Article https://doi.org/10.1038/s41467-022-34113-y The emergence and development of beha- vioral individuality in clonal fish Received: 4 March 2022 Kate L. Laskowski 1,2,6 , David Bierbach1,3,4,6, Jolle W. Jolles 5 , Carolina Doran1 & Max Wolf1,4 Accepted: 13 October 2022 Behavioral individuality is a ubiquitous phenomenon in animal populations, Check for updates yet the origins and developmental trajectories of individuality, especially very early in life, are still a black box. Using a high-resolution tracking system, we 1234567890():,; mapped the behavioral trajectories of genetically identical fish (Poecilia for- 1234567890():,; mosa), separated immediately after birth into identical environments, over the first 10 weeks of their life at 3 s resolution. We find that (i) strong behavioral individuality is present at the very first day after birth, (ii) behavioral differ- ences at day 1 of life predict behavior up to at least 10 weeks later, and (iii) patterns of individuality strengthen gradually over developmental time. Our results establish a null model for how behavioral individuality can develop in the absence of genetic and environmental variation and provide experimental evidence that later-in-life individuality can be strongly shaped by factors pre- dating birth like maternal provisioning, epigenetics and pre-birth develop- mental stochasticity. For decades, human twin studies have been one of the most influential highly inbred animals (which can limit interpretations about the eco- and powerful paradigms to push the boundaries of our understanding logical relevance of the observed variation), and/or experiments that how genetic and environmental factors generate behavioral do not (or cannot due to necessary parental care) control for the social individuality1. In sharp contrast, up to now, little use has been made of environment that individuals experience after birth, in a recent study, the enormous potential of naturally clonal vertebrates as model we have shown that naturally clonal fish (the Amazon molly, Poecilia systems2,3. Experiments with naturally clonal vertebrates provide a formosa), separated on day one of their life into identical environ- unique opportunity for a true replicate-individual approach, with the ments, still exhibited behavioral individuality at seven weeks of age14. study of a large number of “twins” with naturally occurring genotypes To determine the causes and mechanisms that can generate behavioral under highly controlled experimental settings, making it possible to individuality in the absence of genetic and environmental differences, tackle some of the most fundamental open questions concerning the it is essential to first pinpoint when behavioral individuality emerges mechanisms underlying the emergence and development of indivi- and how it continues to unfold after emergence. Birth marks a critical duality early in life. time point: if individuality is present at birth, this points to pre-birth Behavioral individuality is commonly thought to be caused by influences––such as epigenetics15,16, maternal effects17,18, and/or pre- differences in genes and/or environmental conditions, including the birth developmental stochasticity19,20––as being key drivers of indivi- social environment. Challenging this paradigm, there is accumulating duality (Fig. 1A). Alternatively, it could be that individuality primarily evidence that substantial between-individual variation still develops emerges after birth. This emergence could happen both gradually even among genetically identical individuals reared under highly throughout early life, which would suggest that individuality is driven standardized conditions4–13. While much of this evidence is based on by positive feedbacks between behavior and the internal and/or 1 Department of Biology and Ecology of Fishes, Leibniz Institute of Freshwater Ecology and Inland Fisheries, Berlin, Germany. 2Department of Evolution and Ecology, University of California Davis, Davis, CA, USA. 3Faculty of Life Sciences, Albrecht Daniel Thaer-Institute of Agricultural and Horticultural Sciences, Humboldt-Universität zu Berlin, Invalidenstrasse 42, 10115 Berlin, Germany. 4Cluster of Excellence “Science of Intelligence,” Technical University of Berlin, Berlin, Germany, Marchstr. 23, 10587 Berlin, Germany. 5Centre for Ecological Research and Forestry Applications (CREAF), Campus UAB. Edifici C., 08193 Bellaterra, Spain. 6These authors contributed equally: Kate L. Laskowski, David Bierbach. e-mail: klaskowski@ucdavis.edu Nature Communications | (2022)13:6419 1 Article https://doi.org/10.1038/s41467-022-34113-y external environment21,22 (Fig. 1B), or abruptly at particular points early Specifically, we programmed cameras above each of the experimental in life, if it is linked to critical sensitive windows23,24 (Fig. 1C). tanks (using pirecorder26) to take a photo every three seconds during Here we perform an experimental “twin study” that was designed the daylight period (11 h per day) for the first 10 weeks of the animals’ to distinguish between these scenarios. Specifically, using naturally lives. We then convert these photos to a video, one per tank per day, clonal fish, the Amazon molly, we aim to (i) pinpoint the onset of and save the video on an external server (Supplementary Fig. 1). behavioral individuality during post-birth ontogeny, (ii) continuously Altogether over the entire experiment, our system generated close to 1 map its further change during early-life development and (iii) investi- million data points per animal (13,200 photos per day per gate the link between early-life individuality and later-in-life behavior, animal x 70 days of tracking = 924,000 photos). We use a custom thereby also generating a null model for how behavioral individuality tracking software27 to extract the fish’s x and y coordinates in each can develop in the absence of genetic and environmental variation. To frame, and then use these coordinates to compute fish’s swimming do so, we separate genetically identical individuals directly after birth speed and distance to the tank walls for each frame, which we use to into identical environments and then track their behavior at 3 s reso- estimate their median swimming speed, activity (proportion of time lution for the first 70 days of their lives using a custom-built recording spent moving >0.5 cm/s), variation in activity (inter-quartile range of system25,26. Importantly, unlike many other vertebrate model systems, swimming speed over the hour or day) and median distance from the the Amazon molly requires no parental care, thus enabling us to tank walls for each hour and day. We then use Bayesian linear hier- separate the experimental animals (N = 26) immediately after birth and archical models to quantify the magnitude of among-individual allowing near complete experimental control over the environment behavioral variation (i.e. behavioral individuality) and test how pat- the animal experiences during their early life. terns of individual behavioral variation change over time. Experimental fish were reared in highly standardized conditions, Our findings show that substantial behavioral individuality is and housed in identical individual tanks under controlled light and already present at the very first day of life after birth among genetically temperature conditions and with a tightly controlled food regime. To identical individuals, suggesting that pre-birth processes like epige- limit maternal variation, all experimental fish were born to mothers netics, pre-birth developmental stochasticity and/or maternal effects (Nmoth = 8, Supplementary Table 1) that themselves were reared under might play considerably more important roles in shaping behavioral standardized conditions with similar social densities, access to individuality than commonly thought. Importantly, we find that these resources and environmental conditions. Thus, we made every effort early signatures of individuality gradually strengthen over ontogeny to limit potential variation among mothers and provide identical and predict behavior up to at least ten weeks later (the end of our rearing conditions for the experimental offspring. observation period), suggesting that once individuality is triggered, it To follow individual behavior, we built a custom automated sets the starting point for further behavioral differentiation. recording system using Raspberry Pi computers and cameras25. Results A B C Behavioral individuality is present on the first day of life after birth Our first goal was to detect when behavioral individuality first Behavior appeared among our genetically identical and identically reared indi- viduals. We found that even on the very first day after birth individuals showed large consistent individual differences in behavior (repeat- ability (R) of hourly swimming speed = 0.65, 95% Credible Interval = Time Time Time [0.48, 0.80]; activity = 0.71 [0.56, 0.84]; variation in activity = 0.68 [0.52, 0.81]; border distance = 0.80 [0.64, 0.87]; Fig. 2A; Table 1, Sup- Fig. 1 | Three stylized scenarios of how individuality can emerge. Individuality plementary Table 2). This extremely early onset of individuality is a might be present at birth (A) or emerge gradually (B) or in an abrupt fashion (C) robust finding: we find that behavior was also significantly repeatable after birth (see main text for details). on each of the subsequent days of life (days 2–7; Supplementary 3 A B C Median swimming speed (cm/s) R = 0.65 2 1 0 1 2 3 4 5 6 7 8 9 10 11 a b c d e f g h 0.8 0.9 1.0 1.1 1.2 1.3 Hour since lights on Mother identity Total length (cm) Fig. 2 | Individuality is present at day one after birth and is not explained by after birth (C) was not related to an individual’s total length on their first day of life; differences in maternal identity or body size. Repeatability of median swimming see also Supplementary Table 3. Small and large points indicate hourly and daily speed at hourly intervals on the first day after birth (A); each line represents one median swimming speeds for each individual respectively; gray lines indicate individual (N = 26). Maternal identity (B) did not explain variation in swimming posterior estimates for the effects of body size on behavior. Throughout, lines and speeds among individuals. Small and large points indicate the hourly (i.e. 11 data points are colored according to the individual’s behavior in hour one on day one points per individual) and daily median swimming speeds, respectively, of indivi- (yellow represents higher swimming speeds; purple indicates lower swimming duals from each mother on day one after birth; see also Table 1. Behavior on day one speeds). Source Data are provided as a Source Data file. Nature Communications | (2022)13:6419 2 Article https://doi.org/10.1038/s41467-022-34113-y Table 1 | Maternal identity does not explain behavioral Table 2 | Morphological and physiological differences do not individuality explain individuality Model Random effects df DIC DIC DIC Model Effects Estimates [95% CI] R2marg R2cond (day one) (week one) (10 weeks) (DIC) 0 Null 4 611.96 4617.31 3666.76 1 Obs week 0.02 [−.011, 0.16] <0.001 0.36 (288.13) Weekly TL −0.02 [−0.75, 0.56] [0, 0.06] [0.14, 0.62] 1 Intercepts ID 5 335.06 3772.22 2473.94 Intercepts ID 0.09 [0.02, 0.26] Slopes ID 0.004 2 Intercepts mother 5 548.34 3999.10 2823.29 Residual [0.001, 0.009] 3 Intercepts ID 6 335.06 3771.72 2473.57 0.18 [0.14, 0.21] Intercepts mother 2 Obs week 0.02 [−0.02, 0.05] <0.001 0.39 (288.39) Weekly growth 0.09 [−0.66, 0.99] [0, 0.04] [0.14, 0.63] 4 Intercepts ID 6 234.68a 3488.37 2313.16 Intercepts ID 0.11 [0.02, 0.24] Slopes ID Slopes ID 0.003 [0, 0.009] 5 Slopes ID 6 543.77 3900.98 2777.04 Residual 0.18 [0.14, 0.21] Slopes mother 3 Obs week 0.009 [−0.02, 0.05] 0.002 0.40 (286.68) Overall growth 38.36 [−36.60, [0, 0.15] [0.15, 0.65] 6 Intercepts ID 7 234.57 3487.79 2312.49 Intercepts ID 112.13] Slopes ID Slopes ID 0.09 [0.01, 0.27] Intercepts mother Residual 0.003 [0, 0.009] 7 Intercepts ID 8 236.72 3487.93 2312.84 0.17 [0.14, 0.21] Slopes ID 4 Obs week 0.001 [−0.13, 0.19] 0.01 0.42 Intercepts mother (290.46) Weekly TL 0.03 [−0.87, 0.74] [0, 0.16] [0.16, 0.65] Slopes mother Weekly growth 0.08 [−0.98, 1.12] Overall growth 29.48 [−39.18, There is strong support for the inclusion of individual intercepts and slopes both on the first day Intercepts ID 116.96] of life, first week of life and over the entire 10-week experiment. There was no evidence that Slopes ID 0.10 [0.01, 0.27] including maternal identity as a random effect improved the models (reductions in DIC of at least Residual 0.003 [0, 0.009] 5 are considered support for inclusion of the effect in the model). All models included the fixed 0.18 [0.14, 0.22] effects of ‘hour’ (day one models) or ‘observation day’ (week one and 10-week models) and ‘body 5 Obs week 0.03 [−0.10, 0.19] 0.007 0.38 size (TL)’ (mean-centered). (288.79) Overall average TL 0.38 [−1.11, 1.68] [0, 0.12] [0.16, 0.64] a best supported model (lowest DIC) is bolded. Weekly diff TL −0.05 [−0.82, 0.57] Intercepts ID 0.09 [0.02, 0.27] Slopes ID 0.004 [0, 0.009] Residual 0.18 [0.14, 0.21] Comparison of the explanatory power of different aspects of body size (TL = total length) and Table 3) and when considering behavior across the entire first week of growth rates on weekly median swimming speed behavior. Values of less than 0.001 are listed as life combined (R = 0.57 [0.41, 0.50], Supplementary 4). As median ‘0’. R2marg and R2cond are the marginal and conditional R2 values which are the proportion of total swimming speed was tightly correlated with the other behavioral behavioral variation explained by the fixed effects only or by the fixed and random effects combined, respectively. variables throughout the entire observation period (Supplementary Fig. 2), we focus on swimming speed throughout (results for other behaviors show a similar pattern and are reported in Supplementary Table 2). different observation days: the first model was centered on day one, There was no evidence that the observed variation in swimming the second on day seven, the third on day 14 and so on until the 10th speed can be explained by maternal identity, be it on the first day of week after birth (day 70). This revealed a “fanning-out” pattern of life, the first week of life or over the entire 10-week experiment predicted intercepts of individual behavior (Fig. 3A), indicating a gra- (Table 1; Fig. 1B; days 2–7 in Supplementary 5). Similarly, there was no dual increase in among-individual variation and therefore repeatability evidence that body size (automatically extracted from the tracking of behavior throughout the first ten weeks of the fish’s lives (Fig. 3B). photos) explained the observed variation in swimming speed on the We found that individuals consistently differed in their absolute first days of life (effect of total length on day 1 = 1.49 [−0.53, 3.39]; body size over the entire observation period (R = 0.57 [0.37, 0.74]; Fig. 1C; days 2–7 in Supplementary Table 3) nor did body size or Supplementary Fig. 3), however there was no evidence for consistent growth rate appear to influence individual behavior across the first differences in weekly growth rates showing that while some individuals week of life (Supplementary Table 4) or the entire 10-week experiment were always larger (or smaller) than others, individuals were not (Table 2). There was also no evidence that brood size affected indivi- growing at consistently different rates over time (R < 0.01 [0,0.03]). dual body size (effect of brood size = −0.004 [−0.02, 0.01]). Alto- Overall, considering the entire 70-day observation period, there was gether, considering either the first week of life or any of the first seven no evidence that individual behavioral variation was significantly days of life separately, the fixed effects of observation (hour or day), explained by variation in any aspect of body size or growth rates (these body size, and growth rate explained comparatively little of the total fixed factors all explained less than 1% of the behavioral variation (R2m), behavioral variance (marginal R2 of day 1 = 0.13 [0.02, 0.30]; days 2–7 Table 2). all <0.04], Supplementary Tables 3, 4) compared to the random and fixed effects together (conditional R2 first day = 0.74 [0.57, 0.83]; days Early life individuality is predictive of later life individuality 2–7 in Supplementary Table 3), indicating that individual identity was a Our third goal was to investigate how predictive individual differences substantially better predictor of individual behavior than maternal in behavior were, at any given point in time, for behavioral differences identity, body size, growth rate or observation. later in life. We estimated the among-individual behavioral correla- tions in daily median swimming speed across all weeks in the first ten Behavioral individuality gradually strengthens over ontogeny weeks of life. We found that the among-individual correlations were Our second goal was to map the continued development of indivi- significantly and strongly positively correlated across all weeks, duality over the entire 70-day observation period. At this longer including from the first to the tenth week of life (Fig. 4A). Additionally, timescale the support for the inclusion of random slopes at the indi- we found that the among-individual correlation in behavior further vidual level (Table 1), suggests that patterns of individual variation separated in time became stronger as the animals age (interaction systematically changed over time. To account for this effect and enable between week and time interval = 0.011 [0.002, 0.02], Fig. 4B). That is, the estimation of among- and within-individual variation across behavior in a given week is always highly predictive of behavior in the development, we ran a series of models where we centered the data on following week (correlation with behavior one week apart, Fig. 4B), but Nature Communications | (2022)13:6419 3 Article https://doi.org/10.1038/s41467-022-34113-y to DNA methylation patterns or differential resource or hormone A Predicted swimming speed (cm/s) 2.0 allocation, could influence the phenotype of offspring15,17,28. For example, behavioral differences caused by changes in DNA methyla- tion status have previously been related to differences in pre-natal 1.5 maternal experience29. While Amazon mollies do not actively provision their growing embryos through a placenta30, it is plausible that dif- ferences in the internal maternal environment could still influence the 1.0 growing embryos, for example, through differences in maternal resources31 or hormones32. In this study, we limited variation in maternal experience by rearing mothers under similar social densities 0.5 and environmental conditions and indeed did not find evidence that maternal identity explained offspring behavior. Future work like ours that carefully controls both genetic and environmental effects from 0.0 birth but purposefully and systematically varies the environment and/ or conditions the mother fish experience could help further elucidate the potential of maternal effects on offspring behavioral divergence. 0.5 Another non-mutually exclusive hypothesis is that the beha- vioral variation we observed is the result of developmental stochas- ticity, that is, stochastic variation in any molecular, neurological or Repeatability physiological markers that occur over ontogeny12,13,19,20,33. An intri- 0.8 B Amongindividual Withinindividual guing possibility is that the phenotypic variation we observed here––whether arising from epigenetic, maternal, and/or develop- mental stochasticity effects––might itself be adaptive, for example, Variance estimate 0.6 as a potential bet-hedging strategy34. Generating phenotypic varia- tion among one’s offspring by such non-genetic means might be especially relevant in clonal organisms such as the Amazon molly2. There is, for example, evidence in clonal fish35, and poecilid fish 0.4 specifically36, that DNA methylation mechanisms and developmental plasticity more generally37 might be especially sensitive to environ- mental influences, offering a mechanism through which mothers can 0.2 generate variation among their otherwise genetically identical offspring. It is important to note that, as with any experiment, a limitation of 0.0 our study is that it is impossible to perfectly control every aspect of the environment. Thus, while every effort was made to minimize experi- 1 7 14 21 28 35 42 49 56 63 70 ential differences among the newly born individuals, it remains pos- Day sible that they could have experienced some aspects of their environments differently in the first hours of life (i.e. between birth and Fig. 3 | Individuality increases gradually throughout the first 70 days of development. The predicted values of median individual swimming speed diverge transfer into the experimental tanks), triggering both the observed over time (A) leading to gradual increases in the among-individual variance and behavioral individuality on the first day of life and the associated hence repeatability (B) of behavior. These models included only the 26 individuals behavioral divergence later in life, thereby providing an alternative on which we had complete data for the first 10 weeks of life to ensure that absolute explanation for our findings. levels of variation would remain comparable over time. Individual lines in (A) are Regardless of which specific mechanism originally generates colored according to their predicted behavior in week 1 with yellow indicating behavioral differences, we show here that these initial differences greater swimming speeds and purple indicating lower swimming speeds. In (B), continue to diverge and strengthen gradually over development. Such shown are the posterior modes of each variance estimate and its 95% CI. Source ‘fanning out’ patterns are considered indicative of positive feedback Data are provided as a Source Data file. loops, which can occur in response to the internal or external environment21,22,38,39. For example, increasing activity may lead to greater access or intake of resources, in turn leading to further beha- behavior early in life (e.g. weeks 1–3) is not as predictive of behavior vioral change. Phenotypic change can also be generated from internal several weeks later as is behavior later in life (correlation with behavior changes. There is evidence of allelic expression differences even five weeks apart, Fig. 4B). Altogether, our results thus demonstrate among clonal sisters of the Amazon molly40, which could contribute to that individual differences in behavior diverge and stabilize over behavioral variation among individuals. In human monozygotic twins, development. the accumulation of differences in methylation patterns across the genome correlates with patterns of phenotypic discordance over the Discussion lifetime41. Our work suggests that, in whatever way they are initiated, Evidence is accumulating that even genetically identical animals reared these early behavioral differences can influence future trajectories of under near identical conditions develop behavioral individuality4–13, behavioral development, which could have meaningful impacts on yet little is known about when exactly these differences emerge during later fitness. For example, movement-related behaviors such as aver- ontogeny and how they continue to change during early life devel- age and maximum swimming speed, have been shown to be funda- opment. We show that genetically identical individuals already exhibit mental to fishes’ ability to successfully forage, escape predation and substantial behavioral individuality on their first day of life, high- school with conspecifics42,43. Our findings therefore provide support lighting pre-birth influences as being of critical importance to initi- for two hypotheses about the emergence of individuality: individuality alizing durable behavioral differences among individuals. Epigenetic was evident at birth (e.g. Fig. 1A), and while these initial differences set and maternal effects mediated through mechanisms such as changes the seed for individuality, further differentiation in behavior was still Nature Communications | (2022)13:6419 4 Article https://doi.org/10.1038/s41467-022-34113-y Week 1 A Week 2 Week 2 Week 3 Week 3 Week 4 Week 4 Week 5 Week 5 Week 6 Week 6 1.0 Amongindividual correlation B Week 7 Week 7 0.8 Week 8 Week 8 Correlation with 0.6 behavior 1 week later 2 weeks later 3 weeks later Week 9 Week 9 4 weeks later 5 weeks later 0.4 1 2 3 4 5 6 7 8 9 Week Week 10 Fig. 4 | Earlier-in-life individuality predict later-in-life individuality. A Individual among-individual correlation between behavior in those two weeks; correlations behavior was significantly positively correlated across all 10 weeks of observation. were significant across all weeks. B Shown are the among-individual correlations Each plot represents the among-individual correlation in behavior between two between behavior in a given week and behavior 1, 2, 3, 4, or 5 weeks later. The particular weeks in the ten-week experiment. The points in the plot represent the correlations with behavior 2–5 weeks later have significantly steeper slopes than individual’s behavior for that week; throughout, individuals are colored according the correlation with behavior 1 week later showing that the among-individual cor- to their week 1 behavior where yellow represents greater swimming speeds and relation in behavior becomes more predictive of later behavior, later in life. Notice purple represents lower swimming speeds. The line on each plot is the estimated that the y-axis starts above zero. Source Data are provided as a Source Data file. possible (e.g. Fig. 1B). We saw that very early life behavior predicted Experimental breeding and design later life behavior under benign conditions where external environ- The all-female Amazon molly (Poecilia formosa) is a naturally clonal, mental variation was controlled as much as possible. Importantly, this live-bearing fish species that gives birth to broods of genetically does not to suggest that individuals are incapable of more dramatic identical offspring. Like all unisexual vertebrates, Amazon mollies are behavioral change (i.e. rank-order change) when put under greater the result of inter-specific hybridization44,45. As such, this ‘frozen environmental pressure. Future work that manipulates environmental hybrid’ has a heterozygous genome from its ancestral P. mexicana variation can better investigate the patterns and mechanisms deter- mother and P. latipinna father alleviating concerns about reduced mining how individuals adjust to external environmental cues. genetic variation and the resulting inbreeding depression often asso- Behavioral individuality is a fundamental characteristic of ani- ciated with artificially selected isogenic animals. Additionally, despite mal populations; our study provides a null model about how such their clonal nature, the Amazon’s genome shows no evidence of between-individual differences can change and develop in the increased mutation accumulation, genomic decay or transposable absence of genetic and environmental differences. The work pre- element activity suggesting the genomes of these animals are evolving sented here offers critical insights into the drivers of behavioral in similar ways as sexual species46. They reproduce through gyno- individuality: as it is commonly thought, individuality arises mainly genesis where the meiotic process is disrupted so that the eggs contain through genetic differences, differential experiences and particular a full maternal genome. The egg must be fused with a sperm from one ecological conditions; our findings suggest that subtler processes of their ancestral species to stimulate embryogenesis, but this paternal pre-dating birth may play a fundamental role in setting the seed DNA is not incorporated into the egg. This provides the opportunity to around which behavioral individuality crystalizes. Stated simply, control when reproduction occurs by controlling the females’ access individuals are born unique, but this does not preclude further to male sperm donors. behavioral change throughout life. We placed adult females, as potential mothers of experimental fish, in individual (5-gallon) breeding tanks with two Atlantic molly Methods (P. mexicana) males for one week to act as sperm donors. Amazon All animal care and experimental protocols complied with local and mollies give birth to broods of generally ~8-30 individuals. A brood is federal laws and guidelines and were approved by the appropriate born at once (i.e. all individuals are born within minutes of each other) governing body in Berlin, Germany, the Landesamt fur Gesundheit und and birth generally happens early in the day close to dawn. These par- Soziales (LaGeSo G-0224/20). ental fish were lab-bred and themselves sisters, so of the same age and Nature Communications | (2022)13:6419 5 Article https://doi.org/10.1038/s41467-022-34113-y lineage, and were kept at similar social densities and under standardized per week of each compartment, making sure the fish was away from environmental conditions throughout their lives to further minimize the compartment walls and did not show strong body curvature, and potential variation in maternal experience. Each breeding tank con- then used ImageJ software to measure total body length (mm) from tained an artificial plant as refuge and was checked frequently each day the tip of the snout to the end of the body. By averaging the mea- for the presence of offspring, especially during the morning hours when surements of the five images, we acquired one body size measure- births are most likely. Newborn mollies were always found in the ment per week. morning and then singly netted by trained animal caretakers, into individual experimental tanks where their behavior was automatically Error checking recorded for the next 70 days (see below). Moving the fish from the We collected up to 924,000 photos on each individual throughout the maternal tank to the experimental tanks was done in a standardized experimental period resulting in a total of over 24 million data points manner (i.e. individual fish were netted and placed into small dishes of collected on our experimental animals (N = 26 individuals). To ensure water and then placed in the tracking tanks to limit exposure to the air) that our tracking software accurately captured the behavior of our fish, by the same caretakers to minimize variation in experience among we checked for potential tracking errors in two ways. First, we esti- individual fish. Altogether, eight mothers provided offspring that com- mated overall error rates. To do this, we selected at random a starting pleted the entire 10-week experiment (Supplementary Table 1). frame from within a day; then we manually checked each of the sub- Experimental tanks (27 x 27 cm), made of white Perspex, con- sequent 200 frames and identified whether an error was made (fish sisted of four equally sized compartments, and were evenly lit from was not properly located by BioTracker) or not (fish was properly below using 6500K-LEDs. Environmental conditions were highly located) by visual inspection of the videos. We estimated the error rate standardized across tanks: all tanks were on the same 11:13 (L:D) light as the number of errors divided by the total number of checked schedule, water depth was maintained at 10 cm depth, temperature frames. The overall median error rate over the entire observation was maintained at 25 ± 1 °C by a room air conditioning system, and fish period was estimated to be 7%. Error rates increased earlier in the received a standardized amount of powdered flake fish food (Tetra- observation period when the fish were smaller (Supplementary Note I). Min™) twice daily. Opaque blinds surrounded the tanks to further limit As such, as a second step, we manually went through and corrected all outside disturbances. All experimental tanks were connected to the frames for the very first day of tracking (i.e. day 1 post-birth) for all fish same filtration system where water could mix in the sump tank, (~13,200 frames per individual) as this is a critical time period for one of allowing chemical cues to be shared across all experimental fish. Pre- our research questions. This ensured that the resulting behavioral data vious work has shown exposure to just chemical cues of conspecifics is were completely accurate for this day. This manual correction allowed sufficient in preventing the developmental of pathological behavior us the additional opportunity to compare how well our automatically that could be associated with development in complete isolation14. We tracked (i.e. not manually corrected) data performed compared to the initially placed a total of 40 newborn individuals into the tracking manually corrected data. We found that the automatically tracked data tanks. At the end of the 10-week experiment, we were able to achieve re-created near identical estimates of among- and within-individual complete tracking data on 26 individuals; camera malfunctions pre- variance components and most importantly the among-individual vented data collection on four individuals, two individuals jumped into correlation between the automatically tracked and manually corrected neighboring tanks causing the loss of data of all four individuals as we data was over 0.98 for our behavioral variables (Supplementary could not verify their identity; four newborn individuals escaped Note I). This strongly suggests that any errors introduced by our through holes in the water outlet of the tanks; and four individuals died automated tracking software have minimal influence of our behavioral as newborns. All results in the manuscript are on these 26 animals, variables at best and do not affect our interpretation of the results. though including data from all 40 (e.g. patterns of individual variation on the first day post birth) did not change the results or their inter- Statistical analyses pretation (see Supplementary Table 2). We used linear mixed, or hierarchical, models to partition the behavioral variation across different times periods into its among- Behavioral tracking and within-individual components. Throughout we focused our We developed a custom recording system using Raspberry Pi compu- analysis on the 26 individuals for which we had complete data for the ters, which are an upcoming low-cost, highly adaptable solution for entire 10-week observation period to ensure comparable variation many applications in the biological sciences25. Specifically, we created a over time and across models. local network of Raspberry Pi 3B + ’s, each connected to a Raspberry Our first question of interest was to test when individual differ- Pi camera positioned exactly above an experimental tank, com- ences in behavior first appeared over the course of the experiment. We manded by a lab computer, and connected to the server on the started by investigating behavior on the first day post birth (Fig. 1A, institute network (Supplementary Fig. 1). We programmed the Supplementary Table 2) and then planned to proceed in a day-by-day Raspberry Pi’s using pirecorder26 to take timestamped photos every fashion until significant repeatability in behavior was apparent (Sup- 3 s across the daily light period, each day, for 10 weeks, and store plementary Table 3). We used hourly median swimming speed (11 them automatically in dedicated, automatically named folders on the observations for each of 26 individuals) as our response variable and server. Image settings and resolution were thereby optimized to included ‘hour’ and ‘total length (TL)’ as fixed effects and ‘individual’ minimize file size while assuring image quality. After the experi- was included as our random effect of interest. Including TL as a cov- mental period, we created videos of all the recorded images of each ariate allowed us to test whether behavior was related to an offspring’s fish of each day. These videos were subsequently tracked with the body size on its first day of life. We set the first hour of the day as 0 and Biotracker software27, using background subtraction, providing the mean-centered TL as this would allow the among- (and within-) indi- x, y coordinates of each fish in each frame. We then processed the vidual variance components to be estimated at these values (i.e. the data, including scaling and converting the coordinates to mm, and, earliest possible moment from when we could record behavior in the for each frame, computed fish’s swimming speed (cm/s) and distance fish). We estimated the adjusted repeatability of median swimming from the tank walls (cm). We then summarized these variables both speed as the variance attributable to individual identity over the total on an hourly and daily basis to compute fish’s median swimming variance not explained by the fixed effects. We additionally estimated speed, inter-quartile range of swimming speeds, activity (proportion both marginal and conditional R-squared values which estimate the of time spent moving >0.5 cm/s), and median border distance. To variance explained by the fixed effects only and the variance explained quantify fish’s body size over time, we randomly selected five photos by the fixed and random effects combined, respectively. As our Nature Communications | (2022)13:6419 6 Article https://doi.org/10.1038/s41467-022-34113-y individual experimental fish came from different mothers, we first convergence, we initially ran 5 independent chains and inspected explored a number of different variance structures including random posterior trace plots of parameter estimates (Supplementary Note II). intercepts and slopes for both individual ID and maternal ID. This In a preliminary analysis we tested three different prior settings allowed us to test whether maternal identity explained variation in (Supplementary Note II); results did not change with prior settings so individual behavior. However, the most supported model included we chose parameter-expanded priors for all models reported here as random intercepts and slopes for individual ID and not for mother ID, these are generally considered to be more robust. An R Markdown file indicating that our methods to reduce variation among mothers were with all the results presented here is included in Supplementary successful (Table 1). We used median swimming speed as our beha- Note II. vioral variable of interest throughout the main manuscript, as this behavior was tightly correlated with most of our other behavioral Reporting summary variables (Supplementary Fig. 2); though results using the other Further information on research design is available in the Nature behavioral variables yielded the same interpretation (i.e. that sig- Research Reporting Summary linked to this article. nificant individuality in (any) behavior was present on the very first day post-birth; Supplementary Table 2). Data availability Our second research question was to investigate how individual The data generated in this study have been deposited in Dryad Deposi- behavioral variance changed over the course of the entire observation tory [https://doi.org/10.25338/B8XW7G].47 An R markdown file recreat- period (70 days). Again, we first explored several different variance ing the results and figures is included in the Supplemental Information structures to test the importance of maternal identity and/or individual (Supplemental Note II – Code to Reproduce Results). Source data are identity on behavioral variation. We found support for the inclusion of provided with this paper. random slopes at the individual level, but not maternal level (Table 1). This indicates that levels of among- (and within-) individual variation Code availability may differ throughout the observation period. To investigate patterns of All code to reproduce the results are deposited in Dryad (https://doi. change in the variance components, we ran a series of models where we org/10.25338/B8XW7G)47. centered the observation covariate on different days. Individual inter- cepts are estimated when all covariates are set to zero, so this allowed us References to ‘slice’ the data to estimate the among- and within-individual variance 1. Polderman, T. J. C. et al. Meta-analysis of the heritability of human at different time points over the ten weeks. We ran 11 models as we traits based on fifty years of twin studies. Nat. Genet. 47, 702–709 chose to center the data every 7 days (first model was centered on (2015). observation 1; 11th model was centered on observation 70). The pre- 2. Laskowski, K. L., Doran, C., Bierbach, D., Krause, J. & Wolf, M. dicted individual intercepts (best linear unbiased predictors) and esti- Naturally clonal vertebrates are an untapped resource in ecology mated variance components from each model are plotted in Fig. 3. and evolution research. Nat. Ecol. Evolution 3, 161–169 (2019). We also closely investigated any potential influence of body size 3. Vrijenhoek, R. C. Unisexual fish: model systems for studying ecol- and/or growth rate differences on behavioral expression and indivi- ogy and evolution. Annu. Rev. Ecol. Syst. 25, 71–96 (1994). dual behavioral variation in this entire 10-week data set. First, we 4. Gartner, K. A third component causing random variability beside estimated the repeatability of both weekly total length and weekly environment and genotype. A reason for the limited success of a 30 growth rates to determine if individuals consistently differed in these year long effort to standardize laboratory animals? Lab Anim. 24, traits. Then, we ran a series of models with median weekly swimming 71–77 (1990). speed as the response variable and included either weekly total length, 5. Wong, A. H. C., Gottesman, I. I. & Petronis, A. Phenotypic differences weekly growth rate, and/or overall growth rate (estimated over the in genetically identical organisms: the epigenetic perspective. entire 10 weeks), as our fixed effects of interest. Each model also Hum. Mol. Genet 14, R11–R18 (2005). included the random effects of individual intercepts and slopes. 6. Vogt, G. et al. Production of different phenotypes from the same Finally, because body size varies both among individuals (some indi- genotype in the same environment by developmental variation. J. viduals are on average larger than others) and within individuals (as Exp. Biol. 211, 510–523 (2008). they grow), we also performed within-individual centering of total 7. Edenbrow, M. & Croft, D. P. Behavioural types and life history stra- length. In this fifth model, we included each individual’s average total tegies during ontogeny in the mangrove killifish, Kryptolebias length and their weekly deviation from their average length as the two marmoratus. Anim. Behav. 82, 731–741 (2011). fixed effects of interest. Individual identity and slopes were included as 8. Freund, J. et al. Emergence of individuality in genetically identical random effects. For all models, we estimated the variance explained by mice. Science 340, 756–759 (2013). the fixed effects (marginal R2) and the fixed and random effects 9. Schuett, W. et al. Personality variation in a clonal insect: the pea together (conditional R2). These results are reported in Table 2. aphid, Acyrthosiphon pisum. Developmental Psychobiol. 53, For our third and final research question, we tested whether early- 631–640 (2011). life behavior predicted later-life behavior. To test this, we estimated 10. Kain, J. S. et al. Variability in thermal and phototactic preferences in the among-individual correlation (including ‘individual ID’ as our ran- Drosophila may reflect an adaptive bet-hedging strategy. Evolution dom effect) in behavior using multivariate mixed models where the 69, 3171–3185 (2015). daily median swimming speeds in each week were the response vari- 11. Buchanan, S. M., Kain, J. S. & de Bivort, B. L. Neuronal control ables (7 observations per week per individual; 10 weeks total; Fig. 4A). of locomotor handedness in Drosophila. PNAS 112, 6700–6705 Then to investigate how the strength of these correlations may change (2015). over development, we used a linear model to test whether the corre- 12. Honegger, K. & de Bivort, B. Stochasticity, individuality and beha- lation strength was predicted by the interaction between the first week vior. Curr. Biol. 28, R8–R12 (2018). included in the correlation and distance to the next week in the cor- 13. Linneweber, G. A. et al. A neurodevelopmental origin of behavioral relation (1, 2, 3, 4 or 5 weeks away in time; Fig. 4B). individuality in the Drosophila visual system. Science 367, All models were performed using Markov Chain Monte Carlo 1112–1119 (2020). estimation with the MCMCglmm package38 in R v3.6.139. We set our 14. Bierbach, D., Laskowski, K. L. & Wolf, M. Behavioural individuality in models to run 510,000 iterations with a 10,000 burn-in and thinning clonal fish arises despite near-identical rearing conditions. Nat. every 200 iterations. To ensure proper model mixing and Commun. 8, 15361 (2017). Nature Communications | (2022)13:6419 7 Article https://doi.org/10.1038/s41467-022-34113-y 15. Richards, E. J. Inherited epigenetic variation — revisiting soft 40. Lu, Y. et al. Fixation of allelic gene expression landscapes and inheritance. Nat. Rev. Genet 7, 395–401 (2006). expression bias pattern shape the transcriptome of the clonal 16. Kaminsky, Z. A. et al. DNA methylation profiles in monozygotic and Amazon molly. Genome Res. 31, 372–379 (2021). dizygotic twins. Nat. Genet 41, 240–245 (2009). 41. Fraga, M. F. et al. Epigenetic differences arise during the lifetime of 17. Mousseau, T. A. & Fox, C. W. The adaptive significance of maternal monozygotic twins. PNAS 102, 10604–10609 (2005). effects. Trends Ecol. Evolution 13, 403–407 (1998). 42. Katz, Y., Tunstrøm, K., Ioannou, C. C., Huepe, C. & Couzin, I. D. 18. Mulder, E. J. H. et al. Prenatal maternal stress: effects on pregnancy Inferring the structure and dynamics of interactions in schooling and the (unborn) child. Early Hum. Dev. 70, 3–14 (2002). fish. PNAS 108, 18720–18725 (2011). 19. Oates, A. C. What’s all the noise about developmental stochasticity? 43. Jolles, J. W., Boogert, N. J., Sridhar, V. H., Couzin, I. D. & Manica, A. Development 138, 601–607 (2011). Consistent individual differences drive collective behavior and 20. Vogt, G. Stochastic developmental variation, an epigenetic source group functioning of schooling fish. Curr. Biol. 27, 2862–2868.e7 of phenotypic diversity with far-reaching biological consequences. (2017). J. Biosci. 40, 159–204 (2015). 44. Lampert, K. P. & Schartl, M. The origin and evolution of a unisexual 21. Sih, A. et al. Animal personality and state–behaviour feedbacks: a hybrid: Poecilia formosa. Philos. Trans. R. Soc. B: Biol. Sci. 363, review and guide for empiricists. Trends Ecol. Evolution 30, 2901–2909 (2008). 50–60 (2015). 45. Stöck, M., Lampert, K. P., Möller, D., Schlupp, I. & Schartl, M. 22. Stamps, J. A. & Frankenhuis, W. E. Bayesian models of development. Monophyletic origin of multiple clonal lineages in an asexual fish Trends Ecol. Evolution 31, 260–268 (2016). (Poecilia formosa). Mol. Ecol. 19, 5204–5215 (2010). 23. Fawcett, T. W. & Frankenhuis, W. E. Adaptive explanations for sen- 46. Warren, W. C. et al. Clonal polymorphism and high heterozygosity sitive windows in development. Front Zool. 12, S3 (2015). in the celibate genome of the Amazon molly. Nat. Ecol. Evol. 2, 24. Frankenhuis, W. E. & Walasek, N. Modeling the evolution of sensitive 669–679 (2018). periods. Developmental Cogn. Neurosci. 41, 100715 (2020). 47. Laskowski, K. L., Bierbach, D., Jolles, J. W., Doran, C. & Wolf, M. The 25. Jolles, J. W. Broad-scale applications of the Raspberry Pi: a review emergence and development of behavioral individuality in clonal and guide for biologists. Methods Ecol. Evolution 12, 1562–1579 fish. Dryad Depository https://doi.org/10.25338/B8XW7G, (2022). (2021). 26. Jolles, J. pirecorder: Controlled and automated image and video Acknowledgements recording with the raspberry pi. JOSS 5, 2584 (2020). We wish to thank David Lewis and Fiona Rickowski for assistance with 27. Mönck, H. J. et al. BioTracker: An open-source computer vision animal care and husbandry. Funding was provided by the Deutsche framework for visual animal tracking. arXiv:1803.07985 [cs] Forschungsgemeinschaft through overheads on grants BI 1828/2-1 and (2018). BI 1828/3-1 (to D.B.), under Germany’s Excellence Strategy EXC 2002/1, 28. Groothuis, T. G. G., Müller, W., von Engelhardt, N., Carere, C. & “Science of Intelligence” project number 390523135 (to M.W. and D.B.), Eising, C. Maternal hormones as a tool to adjust offspring pheno- from the Alexander von Humboldt Foundation (C.D.), the Zukunftskolleg type in avian species. Neurosci. Biobehav. Rev. 29, 329–352 (2005). Institute of Advanced Studies at the University of Konstanz (J.W.J.), and 29. Monk, C., Spicer, J. & Champagne, F. A. Linking prenatal maternal the National Science Foundation (IOS-2100625 to K.L.L.). adversity to developmental outcomes in infants: the role of epige- netic pathways. Dev. Psychopathol. 24, 1361–1376 (2012). Author contributions 30. Pollux, B. J. A., Pires, M. N., Banet, A. I. & Reznick, D. N. Evolution of K.L.L., D.B. and M.W. conceptualized the study; K.L.L., D.B., J.W.J. and placentas in the fish family Poeciliidae: an empirical study of mac- C.D. performed the methodology and data collection; D.B. and J.W.J. roevolution. Annu. Rev. Ecol., Evolution, Syst. 40, 271–289 (2009). provided or developed software necessary for tracking; K.L.L. per- 31. Trexler, J. C. Resource availability and plasticity in offspring provi- formed the statistical analyses; K.L.L. and M.W. wrote the original draft sioning: embryo nourishment in sailfin mollies. Ecology 78, and all authors reviewed and edited the final version. 1370–1381 (1997). 32. Sopinka, N. M., Capelle, P. M., Semeniuk, C. A. D. & Love, O. P. Funding Glucocorticoids in fish eggs: variation, interactions with the envir- Open Access funding enabled and organized by Projekt DEAL. onment, and the potential to shape offspring fitness. Physiological Biochemical Zool. 90, 15–33 (2017). Competing interests 33. Witvliet, D. et al. Connectomes across development reveal princi- The authors declare no competing interests. ples of brain maturation. Nature 596, 257–261 (2021). 34. Simons, A. M. & Johnston, M. O. Developmental instability as a bet- Additional information hedging strategy. Oikos 80, 401–406 (1997). Supplementary information The online version contains 35. Massicotte, R. & Angers, B. General-purpose penotype or how supplementary material available at epigenetics extend the flexibility of a genotype. Genet. Res. Int. https://doi.org/10.1038/s41467-022-34113-y. 2012, 1–7 (2012). 36. Kelley, J. L. et al. Epigenetic inheritance of DNA methylation chan- Correspondence and requests for materials should be addressed to ges in fish living in hydrogen sulfide–rich springs. PNAS 118, Kate L. Laskowski. e2014929118 (2021). 37. Makowicz, A. M. & Travis, J. Are you more than the sum of your Peer review information Nature Communications thanks the anon- parents’ genes? Phenotypic plasticity in a clonal vertebrate and F1 ymous reviewers for their contribution to the peer review of this work. hybrids of its parental species. Evolution 74, 1124–1141 (2020). 38. Stamps, J. A. Individual differences in behavioural plasticities. Biol. Reprints and permissions information is available at Rev. 91, 534–567 (2016). http://www.nature.com/reprints 39. Kok, E. M. A. et al. Within-individual canalization contributes to age- related increases in trait repeatability: a longitudinal experiment in Publisher’s note Springer Nature remains neutral with regard to jur- red knots. Am. Naturalist 194, 455–469 (2019). isdictional claims in published maps and institutional affiliations. Nature Communications | (2022)13:6419 8 Article https://doi.org/10.1038/s41467-022-34113-y Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons license, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons license and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this license, visit http://creativecommons.org/ licenses/by/4.0/. © The Author(s) 2022 Nature Communications | (2022)13:6419 9