeding Pressure on the Benthos at the
Only Atoll in South Atlantic: Rocas Atoll, NE Brazil. PLoS One. , 10, e0127176.
MATHER, J. A. 1988. Daytime activity of juvenile octopus vulgaris in bermuda. Malacologia,
29, 69-76.
MATHER, J. A. & ANDERSON, R. C. 1993. Personalities of octopuses (Octopus rubescens).
Journal of Comparative Psychology, 107, 336-340.
MATHIS, A., MAMIDANNA, P., CURY, K. M., ABE, T., MURTHY, V. N., MATHIS, M. W. &
BETHGE, M. 2018. DeepLabCut: markerless pose estimation of user-defined body
parts with deep learning. Nat. Neurosci., 21, 1281–1289.
MEISEL, D. V., BYRNE, R. A., MATHER, J. A. & KUBA, M. 2011. Behavioral sleep in Octopus
vulgaris. Vie et Milieu, 61, 185-190.
ROSSUM, G. V. & DRAKE JR., F. L. 2011. Introduction to Python, Bristol, UK, Network
Theory Limited.
ZAR, J. H. 2007. Biostatistical Analysis, New Jersey, Prentice Hall.
