A circuit model of the temporal pattern generator of Caenorhabditis egg-laying behavior
© Zhang et al; licensee BioMed Central Ltd. 2010
Received: 8 May 2009
Accepted: 7 June 2010
Published: 7 June 2010
Egg-laying behavior in the nematode C. elegans displays a distinct clustered temporal pattern: egg-laying events occur primarily in bursts or active phases, separated by inactive phases during which eggs are retained. The onset of the active phase can be modeled as a Poisson process with a time constant of approximately 20 minutes, while egg-laying events within an active phase occur with a faster time constant of approximately 20 seconds. Here we propose a cellular model for how the temporal pattern of egg-laying might be generated, based on genetic and cell-biological experiments and statistical analyses.
We suggest that the HSN neuron is the executive neuron driving egg-laying events. We propose that the VC neurons act as "single egg counters" that inhibit HSN activity for short periods in response to individual egg-laying events. We further propose that the uv1 neuroendocrine cells are "cluster counters", which inhibit HSN activity for longer periods and are responsible for the time constant of the inactive phase. Together they form an integrated circuit that drives the clustered egg-laying pattern.
The detailed predictions derived from this model can now be tested by straightforward validation experiments.
The execution of specific behaviors relies on specialized motor programs, which generate specific spatial and temporal patterns of muscle contraction. The regulatory circuits that control the patterning are in principle independent of the concrete instantiation of the muscular end organ (see for example ). The same logical regulatory structure can be used to control a wide variety of behavioral programs. This also means that studying such pattern generating circuitries in experimental organisms can provide insights into fundamental circuit designs that are re-used in the much more complex brains of higher animals. The neuronal circuits of the worm C. elegans are particularly attractive for this kind of basic studies, because all neurons and muscle cells are individually known, their connectivity is constant, and the transparent body facilitates visualization and manipulation of single cells in living animals. In addition, genetic manipulation is much easier than in vertebrates. In the present paper we take advantage of these experimental options to elucidate the logical connectivity of the circuit controlling the egg-laying behavior of C. elegans and show how it implements basic mathematic operations that control the temporal pattern of egg-laying.
The core components of the neural circuit driving each individual egg-laying event have recently been characterized (; Figure 1B). Two major pairs of neurons contribute directly to the control of egg-laying. The HSN neurons are the central command neurons of the circuit; each autonomous individual Ca2+ spike in HSN can induce contraction of the vulval muscle cells (vm1/vm2) and result in one egg-laying event. The HSNs are modulated by the VC neurons, and also integrate environmentally-controlled neuromodulatory signals that influence egg-laying frequency. For example, posterior body touch transiently inhibits egg-laying by exciting the PLM mechanoreceptor neurons, which in turn directly inhibit HSN activity.
How do these cells create the clustered temporal pattern in egg-laying? Which cells are required to cause the regular transition between active and inactive states? In this paper we propose a model to explain this behavior using evidence from cell biology, genetics and statistics. Ca2+ levels in the HSNs (as well as other neurons of the circuit) have been visualized at high temporal resolution using calcium-sensitive fluorescent dyes, giving us an opportunity to directly monitor the transmission and integration of signals in the circuit. Furthermore, mutants with defects in individual cells make it possible to discriminate correlated excitation from causal links in the circuit. Based on these data, we propose an integrated circuit model that can explain the logic underlying the egg-laying temporal pattern.
HSN is under negative feedback regulation from egg-laying
Interestingly, in egl-15(n484) worms, which are defective in egg-laying behavior due to abnormal vulval muscle development, we found that half of HSN recordings exhibit unrestrained oscillations very similar to the theoretical pattern in Figure 2A, with HSN oscillating about half the time (4 out of 9 recordings) and silent otherwise (Figure 2C). The overall activity of egl-15 HSN is much higher than that of wild-type HSN, due to the hyperactive oscillations in the former. This suggests that egg-laying itself provides inhibitory feedback to the HSN neurons.
Egg-laying provides inhibitory feedback to the VC neurons
To determine the causal relationship between HSN activity and egg-laying muscle contraction, we asked if there is a fixed correlation between the timing of egg-laying and the HSN and VC calcium spikes. Specifically, we investigated whether egg-laying is coupled to any specific point during the calcium transient. A co-variation test (Figure 3B) showed that egg-laying happens when Ca2+ levels are at their peak in the VC neurons. Linear regression shows that egg-laying and the calcium peak happen almost simultaneously, as the slope of the regression line is close to 1 and the intercept close to 0. Thus, egg-laying is tightly coordinated with the end of the calcium transient. From a mechanistic point of view this is surprising: as egg-laying is caused by VC depolarization (which leads to neurotransmitter release and vulval muscle contraction), we might expect egg-laying to occur at a constant delay after the beginning of the depolarization. In contrast, since the duration of the transient (from onset to peak) is highly variable, a correlation between egg-laying and the calcium peak was unanticipated.
As the spike peak reflects the start of repolarization, the simplest explanation for the coupling of egg-laying to spike peaks is that the repolarization of the VC neurons is actually induced by egg-laying. This would establish a negative feedback loop, where depolarization of VC causes egg-laying, which in turn causes termination of the calcium transient and repolarization of the VC neuron.
VC inhibits HSN in the short term
The VC neurons send arborized processes to the surface of the vulval epithelium in addition to their synapses on the vulval muscles. Based on the anatomical arrangement, it can be hypothesized that the function of the processes reaching the vulval epithelium is to sense stretch. Thus, VC neurons are suitably positioned to directly register a successful egg-laying event by sensing muscle or vulval epithelial stretch, and feed this information back to the HSN neurons by inhibiting their activity. Therefore, we hypothesize that individual egg-laying events might be detected by the VC neurons, which in turn provide short-term negative feed back to the HSNs to modulate their activity during the active state.
Long-term feedback may be mediated by the uv1 cells
While the VC neurons appear to provide important inhibitory feedback to the HSN cells, additional cells are likely to provide further modulatory input to the egg-laying circuit. In lin-39 worms, in which the VC neurons are absent, very long silent periods (more than several minutes) are still observed between trains of HSN activity (Figure 4A). This observation suggests that another there may be additional, VC-independent inhibitory inputs to the HSNs that might act on longer time scales .
Since the uv1 s are candidate stretch receptor cells , we propose that they might be activated by movements of the vulva associated with egg-laying. In particular, they might facilitate the long quiescent periods that follow clusters of egg-laying events. Since the uv1 s release peptides and amines in a paracrine rather than a synaptic fashion, they may be well-suited to mediate long-lasting inhibition following multiple egg-laying events. Neurohormones have a much slower action and clearance kinetics compared to neurotransmitters released at synapses and this slow kinetics means that closely timed repeated neurotransmitter releasing events could give rise to an additive effect, building up a strong enough extracellular gradients to act at a longer distance. We therefore hypothesize that the uv1 s release an inhibitory neurohormone in response to a cluster of egg-laying events, inducing long term silencing of HSN activity and egg-laying behavior corresponding to the inactive phase of egg-laying.
Circuit model of egg-laying pattern generator
In our model, the HSN neurons can be seen as an information integration center, processing signals from uv1 and VC cells on top of their own autonomous activity. HSN functions as a digital 'NOR' gate, giving the output of "1" only if both VC and uv1 are inactive. The 'NOR' gate is the most universal logical gate in that it can be used to create all other 15 elementary logical gates (only the 'NAND' gate has the same universality). This means that even the most complex computers (and brains?) can be constructed based solely on a combination of 'NOR' gates . The HSN egg-laying circuit is potentially an elegant example of a biological implementation of a 'NOR' calculation. It will therefore be a valuable system for more detailed studies of this mechanism, which may have intriguing implications in general biological computations.
This model immediately suggests some new characteristics of the temporal pattern that may be worthy of future investigation. According to our model, the number of eggs per cluster will be maintained at a relatively constant level, as it directly controls the switching off of the active state. Furthermore, if our model is correct, the distance between the HSN and uv1 cells should influence the number of eggs laid per cluster, because the longer the distance, the more neurohormone releasing events are required to build up a high enough concentration of hormone to terminate the active phase. Interestingly, HSN cells are first born at the tail and migrate to a location close to the vulva, traveling almost half of the body length (~50 micron). By evolution, the final position of HSN might be naturally selected to produce a certain number of eggs per cluster to achieve the best strategy for long-term species survival. It will be interesting to see if different natural variants of C. elegans have different HSN-vulva distances.
Previously, the clustered temporal pattern in egg-laying was described mathematically as two Poisson processes with different time constants (τ1 = 1/λ1 and τ2 = 1/(p*λ2)) and a switching probability p . The model we developed from biological evidence fits that mathematical description surprisingly well and suggests possible specific biological pathways controlling each parameter. The Poisson-like nature of both long and short interval distributions can be explained by the HSN cells randomly switching back to the egg-laying state after being inhibited by either the VCs or the uv1s; the intra- and inter-cluster time constants correspond to the kinetics of these two switching events, respectively. The switching probability p in the mathematical description corresponds to how many consecutive egg-laying events in a cluster are required to activate the uv1 pathway. The value of p is determined by the distance between the HSN and uv1 cells together with the diffusion constant of uv1 neurohormones. This could potentially be validated in morphogenesis mutants, where the distance between neurons is altered.
Recently it has been found that reversal behavior of worms can also be described as a sum of two exponential processes (unpublished observations; Srivistava N, Clark D, Samuel AD). Therefore, controlled autonomous switching might be is regular feature of many behaviors (e.g., alternation between rest and arousal). For C. elegans egg-laying, two two-cell circuits, uv1-HSN and VC-HSN, control the state alternation. It will be interesting to see whether or not equivalently structured circuits are employed in other behavioral control situations.
In systems biology, the usefulness of a computational model is mainly determined by its predictive power. The cellular and molecular model we present here is the first one for generating a clustered pattern and so far the simplest to explain all available data, including our own extensive genetic manipulation results and behavior recordings. At the same time it makes a number of very concrete, testable predictions that can now be explored by further experimentation, using additional advanced technologies. For instance, we suggest using vulva developmental mutants, which are incapable of accomplishing egg-laying, to confirm the existence of the postulated negative feedback loops. Likewise, the activity of the uv1 cells can be imaged using cameleon Ca2+ indicators to see whether or not they are activated by egg-laying events. If the VCs directly sense egg-laying events, they are predicted to contain touch sensitive channels, which could be identified by electrophysiological studies. Finally, we predict that the egg-number per cluster is determined by the diffusion distance between HSN and the uv1 cells, which can again be tested in mutants or ablated animals. Testing concrete predictions like these will lead to a refined understanding of the molecular and cellular circuitry underlying the complex behavior of this important neurobiological model animal.
All nematodes were grown at 20.5+0.5°C on standard Nematode Growth Medium (NGM) seeded with E. coli strain OP50 as the food source. Nematodes were assayed 24 hours after late L4 stage at 20.5 ± 0.5°C.
Egg-laying behavior assays
M9 egg-laying assays were performed in 96-well plates. 100 μl of melted 2% agarose in M9 were placed in each well and allowed to dry for 1-1.5 hours before starting the behavioral assay. Individual worms were placed on the solid agarose pad in each well and after about 1 min were covered with 50 μl of M9. The number of laid eggs was counted after 1.5 hours.
Ca2+ imaging experiments
The protocol is the same as described in . Combined fluorescent and visible microscopy was used to view both the Ca2+ signal and egg-laying events simultaneously. The microscope equipment was as described earlier ; briefly, a Zeiss Axioskop 2 upright microscope equipped with a Hamamatsu Orca ER CCD camera, a Hamamatsu W-View emission image splitter and a Uniblitz Shutter (Vincent Associates) were used. A minimum of fluorescent light, 20-60% of the original light power with the neutral density filter (ND) of 2.0, was used. Weak visible light, barely revealing egg-laying events, was used to minimize interference with the cameleon fluorescent signal. Fluorescent images were acquired and saved using MetaVue 4.6 (Universal Imaging) at a frequency of either 33 Hz or 10 Hz for muscle and neuron imaging respectively, using a 63× Zeiss Achroplan water immersion objective.
Worms were immobilized with Nexaband S/C cyanoacrylate veterinary glue on a small agarose pad with the buffer of choice freshly made on a microscopy slide. The worms were then quickly covered with the buffer of choice and immediately moved under the microscope for recording. For HSN and vulval muscle imaging, the agarose pad was made in the same buffer as the bath. For VC imaging, a 2% agarose pad in M9 was used, regardless of the bath buffer, due to the fast adaptation of VCs to low osmolarity conditions. Agarose pads were made immediately before slide preparation in order to minimize the loss of focus in long recordings (10 mins), which usually is caused by swelling of the agarose pad. Worms were allowed to equilibrate for 2 to 4 mins before the start of recording.
We thank Yoram Burak (Center for Brain Sciences, Harvard) for critical reading and helpful discussions. This work was supported by the Medical Research Council, UK, and grants from the National Institute of Drug Abuse (DA016445) and the Human Frontier Science Program (W.R.S.).
- Norris BJ, Weaver AL, Morris LG, Wenning A, Garcia PA, Calabrese RL: A central pattern generator producing alternative outputs: temporal pattern of premotor activity. J Neurophysiol. 2006, 96 (1): 309-326. 10.1152/jn.00011.2006View ArticlePubMed
- Zhou GT, Schafer WR, Schafer RW: A three-state biological point process model and its parameter estimation. IEEE Trans on Signal Processing. 1998, 46: 2698-2707. 10.1109/78.720372.View Article
- Zhang M, Chung SH, Fang-Yen C, Craig C, Kerr RA, Suzuki H, Samuel AD, Mazur E, Schafer WR: A self-regulating feed-forward circuit controlling C. elegans egg-laying behavior. Curr Biol. 2008, 18 (19): 1445-1455. 10.1016/j.cub.2008.08.047PubMed CentralView ArticlePubMed
- Chalfie M, Thomson JN: Organization of neuronal microtubules in the nematode Caenorhabditis elegans. J Cell Biol. 1979, 82 (1): 278-289. 10.1083/jcb.82.1.278View ArticlePubMed
- Jose AM, Bany IA, Chase DL, Koelle MR: A specific subset of transient receptor potential vanilloid-type channel subunits in Caenorhabditis elegans endocrine cells function as mixed heteromers to promote neurotransmitter release. Genetics. 2007, 175 (1): 93-105. 10.1534/genetics.106.065516PubMed CentralView ArticlePubMed
- Alkema MJ, Hunter-Ensor M, Ringstad N, Horvitz HR: Tyramine Functions independently of octopamine in the Caenorhabditis elegans nervous system. Neuron. 2005, 46 (2): 247-260. 10.1016/j.neuron.2005.02.024View ArticlePubMed
- Kim K, Li C: Expression and regulation of an FMRFamide-related neuropeptide gene family in Caenorhabditis elegans. J Comp Neurol. 2004, 475 (4): 540-550. 10.1002/cne.20189View ArticlePubMed
- Kurzweil R: The age of intelligent machines. 1990, Cambridge, Mass; London: MIT Press,
- Suzuki H, Kerr R, Bianchi L, Frokjaer-Jensen C, Slone D, Xue J, Gerstbrein B, Driscoll M, Schafer WR: In vivo imaging of C. elegans mechanosensory neurons demonstrates a specific role for the MEC-4 channel in the process of gentle touch sensation. Neuron. 2003, 39 (6): 1005-1017. 10.1016/j.neuron.2003.08.015View ArticlePubMed
This article is published under license to BioMed Central Ltd. This is an Open Access article distributed under the terms of the Creative Commons Attribution License (http://creativecommons.org/licenses/by/2.0), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.