Physiology & Behavior 127 (2014) 54–63

Contents lists available at ScienceDirect

Physiology & Behavior

j ourna l homepage: www.e lsev ie r .com/ locate /phb

Effects of social defeat on sleep and behaviour: Importance of theconfrontational behaviour

Anne Marie Kinn Rød a,⁎, Robert Murison a, Jelena Mrdalj a, Anne Marita Milde a, Finn Konow Jellestad a,Leif Arvid Øvernes a, Janne Grønli a,b

a Department of Biological and Medical Psychology, University of Bergen, Jonas Liesvei 91, 5009 Bergen, Norwayb Norwegian Competence Centre for Sleep Disorders, Haukeland University Hospital, Jonas Liesvei 65, 5009 Bergen, Norway

H I G H L I G H T S

• Social defeat procedures failed to affect sleep or behaviour in rats.• Behaviours during the social confrontation, fighter or submissive, revealed effects.• Fighter rats showed more SWS fragmentation pre and post stress.• Fighter rats showed a longer latency to leave the start box to explore an open field.• Fighter rats failed to show startle response decrement.

⁎ Corresponding author. Tel.: +47 55586002; fax: +47E-mail addresses: anne.kinn@psybp.uib.no (A.M. Kinn

(R. Murison), Jelena.Mrdalj@psybb.uib.no (J. Mrdalj), Ann(A.M. Milde), Finn.Jellestad@psybp.uib.no (F.K. Jellestad),(L.A. Øvernes), Janne.Gronli@psybp.uib.no (J. Grønli).

0031-9384/$ – see front matter © 2014 Elsevier Inc. All rihttp://dx.doi.org/10.1016/j.physbeh.2014.01.010

a b s t r a c t

a r t i c l e i n f o

Article history:Received 22 June 2013Received in revised form 28 October 2013Accepted 14 January 2014Available online 25 January 2014

Keywords:Social defeatSleepStartleCorticosteroneEmergence behaviour

We studied the short- and long-term effects of a double social defeat (SD) on sleep parameters, EEG power,behaviour in the open field emergence test, corticosterone responsiveness, and acoustic startle responses.Pre-stress levels of corticosterone were assessed before all rats were surgically implanted with telemetrictransmitters for sleep recording, and allowed 3 weeks of recovery. Rats in the SD group (n = 10) were ex-posed to 1 hour SD on two consecutive days, while control rats (n = 10) were left undisturbed. Telemetricsleep recordings were performed before SD (day −1), day 1 post SD, and once weekly for 3 weeks thereafter.The open field emergence test was performed on day 9 and weekly for 2 weeks thereafter. Blood samples formeasures of corticosterone responsiveness were drawn after the last emergence test (day 23). Acoustic startleresponses were tested on day 24 post SD. Overall, SD rats as a group were not affected by the social conflict.Effects of SD seemed, however, to vary according to the behaviours that the intruder displayed during the socialconfrontationwith the resident. Compared to those SD rats showingquick submission (SDS, n=5), SD rats fight-ing the resident during one or both SD confrontations before defeat (SDF, n= 5) showedmore fragmented slowwave sleep, both in SWS1and SWS2. They also showed longer latency to leave the start box and spent less time inthe open field arena compared to SDS rats. In the startle test, SDF rats failed to show response decrement at thelowest sound level. Our results indicate that how animals behave during a social confrontation ismore importantthan exposure to the SD procedure itself, and that rapid submission during a social confrontation might bemoreadaptive than fighting back.

© 2014 Elsevier Inc. All rights reserved.

1. Introduction

Psychosocial stress in rats has several behavioural and physiologicaleffects which are often interpreted as parallel to those seen in humans.One commonly used method is the resident–intruder procedure,whereby an intruder rat is exposed to a resident rat in its own home

55589872.Rød), murison@psybp.uib.noe.Milde@psybp.uib.noLeif.Overnes@student.uib.no

ghts reserved.

cage [1–3]. This normally leads to aggressive behaviour by the resident,resulting in the defeat of the intruder (social defeat— SD). The period ofthe psychosocial stress may be prolonged by keeping the intruder ani-mal in the resident's home cage, but protected from injury by wiremesh netting. The effects of psychosocial stress seem to vary accordingto the behaviour which the intruder displays during the confrontation[4–8].

The number and durations of the SD procedures to which theintruder animals are exposed vary. A single or double exposure tosocial stress in rodents induces both short-lasting and long-lastingchanges on physiological, neuroendocrine and behavioural measures[9–12]. Acute effects observed during the social interaction and during

Fig. 1.An overview of the experimental design for the two experimental groups: social de-feat (SD) and control (n = 10 each group). Procedures are identical on all days in bothgroups except on day −1 and day 0 when the stress procedure was conducted. ASR —

acoustic startle response; Cort — blood sampling for corticosterone measure; OF — openfield emergence test; SD — social defeat; Sleep rec. — Sleep recording.

55A.M. Kinn Rød et al. / Physiology & Behavior 127 (2014) 54–63

the hours that follow immediately include increased slowwave activity(SWA) during sleep, increased plasma corticosterone, noradrenalineand adrenaline concentrations, increased core body temperature andincreased heart rate [13–17]. However, all the above mentioned effectsseem to disappear within hours. Other effects may last for days orweeks, for example reduced locomotor activity and increased anxiety-like behaviour in the open field and elevated plus maze tests, reducedsocial interaction, elevated acoustic startle response, reduced sucrosepreference, reduced food intake, suppressed body weight gain, and de-creased circadian variation in core body temperature [3,18–20].

While some immediate effects of SD dissipate over time, othersdevelop. We have previously reported that rats had increased sleepfragmentation and increased amount of slow wave sleep (SWS) 2four days after a double social defeat, but not on the day after the sec-ond defeat [19]. In an earlier experiment, a single SD induced a pro-gressive increase in immobility in repeated sudden silence tests,reaching a maximum after 3 weeks [2]. In general, our knowledgeof the long-term consequences of a stressful social event may be lim-ited because only a few studies employ a sufficient time-span [10].One reason for studying long-term effects lies in the use of SD in ro-dents as an animal model of depression and anxiety [2,21–23], andthe diagnostic criteria and symptomatology of these. Diagnosticcriteria for affective disorders include symptom persistence overtime, such as several weeks [24]. Furthermore, symptoms may taketime to manifest themselves following a precipitating event, e.g. de-layed onset of posttraumatic stress disorder (PTSD).

Disturbed sleep is both a clinical predictor and a symptom of humanaffective disorders such as depression and anxiety disorders (e.g. gener-alized anxiety disorder and posttraumatic stress disorder) [25–29]. Thealterations of sleep parameters include increased latency to sleep, sleepfragmentation (stage shifts and arousals), amount of rapid eye move-ment (REM) sleep, and reduced REM sleep latency, total sleep timeand amount of deep SWS. Changes in sleep electroencephalographic(EEG) activity, EEG power, include reduced power in the delta range(0.2–4 Hz) and increased high frequency power (N20 Hz) [28,30–32].In chronic insomniacs, perceived stress and stress-related avoidance be-haviour have been associatedwith decreased delta power and increasedhigh frequency power (N16 Hz) [33]. Chronic hyperarousal is a state as-sociated with bothmajor depression and anxiety disorders [24], as wellas with functional somatic disorders [34].

Our primary aim in this study was to examine the short-term andlong-term effects of the SD procedure in rats on physiology andbehaviour — sleep parameters, EEG power, startle responses, behav-iour in an open field emergence test as well as the corticosteroneresponse to the emergence test. We expected that socially defeatedrats compared to controls would show high acoustic startle responses,high corticosterone responsiveness, and that in the emergence testthey would show longer latencies to leave the start box and spendless time in the open test arena. Further, that they would show long-term alterations in sleep parameters and EEG power parallel to thoseseen in human affective disorders.

Effects of SD seem to vary according to the behaviour which theintruder displays during the confrontations. Generally, studies haveshown that rats that are passive and show quick submission seemmore affected by the defeat than those that fight back or opposethe resident during the social conflict [4–8]. Rats with this passivecoping strategy during the confrontation display a higher corticoste-rone response to defeat, and a higher level of neuronal activation inthe amygdala and medial prefrontal cortex [4]. Rats that are passivealso show longer-lasting disturbance in diurnal rhythm of heart rate,body temperature and locomotor activity, higher body weight lossand different stress-induced immune changes than those that fightback [5,6].

Our secondary aimwas therefore to investigate whether short-termor long-term effects of the resident–intruder procedure would be relat-ed to the intruder animals' behaviours during the social confrontations.

Here we expected that rats showing rapid submission in the socialdefeat would exhibit the most pronounced alterations in sleep, EEGpower, behaviour and corticosterone response.

All procedures were performed on animals implanted with tele-metric devices to allow long-termmeasurements of EEG and electro-myogram (EMG).

2. Methods

2.1. Ethical evaluation

The experiments described in this article have been approved bythe Norwegian Animal Research Authority and registered by the au-thority. The experiment has thus been conducted in accordance withNorwegian laws and regulations controlling experiments in live ani-mals. Norway has signed and ratified The European Convention forthe Protection of Vertebrate Animals used for Experimental andOther Scientific purposes, of March 18, 1986.

2.2. Design

An overview of the experimental design is shown in Fig. 1.Blood samples for pre-stress levels of corticosterone measures were

taken at least one day before implantation of the telemetric transmitter.After surgery, the animalswere allowed3 weeks of recovery. Rats in theSD group were exposed to social defeat on two consecutive days (−1and 0). Meanwhile, rats in the control group were left undisturbed intheir home cages. Telemetric sleep recordings were performed beforeSD (day −1), day 1 post SD, and once a week for 3 weeks thereafter(on day 7, day 14 and day 21 post SD). The open field emergence testwas performed once a week (on day 9, day 16 and day 23 post SD).Blood samples for measures of corticosterone responsiveness weredrawn 5min after the last emergence test (day 23), and acoustic startleresponses were tested on day 24 post SD.

2.3. Animals and housing

Experiments were performed with 20 outbred male Wistar rats(Taconic, Denmark). On the day after arrival, they were separated,housed individually and allowed 5 days of acclimatisation, before 5days of handling (one minute per day). The cages were individuallyventilated (IVC) polypropylene Euro-standard Type III H cages(425 × 266 × 185 mm — floor area: 800 cm2). Within the cagesthere was an average ambient temperature of 23 °C and an averagerelative humidity of 52%. The rats were exposed to a 12:12 hourlight/dark schedule with lights on at 08:00 h and lights off at20:00 h. A progressive increase in lighting started at 07:00 h andprogressive dimming started at 19:00 h. The rats had free accessto food and water. Bedding (Bee Kay Bedding, Scanbur BK) waschanged once a week. The rats were randomly assigned to SD andcontrol groups (n = 10) after the postoperative period when theyreached their preoperative bodyweight.

56 A.M. Kinn Rød et al. / Physiology & Behavior 127 (2014) 54–63

For the blood sampling, SD procedure and behavioural tests, the ratswere transported one by one in their home cages to and from separate,dedicated rooms.

The resident rats (n = 19, male BDΙΧ, Gades Institute, HaukelandUniversity Hospital, Bergen, Norway) used in the SD procedureswere at least 5 months old and weighed more than 450 g. This strainis bred in nearby animal facilities and the male rats are highly terri-torial and reliably aggressive when faced with an intruder. Theywere pair-housed with ovariectomized females (n = 19, BDΙΧ,Gades Institute, Haukeland University Hospital, Bergen, Norway)for at least two weeks before the SD experiment in order to stimulateinstinctive territorial behaviour. During this time, the females werebrought into oestrous by subcutaneous injections of oestradiol ben-zoate (200 μg/rat in 1ml oil) every fourth day followed by progester-one (0.5 mg/rat in 1 ml oil) 42 h after the oestradiol injection. Theresidents and females were housed in IVC polypropylene Euro-standard Type IV S cages (480 × 375 × 210 mm — floor area:1400 cm2) in a room at another facility. Apart from this, residentshad the same housing conditions as the experimental rats. Theywere habituated to being moved between the housing quarters andthe dedicated room prior to the SD procedures. The bedding was nor-mally changed once a week, but was not changed 2 days prior to thesocial conflict in order to preserve the residents' scent in the bedding.

The residents had been trained to fight for their territory duringat least 5 training sessions with young males (Wistar, Taconic,Denmark). On the last training day, the four most aggressive resi-dents, those that were able to quickly defeat the intruders withoutinjuring them, were selected for use in the present experiment.

2.4. Social defeat procedures

Social defeat procedures were performed during the first four hoursof the animals' active (dark) phase. The residents were transported tothe test room, which was illuminated with a red light, and were givena one hour rest period before the SD procedure. Their female cage-mates were placed in a different cage during this timewindow and dur-ing the social conflict. To have a clear view of the resident–intruder con-flict, the top of the residents' cage was removed and a floorless emptycage was placed upside down as a top.

Social defeat procedures were adapted from Meerlo and collabora-tors [3]. The SD intruder rat (previously implantedwith a telemetric de-vice, see below) was placed in the territory (cage) of the resident rat,andwas attacked and defeated as indicated by fleeing, freezing and sub-missive behaviour. When clear submission behaviour was observed(the intruder lyingmotionless on its back) or the intruder was attackedfor the 5th time, the defeated rat was placed in a small wire-mesh cage(23 × 11.5 × 11.5 cm, mesh width 1.5 × 1.5 cm) that was put back intothe cage of the resident. The SD intruder rat had olfactory, visual andauditory contact with the resident but was protected from repeatedattacks and potential injury. SD rats were exposed to the resident for atotal of 1 h. The phase when the SD intruder was directly exposed tothe resident was filmed for offline behavioural analyses.

Immediately after the defeat session, SD intruder rats were returnedto their home cages, and their own room. SD ratswere exposed to SD ontwo consecutive days and never to the same resident on both days. Thecontrol rats were left undisturbed.

Behaviour of the intruder during the period of direct exposure to theresident was scored manually with the aid of a specialized computersoftware package (Observer XT, Noldus Information Technology,Wageningen, The Netherlands). The following measures were scoredin the SD interaction, based on behavioural parameters from Koolhaaset al. 2013 [1]: Number of received attacks; submissive posture; resi-dent in supine posture; initiated attack, and duration (secs) of: receivedlateral threat; flight; freeze; general activity (including received ano-genital sniffing, social explore, non-social explore, move away and rear-ing); hold resident down; move towards resident; upright posture;

total duration of direct exposure (time from placing the intruder andthe resident together, to time to separation with wire-mesh cage).

2.5. Blood sampling

Blood samples for corticosterone measures were taken between09:00 and 12:00 h.

For blood sampling, the animals were placed in a sealed anaestheticchamber (23 × 12 × 11.5 cm), and anaesthesia was induced withIsofluran (Isoba Vet. Schering-Plough, Denmark). After clear musclerelaxation, they were placed on a table, where one hind limb wasshaved and smeared with Vaseline. The saphenous vein was punc-tured and 40–400 μl blood was collected in BD Microtainer tubes(Medinor, Norway). The sequence of moving the rat from the homecage to complete blood collection took less than 3.5 min. The bloodsamples were left for ½ to 1 h in room temperature for coagulationbefore they were centrifuged at 3500 rpm for 10 min, and serumwas separated and then frozen at −20o C until analysis.

The corticosterone analysis was performed by the means of a RatCorticosterone Enzyme Immunoassay-kit (DSL-10-81100, MedProbe,Norway) with the aid of a plate reader Wallac 1420 Multilabel counter(PerkinElmer, Norway). Serum samples were analysed in duplicateand measures were averaged afterwards. The intra-assay coefficient ofvariation was 12.2% for the low control and 2.4% for the high control.The sensitivity of the assay was 1.6 ng/ml.

2.6. Implantation of telemetric devices

Telemetry transmitters (4ET, Physiotel®, Data Sciences Interna-tional, St. Paul, USA) with biopotential leads for EEG and EMG wereimplanted subcutaneously (s.c.) in SD and control rats for sleep re-cording. For 3 preoperative days, the rats were given antibiotic,5 ml Bactrim per 250 ml drinking water (trimethoprim 8 mg/ml,sulfamethoxazol 40 mg/ml, Roche). The animals, weighing 300 g,were anesthetized by s.c. injection in the neck with a mixture of0.22 ml of Hypnorm (fentanyl citrate 0.315 mg/ml, fluanizone10 mg/ml, Janssen) and 0.22 ml Midazolam B. Braun (midazolam5 mg/ml, Braun), diluted with distilled water. Incisions were madein the dorsomedial lumbar region for the telemetric device and onthe skull and neck for the biopotential leads for EEG and EMG, re-spectively. The EEG deviations used were bilateral fronto-frontal(FF) and fronto-parietal (FP) placed epidurally. The frontal leadswere placed 2 mm anterior to bregma and 2 mm lateral to the midline,and the parietal lead was placed 2 mm anterior to lambda and 2 mmlateral to the midline. All EEG leads were secured to the skull withdental acrylic (GC RELINE, America INC.). The leads for EMG recordingwere attached to the neckmuscle. The incision in the dorsomedial lum-bar region was closed with wound clips. The skin on the head wasclosed with interrupted mattress sutures.

After surgery, the animals received one 0.15 ml analgesic s.c dose ofTemgesic (burprenorphinum 0.3 mg/ml, Reckitt & Colman), followedby 0.1 ml dose twice a day for three days. Antibiotics (Bactrim, Roche)were further given in the drinking water during 3–7 days after surgery.The animals were allowed 3 weeks of recovery, to regain their preoper-ative weight before starting the experiment.

2.7. Telemetric recording and scoring

The telemetric device was turned on by sweeping a magnet alongthe site of the implanted battery. Telemetric recordings took place inthe colony room and animals were in their home cages. At 08:00 hthe cages were gently shaken to wake the rats for the start of record-ing (to calculate latency to sleep onset, SWS2 and REM sleep). Theanimals were then left undisturbed during the 8 h sleep recording.The wireless recording device was calibrated to record signals inthe range −1.25 to +1.25 mV. Telemetry signals were collected

57A.M. Kinn Rød et al. / Physiology & Behavior 127 (2014) 54–63

continuously at the sampling rate of 250 Hz through a receiver (typeRPC-2, Data Sciences International) placed directly beneath the homecage and connected to a data exchange matrix where signals wereconverted and transferred to the acquisition software Dataquest ART(version 4.1, Data Sciences International, St. Paul, USA). Analogueband-pass filters of 1 Hz (high pass filter) and 100 Hz (low pass filter)were used. No additional filtering was used. Both control and SD ani-mals were recorded on the same days.

For the investigation of EEG power spectrum distributions, FastFourier Transform (FFT) analyses were computed offline on the un-filtered FF EEG deviations using Neuroscore software (Version2.0.1, Data Sciences International, St. Paul, USA). Analyses were con-ducted with 10 second epochs and Hamming window overlap of 75%to smooth the transition between every sample. The EEG signalswere visually inspected and all epochs containingmovement or elec-trical artefacts were excluded. EEG power data were calculated fortotal power (0.5–60.0 Hz), nonspecific to sleep or wakefulness. Forthe specific sleep stages, the EEG frequency bands characteristic foreach stage were considered. Wakefulness: beta (19.5–34.5 Hz) andgamma (lower range 35.0–60.0 Hz); REM sleep: theta (5.5–9.5 Hz);SWS (SWS1 + SWS2): delta (0.5–4.5 Hz). Mean values were usedfor statistical analysis.

Sleep was scored using Neuroscore software. For visual display, theFF EEG was high-pass filtered at 3 Hz and low-pass filtered at 35 Hz.The FP EEG was filtered with 0.5 Hz and 35 Hz respectively. The EMGsignal was high-pass filtered at 5 Hz. All the signals were filtered at 50Hz to eliminate power line artefacts.

An automatic scoring algorithm was used on the filtered signalsfor 10 s epochs. For each animal analysis thresholds were adjusted(delta-ratio, theta-ratio, EMG-threshold, activity-threshold) for theautomatic scoring to fit with the manual criteria of Ursin and Larsen[35], and Neckelmann and Ursin [36]. The algorithm did not include athreshold for muscle atonia in REM sleep, leading to incorrect auto-matic scoring of wakefulness and REM sleep. The automatic scoringwas manually re-scored regarding REM sleep and wakefulness. Thethreshold for the delta-ratio was normally satisfactory, thus SWS1and SWS2 from the automatic scoring were only re-scored on a fewoccasions. For manual re-scoring, wakefulness, transition sleep,REM sleep, SWS1 and SWS2 were defined according to Ursin andLarsen [35], and Neckelmann and Ursin [36]. In short, wakefulnesswas scored when there was high-frequency low voltage activity inEEG channels and high to moderate activity in EMG; SWS1 (compa-rable to light sleep in humans) was scored when there was spindleactivity (11–16 Hz) and b50% of delta activity in the EEG channels,and EMG reduced compared to wakefulness; SWS2 (comparable todeep SWS in humans) was scored when there was spindle activityand N50% of delta activity in the EEG channels, and EMG activityequal or lower compared to SWS1; REM sleep was scored whenthere was predominantly EEG theta activity and EMG activity was re-duced to its lowest or abolished (muscle atonia); Transition sleep is astage between SWS and REM sleep characterized by high amplitudespindles, theta activity with higher amplitude than in REM sleep, andlow EMG activity. Transition sleep was manually re-scored as REMsleep when there was muscle atonia and the period was followedby REM sleep. Otherwise, transition sleep was re-scored as SWS1.

An inter-rater reliability between manual and semi-automaticscoring was evaluated in 6 random recordings of 8 h durations (atotal of 2880 epochs, 10 s each), giving a satisfactory mean Kappaof 0.73 ± 0.03. The percentage agreement was 92.7% for wakeful-ness, 96.5% for REM sleep, 85.5% for SWS1 and 92.2% for SWS2,respectively.

The following dependent sleep variables were computed: total sleeptime, duration of wakefulness and sleep stages, sleep fragmentation instages (expressed by number of episodes in wakefulness and sleepstages) and total sleep fragmentation. Sleep onset latencywasmanuallyscored from recording start to stable sleep onset. Stable sleep onset was

defined as the first continuous sleep period (SWS1, SWS2 or REM sleep)lastingmore than 5minwithout being interrupted by awakefulness ep-isode longer than 30 s [37]. SWS2 latency and REM sleep latency weremanually scored from stable sleep onset to the occurrence of the rele-vant sleep stage persisting for at least two epochs.

2.8. Open field emergence test

A separate room with dimmed light was used for administering theemergence test.

The emergence test apparatus consisted of an openfield arenawith ablack square base (100 × 100 cm) with black walls (40 cm high), and ablack start box (17 × 14 × 14 cm, opening: 10 × 10 cm) located in thecentre of the arena. The rats were placed individually in the start box,left undisturbed and activity was recorded for 15 min with a digitalcamera. After completion, the number of faecal droppings was counted,and the rat returned to its home cage for 5 min prior to blood sampling.All the surfaces were thoroughly cleaned with a 5% ethanol solutionprior to testing and between each test.

The position of the rat in the arena was scored manually with theaid of a specialized computer software package (Observer XT, NoldusInformation Technology, Wageningen, The Netherlands). The fol-lowing measures were scored in the emergence test over trials: latencyto exit the start box (secs); percentage time spent in the arena; numberof faecal droppings (in the start box and the area summed).

2.9. Acoustic startle response test

The startle apparatus and startle procedure is described in detailelsewhere [38]. In short, the startle apparatus (SR-LAB Startle ResponseSystem, San Diego Instruments) consisted of a transparent tube placedon a pressure-sensitive plate that registered the animal's gross bodymovements. The cylinder was placed in a sound-attenuated chamberwith normal lighting and a speaker mounted in the ceiling. Acousticstimulus delivery was controlled using the SR-LAB software, which dig-itized, rectified, and recorded the readings.

Four rats at a timewere tested simultaneously in separate chambers.They were left undisturbed for a 5 min habituation period with a back-ground noise level of 67 dB. During the subsequent 10min the ratswereexposed to a series of 30 acoustic stimuli, 10 at each of 95, 105 and115 dB. Each stimulus was presented for 50 ms. Stimuli were presentedpseudo-randomlywith an inter-stimuli interval between 5 and 39 s. Foreach stimulus, maximum response amplitude (Vmax) was recorded.The tubes were thoroughly cleaned prior to and between each testwith a 5% ethanol solution.

2.10. Data analysis

Results are presented as mean ± S.E.M. Statistical analysis wasperformed using Statistica 8.0 (StatSoft, Inc.) with a significancelevel of p b 0.05. Significant effects in the ANOVAs were furtheranalysed with Fisher LSD post-hoc tests.

On the basis of the behaviour in the social defeat confrontation SDrats were split into two subgroups. SD rats that showed no resistanceduring the two SD confrontations were assigned to the SD submis-sive (SDS) subgroup. SD rats that fought back during one or both ofthe SD confrontations, and kept the resident down in a supine pos-ture, were assigned to the SD fighter (SDF) subgroup. Behaviours ofthese two subgroups in the SD confrontation were compared usingone-tailed t-tests. For comparisons of behaviours of the SDS andSDF rats in the SD confrontation, data were adjusted for total durationof direct exposure (frequency — number per 100 seconds direct expo-sure, and percentage duration).

Relationships between behavioural measures during the SD sessionand selected dependent variables (sleep, emergence test, startle) wereassessed using Pearson's r (all n = 10).

Table 1Behaviours in the social defeat confrontation.

Variable SDF SDS t(8)

Total duration of directexposure (s)

373.00 ± 98.11 133.40 ± 32.03 −2.32⁎

General activity (duration %) 50.99 ± 8.67 40.97 ± 13.24 −0.63

Defensive behavioursReceived attack (frequency) 1.57 ± 0.16 5.11 ± 1.03 3.40⁎⁎

Submissive posture (frequency) 0.77 ± 0.24 2.37 ± 0.51 2.85⁎

Flight (duration %) 1.88 ± 1.21 10.21±2.90 2.65⁎

Freeze (duration %) 9.27 ± 2.80 18.97 ± 3.66 2.11⁎

Received lateral threat (duration %) 26.60 ± 6.01 27.83 ± 9.57 0.11

Offensive behavioursResident in supine posture(frequency)

1.21 ± 0.24 0.00 ± 0.00 −4.98⁎⁎⁎

Initiated attack (frequency) 0.19 ± 0.10 0.09 ± 0.09 −0.74Hold resident down (duration %) 2.44 ± 0.80 0.00 ± 0.00 −3.07⁎⁎

Move towards resident(duration %)

1.15 ± 0.70 0.00 ± 0.00 −1.63

Upright posture (duration %) 7.54 ± 3.17 2.29 ± 0.77 −1.61

Summaryof differences between social defeatfighter (SDF) social defeat submissive (SDS)rats' defensive and offensive behaviours in the SD confrontation (group mean ± SEM).Duration % represents total time scored for the behaviour adjusted for total duration of di-rect exposure (s). Frequencies of behaviours represent total number of observations of thebehaviours adjusted for total duration of direct exposure (number per 100 s).⁎p b .05. ⁎⁎p b .01. ⁎⁎⁎p b .001. One-tailed t-tests.

58 A.M. Kinn Rød et al. / Physiology & Behavior 127 (2014) 54–63

For all parameters described below, differences between control andSD, and between control, SDS and SDF, and between SDS and SDF wereanalysed.

Sleep recordings for all the rats were carried out before SD (day−1for baseline), and post SD on day 1, day 7, day 14 and day 21. For the day7 recording, a large part of the data were lost so that the remaining datafor this day were not analysed.

Sleep parameters at baseline were analysed by one-way ANOVAor by repeated measures ANOVA (group × stage). To assess changesacross days (−1, 1, 14 and 21), repeated measures ANOVA wereused (group × day or group × day × sleep stage).

EEGpower bands characteristic forwakefulness, SWSand REM sleepas well as total EEG power, were analysed in separate analyses by re-peated measures ANOVA with ‘group’ or ‘subgroup’ as independentfactor and ‘day’ as repeated measure. For wakefulness (characterisedwith several frequency bands), ‘band’was added as a repeatedmeasure.

The emergence test parameters were analysed between groupsby the Mann–Whitney U Test (SD vs. control and SDS vs. SDF) or bythe Kruskal–Wallis test (SDS vs. SDF vs. control). Changes across dayswithin groups and subgroups were analysed by Friedman ANOVA.

Pre-stress levels and response levels of corticosteronewere analysedseparately by one-way ANOVAs.

For the acoustic startle response test, mean values over all 10 tri-als at each dB level were analysed for maximum response amplitude(Vmax). Decrement of Vmax was defined as the percentage changein Vmax at each dB level from the first to the last (10th) trial. Prelim-inary Levene's test revealed significant heterogeneity of variance inVmax scores. Raw scores were therefore square root transformedand all analyses were performed on these transformed data. Differ-ences between groups and subgroups in Vmax and decrement ofVmax were analysed by one way ANOVA for each parameter andeach dB level in separate analyses.

3. Results

3.1. Behaviour in the social defeat confrontations

All SD rats were eventually defeated as indicated by fleeing, freezingand a submissive supine posture. During the SD confrontations, SDrats showed diverse behaviours and postures (see [1,39]). Some SDrats showed flight and rapidly assumed a submissive supine posture.Others tried to fight back, characterized by a defensive upright andmutual upright posture (boxing/guarding) before eventually show-ing a submissive supine posture. The resident rat in the confronta-tion was always the most aggressive and dominant, characterizedby arched back, piloerection, repeated attacks, lateral threat and anoffensive upright posture. However, some of them showed occasion-al supine posture before they again showed aggressive behaviours.On the basis of the behaviour displayed during the confrontation,SD rats were divided into two subgroups: the SD submissive (SDS)and the SD fighter (SDF) subgroups as already described in the dataanalysis section. Differences in behaviours between the two sub-groups are shown in Table 1. In summary, SDS animals showed a highernumber of submissions than SDF animals, were recipients of more at-tacks, showed more flight behaviour, showed more freezing, held theresident down less and had a shorter duration of direct exposure.

3.2. Effect of social defeat on wakefulness and sleep

One rat from the control group was excluded from the EEG analysesdue to loose EEG leads.

3.2.1. Latency to sleep onset, SWS2 and REM sleepThere were no differences between groups or subgroups at baseline

or effects across days after social defeat on latency to sleep onset, toSWS1, to SWS2 or to REM sleep.

3.2.2. Total sleep time (TST)At baseline, controls and SD rats did not differ. Across days there

was a group × day interaction (F(3,51) = 3.46, p = 0.023), but nomain effect of group. There were no significant differences in TST be-tween the groups on any of the post SD days. Within the SD groupthere was a significant increase in TST from baseline to day 14 (p =0.048). Within the control group TST increased from baseline today 21 (p = 0.006).

Subgroups of SD did not differ from each other or from controls inTST either at baseline or across days.

3.2.3. Duration of wakefulness and sleep stagesAt baseline there were no differences between control and SD rats,

and there were no short-term or long-term effects of social defeat onwakefulness and sleep stages. Descriptively, the subgroups of SD ratsshowed different sleep quality prior to the social stress, although thedifferences were not statistically different (see Fig. 2). SDF rats showedmore SWS2 and less SWS1 compared to SDS, which also persisted afterthe SD confrontations. At baseline only, the SDF rats showed morewakefulness. The SDS rats showed less REM sleep at baseline, but in-creased their time in REM sleep on day 1 and 14. The change acrossdays was not seen in SDF rats.

3.2.4. Sleep fragmentationThere were no differences between groups or subgroups at baseline,

and there were no short-term or long-term effect of social defeat ontotal sleep fragmentation.

SD rats as a group, or separated in SDS and SDF, were not different tocontrols at baseline or across days on fragmentation in wakefulness andsleep stages.

Sleep fragmentation in stages did not differ between the SDS andSDF rats at baseline. However, across days there was an interactionbetween stage and group (F(3,24) = 3.98, p = 0.020), but no maineffect of group or group × day × stage interaction. Over all days,SWS was more fragmented in the SDF rats as they showed morefragmentation in SWS1 (p = 0.012) and SWS2 (p = 0.009) com-pared to SDS (see Fig. 3). Post hoc tests show significantly morefragmentation of SWS1 in SDF than in SDS rats on days 1, 14 and21(p = 0.033, 0.012 and 0.001, respectively). Similarly, post hoc

Fig. 2.Duration of wakefulness and sleep stages expressed as minutes (groupmean± SEM) in control (C, n= 9), social defeat (SD, n= 10) group, and the SD group divided in SD fighter(SDF, n = 5) and SD submissive (SDS, n = 5) subgroups.

59A.M. Kinn Rød et al. / Physiology & Behavior 127 (2014) 54–63

tests show significantly more fragmentation of SWS2 in SDF rats ondays 1, 14 and 21 (p =0.033, 0.006 and 0.001, respectively) (seeFig. 3). SWS1 fragmentation on Day 1 was negatively correlatedwith number of received attacks, r(10) = −0.647, p = 0.043, as

Fig. 3. Sleep fragmentation in stages expressed as number of episodes inwakefulness and sleepSD group divided in SD fighter (SDF, n = 5) and SD submissive (SDS, n = 5) subgroups. *p b 0

was total sleep fragmentation on Day 1, r(10) = −0.648, p =0.043. SWS1 fragmentation on Day 1 was negatively correlated toflight behaviour, r(10) =−0.672, p = 0.033, as was SWS2 fragmen-tation on Day 14, r(10) = −0.633, p = 0.049.

stages (groupmean± SEM) in control (C, n= 9), social defeat (SD, n=10) group, and the.05 SDS vs. SDF, **p b 0.01 SDS vs. SDF.

60 A.M. Kinn Rød et al. / Physiology & Behavior 127 (2014) 54–63

3.3. Effect of social defeat on EEG power in wakefulness and sleep stages

There were no differences at baseline or effects across days of socialdefeat in groups or subgroups on EEG power in the different frequencybands in wakefulness, SWS and REM sleep.

3.4. Effect of social defeat on behaviour in the emergence test

3.4.1. Latency to exit the start boxThere were no differences between the SD and controls. Across

days, both the controls (χ2(2) = 7.94, p = 0.020) and SD rats

(χ2(2) = 7.54, p = 0.023) showed shorter latency to exit the start

box.A Kruskal–Wallis test, which included all three groups, shows

that SDF rats had a longer latency to exit the start box compared toSDS (p = 0.031) on day 23, which was also supported by a Mann–Whitney U Test comparing only SDF and SDS (p = 0.022). Controlsdid not differ from SDF or SDS. Within the subgroups, there was nochange across days (see Fig. 4A).

Emergence latency on the first emergence test was positively corre-latedwith keeping the resident down, r(10)= 0.658, p= 0.039. Emer-gence latency on the third testwas positively correlatedwith resident insupine posture, r(10) = 0.692, p = 0.027 and negatively correlatedwith flight behaviour, r(10) = −0.64, p = 0.046.

3.4.2. Percent time spent in the arenaThere were no significant differences between the SD rats and con-

trol rats on any of the days. The control rats showed increased timespent in the arena across days (χ2

(2) = 12.4 p = 0.002), an effect thatwas not seen in the SD rats.

When SD rats were separated into SDS and SDF subgroups andcompared to controls there were no difference on any of the days.Analysis of only SDS and SDF showed that the SDF rats spent lesstime in the arena compared to the SDS rats on day 16 (p = 0.036)

Fig. 4. Behaviour in the open field emergence test (groupmean± SEM). A) Latency to exitthe start box expressed as seconds (secs) B) Time spent in the arena expressed as percentof total recording time, control (C, n = 10), social defeat (SD, n = 10) group, and theSD group divided in SD fighter (SDF, n = 5) and SD submissive (SDS, n = 5) subgroups.*p b 0.05 SDF vs. SDS (Mann–Whitney U Test or Kruskal–Wallis test).

and day 23 (p = 0.022) (see Fig. 4B). Within the SDF and the SDSrats there were no significant increases in time spent in the arenaacross days.

3.4.3. Number of faecal droppingThere were no differences between groups or subgroups and no sig-

nificant changes across days within groups or subgroups on defecation.

3.5. Effect of social defeat on corticosterone response to open fieldemergence test

One SDS animal was excluded from the analyses of corticosteronedata for the pre-stress sample due to an extreme value. There were nodifferences between groups or subgroups in pre-stress levels of cortico-sterone or in response to the last emergence test (see Table 2 for de-scriptive statistics), although there was a trend for SDS animals tohave higher corticosterone levels compared to controls on the last testday (p = 0.054). All the groups showed an increase from pre-stresssamples to those obtained after the emergence test, F(2, 16) = 40.92,p b 0.0001. The interaction group × sample was not significant.

Corticosterone levels after the third emergence test were negativelycorrelated with movement towards the resident, r(10) =−0.874, p =0.001, and upright posture, r(10) = −0.835, p = 0.003.

3.6. Effect of social defeat on startle response

3.6.1. Maximum response amplitude (Vmax)Comparison between groups and subgroups showednodifference in

Vmax to 95, 105 or 115 dB startle stimuli. See Table 3 for descriptivestatistics.

3.6.2. Startle response decrement over trialsOverall comparison between controls and SD rats showed no dif-

ferences in response decrement to the 95 dB startle stimuli. Whencomparing the control to the SD subgroups, however, the group ef-fect becomes significant (F(2,17) = 6.23, p = 0.009). Post hoc testsshow that decrement for the SDF animals was significantly lessthan for the SDS group (p = 0.005) and for the control group (p =0.008). In fact, the SDF animals failed to show any response decrementto the 95 dB stimuli.

Startle response decrement at 95 dB was positively correlatedwith number of attacks received, r(10) = 0.658, p = 0.039, numberof submissions, r(10) = 0.768, p = 0.01, and negatively correlatedwith keeping the resident down, r(10) = −0.890, p = 0.001, mov-ing towards the resident, r(10) = −0.740, p = 0.014, upright posturer(10) = −0.783, p = 0.007 and total duration of direct exposure,r(10) = −0.886, p = 0.001.

Therewere no differences between groups or subgroups in responsedecrement to the 105 dB or to the 115 dB startle stimuli. See Table 3 fordescriptive statistics.

Table 2Serum corticosterone levels pre-stress and response to emergence test expressed asng/ml.

Valid N Pre-stress cort level Valid N Cort response

C 10 125.30 ± 31.11 10 367.89 ± 53.36SD 9 84.76 ± 23.19 10 488.80 ± 52.26SDF 5 116.29 ± 36.70 5 423.30 ± 95.69SDS 4 45.35 ± 7.08 5 554.30 ± 31.43

Descriptive statistics of serum corticosterone (cort) levels pre-stress and response to openfield emergence test in control (C), social defeat (SD) group, and the SD group divided inSD fighter (SDF) and SD submissive (SDS) subgroups (group mean ± SEM). ANOVAshows no significant differences between groups, an increase from pre-stress levels topost emergence test levels, F(1, 16) = 40.92. p b 0.0001, and no group × sampleinteraction.

Table 3Startle response (Vmax) and response decrement.

Valid N Vmax 95 dB Vmax 105 dB Vmax 115 dB

C 10 11.29 ± 1.34 30.32 ± 3.08 51.83 ± 4.15SD 10 10.89 ± 1.73 31.57 ± 4.35 50.49 ± 5.31SDF 5 12.12 ± 3.48 36.39 ± 7.33 53.27 ± 7.36SDS 5 9.65 ± 0.78 26.74 ± 4.47 47.71 ± 8.29

Valid N Decrement 95 dB Decrement 105dB Decrement 115dB

C 10 27.19 ± 9.12 22.43 ± 27.51 27.31 ± 11.06SD 10 0.92 ± 20.91 −37.21 ± 32.03 25.41 ± 8.88SDF 5 −41.29 ± 30.16 ⁎⁎‡ −46.84 ± 54.76 14.14 ± 4.19SDS 5 43.12 ± 12.95 −27.58 ± 39.63 36.67 ± 16.54

Descriptive statistics of maximum response amplitude (Vmax) and response decrementover trials of Vmax at each dB level in the startle response test in control (C), socialdefeat (SD) group, and the SD group divided in SD fighter (SDF) and SD submissive(SDS) subgroups (group mean ± SEM). ⁎⁎p b 0.01 SDF compared to SDS and ‡p b 0.01SDF compared to C.

61A.M. Kinn Rød et al. / Physiology & Behavior 127 (2014) 54–63

4. Discussion

The present study was primarily aimed at the long-term conse-quences of social defeat stress. We performed the first sleep record-ings 8–12 h after defeat and on a weekly basis thereafter. Most otherbehavioural tests and measures were also collected 1 up to 4 weeksafter defeat. It is therefore not excluded that social defeat had acuteeffects that were not picked up in our measurements. However, thepresent study showed that, overall, defeat caused no major long-term changes in sleep, anxiety-like behaviour and corticosteronelevels.

The present results show that the sleep of SD rats as a group wereunaffected with regard to sleep onset, total sleep time, amount ofdeep SWS and changes in REM sleep, as well as EEG power of the dis-tinct sleep stages. Also contrary to what we expected, neither an in-creased sleep fragmentation nor changes in the deep SWS wereobserved after SD. We have previously shown that defeated rats in-creased sleep fragmentation and amount of SWS2 day 4 after SD[19]. We expected this effect of SD to increase across days in thepresent study as well. However, there was no apparent differenceon day 1, 14 or 21 after the social conflict.

We also expected behavioural changes after SD, specifically elevatedacoustic startle responses and anxiety-like behaviours in the emergencetest. SD did not elevate startle responses to any of the acoustic stimuligiven, nor was there an effect on startle response decrement over trials.SD rats as a group did not differ from the controls in either latency toleave the start box, the time spent in the open arena, the defecationrate or the corticosterone response. One behavioural difference washowever apparent in our study. The control rats showed the expectedincrease in time spent in the arena over repeated sessions as shownby others [40–43], while the SD group did not. This lack of habituationto the test by the SD animals may be an indicator of heightened anxiety.One limitation of the study was the use of undisturbed home cage con-trol animals. However, the absence of overall effects would suggest thatusing handled or yoked animalswould not have significantly altered theoutcome. Rather the opposite, since the two groups would have beenmore alike with regard to treatment.

The absence of effects when all SD animals were included as onegroup may reflect differences in experimental procedures comparedto our previous studies [19,20]. The animals in the present studywere housed in individually ventilated cages (IVCs) (to complywith newer regulations), which involves a greater degree of isolationthan does conventional housing. Little is known about the impact ofIVC housing on behaviour, but rodents housed isolated or in IVCs haveshown anxiety-like behaviour in the elevated plus maze [44,45]. Also,all our animals were implanted with telemetric devices, which involvesa major surgical procedure and necessitates post-operative care at an

individual level. Differential amounts of handling may lead to a largervariation in the startle response and other behaviours, thus weakeningany effects of SD.

Another explanation is that the SD animals here represent a bivari-ate population — fighters and submissives. Other studies have shownthat effects of a social conflict vary according to the behaviour that theintruder displays during the social confrontation [4–8]. In the presentstudy, we observed that the intruders clearly showed variation in be-haviours when confronted by the resident. On the basis of this behav-iour, the SD rats were split into two subgroups; intruders that showedno resistance (SD submissive, SDS) and intruders that fought back dur-ing one or both of the SD confrontations (SD fighters, SDF). It should beemphasised that all SD animals were eventually defeated.

Effects of social defeat varied according to the behaviours that theintruder displayed during the social confrontation with the resident.In the emergence test, rats with an active coping strategy (fightingback during the social confrontation) showed longer latency to leavethe start box and spent less time in the arena compared to those show-ing quick submission. This effect was apparent 16 days after the socialconflict, and was maintained throughout the experiment. Also, theSDF rats showed more fragmented slow wave sleep, both in SWS1and SWS2, an effect which was also more robust after 14 and 21 days.Twenty four days after the social conflict, the SDF rats failed to show re-sponse decrement in the startle test at the lowest sound level.

In the two subgroups of SD animals, we found an interesting patternin the various sleep stages, even though the differences were not sig-nificant. The SDS and SDF rats showed different sleep quality prior toSD. SDF rats were less awake, had more SWS2, less SWS1, and moreREM sleep. The pattern of differences in SWS1 and SWS2 was main-tained throughout the experiment, also at baseline, suggesting thatthis might reflect a trait rather than an effect of social conflict. Withrespect to amount of REM sleep, the SDS group showed an increasecompared to their own baseline prior SD on day 1 and day 14. This re-sult of the SDS group is interesting because a disinhibition of REMsleep is a predictive marker of depression [25]. The SWS continuitywas significantly poorer (greater fragmentation) at the end of theexperiment in the SDFs compared to the SDS rats. In both SWS1and SWS2, the SDF rats showed a more consistent pattern through-out the experiment with a significantly higher fragmentation thanthe SDS. The SDS rats seemed to decrease their fragmentation acrossdays. Although not significant, we believe that these findings are worthpursuing. It should be noted that our sleep recordings were limited to8 h, and to the inactive (light) phase of the dark/light cycle.

In the emergence test, SDF rats exhibited a significantly longer la-tency to leave the start box and they did not increase time spent inthe arena across days. On days 16 and 21 after the social conflictthey spent significantly less time in the arena. Placing a rat in anovel environment evokes competing behavioural tendencies, i.e.anxiety and fear (avoiding the novel open field) vs. exploration andcuriosity [43]. Our result shows that SDF rats avoid the open arena,which may indicate higher anxiety. Our observations also fit inwith the ideas of Koolhaas and colleagues. Over a series of studiesin a number of species, they have observed that aggression is associatedwith some rigidity of routines (intrinsically driven behaviours, proac-tive coping), while non-aggression is more associated with flexibilityof behaviours (extrinsically driven, reactive coping) [46–48]. Our SDFanimals' unwillingness to leave the start box and to explore the openarena might reflect greater rigidity of routines and/or anxiety.

The corticosterone response to the emergence test also appearedto differ somewhat between the subgroups of the SD rats. Comparedto the control rats, the SDS rats had close to a significant (p = 0.054)higher corticosterone response to the third emergence test, and thecorticosterone level was negatively correlated with offensive behav-iours during SD. A similar difference in the corticosterone responseto the SD situation has been reported earlier, with passive oppositionto the resident being associated with a high corticosterone response

62 A.M. Kinn Rød et al. / Physiology & Behavior 127 (2014) 54–63

[4]. Furthermore, high resting levels of corticosterone are seen in ratswith short submission latency before the fourth confrontation in aseries of 7 SD confrontations on consecutive days [7]. In our study,the corticosterone response was not due to the novelty of the emer-gence test, but the SDF appeared to respond with a lower corticoste-rone when habituated to the test.

While the corticosterone data are in line with earlier reports [4,7],our behavioural data are not. Generally, rapid submission in SD hasbeen associated with depressive-like behaviour and endocrine status[7,8]. Our data suggest an increase in anxiety behaviours in the fight-er animals (SDF). A plausible explanation for this discrepancy is thatthe SDF animals spent longer times directly exposed to the residents(see Table 1), invoking therefore being subjected to a greater totalload.

We were surprised that SD rats, either as a group or as subgroups,did not show any differences from controls on the startle responseamplitude to acoustic stimuli at any of the three dB levels. Increasedstartle responses have been seen after inescapable foot shock [38,49]and exposure to predator odour [50]. Also, in our earlier study, a sin-gle social defeat induced a higher startle response to 105 dB and115 dB stimuli when compared to controls and to rats exposed to in-escapable foot shock [20].

In contrast to startle response amplitude, we did observe differencesin decrement of the startle response, but only at 95 dB stimulus in-tensity. SDF rats showed lower decrement than the SDS and controlanimals, which did not differ. Changes in startle amplitude may bewithin a session or between sessions, and thesemay represent differentneurobiological processes [51]. In our case, we only have data fromwithin one session. We hesitate to use the term habituation for this re-sponse decrement since the stimuli were presented at three intensitiesin a pseudo-randomorder. However, it is worth noting that a lack of ha-bituation within startle testing sessions has been reported in PTSD pa-tients [52]. The absence of response decrement in our SDF animalsmay therefore reflect an increased state of anxiety.

Paradoxically, our results to some extent suggest that rapid submis-sion during a social confrontation might bemore adaptive than fightingback. SDS animals had less sleep fragmentation, greater flexibility ofbehaviour in the emergence test, and a decrement in startle responseover repeated trials, as well as a tendency to higher corticosteroneresponsiveness. Such a response pattern is similar to that describedas “reactive” by Koolhaas and his group [46–48]. However, it mustbe recalled that all animals eventually lost their fight. Maybe surren-der after a short fight is more adaptive than surrender after a longerfight. SDS animals were removed from direct exposure to the resi-dents after a shorter time than was the case for the SDF animals, sothat the SDF animals could have been exposed to more total attacksby the resident. However this appears not to be the case. The data actu-ally show that, when not adjusted for total time, the number of attacksreceived by the two subgroups is essentially the same (SDS: 5.6 vs. SDF:5.4), as is also the number of submissions (SDS: 2.6 vs. SDF: 2.0). None-theless, SDF animals had longer direct exposure timewith the residentsthan did the SDS animals, and one might reasonably argue that theywere therefore subjected to a greater stressor or amount of direct threatthan the SDS animals. Against this, the same argument could be appliedto the studies of Wood [7,8], whose results using other outcome mea-sures (CRF-5HT interactions, HPA-dysregulation, Porsolt Forced SwimTest) suggest essentially the opposite of the present study — thatrapid submission is more deleterious than fighting.

In sum, taking into consideration the relatively small numbers of an-imals, we failed to find major overall effects of social defeat procedureson sleep measurements or startle responses, unlike our previous stud-ies, although we did observe changes in the open field emergence testwhich may be interpreted as heightened anxiety. Clearer differencesemerged when behaviour in the social defeat confrontations weretaken into account which suggest that rapid submission might haveadaptive value for some outcome measures, including sleep.

Acknowledgements

This study was performed with support from the University ofBergen, Faculty of Psychology, and the Norwegian Competence Centrefor Sleep Disorders, Haukeland University Hospital, Bergen. The authorsexpress their thanks to Nina Harkestad for her technical assistance withpreoperative preparations and the analysis of corticosterone levels.

References

[1] Koolhaas JM, Coppens CM, de Boer SF, Buwalda B, Meerlo P, Timmermans PJ. Theresident–intruder paradigm: a standardized test for aggression, violence and socialstress. J Vis Exp 2013;77:e4367.

[2] Koolhaas JM, Hermann PM, Kemperman C, Bohus B, Van den Hoofdakker RH. Singlesocial defeat in male rats induces a gradual but long-lasting behavioural change: amodel of depression? Neurosci Res Commun 1990;7:35–41.

[3] Meerlo P, Overkamp GJ, Benning MA, Koolhaas JM, Van den Hoofdakker RH. Long-term changes in open field behaviour following a single social defeat in rats can bereversed by sleep deprivation. Physiol Behav 1996;60(1):115–9.

[4] Walker FR, Masters LM, Dielenberg RA, Day TA. Coping with defeat: acute glucocor-ticoid and forebrain responses to social defeat vary with defeat episode behaviour.Neuroscience 2009;162(2):244–53.

[5] Meerlo P, Sgoifo A, De Boer SF, Koolhaas JM. Long-lasting consequences of a so-cial conflict in rats: behavior during the interaction predicts subsequent changesin daily rhythms of heart rate, temperature, and activity. Behav Neurosci1999;113(6):1283–90.

[6] Stefanski V. Social stress in loser rats: opposite immunological effects in submissiveand subdominant males. Physiol Behav 1998;63(4):605–13.

[7] Wood SK, Walker HE, Valentino RJ, Bhatnagar S. Individual differences in reactivityto social stress predict susceptibility and resilience to a depressive phenotype: roleof corticotropin-releasing factor. Endocrinology 2010;151(4):1795–805.

[8] Wood SK, Zhang XY, Reyes BA, Lee CS, Van Bockstaele EJ, Valentino RJ. Cellular adap-tations of dorsal raphe serotonin neurons associated with the development of activecoping in response to social stress. Biol Psychiatry 2013;73(11):1087–94.

[9] Buwalda B, Kole MH, Veenema AH, Huininga M, de Boer SF, Korte SM, et al. Long-term effects of social stress on brain and behavior: a focus on hippocampal function-ing. Neurosci Biobehav Rev 2005;29(1):83–97.

[10] Koolhaas JM, Meerlo P, De Boer SF, Strubbe JH, Bohus B. The temporal dynamics ofthe stress response. Neurosci Biobehav Rev 1997;21(6):775–82.

[11] Martinez M, Calvo-Torrent A, Pico-Alfonso MA. Social defeat and subordinationas models of social stress in laboratory rodents: a review. Aggress Behav1998;24(4):241–56.

[12] StamR. PTSD and stress sensitisation: a tale of brain and body part 2: animalmodels.Neurosci Biobehav Rev 2007;31(4):558–84.

[13] Meerlo P, de Bruin EA, Strijkstra AM, Daan S. A social conflict increases EEG slow-wave activity during subsequent sleep. Physiol Behav 2001;73(3):331–5.

[14] Meerlo P, Turek FW. Effects of social stimuli on sleep in mice: non-rapid-eye-movement (NREM) sleep is promoted by aggressive interaction but not by sexualinteraction. Brain Res 2001;907(1–2):84–92.

[15] Meerlo P, Pragt BJ, Daan S. Social stress induces high intensity sleep in rats. NeurosciLett 1997;225(1):41–4.

[16] Sgoifo A, de Boer SF, Haller J, Koolhaas JM. Individual differences in plasma catechol-amine and corticosterone stress responses of wild-type rats: relationship with ag-gression. Physiol Behav 1996;60(6):1403–7.

[17] TornatzkyW, Miczek KA. Behavioral and autonomic responses to intermittent socialstress: differential protection by clonidine and metoprolol. Psychopharmacology(Berl) 1994;116(3):346–56.

[18] Meerlo P, De Boer SF, Koolhaas JM, Daan S, Van den Hoofdakker RH. Changes in dailyrhythms of body temperature and activity after a single social defeat in rats. PhysiolBehav 1996;59(4–5):735–9.

[19] Kinn AM, Grønli J, Fiske E, Kuipers S, Ursin R, Murison R, et al. A double exposure tosocial defeat induces sub-chronic effects on sleep and open field behaviour in rats.Physiol Behav 2008;95(4):553–61.

[20] Kinn Rød AM, Milde AM, Grønli J, Jellestad FK, Sundberg H, Murison R. Long-term ef-fects of footshock and social defeat on anxiety-like behaviours in rats: relationshipsto pre-stressor plasma corticosterone concentration. Stress 2012;15(6):658–70.

[21] Pulliam JV, Dawaghreh AM, Alema-Mensah E, Plotsky PM. Social defeat stress pro-duces prolonged alterations in acoustic startle and body weight gain in male LongEvans rats. J Psychiatr Res 2009;44(2):106–11.

[22] Rodgers RJ, Cole JC. Anxiety enhancement in the murine elevated plus maze by im-mediate prior exposure to social stressors. Physiol Behav 1993;53(2):383–8.

[23] Rygula R, Abumaria N, Flugge G, Fuchs E, Ruther E, Havemann-Reinecke U. Anhedo-nia and motivational deficits in rats: impact of chronic social stress. Behav Brain Res2005;162(1):127–34.

[24] American Psychiatric Association. DSM IV-TR: diagnostic and statistical manual ofmental disorders-text revision. 4th ed.Washington, D.C.: American Psychiatric Asso-ciation; 2000

[25] Steiger A, Kimura M. Wake and sleep EEG provide biomarkers in depression. JPsychiatr Res 2010;44(4):242–52.

[26] Kobayashi I, Boarts JM, Delahanty DL. Polysomnographically measured sleep abnor-malities in PTSD: a meta-analytic review. Psychophysiology 2007;44(4):660–9.

[27] Papadimitriou GN, Linkowski P. Sleep disturbance in anxiety disorders. Int Rev Psy-chiatry 2005;17(4):229–36.

63A.M. Kinn Rød et al. / Physiology & Behavior 127 (2014) 54–63

[28] Armitage R. Sleep and circadian rhythms in mood disorders. Acta Psychiatr Scand2007;115:104–15.

[29] Peterson MJ, Benca RM.Mood disorders. In: KrygerMH, Roth T, DementWC, editors.Principles and practice of sleep medicine. 5th ed. St. Louis, Missouri: ElsevierSaunders; 2011. p. 1488–500.

[30] Borbely AA, Tobler I, Loepfe M, Kupfer DJ, Ulrich RF, Grochocinski V, et al. All-nightspectral-analysis of the sleep EEG in untreated depressives and normal controls. Psy-chiatry Res 1984;12(1):27–33.

[31] Tekell JL, Hoffmann R, Hendrickse W, Greene RW, Rush AJ, Armitage R. High fre-quency EEG activity during sleep: characteristics in schizophrenia and depression.Clin EEG Neurosci 2005;36(1):25–35.

[32] Woodward SH, Murburg MM, Bliwise DL. PTSD-related hyperarousal assessed dur-ing sleep. Physiol Behav 2000;70(1–2):197–203.

[33] Hall M, Thayer JF, Germain A, Moul D, Vasko R, Puhl M, et al. Psychological stress isassociated with heightened physiological arousal during NREM sleep in primary in-somnia. Behav Sleep Med 2007;5(3):178–93.

[34] Fink P, Rosendal M. Recent developments in the understanding and management offunctional somatic symptoms in primary care. Curr Opin Psychiatry 2008;21(2):182–8.

[35] Ursin R, Larsen M. Increased sleep following intracerebroventricular injection of thedelta sleep-inducing peptide in rats. Neurosci Lett 1983;40(2):145–9.

[36] Neckelmann D, Ursin R. Sleep stages and EEG power spectrum in relation to acous-tical stimulus arousal threshold in the rat. Sleep 1993;16(5):467–77.

[37] Bjorvatn B, Ursin R. Effects of the selective 5-HT1B agonist, CGS 12066B, on sleep/waking stages and EEG power spectrum in rats. J Sleep Res 1994;3(2):97–105.

[38] Milde AM, Sundberg H, Røseth AG, Murison R. Proactive sensitizing effects of acutestress on acoustic startle responses and experimentally induced colitis in rats: rela-tionship to corticosterone. Stress 2003;6(1):49–57.

[39] Miczek KA. Intraspecies aggression in rats: effects of d-amphetamine and chlordiaz-epoxide. Psychopharmacologia 1974;39(4):275–301.

[40] Archer J. Tests for emotionality in rats and mice: a review. Anim Behav1973;21(2):205–35.

[41] Hall CS. Emotional behavior in the rat. I. Defecation and urination as measures ofindividual differences in emotionality. J Comp Physiol Psychol 1934;18:385–403.

[42] Levay EA, Paolini AG, Govic A, Hazi A, Penman J, Kent S. Anxiety-like behaviour inadult rats perinatally exposed to maternal calorie restriction. Behav Brain Res2008;191(2):164–72.

[43] Ray J, Hansen S. Temperament in the rat: sex differences and hormonalinfluences on harm avoidance and novelty seeking. Behav Neurosci2004;118(3):488–97.

[44] Logge W, Kingham J, Karl T. Behavioural consequences of IVC cages on male and fe-male C57BL/6J mice. Neuroscience 2013;237:285–93.

[45] Pritchard LM, Van Kempen TA, Zimmerberg B. Behavioral effects of repeated han-dling differ in rats reared in social isolation and environmental enrichment. NeurosciLett 2013;536:47–51.

[46] Benus RF, Bohus B, Koolhaas JM, van Oortmerssen GA. Heritable variation foraggression as a reflection of individual coping strategies. Experientia1991;47(10):1008–19.

[47] Koolhaas JM, Korte SM, De Boer SF, Van Der Vegt BJ, Van Reenen CG, Hopster H, et al.Coping styles in animals: current status in behavior and stress-physiology. NeurosciBiobehav Rev 1999;23(7):925–35.

[48] Koolhaas JM, de Boer SF, Coppens CM, Buwalda B. Neuroendocrinology of copingstyles: towards understanding the biology of individual variation. FrontNeuroendocrinol 2010;31(3):307–21.

[49] Garrick T, MorrowN, Shalev AY, Eth S. Stress-induced enhancement of auditory startle:an animal model of posttraumatic stress disorder. Psychiatry 2001;64(4):346–54.

[50] Cohen H, Zohar J, Matar MA, Zeev K, Loewenthal U, Richter-Levin G. Setting apartthe affected: the use of behavioral criteria in animal models of post traumaticstress disorder. Neuropsychopharmacology 2004;29(11):1962–70.

[51] Koch M. The neurobiology of startle. Prog Neurobiol 1999;59:107–28.[52] Jovanovic T, Norrholm SD, Sakoman AJ, Esterajher S, Kozaric-Kovacic D. Altered rest-

ing psychophysiology and startle response in Croatian combat veterans with PTSD.Int J Psychophysiol 2009;71(3):264–8.