|Home | About | Journals | Submit | Contact Us | Français|
The nematode C. elegans is an excellent model organism for a systems-level understanding of neural circuits and behavior. Advances in the quantitative analyses of behavior and neuronal activity, and the development of new technologies to precisely control and monitor the workings of interconnected circuits, now allow investigations into the molecular, cellular and systems-level strategies that transform sensory inputs into precise behavioral outcomes.
The aim of systems neuroscience is to understand how assemblies of neural circuits generate coordinated motor outputs, and how these motor outputs are modified in response to sensory input and experience to effect coherent behavior. C. elegans has long been an ideal animal in which to explore the genetic basis of behavior, due to its experimental amenability, and its small and well-defined nervous system. Here, we review recent advances that are now allowing the use of C. elegans to pursue longstanding questions in systems-level neuroscience. These advances include the development of new behavioral assays that quantify worm motor outputs in defined stimulus environments with high precision, as well as new methods to monitor and manipulate the sensorimotor circuits that produce these behaviors. Thus, the goal of achieving a comprehensive description of the pathways by which a sensory stimulus is transformed through multiple circuit layers into a defined motor response might now be within reach in C. elegans.
Except for simple reflex arcs, sensory inputs do not generate motor outputs in a deterministic manner. Instead, sensory inputs modulate ongoing patterns of neuromuscular activity that regulate motor behavioral outcomes such as locomotion. Thus, understanding the control and production of the motor output is a starting point, and not the last step, in the study of sensory-evoked behaviors. The rhythmic locomotory gait of C. elegans has long been an area of intense focus, informed by classic studies that sought the structure and function of the motor circuits that produce oscillation and undulation in swimming leech and lamprey [1,2]. Unlike in animals with larger nervous systems, however, C. elegans offers an opportunity for a complete understanding of motor circuit function and modulation within the freely moving animal.
C. elegans is the only animal whose complete wiring diagram has been anatomically mapped by serial section electron microscopy . (We are referring to the hermaphrodite; the wiring diagram in the male is not yet complete). Although parts of the motor circuit - in particular, the set of synaptic connections between 58 motor neurons in the ventral nerve cord - were not actually completed in the original reconstruction, these have now been mapped [4•]. The motor circuit for forward movement consists of four types of motoneurons with a consistent connectivity motif along the animal s length (Figure 1). The pattern of connectivity within the motif suggests an elegant mechanism for contralateral inhibition: DB/VB neurons excite dorsal/ventral muscles, and VD/DD neurons in turn inhibit the opposing ventral/dorsal muscles. Contralateral inhibition mediated by this repeating motif helps to coordinate sinusoidal bending by prohibiting simultaneous contraction of the ventral and dorsal musculature.
Beyond contralateral inhibition, a systems-level understanding of the motor circuit is needed to explain the detailed dynamics of the locomotory rhythm. Worms swim in fluids or crawl on agar surfaces with different locomotory gaits [5,6]. Swimming is characterized by large amplitude, high frequency, long wavelength undulations, whereas crawling is characterized by small amplitude, low frequency, short wavelength undulations. When nematodes navigate environments that pose intermediate amounts of mechanical load than those experienced during swimming (low load) or crawling (high load), they display intermediate gaits with continuous variation in amplitude, frequency, and wavelength [7•,8]. Thus, a full understanding of the locomotory rhythm is a multilevel problem in biomechanics, sensory feedback, and the rhythmic activity of neuronal ensembles that produce and propagate rhythmic activity [9,10]. The entry point in a detailed study of the locomotory gait is that it is eminently quantifiable experimentally, and can be modeled mathematically. Progress has been made on both these fronts recently [9,11,12], leading towards a more a detailed description of gait dynamics and the operation of the motor circuit.
The next step in dissecting the motor circuit is to define the contributions of individual neurons or neuronal ensembles to defined aspects of motor output. The transparent, genetically tractable worm is ideal for implementing optogenetic assays for monitoring neuronal or muscle cell activity , or directly manipulating their activity with light-activated ion channels [14–16•]. In a recent study, activity patterns within the muscles of crawling and swimming worms were visualized by calcium imaging in freely moving worms, and it was shown that their distinct gaits correspond to distinct spatiotemporal patterns of activity in the ventral and dorsal musculature . Using light-activated channels and electrophysiology, Liu et al. [17•] simultaneously measured the electrical current evoked in muscle cells upon depolarization or hyperpolarization of motor neurons, and found that levels of neurotransmission at the neuromuscular junction were continuously graded [17•]. This finding indicates that the motor circuit is analog rather than digital, perhaps allowing continuous variation of contraction and relaxation within muscle cells during the propagation of an undulatory wave, or the continuous adaptation of locomotory gait in different mechanical environments.
To date, optogenetic probes have been used to globally alter the locomotory patterns of freely moving animals. Thus, an entire worm has been paralyzed by simultaneous contraction or relaxation of all muscle cells, or by activation of entire classes of motor neurons [14,16•]. However, given that the locomotory gait is generated by precise spatiotemporal patterns of activity in different circuit elements, progress will be driven with the development of new techniques to microstimulate the motor circuit at different points in space and time .
Worms do not swim or crawl continuously in a given direction, but instead exhibit spontaneous transitions between locomotory modes: periods of forward movement (runs), periods of backward movement (reversals), and sharp turns that redirect forward movement without pause [19,20]. These outputs of the motor circuit are governed by command interneurons, whose roles in directing locomotion were elegantly mapped functionally via laser-mediated ablations [3,21]. These command interneurons are stochastically active, and drive spontaneous behaviors even in the absence of incoming sensory inputs [19,20].
The space of possible motor outputs that drive navigation is remarkably small, recently demonstrated in a mathematically rigorous way by Stephens et al [22••] who used principal component analysis to demonstrate that any attainable posture of a crawling worm could be reconstructed from just four fundamental worm shapes, what they called eigenworms. Thus, given any state of motor output, a multiplicity of states of underlying neural activity might correspond to that motor state. Nevertheless, a statistical analysis of transitions between observable states of motor activity - between runs, reversals, and turns - can shed light on the transitions between underlying states of neural activity that correspond to those motor states. For example, Srivastava et al [23••] analyzed the intervals of forward movement exhibited by swimming worms, where successive intervals were separated by the spontaneous occurrence of a sharp turn. A swimming worm will exhibit turns at random points in time, sometimes exhibiting several turns in rapid succession, sometimes waiting long intervals between turns. Detailed analysis of motor statistics using a Hidden Markov Model showed that the spontaneous run/turn behavior of the swimming worm could be described with two run states and one turn state: one run with a high rate of transition to a turn; another run state but with a low rate of transition to a turn; along with “hidden” transitions between the two run states.
In C. elegans, one does not have to settle for inferring hidden transitions between states of neural activity based on observable motor activity. It is now possible to simultaneously monitor neuronal activity and locomotory behavior in behaving nematodes [24,25•]. Microfluidic devices may greatly facilitate the simultaneous imaging of behavior and calcium dynamics in behaving animals. Chronis et al measured intracellular calcium dynamics in the backward command AVA interneurons of worms trapped in the so-called behavior chip with concurrent measurements of propagated body waves [26••]. This work showed that responses in this neuron type were correlated with the initiation, and maintained through the duration, of anterior-traveling waves (and thus, backward locomotion), providing a one-to-one correlation between activity of this neuron and reversal. Additional such measurements can readily generate a correlation map between command interneuron activity and specific behavioral transitions, inferred previously solely from behavioral measurements.
It has long been known that faced with a gradient of a chemical on an agar plate, C. elegans will end up at the peak (if it is an attractive chemical), or at the trough (if it is a repulsive chemical) . When C. elegans navigates gradients in its environment, it modulates the ongoing pattern of runs, reversals, and sharp turns according to rules that yield migration up or down gradients or aggregation near preferred conditions. These rules constitute the computational underpinning of navigational behavior, the mapping of sensory inputs onto motor outputs that enables the worm to reach its goals. Insights into the underlying decision mechanisms that allow animals to move towards or away from the stimulus only arose following high resolution tracking of behaving C. elegans on sensory gradients, along with improvements in the methodologies used to deliver different types of sensory stimuli [28–33, 29•].
Although the activity patterns of command interneurons are deterministically coupled to motor output, the relationship between sensory neuronal activity and motor output is more subtle. Worms climb up or down gradients in their environment by using a biased random walk, a behavioral strategy similar to that used by bacteria to navigate gradients . If the worm encounters improving conditions during a run, it lengthens that run (i.e., it postpones any reorientation maneuvers). If it encounters declining conditions, it shortens that run by turning [28,32]. Thus, the activity of a sensory neuron that is responding to an ambient gradient is related to the probability of stochastic transitions between run states and reorientation states. In other words, the nervous system of the worm drives spontaneous and stochastic dynamics of the motor system even in an isotropic environment, and ambient or past sensory cues simply modulate the statistics of these dynamics [35–38].
The basic computation in performing a biased random walk strategy is the time derivative of the sensory input; the worm has to know whether life is getting better or worse during each run in order to adjust the probability of transitions between run and reorientation. Sensory systems that encode time-derivatives - from bacterial chemotaxis to vertebrate phototransduction - commonly exploit the dynamics of short-term adaptation. In the worm, the locus of short-term adaptation for temporal changes in sensory input appears to be in the sensory neurons themselves. When worms are subjected to step changes in temperature, ambient chemicals or gases, recordings of calcium dynamics in the primary sensory cells exhibit adaptive dynamics with similar time course as the change in reorientation rate of behaving animals [39••–46•]. In some cases, upsteps or downsteps in sensory inputs can both raise or lower calcium levels within a single sensory neuron [26••,43•]. In these cases, sensory neuron activity itself may be sufficient to modulate activity of the downstream navigation circuit to generate biased random walks. However, in other cases, integration must occur at other loci in the circuit. For example, the left/right members of a sensory neuron pair (the ASE chemosensory neurons), or different sensory neurons (BAG and URX) sense upsteps or downsteps in chemical or oxygen concentrations, respectively [40••,44•,45•], suggesting that sensory information is integrated further downstream in the circuit in the calculation of motor decisions.
During navigation via the random walk strategy, not only is the probability of turns random, the direction of reorientation following the turn is also random. In other words, the animal does not steer itself towards a specific direction, but navigates towards its preferred direction by simply modulating the frequency of turns. However, a weak biasing of the turn angle was noted in early analyses of the random walk strategy performed by C. elegans on a chemical gradient . Recent detailed analyses of locomotory behavior data also suggest that, in some cases, the worm can orient itself in a deterministic manner [47••]. In chemotaxis, a gradual steering strategy towards higher concentration appears to work together with the biased random walk strategy to efficiently allow navigation up gradients. In this weathervane strategy, the worm will gradually curve its trajectories to better align itself with the gradient direction during long runs [47••]. A steering strategy is also used by worms to track isotherms at their preferred temperature in thermotaxis behavior . To remain on a track, the worm markedly lowers the probability per unit time of generating a reorientation maneuver that would throw it off the track, and uses temporal variations in temperature recorded at its nose to steer itself along the track . Subsets of the same circuits that mediate random walk behaviors appear to also regulate steering on both chemical and thermal gradients [47••,49].
Almost certainly, more sophisticated patterns of behavior will require more sophisticated computations at the level of sensory and downstream neurons. For instance, on an oxygen gradient, worms navigate not to the peak of the gradient, but to an optimum experience-dependent concentration [50,51]. In a theoretical study, Dunn and Lockery [52••] showed that distinct behavioral strategies are optimal for gradient climbing and goal seeking behaviors, in which an animal seeks the highest or an intermediate point on a gradient such as in chemotaxis or aerotaxis behaviors, respectively. This analysis suggested that these strategies may be encoded in different motifs downstream of the sensory neurons. Examination of neuronal activity in first- and second-layer interneurons, and simultaneous imaging of multiple circuit components may provide hints regarding the mechanisms by which more sophisticated behavioral strategies are encoded by the navigation circuit.
Behavioral responses to stimuli are not always hardwired into the nervous system. Worm navigation behavior is highly flexible and adaptive, and can be modulated by experience. If removed from food and placed in an isotropic environment, the probability of runs and turns made by an animal is dictated by the length of time the animal has been food-deprived [35–38], providing a clear example of modulation of the locomotory circuit by past experience. Similarly, turn rates and run lengths can be modulated to result in avoidance of a normally attractive chemical by prolonged exposure to high concentrations of the chemical, starvation, or coupling with an aversive chemical or experience [53–57].
Where does this plasticity occur in the circuit? The answer appears to be largely at the level of neurons upstream of the command interneurons that directly regulate locomotory behavior. Thus, both the input and/or the output of chemosensory, and first or second layer interneurons are altered by experience to modulate the stochastic dynamics of the downstream core locomotory circuit [36,54•–61••]. Of particular interest is a recent study that indicates that experience-dependent altered neurotransmission of a single sensory neuron type is sufficient to switch an attractive response to an aversive one [61••]. Although the mechanism of this switch has not yet been fully elucidated, not surprisingly, neuromodulation via neuropeptides and hormones has been implicated in regulation of experience-dependent plasticity [54,57,58,60,62].
Recent technical developments and increased sophistication in behavioral analyses and quantification of neuronal activity have been used to describe the neural ensembles that generate motor outputs, and identify the mechanisms by which sensory stimuli modulate these outputs to achieve a defined behavior. To fulfill the promise that the worm system holds to understand behavior at the molecular, cellular, and systems levels, we will need to continue to enhance the resolution of experimental measurements at all levels, as well as enhance the sophistication of the statistical analyses that are used to describe and understand these measurements. We will also need to continue to elucidate the contributions of individual neurons to a behavior via high resolution behavioral analyses, and implement methods to simultaneously image all components of interconnected circuits in live, behaving animals. While a reasonable and practical goal is to investigate these issues in the 2D space of a channel in a nanodevice or on the agar surface of a petri plate, the ultimate goal will be to understand how the nervous system directs behavior in the complex 3D environment of the worm s soil habitat. To paraphrase Barack Obama - it took a lot of blood, sweat, and worms to get to where we are today, but we have just begun.
We thank Eve Marder, Chris Fang-Yen and Quan Wen for comments and discussion. P. Sengupta is supported by NIH R01 GM081639; A. Samuel is supported by NSF PHY-0448289 and by NIH Pioneer Award 5DP1OD004064.
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.