Monoaminergic Orchestration of Motor Programs in a Complex C. elegans Behavior

A single monoamine can orchestrate different phases of a compound motor sequence in C. elegans through the synaptic and extra-synaptic activation of distinct classes of receptors.


Introduction
Complex behaviors require the temporal coordination of independent motor programs in which neurotransmitters and neuromodulators orchestrate the output of neural circuits. How the nervous system directs sequential activation and inhibition of assemblies of neurons, however, is largely unclear. Neurotransmitters can directly activate ligand-gated ion channels at synapses, inducing rapid changes in the electrical activity of postsynaptic cells. Neuromodulators generally act through G-protein coupled receptors (GPCR) that activate intracellular signaling cascades with slower but longer lasting effects. The release of neuromodulators can activate or refine basic motor patterns generated by fast-acting neurotransmitters in a neural network [1][2][3].
In mammals, monoamines such as serotonin, dopamine, and noradrenaline are associated with specific behavioral states. Adrenergic modulation provides one of the most striking examples of the coordination of behavior and physiology to reflect an internal state of stress. Adrenergic transmitters stimulate the amygdala and increase heart rate, muscle tone, oxygen supply to the brain, and the release of glucose from energy stores to prepare for a fight-or-flight response [4]. Noradrenaline and adrenaline are not used by invertebrates, but the structurally related monoamines octopamine and tyramine are often considered as the invertebrate counterparts of these adrenergic transmitters [5]. Octopamine and tyramine have been implicated in subordinate behavior in lobsters [6], the honey bee sting response [7], the energy metabolism of flight in locusts [8], aggression in crickets and fruit flies [9][10][11][12], and the escape response of C. elegans [13]. Pioneering studies in the locust [14], mollusks [15], and crustaceans [16] have supported an ''orchestration hypothesis'' [17] where monoamines control behavioral states through the recruitment of distinct neural circuits. However, the sensory input that triggers the release of monoamines and the molecular and neural coding of these behaviors remain poorly understood.
The C. elegans escape response consists of a behavioral sequence used by the animal to navigate away from a threatening stimulus. C. elegans moves on its side by propagating a sinusoidal wave of ventraldorsal muscle contractions along the length of its body [18]. Locomotion is normally accompanied by exploratory head movements in which the tip of the nose moves rapidly from side to side. Gentle touch to the head elicits a backing response and the suppression of exploratory head movements [13,19]. The backing response is usually followed by a deep ventral head bend, allowing the animal to make a sharp (omega) turn and change its direction of locomotion. The completion of the entire escape response takes approximately 10 s and requires sensory processing, decision-making, and sequential inhibition and activation of distinct motor programs. Therefore, the anterior touch response is a highly orchestrated motor sequence with complexity far beyond that of a simple reflex.
The neural wiring diagram in combination with genetic and laser ablation experiments has provided a framework for the neural circuit that controls the initial phase of the escape response [19][20][21]. C. elegans has a single pair of tyraminergic motor neurons that are essential in coordinating the backing response with suppression of head movements [13]. Synaptic activation of the tyramine-gated chloride channel, LGC-55, inhibits forward locomotion and induces the relaxation of neck muscle [22]. The coordination of these motor programs increases the animals' chances of escaping from predacious fungi that use constricting rings to catch nematodes, illustrating the vital importance of monoaminergic motor control [23,24]. How this initial phase of the escape response is temporally linked to later stages in which the animal makes a sharp turn to navigate away from the stimulus is unknown. To elucidate how monoamines may orchestrate the activity of specific neural circuits in complex behaviors, we analyzed the role of tyramine in C. elegans locomotion during the escape response. We show that the extrasynaptic activation of a Gprotein coupled tyramine receptor generates asymmetry in a locomotion program, thus allowing the animal to execute a deep ventral turn and navigate away from the stimulus.

ser-2 Mutants Are Resistant to Exogenous Tyramine
C. elegans become immobilized on plates containing exogenous tyramine in a dose-dependent manner ( Figure 1A and Figure S1) [22]. Three GPCRs have been shown to bind tyramine with high affinity: TYRA-2, TYRA-3, and SER-2 [25][26][27][28]. To determine whether the effects of tyramine are mediated by these GPCRs, we examined the locomotion of ser-2(pk1357 and ok2103), tyra-2(tm1815), and tyra-3(ok325) deletion mutants on agar plates containing exogenous tyramine. Wild-type, tyra-2, and tyra-3 mutant animals become immobilized on 30 mM tyramine within 5 min ( Figure S2). However, ser-2 mutant animals sustained movement on plates containing exogenous tyramine ( Figure 1A-C and Figure S1). Sensitivity to exogenous tyramine is restored back to wild-type levels in ser-2 mutants containing a ser-2 genomic transgene ( Figure 1B,C and Figure S2). This indicates that exogenous tyramine mediates its paralytic effects through the hyperactivation of endogenous tyramine signaling pathways that are at least in part dependent on SER-2.
We have previously shown that activation of the tyramine-gated chloride channel, LGC-55, inhibits head movements and forward locomotion [22]. Exogenous tyramine initially induced backward locomotion and the inhibition of head movements in both the wild-type and ser-2 mutants. However, unlike the wild type, ser-2 mutants recovered and resumed forward locomotion and head movements within minutes. Body movements of lgc-55 mutants are inhibited, similar to the wild-type, but head movements are sustained on exogenous tyramine ( Figure 1C,D). lgc-55; ser-2 double mutants largely persisted both head and body movements on exogenous tyramine, however locomotion remained slightly uncoordinated. This indicates that while the activation of the ionotropic LGC-55 and metabotropic SER-2 receptor are required for paralysis, exogenous tyramine may affect locomotion through the activation of other receptors.
Head movements and body movements are controlled by distinct groups of muscles and motor neurons [13,20]. In wild-type animals the inhibition of head movements occurred rapidly (half time to immobilization, t i50 = 7762 s) and was followed by the slower inhibition of body movement (t i50 = 14963 s). This suggests that signaling pathways with distinct kinetics contribute to tyramine's effects. The sustained body movements of ser-2 mutants on exogenous tyramine make it difficult to dissect the effects of exogenous tyramine on head movements. Therefore, we analyzed the effect of tyramine on head movements in an unc-3(e151) mutant background. unc-3 mutants display few body movements, but have normal head movements [29]. Head movements of ser-2 unc-3 double mutants were inhibited on exogenous tyramine, like those in the wild-type and unc-3 mutants ( Figure 1D). Consistent with our previous observation [22], head movements were sustained in lgc-55: unc-3 double mutants. The kinetics and distinct inhibition of head and body movements of lgc-55 and ser-2 mutants indicates that exogenous tyramine induces the fast immobilization of head movements mainly through the hyperactivation of the ionotropic tyramine receptor LGC-55 followed by the immobilization of body movements through hyperactivation of the metabotropic tyramine receptor SER-2.

Ga o Signaling Pathway Mutants Are Resistant to Exogenous Tyramine
To identify genes involved in tyramine signaling, we performed a genetic screen for mutants that are resistant to the immobilizing effects of exogenous tyramine on body movements. Four isolates from this screen, zf47, zf97, zf98, and zf133, sustained movement on plates containing 30 mM tyramine. Genetic mapping, complementation tests, and sequence analysis showed that zf47 was an allele of goa-1. goa-1 encodes the C. elegans ortholog of the neural G protein-alpha subunit of the Ga o class. GOA-1/Ga o is expressed throughout the nervous system and is thought to negatively regulate synaptic transmission through the inhibition of EGL-30, the C. elegans Ga q ortholog [30][31][32]. We found that zf97, zf98, and zf133 were alleles of dgk-1, which encodes the C. elegans ortholog of the vertebrate brain diacylglycerol kinase theta (DAGh) [32,33].

Author Summary
How the nervous system controls complex behaviors has intrigued neurobiologists for decades. There are many examples where sequential motor patterns of specific behaviors have been described in great detail. However, the neural mechanisms that orchestrate a full behavioral sequence are poorly understood. Gentle touch to the head of the roundworm C. elegans elicits an escape response in which the animal quickly moves backward. The reversal is followed by a deep turn that allows the animal to change its direction of locomotion and move away from the threatening stimulus. We found that the neurotransmitter tyramine controls the initial reversal phase of the escape response through the activation of a fast-acting ion channel and the later turning phase through the activation of a slow-acting G-protein coupled receptor (GPCR). We show that this tyramine GPCR is expressed in neurons that make contacts with the ventral muscles of the animal. Activation of this receptor facilitates the contraction of ventral muscles and thus allows the animal to turn and resume locomotion in the opposite direction during its escape. Our studies show how a single neurotransmitter coordinates sequential phases of a complex behavior through the activation of distinct classes of receptors.
dgk-1 mutants have a reduced ability to deplete DAG generated by the egl-30/Ga q pathway [33,34]. Previously characterized null mutants for goa-1 and dgk-1 were also resistant to the paralytic effect of tyramine on both head and body movements ( Figure 1B). We found that egl-30/Ga q gain-of-function mutants and mutants for eat-16, which encodes a Regulator of G-protein Signaling (RGS) protein that inhibits EGL-30 [35], are resistant to exogenous tyramine ( Figure 1B). Unlike goa-1 and dgk-1 mutants, which exhibit hyperactive locomotion and egg-laying behavior, ser-2 mutants did not have obvious behavioral defects [36]. Therefore, mutations that reduce GOA-1/Ga o signaling or increase EGL-30/Ga q signaling confer resistance to exogenous tyramine. The resistance of goa-1 and ser-2 mutants to exogenous tyramine suggests that SER-2 signals through GOA-1/Ga o to modulate locomotion.

SER-2 Acts in GABAergic Neurons
Where does SER-2 act to confer resistance to exogenous tyramine? A ser-2 genomic rescuing transgene restored sensitivity of ser-2 mutants to exogenous tyramine, indicating that the transgene is expressed in the cells that confer sensitivity to exogenous tyramine ( Figure 1B and Figure S2). The ser-2 genomic locus is large and complex and encodes several splice variants [25,36]. To analyze the ser-2 expression pattern we expressed mCherry under the control of an 11.8 kb ser-2 promoter sequence that included the four alternative first exons. We found that our Pser-2::mCherry reporter was expressed in head muscles, as well as several neurons in the head and ventral cord. The ser-2 reporter was expressed in the first and second row of muscle cells, which are distinct from neck muscles (third and fourth row of muscle cells) Shown is the percentage of animals immobilized after 10 min on agar plates supplemented with tyramine. Wild-type animals become fully immobilized at concentrations above 30 mM tyramine, while ser-2 mutants continue sustained movement. Each data point represents the mean 6 the standard error of the mean (SEM) for at least three trials, totaling a minimum of 30 animals. (B) G-protein signaling mutants are resistant to the paralytic of effects of exogenous tyramine. Shown is the percentage of animals that become immobilized after 10 min on 30 mM tyramine. Each bar represents the mean 6 SEM for at least four trials totaling a minimum of 40 animals. (Inset) Schematic representation of the Ga o and Ga q signaling pathways that modulate locomotion in C. elegans. The genetic data suggest that SER-2 acts in the Ga o pathway. The names of the human orthologs are shown. Rescue denotes the transgenic line Pser-2::SER-2; ser-2(pk1357). (C, D) Tyramine affects locomotion and head movements through different mechanisms. Shown is the percentage of animals with sustained body (C) or head (D) movements on 30 mM tyramine. ser-2 mutants are partially resistant to the effects of tyramine on body movements, but not head movements. lgc-55 mutants continue to move their heads through the duration of the assay. Each data point represents the mean percentage of animals that become immobilized by tyramine each minute for 20 min 6 SEM for at least six trials, totaling a minimum of 60 animals. Head movements were analyzed in an unc-3 mutant background. unc-3 mutants make few body movements but display normal head movements, and are wild-type for the ser-2 and lgc-55 loci. Statistical significance to wild-type: ***p,0.0001, two-tailed Student's t test. See also Figures S1 and S2. doi:10.1371/journal.pbio.1001529.g001 that express the tyramine-gated chloride channel LGC-55 ( Figure  S3). lgc-55 mutants fail to suppress head movements in response to touch, but head movements were suppressed normally in ser-2 mutants, indicating that only LGC-55 is required for tyramineinduced head relaxation. The ser-2 reporter was not expressed in interneurons that control locomotion, as expression did not overlap with Pglr-1::GFP and Plgc-55::GFP reporters that are expressed in the locomotion command neurons (unpublished data). ser-2 reporter expression was also observed in 13 cells in the ventral cord ( Figure 2A) that send commissures to the dorsal cord. The ventral cord is composed of excitatory cholinergic and inhibitory GABAergic motor neurons that innervate body wall muscles and control locomotion. Coexpression analysis with GFP reporters that specifically label cholinergic or GABAergic neurons showed that Pser-2::mCherry was highly expressed in a subset of GABAergic motor neurons ( Figure 2B-D). The same set of GABAergic motor neurons were labeled in transgenic line (Pser-2a::GFP) that expressed a reporter for the SER-2A isoform (unpublished data) [36]. GABAergic ventral nerve cord neurons are subdivided into 13 VD motor neurons that synapse onto the ventral body wall muscles, and 6 DD motor neurons that synapse onto the dorsal body wall muscles. The cells that highly express the ser-2 reporter do not co-label with a Pflp-13::GFP reporter, which is expressed in the DD motor neurons (unpublished data) [37]. This indicates SER-2 is specifically expressed in the GABAergic VD motor neurons.
Since GABAergic motor neurons are required for normal locomotion, we tested if SER-2 acts in these cells to mediate the inhibitory effects of tyramine on movement. No other promoters have been described that specifically drive expression in the GABAergic VD neurons. The complexity and size of the ser-2 promoter did not allow us to define promoter elements that specifically drive expression in the VD neurons. We therefore expressed ser-2 cDNA in all GABAergic neurons using the unc-47 promoter (Punc-47::SER-2). ser-2 expression in GABAergic neurons of ser-2 mutants restored normal sensitivity to exogenous tyramine ( Figure 2E).
To determine if GABA signaling affects tyramine sensitivity, we tested GABA-deficient mutants for sensitivity to exogenous tyramine. We found that unc-25 mutants, which lack glutamate decarboxylase required for GABA synthesis [38], were slightly hypersensitive to the immobilizing effects of exogenous tyramine ( Figure 2E, Figure S4). This suggests that reduced GABA signaling increases the sensitivity to exogenous tyramine, likely through the hyperactivation of other tyramine receptors, such as LGC-55 and TYRA-2. GABA deficiency suppressed the resistance phenotype of ser-2 mutants, since unc-25; ser-2 double mutants were nearly as sensitive to tyramine as unc-25 single mutants ( Figure 2E). These epistasis experiments indicate that SER-2 acts upstream or in parallel to GABAergic signaling and are consistent with the hypothesis that SER-2 acts in the VD neurons to inhibit GABA signaling and control locomotion.
We performed cell-specific rescue experiments to determine whether G-protein signaling components are required in the GABAergic neurons to mediate sensitivity to the exogenous tyramine. Expression of goa-1 or eat-16 in the GABAergic neurons (Punc-47::GOA-1/Ga o and Punc-47::EAT-16/RGS) of goa-1 and eat-16 mutants, respectively, did not rescue the hyperactive locomotion phenotype. However, goa-1 expression in GABAergic neurons of goa-1 mutants largely restored sensitivity to exogenous tyramine ( Figure 2E, Figure S4). Similarly, rescue expression of eat-16 in GABAergic neurons partly restores sensitivity to exogenous tyramine. The activation of the Ga q pathway in other cells may contribute to tyramine resistance phenotype in G-protein signaling mutants since the sensitivity to tyramine is not completely restored to wild-type levels. Nonetheless, the increased sensitivity of GABA-neuron-specific rescue of G-protein signaling mutants suggest that exogenous tyramine induces body immobilization through the activation of SER-2A and a Ga o pathway in GABAergic neurons.

SER-2 Inhibits Neurotransmitter Release
To test the hypothesis that SER-2 modulates neurotransmitter release from ventral cord motor neurons, we analyzed mutants for their sensitivity to the acetylcholinesterase inhibitor aldicarb. Aldicarb increases acetylcholine (ACh) concentration at the neuromuscular junction (NMJ), causing muscle contraction and eventual paralysis. Mutants with impaired ACh release are resistant to aldicarb-induced paralysis [39]. egl-30/Ga q mutants are resistant to aldicarb, whereas goa-1 mutants are hypersensitive to aldicarb-induced paralysis, indicating that EGL-30/Ga q stimulates and GOA-1/Ga o inhibits ACh release from motor neurons [31][32][33]. Since body wall muscles also receive inhibitory GABA inputs, hypersensitivity to aldicarb can also be caused by decrease in GABA release at the NMJ [40][41][42]. The time course of paralysis of ser-2 mutants induced by aldicarb was similar to the wild-type ( Figure 3A). This may be due to the restricted expression of ser-2 in a subset of GABAergic neurons or insufficient endogenous tyramine signaling to modulate GABA release under regular assay conditions. We therefore generated transgenic lines that overexpressed ser-2 in all cholinergic motor neurons (Pacr-2::SER-2) or GABAergic (Punc-47::SER-2) motor neurons and analyzed the rate of aldicarb-induced paralysis. In the absence of exogenous tyramine, the time course of paralysis of transgenic animals that express SER-2 in cholinergic or GABAergic ventral nerve cord motor neurons was similar to the wild-type ( Figure 3A). However, on plates that contained both aldicarb and tyramine, animals that overexpressed ser-2 in cholinergic neurons (Pacr-2::SER-2) were more resistant to the paralytic effects of aldicarb than the wild-type. Conversely, animals that overexpressed ser-2 in all GABAergic neurons (Punc-47::SER-2) were hypersensitive to paralysis on plates containing aldicarb and tyramine ( Figure 3B). These data are consistent with the hypothesis that SER-2 couples to the GOA-1/Ga io pathway to inhibit neurotransmitter release.

Tyramine-Mediated Reduction in GABA Synaptic Release Requires SER-2
The GABAergic VD motor neurons that express ser-2 make synaptic contacts onto the ventral musculature ( Figure 4A). Therefore, to directly evaluate whether tyramine modulates synaptic release of GABA from motor neurons, we measured the frequency of endogenous inhibitory post-synaptic currents (IPSCs) in whole-cell recordings from ventral body wall muscle cells ( Figure 3C-H). To isolate GABA currents, recordings were made from unc-29; acr-16 double mutants (+/+) that lack excitatory neurotransmission at the NMJ [43,44]. In unc-29; acr-16 double mutants, the only remaining currents are mediated through chloride permeation of the GABA A -like receptor UNC-49 [43,45]. In these animals, we observed high levels of endogenous IPSC activity (,13 events/s) that gradually declined over the time course of the recording period (,10 min). This basal level of inhibitory activity is consistent with previous reports that have used other, nongenetic approaches to isolate IPSCs. After recording an initial 60 s period of basal activity, we switched to a bath solution containing tyramine. Within 30 s of tyramine exposure we noted a clear decrease in IPSC frequency ( Figure 3E,G). The magnitude of this decrease was significantly greater than the slight decrease in IPSC frequency we observed  over the same time course in control experiments without tyramine (+ tyramine, 34%64% decrease; no tyramine, 13%66% decrease). These results indicate that tyramine can inhibit GABA-mediated transmission at the NMJ. The tyramine-induced reduction was not reversible within the time course of our recordings, which may suggest that tyramine is acting through a high affinity receptor. To test whether the reduction in IPSC frequency involved SER-2, we examined the effects of tyramine (Pacr-2::SER-2) or hypersensitivity (Punc-47::SER-2) (B). Aldicarb experiments were conducted on nematode growth media (NGM) plates. 30 mM tyramine dissolved in NGM agar does not inhibit locomotion. SER-2 was expressed in ser-2(pk1357) mutant animals. Each data point represents the mean percentage of animals immobilized by aldicarb scored every 30 min 6 SEM for at least four trials, totaling a minimum of 60 animals. (C, D) Representative endogenous inhibitory postsynaptic currents (IPSCs) recorded from ventral body wall muscles. +/+, unc-29; acr-16 double mutants that lack excitatory neurotransmission at the NMJ are wild-type for the ser-2 locus. ser-2, unc-29; acr-16; ser-2(pk1357) triple mutants. (E, F) Tyramine application decreased the rate of IPSCs in +/+ (n = 6), but not ser-2 mutants (n = 5). Arrow depicts tyramine application time point. Each point represents the IPSC frequency calculated over a 5-s time window as indicated. The red points correspond to the displayed samples in (C) and (D), respectively. Dashed lines show average frequency before tyramine application and during the stabilized tyramine response period. (G) Average IPSC frequency after tyramine application plotted relative to IPSC frequency prior to tyramine exposure. Values were normalized to average frequency observed in control recordings in the absence of tyramine. (H) Average amplitude of IPSCs after tyramine application plotted relative to average amplitude prior to tyramine exposure. Error bars depict SEM. Statistical differences calculated from +/+: *p,0.05, two-tailed Student's t test. doi:10.1371/journal.pbio.1001529.g003 exposure in recordings from unc-29; acr-16; ser-2 triple mutants ( Figure 3F,G). The amplitude of endogenous GABA IPSCs was not significantly different between +/+ animals and ser-2 mutants (+/+, 29.561.4 pA; ser-2, 31.963.9 pA) and was not significantly affected by tyramine exposure for either strain (+/+, 26.961.2 pA; ser-2, 30.764.4 pA), indicating that clustering and function of postsynaptic GABA receptors are normal in ser-2 mutants, and not affected by tyramine ( Figure 3H). The basal IPSC frequency prior to tyramine exposure was also not changed significantly in ser-2 mutants (+/+, 13.261.2 Hz; ser-2, 12.562 Hz). However, the tyramine-mediated reduction in IPSC frequency we observed in +/+ animals was significantly attenuated in ser-2 mutant animals (ser-2, 17%64% reduction; +/+, 35%64% reduction) such that it was indistinguishable from that observed in our control recordings without tyramine (control (no tyramine), 13%66% reduction). After normalization, tyramine application reduced IPSC frequency by 25%65% in +/+ compared to 5%65% reduction in ser-2 mutants ( Figure 3G). Taken together, our data show that tyramine inhibits GABA release onto ventral body wall muscles in a SER-2dependent manner.

Ablation of Subsets of GABAergic Motor Neurons Induces a Navigational Bias
C. elegans moves on its side by propagating a sinusoidal wave of ventral-dorsal flexures along the length of its body. GABAergic VD motor neurons synapse onto ventral muscles and receive synaptic inputs from DA/DB cholinergic motor neurons that synapse onto dorsal body wall muscles ( Figure 4A) [46]. Conversely, GABAergic DD motor neurons synapse onto dorsal muscles and receive synaptic inputs from VA/VB cholinergic motor neurons that synapse onto ventral body wall muscles. This arrangement suggests that a body bend is generated by Achmediated muscle contraction on one side and GABA-mediated relaxation on the contralateral side. This hypothesis is supported by the observation that animals in which the VD and DD neurons are killed by laser ablation move with a reduced wave amplitude [46]. The Pser-2::GFP and Pflp-13::GFP fluorescent markers, which specifically label the 13 VD and 6 DD neurons, respectively, allowed us to further test the role of the VD and DD neurons in laser ablation experiments. Animals in which only the VD neurons were ablated still propagated a sinusoidal wave along the anteriorposterior axis, but displayed deeper ventral than dorsal flexures (Movies S1 and S2). As a consequence VD ablated animals moved in ventrally directed circles (radius 2.2060.29 body lengths, n = 5, Figure 4B,C). Conversely, animals in which the DD neurons were ablated exhibited deeper dorsal than ventral flexures and moved in dorsally directed circles (radius 0.6160.11 body lengths, n = 10). GABA-deficient unc-25 mutants made shallow body bends, but showed no directional bias in their locomotion pattern. Thus, the specific ablation of the VD or DD GABAergic neurons indicate that asymmetric relaxation of either the ventral or dorsal body wall muscles results in a directional bias in locomotion.

Induction of Turning Behavior Through Optogenetic Control of GABAergic Motor Neurons
Our data suggest that the modulation of the activity of either the VD or DD motor neurons allows the animal to bend and steer in either a ventral or dorsal direction. To determine if acute activation or inhibition of a specific subclass of GABAergic neurons could induce turning behavior, we used optogenetic stimulation and inhibition. The complexity of the ser-2 promoter, which contains coding sequences and alternative start sites, did not allow us to highly express light-activated channels in the VD neurons. To test whether the differential activity of the VD and DD motor neurons can induce bending, we generated transgenic animals that co-expressed the light-activated cation channel Channelrhodopsin-2 (ChR2) [47] and light-activated chloride pump Halorhodopsin (NpHR) [48] in the six GABAergic DD motor neurons that synapse onto the dorsal muscles (Pflp-13::ChR2; Pflp-13::NpHR). ChR2 is activated by blue light and depolarizes neurons, while NpHR is activated by green light and hyperpolarizes neurons. We found that blue light activation induced a deep ventral turn ( Figure 5 and Movie S3). In contrast, green light inhibition induced a deep dorsal turn. This turning behavior was not observed in nontransgenic animals or in transgenic animals raised on plates without all-trans-retinal, the chromophore of ChR2 and NpHR. To quantify turning behavior we calculated a bending index as the fraction of animals that turned ventrally or dorsally in response to light exposure ( Figure 5A). A bending index of zero indicates no directional bias, whereas a negative or positive fraction indicates a ventral or dorsal bias, respectively. The bending index of Pflp-13::ChR2/ NpHR transgenic animals was 20.4560.04 with exposure to blue light and 0.4160.06 for animals with exposure to green light. The flp-13 promoter also drives expression in a small set of head neurons in addition to the DD neurons. To determine if the activation or inhibition of the DD neurons was sufficient to induce bending, we used an optogenetic illumination system capable of tracking and stimulating individual regions of a freely moving animal [49]. Targeted illumination of the animal's ventral nerve cord that harbors the DD neuronal cell bodies induced a tight ventral bend in response to blue light activation and a tight dorsal bend in response to green light inhibition ( Figure 5B and Movie S3). Switching between blue and green light exposure enabled the remote control of ventral and dorsal turning behavior in freely moving Pflp-13::ChR2/NpHR transgenic animals. Thus, the acute stimulation of GABA release on the dorsal side induced relaxation of the dorsal muscles resulting in a ventral turn. Conversely, the acute inhibition of GABA release onto the dorsal muscles resulted in hypercontraction of dorsal muscles and a dorsal bend. Our data indicate that modulating the activity of subsets of GABAergic neurons synapsing onto either the dorsal (DD) or ventral side (VD) of the animal can induce navigational bias.

ser-2 Facilitates the Execution of Omega Turns in the Escape Response
The expression of ser-2 in the GABAergic VD neurons suggested a possible role in C. elegans navigation. However, ser-2 mutants moved normally and did not display a ventral or dorsal directional bias ( Figure 4B) possibly through a lack of basal tyramine to activate SER-2 during regular locomotion. We have previously shown that tyramine coordinates the timing of backing locomotion and the suppression of head movements in the escape response [13]. Gentle anterior touch triggers tyramine release from the RIM neurons and the synaptic activation of the tyramine-gated chloride channel, LGC-55 [22]. Is SER-2 also required in the execution of the escape response? Touch-induced reversals are often coupled to a sharp omega turn, which allows the animal to change locomotion in a direction opposite to its original course (Figures 6E and 7A) [18]. The omega turn is initiated by a steep ventral bend of the head when the animal reinitiates forward locomotion. While the sharp bend is propagated posteriorly along the body, the head usually slides along the ventral side of the body. We analyzed turning behavior in response to gentle anterior touch. We found that the likelihood of engaging in an omega turn (.90u turn initiated by the first forward head  . ser-2 mutants [ser-2(ok2103), n = 52; ser-2(pk1357), n = 6 2] touch nose to tail less frequently than wild-type (n = 51) in omega turns induced by both touch (Movies S4 and S5) and blue light in a Pmec-4::ChR2 background [Pmec-4::ChR2, n = 38; ser-2(ok2103); Pmec-4::ChR2, n = 43; ser-2(pk1357); Pmec-4::ChR2, n = 28]. Tyramine/octopamine-deficient tdc-1 mutants touch nose to tail less frequently than wild-type [tdc-1(n3420), n = 144], while swing) was correlated with the length of the reversal ( Figure 6A). Short reversals most often resulted in shallow head bends and modest deflections from the original trajectory. In contrast, escape responses that included reversals of four or more body bends most often ended in an omega turn. These results are consistent with previous studies of reversals where omega turns tend to occur after a long reversal [50][51][52].
Tyramine-deficient tdc-1 mutants or RIM ablated fail to suppress head movements, make short reversals, and less frequently engage in the execution of omega turns [13,51]. In contrast, in response to anterior touch, ser-2 mutants suppressed head movements and reversed similar to wild-type animals ( Figure  S6). Furthermore, ser-2 mutants initiated an omega turn with a steep ventral head bend at the same frequency as the wild-type ( Figure 6A). However, once ser-2 mutants initiated the omega turn, the ventral turn was less deep than the wild-type. Whereas most wild-type animals' heads touched the ventral side of the body (86%63.6%, n = 51) during an omega turn, ser-2 mutant animals failed to fully close the omega turn (27%67.8%, n = 62) ( Figure 6B,C and Movies S4 and S5). Like the ser-2 mutants, tdc-1 mutants that are unable to synthesize tyramine and octopamine [13] failed to fully close the omega turn during the escape response (tdc-1 32%65.5%, n = 144). In contrast, tbh-1 mutants, which only lack octopamine, executed closed omega turns (tbh-1 72%67.9%, n = 153), comparable to the wild-type. We measured the angle of the omega turn (omega angle) from the deepest most contracted region of the body to the closest or touching points in the head and tail ( Figure 6B,D). Wild-type animals typically fully closed their omega bend, while ser-2 mutants often failed to close omega bends, averaging an omega angle of 24u62.3u (n = 62). Animals lacking tyramine and octopamine (tdc-1) averaged an omega angle of 25u63.6u, while animals lacking octopamine (tbh-1) alone closed their omega turn. Since touch stimuli in these assays have some inherent variability, we also induced reversals by optogenetic activation of the touch sensory neurons. Light induces an escape response in Pmec-4::ChR2 transgenic animals that express the ChR2 in the touch sensory neurons [47,49,53]. We analyzed omega turns of Pmec-4::ChR2 transgenic animals in response to blue light in both wild-type and ser-2 mutant backgrounds. The light-induced escape response of ser-2 mutants showed a similar defect in omega turns as with reversals induced by touch and had a lower frequency of closed omega turns than the wild-type ( Figure 6C,D). The turning defects of ser-2 mutants caused an alteration in the direction of reinitiated forward movement, or escape angle. In response to touch, wild-type animals and tbh-1 mutants completely reversed their direction of locomotion with an escape angle of 179u65u (n = 42) and 177u611u (n = 16), respectively. In contrast, in response to touch, ser-2 mutants and tdc-1 mutants made a more shallow escape angle, changing their direction from the point of stimulus by 150u65u (n = 46) and 143u66u (n = 35) ( Figure 6E,F). Genomic rescue lines partially restore the omega angle defect of the mutants and restored the escape angle to wild-type levels. Our data indicate that tyraminergic activation of SER-2 facilitates the execution of a tight ventral bend in the escape.

Discussion
The concept of monoaminergic coding of behaviors originated from work in crustaceans and insects where fully coordinated behavioral sequences can be elicited by the injections of specific monoamines into the nervous system [6,16,17]. However, understanding how sensory inputs recruit the action of monoamines and how changes in circuit properties affect behavior remains a tremendous challenge in the mammalian and even insect nervous systems. Tyramine release from a single pair of neurons, acting on few tyramine receptors in the C. elegans nervous system with single-cell resolution, provides unique insights into how monoamines orchestrate independent motor programs in a complex behavior. In this study we analyzed how the G-protein coupled receptor SER-2 modulates the output of a neural circuit in a compound motor sequence.

Tyramine Inhibits GABA Release
Our genetic data suggest that the tyramine receptor SER-2 acts in a Ga o (GOA-1) pathway. The Ga o /GOA-1 is expressed in all C. elegans neurons where it antagonizes Ga q /EGL-30 function in many behaviors including locomotion, egg laying, and pharyngeal pumping [54,55]. Dopaminergic and serotonergic G-protein coupled receptors are expressed in C. elegans ventral cord neurons and their hyperactivation with exogenous dopamine or serotonin can induce paralysis within minutes [56][57][58][59]. However, ser-2 and other individual biogenic amine receptor mutants have no obvious locomotion defects. This indicates that Ga o /GOA-1 and Ga q / EGL-30 integrate monaminergic signals to modulate neurotransmitter release from the motor neurons to control locomotion ( Figure 1B). Dopamine, serotonin, and tyramine may allow the animal to refine locomotory patterns during different behavioral states. Ga o /GOA-1 activity is thought to reduce the abundance of the synaptic priming protein UNC-13 at the synapse in C. elegans ventral cord neurons [33]. Our data indicate that tyramine reduces GABA release from VD motor neurons in a SER-2dependent manner to augment the reorientation component of the escape response. In this regard, SER-2 shares similarities with mammalian alpha(2)-adrenergic receptors that inhibit neurotransmitter release and cause vasoconstriction during a fight-or-flight response [4].

Tyramine Acts Through Synaptic Activation of Ionotropic Receptors and Extrasynaptic Activation of Metabotropic Receptors
Tyramine can act as a classical neurotransmitter in C. elegans through the synaptic activation of the tyramine-gated chloride channel, LGC-55 [22,60]. While LGC-55 is predominantly expressed in cells that are directly synaptic to tyraminergic RIM neurons, SER-2-expressing cells do not receive direct RIM innervation, indicating that SER-2 activation occurs extrasynaptically [28,36]. In addition, ionotropic and metabotropic receptors have distinct ligand affinity and signaling kinetics. Ligand-gated ion channels like LGC-55 have a relatively low affinity for their ligand (Kd 0.1 to 1 mM) and affect postsynaptic potentials within milliseconds. This allows for fast localized signaling between neurons and their postsynaptic partners. In contrast, G-protein coupled receptors have a high affinity for their ligand (Kd 0.1-1 mM) and operate on timescales from seconds to minutes [61]. Synaptic spillover from the synaptic cleft and diffusion can activate these high affinity receptors that are distant from the release site. As tyramine is released from a single pair of head neurons that extend processes into the nerve ring, the activation of SER-2 in the GABAergic VD neurons depends upon the diffusion of tyramine through the pseudocoelomic space to reach the VD processes along the length of the body. C. elegans has two other G-protein coupled receptors, TYRA-2 and TYRA-3, in addition to SER-2 that bind tyramine with high affinity [26,27]. tyra-2 and tyra-3 are not expressed in cells that receive direct synaptic inputs from the tyraminergic RIM neurons [62]. Furthermore, no tyramine or octopamine reuptake transporter has been identified in either C. elegans or Drosophila, which suggests that tyramine diffusion from the synaptic cleft is part of its mechanism of action. Serotonin, dopamine, and octopamine receptors are also expressed in many C. elegans cells that are not directly postsynaptic to the small number of monoaminergic neurons that release them [36,58]. Similarly, in humans the monoaminergic cells are grouped in relative small nuclei that can affect large areas of the CNS or the periphery that do not receive direct synaptic inputs. Thus, monoamines are not confined to the anatomical connectome and can reconfigure outputs to large neuronal ensembles [63].

Tyramine Temporally Coordinates Independent Motor Programs in a Complex Behavior
The translation of sensory input into goal-directed behaviors requires the temporal coordination of independent motor programs. In response to gentle touch to the head, C. elegans engages in a compound motor sequence that allows the animal to retreat and navigate away from a touch stimulus ( Figure 7A). In the base state, (I) forward locomotion is accompanied by exploratory head movements in which the tip of the nose moves from side to side. Gentle touch to the head of C. elegans elicits an escape response where the animal reverses its direction of locomotion while suppressing its exploratory head movements (II). Classic laser ablation experiments by Chalfie et al. (1985) [19] complemented by the known neural wiring diagram support a model in which the touch sensory neurons (ALM/AVM) inhibit the forward locomotion ''command'' neurons (PVC/AVB) and activate the backward locomotion ''command'' neurons (AVD/ AVA), causing the animal to move backward away from the stimulus ( Figure 7B). The AVA backward locomotion ''command'' neurons activate the electrically coupled tyraminergic RIM motor neurons, which exhibit coactivated calcium transients with the AVA [64,65]. Tyramine release from the RIM induces the suppression of head movements through the activation of the fastacting inhibitory chloride channel, LGC-55, in neck muscles and cholinergic motor neurons that control head movements [13,22]. Activation of LGC-55 in the AVB forward locomotion ''command'' neurons further inhibits forward locomotion, stimulating long reversals that are coupled to the initiation of an omega turn [50,51]. The reversal is followed by a deep ventral head bend (III), allowing the animal to make a sharp (omega) turn (IV) where the head of the animal slides along the ventral side of the body and resumes forward locomotion (I). Our data show that extrasynaptic activation of the G-protein coupled tyramine receptor SER-2 contributes to the proper execution of the omega turn during the last stage of the escape response. The RIV and SMD head motor neurons have been implicated in the ventral head bend that initiates the omega turn [51]. ser-2 mutants properly initiate the turn but bend less deep and the head often fails to touch the tail during the omega turn (IV).
C. elegans locomotion is controlled by cholinergic (A and B) and GABAergic (D) motor neurons that propel a wave of ventraldorsal muscle flexures along the length of the body. We show that SER-2 inhibits neurotransmitter release from the GABAergic VD neurons. VD motor neurons release GABA on ventral body wall muscles and receive inputs from cholinergic DB neurons that synapse onto dorsal muscles. The DD motor neurons receive input from cholinergic VB motor neurons and release GABA on dorsal body wall muscles ( Figure 4A) [20]. This organization indicates that GABAergic motor neurons are essential for contralateral muscle relaxation, thus promoting local bending during wave propagation [46].
While spatial orientation in response to sensory stimuli is a fundamental component of animal behavior, the neural correlates are poorly understood. To change its direction of movement and navigate its environment, an animal needs to generate asymmetry in its locomotion pattern [66,67]. We show that the ablation of GABAergic VD neurons induces ventrally directed movement likely through the failure to properly relax the ventral muscles. Conversely, ablation of GABAergic DD neurons induces dorsally directed movement. Our experiments provide functional evidence that VD and DD neurons inhibit ventral and dorsal bending. Strikingly, the acute inhibition or activation of the DD motor neurons induces a sharp dorsal or ventral turn, respectively. Optogenetic neuronal stimulation can elicit distinct motor responses in multiple systems [47,[68][69][70]. The optogenetic control of a single class of GABAergic neurons induces bi-directional steering and the remote control of animal navigation.
The ventral omega turn in the escape response occurs several seconds after the touch-induced reversal, which is within the time scale that GPCRs modulate neuronal activity. To make a tight turn the animal needs to generate an asymmetry in the excitation and inhibition of the ventral and dorsal muscle quadrants. Our data indicate that SER-2-mediated inhibition of GABAergic release onto the ventral muscles stimulates ventral muscle contraction and the proper propagation of the ventral body bend during the omega turn ( Figure 7B). The kinetics of synaptic activation of the ion channel LGC-55 and extrasynaptic activation of the G-protein coupled receptor SER-2 temporally coordinates different phases of the escape response. Interestingly, the tyramine GPCR TYRA-3 modulates responses to pain-like stimuli and decision making [27,62]. This indicates that tyramine, much like the mammalian adrenergic signaling, coordinates different aspects of a flight response.
Our results provide molecular and neural insights on how monoamines reconfigure the output of neural circuits and orchestrate complex behaviors. Many neurotransmitters, including serotonin, GABA, acetylcholine, and glutamate, act through both ion channels and GPCRs in animals ranging from worms to man. We hypothesize that the temporally coordinated activation of ionotropic and metabotropic receptors may be a common signaling motif employed across organisms to orchestrate behavioral responses. Furthermore, the different monoamines appear to play a strikingly conserved role in the specification of distinct internal states, suggesting that the principles of neuronal modulation are conserved across species.

Behavioral Assays
Behavioral assays were performed at room temperature. Drug assays were conducted on young adult animals aged 24 h post-L4 larval stage. Locomotion assays were performed on agar plates containing 2 mM acetic acid with or without tyramine hydrochloride (Sigma-Aldrich). Approximately 10 animals were transferred to assay plates and scored for locomotion every minute over a 20-minute period. Animals were scored as immobilized if there was no sustained forward or backward locomotion in a 5-s interval. Aldicarb drug assays were performed using NGM agar plates supplemented with 0.5 mM aldicarb (Sigma-Aldrich) with or without 30 mM tyramine. Locomotion is not obviously affected on plates with 30 mM tyramine dissolved in NGM agar instead of the agar used in exogenous tyramine paralysis assays. Animals were scored as paralyzed when they did not move when prodded with a platinum wire.
Optogenetic blue and green light-induced bending assays were performed with Pflp-13::ChR2::GFP; Pflp-13::NpHR::CFP; lite-1(ce314) transgenic animals. Bending behavior movies, worm tracking traces, and kymographs were generated using the CoLBeRT worm tracking system as previously described [49]. Animals were placed in between two glass slides in 200 ml of NGM containing 30% dextran. The space between the slides was approximately 0.127 mm and limited locomotion to two dimensions. The ventral nerve cord was illuminated with blue or green light using the micromirror control of the CoLBeRT system, and the behavior was analyzed using custom tracking software written in MATLAB. For assays used to calculate a bending index, young adult animals were transferred to NGM plates without food for 45 min. Food deprivation stimulated long forward runs, which facilitated the analysis of light-induced bending. Animals were exposed to blue or green light, using GFP (525 nm) and Rhodamine (550 nm) filters for 3 s. A dorsal or ventral bend were scored if the bend was larger than 45u. The bending index was calculated as the fraction of dorsally bending worms minus the fraction of ventrally bending worms. A positive fraction indicates a dorsal bias, a negative fraction indicates a ventral bias, and a zero value represents no directional bias or no response to light exposures.
Omega turns were analyzed on NGM agar plates 2 d postpouring to control for assay plate humidity. Assay plates (60 mm diameter) were seeded with 40 ml OP50 E. coli and grown overnight at 37uC to produce a thin bacterial lawn. Young adult worms were transferred to an omega assay plate and allowed to acclimate for at least 10 min. Omega turns were induced by gentle anterior touch with fine eyebrow hair while recording using a FireWire camera and Astro IIDC software. For Pmec-4::ChR2::YFP-induced omega turns, animals were exposed to blue light for 5 s to induce a reversal. Bending angles and locomotion trajectories were calculated using Image J software analysis of movie stills. An omega turn was classified as a sharp turn larger than 90u from the initial trajectory, following a reversal of three or more body bends. The omega angle was measured using the deepest part of the bend as the apex with vectors extending to the closest points along the body. Angles larger than 60u were not scored. The escape angle was measured as the angle between the reversal trajectory and the trajectory of reinitiation of forward locomotion after the omega turn.

Laser Ablations
Animals were mounted on agar pads and anesthetized with 20 mM sodium azide. Laser ablations were done using standard methods [74]. DD and VD motor neurons were identified in Pflp-13::GFP animals in the L2 larval stage and Pser-2::GFP animals in the L3-L4 larval stage, respectively. Following a recovery period of 1 to 3 d postablation, locomotion and omega turn assays were conducted on young adult animals. Locomotion patterns of animals that exhibited coordinated long runs were recorded at 7.5 fps. Movie analysis was done using MATLAB and the MATLAB Image Acquisition Toolbox [75]. To determine directionality for each locomotion trace, the slope of instantaneous direction over time was measured for individual 360 degree turning events. Laser ablation of motor neurons was confirmed by lack of GFP expression in the cell body positions in adult animals following behavioral experiments.

Electrophysiology
Endogenous postsynaptic currents were recorded from body wall muscles as previously described [43]. All electrophysiology experiments were carried out at room temperature. Adult animals held at drop of bath solution were glued down to the sylgard coated glass coverslip with cyanoacrylate tissue adhesive (Skinstitch Corp.) applied along the dorsal side of the body. A longitudinal incision was made by sharp glass electrode tip in the dorsolateral area, the intestine and gonad were removed, and the cuticle flap along the incision was glued down in order to expose the ventral medial body wall muscles along the ventral nerve cord. The preparation was then washed briefly for ,20 s with a solution of collagenase type IV from Clostridium hystolyticum (Sigma-Aldrich) in extracellular bath solution (at a concentration of 1 mg/ ml) in order to remove the basement membrane overlying the muscles.
The extracellular solution consisted of 150 mM NaCl, 5 mM KCl, 4 mM MgCl 2 , 1 mM CaCl 2 , 15 mM HEPES, and 10 mM glucose (pH 7.4, osmolarity adjusted with 20 mM sucrose). The intracellular fluid (ICF) consisted of 25 mM K-gluconate, 115 mM KCl, 0.1 mM CaCl 2 , 50 mM HEPES, 5 mM Mg-ATP, 0.5 mM Na-GTP, 0.5 mM cGMP, 0.5 mM cAMP, and 1 mM BAPTA (pH 7.4, osmolarity adjusted with 10 mM sucrose). Whole-cell voltage clamp recordings from C. elegans body wall muscle cells (group of ventral medial muscles 9, 11, 13) from +/+ and ser-2 mutant strains were performed as previously described [43] using an EPC-10 amplifier (HEKA). Data acquisition and voltage protocols were controlled by HEKA Patchmaster software. Patchclamp electrodes were pulled from borosilicate glass with filament (Sutter Instrument), fire-polished to a resistance of 4-6 MV, and filled with an internal solution. 1 M KCl electrode with agarose bridge served as reference electrode. Leak currents and liquid junction potentials were not compensated, though pipette offset was zeroed immediately before getting a gigaohm seal. The membrane potential was clamped at 260 mV. Data were digitized at 6.67 kHz and low pass filtered at 3.3 kHz. Membrane capacitance and the series resistance (at least 20%, up to 60%) were compensated, and only recordings in which the series resistance was stable throughout the course of the recording were included. Endogenous synaptic activity data were collected in continuous mode (saved as 30-s recording sweeps). Typically, 60-90 s of control endogenous activity was recorded followed by a 90s bath tyramine application (100 mM), and subsequent tyramine wash-out (occasional). The muscle preparation was continuously perfused by extracellular solution with or without tyramine by gravity-flow at a perfusion rate of 2 ml/min. Cells were excluded from analysis if a leak current .300 pA was observed. Only recordings with series resistance R s ,15 MV were included in the analysis. Data analysis and graphing were performed using Excel (Microsoft), Igor Pro (WaveMetrics Inc.), and GraphPrism (GraphPad Software). Mini Analysis software (Synaptosoft Inc.) was used to detect and analyze the endogenous events off-line. Parameters for detection of events were set as follows: amplitude threshold 12 pA, period to search local minimum 80 ms, time before a peak for baseline 20 ms, period to search a decay time 30 ms, fraction of peak to find a decay time 0.37, period to average a baseline 10 ms, area threshold 10, and number of points to average for peak 3. In addition, following automatic detection of endogenous postsynaptic currents, traces were visually inspected, and all events were manually verified and accepted/rejected or recalculated, if necessary. Sixty seconds of continuous data were used in analysis. Segments of 60-s recording prior to tyramine application served as a control. The tyramine effect was evaluated in the last 60 s of a 90-s perfusion-that is, initial 30 s of recording with tyramine in bath solution were eliminated from analysis in order to allow enough time for adequate tyramine perfusion and stabilization effect. Data from each group were averaged, and statistical significance between the strains was determined by two-tailed unpaired Student's t test. All chemicals were purchased from Sigma-Aldrich. Figure S1 C. elegans become immobilized on exogenous tyramine in a dose-dependent manner. (A) Wild-type animals become immobilized within 5 min on 30 mM tyramine (also see Pirri et al., 2009 [22]). (B) ser-2 mutants are resistant to body immobilization compared to wild-type, but become immobilized by 60 mM tyramine. Each data point represents the mean 6 SEM for at least four trials totaling a minimum of 40 animals. (TIF) Figure S2 tyra-2 and tyra-3 mutant animals paralyze on exogenous tyramine. (A) tyra-2 and tyra-3 mutants become immobilized on plates containing 30 mM exogenous tyramine similar to wild-type. (B) Two additional SER-2 rescue strains (10 ng/ml injection) also rescue the immobilization resistance phenotype of ser-2 mutants. Rescue denotes the transgenic line Pser-2::SER-2; ser-2(pk1357). Each data point represents the mean 6 SEM for at least five trials totaling a minimum of 50 animals. Movie S3 Tracking of a [Pflp-13::ChR2; Pflp-13::NpHR] single worm's locomotion and spatial illumination of the ventral nerve cord using the CoLBeRT system [49]. Inhibition of the DD motor neurons with green light activation of NpHR causes a dorsal bend. Activation of the DD motor neurons with blue light activation of ChR2 causes a ventral bend. (MOV)

Supporting Information
Movie S4 A wild-type animal executing a complete omega turn in response to an anterior touch. Following a long reversal, the animal makes a deep ventral turn where the head of the animal touches and slides along the ventral side of the body and resumes forward locomotion in the direction opposite its original trajectory. (MOV) Movie S5 A ser-2 mutant animal does not execute a complete omega turn in response to an anterior touch. Following a long reversal, the animal's head fails to touch the ventral side of the body. (MOV)