Search

We present a method for automated nucleus identification and tracking in time-lapse microscopy recordings of entire developing embryos. Our method combines deep learning and global optimization to enable complete lineage reconstruction from sparse point annotations, and uses parallelization to process multi-terabyte light-sheet recordings, which we demonstrate on three common model organisms: mouse, zebrafish,Drosophila. On the most difficult dataset (mouse), our method correctly reconstructs 75.8% of cell lineages spanning 1 hour, compared to 31.8% for the previous state of the art, thus enabling biologists to determine where and when cell fate decisions are made in developing embryos, tissues, and organs.

Summary The thirteen nuclear cleavages that give rise to the Drosophila blastoderm are some of the fastest known cell cycles [ 1 ]. Surprisingly, the fertilized egg is provided with at most one-third of the dNTPs needed to complete the thirteen rounds of DNA replication [ 2 ]. The rest must be synthesized by the embryo, concurrent with cleavage divisions. What is the reason for the limited supply of DNA building blocks? We propose that frugal control of dNTP synthesis contributes to the well-characterized deceleration of the cleavage cycles and is needed for robust accumulation of zygotic gene products. In support of this model, we demonstrate that when the levels of dNTPs are abnormally high, nuclear cleavages fail to sufficiently decelerate, the levels of zygotic transcription are dramatically reduced, and the embryo catastrophically fails early in gastrulation. Our work reveals a direct connection between metabolism, the cell cycle, and zygotic transcription.

In vivo imaging applications typically require carefully balancing conflicting parameters. Often it is necessary to achieve high imaging speed, low photo-bleaching, and photo-toxicity, good three-dimensional resolution, high signal-to-noise ratio, and excellent physical coverage at the same time. Light-sheet microscopy provides good performance in all of these categories, and is thus emerging as a particularly powerful live imaging method for the life sciences. We see an outstanding potential for applying light-sheet microscopy to the study of development and function of the early nervous system in vertebrates and higher invertebrates. Here, we review state-of-the-art approaches to live imaging of early development, and show how the unique capabilities of light-sheet microscopy can further advance our understanding of the development and function of the nervous system. We discuss key considerations in the design of light-sheet microscopy experiments, including sample preparation and fluorescent marker strategies, and provide an outlook for future directions in the field.

Optimal image quality in light-sheet microscopy requires a perfect overlap between the illuminating light sheet and the focal plane of the detection objective. However, mismatches between the light-sheet and detection planes are common owing to the spatiotemporally varying optical properties of living specimens. Here we present the AutoPilot framework, an automated method for spatiotemporally adaptive imaging that integrates (i) a multi-view light-sheet microscope capable of digitally translating and rotating light-sheet and detection planes in three dimensions and (ii) a computational method that continuously optimizes spatial resolution across the specimen volume in real time. We demonstrate long-term adaptive imaging of entire developing zebrafish (Danio rerio) and Drosophila melanogaster embryos and perform adaptive whole-brain functional imaging in larval zebrafish. Our method improves spatial resolution and signal strength two to five-fold, recovers cellular and sub-cellular structures in many regions that are not resolved by non-adaptive imaging, adapts to spatiotemporal dynamics of genetically encoded fluorescent markers and robustly optimizes imaging performance during large-scale morphogenetic changes in living organisms.

Understanding how the brain works in tight concert with the rest of the central nervous system (CNS) hinges upon knowledge of coordinated activity patterns across the whole CNS. We present a method for measuring activity in an entire, non-transparent CNS with high spatiotemporal resolution. We combine a light-sheet microscope capable of simultaneous multi-view imaging at volumetric speeds 25-fold faster than the state-of-the-art, a whole-CNS imaging assay for the isolated Drosophila larval CNS and a computational framework for analysing multi-view, whole-CNS calcium imaging data. We image both brain and ventral nerve cord, covering the entire CNS at 2 or 5 Hz with two- or one-photon excitation, respectively. By mapping network activity during fictive behaviours and quantitatively comparing high-resolution whole-CNS activity maps across individuals, we predict functional connections between CNS regions and reveal neurons in the brain that identify type and temporal state of motor programs executed in the ventral nerve cord., To understand how neuronal networks function, it is important to measure neuronal network activity at the systems level. Here Lemon et al. develop a framework that combines a high-speed multi-view light-sheet microscope, a whole-CNS imaging assay and computational tools to demonstrate simultaneous functional imaging across the entire isolated Drosophila larval CNS.

Imaging fast cellular dynamics across large specimens requires high resolution in all dimensions, high imaging speeds, good physical coverage and low photo-damage. To meet these requirements, we developed isotropic multiview (IsoView) light-sheet microscopy, which rapidly images large specimens via simultaneous light-sheet illumination and fluorescence detection along four orthogonal directions. Combining these four views by means of high-throughput multiview deconvolution yields images with high resolution in all three dimensions. We demonstrate whole-animal functional imaging of Drosophila larvae at a spatial resolution of 1.1-2.5 μm and temporal resolution of 2 Hz for several hours. We also present spatially isotropic whole-brain functional imaging in Danio rerio larvae and spatially isotropic multicolor imaging of fast cellular dynamics across gastrulating Drosophila embryos. Compared with conventional light-sheet microscopy, IsoView microscopy improves spatial resolution at least sevenfold and decreases resolution anisotropy at least threefold. Compared with existing high-resolution light-sheet techniques, IsoView microscopy effectively doubles the penetration depth and provides subsecond temporal resolution for specimens 400-fold larger than could previously be imaged.

Coupling of autonomous cellular oscillators is an essential aspect of circadian clock function but little is known about its circuit requirements. Functional ablation of the pigment-dispersing factor-expressing lateral ventral subset (LNV) ofDrosophilaclock neurons abolishes circadian rhythms of locomotor activity. The hypothesis that LNVs synchronize oscillations in downstream clock neurons was tested by rendering the LNVs hyperexcitable via transgenic expression of a low activation threshold voltage-gated sodium channel. When the LNVs are made hyperexcitable, free-running behavioral rhythms decompose into multiple independent superimposed oscillations and the clock protein oscillations in the dorsal neuron 1 and 2 subgroups of clock neurons are phase-shifted. Thus, regulated electrical activity of the LNVs synchronize multiple oscillators in the fly circadian pacemaker circuit.

The fruit fly is an excellent model system for investigating the sequence of epithelial tissue invaginations constituting the process of gastrulation. By combining recent advancements in light sheet fluorescence microscopy (LSFM) and image processing, the three-dimensional fly embryo morphology and relevant gene expression patterns can be accurately recorded throughout the entire process of embryogenesis. LSFM provides exceptionally high imaging speed, high signal-to-noise ratio, low level of photoinduced damage, and good optical penetration depth. This powerful combination of capabilities makes LSFM particularly suitable for live imaging of the fly embryo.The resulting high-information-content image data are subsequently processed to obtain the outlines of cells and cell nuclei, as well as the geometry of the whole embryo tissue by image segmentation. Furthermore, morphodynamics information is extracted by computationally tracking objects in the image. Towards that goal we describe the successful implementation of a fast fitting strategy of Gaussian mixture models.The data obtained by image processing is well-suited for hypothesis testing of the detailed biomechanics of the gastrulating embryo. Typically this involves constructing computational mechanics models that consist of an objective function providing an estimate of strain energy for a given morphological configuration of the tissue, and a numerical minimization mechanism of this energy, achieved by varying morphological parameters.In this chapter, we provide an overview of in vivo imaging of fruit fly embryos using LSFM, computational tools suitable for processing the resulting images, and examples of computational biomechanical simulations of fly embryo gastrulation.

Axonal branching allows a neuron to connect to several targets, increasing neuronal circuit complexity. While axonal branching is well described, the mechanisms that control it remain largely unknown. We find that in the Drosophila CNS branches develop through a process of excessive growth followed by pruning. In vivo high-resolution live imaging of developing brains as well as loss and gain of function experiments show that activation of Epidermal Growth Factor Receptor (EGFR) is necessary for branch dynamics and the final branching pattern. Live imaging also reveals that intrinsic asymmetry in EGFR localization regulates the balance between dynamic and static filopodia. Elimination of signaling asymmetry by either loss or gain of EGFR function results in reduced dynamics leading to excessive branch formation. In summary, we propose that the dynamic process of axon branch development is mediated by differential local distribution of signaling receptors. DOI: http://dx.doi.org/10.7554/eLife.01699.001, eLife digest In the human brain, 100 billion neurons form 100 trillion connections. Each neuron consists of a cell body with numerous small branch-like projections known as dendrites (from the Greek word for ‘tree’), plus a long cable-like structure called the axon. Neurons receive electrical inputs from neighboring cells via their dendrites, and then relay these signals onto other cells in their network via their axons. The development of the brain relies on new neurons integrating successfully into existing networks. Axon branching helps with this by enabling a single neuron to establish connections with several cells, but it is unclear how individual neurons decide when and where to form branches. Now, Zschätzsch et al. have revealed the mechanism behind this process in the fruit fly, Drosophila. Mutant flies that lack a protein called EGFR produce abnormal numbers of axon branches, suggesting that this molecule regulates branch formation. Indeed in fruit flies, just as in mammals, the developing brain initially produces excessive numbers of branches, which are subsequently pruned to leave only those that have formed appropriate connections. In Drosophila, an uneven distribution of EGFR between branches belonging to the same axon acts as a signal to regulate this pruning process. To examine this mechanism in more detail, high-resolution four-dimensional imaging was used to study brains that had been removed from Drosophila pupae and kept alive in special culture chambers. Axon branching and loss could now be followed in real time, and were found to occur more slowly in brains that lacked EGFR. The receptor controlled the branching of axons by influencing the distribution of another protein called actin, which is a key component of the internal skeleton that gives cells their structure. In addition to providing new insights into a fundamental aspect of brain development, the work of Zschätzsch et al. also highlights the importance of stochastic events in shaping the network of connections within the developing brain. These findings may well be relevant to ongoing efforts to map the human brain ‘connectome’. DOI: http://dx.doi.org/10.7554/eLife.01699.002

Insects attempting to identify and classify general, food-related odors in a natural environment are faced with the difficult problem of turning a spatially and temporally noisy, multicomponent olfactory stimulus into a reliable neural code that is independent of spurious changes in concentration or the temporal structure of the odor plume. The insect nervous system accomplishes this by representing olfactory information in different forms in different parts of the olfactory sensory pathway. At the periphery, olfactory information is encoded as a simple rate code expressed in a large population of broadly tuned receptor neurons. Some information processing may take place as a result of interactions between olfactory receptor neurons located in the same sensillum of the insect antenna. The olfactory information is conducted to intrinsic interneurons within the antennal lobe of the brain. Olfactory information is subsequently transmitted from intrinsic interneurons to antennal lobe projection neurons that carry a condensed form of olfactory information to the mushroom bodies and other higher brain centers. Projection neurons appear to encode olfactory information in a complex temporal code comprising spike trains with variable interspike intervals. The precise timing of these action potentials may provide higher brain centers with information about the identity, concentration, and duration of the stimulus. The olfactory glomeruli appear to translate the simple rate code input into a complex temporal code output that may involve transient synchronization of the action potentials and oscillations of field potentials in the brain.

The comprehensive reconstruction of cell lineages in complex multicellular organisms is a central goal of developmental biology. We present an open-source computational framework for the segmentation and tracking of cell nuclei with high accuracy and speed. We demonstrate its (i) generality by reconstructing cell lineages in four-dimensional, terabyte-sized image data sets of fruit fly, zebrafish and mouse embryos acquired with three types of fluorescence microscopes, (ii) scalability by analyzing advanced stages of development with up to 20,000 cells per time point at 26,000 cells min(-1) on a single computer workstation and (iii) ease of use by adjusting only two parameters across all data sets and providing visualization and editing tools for efficient data curation. Our approach achieves on average 97.0% linkage accuracy across all species and imaging modalities. Using our system, we performed the first cell lineage reconstruction of early Drosophila melanogaster nervous system development, revealing neuroblast dynamics throughout an entire embryo.

Behavioral and physiological evidence indicates that insect pheromone sensory neurons are able to resolve pulses of pheromone concentration as they occur downwind from a point source, but the abilities of insect sensory neurons that are sensitive to general odors to respond to pulsatile stimuli are unknown. The temporal response characteristics of olfactory sensory neurons of female American cockroaches Periplaneta americana in response to general odors were measured using a series of short odor pulses (20–400 ms). Odor pulses were delivered to olfactory sensilla in a moving airstream controlled by electromagnetic valves. The responses of sensory neurons were recorded using a tungsten electrode placed at the base of the sensillum. The temporal responses of sensory neurons followed the temporal changes in stimulus concentration, which were estimated by replacing the odorant with oil smoke and measuring the concentration of smoke passing through a light beam. Spike frequency varied with odorant concentration with surprisingly fine temporal resolution. Cockroach olfactory sensory neurons were able reliably to follow 25 ms pulses of the pure odorant 1-hexanol and 50 ms pulses of the complex odor blend coconut oil. Lower concentrations of odorants elicited responses with lower peak spike frequencies that still retained the temporal resolution of the stimulus pulses. Thus, responses of olfactory sensory neurons can reflect the fine structures of non-uniform distributions of general odorants in a turbulent odor plume as well as the average odorant concentration.

Recent studies that have shown that variation in avian foraging behavior can produce changes in reproductive success and survival have raised the question of the role of energy budgets in determining reproductive behavior and other life-history traits of birds. The study reported here describes how the daily energy budgets of breeding populations of the zebra finch Taeniopygia guttata influenced the rate at which broods were produced, the size of broods that could be success-fully supported by their parents, and the subsequent survival of the parents. This was examined by experimentally altering the net rate of energy gain of four populations of randomly chosen zebra finches. The experimental design increased the amount of time spent searching for food without altering the amount of food available or consumed, which resulted in differences in the daily energy budget. Zebra finches with low rates of energy gain spent more time and energy foraging and had less energy available for reproduction than did finc...

I measured the heritability of foraging patch choice in a laboratory population of zebra finches (Taeniopygia guttata). Mothers and offspring were tested for their ability to discriminate between four foraging patches which provided four different rates of energy gain. Use of a foraging patch with a high rate of energy gain has been shown to confer a selective advantage on zebra finches in a similar experimental system. In this population of zebra finches there was a large amount of variation in foraging patch choice behaviour both within and among individuals. I determined that foraging patch choice was a phenotypically labile trait with a degree of stereotypy or repeatability, much lower than those typically recorded for morphological traits. The mating behaviour of zebra finches required that heritability be determined from a mother—offspring regression, which showed that narrow sense heritability of foraging patch choice was approximately 0.346. This heritability was significantly different than zero, as was heritability when it was limited by repeatability to 0.246. Foraging patch choice, a behaviour that has a demonstrated fitness consequence, had a heritable component in this laboratory population of zebra finches.

The critical assumption of optimal foraging theory is that foragers with high net rates of energy gain have higher fitness than foragers with low net rates of energy gain. To test this assumption 60 pairs of zebra finches, Taeniopygia guttata, were randomly assigned to four experimental treatments in which their rate of energy gain was altered by increasing search time, without changing the amount of food available or the amount of food consumed. Reproductive success was monitored for the lifetime of the birds. Reproductive success decreased as the net rate of energy gain decreased. Decreased reproductive success was caused by a small decrease in brood size, an increase in the time intervals between broods, and increased adult mortality. The result that reproductive success is positively correlated with the rate of energy gain while foraging supports the prevalent use of energy measurements to predict fitness.

A subset ofDrosophilaneurons that expresses crustacean cardioactive peptide (CCAP) has been shown previously to make the hormone bursicon, which is required for cuticle tanning and wing expansion after eclosion. Here we present evidence that CCAP-expressing neurons (NCCAP) consist of two functionally distinct groups, one of which releases bursicon into the hemolymph and the other of which regulates its release. The first group, which we call NCCAP-c929, includes 14 bursicon-expressing neurons of the abdominal ganglion that lie within the expression pattern of the enhancer-trap line c929-Gal4. We show that suppression of activity within this group blocks bursicon release into the hemolymph together with tanning and wing expansion. The second group, which we call NCCAP-R, consists of NCCAPneurons outside the c929-Gal4 pattern. Because suppression of synaptic transmission and protein kinase A (PKA) activity throughout NCCAP, but not in NCCAP-c929, also blocks tanning and wing expansion, we conclude that neurotransmission and PKA are required in NCCAP-R to regulate bursicon secretion from NCCAP-c929. Enhancement of electrical activity in NCCAP-R by expression of the bacterial sodium channel NaChBac also blocks tanning and wing expansion and leads to depletion of bursicon from central processes. NaChBac expression in NCCAP-c929 is without effect, suggesting that the abdominal bursicon-secreting neurons are likely to be silent until stimulated to release the hormone. Our results suggest that NCCAPform an interacting neuronal network responsible for the regulation and release of bursicon and suggest a model in which PKA-mediated stimulation of inputs to normally quiescent bursicon-expressing neurons activates release of the hormone.

The experiments presented here were designed to determine the origin of the temporally complex activity of antennal lobe projection neurons in the cockroach olfactory system. We determined this through the use of complex chemical stimuli that evoked neural activity recorded extracellularly from olfactory sensory neurons and intracellularly from antennal lobe projection neurons in the cockroach Periplaneta americana. Olfactory information was represented by a simple, short time-scale rate code in the olfactory sensory neurons. This rate code input from the sensory neurons was processed by the antennal lobe and transformed into a longer time-scale, temporally encoded output expressed across a smaller population of antennal lobe projection neurons. The projection neuron responses comprised temporal patterns of increases and decreases in spike frequency that differed among projection neurons and were consistent among repeated presentations of the same stimulus. Presentation of simple and complex chemical stimuli showed that the complexity of projection neuron activity was a product of the antennal lobes and was not associated with the chemical complexity of the stimulus. To characterize the encoding schemes used by each class of neurons, the responses were decomposed into their principal components. The stimulus was correlated with only the first principal component of the activity of sensory neurons, which is consistent with a rate encoding scheme. The stimulus was correlated with higher order principal components of the activity of projection neurons, which is consistent with a temporal encoding scheme.

Here we consider the issue of choice and how neural systems can be used to investigate the processes of discrimination, as well as the evolution of different kinds of choice-related behavior in animals. We develop these ideas in the context of three studies, among others. The first study is on the evolution of specialization in animals using locust feeding behavior as the leitmotif, where decision making in individuals is modeled by a 3-layer-perceptron. In this study the fitness of individuals depends on their response to signals from plants and the density of individuals using those plants [1]. The second is a study that investigates the evolution of species recognition in sympatric taxa using female mate choice in frogs as the leitmotif [2]. Here individuals are modeled by Elman nets (3-layered perceptrons with feedback) and their fitness is determined by their ability to discriminate conspecifics from heterospecifics. The third is a study of the response characteristics of a recurrent Hopfield-type neural network to input that represents olfactory stimuli. The connectivity of this net reflects the basic architectural features of the neuron in the insect antennal lobe, as typified by cockroaches or bees [3].

OPTIMAL foraging theory is based on the assumption that natural selection favours animals that forage most efficiently1–3. But such selection does not act directly on foraging efficiency, but rather indirectly by favouring animals that survive and reproduce most successfully. Studies that use optimal foraging models often assume that maximization of some behavioural currency, such as the animal's net rate of energy gain, maximizes the animal's fitness4, but rarely is an attempt made to test this assumption5–8. Most studies of the effects of foraging behaviour on fitness fail to control for the amount of energy gained by the foraging animals5–8, and lead to the obvious conclusion that animals that eat more reproduce more. Often the studies do not control for characters correlated with foraging behaviour6 or compare traits assumed to be correlated with fitness7,8. A better method would be to assign net rates of energy gain to randomly chosen individuals for their entire lifetimes in a controlled environment and measure fitness directly. Variation in the amount of energy consumed would be controlled by using individuals that employ a time-minimizing foraging strategy9 and would alter the time taken to satisfy their daily energy requirements, while obtaining the same absolute amount of energy. I have now manipulated the net rate of energy gain in four populations of the zebra finch Taeniopygia guttata, and show that fitness, as measured by population growth rate, is indeed positively and significantly correlated with the net rate of energy gain.

During the metamorphosis of Manduca sexta the larval nervous system is reorganized to allow the generation of behaviors that are specific to the pupal and adult stages. In some instances, metamorphic changes in neurons that persist from the larval stage are segment-specific and lead to expression of segment-specific behavior in later stages. At the larval-pupal transition, the larval abdominal bending behavior, which is distributed throughout the abdomen, changes to the pupal gin trap behavior which is restricted to three abdominal segments. This study suggests that the neural circuit that underlies larval bending undergoes segment specific modifications to produce the segmentally restricted gin trap behavior. We show, however, that non-gin trap segments go through a developmental change similar to that seen in gin trap segments. Pupal-specific motor patterns are produced by stimulation of sensory neurons in abdominal segments that do not have gin traps and cannot produce the gin trap behavior. In particular, sensory stimulation in non-gin trap pupal segments evokes a motor response that is faster than the larval response and that displays the triphasic contralateral-ipsilateral-contralateral activity pattern that is typical of the pupal gin trap behavior. Despite the alteration of reflex activity in all segments, developmental changes in sensory neuron morphology are restricted to those segments that form gin traps. In non-gin trap segments, persistent sensory neurons do not expand their terminal arbors, as do sensory neurons in gin trap segments, yet are capable of eliciting gin trap-like motor responses.

A classical conditioning paradigm was used to test the ability of Sternopygus macrurus to detect EOD-like stimuli (sine waves) of different frequencies. The behavioral tuning curves were quite close in shape to tuning curves based on single-unit recordings of T units, although the sensitivity at all frequencies was much greater. The behavioral curves showed notches of greatly reduced sensitivity when the test frequency was equal to, or twice the EOD frequency. The EOD of each of the fish was eliminated by lesioning the medullary pacemaker nucleus, and the fish were retested. The resulting tuning curves were nearly the same in shape as those of the EOD-intact individuals, but the PMN-lesioned fish showed an overall reduction of sensitivity of 30 dB. The EOD appears to enhance sensitivity by placing the summed stimulus (test stimulus + fish's EOD) at an amplitude where T units are maximally sensitive to small temporal modulations in the fish's own EOD. Peripheral tuning appears to limit the ability of males to detect the EOD of females, since these are, on average, an octave higher in frequency than the male EOD, while the peak sensitivity of the male occurs 5–10 Hz above its own EOD frequency.

SummaryWe present the Real-time Accurate Cell-shape Extractor (RACE), a high-throughput image analysis framework for automated three-dimensional cell segmentation in large-scale images. RACE is 55–330 times faster and 2–5 times more accurate than state-of-the-art methods. We demonstrate the generality of RACE by extracting cell-shape information from entire Drosophila, zebrafish, and mouse embryos imaged with confocal and light-sheet microscopes. Using RACE, we automatically reconstructed cellular-resolution tissue anisotropy maps across developing Drosophila embryos and quantified differences in cell-shape dynamics in wild-type and mutant embryos. We furthermore integrated RACE with our framework for automated cell lineaging and performed joint segmentation and cell tracking in entire Drosophila embryos. RACE processed these terabyte-sized datasets on a single computer within 1.4 days. RACE is easy to use, as it requires adjustment of only three parameters, takes full advantage of state-of-the-art multi-core processors and graphics cards, and is available as open-source software for Windows, Linux, and Mac OS.