Abstract
The genome versus experience, or “Nature versus Nurture”, debate has dominated our understanding of individual behavioral variation. A third factor, namely variation in complex behavior potentially due to non-heritable “developmental noise” in brain development, has been largely ignored. Using the Drosophila vinegar fly we demonstrate a causal link between variation in brain wiring due to developmental noise, and behavioral individuality. A population of visual system neurons called DCNs shows non-heritable, inter-individual variation in right/left wiring asymmetry, and control object orientation in freely walking flies. We show that DCN wiring asymmetry predicts an individual’s object responses: the greater the asymmetry, the better the individual orients. Silencing DCNs abolishes correlations between anatomy and behavior, while inducing visual asymmetry via monocular deprivation “rescues” object orientation in DCN-symmetric individuals.
One Sentence Summary Non-heritable individual variation in neural circuit development underlies individual variability in behavior.
Main Text
Individual variability in external and internal organ morphology is highly abundant in all organisms, including among genetically identical individuals, such as human identical twins and species that reproduce by parthenogenesis (1-3). In this regard, the brain is no exception. The simplest examples of individual brain variation include differences of size and weight of human brains (4) but also variance of more complex traits, like neuroanatomical parcellations, have been described (5, 6). The same variability is also found at a deeper organizational level in the number of neurons (7). In invertebrate model organisms, it was shown that individual neurons show varying morphology and wiring (8), variable and plastic synaptic morphology and molecular composition (9-12).
The same level of individual variability can also be found in behavior, the main output of nervous system function (13). Complex innate behaviors, like selective attention to stimuli, show significant individual variation even amongst genetically similar or even identical individuals (14-16). Stability of these individual differences over time allows to define these behavioral idiosyncrasies as animal individuality (17, 18). Analysis of neural circuits in various genetic model organisms has led to the widely-held view that variability in innate behavior is largely due to neuromodulation of otherwise anatomically hard-wired neuronal circuits (19, 20). This includes stable behavioral traits in C. elegans foraging that depend on serotonin modulation (19) and serotonergic and dopaminergic regulation of the zebrafish larval startle response (20). More recently it was shown that the developmental plasticity of higher order neural circuits, partly driven by stochastic mechanisms, makes them intrinsically variable, resulting in a range of possible circuit diagrams even amongst genetically identical individuals. This variability in developmental wiring is regulated by cell-cell signaling events and results in differential gene expression profiles amongst otherwise indistinguishable neurons (21, 22). However, despite much interest, debate and speculation, the contribution of non-heritable developmental variability in neural circuit wiring to individual behavioral variation remains almost completely unexplored (23, 24).
To test whether probabilistic wiring of neural circuits affects behavioral variation, we used a Drosophila higher order visual circuit, called the Dorsal Cluster Neurons (DCN) (25). DCNs exhibit up to 30% wiring variability of their axonal projection not only between individuals, but also between the left and right hemispheres of the same brain (26). DCNs in each hemisphere derive from a single neural stem cell (25, 27) and their axons innervate two alternative target areas in the fly visual system called the lobula and the medulla, respectively (25). The medulla processes visual motion information (28), while the lobula is implicated in the integration of visual and motor information. Initial characterization of DCN anatomy showed that about 25% of DCN axons have their terminal presynaptic arbors in the contralateral medulla (M-DCNs), while about 75% have their terminal presynaptic arbors in the contralateral lobula (L-DCNs). L-DCN axons also project onto M-DCN dendrites, while M-DCN axons project onto medulla neurons. Because the decision that each DCN makes between being M-DCN or L-DCN is determined by an intrinsically stochastic lateral inhibition mechanism (21).
In order to test a quantitative relationship between a behavior and non-heritable brain wiring variability, a visual behavioral assay was required that allows testing individual flies repeatedly over a significant stretch of time. For this purpose, the robust and easily accessible Buridan’s paradigm was chosen (29). In this assay a single fly walks between two equally attractive visual targets in the form of high contrast black stripes placed at 180 degrees from each other in an otherwise uniformly illuminated arena (30). The two stripes are unreachable by the fly because the walking arena is separated from the stripes by a trench of water, inducing the fly to walk back and forth between the two stripes for the duration of the assay. A major advantage of open world assays like Buridan’s paradigm is the ease of repeated measurement of a large number of behavioral parameters on freely walking individuals. This assay is named after a medieval philosophical paradox meant to highlight the importance of intrinsic bias when external conditions are completely symmetrical. Interestingly, the problem of how to resolve identical options has been proposed as one of the advantages of a non-deterministic and noisy brain (31).
Using this behavioral paradigm, we find that flies show highly idiosyncratic responses that are very stable over a long period of time. In particular, the width of the path that a fly walks between the two stripes, a parameter we call “absolute stripe deviation”, is a unique and stable feature of a given fly that shows a normal distribution of variability within the population. We show that behavioral individuality of stripe deviation is non-heritable and is not reduced through inbreeding. Using unbiased anatomy-to-behavior correlation mapping, we find that the degree in left/right DCN wiring asymmetry is a robust predictor of behavioral performance of an individual fly and its variance across the population. The more asymmetric the DCN wiring pattern, the narrower the path a fly walks between the two stripes. DCN activity is necessary for this correlation, while artificially inducing asymmetry in the visual system is sufficient to change the response of an individual. This establishes a causal link between variability in the development of the brain and the emergence of individuality of animal behavior.
Results
While analyzing object orientation responses in wildtype Canton S (CS) flies using Buridan’s paradigm (Fig. 1 A and Movies S1-3) we noted significant degree of inter-individual variability in their trajectories. Given that males and females display dimorphic behavioral traits (32) and are genetically different from each other, we first tested whether this variation was principally due to sex differences. However, at the population level males and females displayed similar responses towards the high contrast stripes, as shown by the occupancy heatmaps and individuals with equally variable responses were found in both sexes (Fig. 1 B-C). Next, we used a simple and robust parameter called “absolute stripe deviation” which shows how much a fly deviates from an idealized narrowest possible path between the stripes. Thus, a low stripe deviation score indicates a fly that walks a narrow and straight path between the two stripes, while a high stripe deviation score indicates a fly that walks a broad and meandering path between the two stripes. We found that males and females display a similar degree of inter-individual variation in stripe fixation (Fig. 1 D). Therefore, we combined the responses of the CS population in one histogram displaying the variability for absolute stripe deviation (Fig. 1 E) and continued our studies with combined male and female populations.
Individual variability of object orientation responses is independent of genetic diversity
Next, we asked if the behavioral variability correlates with genetic diversity. If genetic diversity results in behavioral variability, then strains with low genetic diversity should show reduced object orientation variability. To test this idea, we compared strains with high genetic diversity to strains with low diversity. First, we screened a subset (N=10) of the Drosophila genomic reference panel (DGRP) for strains that showed either very strong or very weak object orientation responses at the population level. DGRP lines are fully sequenced and they have been inbred for 20 generations, which should make all individuals of one strain as genetically homogenous as it is compatible with viability (33). We identified two strains with opposing behavioral phenotypes at the population level. DGRP-639 showed near wildtype absolute stripe deviation (Fig. 1 F-G), while DGRP-859 showed strongly increased stripe deviation (Fig. 1 G-I). Importantly, these population level differences were not confined to the stripe deviation index as statistical analysis of eight representative behavioral parameters showed that the outbred CS control strain differed significantly in the mean from the inbred lines (Fig. S1 A), showing that these were truly behaviorally different populations. Next, we asked if the inbred DGRP lines showed a reduction in behavioral variability along with their reduced genetic diversity. We find that the degree of individual variation of object orientation responses, as indicated by the distribution of the stripe deviation index, was not reduced (Fig. 1 G, I). In fact, if anything, the DGRP-639, which as a population has reduced stripe deviation, showed increased behavioral variability (Fig. 1 G, Fig. S1 A). Therefore, reduction of genetic diversity did not reduce phenotypic behavioral variation.
Individual object orientation responses are non-heritable
We next wanted to know whether a specific genotype in an individual is associated with reproducible behavior of that individual. If that would be the case, we should be able to breed specific alleles by choosing parental animals with certain behavioral traits. From a CS population of 47 male and 37 virgin female flies, we selected and mated the three pairs with the lowest and highest stripe deviation scores, respectively (Fig. 2 A). Object orientation responses in the offspring of these pairs were measured in the Buridan arena (Fig. 2 B, C). We find no differences between the two sets of offspring in stripe deviation scores as well as six other parameters tested (Fig. 2S A). The same was true for the offspring of a single pair with low and high stripe deviation (Fig. 2S B-C). More importantly, the individual variation of stripe deviation scores of the two sets of offspring matched each other, as shown in the histograms, and matched the original variability of the parental CS population. Therefore, an individual’s behavioral profile is non-heritable and offspring of similar parents recreate the individual variability profile of the starting population.
Individual object orientation responses are stable over time
A specific behavioral profile may not be heritable for two reasons. First, because it is driven by current state modulations and is thus not a stable trait of an individual. Second, because it is a stable trait driven by intrinsic but non-heritable mechanisms. To distinguish these possibilities, we asked if the object orientation responses of an individual are stable over extended periods of time. We tested the same individual CS flies once every other day for three days, in the Buridan arena. We find that an individual’s behavior is virtually identical over the three trials for animals with low (Fig. 3 A) as well as high (Fig. 3 B) stripe deviation scores. Statistical analysis of absolute stripe deviation showed that the individual responses of CS flies on different days were strongly and highly significantly correlated (Pearson correlation coefficients ranging from 0.74-0.77, Fig. 3 C). Furthermore, the same was true for path details like left or right shifted angles (Fig. 3 A and Fig. S3 A for statistical analysis, including angle parameters). Similar results were obtained for many other behavioral parameters including: distance, full walks, meander, absolute horizon deviation, absolute angle deviation, angle deviation and center deviation (Fig. S3 C). The stability of behavioral responses over several days argued strongly against current state modulations and in favor of individual properties. Next, we asked how temporally stable object orientation responses are by testing four-week-old flies repeatedly in our arena. Aged animals, exactly like their young counterparts showed highly stable individual responses (Fig. S3 D). Finally, we asked whether reduced genetic diversity might impact behavioral stability. We performed repeated testing of DGRP-639 and DGRP-859 individual flies and found that both inbred strains showed temporally stable individual stripe deviation responses (Fig. 3 D-E; Fig. S3 A-B), as well as many of the other behavioural parameters for both DGRP lines (Fig. S3 C).
Altogether, our behavioral analyses suggest that individual variability in object orientation responses is an intrinsic, non-heritable, temporally stable trait that is independent of sex and genetic background and that is not eliminated by reduced genetic diversity. What in the brain might underlie such individuality in visual behavior?
The DCNs are a highly variable set of commissural visual interneurons
In a classic object orientation paper Bülthoff (34) suggested, based on earlier work by Zimmermann and Götz (35, 36), that object position processing in Drosophila and other insects (37) might in part rely on what was referred to as qualitative asymmetry between front-to-back and back-to-front motion. This simply refers to the fact that as a fly turns towards an object, this object moves from front-to-back on one retina, and from back to front on the other retina. Asymmetry of this percept would allow the fly to better center the object in the frontal part of its visual field. However, direct evidence for this notion remains lacking, and it is unclear where the source of such visual system asymmetry would be, especially that the sizes of the left and right eyes of the same fly, though not identical, are nonetheless very highly correlated (38). In 1986, Heisenberg and colleagues suggested that binocular interactions, perhaps through commissural interneurons further upstream in the visual system may be required for object orientation in the frontal visual field (39). Interestingly, the DCNs described above precisely match this predicted circuit. DCNs, whose function has thus far remained unknown show a highly variable wiring diagram between individuals and between the left and right sides of the same individual, due to intrinsically stochastic developmental wiring mechanisms (21, 22). DCN soma are located on the dorsal side of the fly visual center called the optic lobe. Each DCN projects a single ventral neurite that splits into an ipsilateral dendrite and a contralateral axon. On the contralateral side the axon terminates in either the lobula, thus defining the neuron as a L-DCN, or the medulla for the M-DCNs (Fig. 4 A). The DCNs show variability on all three levels: the number of cell bodies, the number of axons innervating the lobula and the number of axons innervating the medulla (21). All three levels of variability are observed between individuals as well as between the left and right hemispheres of the same individual.
We first extended the quantitative analysis of DCN wiring variability using an automatic neuronal reconstruction method on a much larger number of individuals than previously analyzed (N = 103). Our data revealed that we had in fact previously underestimated DCN variability (21). Specifically, we found that the number of DCNs varies from 22-68 cells, with a range of 11-55 L-DCNs and 6-23 M-DCNs (Fig. 4 B; Fig. 4S A). In addition to the obvious asymmetries in the number of neurons the majority of which are L-DCNs, we observe a distribution of variation in medulla targeting asymmetry by M-DCNs (Fig. 4 B, histogram distributions; Fig. 4S A). The distribution of all DCN asymmetries showed a peak of low asymmetries, while extreme asymmetries were present but rare. Importantly, this extended analysis confirmed the previous observation that the number of axons in the medulla is not correlated to the total number of DCNs that give rise to them (Fig 4S A). In other words, the number of DCNs in the right hemisphere does not predict the number of DCN axons in the left medulla and vice versa. Finally, 3D reconstruction shows that M-DCN axons terminate specifically in the posterior medulla (Movies S4-6), where visual columns form the anterior (i.e. frontal) visual field are located, and the DCN wiring pattern in the medulla does not change in the adult (Movie S7). In summary, the DCNs innervate the frontal visual columns in the medulla in a fashion that shows a variable degree of asymmetry between individuals which is stable over time in any given individual.
Individual variability in DCN wiring asymmetry drives individual variability in object orientation behavior
The data above suggest that DCNs represent an ideal candidate for an intrinsically asymmetric population of contralateral higher order interneurons that may mediate binocular interactions relevant to the anterior visual field (39). We therefore hypothesized that the DCNs represent a circuit that explains how binocular asymmetries in the visual system regulate object orientation responses. To test this hypothesis, we first asked whether the DCNs were required for object orientation in the Buridan assay (29). Silencing either all DCNs, or only M-DCNs, by expressing either the potassium ion channel Kir2.1 (40) or the tetanus toxin light chain (TNT) (41) resulted in a strong disruption of object orientation behavior. Specifically, DCN silenced flies showed a strong tendency to approach the two stripes from the extreme edges of the arena, as opposed to through the middle, as revealed by the change in the occupancy heat maps and a significant increase in the stripe deviation index (Fig. 4 C, D and Fig. 4S B, C).
Next, we queried the relationship between individual variability in object orientation behavior and individual variability in DCN cell numbers and wiring diagram. To this end, we expressed nuclear and membrane markers in DCNs and quantitatively measured individual object orientation behavior of these flies (N = 103), followed by high-resolution confocal imaging and semi-automatic neuronal reconstructions of each individual, and finally, an unbiased correlation analysis between 36 key behavioral parameters and 37 prominent anatomical features of DCNs in each individual (Fig. 5 A, Fig. 5S A-C). We found that left/right asymmetry in medulla innervation by DCNs strongly and specifically correlated with an individual’s stripe deviation index (Fig. 5 B, r = −0.67) and other inter-dependent parameters, such as absolute angle deviation and center deviation, but not unrelated parameters such as total distance or the number of full walks between stripes (Fig. S5 A). Individuals with high M-DCN asymmetry tend to have a low stripe deviation index (i.e. walk a narrow path between stripes), while individuals with symmetric M-DCN have a high stripe deviation index (Fig. 5 C).
If DCN wiring asymmetry is a functional driver of individual object orientation behavior, then silencing DCN activity should abolish the correlation between anatomy and behavior in individuals. To test this, we expressed TNT in DCNs labeled with GFP and repeated the individual behavioral-anatomy correlation analyses as described above. We observed that in animals that express TNT in DCN neurons, no significant correlation exists between M-DCN asymmetry and absolute Stripe deviation (Fig. 5 D, r = −0.002), demonstrating a requirement for DCN activity for a link between wiring asymmetry and behavior in individuals.
Visual asymmetry determines object orientation in individuals
Our data show that under normal conditions intrinsic, non-heritable, developmental variation in DCN wiring asymmetry is necessary for creating variability in object orientation behavior across individuals. This supports Götz’s original hypothesis that object orientation depends on asymmetry in the processing of visual input. To directly test if generating asymmetry anywhere in visual information is sufficient to change an individual’s behavior, we tested 79 wildtype CS flies in the Buridan arena. Next, we selected the 20 flies with stripe fixation indices above 40 – thus those that tend to have lower asymmetry in DCN wiring – and then performed monocular deprivation and tested them again. We find that monocular deprivation resulted in a significant reduction of the stripe deviation index in these flies (Fig. 5 E). Remarkably, this was also true for the entire population (Fig. 5S D). Therefore, variability in the asymmetry of visual processing causally contributes to the behavioral variability in object orientation behavior.
Discussion
The question for the origins of individual behavioral variation a central open question in the neurosciences and psychology. Obviously both heritable and environmental factors shape any given individual, yet mechanisms that explain behavioral individual variation in most cases remain elusive. The discovery of stable individual traits in non-human vertebrates (20) and invertebrate species facilitated research in numerous species on behavioral variation (14, 17). In recent years, due to experimental advantages, invertebrate model systems like the fruit fly Drosophila melanogaster and the nematode Caenorhabditis elegans have become increasingly popular to answer important questions about the origins of individual behavioral variability (15, 16, 19, 42). These seminal papers clearly demonstrate that individuality can be found in genetic invertebrate model organisms and they offer both genetic (15, 16, 42) and neuromodulatory (15, 19, 42) explanations for the observed idiosyncrasies.
The work presented here demonstrates an entirely new mechanism to explain aspects of individual behavioral variation through stochastic variability in nervous system development. We had previously reported that significant wiring variability in a population of higher order visual system neurons called the DCNs arises through fundamentally stochastic cell-cell interaction mechanisms such as lateral inhibition and cell surface receptor recycling during filopodial growth and retraction (21, 22). This study further shows that these mechanisms also give rise to significant intra-individual variation in the shape of a broad distribution of left/right asymmetry in the innervation of the visual areas by contralateral DCN axons. Such extensive individual variability in neural circuit development raises the question of whether individual variation in circuit morphology has implications for individual variation in visually guided innate behavior. We addressed this question using the Buridan paradigm (29), a robust object orientation assay that allows repeated testing of the same individual. Our data show that flies display temporally stable, non-heritable, individual behavioral differences in object responses. When considered in the context of previous reports showing individuality in similar to locomotor handedness (16, 42) and phototactic (15) behavior in flies, our work shows that flies have innate individuality traits across a complex range of visually guided behaviors. The amenability of the relatively complex Drosophila brain to multiscale analysis, from the molecular to the behavioral, at single animal resolution makes it an ideal model for understanding the emergence of individuality at each of these scales.
Our work demonstrates that variability in neuronal wiring arising during development strongly correlates with behavioral individuality. Specifically, using an unbiased approach in which object orientation behavioral parameters were correlated with DCN wiring morphology, we show that the variation in left/right asymmetry across individuals in a population explains the variation in object orientation responses. In this regard, two observations merit further comment. First, unbiased analysis revealed a highly specific correlation between the shape of the path a walking fly takes towards an object and the asymmetry of DCN innervation of a particular visual area called the posterior medulla, through which information form the frontal visual field flows. Individuals with high asymmetry show much stronger object orientation responses than animals with high symmetry. Second, silencing DCN activity did not abolish object responses per se, but rather completely abolished the correlation between wiring variation and behavioral variation, meaning that the anatomical asymmetry is functionally relevant. Given the very little asymmetry observed in eye size in Drosophila lab strains, our work suggests that in the fly visual system asymmetries relevant to object orientation arise in the brain and not in the periphery. To query whether peripheral asymmetry has the same effect, we blocked a single eye of animals with weak orientation responses and found that this strongly enhanced their object orientation, meaning that asymmetry in visual information is in and of itself sufficient for improving object orientation. This is important because classic work in Drosophila visual behavioral neuroscience lead to the proposal that asymmetry in visual information processing influences object responses. Where such functional asymmetry lay and how it might arise has, until now, remained unclear. Independently, the study of object responses in motion blind mutants led Heisenberg and colleagues to propose a hypothetical contralateral circuit dedicated to object responses in the frontal visual field (30, 34-36, 39). We therefore propose that the DCNs are the neurons that explain both of these observations: a contralateral asymmetric visual circuit that regulates object orientation in the frontal visual field. Future work will reveal the exact physiological consequences of morphological asymmetry, such as whether wring asymmetry induces timing differences as in auditory navigation (43) or whether the absolute differences are simply summed up.
Here we show that inherently stochastic cellular mechanisms lead to probabilistic wiring diagrams across a population which in turn underlies significant individual variability in behavior in an invertebrate model. This is consistent with the fact that variations in fluctuating morphological asymmetry arise during development even in genetically identical individuals of parthenogenic species raised in the same microenvironment (44). Studies in humans clearly show similar correlations between variations in brain morphology and individual variations in behavior and personality (27, 45). More recently, some studies have begun to focus on those anatomical features that can be definitively traced back to developmental events. Work focused on differences in reading capabilities shows that continuous versus interrupted morphology of the human sulcus in the visual word form area, which arises during fetal development, predicts reading skills in adults (46). This, combined with differences in brain anatomy between identical human twins, strongly suggests that stochastic developmental variation in neural network formation is, in addition to genetic differences and environmental factors, a determining factor of individual behavioral variation and personality. The amenability of invertebrate models to highly detailed multiscale analysis, from the molecular to the behavioral, at single animal resolution of the causal links between genes, development and the environment in generating personality traits holds great promise for further breakthrough discoveries in the field, that the history of science shows will likely translate to human biology.
Funding
This work was supported by ICM, the program “Investissements d’avenir” ANR-10-IAIHU-06, The Einstein-BIH program, the Paul G. Allen Frontiers Group, VIB, the WiBrain Interuniversity Attraction Pole network (Belspo) and Fonds Wetenschappelijke Onderzoeks (FWO) grants G.0543.08, G.0680.10, G.0681.10 and G.0503.12 (B.A.H.), as well as EMBO Long Term Postdoctoral Fellowships (G.A.L. and R.E.), a VIB Omics postdoctoral fellowship (R.E.), and the Marie Sklodowska Curie actions in FP7 and Horizon 2020 (G.A.L). B.A.H is an Allen Distinguished Investigator and an Einstein Visiting Fellow
Author contributions
G.A.L. and B.A.H conceived the study, designed the experiments and wrote the manuscript. G.A.L. conducted all behavioral experiments, immunohistochemistry and all data analysis. M.A., S.D. and L.H. helped with the neuronal reconstructions. G.L. and R.K.E provided technical expertise and R.K.E wrote the python analysis software. L.M.F and A.D.S shared data before publication and provided technical expertise. M.W. and P.R.H. provided expertise and equipment and helped writing the manuscript
Competing interests
Authors declare no competing interests.; and Data and materials availability All data is available in the main text or the supplementary materials.
Acknowledgments
We thank the Bloomington stock center (NIH P40OD018537) and Gilad Barnea for providing flies used in this study. We thank Bruno van Swinderen and all members of the Hassan, Hiesinger and Wernet lab for support and providing valuable comments in the course of this project. We further would like to thank the VIB Bio Imaging Core for support
References and Notes
Supplementary citations
- 1.
- 2.
- 3.
- 4.
- 5.
- 6.
- 7.
- 8.
- 9.
- 10.
- 11.