Strains differences in the collective behaviour of zebrafish (Danio rerio) in heterogeneous environment

8.3 MB·

by  Axel Séguret

Strains differences in the collective behaviour of zebrafish (Danio rerio) in
heterogeneous environment
Axel S´eguret,∗ Bertrand Collignon, and Jos´e Halloy†
Univ Paris Diderot, Sorbonne Paris Cit´e
LIED, UMR 8236, 75013, Paris, France
Recent studies show differences in individual motion and shoaling tendency between strains of
the same species. Here, we analyse the collective motion and the response to visual stimuli in two
morphologically different strains (TL and AB) of zebrafish. For both strains, we observe 10 groups
of 5 and 10 zebrafish swimming freely in a large experimental tank with two identical attractive
landmarks (cylinders or disks) for one hour. We track the positions of the fish by an automated
tracking method and compute several metrics at the group level. First, the probability of presence
shows that both strains avoid free space and are more likely to swim in the vicinity of the walls of
6
1 thetankandtheattractivelandmarks. Second,theanalysisoflandmarksoccupancyshowsthatAB
0 zebrafisharemorepresentintheirvicinitythanTLzebrafishandthatbothstrainsregularlytransit
2 fromonelandmarktotheotherwithnopreferenceonthelongduration. Finally,TLzebrafishshow
ahighercohesionthanABzebrafish. Thus,landmarksanddurationoftherepicatesallowtoreveal
n
collectivebehaviouralvariabilitiesamongdifferentstrainsofzebrafish. Theseresultsprovideanew
u
insight into the need to take into account individual variability of zebrafish strains for studying
J
collective behaviour.
7
2 Keywords: Collectivebehaviour—Decisionmaking—Zebrafish—Phenotypes

]
M I. INTRODUCTION trial[18]). Thesestudieshaveshownforexamplethat
landmarks in a bare tank arouse interest and attract
Q
the fish [18, 19] and that variations of the shape of
o. Collective decision making has been evidenced in theselandmarkscanchangeterritoryfeatures[20,21].
i many animal species and contexts [1] including food While these studies provide numerous insights on
b collection [2], problem-solving [3, 4], collective move- theindividualandcollectivepreferencesinfishgroups,
-
q ment [5–11] or nest site selection [12, 13]. In this they generally rely on the comparison between two or
[ later case, social animals have to select a resting sites more qualitatively different alternatives. Thus, the
among severals potential options in a complex envi- selection of one option is often based on an intrin-
3
v ronment. This selection can be made either through sic preference for a particular feature in comparison
4 individual decisions or complex decision-making pro- with the others. Such asymmetric choices may hide
2 cessesinvolvingtheparticipationofallindividuals[14] the collective decision that results from the internal
0 and can be temporary or permanent according to the processes of decision-making of the group.
0 needs and living style of the considered species. Here, our aim is to characterise the collective be-
0
haviourofgroupsofzebrafishswimminginanenviron-
. While this process of collective decision has been
2 ment with identical landmarks. We observe the col-
studied for a long time in social animals that select
0 lective motion of small shoals of different group sizes
a permanent home (social insects for example), only
6
(5 and 10 fish) and of two different zebrafish strains
1 fewstudiesaddressthisprobleminthecaseofnomad
(wild type AB or TL). Zebrafish are gregarious ver-
: speciesthatconstantlyexploretheirenvironmentsuch
v tebrate model organisms to study the cohesion of the
asbirdsorfish. Inparticular,experimentsonfishhave
Xi beengenerallydesignedtoobservepreferencesforpar- group and its decision making [22, 23]. Originating
from India where they live in small groups in shallow
r ticular environmental features or landmarks during a
a relatively short experimental time (few seconds [15], freshwaters[24],thezebrafishisadiurnalspeciesthat
prefers staying in groups both in nature and in the
5 minutes [16], 10 minutes [17], up to 30 minutes per
laboratory [7, 25, 26]. To be closer to their natural
swimming habits, our experimental method relies on
theobservationoffishfreelyswimminginanopenen-
∗Electronicaddress: [email protected] vironment rather than in a constraining set-up (i.e.
†Electronicaddress: [email protected] mazeasusedin[15–17]). Weobserveduringonehour
2
each group of fish swimming in a large experimental less observed near the cylinders but are still present
tank with two spots of interest (landmarks). along the walls of the tank (Figure 1). The proba-
The landmarks consist of two striped yellow-green bilities of presence computed for each experiment are
opaque plastic cylinders placed in the water column shown in the annexe (Figure 11, Figure 12, Figure 13
ortwobluetransparentfloatingPerspexdisksprovid- and Figure 14).
ing shadow. We choose these colours in the visible
spectrum of the zebrafish according to the results of
[27]. We expect that these landmarks placed in a ho-
mogeneous environment could induce a choice of one
preferedoptionbythezebrafishasevidencedforother
speciesfacedwithidenticalressources[13,28,29]. On
the one hand, we test with cylinders the effect of vi-
sual and physical cues in the water column on collec-
tive choices. Since zebrafish are known to swim along
thewallsofexperimentaltanks,cylinderscouldactas
such walls in the water column. On the other hand,
we test with floating disks the impact of visual and
physical cues above water, on collective choices. We
placed disks and cylinders landmarks to see whether
and how the two strains of zebrafish will adapt their
group behaviour and their preferences for landmarks.
We quantified behavioural properties of groups of
zebrafishfromhighresolutionvideoswiththehelpofa FIG.1: Probabilityofpresenceof(A)10ABzebrafish,
trackingsystemandanautomatedanalysis. Following (B) 10 TL zebrafish, (C) 5 AB zebrafish and (D) 5 TL
themethodof[30],wetrackinrealtimethepositions zebrafish in a tank with two cylinders. The probability
of the fish and automatically compute their prob- is calculated on the positions of all zebrafish (i. e. 5 or
abilities of presence in the tank, their interindivid- 10 individuals) observed during one hour and cumulated
ual distances, the number of individual present near for 10 trials. The response to the landmarks is strain and
groupsizedependant: while5ABand5TLzebrafishshow
the landmarks as well as the duration of their visits.
similarprobabilityofpresencenearthecylinders,alarger
Then, we compare the distributions obtained for each
group size of AB increases the response to the landmarks
group size and landmark to highlight differences be-
but decreases the response of groups of 10 TL.
tween strains of zebrafish at the collective level. This
methodologybasedonmassivedatagatheringhasnow
To highlight the differential effect of the cylinders
become standard in studies on animal collective be-
on the tested groups, we measured the proportion of
haviour with flies, Drosophila melanogaster [31, 32],
positionsthatweredetectedat 25cmfromthecentre
birds, Sturnus vulgaris [33–35] and fish, Notemigonus
of the cylinders (Figure 2). This measure confirms
crysoleucas [36].
thatgroupsof10ABzebrafishweremorepresentnear
the cylinders than groups of 5 AB. In contrast, while
groups of 5 TL responded similarly to groups of 5
II. RESULTS
AB,groupsof10TLzebrafishwerelessdetectednear
the cylinders. A two-way ANOVA (group size, strain,
A. Collective behaviour in heterogeneous n = 10 for each experimental conditions) indicated a
environment with cylinders non-significativeeffectofthegroupsize(p=0.07,F=
3.59, df= 1)but asignificanteffect ofthe strain(p<
In the presence of two striped yellow and green 0.001, F=43.17, df=1)andasignificantinteraction
cylinders(seeMethods),thegroupsof10ABzebrafish strain/group size (p < 0.001, F = 35.38, df = 1) on
are mainly present along the wall of the tank and theattractivityofthecylinders. Theinteractioneffect
aroundthecylinders, asshownbytheiraverageprob- indicates here a strain-specific effect of the group size
abilityofpresencecomputedfortheonehourobserva- on the time spent near the landmarks : groups of 10
tion period and 10 replicates (Figure 1). On the con- AB are more attracted by the cylinders than groups
trary, groups of 5 TL, 10 TL and 5 AB zebrafish are of 5 AB but on the contrary, 10 TL are less detected
3
Whitney U test, U = 48, p = 0.455 ; outside, Mann-
1.0 Whitney U test, U = 38, p=0.192).
Group size effect: p = 0.07 (ns) AB strain
s Strain effect: p < 0.001 (***) TL strain Then, we characterised the transitions of the fish
ot Interaction: p < 0.001 (***) from one landmark to the other by analysing the suc-
p0.8
s cession of majority events. In particular, we counted
r
a the number of ”Collective transitions” (two major-
e
e n0.6 ity events nearby different cylinders separated by a
b majority outside), the number of ”One-by-one tran-
o
t sitions” (succession of a majority event in one cylin-
y 0.4
t der and a majority event in the other cylinder) and
bili finally the number of ”Collective U-turns” (two ma-
a
ob0.2 jority events in the same cylinder separated by a ma-
r
P jorityoutside). Thisrevealsthatthemaintransitions
occurringforABzebrafisharethecollectiveoneswhile
0.0
5 AB 10 AB 5 TL 10 TL somecollectiveU-turnsandalmostnoindividualtran-
sitions were detected (Figure 4B red). Similarly, al-
mostnoindividualtransitionwereobservedfortheTL
FIG.2: Probability to be at 25 cm from the center
zebrafish that perform mainly collective transitions
of the cylinders for 10 trials with 5 AB, 10 AB, 5 TL
or10TLzebrafishina1m2 tankwithtwocylinders. The (Figure 4B green). TL zebrafish performed also nu-
black line shows the median, the red square shows the merous collective U-turns. The absence of individual
mean. Statistical tests show that groups of 10 AB are transitions reveals that both strains are mostly swim-
more attracted by the cylinders than groups of 5 AB and mingingroupsbutwithdifferentcollectivedynamics.
5 TL which are more attracted by cylinders than groups WecomparedtheresultswithMann-WhitneyUtests:
of10TL.N =10measuresfor5AB,10AB,5TLand10 ”One-by-one transitions” (U = 3.0, p < 0.001) and
TL. * = p < 0.05, ** = p < 0.01, *** = p < 0.001, ns = ”Collective U-turns” (U = 27.5, p < 0.050) between
non significant.
AB and TL are significantly different when ”Collec-
tive transitions” between AB and TL are not (U =
40, p=0.236).
near the cylinders than groups of 5 TL.
Finally, we computed the interindividual distances
Then, we studied in more details the dynamics of between all fish to characterize the cohesion of the
the presence near the cylinders of the zebrafish swim- group for both strains and group sizes. The distribu-
ming in group of 10 individuals. In the AB strain, tion of all interindividual distances (Figure 5) shows
the fish form a cohesive group that regularly transits that groups of TL zebrafish have a stronger cohesion
from one landmark to the other at the beginning of (5TL:Median=0.12mand10TL:Median=0.14m)
the trial. Then, the group starts to split in multiple than the groups of AB zebrafish (5 AB: Median =
subgroups and the periodicity of the visits becomes 0.27m and 10 AB: Median = 0.23m). The intra-
less regular (Figure 3A). These oscillations are also strain comparison for the two group sizes shows that
observed for groups of the TL strain but contrarily the distribution significantly differs from each other
to AB zebrafish, this phenomenon is observed for the (Kolmogorov-Smirnovtest,5ABvs10AB,D=0.102,
whole experimental time (Figure 3B). p < 0.001; 5 TL vs 10 TL, D = 0.080, p < 0.001).
To quantify the dynamics of these transitions, we The inter-strain comparison for similar group sizes
computed the number of majority events detected also reveals a statistical difference between the distri-
near one of the two landmarks or outside of them. butions (Kolmogorov-Smirnov test, 10 AB vs 10 TL,
A majority event was counted when 7 or more indi- D = 0.184, p < 0.001; 5 AB vs 5 TL, D = 0.161,
viduals were simultaneously present in the same zone p<0.001). Thedistributionsoftheaverageinterindi-
independently of the duration of this majority event. vidual distance measured at each time step confirms
Figure 4A shows that the median and mean number these results (Figure 6, Kolmogorov-Smirnov test, 5
of majority event is always smaller in the TL strain, AB vs 10 AB, D = 0.433, p < 0.001; 5 TL vs 10 TL,
butthisdifferenceisonlysignificantforthemajorities D = 0.051, p < 0.001; 10 AB vs 10 TL, D = 0.333,
detected near one of the cylinder (cylinder 1, Mann- p < 0.001; 5 AB vs 5 TL, D = 0.464, p < 0.001. In
WhitneyUtest,U=26,p=0.038;cylinder2,Mann- addition, the analysis of the evolution of the average
4
1 trial of 10 AB zebrafish
A
10
5
0
0 100 200 300 400 500 600
s 10
l
a 5
u
0
d 600 700 800 900 1000 1100 1200
vi 10
di 5
n 0
i 1200 1300 1400 1500 1600 1700 1800

f 10
o
5
r
e 0
1800 1900 2000 2100 2200 2300 2400
b
m 10
5
u
N 0
2400 2500 2600 2700 2800 2900 3000
10
5
0
3000 3100 3200 3300 3400 3500 3600
Time (s) Outside
Cylinder 1
Cylinder 2
1 trial of 10 TL zebrafish
B
10
5
0
0 100 200 300 400 500 600
s 10
l
a 5
u
0
d 600 700 800 900 1000 1100 1200
vi 10
di 5
n 0
i 1200 1300 1400 1500 1600 1700 1800

f 10
o
5
r
e 0
1800 1900 2000 2100 2200 2300 2400
b
m 10
5
u
N 0
2400 2500 2600 2700 2800 2900 3000
10
5
0
3000 3100 3200 3300 3400 3500 3600
Time (s)
FIG. 3: Landmark occupancy time series for 1 trial of 10 individuals of (A) AB zebrafish and (B) TL zebrafish in
a tank with two cylinders. For readability, time series are divided in 6 consecutive subplots. Y-axis report the number
of individuals. Green line represents individuals in the zone of interest (less than 25 cm away from the center of the
landmark) of landmark 1, red line the individuals in the zone of interest of landmark 2 and blue line the individuals
outside both zones of interest. The results show that for each strain, the number of fish present in each zone fluctuate
in a short time range. The time series of all trials can be seen in the annexe (Figure 17 and Figure 18).
interindividual distance revealed that the cohesion of
thefishdecreasesforbothstrainsandbothgroupsizes
through the time (Figure 21 of the annexe).
5
A B
10LgroupsLofL10Lzebrafish 10LgroupsLofL10Lzebrafish
250 ns 1100ATLB 250 ns 1100ATLB
nts200 ents200
e v
orityLev150 * ns berLofLe150 *
aj100 m100
M u
N ***
50 50
0 0
CylinderL1 Outside CylinderL2 One-by-one Collective Collective
transitions transitions U-turns
FIG. 4: Landmark occupancy and transitions for 10 trials of 10 AB zebrafish and 10 TL zebrafish. (A) Number
of majority events occurring around the cylinders and outside. A majority event was considered as soon as more than
or equal to 7 fish are aggregated in the same zone. (B) Number of transitions of the majority from one zone to another
one. We mainly looked at ”One by one” transitions (the fish transit one by one from one cylinder to the other),
Collective transitions (thewholegrouptransitsbetweenbothcylindersthroughtheoutsidearea)andCollective U-turns
transitions (the group go back to the previous cylinder). For both strain, while several U-turns were observed, the
majority of transitions were made in groups (Collective transitions) and only a few were made one-by-one. The TL
strainsignificantlydiffersfromtheABstrainbyperformingsignificantlylessOne-by-onetransitionsandmoreCollective
U-turns. Eachboxplotiscomposedof10valuesofnumbersofevents. *=p<0.05, **=p<0.01, ***=p<0.001, ns
= non significant.
A 00.6.6 10:trials:of:10:AB B 00.6.6 10:trials:of:10:TL
MediaMnea:n=:±0s.2td3::::0m.34:+:0.28 MedMiaeann:=±:0st.d1:::40:.3m:+:0.32
n 0.5 Max:=Ma:1x:.=3:61.:3m4,:,M:Min:i=n:0:=.0:0:m n 0.5 MaxM:=ax::1=.:31.737:m,:M,i:nM:=i:n0.:0=:0:m:
ortioortion000.4..43 ortioortion000.4..43
pop pop
oPr00.2.2 oPr00.2.2
r r
P P
0.1 0.1
00..000.0 00..22 00..44 00..66 00..88 11..00 11..22 11..44 00..000.0 00..22 00..44 00..66 00..88 11..00 11..22 11..44
InteInrtienrdinidviivdiduuaall::ddiisstatanncec:Tem:TLmL InteInrtienrdinidviivdiduuaall::ddiisstatancnec:Tem:TLmL
C 00.6.6 10:trials:of:5:AB D 00.6.6 10:trials:of:5:TL
MediaMnea:n=:±0s.2td7::::0m.38:+:0.34 MedMiaeann:=±:0st.d1::2:0:.3m:+:0.33
n 0.5 Max:=M:a1x:.=3:61.:3m6,:,M:Min:i=n::0=.0:0:m n 0.5 MaxM:=ax::1=.:31.434:m,:M,i:nM:=i:n0.:0=:0:m
ortioortion000.4..43 ortioortion000.4..43
ropProp00.2.2 ropProp00.2.2
P P
0.1 0.1
00..000.0 00..22 00..44 00..66 00..88 11..00 11..22 11..44 00..000.0 00..22 00..44 00..66 00..88 11..00 11..22 11..44
InteInrtienrdinidviivdiduuaall::ddiisstatancnec:Tem:TLmL InteInrtienrdinidviivdiduuaall::ddiisstatancnec:Tem:TLmL
FIG. 5: Interindividual distances for 10 trials with (A) 10 AB zebrafish (N = 22 480 363 distances), (B) 10 TL
zebrafish(N=20981798distances), (C)5ABzebrafish(N=5169077distances)and(D)5TLzebrafish(N=4578
036 distances). N is the number of measured distances. The dashed lines show the medians. The distributions show
thatgroupsofTLzebrafishhaveastrongercohesionthangroupsofABzebrafish. Althoughgroupsizedoesnotchange
this distribution in groups of TL strain, it has a strong impact on the AB strain. Distributions of the interindividual
distances of 10 AB and 10 TL are different, 5 AB and 5 TL are different, 5 AB and 10 AB are different and 5 TL and
10 TL are different (see in the text the results of the statistical tests).
6
A 10LABLvs.L10LTL
0.25
10AB
10TL
0.20
n
o0.15
i
t
r
o
p
o0.10
Pr
0.05
0.00
0.0 0.1 0.2 0.3 0.4 0.5 0.6
AverageLinterindividualLdistanceL(m)
B 5LABLvs.L5LTL
0.25
5AB
5TL
0.20
n
o0.15
i
t
r
o
p
o0.10
Pr
0.05
0.00
0.0 0.1 0.2 0.3 0.4 0.5 0.6
AverageLinterindividualLdistanceL(m)
FIG. 6: Average interindividual distance (A) for 10
AB zebrafish (N = 518960 measures) versus 10 TL ze-
brafish (N= 500794measures). The red distribution rep-
resents 10 trials with groups of 10 AB zebrafish and the
green represents 10 trials with groups of 10 TL zebrafish
; (B) for 5 AB zebrafish (N = 528357 measures) versus 5
TLzebrafish(N=495659measures). Thereddistribution
represents10trialswithgroupsof5ABzebrafishandthe
greenisfor10trialswithgroupsof5TLzebrafish. Dashed
lines represent medians. TL zebrafish are more cohesive
than AB zebrafish. Smaller groups of AB zebrafish show
a shift to higher values. The distributions of the average
interindividualdistancesof10ABand10TLaredifferent,
5ABand5TLaredifferent,5ABand10ABaredifferent
and 5 TL and 10 TL are different.
7
B. Collective behaviour in heterogeneous
environment with disks 1.0
Spot effect: p < 0.001 (***) 10 AB strain
s Strain effect: p < 0.001 (***) 10 TL strain
We also placed symmetrically two floa√ting identi- pot0.8 Interaction: p = 0.02 (*)
s
cal perspex transparent blue disks, at 25 2 cm from r
a
two opposite corners along the diagonal. Acting as e
n0.6
roofs on the water, they create shades that could at- e
b
tract zebrafish. We did 10 trials on groups of 10 AB o
and 10 TL zebrafish. We observed similar behaviours y t0.4
t
than in the previous experiments with cylinders. The bili
maximumprobabilityofpresenceunderdiskswithAB a
zebrafish reaches 4.5×10−3 when for TL zebrafish it ob0.2
r
reaches only 1×10−3 (Figure 7). Again, TL zebrafish P
spent the majority of their time near the borders of
0.0
AB Disks TL Disks
the tank (the probabilities of presence for each ex- AB Cylinder TL Cylinder
periment are shown in the Figure 15 and Figure 16
of the annexe). We compared the probability to be FIG.8: Probability to be at 25 cm from the center
nearthespotsforbothstrainsandbothtypesofland- of cylinders or disks for 10 trials. Groups of 10 AB or
markswithatwo-wayANOVA(Figure8). Itrevealed 10TLzebrafishinatankwithtwocylindersortwodisks.
that the type of landmarks (p<0.001, F = 11.37, df The black line markes the median, the red square markes
= 1) and the strain of zebrafish affect the attraction the mean. A serie of tests shows that groups of 10 AB
(p < 0.001, F = 102.95, df = 1) while there is also are attracted by cylinders as much as disks, groups of 10
TL are more attracted by cylinders than disks, groups of
an evidence of an interaction effect between type of
10 AB are more attracted by the cylinders and the disks
landmarks and strains (p = 0.02, F = 5.67, df = 1).
than groups of 10 TL. N = 10 measures for AB disk, AB
Thus, while groups of 10 AB are attracted by cylin-
cylinder, TL disk and TL cylinder. * = p < 0.05, ** =
ders as much as by disks, groups of 10 TL are more
p<0.01, *** = p<0.001, ns = non significant.
attracted by cylinders than disks. Eventually, TL ze-
brafish were also significantly more cohesive than the
AB zebrafish in the presence of the floating disks, as
III. DISCUSSION
shown by the distribution of the interindividual dis-
tances(mediansfor10AB:0.35m,for10TL:0.23m,
Figure 9, Kolmogorov-Smirnov test, D = 0.135 and We investigated whether collective motion and col-
p<0.001). lective choice can differ in groups of 5 and 10 indi-
viduals of the same species (zebrafish Danio rerio)
but of different strains (AB versus TL). Both strains
were laboratory wild types, had the same age and
were raised under the same conditions. We observed
zebrafish swimming for one hour in the presence of
two identical landmarks (immersed cylinders or float-
ing disks) that were put symmetrically in the tank.
Although one hour observations show that the ze-
brafishdonotselectoneofthetwolandmarks,thefish
weremainlyswimmingtogetherandoscillatefromone
landmarktotheotherwithshortrestingtimes. Thus,
FIG.7: Probability of presenceof(A)10ABzebrafish while all individuals can be punctually present at the
in a tank with two disks, (B) 10 TL zebrafish in a tank same landmark (Figure 3, Figure 17 and Figure 18
withtwodisks. Attractivitytolandmarksisstraindepen-
of the annexe), the probability of presence computed
dant: theprobabilityofpresenceoffinding10ABzebrafish
for the entire experimental time shows that the fish
aroundthelandmarksis2timeshigherthanthatof10TL
were equally present at both stimuli. Therefore, no
zebrafish.
collective choice emerged on the long time for both
strains of zebrafish and group size. Hence, long time
8
A 10BtrialsBofB10BAB fish in the group could favour the transition from one
0.6
MedianB=B0.35Bm spot to the other while groups composed only by shy
Mean ± stdB:B0.44B+B0.33
0.5 MaxMB=axB1B=.B61.36B3m,BM,iBnMB=iBn0.B0=B0Bm individuals could show less frequent departures [40].
n0.4 A more detailed analysis show quantitative differ-
o
rti0.3 ences among the two studied strains and group sizes.
o
p Concerning the response of the fish to the landmarks,
ro0.2 we highlight that groups of 5 AB and TL zebrafish
P
0.1 show the same attraction for the cylinders by com-
putingtheprobabilityforthefishtobeobservednear
0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 these landmarks. This attraction increases for groups
InterindividualBdistanceBPmp of 10 AB zebrafish but decreases for groups of 10 TL
B 10BtrialsBofB10BTL zebrafish. Thisstraindifferenceisalsoobservedinthe
0.6
MedMiaenanB=±B0st.d2B:3B0B.m41B+B0.38 experiments with floating disks. In addition, the type
0.5 MaxMB=axB1B=.B615.6B5m,BM,iBnMB=iBn0.B0=B0Bm of the landmarks seems to be determinant for TL ze-
n0.4 brafish as they prefer objects immersed in the water
o
rti0.3 columnthanobjectslyingonthesurfaceofthewater.
po Hence,itexistsadifferenceofcollectivebehaviourbe-
ro0.2 tween the two tested strains of zebrafish.
P
0.1 This different response to heterogeneities may be
basedontheintrinsicpreferenceofthefishofapartic-
0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 ularstrainforcongenersorforlandmarks. Suchdiffer-
InterindividualBdistanceBPmp
encehasalreadybeenshowninshoalingtendencybe-
tweenseveralstrainsofguppies[41]andzebrafish[42],
FIG. 9: Interindividual distances for 10 trials with
[43]. In their natural environment, fish have to bal-
groups of (A) 10 AB zebrafish (N = 1 620 450 distances),
ancethecostsofrisksandbenefitsofmovingingroups
(B) 10 TL zebrafish (N = 1 620 450 distances) in a tank
orstayingnearlandmarks[44]. Movingingroupsfirst
with two disks. N is the number of measure of dis-
tance. Distribution of the interindividual distance shows of all prevents the fish to be static preys, and second
thatgroupsofTLzebrafishhaveastrongercohesionthan allow spatial recognition and easier food and preda-
groupsofABzebrafish. Maxandminshowthemaximum tors detection. The drawback is that the takeover of
orminimumdistancesfoundbetweenfish. AKolmogorov- the fish on the territory is punctual and they have
Smirnovtestshowsthatthedistributionsoftheinterindi- lesschancetofindareaswheretheycanhide. Staying
vidual distances of 10 AB and 10 TL are different. around landmarks gives the fish a feeling of control of
the territory and the possibility to hide from preda-
tors. In that case the drawback is that the preys will
and short time range analyses reveal opposite results rarely cross the territory of the fish.
and conclusions on collective motions. Regardingthestructureofthegroup,wenoticethat
Our methodology is complementary to typical Y- whatever the group size of TL zebrafish, the cohesion
mazeexperiments. Weextendandcomparetheircon- of the group does not change and is always stronger
clusionstoourobservationswithrepeatedinteractions thanthoseofgroupsofABzebrafish. Also,thebigger
between the fish and their environment. During a the group of AB zebrafish, the stronger the cohesion.
hour, the collective behaviour of zebrafish contrasts These differences of group cohesion may be based on
with other collective species in which spatial fidelity physical features differences between AB and TL ze-
emergesfromtheinteractionsbetweentheindividuals brafish. Cohesion differences could be explained by
thattakeplaceintherestingsites(incockroaches[37], phenotype differences between the two strains. Some
inhymenoptera[12]). Theseoscillationsfromonesite studies demonstrated that a large variability exists in
totheothercouldoriginatefromindividualdifferences theindividualmotionandshoalingtendencyoftheze-
amonggroupmembers: bold andshy behaviouralpro- brafish according to their age or strain. For example,
fileshavebeenevidencedinzebrafishaccordingtothe adults AB and casper zebrafish swim longer distance
intrinsic propensity of each fish to explore new envi- than ABstrg, EK, TU or WIK zebrafish [45]. Like-
ronments [38]. It also has been identified in other fish wise,theinterindividualdistancebetweenshoalmem-
species [39]. In this context, the presence of bolder bersdecreasesfrom16bodylengthto3.5bodylength
9
between day 7 and 5 months after fertilization [46]. each trial.
Also,ithasbeendemonstratedthatthefinsizehasan
impact on the swimming performance and behaviour
of the zebrafish [47]. Starting from the assumptions IV. METHODS
that AB and TL zebrafish have the same visual acu-
ity and the same lateral line performance, the only
1. Ethics statement
sourceofseparationbetweenbothstrainsaretheirfin
lengths and the patterns on the skin: TL zebrafish
are homozygous for leot1 and lofdt2, where leot1 is a The experiments performed in this study were con-
ducted under the authorisation of the Buffon Ethical
recessive mutationcausingspottinginadultzebrafish
and lofdt2 is a dominant homozygous viable mutation Committee(registeredtotheFrenchNationalEthical
Committee for Animal Experiments #40) after sub-
causing long fins [48], [49]. Thus, AB zebrafish have
mission to the state ethical board for animal experi-
shortfinsandTLzebrafishshowlongfins. Itmaysug-
ments.
gestthatTLzebrafishmoveahigherquantityofwater
with their long fins when swimming and thus emit a
stronger signal of presence (hydrodynamical signal).
Thus, it may be easier for conspecifics in the moving 2. Fish and housing
shoal to perceive the signal through their lateral line
andrealignthemselvesaccordingtotheirconspecifics. We acquired 500 adult wild-type zebrafish (200 AB
Iftherealignmentbecomeseasier, itissimplerforTL strain and 300 TL strain) from Institut Curie (Paris)
zebrafishtokeeptheirpositionintheshoal, whichin- and raised them in tanks of 60L by groups of 50. The
creases its cohesion. Following a similar hypothesis, zebrafish AB line show a zebra skin, short tail and
the signal of presence is weaker for AB zebrafish due fin. The zebrafish TL line show a spotted skin, long
totheirshorterfins. Thus,realignmentinthemoving tail and fin and barbel. Both strains are 3.5 cm long.
shoal is less performant and their cohesion decreases. The zebrafish used for the experiments are adult fish
Each of the two hypotheses could explain the collec- between 5 months and 18 months of age. During this
tive behaviours observed during the experiments and period, zebrafish show a shoaling tendency allowing
nothing prevents merging both of them. study of their collective behaviours. We kept fish un-
derlaboratorycondition, 27◦C, 500µSsalinitywitha
In conclusion, this study demonstrates that be-
9:15 day:night light cycle. The fish were reared in 55
havioural differences exist at the individual and col-
litres tanks and were fed two times per day (Special
lective levels in the same species of animal. The
Diets Services SDS-400 Scientic Fish Food). Water
analysis of the dynamics reveals that AB and TL ze-
pH is maintained at 7.5 and nitrites (NO2−) are be-
brafishmainlyoscillateingroupsbetweenlandmarks.
low 0.3 mg/l. We measured the size of the caudal fins
In addition, increasing the size of the group leads
of 10 AB (about 0.4 cm) and 10 TL (about 1.1 cm)
to opposite results for the two strains: groups of 10
zebrafish.
AB zebrafish are proportionally more detected near
the landmarks than groups of 5 AB while groups of
10 TL zebrafish are less attracted by the landmarks
than groups of 5 TL. Finally, the two tested zebrafish 3. Experimental setup
strains show differences at the structural level: (1)
groups of TL zebrafish are more cohesive than groups The experimental tank consists in a 1.2 m x 1.2
of AB zebrafish and (2) AB zebrafish collective re- m tank confined in a 2 m x 2 m x 2.35 m experi-
sponses to landmarks show that they are generally mental area surrounded by a white sheet, in order to
more present near the cylinders and floating disks isolate the experiments and homogenise luminosity.
than TL zebrafish. Thus, this study provides evi- The wall of the experimental tank were covered with
dences that zebrafish do not select resting site on the white tape and the water column is 6 cm. Water pH
midtermandhighlightsbehaviouraldifferencesatthe is maintained at 7.5 and Nitrites (NO2−) are below
individual and collective levels among the two tested 0.3mg/l. Theexperimentswithdiskswereperformed
strains of zebrafish. Future studies of collective be- in the experimental tank while those with cylinders
haviour should consider the tested strains, the intra- were performed in a white square arena (1 m x 1 m x
strain composition of the shoals and the duration of 0.15 m) placed in the experimental tank. Groups of
10
A B
m
c
5
1
10 cm
m
1
1 m
FIG. 10: (A) Experimental setup (1m x 1m) with two cylinders symmetrically placed and a magnified cylinder (φ = 10
cm, Height = 15 cm). (B) Experimental setup (1.2m x 1.2m) with two blue perspex disks symmetrically placed and a
disk (φ = 20 cm). In the experimental tank, the water column is 6 cm. Luminosity is ensured by 4 fluorescents lamps
of 80W placed on each side of the tank, on the floor and directed towards the walls to provide indirect lightning. The
wholeset-upisconfinedina2mx2mx2.35mexperimentalchamber(cage)surroundedbywhitesheetstoisolatethe
experiments and to homogenise luminosity.
zebrafish were randomly formed at the beginning of trials, the attractive landmarks are put in the setup
the experiments. and fish are placed with a hand net in a cylindrical
Theexperimentswererecordedbyahighresolution arena (20 cm diameter) made of Plexiglas placed in
camera(2048x2048px,BaslerScoutacA2040-25gm) the centre of our tank. Following a five minutes accli-
placed above the experimental tank and recording at matisation period, this arena is removed and the fish
15fps(framepersecond). Luminosityisensuredby4 areabletofreelyswimintheexperimentalarena. We
fluorescents lamps of 80W placed on each side of the performed 10 trials for each strain with the floating
tank, on the floor and directed towards the walls to disks and 10 trials for each combination of parame-
provide indirect lightning. ters (number of fish x strain) with the cylinders for a
To trigger interest of fish, we placed symmetrically total of 60 experiments. Each fish was never tested
in the set-up either two floating disks (φ = 20 cm) twice in the same experimental condition.
or two cylinders (φ = 10 cm, Height = 15 cm) sur-
rounded by yellow and green striped tape [50–52]. To
avoid the presence of a blind zone, the cylinders were
5. Tracking and Data analysis
slightlytiltedtowardthecentreofthetank. Thecen-
√
ter of both disks and cylinders are at 25 2 cm from
two opposite corners along the diagonal of the tank The experiments with cylinders were recorded at
(Figure 10). 15 fps and tracked online by a custom made track-
ing system based on blob detection. We call a batch
a group of 10 experiments. For these batchs, each
experiment consists of 540000 positions (10 zebrafish
4. Experimental procedure
x 54000 frames) and 270000 positions (5 zebrafish x
54000 frames). For experiments with disks, we faced
We recorded the behaviour of zebrafish swimming tracking troubles. Since the fish below the floating
in the experimental tank during one hour. Before the disks were difficult to distinguish by the program due