Ablating ADL differently affects avoidance of high osmolality by 2 M glycerol in nematode Pristionchus pacificus and C. elegans. ADL ablation in the former species leads to a significant reduction of hyperosmotic avoidance, less intensively than ablating both ASH and ADL. In contrast, ADL ablation in the latter species does not affect the avoidance behavior (Srinivasan et al., 2008). A possible explanation for this difference is that the two species may have varied intrinsic sensitivity to osmolality. The ADL functions in the sense and avoidance of hyperosmolality in C. elegans need study.

To evaluate the ADL’s role in the avoidance of different osmolality, we used the drop test shown in Supplementary Figure S1A to assay avoidance of varied osmolality shocks in wild-type (WT) N2 as a control and ADL-specific neurotransmission-ablated animals. Neuron-specific neurotransmission ablation was conducted by specific expression of neuronal toxin TeTx in tested neurons. TeTx, a light chain of tetanus toxin, is a specific protease of a vesicular SNARE protein synaptobrevin essential for vesicle fusion with the plasma membrane. It hydrolyzes synaptobrevin and blocks vesicle fusion with the plasma membrane and, thus, synaptic transmission (Schiavo et al., 1992). It is used to successfully eliminate neurotransmission in tested neurons (neurons::TeTx in short; Macosko et al., 2009; Guo et al., 2015; Liu et al., 2019; Wen et al., 2020; Wu et al., 2022). In the test, the M13 buffer and different solutions of varied osmolality were used as a control and stimuli, respectively. Solutes of glycerol (99%), sodium chloride (NaCl), fructose (99%), and sorbitol (98%) were put into the M13 buffer to obtain varied concentrations. The measured osmolality of different solutions was as follows. M13 buffer, 0.28 Osm; glycerol/M13 solutions: 0.1 M, 0.41 Osm; 0.5 M, 0.88 Osm; 1.0 M, 1.37 Osm; 2.0 M, 2.29 Osm; 0.25 M NaCl/M13, 0.77 Osm; 0.5 M fructose/M13, 0.87 Osm; 0.5 M sorbitol/M13, 0.90 Osm (Supplementary Figure S1B). The hyperosmotic avoidance responses indicated by the percentage or ratio in the WT N2 and ADL (ver-2p, 2.7 kb, ADL specific)::TeTx transgenic animals were positively related to the osmolality by glycerol/M13 solutions of varied concentrations (Supplementary Figure S1C). The relationship was well fitted by a Hill function, with a half maximal effect (ED₅₀) 0.8614 and 0.9775 Osm in the WT and ADL::TeTx animals, respectively. The ADL neurotransmission elimination significantly reduced animal avoidance of mild and medium hyperosmolality of 0.41 and 0.88 Osm by 0.1 and 0.5 M glycerol/M13 solutions. However, it did not affect the avoidance of high osmolality of 1.37 and 2.29 Osm by 1 M or 2 M glycerol/M13 solutions, compared with its effect in the WT N2. The difference was most significant at 0.88 Osm. We thus used 0.5 M glycerol/M13 solution (0.88 Osm) for the following experiments in this study unless otherwise indicated. We further used 0.25 M NaCl/M13 (0.77 Osm), 0.5 M fructose/M13 (0.87 Osm), and 0.5 M sorbitol/M13 (0.90 Osm) solutions to validate the observed behavioral response to 0.8 Osm glycerol/M13 is hyperosmotic avoidance. As expected, WT N2 animal displayed similar hyperosmotic avoidance of these solutions (Supplementary Figure S1D), supporting the behavior is the avoidance of hyperosmolality, not of a given solute.

Continual inhibition of neurotransmission by TeTx beginning in embryonic periods may interfere with nervous system development. We next used chemogenetics to acutely inhibit the tested neurons and examine their neuronal functions in adult animals with HisCl1 channels and 10 mM histamine (neuron::chemogenetic inhibition, in short). HisCl1 is a histamine-gated chloride channel subunit from Drosophila that is effective for silencing neurons when activated by exogenous histamine (Pokala et al., 2014; Guo et al., 2015; Liu et al., 2019; Ge et al., 2020). Our results showed that ADL::TeTx and ADL::chemogenetic inhibition similarly reduced animals’ hyperosmotic avoidance (Figure 1A), supporting that ADL upregulates avoidance of medium hyperosmolality in C. elegans.

Given that ADL regulates hyperosmotic avoidance, it should respond to hyperosmotic stimulation cell-or non-cell-autonomously. We thus examined ADL calcium (Ca²⁺) responses to the medium hyperosmolality of 0.88 Osm by 0.5 M glycerol/M13 solution, using fluorescent Ca²⁺ imaging with GCaMP3.0 as a Ca²⁺ indicator (Tian et al., 2009) combined with microfluidic control of stimulation and animal movement (Chronis et al., 2007; Guo et al., 2015; Wang et al., 2015; Liu et al., 2019; Ge et al., 2020; Wen et al., 2020; Wu et al., 2022). ADL neurons in the WT N2 worm displayed robust Ca²⁺ responses to 0.88 Osm osmolality (Figures 1B,C; Supplementary Figure S1E). The changes in ADL Ca²⁺ signals may result from neurotransmission from other neurons. That is, ADL Ca²⁺ responses may be non-cell autonomous. We thus used unc-13(e1091) and unc-31(e928) mutant animals to assay the cell autonomy of the ADL hyperosmotic responses. Unc-13 (UNCoordinated) encodes syntaxin-1 binding protein UNC-13, which is required for synaptic vesicle fusion with the presynaptic membrane and thus essential for neurotransmitter release (Richmond et al., 1999; Tokumaru and Augustine, 1999). The gene unc-31 encodes UNC-31 protein, an ortholog of human CAPS (calcium-dependent secretion activator), which is essential for the exocytosis of dense cored vesicles and thus the release of neuropeptides (Avery et al., 1993; Lin et al., 2010). In unc-13 or/and unc-31 mutant animal/s, if the sensory response of a tested sensory neuron is significantly reduced or even disappears, then the neuronal response is non-cell autonomous; if the response remains unchanged, then it is cell autonomous; if the response increases, then it is cell autonomous and is inhibited by other neuron/s. Our result showed that unc-13 and unc-31 mutant animals displayed significantly augmented and WT ADL Ca²⁺ responses, respectively (Figures 1B,C; Supplementary Figure S1E). This result indicates that ADL Ca²⁺ responses to hyperosmolality are cell-autonomous and may be inhibited by neurotransmission mediated by classical neurotransmitter/s from other neurons.

ASH is the known main hyperosmolality-sensitive neuron so it may be the source of ADL inhibition. We then employed ASH::TeTx transgenic animals to determine the ASH source of ADL inhibition. ASH-specific expression of TeTx was driven by the srv-11 promoter (1.9 kb upstream of the start codon; Taniguchi et al., 2014; O'Donnell et al., 2020). The ASH::TeTx transgenic animal displayed significantly increased ADL Ca²⁺ responses to 0.88 Osm osmolality, similar to unc-13(e1091) animal (Figures 1D,E; Supplementary Figures S1F,G). The result suggests that ASH inhibits the ADL hyperosmotic sensory response.

Given that ASH and ADL are primary and secondary hyperosmolality-sensitive neurons. Genetically ablating neurotransmission in these two types of neurons should display different impacts on the animal’s hyperosmotic avoidance behavior. We thus examined the hyperosmotic avoidance in transgenic animals of ASH::TeTx, ADL::TeTx, and ASH/ADL::TeTx. The expression of TeTx in ASH and ADL was directed by promoters srv-11p (in ASH), ver-2p (2.7 kb, in ADL), and gap-11p (3.3 kb, in ASH and ADL), respectively. Interestingly, genetic inhibition of neurotransmission in ADL, ASH, and ADL/ASH neurons similarly reduced the hyperosmotic avoidance of 0.88 Osm hyperosmolality by glycerol/M13 solution in transgenic animals (Figure 1F) in the wet drop test. This suggests that ASH and ADL may function in the same pathway and form a feedback circuit to regulate hyperosmotic avoidance. Based on the above results, ASH inhibits ADL Ca²⁺ responses to the medium hyperosmolality of 0.88 Osm. ADL in a possible feedback circuit may upregulate ASH Ca²⁺ responses to hyperosmolality. Indeed, genetically eliminating ADL neurotransmission by TeTx reduced ASH somal Ca²⁺ transients in response to the 0.88 Osm hyperosmolality stimulation (Figures 1G,H; Supplementary Figure S1H), suggesting that ADL excites ASH.

The above results suggest that an interaction between ASH and ADL sensory neurons regulates hyperosmotic sensation and avoidance behavior in C. elegans. In this interaction, ADL excites ASH; in contrast, ASH inhibits ADL.

What is the mechanism for ADL exciting ASH? ADL is glutamatergic, neuropeptidergic, and presynaptic to ASH (White et al., 1986; Li and Kim, 2008; Cook et al., 2019). ASH expresses neuropeptidergic receptors NPR-1, a receptor for FMRF-Like peptides FLP-18 and FLP-21, but not known glutamatergic receptors.¹ Thus, ADL likely excites ASH by neuropeptide signaling. We first used egl-3 and unc-31 knocked-down animals by ADL-specific RNA interference (RNAi) to identify the role of neuropeptide signaling in ADL regulating ASH. The egl-3 gene encodes a homolog of mammalian proprotein convertase EGL (EGg Laying defective)-3 essential for neuropeptides biosynthesis (Li and Kim, 2008; Nkambeu et al., 2019). The unc-31 is necessary for neuropeptide release from dense cored vesicles (Avery et al., 1993; Lin et al., 2010). As expected, ADL-specific RNAi knockdown of unc-31 and egl-3 significantly reduced the avoidance behavior as strongly as TeTx-mediated ablation of ADL neurotransmission (Figure 2A). We then used Ca²⁺ imaging to examine the effects of ADL::TeTx, ADL::unc-31(RNAi), and ADL::egl-3(RNAi) genetic manipulations on ASH Ca²⁺ responses to the 0.88 Osm stimulation. ADL::TeTx, ADL::unc-31(RNAi), and ADL::egl-3(RNAi) similarly decreased ASH Ca²⁺ responses to the hyperosmolality (Figures 2B,C; Supplementary Figure S2A). The data suggest that neuropeptide/s released from ADL mediate/s ADL excitation of ASH. ADL expresses FLP-4, FLP-21, neuropeptide-like peptides NLP-7, NLP-8, and NLP-10 (see footnote 1; Li and Kim, 2008). Thus, we screened the neuropeptide/s involved in the hyperosmotic avoidance regulation and the interaction between ADL and ASH, using flp-4, flp-21, nlp-7, nlp-8, and nlp-10 mutant. Among these mutant animals, only the flp-4 mutant animal displayed significantly reduced hyperosmotic avoidance compared with the WT N2. The flp-4 genetic rescue expression in its expression neurons and ADL alone, driven by the flp-4 promoter and ver-2 promoter, fully restored the WT behavioral phenotype (Figure 2D). In contrast, flp-21(ok889) mutant showed augmented hyperosmotic avoidance. However, the behavioral phenotype could not be restored to the WT by the reconstitution of flp-21 genomic DNA driven by its promoter of 4.1 kb flp-21p (Macosko et al., 2009) and ver-2p of 2.7 kb (Figure 2D). The results suggest that FLP-4 may mediate ADL excitation of ASH.

NPR-4 and NPR-5 are known FLP-4 receptors (Cohen et al., 2009; Frooninckx et al., 2012). The gene npr-5 displays trace expression in ASH neurons by single cell RNA sequencing.² We thus used npr-4 and npr-5 mutant animals to test animals’ hyperosmotic avoidance. Loss-of-function (lof) mutant animals of npr-4(tm1782) and npr-5(ok1583) showed significantly decreased hyperosmotic avoidance compared to WT N2 (Figure 2E; Supplementary Figure S2B). npr-4 and npr-5 rescue expression in its expression cells driven by its promoters npr-4p and npr-5p were able to restore WT hyperosmotic avoidance, respectively (Figure 2E; Supplementary Figure S2B). However, only npr-5 genetic rescue in ASH restored the WT hyperosmotic avoidance. Furthermore, animals of the flp-4(sy1606), npr-5(ok1583), and double knockout (DKo) of flp-4(sy1606); npr-5(ok1583), displayed almost similar defects in hyperosmotic avoidance. Only double gene rescued animal of ADL::flp-4; ASH::npr-5; DKo restored WT behavior; both single gene rescue animals of ADL::flp-4; DKo and ASH::npr-5; DKo phenotypically copied animals of single flp-4 and npr-5 mutant and the double mutant (Figure 2E). The results suggest that FLP-4 and NPR-5 function in the same signal pathway to regulate hyperosmotic avoidance.

As FLP-4/NPR-5 signaling between ADL and ASH mediates hyperosmotic avoidance upregulation, it should act to augment the ASH Ca²⁺ responses to hyperosmolality. As expected, flp-4(sy1606) and npr-5(ok1583) animals showed similar decreased ASH Ca²⁺ responses to the stimulation of 0.88 Osm glycerol/M13 solution. In contrast, ADL::flp-4; flp-4 and ASH::npr-5; npr-5 rescue animals showed WT Ca²⁺ responses (Figures 2F,G; Supplementary Figure S2C). In addition, chemogenic inhibition of ADL significantly reduced ASH Ca²⁺ responses to medium hyperosmolality of 0.88 Osm (Figures 2H,I; Supplementary Figure S2D).

The above results support that ADL positively regulates C. elegans’ avoidance of medium hyperosmolality and augments ASH sensory response via the FLP-4/NPR-5 signaling pathway.

Our results indicate that ASH inhibits ADL Ca²⁺ responses to medium hyperosmolality of 0.88 Osm glycerol/M13, and that there is an interaction between ASH and ADL nociceptive neurons. ADL upregulated ASH sensory response to the medium hyperosmolality by FLP-4 signaling. ASH inhibits ADL hyperosmotic sensory response. However, the underlying mechanism still needs study. ASH is a glutamatergic and neuropeptidergic neuron. It is known to release glutamate and several neuropeptides, FLP-21, INS (INSulin related)-1, NLP-3, and NLP-15, for neurotransmission (see footnote 1; Li and Kim, 2008; Serrano-Saiz et al., 2013). Thus, we identified the neurotransmitter/s, neuropeptide/s or/and glutamate, by which ASH generates avoidance of hyperosmotic shock and downregulates ADL activity. We first used mutant animals of neuropeptides released by ASH and EAT-4 to assay hyperosmotic avoidance behavior. EAT-4 is a vesicular L-glutamate transporter essential for filling glutamate into synaptic vesicles and, thus, glutamatergic neurotransmission (Lee et al., 1999; Bellocchio et al., 2000). Among mutant animals of ins-1, nlp-3, nlp-15, and flp-21, only the flp-21(ok889) displayed augmented hyperosmotic avoidance that contradicts the expected phenotype. In addition, the behavioral phenotype was not restored to the WT in transgenic animals expressing flp-21 in ASH/ADL and ASH/ASI (Figure 3A). Thus, ASH is not likely to play its role by neuropeptidergic signaling. It may act via glutamatergic signaling. As expected, our data showed that the eat-4(ky5) animals were less sensitive to medium hyperosmolality of 0.88 Osm. Although eat-4 rescue expression in its expression cells driven by its promoter eat-4p (5.6 kb) restored WT hyperosmotic avoidance in the transgenic animal, eat-4 re-expression in ASH/ASI and ASH/ADL showed no effect on the behavior (Figure 3B). eat-4 is widely expressed in the C. elegans nervous system, including sensory neurons, interneurons, and motor neurons.³ It sounds reasonable that eat-4 reconstitution in ASH alone cannot restore the WT hyperosmotic avoidance behavior. We thus used ASH-specific RNAi knockdown of eat-4 by cell-selective promoters, srv-11p, gpa-11p, and sra-6p, to identify the role of glutamate signaling in hyperosmotic avoidance. As expected, the genetically knocked-down animals of ASH (driven by srv-11p of 1.9 kb)::eat-4(RNAi), ASH/ADL::eat-4(RNAi), and ASH/ASI::eat-4(RNAi) showed a significantly increased avoidance ratio (Figure 3B). All three transgenic animals displayed similar behavioral phenotypes, supporting ASH acting in hyperosmotic avoidance through glutamate signaling.

The inconsistency of behavioral phenotype in animals of eat-4 genetic rescue and knockdown in ASH suggests that glutamatergic sensory neurons besides ASH and ADL may be involved in hyperosmotic avoidance. We then used TeTx neurotransmission inhibition to screen glutamatergic sensory neurons preliminarily. TeTx neurotransmission inhibition or elimination in AWC, ASE, ASG, AQR, and PQR, but not in ASK, BAG (BAG-like Dendritic Ending), ASJ (Amphid Single Cilium J), ADF (Amphid Dual Ciliated Ending F), AFD (Amphid Finger-like Endings D), and AWB (Amphid Wing Neuron B), reduced hyperosmotic avoidance (Figure 3C). We further used neuron-specific knockdown of eat-4 to confirm the result of genetic neurotransmission inhibition. As shown in Figure 3D, the eat-4 knockdown in the candidate sensory neurons identified by TeTx manipulation diminished hyperosmotic avoidance. The results indicate that in addition to ASH and ADL, multiple sensory neurons are involved in hyperosmotic sensation and avoidance behavior.

We focused the present study on the mechanism of interaction between ADL and ASH and the function of this interaction. We next identified the glutamatergic receptor/s involved in ASH regulation of hyperosmotic avoidance. Among the mutant animals we tested, nmr-2(ok3324) and nmr-2(tm3785) displayed notably augmented hyperosmotic avoidance. The reconstitution of nmr-2 in its expression cells driven by its promoter nmr-2p (4.9 kb) restored the WT behavioral phenotype (Figure 3E). NMR-2 is an NMDA glutamate receptor (Kano et al., 2008). It is expressed in sensory neurons, motor neurons, and interneurons RIM (see footnote 3). ASH may use the same neurotransmitter for functioning in animal hyperosmotic avoidance and negative regulation of ADL. ASH is postsynaptic but not presynaptic to ADL. ADL expresses a tyramine receptor TYRA-3, an octopamine receptor SER (SERotonin/octopamine receptor)-6, but no glutamate receptor (see footnotes 1, 3). Thus, ASH may inhibit ADL activity through the relay by intermediate neuron/s. RIM expresses glutamatergic NMR-2 and releases tyramine and glutamate (see footnote 1). Therefore, we focused on testing RIM. We used nmr-2 genetic rescue and RIM-specific TeTx neurotransmission inhibition to test the RIM function in hyperosmotic avoidance and ADL activity. As expected, the RIM-specific (driven by gcy-13p of 2.3 kb) nmr-2 rescue and TeTx expression restored the WT behavior and displayed augmented hyperosmotic avoidance, respectively (Figure 3E). We further employed RIM::chemogenetic inhibition to examine RIM function in adult animal behavior. Specific RIM inhibition by expressing HisCl1 and applying 10 mM histamine made the animal more sensitive to hyperosmolality by 0.88 Osm glycerol/M13 than the WT (Supplementary Figure S3A). These results support that the NMR-2 signaling in RIM interneurons mediates ASH function in hyperosmotic avoidance and possibly downregulation of ADL activity.

We next examined RIM somal Ca²⁺ responses to 0.88 Osm glycerol/M13 solution in eat-4 and nmr-2 mutant and genetically rescued animals to confirm glutamate/NMR-2 signaling for neurotransmission from ASH to RIM. The RIM Ca²⁺ responses to 0.88 Osm osmolality in the eat-4(ky5) animal significantly decreased. The signals were fully restored to the WT by the extrachromosomal re-expression of eat-4 cDNA in ASH, using two cell-specific promoters, gpa-11p (in ASH and ADL) and sra-6p (in ASH and ASI) with an expression overlap is ASH. However, reconstitution of eat-4 in ADL, which is postsynaptic to RIM, did not rescue the RIM Ca²⁺ responses to 0.88 Osm osmolality (Figures 3F,G; Supplementary Figure S3B). In addition, the 0.88 Osm hyperosmolality-evoked RIM Ca²⁺ responses in the nmr-2(ok3324) significantly decreased and restored to the WT in the RIM-specific nmr-2 rescued animal (Figures 3H,I; Supplementary Figure S3C). There is an inconsistency between behavioral assay (Figure 3B) and calcium imaging (Figures 3F,G) results in animals of eat-4 rescue in ASH/ADL or ASH/ASI. This inconsistency is expected. Because multiple glutamatergic neurons are required for normal hyperosmotic avoidance, and ASH activates its postsynaptic RIM neurons by chemical synapse connection, the process does not need other glutamatergic neurons. These data support that RIM relays ASH function in animal hyperosmotic avoidance and possibly ASH down-regulation of ADL.

The above results suggest that RIM interneurons relay ASH inhibition to ADL. RIM is tyraminergic, and ADL expresses the tyraminergic receptor TYRA-3 (see footnotes 1, 3). Thus, RIM modulates ADL activities, possibly through tyramine/TYRA-3 signaling. Tyramine is a monoamine neuromodulator (Alkema et al., 2005). It is synthesized from tyrosine catalyzed by tyrosine decarboxylase TDC-1. We thus used tdc-1 mutant and RIM-specific knocked-down animals to examine the functions of tyraminergic signaling in hyperosmotic avoidance and ADL Ca²⁺ responses to hyperosmotic stimulation. The tdc-1(n3419) animal displayed significantly augmented avoidance of 0.88 Osm glycerol/M13 solution. This behavioral phenotype was fully restored to the WT by tdc-1 genetic rescue expression in its expression neurons (RIM and RIC) and RIM alone, driven by tdc-1 promoter (3.0 kb) and gcy-13p (2.3 kb), and by administration of exogenous tyramine (5 mM). Furthermore, the RIM-specific tdc-1 knockdown showed the same behavioral effect as the tdc-1 mutation (Figure 4A; Supplementary Figure S4A). This result supports that RIM modulates hyperosmotic avoidance by releasing tyramine. However, the postsynaptic tyraminergic receptor/s need/s to be identified. Caenorhabditis elegans expresses tyraminergic receptors SER-2, TYRA-2, TYRA-3, and LGC (Ligand-Gated ion Channel)-55. We thus used the receptor mutant animals to test which receptor/s function/s in animal hyperosmotic avoidance. Among ser-2, tyra-2, tyra-3, and lgc-55 mutant animals, only the tyra-3(ok325) showed non-WT behavior, a significantly augmented hyperosmotic avoidance (Supplementary Figure S4B). The hyper-avoidance was restored to the WT by tyra-3 rescue expression in its expression cells and ADL alone, but not in ASK, AWC, and BAG neurons, or neuronal sets ADE (Anterior DEirid Neuron)/CEP (CEPhalic Sensory Neuron)/PDE (Posterior DEirid) and AIM (Anterior Interneuron M)/AFD (Amphid Dual Ciliated Ending F; Figure 4B). These results indicate that RIM regulates C. elegans hyperosmotic avoidance by neurotransmission to ADL through tyramine/TYRA-3 signaling.

RIM may use the same signaling pathway to regulate ADL activities as it does to modulate hyperosmotic avoidance. We first examined the tyramine role in ADL Ca²⁺ responses to 0.88 Osm glycerol/M13. ADL Ca²⁺ signals in tdc-1(n3419) and RIM::TeTx neurotransmission-eliminated animals were augmented similarly. The increased Ca²⁺ response in the mutant animal was restored to the WT by the RIM-specific tdc-1 genetic rescue and the treatment with 50 μM exogenous tyramine (Figures 4C,D; Supplementary Figure S4C). This suggests that RIM inhibits ADL through tyramine signaling. Logically, ADL Ca²⁺ responses to 0.88 Osm glycerol/M13 should change similarly in tyra-3(ok325) and tdc-1(n3419) animals. As expected, the ADL Ca²⁺ signals in the tyra-3(ok325) significantly increased and were restored to the WT by ADL::tyra-3 genetic rescue but not by the application of exogenous tyramine (Figures 4E,F; Supplementary Figure S4D). In addition, RIM-specific chemogenetic inhibition increased the ADL Ca²⁺ signals (Figures 4G,H; Supplementary Figure S4E). These results indicate that RIM interneurons inhibit ADL sensory neurons by tyramine/TYRA-3 signaling.

We demonstrated that ASH excites RIM by glutamate/NMR-2 signaling, RIM inhibits ADL via the neurohumoral tyramine/TYRA-3 pathway, and ADL excites ASH through FLP-4/NPR-5 signaling. Three types of neurons form a negative feedback circuit. Then, RIM should decrease Ca²⁺ responses to hyperosmolality in ASH as in ADL. We thus employed an experimental strategy like the one we used for examining the RIM effect on ADL to test the RIM regulation of ASH. Indeed, the changes in ASH Ca²⁺ responses to 0.88 Osm glycerol/M13 in tdc-1 and tyra-3 mutant, the genetically rescued, and tyramine-treated animals, almost entirely copied the changes in ADL Ca²⁺ signals (Figures 4I–L; Supplementary Figures S4F,G). These results support that ASH, RIM, and ADL form a negative feedback circuit. In this circuit, ASH excites RIM; RIM inhibits ADL, and ADL excites ASH. Thus, ASH exciting RIM removes or reduces ADL exciting ASH. This neuronal signal integration modality is another illustration of disexcitation reported in our previous study (Liu et al., 2019).

In the natural habitat of C. elegans, the osmolality changes are usually gradual instead of acute like in the drop test. Furthermore, environmental hyperosmolality may increase nematode internal osmolality. Caenorhabditis elegans senses osmotic upshifts via signaling that requires the cGMP-gated sensory channel subunit TAX-2 in body cavity sensory neuron URX (Unknown Receptor, not Ciliated X) and that generates increased aversion behavior (Yu et al., 2017). We performed a similar droplet assay used by this previous study to test whether the ASH/RIM/ADL circuit function in the behavioral response of internal osmolality upshifts (Yu et al., 2017). The WT N2 animal displayed a rapid increase of aversion behavior in 1 min in response to hyperosmolality (in Osm) of 0.41, 0.49, 0.62, 0.74, and 0.88 by glycerol/M13 solutions and was quickly paralyzed by immersion in solutions of higher hyperosmolality of 0.74 Osm and 0.88 Osm (Supplementary Figures S5A,B). We further used transgenic animals of ASH::TeTx, ADL::TeTx, and RIM::TeTx to test hyperosmotic behavioral response to the treatment of 0.62 Osm glycerol/M13. Our data showed that all transgenic animals responded to the hyperosmolality treatment similarly to the WT N2 (Supplementary Figures S5C,D). This result does not support that ASH/RIM/ADL circuit functions in aversive behavior responding to the upshift of internal osmolality.

So far, our results indicate that ASH down-regulates ADL sensory responses to and C. elegans avoidance of medium osmolality through the ASH/RIM/ADL negative feedback circuit. Since ASH is the main osmolality-sensitive sensory neuron, it should generate and regulate hyperosmotic avoidance via other neuronal circuits. ASH synaptically connects with forward and backward command interneurons, such as AVA (Anterior Ventral Process A), AVB (Anterior Ventral Process B), AVD (Anterior Ventral Process D), and AVE (Anterior Ventral Process E). ASH connects with interneurons controlling or modulating locomotion by chemical and electric synapses (White et al., 1986; Cook et al., 2019).⁴ Among the interneurons postsynaptic to ASH, RIC regulates the avoidance of heavy metal ions Cu²⁺ (Guo et al., 2015, 2018) and aversive odor 2-nonanone (Kimura et al., 2010) by neurohumoral octopamine. ASH may regulate hyperosmotic avoidance through interneuron RIC. We thus evaluated the RIC function in hyperosmotic avoidance.

We first eliminated RIC neurotransmission and inhibited its activity by RIC::TeTx and RIC::chemogenetic inhibition to test RIC’s role in hyperosmotic avoidance. These two types of neuronal manipulations similarly reduced animal hyperosmotic avoidance, supporting to the RIC enhancement effect on the behavior (Figure 5A). RIC should display a response to hyperosmolality, possibly by excitation from ASH through electrical synapses. We next used Ca²⁺ imaging to assay the RIC response. As expected, RIC displayed robust Ca²⁺ responses to hyperosmolality of 0.88 Osm in the WT N2 animal (Figures 5B,C; Supplementary Figure S6A). ASH and RIC express the innexin family of gap junction proteins INX-4 (in both neurons) and UNC-9 (in RIC; Altun et al., 2015; see footnote 4). A strong TeTx expression in ASH disrupts ASH’s gap junctions and, thus, RIC excitation in response to noxious Cu²⁺ stimulation (Guo et al., 2015). If the RIC hyperosmotic response resources directly from ASH by gap junctions, it should majorly decrease or even disappear in the mutant animals of these two innexin proteins and ASH::TeTx transgenic worm. As expected, the RIC Ca²⁺ responses to the medium hyperosmolality almost disappeared in unc-9(fc16), inx-4(e1128), and ASH::TeTx, and were restored to the WT in transgenic animals of the RIC::unc-9;unc-9 and ASH::inx-4; RIC::inx-4; inx-4 (Figures 5B–G; Supplementary Figures S6A–C). All these results support that ASH excites RIC via gap junction.

RIC is octopaminergic. It may modulate hyperosmotic avoidance through octopaminergic signaling. Octopamine is biosynthesized from tyramine catalyzed by tyramine β-hydroxylase TBH-1 (Chase and Koelle, 2007). We then used tbh-1 mutant and rescued animals to test the octopamine role in hyperosmotic avoidance. The tbh-1(n3247) mutant displayed reduced hyperosmotic avoidance; the RIC-specific tbh-1 rescued and exogenous octopamine-treated tbh-1(n3247) animals showed the WT behavioral phenotype (Figure 5H; Supplementary Figure S6D). Moreover, the RIC-specific tbh-1 knocked-down animal phenocopied the tbh-1 mutant (Figure 5H). Caenorhabditis elegans expresses octopaminergic receptors OCTR-1 (OCTopamine Receptor 1), SER (SERotonin/octopamine receptor)-3, and SER-6. We next employed receptor mutant animals to identify the octopamine receptor/s involved in hyperosmotic avoidance. Ser-6 and octr-1 mutant animals displayed significantly reduced hyperosmotic avoidance. The behavioral defects were eliminated by the gene rescue driven by their promoters (Figure 5I). SER-6 and OCTR-1 are expressed in neurons including RIC and AIY, respectively (see footnote 4). Octopamine from RIC may regulate hyperosmotic avoidance by two signaling pathways: autocrine SER-6 and OCTR-1 in AIY. As expected, the genetic reconstitution of ser-6 in RIC and octr-1 in AIY but no other ser-6-expressing or octr-1-expressing neurons restored the WT behavioral phenotype in transgenic animals (Figure 5I). In addition, AIY::TeTx and AIY::chemogenetic inhibition similarly suppressed animals’ hyperosmotic avoidance (Figure 5J; Supplementary Figure S6E). These behavioral test results support the hypothesis of two signaling pathways.

We next monitored RIC and AIY Ca²⁺ responses to the medium hyperosmolality to support the behavioral conclusion. Without exception, all Ca²⁺ image data were as expected. The hyperosmolality-elicited somal calcium transients of RIC significantly decreased in the tbh-1(n3247) animal and restored to the WT in the RIC-specific gene rescued animal (Figures 5K,L; Supplementary Figure S6F). In the same way, the RIC Ca²⁺ signals changed in the ser-6(tm2146) and RIC-specific ser-6 rescued animals (Figures 5M,N; Supplementary Figure S6G). The AIY Ca²⁺ responses to the hyperosmolality were nearly eliminated in tbh-1(n3247) and RIC::TeTx animals and restored to the WT in the RIC-specifically tbh-1 rescued animal (Figures 5O,P; Supplementary Figure S7A). AIY Ca²⁺ responses in octr-1 mutant and AIY rescued animals changed similarly to the RIC signals in ser-6 mutant and RIC-rescued animals (Figures 5Q,R; Supplementary Figure S7B). In addition, RIC chemogenetic inhibition reduced the AIY Ca²⁺ signals (Figures 5S,T; Supplementary Figure S7C).

All the above results support that ASH, RIC, and AIY form a feedforward circuit to upregulate hyperosmotic avoidance. The signaling pathways in this circuit are direct signal flow from ASH to RIC through electric synapses, autocrine octopamine/SER-6 signaling in RIC, and octopamine/OCTR-1 pathway between RIC and AIY.

Animals of all C. elegans strains were cultured on nematode growth media (NGM) plates at 20°C using E. coli bacteria OP50 as food by standard procedures (Brenner, 1974). The strains were obtained from the CGC⁵ or the National Bio-Resources Project.⁶ All transgenic animals were generated with standard microinjection techniques (Mello et al., 1991). The injection pressure was controlled by a DMP-300 digital pneumatic microinjection pump (Micrology Precision Instruments, Ltd, Wuhan, China). Plasmids were injected at 50 ng/μL together with lin-44p::GFP as a coinjection marker at 10 ng/μL. All strains used in this study are listed in Supplementary Table S1. We used two or three lines of transgenic worms to conduct experiments and summarized all data for statistical analysis.

All expression constructs were generated with a Three-Fragment Multisite Gateway® system (Invitrogen™, Thermo Fisher Scientific, Waltham, MA, United States). Briefly, three entry clones comprising three PCR products (promoter, the gene of interest, and sl2e::TagRFP-t, sl2d::GFP, or unc-54 3’ UTR, in the name of slot1, slot2, and slot3, respectively) were recombined into the pDEST™ R4-R3 Vector II or custom-modified destination vectors using attL-attR (LR) recombination reactions to generate expression clones.

We constructed an “A” entry clone containing a sequence of promoters used in this study by the In-Fusion method. In short, a modified attL4-attR1 entry clone was linearized by PCR. Then, the linearized product and a promoter PCR product were used to generate an “A” entry clone using the ClonExpress®II One Step Cloning Kit (Vazyme Biotech Co., Ltd., Nanjing, China). The PCR products of promoters were amplified from C. elegans genomic DNA with primers containing 15–20 bp sequences that carry some sequence of attL4 and attR1 recombination sites. Alternatively, promoter vectors in the C. elegans Promoters Library (Thermo Fisher Scientific, Waltham, MA, United States) were directly used. The length of each promoter used in this study is as follows: ver-2p 2.7 kb (in ADL), gpa-11p 3.3 kb (in ASH and ADL), sra-6p 3.8 kb (in ASH and ASI), srv-11p 1.9 kb (in ASH), flp-21p 6.6 kb, npr-5p 3.0 kb, npr-4p 4.0 kb, eat-4p 5.6 kb, sra-9p 4 kb (in ASK), flp-17p 3.3 kb (in BAG), srh-11p 0.7 kb (in ASJ), srh-142p 3.5 kb (in ADF), gcy-32p 0.8 kb (in AQR, PQR, and URX), str-1p 4.0 kb (in AWB), str-2p 3.7 kb (in AWC), srsx-3p 0.9 kb (in AWB and AWC), gcy-5p 3.2 kb (in ASER), gcy-7p 1.3 kb (in ASEL), gcy-15p 0.8 kb (in ASG), nmr-2p 4.9 kb, gcy-13p 2.3 kb (in RIM), tdc-1p 3.0 kb (in RIM and RIC), tyra-3p 4.2 kb, dat-1p 0.7 kb (in ADE, CEP, and PDE), gcy-18p 0.8 kb (in AFD and AIM), tbh-1p 4.5 kb (in RIC), ser-6p 3.5 kb, octr-1p 3.9 kb, and ttx-3p 3.1 kb (in AIY, Hobert et al., 1997), flp-4p 3.3 kb, respectively.

BP recombination reactions were used to generate entry clones B (containing a sequence of a tested gene) and C (containing sl2e::TagRFP-t, sl2d::GFP, or unc-54 3’ UTR). The following genes were used to create the B entry clones. The flp-21, npr-4, npr-5, tdc-1, octr-1, and ser-6 genes were amplified from the genomic DNA of wild-type N2 worms. Nmr-2, eat-4, tyra-3, ser-6, and tbh-1 cDNA were amplified by reverse transcription-PCR (RT-PCR) from C. elegans mixed-stage RNA. Their PCR products flanked by attB1 and attB2 were recombined with the pDONR-221 vector containing attP1 and attP2. To generate entry clone C, the BP reaction sites attB2r and attB3 were inserted into the sequences of the sl2e::TagRFP-t, sl2d::GFP, or unc-54 3’ UTR and recombined with the PDONR-P2R-P3 vector. All primers for cloning these promoters and genes are listed in Supplementary Table S2.

Data of avoidance ratio are displayed as box plots, with each dot representing the data from each individual tested animal or each test. The Ca²⁺ signal data are expressed as heatmaps, box plots, or as the means ± SEM indicated by solid traces ± gray shading. Data were statistically analyzed using software packages in GraphPad Prism 8 (GraphPad Software, Inc., San Diego, CA, United States). When the comparison was limited to 2 groups, an unpaired t-test was used to analyze differences and calculate p-values. When more than two groups of data were compared, data were analyzed by ordinary one-way or two-way analysis of variance (ANOVA), with recommended post hoc tests in the GraphPad Prism 8 software package. Dunnett’s multiple comparison correction was applied when multiple samples were compared to a single sample, i.e., wild-type N2 or other controls. Tukey’s multiple comparison correction was used when multiple samples were compared. The p-value is indicated as follows: ns, not significant, ^*p < 0.05, ^**p < 0.01, ^***p < 0.001, ^****p < 0.0001, and in different colors for varied comparisons.