j ourna l homepage: www.e lsev ie r.com/ locate /phb
Behavioural battery testing: Evaluation and behavioural outcomes in 8 inbredmouse strains
Heena Vanmalibhai Lad, Lin Liu, Jose Luis Paya-Cano, Michael James Parsons, Rachel Kember,Cathy Fernandes, Leonard Cornelis Schalkwyk ⁎Social, Genetic and Developmental Psychiatry Centre, Institute of Psychiatry, Kings College London, De Crespigny Park, London SE5 8AF, UK
The use of large scale behavioural batteries for the discovery of novel genes underlying behavioural variationhas considerable potential. Building a broad behavioural profile serves to better understand the complexinterplay of overlapping genetic factors contributing to various paradigms, underpinning a systems biologyapproach. We devised a battery of tests to dissect and characterise the genetic bases of behaviouralphenotypes, but firstly undertook to evaluate several aspects considered potentially confounding formapping quantitative traits. These included investigating: individual versus sibling housing; testing atdifferent times during the day; battery versus non-battery testing; and initial placement within the light–dark box. Furthermore, we assessed how behavioural profiles differed in our battery across 8 inbred strains.Overall, we found the behavioural battery was most sensitive to paired-housing effects, where weight andsome measures in the open field, elevated plus maze and light–dark box differed significantly betweensibling housed and singly housed mice. Few large effects were found for testing at different times of day andbattery versus non-battery testing. Placement in the light–dark box influenced activity and durationmeasures, which profoundly affected the analysis outcome. Behavioural profiles across eight inbred strains(C57BL/6J, 129S1/SvImJ, A/J, BALB/cByJ, C3H/HeJ, DBA/2J, FVB/NJ, and SJL/J) demonstrated some robust strainranking differences for measures in the open field and light–dark tests in our battery. However, some testssuch as the elevated plus maze produced incongruous strain ranking effects across measures. The findingsreported herein bear out the promise of behavioural batteries for mapping naturally occurring variation inmouse reference populations.
A major difficulty in using mouse models to map QTLs for complextraits, such as behaviour, is the large number of animals required for theundertaking to be successful. This complication is further compoundedby the paradox of dissecting complex traits into manageable compo-nents without losing sight of the multitude of underlying factors thatmay be interacting and attributable for the trait in question.Behavioural phenotyping assays performed in isolation may oversim-plify and fail to account for the complex networks that are involved, somulti-scale phenotyping approaches are desirable as they providemuch more information on many levels and enable us to generate acomprehensive profile of a phenotype [1–3]. Studies that aim to dissectand map behaviours using these approaches have become popular inthe post-genomic era [4–8]. While comparisons between behaviouralbattery tested and naïve mice have been reported to demonstrate task-
dependent differences, in the main behavioural profiles are found to becomparable [9]. In battery testing we can make use of a composite ofmeasures across multiple tests and look for their correlation withoverlapping behavioural phenotypes, which can in turn be comple-mented by associationwith biological markers to gain a broad appraisalof the underlying mechanisms [10–12]. A further advantage is thatsince screening can be performed using the same animals throughout acarefully devised phenotyping platform, these studies can actuallyserve to reduce the number of animals needed to effectively map QTLsfor behaviours.
Mapping studies that make use of a battery of tests can only befruitful if the experimental design is given careful consideration.Homogeneous test groups that account for age and weight acrossanimals are an important starting point. However, standardisingenvironmental variables in particular, which mouse behaviouralbatteries may be sensitive to, such as ambient conditions (noise,temperature, humidity), pH of drinking water, diet, single or grouphousing, and methods of animal husbandry, are critical since they canprofoundly affect test outcome [13–17]. Behavioural differences havebeen shown to be vulnerable to housing parameters across certaininbred strains, which one group [18] demonstrated when they
compared mice that were housed individually with group-housedmice in a behavioural test battery. In particular they found that insome strains, singly housed mice habituated to test conditions fasterwhen the assay was measuring activity and exploratory behaviours,but showed more anxiety-like behaviours in the light–dark box andhyponeophagia tests, and less so in the elevated plus maze [18]. Thereis also a complex interplay of gene-environment with different formsof cage enrichment [19,20] that is apparent even with subtlemodification, which was found to significantly alter behaviours acrossinbred strains in specific tests [21]. Previous efforts of standardisationhave proven challenging and identified several parameters that areuncontrollable. A key study [15] highlighted the limitations withcross-validation between laboratories, when they demonstrated thatresults were relatively reproducible but, some parameters could notbe absolutely replicated between the three participating centresdespite equating experimental design and test apparatus.
Standardising environmental variables is necessary, but equalconsideration must be given to the criteria for experimental designespecially when devising behavioural batteries since sequence oftesting should take into account the sensitivity of specific tests thatcould significantly influence the outcome of subsequent tests. Onegroup [9] showed that some tests were more susceptible to test orderthan others in their test battery. Measures in the light–dark box wereparticularly sensitive to test order in their investigation, and acousticstartle response in C57BL/6J was also affected. Circadian rhythms canfurther influence performance in some behavioural tests [22–24],such as those that measure activity and anxiety. These measures maybe affected by the time of day as well as the cycle-phase in which thetest is carried out, particularly since mice are nocturnal mammals andtherefore testing during their alternate phase could affect the naturalresponse observed across behavioural tests.
Even when taking into account all possible confounding factors,the inherent differences that are commonly reported within andbetween laboratories in mouse phenotyping studies are likely to beattributable to subtle and specific inter-laboratory practices, yet inorder to obtain results from which we can make meaningfulinferences, adopting common standards are a prerequisite for itssuccess. The use of clearly defined objectives and employing a robustand reliable phenotyping battery at the outset will ultimately enableresults to be obtained with a reasonable level of accuracy. Ademonstrable behavioural battery [25] that was developed andvalidated across five test centres throughout Europe, showed thevalue of well defined Standard Operating Procedures (SOPs) withintheir behavioural battery, which allowed them to identify thepotential sources of variation, and where necessary refine proceduresto attain a good degree of reliable and reproducible strain rankingeffects between participating centres.
We devised a high throughput behavioural battery that aimed toindex a broad and comprehensive range of behaviours such asanxiety, locomotor activity, learning and memory, in order to mapunderlying QTLs and to further characterise the biological pathwaysinvolved within a recombinant inbred panel (BXD) and an outbredpopulation (Heterogeneous Stock) of mice. Central to our success inthis task was evaluation of the behavioural battery design, whichwas performed using the C57BL/6J strain, a progenitor of both BXDand HS. The specific aims of the evaluation exercise were to assesshow robust our experimental design was by means of the potentialconfounds frequently associated with behavioural batteries. Weinvestigated the effects of several aspects of our experimentaldesign: individually versus sibling (pair) housed; testing at differenttimes during the day (am/pm); the test run through the batteryversus testing in isolation; and the initial placement within thelight–dark box. In addition, we assessed how these behaviouralprofiles differ in this battery of tests across 8 well characterisedinbred strains, some of which are also progenitors of the HS mice weused in our study.
2. Methods
2.1. Animals
Male mice [C57BL/6J (n=111), 129S1/SvImJ (n=11), A/J (n=11),BALB/cByJ (n=11), C3H/HeJ (n=10), DBA/2J (n=11), FVB/NJ (n=11),SJL/J (n=11)] were generated in the Comparative Biology Unit animalfacilities at the Institute of Psychiatry using original stocks [respec-tive stock numbers: 000664, 002448, 000646, 001026, 000659, 000671,000671, and000686]purchased fromThe JacksonLaboratory (BarHarbor,ME, USA). We tested male mice only throughout our battery to avoid thepossible confounds of oestrous cycle effects [26,27]. Mice were weanedat 3 weeks of age and transferred at approximately 8 weeks of age to aseparate housing facility where all mice were singly housed; except for agroup (n=10) of C57BL/6J mice that were sibling housed (2 per cage) toinvestigate the effect of paired-housing on behavioural outcome in thebattery of tests. All mice were allowed to habituate for 2 weeks beforeundergoing the battery of behavioural tests.
2.2. Housing conditions
Mice were housed in standard cages measuring 30.5×13×11 cm,with food (Rat andMouse No. 1Maintenance Diet, Special Diet Services,Essex, UK) and water available ad libitum. The housing room wasmaintained on a reversed 12:12 light cycle with white lights on from20:00 to 8:00 h and red light on during the dark cycle. Behavioural testswere performed during the dark cycle between 09:30 and 19:00 h;except for the two groups in which time of day was being investigated,where theywere tested between 09:00 and 12:00 for the amgroup, andbetween 15:00 and 19:00 h for the pm group. Light intensity in thehousing roomwas 400 lx (lux) during the lights-on period and less than2 lx during the dark period [28]. Four red cluster lights (LED cluster redlight No. 310-6757; RS Components Northants, UK) of approximatewavelength 705 nm provided minimal red light during the dark phase,allowing experimenters to work with themice during their dark phase.Ambient temperature in all roomswasmaintained at 21±2 °Cwith 45%humidity level. Sawdust and nesting materials in each cage werechangedonce aweek, but never on the daybefore or theday of testing tominimize the disruptive effect of cage cleaning on behaviour. Allhousing and experimental procedures were performed in compliancewith the UK Home Office Animals Scientific Procedures Act 1986.
2.3. Experimental procedures and analysis
Table 1 illustrates the 12 test groups included to evaluate thebehavioural battery designed. Behavioural tests in the battery (Fig. 1)wereperformed in the following order startingwith those considered tobe least stressful: barrier test, home-cage activity (HC), open field (OF),novel object exploration (NO), elevated plus maze (EPM), light–darkbox (LD), primary screen of SHIRPA (SmithKline Beecham Pharma-ceuticals; Harwell, MRC Mammalian Genetics Unit; Imperial CollegeSchool of Medicine at St Mary's; Royal London Hospital, StBartholomew's and Royal London School of Medicine; PhenotypeAssessment), puzzle box (PB), Morris water maze (MWM), and tailsuspension test (TS). Tests in the non-battery tested groups wereperformed at the equivalent stage within the battery. Mice weretested in a pseudorandom order and were moved to the behaviouralsuite adjacent to the housing room immediately before testingwith aminimal transfer time. Each apparatus was wiped clean with 1%Trigene® between subjects to avoid olfactory cueing influencingbehaviours. Behaviours for all tests were recorded on videotapes forfurther detailed analysis. For albino mice, automated tracking usingEthoVision software [29,30] was not possible and so a comparablemethod of hand coding was used to score locomotor activity for theOF and LD across these mice. Mice were returned to their home cageat the end of each test.
Table 1
Groups Tests Validate Comparison group(s) n
Control OF, NO, EPM, LD dark, SHIRPA, PB, MWM, TSa Single housing Sibling housed 10Sibling housed OF, NO, EPM, LD dark, SHIRPA, PB, MWM, TS Sibling housing Control 10am OF, NO, EPM, LD dark, SHIRPA, PB, MWM, TS Time of day pm 5pm OF, NO, EPM, LD dark, SHIRPA, PB, MWM, TS Time of day am 5Open field OF Barrier test effect on OF Control 10Elevated plus maze EPM Battery versus non-battery Control 10Light–dark light LD light Initial placement in LD box Light–dark dark 8Light–dark dark LD dark Initial placement in LD box Light–dark light 8Cognition OF, NO, PB, MWM, Effects of anxiety battery on cognitive tests Control 9Puzzle box PB Battery versus non-battery Control 9Morris water maze MWM Battery versus non-battery Control 9Tail suspension TS Battery versus non-battery Control 7
Home cage preceded all tests in the battery and non-battery tested groups.a OF—open field, NO—Novel Object, EPM—elevated plus maze, LD—Light–Dark, SHIRPA—primary SHIRPA screen, PB—puzzle box, MWM—Morris water maze, TS—tail suspension.
A simple assay was carried out to observe spontaneous exploratorybehaviour in the barrier test, based on others principles [31]. The testserved tohabituate themouse to being removed from its home cage andplaced in a novel test arena. A standard cage (30×13×12 cm) waspartitioned into equal compartments (15×13×12 cm) using an acrylicsheet (4 cm high) as a barrier. Lighting inside the cage was set to 200–300 lx. Each mouse was placed into a compartment and the latency toclimb over the barrier recorded, after which themouse was removed. Amaximumof 3 minwas allowed for themouse to climb over the barrier.
2.5. Home-cage activity
Adapted using standard sized breeding cages (40×25×12 cm), thistest was designed to observe spontaneous behaviours mimicked in a‘home-cage’ environment [32]. The cageswere custom-built, containing
Fig. 1. Test schedule for phenotyping (left) and illustrations (right) of the behavioural testscages during experimentation. b) View above the open field apparatus. c) A novel objectbeneath the elevated plus maze apparatus showing the transparent acrylic closed arms (‘sshowing brightly lit ‘light chamber’ at the bottom and the ‘dark chamber’ in which the mousthe puzzle box apparatus; mouse enters via underpass into covered compartment. h) Morristhe tail suspension test.
a food hopper and a fixture to hold a water bottle externally at one endof the cage, which could be accessed ad libitum from inside the cage(Fig. 1a). The test room lighting (lux) and cycle matched that of thehousing room. Fresh sawdust was placed inside each clean cage as wellas regular food andwater before the test.Micewere transferred tohomecages during themiddle of their dark cycle, between 13:00 and 14:00 h,and a Plexiglas lid (42×25.5 cm) was placed on top of the cage toprevent escaping behaviours. Activity was recorded for the first hourfollowing the transfer, in order to monitor the period to habituation.Another hourwas recorded at 22.5 h after transfer, duringwhich time itwas expected that themouse had habituated— resembling ‘home-cage’behaviour.
Video recordings were analysed in detail using the automatedtracking EthoVision software [29,30], from which mean speed (cm/s),movement frequency, duration (s) and number of visits to thefeeding/drinking area were deduced independently for the ‘transfer’and ‘home-cage’ periods.
in the battery run in order a–i, from the least to most stressful. a) View above 6 homeplaced within the open field apparatus for the novel object exploration task. d) Viewafe zone’) and the open arms (‘exploratory zone’). e) View above the light–dark boxe is residing. f) Viewing jar in the primary SHIRPA screen. g) Brightly lit compartment ofwater maze container in the visible platform session. i) Mouse suspended from a line in
The open field [33] arena used was square and custom-built ofwhite acrylic with dimensions 72×72×33 cm (Fig. 1b). White light of300–330 lx was evenly distributed across the arena during testing.
Each mouse was taken from its home cage and placed into acorner, wall-facing, of the open field arena at the start of the testsession, and open field activity was video-recorded for 5 min forfurther analysis using EthoVision. The number of faecal boli and thepresence of urination were recorded at the end of the test.
In EthoVision a square of equal distance from the periphery(36×36 cm) was defined as the ‘central zone’ in order to determinecentre locomotor activity (cm), duration (s) and frequency in the openfield. Latency (s) to enter the centre as well as peripheral locomotoractivity (cm) was measured.
2.7. Novel object exploration
The novel object test [34,35] was performed 48 h following theopen field test using the same apparatus (see above for details) tomeasure the exploratory activity of a mouse in response to a novelobject placed within the centre of the arena during the test session. Abluemetal cylinder, 20 cm height and 7 cm diameter, the top of whichhad a white surface for tracking, was the novel object introduced intothe arena (Fig. 1c). Low white lighting, at approximately 30 lx, wasevenly distributed across the arena during testing.
Each mouse was taken from its home cage and placed into acorner, wall-facing, of the open field arena at the start of the testsession. The novel object was placed in the centre of the open fieldarena 2 min into the test session, after which the mouse was allowedto explore the object freely for the final 3 min of the test.
A circular area around the object was defined as the ‘explorationzone’ in EthoVision, measuring 20 cm in diameter. Latency (s) toinitial exploration of the object, as well as the frequency and duration(s) to explore the object, and rearing behaviours during the 3 minwere measured.
2.8. Elevated plus maze
An elevated plus maze [36] constructed from black opaque acrylicwas used with each armmeasuring 30×5 cm and the central platform5×5 cm. One set of arms, opposing one another, were enclosedcompletely by a wall of transparent acrylic, 15 cm high, while theother set was open with a ledge of 0.5 cm either side of the arms(Fig. 1d). The maze was elevated 40 cm from the ground on atransparent acrylic stand. Light intensity around the maze was setbetween 300 and 330 lx.
The mouse was placed on the central platform, facing towards aclosed arm, and allowed to explore the maze freely for 5 min. Thenumber of faecal boli and the presence of urination were recorded atthe end of the test.
EthoLog version 2.25 [37] was used to score behaviours fromvideotapes of the EPM. An arm entry counted when all four paws hadentered an arm, and so was considered on the central platform whenthe forepaws were out on an arm. An arm exit was considered whenall four paws had left the arm. Behavioural indices scored included:open and closed arm entries; open and closed arm duration; latencyto first enter an open arm; and time spent on the central platform afterinitial placement.
2.9. Light–dark box
For the light–dark test [38], a custom-built box of white acrylic wasused with dimensions 44×21×21 cm. The box was partitionedunequally with a sheet of white acrylic, 21×50 cm, so that approxi-mately one-third (15×21×21 cm) of the total area was under low
lighting (10–15 lx) representing the ‘dark chamber,’ and the remainingtwo thirdsbrightly lit (300–330 lx)withwhite light,whichservedas the‘light chamber’ (Fig. 1e). A small entry within the partition, 5×7 cm,allowed mice to move between chambers freely.
Eachmouse was taken from its home cage and placed into the darkchamber facing the end wall (parallel to the partition) except for thegroup that was used to investigate initial placement in the light, whichwere placed facing the end wall (parallel to the partition) within thelight chamber. Activity in the light–dark box was video-recorded for5 min. The number of faecal boli and the presence of urination wererecorded at the end of the test.
Using EthoLog, the latency (s) to emerge from the dark chamberinto the light chamber was recorded for each mouse; for the groupplaced in the light chamber at the start of the test, latency to moveinto the darkwas recorded. Duration in each chamber and the numberof light–dark transitions were also hand coded using Etholog. A singletransition was counted when all four paws had entered a chamber. Inaddition, activity within the dark chamber and light chamber (cm)were measured using Ethovision and mean speed (cm/s) was derivedfrom these measures.
2.10. Primary screen of SHIRPA
An observational assessment of each mouse was obtained using theprimary screen of SHIRPA [39,40] that formed a semi-quantitativebehavioural profile. A transparent acrylic cylinder, height 13 cm anddiameter 13 cm, was used for viewing the mouse, and a black acrylicsheet, 20×14.5 cm,wasplacedunderneath the jar to transfer themouseinto the arena. The arena, 53×36×18 cm,floorwasdivided into a grid of10×10 cm squares and was used to determine locomotor activity; thenumber of square-crossings (of all four paws) indexed locomotoractivity.
Lighting in the test room was approximately 330 lx. Mice weretaken from the housing suite to the test room and allowed tohabituate to the room for 30 min prior to examination. Each mousewas placed in the viewing jar (Fig. 1f) for an initial assessment. Thiswas followed by a series of behavioural assessments in response toexperimenter induction; first by transfer into the arena, 30 cm aboveit, in which the number of squares crossedwas observed and recordedduring the first 30 s to determine locomotor activity, with a finalassessment of sensorimotor reflexes above and around the arena(describedmore fully in [39]).We excluded observation of behaviourswithin the supine restraint category to reduce excessive handling ofmice; additionally spontaneous activity in the viewing jar was notconsidered.
2.11. Puzzle box
Adapted from the problem solving paradigm for rats [41] in whicha series of tasks are presented in order of difficulty, the puzzle boxwasdevised to demonstrate the motivation of a mouse to solve problemswhen exposed to a brightly lit arena. Previous results from our groupsuggested that mice employ contextual memory and spatial naviga-tion to solve problems within the puzzle box [42,43].
The apparatus was custom-built of white acrylic with dimensions73×28×27.5 cm. The box was unequally partitioned into 2 compart-ments with a movable black sheet of acrylic, 28×27.5 cm, whichdesignated a ‘start-box’ on one side, 58×28×27.5 cm, and a smaller‘goal box’ on the other, 14×28×27.5 cm (Fig. 1g). Two differentpartitions were used in this test: the first (partition 1) had an entrancecut into the base of it, 4×4 cm, which enabled visibility of — andaccess to — the goal box from the start-box (Fig. 2, day 1 and trial 1);the second (partition 2) was merely a partition. An underpass,15×4×2 cm, was cut into the 2.5 cm thick-base of the apparatus,spanning the partition and allowed mice to move between compart-ments when it was in place. The test takes advantage of the
Fig. 2. Puzzle box test schedule. Run over 3 separate days, 3 trials were carried out in succession for the puzzle box tasks as indicated. An initial training phase on day 1, for learningacquisition, was followed by two different problem solving tasks (burrowing and plug) on days 2 and 3.
spontaneous exploratory behaviour of a mouse with the drive toescape an open, brightly lit area. The mouse is placed within the start-box and the latency to move into the goal box via the underpassrecorded over successive trials. Light intensity was set between 600and 700 lx across the start-box during testing, whilst the goal box wascovered with a sheet of white acrylic (b1 lx).
The puzzle box test was run over 3 consecutive days, whichincluded an initial training session on day 1 and two different problemsolving sessions on days 2 and 3 (Fig. 2). Each test session consisted of3 trials, carried out in succession, with an increased level of difficulty.
Each mouse was taken from its home cage and placed into thestart-box facing the end wall parallel to the partition, and the latencyto enter into the goal box, when all four paws were contained, wasrecorded for each trial. The mouse was returned to its home cagebetween trials. Day 1— training session: for trial 1 only partition 1 wasin place (see details above) to separate the start-box from the goalbox, through which the goal box was visible. In trials 2 and 3, fromwhich point the partition 2 was in place for the remaining trials (seeFig. 2), the objective was for the mouse to learn how to access the goalbox via the underpass. Day 2 — burrowing session: in trial 1 the mousehad to enter the goal box via the open underpass. The underpass wasthen filled with sawdust for trials 2 and 3, so themouse had to burrowin order to reach the goal box. Day 3 — plug session: in trial 1 themouse had to enter the goal box via the open underpass. A rectangularpiece of corrugated cardboard, 7.5×2.5×0.5 cm, was mounted on acardboard block, 2×3×1.5 cm, which was placed across the under-pass creating a ‘plug’. The mouse had to learn to remove the plug inorder to access the goal box for trials 2 and 3.
Mean latencies were calculated for the training, burrowing and plugtrials as an index of the problem solving abilities of each mouse acrossthe tasks.
2.12. Morris water maze
This test uses visual cues placed around a water-filled maze todemonstrate the spatial learning ability and memory of the mouseacross several trials run over a number of days [44,45]. Themouse wastrained to remember the location of an initially visible platform thatwas subsequently submerged in water during acquisition. A reducedlatency over successive trials indexed learning. In addition, on thefinal day a ‘reversal’ learning task was included when the location ofthe platformwasmoved to a different quadrant of the pool tomeasurethe flexibility of spatial orientation — ability to rapidly and accuratelylearn the new location of the platform [46].
A blue circular container made of plastic, diameter 70 cm and depth28 cm, was used for the test (Fig. 1h). The container was divided equallyinto quadrants and the edge marked to designate the following fordetailed tracking inEthoVision: right (R), left (L), target (T), andopposite(O). Each quadrant was used to determine the placement of the mouseinto the maze which alternated for successive trials. Tap water (21–22 °C) was used to fill the container up to a depth of 22 cm, giving awater surface diameter of 60 cm. A non-toxic whitener (80 g of milkpowder) was used to colour thewater. Light intensity across the surfaceof thewaterwasset to100–150 lx. A transparentplastic squareplatform,6×6×21 cm, was placed in middle of quadrant T (1 cm above for thevisible trials and 1 cmbelow the water during the acquisition phase) forall trials except reversal. For the reversal session, the platform wassubmerged 1 cm in the middle of quadrant O.
Testing for the Morris water maze was carried out over 4consecutive days, and per day each mouse underwent 2 sessions,each consisting of 4 trials (Fig. 3). An additional acquisition sessionwas includedwithin the inbred strain study. A quadrant sequencewaspre-determined for placement of the mouse into the maze acrosstrials per session e.g. TLRO, which was altered each day. Each mousewas taken from its home cage and placed into the water maze at thedesignated quadrant, wall-facing, and allowed a maximum of 60 s tolocate the platform, after which themouse was guided to the platformwith the use of a wooden plank (26×4.5×1 cm), and returned to itshome cage. The inter-trial interval was approximately 6 min. Latencyto reach the platformwas recorded for eachmousewhen all four pawswere on the platform. In session 1 (day 1), the platform was visible1 cm above the surface of the water, for each trial, which served as thetraining phase. The platform was subsequently submerged 1 cmbeneath the surface of the water for all trials in sessions 2–6 (days1–3). For sessions 7 and 8 (day 4), the platformwas placed in quadrantO, diagonally opposite the original placement quadrant T (see Fig. 3);and thewooden plankwas initially placed on the platform for trial 1 ofsession 7 to indicate its new location in the reversal task.
Mean latencies were calculated across the trials per session foreach mouse, in order to obtain session performance per mouse, fromwhich a group/strain mean was determined to give an overall meanlatency per group/strain for each session.
2.13. Tail suspension test
Two separate trials of the tail suspension test [47,48] wereperformed on each mouse at approximately the same time of dayon consecutive days, with a minimum of 24 h between trials.
Fig. 3. Test schedule for the Morris water maze. Each mouse underwent 32 trials in total; 4 trials/session and 8 trials/day over 4 days. The maze was divided into 4 equal quadrants asindicated by each circle: right (R), left (L), target (T), and opposite (O), which was used to designate the initial placement of the mouse into the maze but alternated within a session.A sequence was pre-determined and remained the same across both sessions on the day e.g. on day 1 the sequence ran TLRO. A square platform was placed in the T quadrant for alltrials except for the reversal sessions on day 4 where it was placed in quadrant O. The platform was 1 cm above the surface of the water during training, when it was visible(black square), and 1 cm below the surface during acquisition (grey square).
Lighting under test conditions was set to 350 lx. A cord (3 mmdiameter) was extended and secured between two legs of anupturned chair, at least 30 cm height from the base. A cardboardcone was placed around the tail of each mouse immediately beforethe test, with its tail extending through the tip, to prevent tailclimbing behaviours.
The mouse was suspended at approximately one-third from theend of its tail, using soft padding around the area to protect the tail,and plastic clothes pegs were used to secure the mouse to the line(Fig. 1i). Each trial was 5 min long and analysed using the mobilitydetection module in EthoVision.
2.14. Statistical analysis
The Welch two sample t-test was used to assess groupdifferences in the evaluation exercise across measures that wereparametric. The Wilcoxon rank-sum statistic was used to comparenon-parametric (boli, majority of SHIRPA scores and, latencies insome tests) results between groups. Power tests were calculatedusing a delta (true difference in means) value of 2 and standarddeviation of 1. Repeated measures ANOVA were used to calculatedifferences across sessions in the MWM and TST data. ANOVAregressed against age and weight was used to calculate inbredstrain differences across behavioural measures. Genetic effect sizeswere estimated using eta2, SSstrain/(SSstrain+SSresidual), where SSare the sums of squared deviations using a linear model andANOVA, which also included regression against age and weight. Allstatistical analyses were performed using the R statistical environ-ment [49].
3. Results
3.1. Behavioural battery evaluation
Themean (±standard deviation) age andweight across all groups, atthe start of the evaluation battery, was 86.8±4.6 days and26.0±1.5 (g),respectively. Home cage measures across all comparison groups werecomparable and showed no significant differences.
3.1.1. Control versus sibling housedThe sibling housed group differed from the control (singly housed)
group for several specific measures within some of the tests: primarySHIRPA screen; open field; light–dark box; elevated plus maze andMorris water maze. Weight (g) taken in the primary screen of SHIRPA(see Supplemental Figure a in the Appendix) indicated that siblinghoused mice were significantly heavier (pb0.05) than control micedespite no differences found between their weights measured at thestart of the behavioural battery. Touch escape rating (SupplementalFigure b in the Appendix) also in the primary screen of SHIRPA,suggested that the sibling housed group exhibited a reduced responseto touch (pb0.05). In the open field (Fig. 4a) however, the siblinghoused group to some extent indicated more anxiety-like behaviourswith an increased latency to enter the central zone of the open field,and a reduced frequency, activity (cm) and time spent (s) in thecentre compared to control mice. Similar observations of anxiety-likebehaviours were also partially reflected in the results found across theelevated plus maze (Fig. 4b) where sibling housed mice spent muchmore time on the closed arms (pb0.005), and comparatively less timeon the open arms. In contrast, duration and activity within the lightchamber in the light–dark test (Fig. 4c) demonstrated few observable
Fig. 4. Control (single housed) versus sibling housed group differences. Bars represent mean values±standard error of the mean (SEM). Here latency (ln) represents the natural logof latency (s) values where data is not normally distributed. a. Sibling housed mice displayed an overall reduced locomotor activity compared with the control (singly housed) micein the OF and, peripheral activity was significantly different between groups (p=6.87e−03*). Small differences for transitions into— and time spent within— the central area of thearena were found, and these were somewhat reduced in the sibling housed group. b. EPM activity on the open and closed arms suggest increased anxiety across the sibling housedmice, and a significantly increased duration on the closed arms (p=1.53e−03⁎⁎), which was higher than open arm duration in this group. Central platform duration was alsoreduced (p=2.94e−03⁎⁎) in the sibling housed mice c. The sibling housed group were less active, with significant differences observed for transitions between the light–darkchambers (p=2.06e−02⁎) as well as activity within the dark chamber (p=3.95e−03⁎⁎). There was little difference for activity and duration within the light chamber but asomewhat increased latency to enter the light in the sibling housed group.
differences between these groups. However there was a slightincrease in the latency to enter the light chamber for the siblinghoused group. Overall locomotor activity measures were markedlyreduced in the sibling housed mice with significant differences foundfor peripheral activity (cm) in the open field (pb0.05), transitionsbetween the light–dark chambers (pb0.05) and activity within thedark chamber (pb0.005) of the light–dark test, which couldpotentially explain the increased weight observed for this group inthe primary screen of SHIRPA. Sibling housed mice spent much lesstime on the central platform (pb0.005) in the elevated plus maze
which taken together with other measures in the EPM could be anindex of exploratory activity. In the MWM (Fig. 7) main group-effectswere only seen for the reversal platform sessions (pb0.05), howeverthe between-subject difference in the sibling housed group could beattributable to this effect.
3.1.2. Time of day: am versus pmThere were no significant differences found between mice tested
during the morning compared with those tested in the afternoon forany of the behavioural parameters.
3.1.3. Barrier test effect on OFThe barrier test preceded the battery of tests in the control group and
wasnot found tohave any significant impact on the outcomeof openfieldmeasures when compared with the non-battery tested open field group.
3.1.4. Control (battery tested) versus non-battery tested EPMFew large differences were observed for open and closed arm activity
between the control and non-battery tested groups (Fig. 5a) howeverthere was some indication that the non-battery tested mice were lessactiveorexploratoryoverall on themaze.Aclear andsignificantdifference(pb0.005) was found for time spent on the central platform between thegroups, which was much reduced in non-battery tested mice.
3.1.5. Light–dark box: control (battery) versus non-battery tested, andinitial placement
Fig. 6 illustrates the influence of initial placement of mice withinthe light–dark box for the control, LDd (initial placement in the dark),
Fig. 5. Control (battery tested) versus non-battery testedmice in the EPM and PB. Bars represwhere data is not normally distributed. a. No significant differences were found in the EPM fobut some index that the non-battery testedmice were less active. A significant difference (p=goal box in the training task for non-battery tested mice was greater than in the control (batplug tasks between the two groups.
and LDl (initial placement in the light) groups. The control groupdisplayed behaviours similar to the LDd group as expected since micewere also placed initially within the dark chamber in this group. As aresult of battery versus non-battery testing, no large effects werefound between these two groups. Placement in the light–dark boxclearly influenced measures within respective start chambers, andmost noticeably mice spent an increased time in the chamber itstarted in. Duration in the light chamber was found to be muchgreater (pb0.0005) within the LDl group compared to the LDd group,which had spent more time in the dark chamber (pb0.0005). Activity(cm) in the dark chamberwas also found to be significantly reduced inthe LDl group that paralleled the duration measures however, therewere no between-group differences seen for activity within the lightchamber. Interestingly, although the LDl group spent more time in thelight chamber, their activity in the light chamber was modest andmean speed (cm/s) furthermore suggested that they were coveringless distance across time (pb0.05) compared to the LDd group. In
ent mean values±SEM. Here latency (ln) represents the natural log of latency (s) valuesr time spent on open and closed arms between the battery vs. non-battery tested groups5.36e−03⁎⁎) found for time spent on the central platform. b. Mean latency to reach the
tery tested) mice (p=4.57e−02⁎), but no large differences found in the burrowing and
Fig. 6. Placement in the light–dark box across the control, LDd (initial placement in thedark) and LDl (initial placement in the light) groups. Bars represent mean values±SEM.The control and LDd groups had a comparable response in the light–dark box and, nolarge effects found between these two groups. Duration in the light chamber wassignificantly increased in the LDl group (p=2.49e−04⁎⁎⁎) compared to the LDd group,which spent significantly (p=2.40e−04⁎⁎⁎)more time in the dark chamber. Activity inthe light chamber showed no major differences between LDd and LDl groups, whilstactivity in the dark in the LDl group (p=4.07e−04⁎⁎⁎) was much reduced parallelingduration measures. Absolute measures of activity were comparable for the control andLDd groups within each chamber, but significantly different (p=6.68e−05) betweenchambers for the LDl group. The mean speed (cm/s) was relative within each chamberfor the LDl group however, was significantly lower (p=1.45e−02⁎) in the lightchamber compared to the LDd group.
Fig. 7.Mean latency to reach the platform plotted across each session in the MWM for 6groups in the evaluation of the behavioural battery across C57BL/6J (visible = visibleplatform; h1–h5 = hidden platform; r1–2 = reversal platform). Mean values±SEM.Overall no significant differences were found between the groups per session thatperformed theMWM task. One-way ANOVA for the visible platform session revealed nosignificant between-group difference. Repeated measures ANOVA calculated forwithin-group session-differences indicated a main session-difference in all groups,and a learning effect across the hidden platform sessions (control: p=6.04e−05; Sibs:p=1.14e−02; am: p=1.08e−03; pm: p=3.00e−02; MWM: p=9.77e−03; andCognition: p=1.33e−03), but no significant group differences were found betweeneach of the comparison groups (control vs. Sibs; am vs. pm; control vs. MWM, andcontrol vs. Cognition) for the hidden platform sessions.Within-group session-differencefor the reversal platform was significant for two groups (am: p=3.74e−03 andCognition: p=2.60e−02). A group- and session-effect was found between control andSibs (p=1.83e−02 and p=8.00e−05, respectively) for reversal platform sessions,however there was also a between-subject difference in the Sibs group (p=2.43e−03)for these sessions.
addition, themean speed of the LDl group in the light chamber did notdiffer significantly from the dark chamber. In contrast, mean speed forthe LDd group in the dark chamberwasmarkedly reduced (p=0.001)compared with the light chamber, which was indicative of anxiety-like behaviours.
3.1.6. Effect of anxiety tests on the cognitive batteryMice which performed cognitive tests only from the test battery
compared with the control (battery tested) group, indicated that theanxiety tests within the battery did not significantly affect the outcomeof measures for the cognitive tests in the behavioural battery.
3.1.7. Control (battery tested) versus non-battery tested PBMean latencies (s) calculated across the sessions in the puzzle box
gave three indices: training mean latency; burrowing mean latency;and plug mean latency. In the training task (Fig. 5b), non-batterytestedmice took longer to reach the goal box (pb0.05) comparedwiththe control group. Modest but non-significant group differences were
seen between the groups for the burrowing and plug puzzle tasks.Mean latency across tasks overall reflected the level of increasingdifficulty better in the battery than non-battery tested mice.
3.1.8. Control (battery tested) versus non-battery tested MWMMean latency to reach the platform, across trials in each session,
are illustrated for all groups that performed the MWM in Fig. 7.Repeated measures ANOVA calculated for between-group session-differences showed that there was no main group effect observedacross the mean latency to reach the hidden or reversal platformsessions between the control and non-battery groups in the MWM.The within-group session-differences were significant for each groupin the hidden platform sessions (control: p=6.04e−05; and MWM:p=9.77e−03) suggesting learning acquisition, but neither groupshowed any significant session-difference for the reversal platformtask.
3.1.9. Control (battery tested) versus isolation tested TSDuration of immobility in the tail suspension test was comparable
in control (battery tested) versus non-battery tested groups, andwhilst there was an inter-trial difference apparent in both groupsthere were no significant differences found between groups.
3.2. Behavioural differences in the battery of tests across 8 inbred strains
3.2.1. Open fieldRobust strain ranking effects were observed in the open field test
(Fig. 8a) for latency and duration in the centre of the arena, as well asperipheral LMA (locomotor activity), which more or less resulted in
C57NC3HN129NFVBNDBANBALBNSJLNA for duration in the centreand peripheral LMAmeasures with a corresponding opposite effect onthe latency to enter the centre. Broad sense heritability estimatescalculated, eta2, for these measures suggested that between 44 and59% of the phenotypic variance was attributable to genetic effects anda main effect of strain (p=1e−07) was found from ANOVA regressedagainst weight and age for these measures. Centre LMA across strainsshowed less consistency in strain ranking effects and eta2 was small incomparison (13%) with no main effect of strain.
3.2.2. Novel object explorationPhenotypes in the novel object exploration test (Fig. 8b) demon-
strated some variability in terms of strain ranking, with few robusteffects seen across all measures; however strain order (DBANC57N129NFVBNC3HNSJLNBALBNA) was comparable for frequency andduration of exploration. Latency to the first exploratory episode andduration to explore the novel object indicated that strain modestlyaffected the phenotype, explaining between 30 and 50% of thevariance, and a corresponding main effect of strain (pb0.0001)matched findings for these measures. Rearing frequency was found tobe highly heritable at around 70%, and also weighted largely on strain.
Fig. 8. Strain effects in the open field and novel object exploration. Bars represent mean valueswell as peripheral LMA (locomotor activity). ANOVA regressed against weight and age also dep=8.47e−08, and peripheral LMA: p=3.35e−09). Strain effects for centre LMA were incongacrossmeasures in the novel object exploration task. Strain effectswere significant for latency (weighted largely on strain effects (p=8.74e−17).
3.2.3. Elevated plus mazeLatency to open arm entry and other open arm measures (Fig. 9,
top panel) demonstrated a relatively robust strain ranking effect,where for the most part C57N129NC3HNFVBNDBANSJLNBALBNA, forfrequency and time spent on the open armswith an opposite effect forlatency to enter the open arm. Heritability estimates were modest(33–44%) for these measures and ANOVA indicated that there weresignificant effects of strain (pb0.00001). Closed arm frequency wasreasonably heritable (40%) and the phenotype was significantlydifferent across strains, but displayed an altered strain ranking effectto that found in the open arm measures. Duration on the closed armsand central platform were highly variable showing little or nodifference across strains, corresponding with much lower heritabil-ities (b20%) and no distinct strain ranking.
3.2.4. Light–dark boxMeasures in the light–dark test illustrated a clear strain ranking
effect (Fig. 10a), with C57N129NFVBNSJLNC3HNDBANBALBNA forduration in the light and transitions between the light–darkcompartments, with the opposite effect found for latency to enterthe light. Estimates of genetic variance, eta2, indicated that these
±SEM. a. There was a robust strain ranking effect for latency and duration in the centre, asmonstrated a main effect of strain for these measures (latency: p=1.06e−12, duration:ruous, with no main strain effect (pN0.05). b. Strain ranking effects were less consistentp=2.44e−09) and exploration of the novel object (p=2.36e−06). Rearing frequencywas
Fig. 9. EPM measures across 8 inbred strains. Bars represent mean values±SEM. Latency to open arm entry, duration and frequency on the open arm (top panel) demonstrate relatively robust strain ranking effects. ANOVA indicated thatstrain effects were significant (latency: p=1.69e–07 duration: p=7.26e–05 and frequency: p=1.35e−07). Closed arm frequency was significantly different (p=1.22e−06) across strains but, strain ranking differed from open armmeasures. Duration on the closed arms and central platform were highly variable showing little (p=8.25e−03) and no difference (pN0.05) respectively, across strains with no distinct strain ranking effects.
phenotypes were reasonably heritable (∼40–60%), and ANOVAfurther corroborate that the variance was due to strain factors(pb1e−05).
3.2.5. Puzzle boxTraining mean latency in the puzzle box was found to be attributable
to strain effects (Fig. 10b) andwashighlyheritable (∼60%), howeverpost-hoc analysis revealed that this difference was only seen for comparisonswith strain A. Burrowing latency demonstrated a strain ranking effect(C57NDBANFVBNC3HN129NBALBNSJLNA) that was relatively modest,and this was also reflected in eta2 (∼40%), with ANOVA indicating thatstrain accounted for most of the variance (pb1e−05). Strain order wasaltered for mean latency in the plug task, comparedwith the training andburrowing tasks, however these differences were significant for straineffects with demonstrable heritability (∼60%).
3.2.6. Morris water mazeFig. 11a illustrates the mean latency to reach the platform, across
the trials in each session, in the Morris water maze for each inbredstrain. One-way ANOVA, calculated for latency to reach the visible
Fig. 10. Inbred strain effects in the light–dark and puzzle box. Bars represent mean values±Sconfirmed that these differences were largely weighted on strain effects (latency: p=7.01etraining mean latency in the puzzle box loaded heavily on strain (p=2.75e−13); Tukey platency demonstrated strain ranking effects that were relatively modest; ANOVA suggested atask, and these differences were significant (p=1.52e−11) for strain.
platform using strain as a factor, showed that there was a significanteffect of strain (p=1.88e−11), however Tukey post-hoc analysisrevealed that this result was apparent only with respect tocomparisons made with strain A. Repeated measures ANOVAindicated that there was a main effect of strain (p=5.64e−14) inthe hidden platform task, and there was also a significant difference(p=3.09e−10) found across sessions, but no interaction was foundbetween the strain and sessions. Looking at the hidden platformsession-difference for each strain separately, a learning effect wasobserved in all strains except C3H and BALB. In the reversal task, therewas also a main effect of strain (p=2.24e−06) but no significantdifferences were found between the 2 sessions or interaction of thestrain by sessions but, within-strain analyses revealed a reversalplatform session-difference for the C57 and 129 strains. Overall,heritability estimates across the hidden and reversed platformsessions were modest (30–45%).
3.2.7. Tail suspensionSubtle but significant effects of strain (pb0.05) were found across
both trials in the tail suspension test (Fig. 11b) and accordingly eta2
EM. a. Robust strain differences were found for measures in the light–dark test. ANOVA−07, duration: p=1.50e−06, and transitions: p=2.49e−12). b. ANOVA indicated thatost-hoc analysis found this true for comparisons made with strain A only. Burrowingmain strain effect (p=8.65e−06). An altered strain ranking was observed for the plug
Fig. 11. Morris water maze and the tail suspension test across inbred strains (visible = visible platform; h1–h6 = hidden platform; and r1–2 = reversal platform). a. Mean latencyplotted across each session for the inbred strains in the MWM. One-way ANOVA revealed that there was a significant strain difference (p=1.88e−11) for the visible platform task;Tukey post-hoc analysis showed that differences were only seen for comparisons with strain A. Repeated measures ANOVA for the hidden platform task of the MWM indicated thatthere was amain effect of strain (p=5.65e−14), and there was also a significant difference (p=3.09e−10) found across sessions but, no interaction of strain by session. There was amain strain effect (p=2.24e−06) in the reversal platform task but, no significant differences found between the 2 sessions. b. Bars represent mean values±SEM. A small straineffect was found for trial 1 (p=0.02) and trial 2 (p=0.02) for immobility in the tail suspension test. A significant (p=1.76e−04) inter-trial difference was found that also alteredthe strain ranking observed between trials but, there was no strain by trial interaction.
estimated that approximately less than 15% of the variance wasattributable to genetic effects. There was an inter-trial difference(p=1.76e−04) and an altered strain ranking was observed acrossboth trials but, there were no significant interaction effects betweenstrain and trial found.
4. Discussion
The development of a number of high throughput mouse beha-vioural phenotyping batteries, which aim to better characterise theirgenetic underpinnings, has accelerated with a greater emphasis placedon studying the underlying complex networks [3,4,6,7,17,25,35,50,51].Fundamental to fulfilling this challenge is the need for robust assaysthat account for a multitude of variables, including the subtle environ-mental stimuli which may be interacting with genetic factors [8,21,52],as well as the discrete aspects of the experimental design. These willultimately shape the accuracy of datasets generated [53,54] and deter-mine their application for translational research. To obtain an accurateand comprehensive phenotypic profile of a range of behaviours in arecombinant inbred panel and an outbred population using a battery oftests, we had to firstly evaluate the sensitivity of our proposed design,and secondly investigate how baseline behaviours differed across eightinbred strains using our behavioural battery. Table 2 summarises themain outcomes found from these investigations. The results from theevaluation exercise demonstrated that many aspects of our experimen-tal design are robustwith fewobservable differences found as a result ofthe behavioural battery, and data was largely comparable betweenbattery and non-battery tested groups for different assays. Behaviouraloutcome however, proved to be sensitive to paired-housing effectsparticularly for activity and anxiety measures across tests. Thebehavioural profiles across eight inbred strains for these tests in
contrast, showed some large and significant strain ranking effects thatwere often consistent for specific behaviours within an assay.
Weight (g) taken at weaningwas homogeneous across test groups,yet this measure differed in the primary SHIRPA screen betweencontrol (single housed) and sibling housed mice. Our findings aresupported by other studies [18,55], in which C57BL/6J mice werehoused singly or in groups of 5, where they also found that singlehoused mice were smaller compared to group-housed mice that hadgained weight. Locomotor activity was somewhat reduced in thesibling housed group across tests in our battery, which could explainthe weight difference observed here. Performance in the open fieldand elevated plus maze tests suggested that for parameters whichindex anxiety, sibling housed mice spent much less time within theexposed zones of the apparatus compared to the control (singlyhoused) group. However these results were not apparent formeasures in the light–dark box, in which few differences werefound between groups. Similar effects of individual housing have beenreported for the elevated plus maze and light–dark box [18], wherereduced anxiety-like behaviours were seen on the elevated plus mazebut, no significant differences were found between housing groups fortime spent in the light chamber. It is possible that these effects areparadigm specific where the magnitude of anxiety scores may belimited in the light–dark test by exploratory behaviours within eachdistinct and separate sub-compartment of the apparatus; facets whichare less distinguishable in the open field and elevated plus maze tests.To some degree we see this effect of exploration from activitymeasures in the light–dark box (Fig. 4c), which demonstrates littleabsolute difference between the dark and light chambers within eachgroup. Overall activity measures were much reduced in the siblinghoused group, and they displayed a somewhat increased latency toexplore the zones of the test apparatus considered to be exposed or
Table 2
Single vs. siblinghoused
Battery vs.non-battery
am vs.pm
Heritabilityeta2
Open fieldLatency to enter centre = = = HighCentre frequency = = = HighDuration in the centre = = = ModestCentre activity = = = LowPeripheral activity ↑ vs. ↓ = = Modest
Novel objectLatency to explore object = = ModestFrequency of exploration = = ModestDuration of exploring = = Modest
Elevated plus mazeLatency to open arm = = = ModestOpen arm entries = = = ModestOpen arm duration = = = ModestCentral platform duration ↑ vs. ↓ ↑ vs. ↓ = LowClosed arm duration ↓ vs. ↑ = = Low
Light–dark boxLatency to enter light = = = ModestLight/dark transitions ↑ vs. ↓ = = HighDuration in the light = = = ModestActivity in the light = = = HighActivity in the dark ↑ vs. ↓ = = Modest
unsafe. These results found are in line with those reported in anotherstudy [18], and may suggest an effect of single versus paired-housingon habituation to the test situation. There was a difference foundbetween these groups for the reversal platform sessions in the Morriswater maze however, this was due to the large variance found withinthe sibling housed group for these sessions.
Testing at different times of day did not have any major effects onthe behavioural outcome for tests in our battery; however theprobability ofmaking a type II error for these groups (n=5) is 0.2, andso the power to detect an effect 0.80 (δ=2; standard deviation=1),which is large enough to detect big differences but perhaps unable todetect subtle effects.
Generally, comparisons made between battery tested and non-battery tested groups for most tests showed very few largedifferences, which is similar to results reported in another study [9]where only specific measures were sensitive to the battery testing. Inthe elevated plus maze, central platform duration was significantlylower in non-battery tested mice, however in accordance with openand closed arm measures there was some suggestion that this groupwere less active overall that could explain this difference. Theseresults are contrary to those found by another group [56] in whichexperimentally naïve C57BL/6JOlaHsd were shown to be much more
active on the EPM compared with battery tested mice. The differentsub-strain used here could reflect the altered observation from theresults found in our study. A small but significant differencewas foundfor training mean latency in the puzzle box, and indicated that non-battery tested mice initially took longer to reach the goal box but,there were no group differences found for the more complexburrowing or plug tasks. In the Morris water maze, while the variancewas large between the groups in the visible platform session, nodifferences were found between battery and non-battery tested micefor the hidden and reversal platform tasks. Furthermore, cognitivetests performed without the anxiety assays preceding them showedno marked differences when compared to battery tested mice.
Methods regarding the placement of mice within the light–darkapparatus have varied with most starting mice in the light chamber[1,9,27,35,57–61], since the test was originally devised [38], and a fewstarting mice in the dark [18,62,63]. Our laboratory also routinelystarts mice within the dark chamber for the light–dark test.Investigating initial placement within the light–dark box did highlightsome important aspects of the paradigm that could profoundly affectthe inferencesmade from the test. Firstly, placement of themouse hada significant influence on both activity and duration measures, whichour results demonstrated was clearly biased within the chamber inwhich the mouse started. As a result, the group initially placed in thelight had a higher baseline duration within the light chamber, butinterestingly only a modest increase in activity was seen and theirmean velocity suggested that overall they moved much less in thelight chamber than the group that started in the dark. In contrast, thegroup that started in the dark spent significantly more time in thedark chamber in which they were highly active, with a mean speedthat was comparable to the group that started in the light.Nonetheless, this was much reduced from their mean speed withinthe light chamber. Activity measures for the group starting in the darkdemonstrated on the whole that there was no major differencebetween chambers. These results possibly argue in favour ofplacement within the dark since the mice that started in the darkspent much less time in the light chamber yet the distance travelledwas equivalent to that in the dark chamber but, their higher meanspeed in the light reflected more anxiety-like behaviours. The groupplaced in the light on the other hand displayed behaviours thatconflicted aspects of the paradigm, with an increased duration in thelight and a marked reduction in the absolute measures of activity inthe dark chamber, which equated to a mean speed that was not easilydifferentiated between chambers.
Comparing the behavioural profiles of 8 inbred strains for the testsin our battery clearly demonstrated the effect that genetic backgroundhas on some behaviours. Heritability estimates indicated that thevariance seen across measures was mainly due to the strain effects,which was also confirmed by ANOVA. Whilst some robust strainranking differences were seen across all measures in tests such as theopen field and the light–dark box, for the elevated plus maze andnovel object exploration only some of the measures showedcongruous strain ranking effects. These results are likely to be aproduct of paradigm specific differences. Remarkably, there were 3strains that repeatedly displayed reduced levels of activity andduration within the exposed zones of the apparatus for many of theanxiety based tasks, with a strain ranking BALBNSJLNA. Howeverstrain A strain was often a non-performer in some measures thatinflated the variance seen across a test. The C57 strain invariablyscored at the extreme, typically high performing end, within thesestrains for the majority of all tests in the battery, with the remainingstrains showing less consistent and altered effects of ranking acrossthe different tests.
Open field results across the inbred strains in our study indicated astrain ranking of C57NC3HN129NFVBNDBANBALBNSJLNA for mea-sures of peripheral locomotor activity, centre latency and duration.Other studies have reported similar ranking effects (C57BL/6JNC3H/
HeJNDBA/2JNBALB/cJNA/J) across measures of activity [64] andcentre time [6] in the open field. Variations in these rankings havealso been found for these strains [62,65] although it appears thatstrains which display lower levels of anxiety-like behaviours in theopen field are generally changeable for ranking effects. Novel objectexploration and frequency across the inbred strains in our studydemonstrated that the three low active strains (BALB, SJL and A)displayed little or no exploration of the object whilst DBA and C57were found to be the most exploratory strains. A study conductedacross five inbred strains [66] showed that BALB/cJ and DBA/2Jdisplayed the highest scores of exploration in response to a novelobject; however variation in findings here is more likely becausedistinct methods and type of the novel object used for the task. In theelevated plus maze, scores for time spent on the open arms andlatency to first entry demonstrated stable strain differences, andwhilst there were some effects of strain seen for closed arm frequency,an altered effect of ranking was seen for closed arm and centralplatform duration. This could indicate that closed arm frequency ismore likely an index of activity-related behaviours rather than anxiety[67]. Few studies have shown any consistent strain ranking effects forthe elevated plus maze [6,27,35,58,64,65], whichmay be attributable toseveral elements of the elevated plus maze apparatus, such astransparent/opaque arm runways and/or walls on the closed arms,that could be contributing to some of the variable effects observed. Infact, the test is found to be notorious at showing unreliable straineffects, which one group noted when they investigated the stability ofinbred strain behaviours between laboratories and across severaldecades [68]. The light–dark test revealed robust strain ranking effectswith C57N129NFVBNSJLNC3HNDBANBALBNA for time spent in thelight-box but, methodological differences in this test mean thatcomparisons of strain distribution patterns can only be fairly assessedwith studies that start mice in the dark chamber. These are under-represented in the literature for strain differences. Nevertheless, onestudy which used a comparable method in the light–dark test [62]reported strain ranking effects (FVBNC57NC3HNDBANAN129) similarto ours for percentage time spent in the light. The altered position of129 here may be explained by a different sub-strain being used in theirstudy. There were no consistent strain ranking effects seen across thedifferent tasks in the puzzle box, and discounting the skewed effects ofstrain A in the training mean latency, genetic effects appeared toaccount formost of the phenotypic variance. A small Morris watermazewas used in the current study allowing sufficiently reliable estimates ofperformance but not so large as to be aversive [17]. Furthermore,simple protocols were used to maximise performance of mice giventhat a range of inbred strains with varying abilities were tested. TheMorris water maze results showed that again strain A was largelyattributable for the phenotypic variance seen between strains acrossthe visible platform session. For most of the successive hidden platformsessions, strain A remained outside the range of the other inbredstrains, and a relative effect of strain ranking of C57NDBAN129NBALBNSJLNFVBNC3HNA reflecting success in performance was ob-served. These strain rankings may relate to the visual discriminationeffects reported across these strains in other visuo-spatial learning andmemory tasks [69], given the inclusion of several strains that have poorvisual acuity (A/J; BALB/cByJ) and susceptibility to develop retinaldegeneration (C3H/HeJ; FVB/NJ; SJL/J). In the reversal platform sessions,the ranking performance altered somewhat although the variance wasless pronounced between the other strains and strain A. Whilstheritability was found to be modest across these measures, severalothers have reported similar ranking effects [65,70,71] for acquisition inthe Morris water maze for some of these strains. Strain effects forimmobile duration in the tail suspension test were subtle but significantfor both trials, and paralleled estimates of heritability which weresmall. There was a significant inter-trial difference and strainrankings were also affected. Other studies [72–74] have observed asimilar level of inbred strain differences for baseline TST but strain
rankings are variable across all studies. Methodological differencesin scoring immobility are likely to be responsible for incongruousresults found here [47].
Male mice only were tested in this behavioural battery to avoidpossible confounds introduced by oestrous cycle effects, however it isplausible the outcome in female mice may vary for these tests sincethere is evidence that strains display sex differences in a task specificmanner [26,27]. Based on the sensitivity of our behavioural batterydesign,whichproved to be robust formost of the behaviours,wepredictthat sex effects are likely to be minimal and genetic background wouldbe responsible for most of the observed differences.
Taken as a whole, our study has demonstrated the value thatevaluating a behavioural battery design can serve and further sub-stantiates the sensitivity that some aspects, such as single versus paired-housing effects and details within experimental procedures, have onbehavioural outcome. Many studies have investigated differences inbehaviour across inbred strains and show that absolute replication is alimiting factor across laboratories [15] but in the main relatively robuststrain ranking effects can be found [25]. We too showed across eightinbred strains that consistent ranking effects can be obtained for somebehavioural scores. These findings bear out the promise of behaviouralbatteries for mapping naturally occurring variation in mouse referencepopulations.
Acknowledgements
Funding for this work was provided by the Medical ResearchCouncil (G0000170). The authors thank S. Whatley for supplying theC57BL/6J mice.
Appendix A. Supplementary data
Supplementary data associated with this article can be found, inthe online version, at doi:10.1016/j.physbeh.2009.11.007.
References
[1] Bouwknecht JA, Paylor R. Behavioral and physiological mouse assays for anxiety: asurvey in nine mouse strains. Behav Brain Res Nov 15, 2002;136(2):489–501.
[2] Fernandes C, Paya-Cano JL, Sluyter F, D'Souza U, Plomin R, Schalkwyk LC.Hippocampal gene expression profiling across eight mouse inbred strains:towards understanding the molecular basis for behaviour. Eur J NeuroSci May2004;19(9):2576–82.
[3] Valdar WS, Flint J, Mott R. QTL fine-mapping with recombinant-inbredheterogeneous stocks and in vitro heterogeneous stocks. Mamm Genome Dec2003;14(12):830–8.
[4] Flint J. Analysis of quantitative trait loci that influence animal behavior. J NeurobiolJan 2003;54(1):46–77.
[5] Henderson ND, Turri MG, DeFries JC, Flint J. QTL analysis of multiple behavioralmeasures of anxiety in mice. Behav Genet May 2004;34(3):267–93.
[6] Solberg LC, Valdar W, Gauguier D, Nunez G, Taylor A, Burnett S, et al. A protocol forhigh-throughput phenotyping, suitable for quantitative trait analysis in mice.Mamm Genome Feb 2006;17(2):129–46.
[7] Turri MG, Datta SR, DeFries J, Henderson ND, Flint J. QTL analysis identifiesmultiple behavioral dimensions in ethological tests of anxiety in laboratory mice.Curr Biol May 15, 2001;11(10):725–34.
[8] ValdarW, Solberg LC, Gauguier D, CooksonWO, Rawlins JN,Mott R, et al. Genetic andenvironmental effects on complex traits in mice. Genetics Oct 2006;174(2):959–84.
[9] McIlwain KL, Merriweather MY, Yuva-Paylor LA, Paylor R. The use of behavioraltest batteries: effects of training history. Physiol Behav Aug 2001;73(5):705–17.
[10] Arendash GW, King DL. Intra- and intertask relationships in a behavioral testbattery given to Tg2576 transgenic mice and controls. Physiol Behav Apr 15,2002;75(5):643–52.
[11] Cervino AC, Darvasi A, Fallahi M, Mader CC, Tsinoremas NF. An integrated in silicogene mapping strategy in inbred mice. Genetics Jan 2007;175(1):321–33.
[12] Chesler EJ, Wang J, Lu L, Qu Y, Manly KF, Williams RW. Genetic correlates of geneexpression in recombinant inbred strains: a relational model system to exploreneurobehavioral phenotypes. Neuroinformatics 2003;1(4):343–57.
[16] Crawley JN. Behavioral phenotyping of transgenic and knockout mice: experi-mental design and evaluation of general health, sensory functions, motor abilities,and specific behavioral tests. Brain Res Jul 17, 1999;835(1):18–26.
[17] van der Staay FJ, Steckler T. Behavioural phenotyping of mouse mutants. BehavBrain Res Nov 1, 2001;125(1–2):3–12.
[18] Voikar V, Polus A, Vasar E, Rauvala H. Long-term individual housing in C57BL/6Jand DBA/2 mice: assessment of behavioral consequences. Genes Brain Behav Jun2005;4(4):240–52.
[19] Wurbel H. Ideal homes? Housing effects on rodent brain and behaviour. TrendsNeurosci Apr 2001;24(4):207–11.
[20] Zhu SW, Yee BK, Nyffeler M, Winblad B, Feldon J, Mohammed AH. Influence ofdifferential housing on emotional behaviour and neurotrophin levels in mice.Behav Brain Res Apr 25, 2006;169(1):10–20.
[21] Tucci V, Lad HV, Parker A, Polley S, Brown SD, Nolan PM. Gene-environmentinteractions differentially affect mouse strain behavioral parameters. MammGenome Nov 2006;17(11):1113–20.
[22] Brown RE, Stanford L, Schellinck HM. Developing standardized behavioral tests forknockout and mutant mice. Ilar J 2000;41(3):163–74.
[23] Hossain SM, Wong BK, Simpson EM. The dark phase improves geneticdiscrimination for some high throughput mouse behavioral phenotyping. GenesBrain Behav Jun 2004;3(3):167–77.
[24] Roedel A, Storch C, Holsboer F, Ohl F. Effects of light or dark phase testing onbehavioural and cognitive performance in DBA mice. Lab Anim Oct 2006;40(4):371–81.
[25] Mandillo S, Tucci V, Holter SM, Meziane H, Banchaabouchi MA, Kallnik M, et al.Reliability, robustness, and reproducibility in mouse behavioral phenotyping: across-laboratory study. Physiol Genomics Aug 15, 2008;34(3):243–55.
[26] Meziane H, Ouagazzal AM, Aubert L, Wietrzych M, Krezel W. Estrous cycle effectson behavior of C57BL/6J and BALB/cByJ female mice: implications for phenotypingstrategies. Genes Brain Behav Mar 2007;6(2):192–200.
[27] Voikar V, Koks S, Vasar E, Rauvala H. Strain and gender differences in the behavior ofmouse lines commonly used in transgenic studies. Physiol Behav Jan 2001;72(1–2):271–81.
[28] Jeon CJ, Strettoi E, Masland RH. The major cell populations of the mouse retina.J Neurosci Nov 1, 1998;18(21):8936–46.
[29] Noldus LP, Spink AJ, Tegelenbosch RA. EthoVision: a versatile video trackingsystem for automation of behavioral experiments. Behav Res Methods InstrumComput Aug 2001;33(3):398–414.
[30] Spink AJ, Tegelenbosch RA, Buma MO, Noldus LP. The EthoVision video trackingsystem—a tool for behavioral phenotyping of transgenic mice. Physiol Behav Aug2001;73(5):731–44.
[31] Prior H, Sachser N. Effects of enriched housing environment on the behaviour ofyoung male and female mice in four exploratory tasks. J Exp Anim Sci 1995;37:57–68.
[32] Mill J, GalsworthyMJ, Paya-Cano JL, Sluyter F, Schalkwyk LC, Plomin R, et al. Home-cage activity in heterogeneous stock (HS) mice as a model of baseline activity.Genes Brain Behav Aug 2002;1(3):166–73.
[33] Hall CS. Emotional behavior in the rat: VI. The relationship between persistenceand emotionality. J Comp Psychol 1934;18:385–403.
[34] Misslin R, Ropartz P. Responses in mice to a novel object. Behaviour 1981;78:169–77.
[35] van Gaalen MM, Steckler T. Behavioural analysis of four mouse strains in ananxiety test battery. Behav Brain Res Oct 2000;115(1):95–106.
[36] Lister RG. The use of a plus-maze to measure anxiety in the mouse.Psychopharmacology 1987;92:180–5.
[37] Ottoni EB. EthoLog 2.2: a tool for the transcription and timing of behaviorobservation sessions. Behav Res Methods Instrum Comput Aug 2000;32(3):446–9.
[38] Crawley J, Goodwin FK. Preliminary report of a simple animal behaviormodel for theanxiolytic effects of benzodiazepines. Pharmacol Biochem Behav Aug 1980;13(2):167–70.
[39] Hatcher JP, Jones DN, Rogers DC, Hatcher PD, Reavill C, Hagan JJ, et al.Development of SHIRPA to characterise the phenotype of gene-targeted mice.Behav Brain Res Nov 1, 2001;125(1–2):43–7.
[40] Rogers DC, Fisher EM, Brown SD, Peters J, Hunter AJ, Martin JE. Behavioral andfunctional analysis of mouse phenotype: SHIRPA, a proposed protocol forcomprehensive phenotype assessment. Mamm Genome Oct 1997;8(10):711–3.
[41] Crinella FM, Yu J. Brain mechanisms in problem solving and intelligence: areplication and extension. Intelligence 1995;21(2):225–46.
[42] Galsworthy MJ, Paya-Cano JL, Liu L, Monleon S, Gregoryan G, Fernandes C, et al.Assessing reliability, heritability and general cognitive ability in a battery ofcognitive tasks for laboratory mice. Behav Genet Sep 2005;35(5):675–92.
[43] Galsworthy MJ, Paya-Cano JL, Monleon S, Plomin R. Evidence for general cognitiveability (g) in heterogeneous stock mice and an analysis of potential confounds.Genes Brain Behav May 2002;1(2):88–95.
[44] Fernandes C, Liu L, Paya-Cano JL, Gregorova S, Forejt J, Schalkwyk LC. Behavioralcharacterization of wild derived male mice (Mus musculus musculus) of the PWD/
Ph inbred strain: high exploration compared to C57BL/6J. Behav Genet Nov2004;34(6):621–30.
[45] Morris R. Developments of a water-maze procedure for studying spatial learningin the rat. J Neurosci Methods May 1984;11(1):47–60.
[46] Wolfer DP, Stagljar-Bozicevic M, Errington ML, Lipp HP. Spatial memory andlearning in transgenic mice: fact or artifact? News Physiol Sci Jun 1998;13:118–23.
[47] Lad HV, Liu L, Paya-Cano JL, Fernandes C, Schalkwyk LC. Quantitative traits for thetail suspension test: automation, optimization, and BXD RI mapping. MammGenome Jul 2007;18(6–7):482–91.
[48] Steru L, Chermat R, Thierry B, Simon P. The tail suspension test: a new method forscreening antidepressants inmice. Psychopharmacology (Berl) 1985;85(3): 367–70.
[49] R.D.C. Team. R: a language and environment for statistical computing, in Book R: alanguage and environment for statistical computing, Series R: a language andenvironment for statistical computing, Editor ed.^eds., City: R Foundation forStatistical Computing Vienna, Austria, 2004.
[50] Wehner JM, Radcliffe RA, Rosmann ST, Christensen SC, Rasmussen DL, Fulker DW,et al. Quantitative trait locus analysis of contextual fear conditioning in mice. NatGenet Nov 1997;17(3):331–4.
[51] Yang J, Zhu J, Williams RW. Mapping the genetic architecture of complex traits inexperimental populations. Bioinformatics Jun 15, 2007;23(12):1527–36.
[52] Wurbel H. Behavioral phenotyping enhanced—beyond (environmental) standardiza-tion. Genes Brain Behav Jan 2002;1(1):3–8.
[53] Bogue MA, Grubb SC, Maddatu TP, Bult CJ. Mouse phenome database (MPD).Nucleic Acids Res Jan 2007;35(no Database issue):D643–9.
[54] Grubb SC, Maddatu TP, Bult CJ, Bogue MA. Mouse phenome database. NucleicAcids Res Jan 2009;37(no Database issue):D720–30.
[55] Nagy TR, Krzywanski D, Li J, Meleth S, Desmond R. Effect of group vs. single housingon phenotypic variance in C57BL/6J mice. Obes Res May 2002;10(no 5):412–5.
[56] Voikar V, Vasar E, Rauvala H. Behavioral alterations induced by repeated testing inC57BL/6J and 129S2/Sv mice: implications for phenotyping screens. Genes BrainBehav Feb 2004;3(no 1):27–38.
[57] Belzung C, Le Pape G. Comparison of different behavioral test situations used inpsychopharmacology for measurement of anxiety. Physiol Behav Sep 1994;56(no 3):623–8.
[58] Griebel G, Belzung C, Perrault G, Sanger DJ. Differences in anxiety-relatedbehaviours and in sensitivity to diazepam in inbred and outbred strains of mice.Psychopharmacology (Berl) Feb 2000;148(2):164–70.
[59] Hascoet M, Bourin M. A new approach to the light/dark test procedure in mice.Pharmacol Biochem Behav Jul 1998;60(no 3):645–53.
[60] Holmes A, Wrenn CC, Harris AP, Thayer KE, Crawley JN. Behavioral profiles ofinbred strains on novel olfactory, spatial and emotional tests for referencememory in mice. Genes Brain Behav Jan 2002;1(no 1):55–69.
[61] Liu X, Gershenfeld HK. An exploratory factor analysis of the Tail Suspension Test in12 inbred strains of mice and an F2 intercross. Brain Res Bull May 15, 2003;60(3):223–31.
[62] Hovatta I, Tennant RS, Helton R, Marr RA, Singer O, Redwine JM, et al. Glyoxalase 1and glutathione reductase 1 regulate anxiety inmice.Nature Dec 1, 2005;438(7068):662–6.
[63] Takao K, Miyakawa T. Light/dark transition test for mice. J Vis Exp Dec 2006(1):104.[64] Trullas R, Skolnick P. Differences in fear motivated behaviors among inbred mouse
strains. Psychopharmacology (Berl) 1993;111(3):323–31.[65] Moy SS, Nadler JJ, Young NB, Perez A, Holloway LP, Barbaro RP, et al. Mouse
behavioral tasks relevant to autism: phenotypes of 10 inbred strains. Behav BrainRes Jan 10, 2007;176(1):4–20.
[66] Kim D, Chae S, Lee J, Yang H, Shin HS. Variations in the behaviors to novel objectsamong five inbred strains of mice. Genes Brain Behav Jul 2005;4(5):302–6.
[67] Fernandes C, File SE. The influence of open arm ledges and maze experience in theelevated plus-maze. Pharmacol Biochem Behav May 1996;54(1):31–40.
[68] Wahlsten D, Bachmanov A, Finn DA, Crabbe JC. Stability of inbred mouse straindifferences in behavior and brain size between laboratories and across decades.Proc Natl Acad Sci U S A Oct 31, 2006;103(44):16,364–9.
[69] Wong AA, Brown RE. Visual detection, pattern discrimination and visual acuity in14 strains of mice. Genes Brain Behav Jul 2006;5(5):389–403.
[70] Brown RE, Wong AA. The influence of visual ability on learning and memoryperformance in 13 strains of mice. Learn Mem Mar 2007;14(3):134–44.
[71] Wahlsten D, Cooper SF, Crabbe JC. Different rankings of inbred mouse strains onthe Morris maze and a refined 4-arm water escape task. Behav Brain Res Nov 30,2005;165(1):36–51.
[72] Liu X, Peprah D, Gershenfeld HK. Tail-suspension induced hyperthermia: a newmeasure of stress reactivity. J Psychiatr Res May–Jun 2003;37(3):249–59.
[73] Ripoll N, David DJ, Dailly E, Hascoet M, Bourin M. Antidepressant-like effects invariousmice strains in the tail suspension test. Behav Brain Res Aug 14, 2003;143(2):193–200.
[74] Trullas R, Jackson B, Skolnick P. Genetic differences in a tail suspension test forevaluating antidepressant activity. Psychopharmacology (Berl) 1989;99(2):287–8.
