Dissection of neuronal gap junction circuits that regulate social behavior in Caenorhabditis elegans

Significance

Electrical coupling through gap junctions is an important form of neuronal communication, but methods to interrogate electrical coupling are limited. We identified the UNC-9 gap junction protein (innexin) as an essential component of a circuit that regulates social aggregation and related behaviors in Caenorhabditis elegans. By transgenic expression of dominant-negative UNC-1 stomatin and rescue with gain-of-function UNC-9::GFP, we selectively interrogated and restored gap junction coupling between neurons, allowing fine dissection of the neuronal circuit for social behavior.

Abstract

A hub-and-spoke circuit of neurons connected by gap junctions controls aggregation behavior and related behavioral responses to oxygen, pheromones, and food in Caenorhabditis elegans. The molecular composition of the gap junctions connecting RMG hub neurons with sensory spoke neurons is unknown. We show here that the innexin gene unc-9 is required in RMG hub neurons to drive aggregation and related behaviors, indicating that UNC-9–containing gap junctions mediate RMG signaling. To dissect the circuit in detail, we developed methods to inhibit unc-9–based gap junctions with dominant-negative unc-1 transgenes. unc-1(dn) alters a stomatin-like protein that regulates unc-9 electrical signaling; its disruptive effects can be rescued by a constitutively active UNC-9::GFP protein, demonstrating specificity. Expression of unc-1(dn) in RMG hub neurons, ADL or ASK pheromone-sensing neurons, or URX oxygen-sensing neurons disrupts specific elements of aggregation-related behaviors. In ADL, unc-1(dn) has effects opposite to those of tetanus toxin light chain, separating the roles of ADL electrical and chemical synapses. These results reveal roles of gap junctions in a complex behavior at cellular resolution and provide a tool for similar exploration of other gap junction circuits.

Electrical synapses and chemical synapses have complementary functions in the nervous system. Separate regulation of electrical and chemical synapses creates multifunctionality in neural circuits, as illustrated in the vertebrate retina, where regulation of electrical synapses rewires information flow at different light levels (1, 2). Electrical synapses also have essential roles in the developing brain, where they regulate the formation of chemical synapses (3–5). In humans, mutations that affect electrical synapses are associated with neurological conditions such as epilepsy, deafness, and peripheral neuropathies (6–8). However, it is difficult to study electrical synapses in detail because there are few simple functional assays or pharmacological tools that selectively target their activity.

The structural components of electrical synapses are intercellular gap junctions consisting of two aligned hemichannels, each composed of six connexin subunits in vertebrates or eight innexin subunits in invertebrates (9–11). The apposition and opening of these hemichannels form a small protein-lined pore that allows bidirectional ionic flow and electrical coupling between two cells (12). Although the ions always flow down their gradients, the opening of gap junctions can be regulated by membrane potential, transjunctional potential, and protein phosphorylation (12). Gap junctions also allow bidirectional flow of ATP and other intracellular metabolites and function in nonexcitable cells as well as neurons and muscle (12).

The nematode Caenorhabditis elegans has gap junctions in neurons and in nonneuronal tissues. Its pharyngeal muscles coordinate their contraction through gap junctions; like the vertebrate heart, its body wall muscles use gap junctions to couple their excitability; and its intestinal cells propagate calcium waves through gap junctions to set the digestive rhythm (13–15). In addition, the anatomical wiring diagram of the C. elegans nervous system derived from electron micrographs has ∼600 gap junctions among its 302 neurons, representing about 10% of all synapses (16–19). Circuit manipulations and genetic studies indicate that gap junctions are essential in locomotion circuits and escape behaviors (20, 21). For example, the broadly expressed innexin subunits unc-7 and unc-9 are required for coordinated locomotion and link command interneurons with motor neurons to balance forward and backward movement states (21, 22), and unc-7 also mediates electrical coupling between neurons in the gap junction circuit for nose-touch avoidance behavior (23). However, most neuronal gap junctions are solely predicted by electron micrographs and have not been linked to particular genes or functions.

In C. elegans, a network of neurons connected by gap junctions regulates aggregation and a suite of related behaviors. C. elegans aggregates into social groups or “clumps” under stressed or crowded conditions (24, 25). This behavior is prominent in wild C. elegans strains, which aggregate in response to the high ambient oxygen levels that they encounter in the laboratory (24, 26, 27). Early in its laboratory cultivation, the N2 strain of C. elegans acquired a gain-of-function mutation in the neuropeptide receptor NPR-1 that reconciled it to high oxygen levels (27). The npr-1(gf) mutation suppresses aggregation and oxygen avoidance, remodels pheromone responses, decreases locomotion speed on food, and results in a selective advantage in laboratory growth (26, 28–33). These behaviors are controlled by a hub neuron called RMG, which has gap junctions with sensory neurons that detect ambient oxygen, pheromones, and other environmental cues (28). In wild strains or npr-1(lf) mutants, activity of the RMG circuit promotes aggregation and related responses to oxygen, pheromones, and food. In N2 animals, high npr-1(gf) activity suppresses the RMG circuit and related behaviors.

The gap junctions are attractive candidates for mediating cell communication in the RMG circuit, but their significance is inferred indirectly from their anatomical presence in electron micrographs. Alternative models for RMG signaling are possible, as all neurons in the circuit also have chemical synapses, and npr-1 also affects gene expression and neuropeptide signaling (32–34). Here we identify the unc-9 innexin gene as an essential component of the RMG circuit. Using cell-specific manipulation of unc-9 activity with Cre recombinase and a newly developed dominant-negative unc-1(dn) reagent, we define the roles of gap junctions in specific neurons in generating aggregation and related behaviors. These results demonstrate both localized and distributed roles of neuronal gap junctions in mediating complex behaviors.

Results

unc-9–Based Gap Junctions Are Required for Aggregation Behavior.

The anatomical circuit diagram of C. elegans suggests that RMG is the hub of a gap junction circuit composed of unidentified molecular components, probably innexins (Fig. 1A) (16–19, 28). We hypothesized that mutations affecting gap junctions in the RMG circuit would suppress the aggregation phenotype of npr-1(lf) mutants. Double mutants were generated between npr-1(lf) and 21 of the 25 predicted innexin genes, excluding the innexin genes that are inviable as null mutants, and aggregation behavior was scored in the double-mutant strains (Table S1). This screen identified unc-9 and unc-7 as strong suppressors and inx-4, inx-6, inx-17, and inx-22 as weak suppressors of the npr-1 aggregation phenotype. An independent study of a different npr-1(lf) phenotype, regulated expression of the srh-234 chemoreceptor gene, also demonstrated a requirement for unc-9 and unc-7 (34).

Fig. 1.

unc-9 gap junctions are required for aggregation in npr-1 mutants. (A, Left) The RMG gap junction circuit for aggregation behavior (16, 17, 19, 28). RMG forms anatomically defined gap junctions with multiple sensory neurons (shown) and some interneurons and motor neurons. (Right) Expression patterns of npr-1 and flp-21 promoters used in C–E; complete expression patterns are in Table S3. (B) Cre-Lox reagents for cell-specific knockout of unc-9 in Cre-Lox strains using ncs-1, flp-5 (“RMG”), and flp-8 (“URX”) promoters. The flp-5 promoter drives expression in RMG, ASE, PVT, I4, and M4 neurons and pharyngeal muscles, and the flp-8 promoter drives expression in URX, AUA, and PVM neurons (see also Table S3). (C) Swimming behavior. Each dot represents the number of body bends per minute measured from a single animal; bars represent mean and SEM. n = 5–18 animals per genotype. (D) Aggregation behavior. Boxes represent median and first and third quartiles, and whiskers represent 5th–95th percentiles. n = 3–18 assays per genotype. (E) Locomotion speed on a bacterial lawn. Bars represent mean ± SEM. n = 2–6 assays per genotype. For all panels, *P < 0.05, **P < 0.01, ***P < 0.001, ns (nonsignificant) by one-way ANOVA with Dunnett’s correction for multiple comparisons.

Table S1.

Suppression of npr-1 behaviors by innexin mutants

Genotype	Aggregation	Bordering
N2	–	–
npr-1(ad609)	++	++
inx-1(tm3524) npr-1(ad609)	++	+
inx-2(ok376) npr-1(ad609)	++	++
inx-4(ok2373); npr-1(ad609)	+	++
inx-6(rr5); npr-1(ad609)	+	+
inx-7(tm2738); npr-1(ad609)	++	++
inx-8(gk42); npr-1(ad609)	++	++
inx-9(ok1502); npr-1(ad609)	++	++
inx-10(tm3393); npr-1(ad609)	++	++
inx-11(ok2783); npr-1(ad609)	++	+
inx-14(ag17); npr-1(ad609)	++	++
inx-15(tm3394); npr-1(ad609)	++	++
inx-16(tm1589); npr-1(ad609)	++	+
inx-17(tm3829); npr-1(ad609)	+	+
inx-18(ok2454); npr-1(ad609)	++	++
inx-19(tm1896); npr-1(ad609)	++	++
inx-20(ok681); npr-1(ad609)	++	++
inx-21(tm4591); npr-1(ad609)	++	++
inx-22(tm1661); npr-1(ad609)	+	++
unc-7(e5) npr-1(ad609)	–	–
unc-9(e101) npr-1(ad609)	–	–
eat-5(ad464); npr-1(ad609)	++	++

– represents aggregation or bordering indistinguishable from N2 and significantly different from npr-1. + represents intermediate aggregation or bordering significantly different from both N2 and npr-1. ++ represents aggregation or bordering indistinguishable from npr-1 single mutants (one-way ANOVA with Dunnett’s correction for multiple comparisons or student’s t test with Bonferroni’s correction for multiple comparisons).

Table S3.

Complete list of promoters and their neuronal expression

Promoter	Neuronal expression
Pncs-1	RMG AWB ADL AVE ASI AWC ASG ASE BAG AFD AIY AVK; variable unidentified head and tail neurons
Pflp-21	RMG URX ADL ASK ASH ASJ ADF M2 ASI ASG (FLP) (URA); variable unidentified tail neurons
Pnpr-1	RMG URX ASG ASE AUA RIG AQR SMBD; variable unidentified head and tail neurons
Pflp-5	RMG ASE PVT I4 M4 neurons, and pharyngeal muscle
Pflp-8	URX AUA PVM
Psre-1	ADL
Psrh-220	ADL
Psra-9	ASK
Psra-6	ASH ASI PVQ

Both unc-9 and unc-7 are expressed broadly in the nervous system, and null mutants in these genes have strong locomotion defects; unc-9 is also expressed in body wall muscles (35–37). To bypass indirect effects of locomotion on aggregation and narrow down sites of unc-9 action, we targeted unc-9 using a cell-specific Cre-Lox strategy (Fig. 1B). A transgene with an unc-9 cDNA flanked by two loxP sites under the unc-9 promoter was introduced into unc-9 npr-1 double mutants. This transgene fully rescued locomotion speed in a swimming assay (Fig. 1C, in green) and also rescued aggregation behavior (Fig. 1D, in green; controls in Fig. S1). Therefore, the transgene is expressed at the sites essential for locomotion and aggregation. Expression of Cre recombinase under cell-specific promoters was then used to eliminate unc-9 activity in targeted cells (Materials and Methods). Pan-neuronal deletion of unc-9 with a tag-168::Cre transgene resulted in locomotion defects and a loss of aggregation, suggesting that the observed unc-9 effects are mediated by neurons (Fig. 1 C and D). Deletion of unc-9 in several neurons of the RMG circuit with a ncs-1::Cre transgene suppressed aggregation but allowed normal locomotion (Fig. 1 C and D). Suppression of aggregation was also observed after selectively targeting RMG with a flp-5::Cre transgene (RMG::Cre; Fig. 1D) but not after targeting URX with a flp-8::Cre transgene (URX::Cre; Fig. 1D). Other than RMG, the only site of overlap of ncs-1 and flp-5 is the ASE sensory neuron, which is not predicted to form gap junctions in the wiring diagram (16, 19) (Fig. 1 A and B). Although the Cre recombinase might be expressed in additional neurons that we did not detect, these results strongly suggest that RMG-specific deletion of unc-9 gap junctions suppresses aggregation behavior while preserving locomotion.

Fig. S1.

Cre recombinase does not affect behaviors of unc-9 npr-1 mutants. (A) Swimming behavior. Each dot represents the number of body bends per minute measured from a single animal; bars represent mean ± SEM. n = 5–18 animals per genotype. (B) Aggregation behavior. Boxes represent median and first and third quartiles, and whiskers represent 5th–95th percentiles. n = 3–18 assays per genotype. (C) Locomotion speed on a bacterial lawn. Bars represent mean ± SEM. n = 2–6 assays per genotype. For all panels, ns (nonsignificant) by one-way ANOVA with Dunnett’s correction for multiple comparisons.

npr-1 animals have a high crawling speed on a bacterial lawn compared with N2 animals, a foraging behavior that is regulated by the RMG circuit (28). Foraging speed was intermediate between N2 and npr-1 when floxed unc-9 was deleted from RMG (Fig. 1E). This result suggests that unc-9 acts in RMG as well as other neurons to regulate locomotion speed on food.

A Dominant-Negative unc-1 Mutation Suppresses Aggregation in an unc-9–Dependent Manner.

unc-9 has strong genetic interactions with a second gene, unc-1, whose mutation leads to a similar uncoordinated and anesthesia-resistant phenotype (38). unc-1 encodes a stomatin-like integral membrane protein that colocalizes with UNC-9 proteins and modulates the gating of UNC-9 gap junctions in muscle (Fig. 2A) (37). Although most alleles of unc-1 are recessive, a few are dominant and might be dominant-negative alleles (39). To examine this possibility, we cloned cDNAs for two unc-1 alleles, n774 and n494, and expressed them pan-neuronally in a wild-type background. Both unc-1(n774) and unc-1(n494) transgenes reduced swimming speeds in buffer (Fig. 2B, in red). The stronger unc-1(n494) transgene resembled unc-1– or unc-9–null mutant animals (Fig. 2B), consistent with its strong phenotype as a genetic mutant (39). These results suggested that unc-1(n494) might be able to inhibit unc-1 or unc-9 function in specific neurons.

Fig. 2.

A dominant-negative unc-1 cDNA that suppresses aggregation is rescued by UNC-9::GFP. (A) Cartoon of UNC-1 (285 aa) showing location of dominant-negative n774 and n494 mutations and its inferred relationship with two aligned gap junction hemichannels composed of eight UNC-9 subunits each from different cells. UNC-9::GFP gap junction can function independently of unc-1 (37). (B) Swimming behavior in wild-type N2 animals expressing wild-type or mutant unc-1 cDNAs under the pan-neuronal tag-168 promoter and in unc-1–null mutants. Each dot represents the number of body bends per minute measured from a single animal; bars represent mean and SEM. n = 10–19 animals per genotype. (C) npr-1 mutants expressing wild-type or mutant unc-1 cDNAs in the RMG circuit. (Top) Aggregation behavior. n = 6–17 assays per genotype. (Bottom) Locomotion speed in the presence of food. n = 2–8 assays per genotype. (D) npr-1 mutants expressing unc-1(n494) and UNC-9::GFP fusion protein in different subsets of neurons in the RMG circuit. (Top) Aggregation behavior. n = 5–17 assays per genotype. (Bottom) Locomotion speed on a bacterial lawn. n = 3–5 assays per genotype. (E) npr-1 mutants expressing unc-1(n494) cDNA in RMG (Pflp-5), URX (Pflp-8), or both neurons. (Top) Aggregation behavior. n = 6–10 assays per genotype. (Bottom) Locomotion speed on a bacterial lawn. n = 3–8 assays per genotype. For Top panels in C–E, boxes represent median and first and third quartiles, and whiskers represent 5th–95th percentiles. For Bottom panels in C–E, bars represent mean ± SEM. For all panels, *P < 0.05, **P < 0.01, ***P < 0.001, ns (nonsignificant) by one-way ANOVA with Dunnett’s correction for multiple comparisons.

Expressing unc-1(n494) broadly in the RMG circuit using the flp-21 promoter suppressed the aggregation behavior of npr-1 mutants and their high foraging speed on food, an effect similar to knocking out unc-9 in those neurons (Fig. 2C). Locomotion in the flp-21::unc-1(n494) strain was normal in the absence of food, indicating that the transgene did not disrupt motor function (Fig. S2A). Expressing unc-1(wt) or unc-1(n774) cDNAs under the same promoter only slightly decreased aggregation and speed on food (Fig. 2C). Thus, transgenic unc-1(n494) interferes with the RMG circuit in ways that resemble inactivation of unc-9.

Fig. S2.

Behaviors of animals that express unc-1(n494) and/or UNC-9::GFP. (A) Locomotion speed in the absence of food in npr-1 mutants expressing wild-type or mutant unc-1 cDNAs in the RMG circuit. n = 2–8 assays per genotype. (B) Locomotion speed in the absence of food in npr-1 mutants expressing UNC-9::GFP or UNC-9::SL2::GFP in different subsets of neurons in the RMG circuit. n = 3–4 assays per genotype. (C) Locomotion speed in the absence of food in npr-1 mutants expressing unc-1(n494) cDNA in RMG (Pflp-5), URX (Pflp-8), or both neurons. n = 2–4 assays per genotype. (D) Aggregation behavior of npr-1 or N2 animals expressing UNC-9::GFP under different promoters. n = 6–9 assays per genotype. (E) Locomotion speed on the bacterial lawn in npr-1 mutants expressing UNC-9::GFP or a bicistronic UNC-9::SL2::GFP transcript in a subset of neurons in the RMG circuit. n = 4 assays per genotype. (F) Locomotion speed in the absence of food in npr-1 mutants expressing UNC-9::GFP, a bicistronic UNC-9::SL2::GFP transcript in a subset of neurons in the RMG circuit. n = 3–5 assays per genotype. For A-C, E, and F, bars represent mean ± SEM. For D, boxes represent median and first and third quartiles, and whiskers represent 5th–95th percentiles. For all panels, *P < 0.05, **P < 0.01, ns (nonsignificant) by one-way ANOVA with Dunnett’s correction for multiple comparisons.

Previous studies demonstrated that an UNC-9::GFP fusion protein can bypass the endogenous function of unc-1 in body wall muscles, apparently by generating constitutively active gap junctions (37). We found that expressing the UNC-9::GFP fusion protein reversed the effect of the unc-1(n494) transgene in npr-1 mutants and restored aggregation behavior and high locomotion speed (Fig. 2D). In a control experiment, a bicistronic unc-9::SL2::GFP transgene that generated separate UNC-9 and GFP protein products did not restore aggregation behavior (Fig. 2D). These results suggest that unc-1(n494) suppresses aggregation by inhibiting unc-9 and not by causing general neuronal defects. Conversely, the behavioral effects of UNC-9::GFP were only observed in the unc-1(n494) context; expressing UNC-9::GFP in an N2 background under either of two promoters did not elicit aggregation behavior, nor did expressing UNC-9::GFP in the npr-1 background alter aggregation (Fig. S2D). Thus, UNC-9::GFP did not hyperactivate the RMG circuit on its own or interfere with npr-1 function. The mutual dependence of unc-1(n494) and UNC-9::GFP supports a close interaction between these genes.

UNC-1 is a regulatory subunit and not a structural subunit of gap junctions, and the sites of unc-1(n494) activity were not always identical to those identified for unc-9 in the Cre-Lox experiments. In general, unc-1(n494) had to be expressed in a greater number of cells to have an effect, suggesting that it might attenuate function best when expressed by both gap junction-coupled neurons. For example, Cre-Lox deletion of unc-9 in RMG suppressed aggregation in npr-1 animals (Fig. 1D), but expression of unc-1(n494) from the same promoter did not (Fig. 2E). However, expressing unc-1(n494) simultaneously in the oxygen-sensing URX neurons and the RMG neuron suppressed aggregation (Fig. 2E), in agreement with previous studies suggesting that RMG and URX both contribute to this behavior (28).

Rescue with UNC-9::GFP provided a view of the contributing neurons that was more similar to unc-9 knockout. Expressing the UNC-9::GFP fusion protein in RMG neurons was sufficient to rescue aggregation in the flp-21::unc-1(n494) strain, but expression in URX was not (Fig. 2D). Although there are subtleties, the results of these manipulations consistently point to RMG as the site at which unc-9 and unc-1(n494) exert their strongest effects on aggregation.

Locomotion speed on food was strongly suppressed by unc-1(n494) expression from the flp-21 promoter, and this effect was rescued by expressing UNC-9::GFP in either RMG or URX (Fig. 2D). However, locomotion speed was only weakly suppressed by unc-1(n494) expression in both RMG and URX (Fig. 2E). This intermediate result resembles the intermediate effects of cell-specific unc-9 knockout on locomotion (Fig. 1E) and indicates that RMG, URX, and other flp-21–expressing neurons regulate locomotion speed on food.

Cell-Specific Expression of unc-1(n494) in the RMG Circuit Modifies Individual Sensory Behaviors.

Previous experiments with the RMG circuit have examined the effects of silencing sensory transduction with tax-4 or osm-9 mutations or silencing synaptic transmission with a tetanus toxin light chain (24, 28, 40). The unc-1(n494) transgene presented an opportunity to ask where gap junctions might act to regulate individual aggregation-related behaviors, which was pursued by expressing unc-1(n494) alone or in combination with UNC-9::GFP using different cell-specific promoters.

Aggregation in npr-1 animals is strongly associated with avoidance of high oxygen levels (26). The URX neurons sense increasing oxygen concentrations directly and are required for behavioral responses to oxygen upshifts that include reversals, turning, and increased locomotor speed (31, 41, 42). Expressing unc-1(n494) in the RMG circuit with the flp-21 promoter shifted the oxygen-dependent locomotion speed and reversal rates of npr-1 animals toward those of N2 animals (Fig. 3 A–D). The effects of unc-1(n494) on speed were partly reversed by expressing UNC-9::GFP either in the URX or in the RMG neurons or in both (Fig. 3 A–D). These results suggest that oxygen responses are regulated by gap junctions in several neurons of the RMG circuit, including RMG, URX, and other flp-21–expressing neurons.

Fig. 3.

Gap junction signaling in the RMG circuit is required for npr-1 modulation of O₂ chemosensory responses. (A) Population mean locomotion speed (±SEM) in response to an 18%–7%–18% oxygen switch. Red bars indicate the 80-s time windows used for quantification in B, and gray areas indicate periods at 18% oxygen. (B) Population mean frequency of reversal initiation events (±SEM) in response to a 18%–7%–18% oxygen switch. Red and blue bars indicate the time windows compared for quantification in D, and gray areas indicate periods at 18% oxygen. (C) Trial means of locomotion speed (±SEM) at 18% oxygen represented as bar graphs. (D) Trial means of reversal frequency change (post–pre) (±SEM) in response to a 7%–18% oxygen switch from the same experiments. (E) Calcium imaging of URX in response to oxygen upshift (7%–18%). Mean and individual traces are shown as black and gray lines, respectively. Traces show G-CaMP5K fluorescence relative to baseline (mean fluorescence from 1 to 10 s). Genotypes and number of animals are indicated in each panel. Data in A and C are binned by 1 s and 15 s, respectively. Each trial was performed on ∼50–90 animals. n = 6 trials, except n = 8 for N2 and n = 12 for npr-1; Pflp-21::unc-1(n494). Population mean is an average across all recorded animals in each bin. Trial mean is the average of population means from each trial calculated from the periods indicated by red bars in A and C. *P < 0.05, **P < 0.01, ***P < 0.001, ns (nonsignificant) by one-way-ANOVA with Holm–Sidak’s multicomparison correction.

The activity of the URX sensory neurons is regulated by oxygen, but this primary sensory response has not been compared between N2 and npr-1 animals. Using functional calcium imaging, we did not detect significant differences in URX oxygen responses when comparing N2, npr-1, and unc-1(n494) strains under conditions similar to the ones at which these strains had distinct behaviors (Fig. 3E). Although more detailed analysis may yield differences in URX calcium responses, these results suggest that oxygen-sensitive behavior might be regulated by npr-1 and gap junctions at a step downstream of or parallel to the initial URX calcium response (33).

In another npr-1–regulated behavior, N2 animals are repelled by the pheromone ascr#3 (asc-ΔC9), but npr-1 animals are not (Fig. 4 A–C) (43). ascr#3 is sensed by the ADL neurons in the RMG circuit (43). Expressing unc-1(n494) in ADL in npr-1 animals caused them to avoid ascr#3 (Fig. 4C, in red), as did expressing unc-1(n494) in either RMG or URX (Fig. 4 D and E). As URX does not form any direct gap junctions with ADL, this result suggests that ascr#3 avoidance may reflect both direct and indirect effects of the RMG gap junctions on ADL-mediated behaviors (16, 17, 19). Control experiments demonstrated that a unc-1(wt) cDNA had no effect on ascr#3 avoidance (Fig. 4 C–E) and that unc-1(n494) expression did not enhance ascr#3 avoidance in an N2 background (Fig. S3A).

Fig. 4.

Gap junction signaling in the RMG circuit is required for the npr-1 modulation of sensory responses to pheromone. (A, Left) Diagram of drop test. A drop of diluted ascr#3 (asc-ΔC9) is delivered by a glass capillary near the nose of the forward-moving animal on the homogeneous bacterial lawn. (B) Sites at which unc-9 affects specific behaviors (Figs. 1–4). Circuit for acute reversal to ascr#3 (C–E); circuit for regulating locomotion speed on food (F and G, also Figs. 1E, 2 C–E, and 3 A and C); circuit for aggregation (H, also Figs. 1D and 2 C–E). (C) Avoidance of 25 nM ascr#3 by npr-1 mutants expressing unc-1(wt) or unc-1(n494) in ADL neurons. n = 14–28 animals per genotype. (D) Avoidance of 25 nM ascr#3 by npr-1 mutants expressing unc-1(wt) or unc-1(n494) in RMG neurons. n = 6–17 animals per genotype. (E) Avoidance of 25 nM ascr#3 by npr-1 mutants expressing unc-1(wt) or unc-1(n494) in URX neurons. n = 8–22 animals per genotype. (F) Locomotion speed on a bacterial lawn in npr-1 mutants expressing unc-1(wt) or unc-1(n494) in ASK or ADL. n = 3–7 assays per genotype. (G) Locomotion speed in the absence of food in the npr-1 mutants expressing unc-1(wt) or unc-1(n494) in ASK. n = 2–6 assays per genotype. (H) Aggregation behavior in npr-1 mutants expressing unc-1(wt) or unc-1(n494) in ASK or ADL. n = 4–8 assays per genotype. For C–E, boxes represent median and first and third quartiles, and whiskers represent 10th–90th percentiles. For F and G, bars represent mean ± SEM. For H, boxes represent median and first and third quartiles, and whiskers represent 5th–95th percentiles. For all panels, *P < 0.05, **P < 0.01, ***P < 0.001, ns (nonsignificant) by one-way ANOVA with Dunnett’s correction for multiple comparisons.

Fig. S3.

Cell-specific manipulation of gap junctions or synaptic transmission. (A) Acute avoidance of 25 nM ascr#3 of N2 animals expressing unc-1(n494) cDNA in a subset of neurons in the RMG circuit. N2 animals expressing tetanus toxin in ADL have been reported previously (43) and were tested in parallel as controls. n = 4–6 assays per genotype. (B) Aggregation behavior of npr-1 mutants expressing unc-1(wt) or unc-1(n494) in both ADL and ASH neurons simultaneously. n = 3–5 assays per genotype. (C) Aggregation behavior of npr-1 mutants expressing tetanus toxin (TeTx) or pkc-1(gf) in ADL, ASH, or both neurons. TeTx blocks synaptic vesicle release, and pkc-1(gf) enhances release of synaptic vesicles and neuropeptides. n = 3–26 assays per genotype. For A, boxes represent median and first and third quartiles, and whiskers represent 10th–90th percentiles. For B and C, boxes represent median and first and third quartiles, and whiskers represent 5th–95th percentiles. For all panels, *P < 0.05, **P < 0.01, ***P < 0.001, ns (nonsignificant) by one-way ANOVA with Dunnett’s correction for multiple comparisons.

In examining additional candidates from the RMG circuit, we found that expressing unc-1(n494) in ASK neurons significantly reduced locomotion speed on food, like unc-1(n494) expression in RMG and URX (Fig. 4F). However, expressing unc-1(n494) in ASK did not reduce speed off food or aggregation (Fig. 4 G and H).

unc-1(n494) also had little effect on aggregation when expressed simultaneously in ASH and ADL, two neurons whose sensory activity is required for npr-1–related behaviors (Fig. S3B) (24). However, simultaneous manipulation of chemical synapses in ASH and ADL reciprocally increased or suppressed aggregation, suggesting that these neurons interact with aggregation behavior through their chemical synapses (Fig. S3C).

ADL Sensory Activity Reflects Gap Junction Activity in the RMG Circuit.

Ascr#3 induces a calcium increase in ADL neurons that differs between N2 and npr-1 animals (43). To ask how unc-1(n494) affects ADL ascr#3 signaling, we examined sensory dynamics using the G-CaMP5K sensor, whose calcium affinity is higher than the G-CaMP3 reporter used in previous experiments (44). ADL neurons were subjected to a series of five ascr#3 pulses, a protocol that assessed sensory adaptation and recovery as well as the initial sensory response. In agreement with previous studies (43), the ADL responses of N2 adapted more quickly than those of npr-1 (t 1/2 ∼3.6 s for N2 vs. 9.2 s for npr-1) and recovered less between successive ascr#3 pulses (Fig. 5A).

Fig. 5.

ADL calcium responses to ascr#3 in unc-1(n494) strains. (A) Average calcium responses detected with G-CaMP5K in N2, npr-1, npr-1;ADL::unc-1(wt), and npr-1;ADL::unc-1(n494) animals. Each trace was scaled to its maximum value F_max before averaging (Materials and Methods). Gray areas indicate five 20-s pulses of 100 nM ascr#3 (asc-ΔC9). n = 20 animals for N2, 16 for npr-1, 16 for npr-1;ADL::unc-1(wt), and 15 for npr-1;ADL::unc-1(n494). (B) Visual representation of the first principal component explaining 32% of the variance in 10 tested genotypes (Fig. S5). (C) PCA of ADL calcium responses. The graph shows the distribution of eigenvalues of the first principal component (E^PC1) in each genotype, with mean value and SEM. Each dot represents an eigenvalue from a single trace. ***P < 0.001, ns (nonsignificant) by one-way ANOVA with Dunnett’s correction for multiple comparisons.

The analysis of ADL responses in animals expressing unc-1(n494) in various neurons was challenging, because individual ADL calcium responses were highly variable between animals with the same genotype, and even between individual ascr#3 pulses (Fig. S4). This variability could be due to genetic, environmental, or technical differences between the experiments. To better understand the features of the ADL response that might vary between genotypes, we used principal component analysis (PCA) to analyze a full ADL imaging dataset from 10 different strains (Fig. S5 and Materials and Methods). Intuitively, the aim of this approach was to extract filters that best separated the input traces and subsequently ask whether such unbiased filters were preferentially associated with particular genotypes. Each principal component represents a time course that explains a fraction of the variance between traces. The first principal component identified in this analysis explained 32% of the total variance between traces, followed distantly by the second (11%) and third (7.2%) components (Fig. S5D).

Fig. S4.

Individual ADL calcium traces of N2, npr-1, and npr-1 animals expressing ADL::unc-1. (A) Individual N2 traces. (B) Individual npr-1 traces. (C) Individual npr-1;ADL::unc-1(wt) traces. (D) Individual npr-1;ADL::unc-1(n494) traces. Black lines represent actual traces. Red lines represent reconstructions of each trace based only on the first principal component, as defined in Fig. S5. ΔF/F₀ was calculated by first scaling all traces on a 0–1 scale, because the magnitude of the response did not vary by genotype (Materials and Methods). Gray area indicates five pulses of 100 nM of ascr#3 (asc-ΔC9) for 20 s. Titles represent the eigenvalues of first principal component of each trace from PCA of 10 genotypes.

Fig. S5.

PCA on calcium traces of N2, npr-1, and unc-1-expressing N2 and npr-1 animals. (A) Average calcium responses detected with G-CaMP5K. N2, n = 20 animals; N2 expressing unc-1(wt) and unc-1(n494) in ADL, n = 19 animals each; npr-1, n = 16 animals; npr-1 expressing unc-1(wt) and unc-1(n494) in ADL, n = 16 and 15 animals, respectively; npr-1 expressing unc-1(wt) and unc-1(n494) in RMG, n = 13 animals each; npr-1 expressing unc-1(wt) and unc-1(n494) in URX, n = 10 and 11 animals, respectively. For all panels, each trace was scaled to its maximum value F_max before averaging (Materials and Methods). Gray areas indicate the five 20-s pulses of 100 nM of ascr#3 (asc-ΔC9). (B–E) PCA of ADL calcium responses. (B) The distribution of eigenvalues of the first principal component (E^PC1) in each genotype. Each dot represents an eigenvalue from a single trace. *P < 0.05, ***P < 0.001, ns (nonsignificant) by one-way ANOVA with Dunnett’s correction for multiple comparisons. (C) The distribution of eigenvalues of the second principal component (E^PC2) in each genotype. Each dot represents an eigenvalue from a single trace. *P < 0.05 compared with npr-1. No other significant difference was found by one-way ANOVA with Dunnett’s correction for multiple comparisons. (D) Percent variance explained by each principal component (152 input traces). (E) Visual representations of PC1–PC3. The top curve shows the mean of all ADL calcium traces from the 10 different genotypes shown in A–C. PC1 separates traces based on their adaptation within and between pheromone pulses and correlates with genotype and npr-1–like responses to ascr#3. PC2 separates traces based on the magnitude of the initial peak response. PC3 is complex. Gray areas indicate the five 20-s pulses of 100 nM ascr#3 (asc-ΔC9).

Direct inspection of the first principal component indicates that it largely reflects the rate of adaptation in the calcium response within and between ascr#3 pulses (Fig. 5B and Fig. S5E). Reconstruction of traces based on the eigenvalues of the first principal component supported that conclusion (Fig. S4), and statistical analysis of its eigenvalues identified a substantial and significant separation of N2 and npr-1 traces based on genotypes (Fig. 5C and Materials and Methods).

Expression of unc-1(n494) in ADL, RMG, or URX neurons of npr-1 animals resulted in a significant shift toward N2-like responses in this first principal component, matching the behavioral result (Fig. 5 A and B and Fig. S5). In control experiments, expressing unc-1(wt) in ADL or URX had no effect, although unc-1(wt) did have a partial effect when expressed in RMG (Fig. S5 A and B). ADL expression of unc-1(n494) or unc-1(wt) in the N2 background did not alter ADL responses, consistent with the behavioral data indicating that unc-1(n494) only affects npr-1 animals (Fig. S5 A and B). Together, these results suggest that the first principal component in this analysis represents features of the ADL calcium response that are significantly affected by the npr-1 genotype and similarly affected by unc-1(n494). The observed effect on ADL adaptation is reminiscent of the role of npr-1 in the ASH neurons, where it enhances adaptation to noxious stimuli (45).

Discussion

The innexin encoded by unc-9 is essential in the RMG hub-and-spoke circuit that regulates aggregation, foraging speed on food, and behavioral responses to oxygen and pheromones. The genetic requirement for an innexin provides functional support for the model of the RMG circuit suggested by the C. elegans wiring diagram, whose central feature is the communication of multiple sensory neurons with an RMG hub through gap junctions. Chemical synapses of many neurons also modify these behaviors (28), demonstrating that gap junctions and chemical synapses in the RMG circuit have complementary rather than redundant roles.

After demonstrating an effect of unc-9 in the RMG circuit, we showed that a dominant-negative unc-1(n494) transgene can antagonize the function of UNC-9–containing gap junctions. The rescue of unc-1(n494) by UNC-9::GFP supports a close link between unc-1 and unc-9 in the RMG circuit. Further use of unc-1(n494) transgenes may help to establish the roles of unc-9 gap junctions in other circuits and may be useful to bypass the strong locomotion defect in unc-9–null mutants. Such studies should, however, include the UNC-9::GFP control to exclude off-target effects. The specificity of unc-1 is not known, and it might also interact with gap junctions encoded by other C. elegans innexins. Even within the RMG circuit, unc-9 is not the only candidate innexin; RMG expresses unc-7, and ADL and ASK express inx-18 (35). Moreover, although stomatin proteins such as UNC-1 can regulate electrical synapses in C. elegans, the family has other functions. Some C. elegans and vertebrate stomatin proteins are regulators of ENaC/Deg channels, and most stomatins are uncharacterized (37, 46, 47). Future studies will determine whether UNC-1 selectively and specifically modulates electrical synapses.

The expression of unc-1(n494) in the RMG circuit had a dramatic effect on aggregation and related behavior in npr-1 mutants but no significant effect on N2 animals. Therefore, unc-1(n494) interacts specifically with the npr-1–sensitive features of the RMG circuit and its component neurons; it does not, for example, interfere with the basic oxygen-sensing properties of URX or pheromone-sensing properties of ADL. NPR-1 is a G protein-coupled receptor thought to be coupled to the Gi/o pathway, but its ultimate molecular target is not known (30, 48). It is possible that NPR-1 regulates UNC-9–containing gap junctions directly, but it might also silence the RMG circuit by affecting excitability, synaptic transmission, or neuropeptide signaling (28, 33, 34, 49).

The importance of unc-9/unc-1–sensitive gap junctions varied considerably among different neurons, depending on the relevant behaviors. Aggregation behavior was closely associated with unc-9 activity in the RMG neuron, consistent with its identity as a hub neuron and its many gap junctions, and also included a contribution from URX neurons. The regulation of locomotion speed was distributed among several neurons, including RMG, URX, and ASK. Sensory responses to oxygen levels also involved RMG, URX, and other neurons. Some of these responses are rapidly modified by acute expression of npr-1 from a heat shock promoter, suggesting that they are acute and primary readouts of circuit activity (40, 43).

The results of these experiments help to distinguish between alternative models of RMG circuit function. For example, ASH and ADL have been proposed to act as current sinks for RMG that only indirectly contribute to the function of the RMG circuit (33). However, unc-1(n494) interference with gap junctions, which would be expected to negate this function, did not suppress aggregation, and modulation of their synaptic output could affect aggregation bidirectionally, suggesting an instructive role for these neurons. That said, the genetic results are consistent with many possibilities, and a full understanding awaits fast and direct assays for gap junctions and synaptic activity in vivo.

Ascr#3 pheromone avoidance was sensitive to unc-1(n494) expression at multiple locations, including ADL, RMG, and the indirectly connected URX neuron. Thus, both direct and indirect effects of npr-1 on ADL are mediated by the RMG circuit. In another example of an npr-1–dependent but indirect effect, expression of the ADL chemoreceptor gene srh-234 is indirectly regulated by unc-7 and unc-9 expression in other neurons (34) and is likely to represent a slow pathway for reconfiguring the circuit. npr-1 mutants show an altered gene expression landscape at the whole-animal level affecting neuropeptides, insulin signaling, and metabolic processes that are important during embryonic and larval development (32). The chronic behavioral state induced by npr-1 may alter circuit physiology and gene expression over a variety of spatial and temporal scales, which may be more easily resolved with new tools for manipulating gap junctions.

Materials and Methods

Genetics.

In Cre-Lox experiments (Fig. 1 and Fig. S1), two separate transgenes were made by microinjection and then crossed with one another to generate animals bearing both transgenes that were tested for behavior. This design was used to prevent recombination or transactivation between the tested plasmids. Cell-specific expression of Cre recombinase and subsequent elimination of unc-9 activity was verified by the absence of GFP and/or the presence of mCherry in targeted cells. A full strain list appears in Table S2.

Table S2.

List of strains and corresponding figures

Strain	Genotype	Figures
N2	Wild-type Bristol strain	Figs. 1 C–E, 2 B–E, 3 A–D, and 4 C–H and Figs. S1 A–C, S2 A–F, and S3A
CX13663	npr-1(ad609) X	Figs. 1 C–E, 2 B–E, 3 A–D, and 4 C–H and Figs. S1 A–C, S2 A–F, and S3 B and C
CX10041	unc-9(e101) npr-1(ad609) X	Fig. 1 C–E and Fig. S1 A–C
CB719	unc-1(e719) X	Fig. 2B
CX15925	unc-9(e101) npr-1(ad609) X; kyEx5390 [pHE01 (Punc-9::loxP::unc-9::SL2::GFP::loxP::mCherry), 2 ng/µL]	Fig. 1 C–E
CX15926	unc-9(e101) npr-1(ad609) X; kyEx5391 [pHE01 (Punc-9::loxP::unc-9::SL2::GFP::loxP::mCherry), 2 ng/µL]	Fig. 1 C–E
CX15927	unc-9(e101) npr-1(ad609) X; kyEx5392 [pHE01 (Punc-9::loxP::unc-9::SL2::GFP::loxP::mCherry), 2 ng/µL]	Fig. 1 C–E
CX16150	unc-9(e101) npr-1(ad609) X; kyIs628 [pEM03 (Pncs-1::Cre), 20 ng/µL]	Fig. 1 C–E and Fig. S1 A–C
CX16173	unc-9(e101) npr-1(ad609) X; kyEx5526 [pHE03 (Pflp-5::Cre), 15 ng/µL]	Fig. 1 C–E and Fig. S1 A–C
CX16174	unc-9(e101) npr-1(ad609) X; kyEx5518 [pHE04 (Ptag-168::Cre), 15 ng/µL]	Fig. 1 C–E and Fig. S1 A–C
CX17160	unc-9(e101) npr-1(ad609) X; kyEx5994 [pHE04 (Pflp-8::Cre), 10 ng/µL]	Fig. 1 C–E and Fig. S1 A–C
CX14160	kyEx4421 [pSF239 (Ptag-168::unc-1(wt)::SL2::GFP), 10 ng/µL]	Fig. 2B
CX14161	kyEx4422 [pSF239 (Ptag-168::unc-1(wt)::SL2::GFP), 10 ng/µL]	Fig. 2B
CX14106	kyEx4389 [pSF102 (Ptag-168::unc-1(n494)::SL2::GFP), 10 ng/µL]	Fig. 2B
CX14107	kyEx4390 [pSF102 (Ptag-168::unc-1(n494)::SL2::GFP), 10 ng/µL]	Fig. 2B
CX14108	kyEx4391 [pSF103 (Ptag-168::unc-1(n774)::SL2::GFP), 10 ng/µL]	Fig. 2B
CX14109	kyEx4393 [pSF103 (Ptag-168::unc-1(n774)::SL2::GFP), 10 ng/µL]	Fig. 2B
CX12328	kyEx3438 [pHE22 (Psre-1::TeTx::mCherry), 30 ng/µL]	Fig. S3A
CX15221	npr-1(ad609) X; kyEx5110 [pHE05 (Pflp-21::unc-1(wt)::SL2::mCherry), 15 ng/µL]	Fig. 2C and Fig. S2A
CX15222	npr-1(ad609) X; kyEx5111 [pHE05 (Pflp-21::unc-1(wt)::SL2::mCherry), 15 ng/µL]	Fig. 2C and Fig. S2A
CX15226	npr-1(ad609) X; kyEx5115 [pHE06 (Pflp-21::unc-1(n774)::SL2::mCherry), 15 ng/µL]	Figs. 2C and 3 A–D and Fig. S2A
CX15227	npr-1(ad609) X; kyEx5116 [pHE06 (Pflp-21::unc-1(n774)::SL2::mCherry), 15 ng/µL]	Fig. 2C and Fig. S2A
CX15228	npr-1(ad609) X; kyEx5117 [pHE06 (Pflp-21::unc-1(n774)::SL2::mCherry), 15 ng/µL]	Fig. 2C and Fig. S2A
CX15223	npr-1(ad609) X; kyEx5112 [pHE07 (Pflp-21::unc-1(n494)::SL2::mCherry), 15 ng/µL]	Fig. 2 C and D and Figs. S2 A and B and S3A
CX15224	npr-1(ad609) X; kyEx5113 [pHE07 (Pflp-21::unc-1(n494)::SL2::mCherry), 15 ng/µL]	Figs. 2 C and D and 3 A–D and Figs. S2 A and B and S3A
CX15225	npr-1(ad609) X; kyEx5114 [pHE07 (Pflp-21::unc-1(n494)::SL2::mCherry), 15 ng/µL]	Fig. 2 C and D and Fig. S2 A and B
CX15669	npr-1(ad609) X; kyEx5281 [pHE07 (Pflp-21::unc-1(n494)::SL2::mCherry), 15 ng/µL]	Fig. 3 A–D
CX15670	npr-1(ad609) X; kyEx5282 [pHE07 (Pflp-21::unc-1(n494)::SL2::mCherry), 15 ng/µL]	Fig. 3 A–D
CX15295	npr-1(ad609) X; kyEx5140 [pHE07 (Pflp-21::unc-1(n494)::SL2::mCherry), 15 ng/µL + pHE08 (Pnpr-1::unc-9::GFP), 30 ng/µL]	Fig. 2D and Fig. S2B
CX15296	npr-1(ad609) X; kyEx5141 [pHE07 (Pflp-21::unc-1(n494)::SL2::mCherry), 15 ng/µL + pHE08 (Pnpr-1::unc-9::GFP), 30 ng/µL]	Fig. 2D and Fig. S2B
CX15297	npr-1(ad609) X; kyEx5142 [pHE07 (Pflp-21::unc-1(n494) SL2 mCherry), 15 ng/µL + pHE08 (Pnpr-1::unc-9::GFP), 30 ng/µL]	Fig. 2D and Fig. S2B
CX15289	npr-1(ad609) X; kyEx5134 [pHE07 (Pflp-21::unc-1(n494)::SL2::mCherry), 15 ng/µL + pHE09 (Pnpr-1::unc-9::SL2::GFP), 30 ng/µL]	Fig. 2D and Fig. S2B
CX15290	npr-1(ad609) X; kyEx5135 [pHE07 (Pflp-21::unc-1(n494)::SL2::mCherry), 15 ng/µL + pHE09 (Pnpr-1::unc-9::SL2::GFP), 30 ng/µL]	Fig. 2D and Fig. S2B
CX15291	npr-1(ad609) X; kyEx5136 [pHE07 (Pflp-21::unc-1(n494)::SL2::mCherry), 15 ng/µL + pHE09 (Pnpr-1::unc-9::SL2::GFP), 30 ng/µL]	Fig. 2D and Fig. S2B
CX15335	kyEx5155 [pHE08 (Pnpr-1::unc-9::GFP), 30 ng/µL]	Fig. S2D
CX15336	kyEx5156 [pHE08 (Pnpr-1::unc-9::GFP), 30 ng/µL]	Fig. S2D
CX15337	kyEx5157 [pHE08 (Pnpr-1::unc-9::GFP), 30 ng/µL]	Fig. S2D
CX15405	kyEx5186 [pHE10 (Pflp-21::unc-9::GFP), 80 ng/µL]	Fig. S2D
CX15406	kyEx5187 [pHE10 (Pflp-21::unc-9::GFP), 80 ng/µL]	Fig. S2D
CX15407	kyEx5188 [pHE10 (Pflp-21::unc-9::GFP), 80 ng/µL]	Fig. S2D
CX15298	npr-1(ad609) X; kyEx5143 [pHE08 (Pnpr-1::unc-9::GFP), 30 ng/µL]	Fig. S2 D–F
CX15299	npr-1(ad609) X; kyEx5144 [pHE08 (Pnpr-1::unc-9::GFP), 30 ng/µL]	Fig. S2 D–F
CX15300	npr-1(ad609) X; kyEx5145 [pHE08 (Pnpr-1::unc-9::GFP), 30 ng/µL]	Fig. S2 D–F
CX15293	npr-1(ad609) X; kyEx5138 [pHE09 (Pnpr-1::unc-9::SL2::GFP), 30 ng/µL]	Fig. S2 D–F
CX15294	npr-1(ad609) X; kyEx5139 [pHE09 (Pnpr-1::unc-9::SL2::GFP), 30 ng/µL]	Fig. S2 D–F
CX16100	npr-1(ad609) X; kyEx4530 [pHE07 (Pflp-21::unc-1(n494)::SL2::mCherry), 15 ng/µL + pHE11 (Pflp-5::unc-9::GFP), 30 ng/µL]	Figs. 2D and 3 A–D and Fig. S2B
CX16101	npr-1(ad609) X; kyEx4531 [pHE07 (Pflp-21::unc-1(n494)::SL2::mCherry), 15 ng/µL + pHE11 (Pflp-5::unc-9::GFP), 30 ng/µL]	Fig. 2D and Fig. S2B
CX16102	npr-1(ad609) X; kyEx5038 [pHE07 (Pflp-21::unc-1(n494)::SL2::mCherry), 15 ng/µL + pHE11 (Pflp-5::unc-9::GFP), 30 ng/µL]	Fig. 2D and Fig. S2B
CX16103	npr-1(ad609) X; kyEx5497 [pHE07 (Pflp-21::unc-1(n494)::SL2::mCherry), 15 ng/µL + pHE12 (Pflp-8::unc-9::GFP), 30 ng/µL]	Figs. 2D and 3 A–D and Fig. S2B
CX16104	npr-1(ad609) X; kyEx5498 [pHE07 (Pflp-21::unc-1(n494)::SL2::mCherry), 15 ng/µL + pHE12 (Pflp-8::unc-9::GFP), 30 ng/µL]	Fig. 2D and Fig. S2B
CX16098	npr-1(ad609) X; kyEx4528 [pHE07 (Pflp-21::unc-1(n494)::SL2::mCherry), 15 ng/µL + pHE11 (Pflp-5::unc-9::GFP), 30 ng/µL + pHE12 (Pflp-8::unc-9::GFP), 30 ng/µL]	Fig. 2D and Fig. S2B
CX16099	npr-1(ad609) X; kyEx4529 [pHE07 (Pflp-21::unc-1(n494)::SL2::mCherry), 15 ng/µL + pHE11 (Pflp-5::unc-9::GFP), 30 ng/µL + pHE12 (Pflp-8::unc-9::GFP), 30 ng/µL]	Fig. 3 A–D
CX14717	npr-1(ad609) X; kyEx4824 [pHE13 (Psrh-220::unc-1(wt)::SL2::mCherry), 30 ng/µL]	Fig. 4 F and H
CX14718	npr-1(ad609) X; kyEx4825 [pHE13 (Psrh-220::unc-1(wt)::SL2::mCherry), 30 ng/µL]	Fig. 4 F and H
CX14719	npr-1(ad609) X; kyEx4824 [pHE13 (Psrh-220::unc-1(wt)::SL2::mCherry), 30 ng/µL]	Fig. 4 F and H
CX14723	npr-1(ad609) X; kyEx4830 [pHE14 (Psrh-220::unc-1(n494)::SL2::mCherry), 30 ng/µL]	Fig. 4 F and H
CX14724	npr-1(ad609) X; kyEx4831 [pHE14 (Psrh-220::unc-1(n494)::SL2::mCherry), 30 ng/µL]	Fig. 4 F and H
CX14725	npr-1(ad609) X; kyEx4832 [pHE14 (Psrh-220::unc-1(n494)::SL2::mCherry), 30 ng/µL]	Fig. 4 F and H
CX15155	npr-1(ad609) X; kyEx5067 [pHE13 (Psrh-220::unc-1(wt)::SL2::mCherry), 5 ng/µL]	Fig. 4C
CX15156	npr-1(ad609) X; kyEx5068 [pHE13 (Psrh-220::unc-1(wt)::SL2::mCherry), 5 ng/µL]	Fig. 4C
CX15157	npr-1(ad609) X; kyEx5069 [pHE13 (Psrh-220::unc-1(wt)::SL2::mCherry), 5 ng/µL]	Fig. 4C
CX15158	npr-1(ad609) X; kyEx5070 [pHE14 (Psrh-220::unc-1(n494)::SL2::mCherry), 5 ng/µL]	Fig. 4C
CX15159	npr-1(ad609) X; kyEx5071 [pHE14 (Psrh-220::unc-1(n494)::SL2::mCherry), 5 ng/µL]	Fig. 4C
CX16105	npr-1(ad609) X; kyEx5499 [pHE16 (Pflp-5::unc-1(n494)::SL2::mCherry), 15 ng/µL + pHE18 (Pflp-8::unc-1(n494) SL2 mCherry), 15 ng/µL]	Fig. 2E and Fig. S2C
CX16106	npr-1(ad609) X; kyEx5500 [pHE16 (Pflp-5::unc-1(n494)::SL2::mCherry), 15 ng/µL + pHE18 (Pflp-8::unc-1(n494) SL2 mCherry), 15 ng/µL]	Fig. 2E and Fig. S2C
CX16107	npr-1(ad609) X; kyEx5501 [pHE16 (Pflp-5::unc-1(n494)::SL2::mCherry), 15 ng/µL + pHE18 (Pflp-8::unc-1(n494)::SL2::mCherry), 15 ng/µL]	Fig. 2E and Fig. S2C
CX16144	npr-1(ad609) X; kyEx5507 [pHE16 (Pflp-5::unc-1(n494)::SL2::mCherry), 15 ng/µL]	Fig. 2E and Fig. S2C
CX16145	npr-1(ad609) X; kyEx5508 [pHE16 (Pflp-5::unc-1(n494)::SL2::mCherry), 15 ng/µL]	Fig. 2E and Fig. S2C
CX16146	npr-1(ad609) X; kyEx5509 [pHE16 (Pflp-5::unc-1(n494)::SL2::mCherry), 15 ng/µL]	Fig. 2E and Fig. S2C
CX16724	npr-1(ad609) X; kyEx5750 [pHE15 (Pflp-5::unc-1(wt)::SL2::mCherry), 5 ng/µL]	Fig. 4D
CX16725	npr-1(ad609) X; kyEx5751 [pHE15 (Pflp-5::unc-1(wt)::SL2::mCherry), 5 ng/µL]	Fig. 4D
CX16721	npr-1(ad609) X; kyEx5747 [pHE16 (Pflp-5::unc-1(n494)::SL2::mCherry), 5 ng/µL]	Fig. 4D
CX16722	npr-1(ad609) X; kyEx5748 [pHE16 (Pflp-5::unc-1(n494)::SL2::mCherry), 5 ng/µL]	Fig. 4D
CX16723	npr-1(ad609) X; kyEx5749 [pHE16 (Pflp-5::unc-1(n494)::SL2::mCherry), 5 ng/µL]	Fig. 4D
CX16122	npr-1(ad609) X; kyEx5503 [pHE18 (Pflp-8::unc-1(n494)::SL2::mCherry), 15 ng/µL]	Fig. 2E and Fig. S2C
CX16123	npr-1(ad609) X; kyEx5504 [pHE18 (Pflp-8::unc-1(n494)::SL2::mCherry), 15 ng/µL]	Fig. 2E and Fig. S2C
CX16124	npr-1(ad609) X; kyEx5505 [pHE18 (Pflp-8::unc-1(n494)::SL2::mCherry), 15 ng/µL]	Fig. 2E and Fig. S2C
CX16746	npr-1(ad609) X; kyEx5770 [pHE17 (Pflp-8::unc-1(wt)::SL2::mCherry), 5 ng/µL]	Fig. 4E
CX16747	npr-1(ad609) X; kyEx5771 [pHE17 (Pflp-8::unc-1(wt)::SL2::mCherry), 5 ng/µL]	Fig. 4E
CX16748	npr-1(ad609) X; kyEx5772 [pHE17 (Pflp-8::unc-1(wt)::SL2::mCherry), 5 ng/µL]	Fig. 4E
CX16749	npr-1(ad609) X; kyEx5773 [pHE18 (Pflp-8::unc-1(n494)::SL2::mCherry), 5 ng/µL]	Fig. 4E
CX16750	npr-1(ad609) X; kyEx5773 [pHE18 (Pflp-8::unc-1(n494)::SL2::mCherry), 5 ng/µL]	Fig. 4E
CX16466	npr-1(ad609) X; kyEx5617 [pHE19 (Psra-9::unc-1(wt)::SL2::mCherry), 15 ng/µL]	Fig. 4 F–H
CX16468	npr-1(ad609) X; kyEx5619 [pHE19 (Psra-9::unc-1(wt)::SL2::mCherry), 15 ng/µL]	Fig. 4 F–H
CX16469	npr-1(ad609) X; kyEx5620 [pHE19 (Psra-9::unc-1(wt)::SL2::mCherry), 15 ng/µL]	Fig. 4 F–H
CX16467	npr-1(ad609) X; kyEx5618 [pHE19 (Psra-9::unc-1(n494)::SL2::mCherry), 15 ng/µL]	Fig. 4 F–H
CX16498	npr-1(ad609) X; kyEx5632 [pHE19 (Psra-9::unc-1(n494)::SL2::mCherry), 15 ng/µL]	Fig. 4 F–H
CX12645	npr-1(ad609) X; kyEx3541 [pHE24 (Psre-1::pkc-1(gf)::SL2::GFP), 30 ng/µL + pHE26 (Psra-6::pkc-1(gf)::SL2::GFP), 30 ng/µL]	Fig. S3C
CX12646	npr-1(ad609) X; kyEx3542 [pHE24 (Psre-1::pkc-1(gf)::SL2::GFP), 30 ng/µL + pHE26 (Psra-6::pkc-1(gf)::SL2::GFP), 30 ng/µL]	Fig. S3C
CX16616	mzmEx410 [Psre-1::G-CaMP5Kopt, 10 ng/µL]	Fig. 5 A and B and Figs. S4A and S5 A–E
CX16818	kyEx5067 [pHE13 (Psrh-220::unc-1(wt)::SL2::mCherry), 5 ng/µL]; mzmEx410 [Psre-1::G-CaMP5Kopt, 10 ng/µL]	Fig. S5 A–E
CX16819	kyEx5071 [pHE14 (Psrh-220::unc-1(n494)::SL2::mCherry), 5 ng/µL]; mzmEx410 [Psre-1::G-CaMP5Kopt, 10 ng/µL]	Fig. S5 A–E
CX16560	npr-1(ad609) X; mzmEx410 [Psre-1::G-CaMP5Kopt, 10 ng/µL]	Fig. 5 A–C and Fig. S5 A–E
CX16818	npr-1(ad609) X; kyEx5067 [pHE13 (Psrh-220::unc-1(wt)::SL2::mCherry), 5 ng/µL]; mzmEx410 [Psre-1::G-CaMP5Kopt, 10 ng/µL]	Fig. 5 A–C and Fig. S5 A–E
CX16819	npr-1(ad609) X; kyEx5071 [pHE14 (Psrh-220::unc-1(n494)::SL2::mCherry), 5 ng/µL]; mzmEx410 [Psre-1::G-CaMP5Kopt, 10 ng/µL]	Fig. 5 A–C and Fig. S5 A–E
CX16728	npr-1(ad609) X; kyEx5750 [pHE15 (Pflp-5::unc-1(wt)::SL2::mCherry), 5 ng/µL + Punc-122::dsRed, 25 ng/µL]; mzmEx410 [Psre-1::G-CaMP5Kopt, 10 ng/µL + Psre-1::tagRFP, 50 ng/µL]	Fig. S5 A–E
CX16729	npr-1(ad609) X; kyEx5748 [pHE16 (Pflp-5::unc-1(n494)::SL2::mCherry), 5 ng/µL + Punc-122::dsRed, 25 ng/µL]; mzmEx410 [Psre-1::G-CaMP5Kopt, 10 ng/µL + Psre-1::tagRFP, 50 ng/µL]	Fig. S5 A–E
CX16744	npr-1(ad609) X; kyEx5768 [pHE17 (Pflp-8::unc-1(wt)::SL2::mCherry), 5 ng/µL + Punc-122::dsRed, 25 ng/µL]; mzmEx410 [Psre-1::G-CaMP5Kopt, 10 ng/µL + Psre-1::tagRFP, 50 ng/µL]	Fig. S5 A–E
CX16745	npr-1(ad609) X; kyEx5769 [pHE18 (Pflp-8::unc-1(n494)::SL2::mCherry), 5 ng/µL + Punc-122::dsRed, 25 ng/µL]; mzmEx410 [Psre-1::G-CaMP5Kopt, 10 ng/µL + Psre-1::tagRFP, 50 ng/µL]	Fig. S5 A–E
ZIM460	mzmEx291 [Pflp-8::GCaMP5k, 5 ng/μL; Pflp-8::mCherry, 2 ng/μL]	Fig. 3E
ZIM446	npr-1(ad609)X; mzmEx291 [Pflp-8::GCaMP5k, 5 ng/μL; Pflp-8::mCherry, 2 ng/μL]	Fig. 3E
ZIM744	npr-1(ad609) X; mzmEx291 [Pflp-8::GCaMP5k, 5 ng/μL; Pflp-8::mCherry, 2 ng/μL]; kyEx5289 [pHE07 (Pflp-21::unc-1(wt)::SL2::mCherry), 15 ng/μL]	Fig. 3E
ZIM740	npr-1(ad609) X; mzmEx291 [Pflp-8::GCaMP5k, 5 ng/μL; Pflp-8::mCherry, 2 ng/μL]; kyEx5282 [pHE07 (Pflp-21::unc-1(n494)::SL2::mCherry), 15 ng/μL]	Fig. 3E
ZIM741	npr-1(ad609) X; mzmEx291 [Pflp-8::GCaMP5k, 5 ng/μL; Pflp-8::mCherry, 2 ng/μL]; kyEx5281 [pHE07 (Pflp-21::unc-1(n494)::SL2::mCherry), 15 ng/μL]	Fig. 3E

In most cases, transgenic strains were coinjected with Pmyo-3::mCherry (5 ng/µL), Punc-122::dsRed (20–25 ng/µL), or Punc-122::GFP (10–25 ng/µL).

ADL imaging strains were coinjected with Psre-1::tagRFP (50 ng/µL), and URX imaging strains were coinjected with Pflp-8::mCherry (2 ng/µL). Empty pSM vector was added to bring the total concentration of injected DNA to 60–100 ng/µL.

Molecular Biology.

Coding sequences.

The wild-type unc-1 cDNA was amplified by RT-PCR from total N2 RNA. Primers used for amplification were 5′-attattagCCTAGGatgtcaaacaaggaaagaacaga-3′ and 5′-attattagGGTACCttattggtcttttttcataaatgctcc-3′, and the resulting PCR product was cloned into the pSM-SL2::GFP vector via the NheI and KpnI sites. This 858 bp-long cDNA translates into UNC-1 isoform A with 285 amino acids. To generate mutant cDNAs of unc-1, the wild-type unc-1 construct was mutated using overlap extension PCR. For the n774 mutation, the 178th nucleotide was mutated to introduce the E64K mutation. For the n494 mutation, the 544th nucleotide was mutated to introduce the G186R mutation.

The 1,164 kb-long unc-9 cDNA codon-optimized for C. elegans expression (a generous gift from Loren Looger, Howard Hughes Medical Institute Janelia Campus, Ashburn, VA) was used for floxed unc-9 rescue and expression of unc-9::SL2::GFP. For expression of unc-9::GFP fusion protein, an unc-9::GFP fragment (a generous gift from Zhao-Wen Wang, University of Connecticut, Farmington, CT) (37) was cloned into the pSM vector via the XbaI and NheI sites.

G-CaMP5Kopt was modified for optimal codon use in C. elegans using the OptimumGene method (GenScript).

Promoters.

To drive the expression of floxed unc-9 rescue construct, 9 kb upstream, the start codon of unc-9 was amplified from the genomic DNA of N2 animals and cloned into the pSM vector using the FseI and AscI sites. Oligonucleotides used for cloning were 5′-attataGGCCGGCCagaagaatagctctcgcatatgatgtcttg-3′ and 5′-attaatGGCGCGCCcctaagtcatgtcgtgacggcg-3′.

In other transgenes, the following promoters were used: npr-1 promoter (7,196 bp), 5′-CTCGAGTTCTCGTGTTTGTG-3′ and 5′-TTTTTTAGTCTAATGGAGGA-3′; flp-5 promoter (4,250 bp), 5′-AAAGATTACAAGTGACGTGA-3′ and 5′-GGCTTGCCAACTGGCTCCGC-3′; sre-1 promoter (1,000 bp), 5′-GTCACACGAGTGT-3′ and 5′-GTTGGAATGAAAA-3′; and srh-220 promoter (2,449 bp), 5′-GGCTGATAGATAGACCGTCC-3′ and 5′-ACTTGAGTTTGGACCGAAAA-3′.

• tag-168, ncs-1, flp-21, flp-8, sra-9, and sra-6 promoters were as described (28, 50).

Behavioral Assays.

GraphPad Prism software was used for all statistical analyses. Primary behavioral data appears in Dataset S1.

Swimming behavior.

A single animal was picked using minimal amount of bacterial food and gently transferred into a well filled with a drop of 50 μL M9 in a 96-well plate. Animals started thrashing immediately after transfer. After 1 min, the number of complete body bends was counted by eye for the following 1-min period to generate thrashing rate per minute.

Aggregation and bordering behavior.

Assays were performed simultaneously as described (25, 27). Assay plates were prepared with a circular lawn about 10 mm in diameter by seeding 50 μL Escherichia coli OP50 on an NGM plate and growing for 4 d at room temperature. Sixty young adult animals grown at room temperature were picked onto the off-food area of the assay plate, and the aggregating and bordering were scored after 1 h. Aggregation was measured by calculating the fraction of animals that are in contact with two or more animals along at least half of their body length. Bordering was measured by calculating the fraction of animals that resided within 2 mm of the edge of the bacterial lawn. Fraction aggregating or bordering obtained was then analyzed with one-way ANOVA with Dunnett’s correction for multiple comparisons. For inx-6(rr5);npr-1(ad609) mutants, animals were grown at room temperature and incubated overnight at 25 °C, the restrictive temperature for rr5 temperature-sensitive allele, before being assayed at 25 °C.

Locomotion speed.

Speed was measured as described (50). Twenty to thirty young adult animals were transferred onto a food-free assay plate or a food assay plate with a uniform E. coli OP50 lawn grown overnight at room temperature. Animals were allowed to crawl freely within a 28 mm × 28 mm arena confined by a filter paper soaked with 10 mM CuCl₂ solution. At 30 min after transfer, the arena was recorded for 5–10 min. Video recordings were made at 3 frames per second using a Pixellink camera and Streampix software, followed by analysis with custom software written in MATLAB (Mathworks). Mean speed was obtained from each assay plate with 20–30 animals, and values from two to eight independent assay plates were averaged to generate data for each genotype and condition.

Oxygen flow behavioral assays.

Locomotion responses to changing O₂ concentrations were determined in a population behavioral assay as described previously (41, 51) with some modifications. For preparation of assay plates, 14 cm-diameter NGM plates were inoculated with a dense solution of E. coli OP50 bacteria and left for 1 d at room temperature to yield a homogeneous bacterial lawn. A 56 mm × 56 mm assay arena was defined by a Whatman paper soaked with 20 mM CuCl₂, a repellent, to prevent animals from leaving the arena. Fifty to ninety adult animals were transferred to the arena and left unperturbed for 1 h. At 5 min before recording onset, a transparent Plexiglas device with a flow arena of 56 mm × 56 mm × 0.7 mm was placed on top of the assay arena. O₂ concentrations were controlled using a static gas mixer connected to mass flow controllers (Vögtling Instruments) operated by LabView software. Total gas flow was set to 100 mL/min. During illumination with a red flat LED, animals were recorded at 3 frames per second using a 4 megapixel CCD (Jai) camera and Streampix software. Movie analysis, speed measurements, and turn detection were performed using a modified MATLAB-based tracking software (52), available at med.stanford.edu/wormsense/tracker/. For speed calculation, only periods of continuous forward movement were used and data were binned by taking the mean of three consecutive frames (1-s binning). Reversal initiation events were detected by measuring characteristic changes in angular velocity, and data were binned by taking the sum of events in 45 consecutive frames (15-s binning). The event frequency (min⁻¹) was calculated for each bin. For quantifications, we accounted for experiment-to-experiment variation by calculating trial means; that is, for each individual experiment, the population means within the indicated time windows were calculated.

Acute pheromone avoidance.

Ascr#3 (asc-ΔC9) was a generous gift from Rebecca Butcher, University of Florida, Gainesville, FL. The drop test assay was performed essentially as described to address acute avoidance of pheromone (53). All assays were performed in the presence of food, on uniform E. coli OP50 lawn on NGM plates (100 mm × 15 mm) that were grown overnight at 37 °C. Young adult animals that were grown in room temperature were transferred to the assay plate at room temperature to equilibrate for 30–60 min. We delivered 25 nM of ascr#3 prepared in M13 buffer (30 mM Tris, pH 7.5, 100 mM NaCl, 10 mM KCl) by mouth-pipetting to individual animals moving forward via a 10-μL glass capillary that was previously pulled by laser-based micropipette puller (P-2000, Sutter Instrument) and whose tip was gently broken. Responses were scored as reversals if animals initiated backward movements longer than half their body length within 4 s. Effect sizes (shown in figures as fraction reversing) represent the increase or decrease in fraction of animals reversing in response to pheromone, compared with the fraction reversing in response to M13 buffer alone.

ADL calcium imaging and PCA.

For ADL imaging, transgenic animals expressed G-CaMP5K sequence-optimized for expression in C. elegans under the ADL-specific sre-1 promoter. Animals were imaged in custom-fabricated polydimethylsiloxane (PDMS) chamber (54). A single well-fed animal was loaded onto the chamber and allowed to equilibrate for 5 min, followed by photo-bleaching for 2 min before imaging to silence the light response of ADL. We prepared 10 mM ascr #3 (asc-ΔC9) in ethanol stored at –20 °C at 100 nM in S-basal buffer (stimulus), and the equivalent amount of ethanol was prepared in S-basal buffer (nonstimulus control). Stimuli were delivered by using a vacuum system and automated switching between two fluid streams with a three-way valve, a ValveBank controller, and Metamorph software. Each animal was imaged once. Image stacks were obtained at 10 frames per second using an Andor iXon3 camera, Metamorph Software, and a 40× objective on the Zeiss Axioplan microscope. Stacks were then processed with a custom-written script using ImageJ. A first-level analysis of traces indicated no difference in maximum ∆F/F0 response between N2 and npr-1. However, variations in transgene expression and a low basal fluorescence level compared with background autofluorescence generated apparent ∆F/F0 values that varied widely between traces of each genotype. To minimize this effect of transgene expression, all traces were individually normalized on a 0 to 1 scale. The median values of responses during 50–60, 90–100, 130–140, 170–180, and 210–220 s (the last 10 s of second to sixth off-stimuli) were used as F0 baseline and normalized F values to generate ∆F/F0.

PCA was performed across ADL calcium traces from 10 genotypes (shown in Fig. S5) using a custom MATLAB script. All calcium traces individually scaled on a 0 to 1 axis were subjected to PCA over the time domain so that each principal component represents a time vector. Time points beginning 2 s before the first ascr#3 stimulus through 20 s after the last ascr#3 stimulus were used for PCA (total 202 s). The eigenvalues of the first principal component (E^PC1) were compared between different genotypes by ANOVA. Reconstruction in Fig. S4 was made by (mean) + (E^PC1) × (PC1).

URX calcium imaging and analysis.

Ca²⁺ imaging of URX neurons was performed using a microfluidic device for controlling environmental O₂ levels as described previously with some modifications (31, 41, 42). Animals were imaged in the presence of E. coli OP50 bacteria suspended in S-basal at 0.6 OD, to mimic on-food conditions that were used in the behavioral assays, and 1 mM of the muscle paralyzing compound tetramisole, to fully suppress movement. O₂ concentrations were controlled with a static gas mixer connected to mass flow controllers (Vögtling Instruments) operated by LabView software. Total gas flow at the inlet to the device was set to 50 mL/min. Fluorescence of G-CaMP5K expressed in URX was recorded with an inverted epifluorescence microscope equipped with a CoolLED pE-2 excitation system (470 nm) and Piston GFP bandpass filter set. Imaging data were acquired with a Photometrics Evolve 512 EMCCD camera with 100 ms exposure time, streaming images at 10 Hz acquisition rate to VisiView software. Fluorescence intensity time series were extracted from the image sequences using a custom-written script in Metamorph software. A region of interest (ROI) and a neighboring background region in each frame was were defined based on thresholding and tracked using the Metamorph track objects function, the total integrated fluorescence intensity of the ROI was determined, and background fluorescence was subtracted in each frame. We report and quantify the fluorescence relative to a baseline: ∆F/F = (F – F0)/F0. F0 is the median fluorescence within 1–99 s from recording onset.