The transcription factor unc-130/FOXD3/4 contributes to the biphasic calcium response required to optimize avoidance behavior

The central neural network optimizes avoidance behavior depending on the nociceptive stimulation intensity and is essential for survival. How the property of hub neurons that enables the selection of behaviors is genetically defined is not well understood. We show that the transcription factor unc-130, a human FOXD3/4 ortholog, is required to optimize avoidance behavior depending on stimulus strength in Caenorhabditis elegans. unc-130 is necessary for both ON responses (calcium decreases) and OFF responses (calcium increases) in AIBs, central neurons of avoidance optimization. Ablation of predicted upstream inhibitory neurons reduces the frequency of turn behavior, suggesting that optimization needs both calcium responses. At the molecular level, unc-130 upregulates the expression of at least three genes: nca-2, a homolog of the vertebrate cation leak channel NALCN; glr-1, an AMPA-type glutamate receptor; and eat-4, a hypothetical L-glutamate transmembrane transporter in the central neurons of optimization. unc-130 shows more limited regulation in optimizing behavior than an atonal homolog lin-32, and unc-130 and lin-32 appear to act in parallel molecular pathways. Our findings suggest that unc-130 is required for the establishment of some AIB identities to optimize avoidance behavior.


Results
A FOXD3/4 ortholog, unc-130, affects the frequency of turn behavior during avoidance optimization. C. elegans tends to show reversal behavior in response to mild osmolarity changes but is biased toward omega turns in a manner correlating with stimulus strength (Fig. 1a) 6 . In our previous paper, we mentioned that unc-130 null mutants exhibited reduced avoidance frequency in response to noxious stimuli 6 . However, it remained unclear whether unc-130 affected the choice of appropriate behavior or simple locomotion. To clarify this issue, we analyzed the avoidance behavior patterns in response to four gradient osmotic concentrations of noxious stimuli. The results clarified that the unc-130 mutants were less likely to take omega turns upon hyperosmotic stimulation than wild-type animals ( Fig. 1b and c; Fig. S1a-h). Instead, the number of reversals increased in the mutants (Fig. 1c). The critical point is that the unc-130 mutants had no significant changes in the sum of omega turn and reversal frequencies, suggesting that they can avoid stimuli but not optimize how to do so ( Fig. 1b and c; Fig. S1a-h).
Meanwhile, such behavioral phenotype left the possibility that the unc-130 mutant was unable to make the omega turn because of its uncoordinated movements. We hypothesized that the omega turn might occur by a more intense stimulus. If this were the case, we could conclude that the uncoordinated movements were not the reason for the inability to turn. In the drop test, it was difficult to use a sorbitol solution thicker than 6 M due to the viscosity. Therefore, to investigate whether the unc-130 mutant can mediate the omega turn as well as the wild-type animals, we established and performed an original harsh tap assay with a platinum wire. In this Figure 1. unc-130 is necessary for turning behavior during optimization of avoidance behavior. (a) A schematic of optimization of avoidance behavior in C. elegans. During exposure to a mild osmotic stimulus, the animal exhibits a reversal behavior, such as body retraction or backward movement. Strong stimulation induces body bending, causing an omega turn so that the animal can return to its original location 26 . (b) In wild-type C. elegans, the total frequency of avoidance increased with increasing sorbitol concentration (P < 0.001 in 0 M vs. 2-6 M, 1 M vs. 2-4 M, n = 11, 12, 15, 17, 20 (0-6 M, respectively)). In particular, the omega turn frequency was significantly higher under 2-6 M compared to 0 M, under 4 M compared to 1 M, and under 6 M compared to 1 M (P < 0.001, P < 0.01, P < 0.001). (c) Nematodes with a null mutation in the unc-130 had a lower omega turn frequency than wild-type animals (Fig. 1b) (P < 0.05 in 1 M and 6 M; n = 12, 12, 14, 20). On the other hand, the total avoidance rate (sum of reversals and omega turns) was similar to that of wild-type animals because of the increase in reversal frequency in all conditions (P > 0.05). (d) A schematic of the harsh tap assay. A platinum wire placed in the forward direction is lightly applied to the tip of the nematode's nose. e unc-130 mutants showed the omega turn rate comparable to the wild-type animals. P = 0.392 (n = 10, 10, t-test). (f) Schematic of the analysis method using channelrhodopsin 2 (ChR2(H134)). ChR2 (H134) was selectively expressed in primary nociceptive sensory neurons (ASH sensory neurons) 27 . Exposure to blue light can selectively activate ASH sensory neurons and their downstream neurons. (g) Adding all-trans retinoic acid (ATR) and blue light triggered turn behavior. unc-130 mutants had a significantly lower turn frequency than the wild-type animals. *** indicates P < 0.001, * indicates P < 0.05 (n = 13,14,14,14). (h) Self-promoter-driven UNC-130 expression rescued the decreased turn frequency of unc-130 mutants. See Fig. S1i-l. *** indicates P < 0.001, * indicates P < 0.05, ns indicates P > 0.05 (n = 6, 6, 4). All statistical analyses were performed by Kruskal-Wallis test with Dunn's multiple comparisons test as a post hoc test. The error bars in the figures represent the ± SEM. www.nature.com/scientificreports/ experiment, we tapped nematodes on the tip of their nose with the platinum wire on NGM plates, and observed their avoidance behavior (Fig. 1d). As a result, the unc-130 mutant also mediated omega turns in an average of 84.0%, and there was no statistically significant difference compared to the wild-type animals, which mediated omega turns in 89.0% (P = 0.392, n = 10, 10, t-test) (Fig. 1e). This data clearly shows that unc-130 mutants have the physical ability to perform omega turn. Based on the above results, we concluded that unc-130 affects the behavioral choice. The behavioral phenotype in the null mutants corresponds to "hypoesthesia" in psychological terms 28 .
In the drop test, a sorbitol solution accompanying touch or mechanical stimuli might trigger sensory neurons other than ASHs. To confirm whether the optimization defect in unc-130 mutants is dependent on noxious stimuli, we observed the behavior induced by selective excitation of ASH sensory neurons using channelrhodopsin-2 ( Fig. 1f) 29 . Optogenetic excitation separate from touch or mechanical stimuli reproduced the optimization defect (P < 0.001, n = 13, 14, 14, 14, Kruskal-Wallis test with Dunn's multiple comparisons test as a post hoc test) (Fig. 1g), suggesting that unc-130 optimizes avoidance patterns in response to the strength of the noxious stimuli. UNC-130 driven by its promoter rescued the behavior of the null mutants, indicating that the optimization defect was due to the loss of unc-130 gene function (P < 0.05, n = 6, 6, 4, Kruskal-Wallis test with Dunn's multiple comparisons test as a post hoc test) ( Fig. 1h; Figs. S1i-l and S2a-e).

unc-130 is necessary for biphasic calcium responses of the central neurons for behavioral optimization.
Previously, we showed that AIB neurons are the central neurons for turning and the biphasic response involving a calcium reduction during osmotic stimulation and an increase after stimulation 6 . Considering that unc-130 expression is relatively restricted in cell lineages including AIB cells (ABplaapa, ABpraapa), during embryogenesis 30 , we hypothesized that the optimization defect of unc-130 mutants likely resulted from AIB neuronal dysfunction. To test this hypothesis, we attempted to analyze AIB neural responses, including calcium reductions during stimulation, using the calcium indicator inverse-pericam 2.0 (IP2.0), whose fluorescence intensity increases as the calcium concentration decreases 31 . As we expected, the fluorescence intensity of IP2.0 increased during osmotic stimulation and decreased after stimulation in wild-type animals, showing that AIBs received inhibitory input during stimulation and excitatory input after the stimulus was removed ( Fig. 2a and Fig. S3a). In contrast, the unc-130 mutants showed no responses (Fig. 2a). Baseline fluorescence values before correction was close to minimum IP2.0 fluorescence values ( Fig. S3bc and c) (P = 0.077, n = 14, 18, t-test), implying relatively high resting calcium level in both wild-type and unc-130 animals (P > 0.05, n = 14, 18, t-test). These data suggest that reduced responses in AIB neurons cause the optimization defect in unc-130 mutants.
Increases in calcium in AIBs induce turning 6,18 , but it remains unclear whether the AIB ON response (calcium decrease) contributes to the turning behavior. In C. elegans, ASH sensory neurons indirectly inhibit AIBs via predicted AIA inhibitory interneurons (Fig. 2b) 32 . Therefore, we analyzed whether ablation of AIAs reduces turning. Animals in which AIAs were ablated by caspase CED-3-mediated cell death 33 exhibited significantly decreased turn frequencies (P < 0.01, n = 12, 8, t-test), and AIA-ablated unc-130 mutants exhibited indifferent turn frequencies ( Fig. 2c; Fig. S3d). These results imply that AIA neurons are required for turning, potentially via inhibition of AIBs during the stimulus.
We compared the responses in ASHs using the calcium indicator GCaMP6s 34,35 to confirm that the other neuron responses in the avoidance circuit remained unaltered, and we found no differences between unc-130 mutants and wild-type animals (Fig. 2d). This result supports the occurrence of "secondary hypoesthesia" based on a central neuron defect rather than "primary hypoalgesia, " a simple decrease in perception. We also analyzed the calcium responses of the AIAs using GCaMP6s. Since calcium responses in AIAs to noxious osmolarity stimuli have not yet been reported, we started by observing the detailed patterns in wild-type animals. We found increased calcium concentrations in AIA neurites during osmotic stimulation ( Fig. 2e; Fig. S3e). In previous studies, calcium imaging has been performed in AIA neurites 36 . One reason is that AIA neurites, but not the cell bodies, show large calcium fluctuations. Fortunately, we were able to observe calcium responses in AIA cell bodies using our original transgenic strain (Fig. S3f). We found that AIA cell bodies show a triphasic pattern: a sustained increase in calcium concentration during stimulation, a decrease for 30 s after the end of stimulation, and a subsequent increase (Fig. S3f). This response pattern appears to correlate more strongly with calcium decrease responses in AIB neurites than with responses in AIA neurites (Fig. S3a, e and f). Considering that the AIA-AIB synapse is a predicted inhibitory circuit, the responses in the AIA cell body might affect the biphasic .0 is an indicator to monitor the calcium decrease. In wild-type animals, an ON response (calcium decrease) occurred during stimulation (5-35 s; light blue area in the figure) and was followed by an OFF response (calcium increase) immediately after the stimulus (n = 16). In unc-130 mutants, both of these ON and OFF calcium responses were absent (n = 15). (b) Detailed schematic of the downstream neural circuit of ASH sensory neurons. ASH sensory neurons sense osmotic stimuli at receptors and directly excite two types of interneurons (AIBs and AIAs) via glutamate release. Cholinergic AIAs inhibit calcium induction of downstream AIBs. AIBs release glutamate or signal via electrical synapses to excite downstream motor/interneurons, resulting in turn behavior. (c) Cell ablation of inhibitory AIA interneurons resulted in reduced turn frequency. ** indicates P < 0.01 (n = 12, 8, t-test). (d) Analysis of calcium responses using G-CaMP6s in ASH sensory neurons (n = 20,19). (e) Calcium imaging of AIA neurites using G-CaMP6s. There was no significant difference in the peak of ON calcium increase in the wild-type animals compared with unc-130 mutants (P = 0.892, n = 21, 23, t-test). (f) Calcium imaging of AVA neurons, one of the types of downstream AIB neurons. Wild-type animals showed a small ON calcium increase during stimulation and a large increase after stimulation (n = 20). In the unc-130 mutants, OFF calcium responses were abolished (n = 19). The area in orange color is enlarged in the right view. The error bars in these figures represent the ± SEM values. www.nature.com/scientificreports/ response in the AIB. Then, we observed the AIA reaction in the unc-130 mutants and clarified that there was no statistically significant variation in peak ΔF/F amplitude (P = 0.892, n = 21, 23, t-test), although the response was slightly dampened in mutants compared within wild-type animals (Fig. 2e). Finally, to understand the output function of AIBs on the avoidance circuit, we observed the calcium response in AVAs. AVAs receive input from the presynaptic AIB neurons (Fig. 2b). As expected, unc-130 mutants showed remarkably impaired OFF responses (calcium increases) in AVAs (Fig. 2f). In our previous paper, we discussed how low ON responses (calcium increases) in AVAs during osmotic stimulation might induce reversal behavior independent of AIBs 6 . Since the unc-130 mutants can perform reversal behavior ( Fig. 1c; Fig. S1f and g), we explored whether the AVA ON response (calcium increase) remains in the unc-130 mutants. The unc-130 mutants showed an ON response (calcium increase), although it was reduced in size (see the enlarged view of the area in Fig. 2f, inset), consistent with our hypothesis. These results suggest that AIBs are critical neurons for determining appropriate behaviors in response to various stimuli.
The inhibitory synaptic receptors on AIBs have not yet been identified. Since AIAs are cholinergic neurons 37 , we first considered the possibility of the downregulation of inhibitory acetylcholine-gated chloride channels expressed on AIBs in the unc-130 mutants. We observed the expression patterns of six hypothetical acetylcholinegated or ligand-gated chloride channels to focus on the inhibitory channels expressed on AIB neurons. In this experiment, we found that all unc-130 mutant animals expressed innexin 1 (inx-1) on AIBs (n = 25); we used the inx-1 promoter to mark AIBs and identify AIB-specific gene expression. AIBs expressed acc-1 (Acetylcholinegated Chloride Channel 1) 38 , lgc-46 (Ligand-Gated ion Channel 46, predicted to have chloride channel activity), and lgc-49 (Ligand-Gated ion Channel 49, predicted to have chloride channel activity) 39 promoter-driven GFP (Fig. S4a). However, AIBs did not express acc-2, acc-3, or acc-4 (Acetylcholine-gated Chloride Channel 2, 3, or 4, respectively) 38 (Fig. S4b), nor did they express lgc-47 or lgc-48 (Ligand-Gated ion Channel 47 or 48, respectively, predicted to have chloride channel activity) 10 . We analyzed the turn behavior of deletion mutants of acc-1, lgc-46, and lgc-49, but there were no differences between the single or double mutants and wild-type animals ( Fig. S4c and d).
We next assumed a contribution of G protein-coupled acetylcholine receptors. We observed turn behavior in deletion mutants of goa-1 (G protein O, alpha subunit 1; an ortholog of human GNAO1 10 ), which exhibits G protein-coupled acetylcholine receptor activity 40 . In C. elegans, only goa-1 encodes a member of the mammalian Gi/o class of Gα subunits, and the predicted amino acid sequence of C. elegans GOA-1 is over 80% identical to that of mammalian Gαo 40 . Hypothetical goa-1 null mutants showed a lower turn frequency (13.3 ± 4.41%) than wild-type animals (56.3 ± 5.96) (P < 0.001, n = 9, 8, t-test) (Fig. S5a), but the expression levels of goa-1 promoterdriven GFP in AIBs were comparable in both strains ( Fig. S5b and c) (P = 0.450, n = 10, 9, t-test). Thus, we concluded that the reduction in turning was independent of the goa-1 function in AIB neurons. In C. elegans, there are three types of muscarinic-type acetylcholine receptors: gar-1 (G-protein-linked Acetylcholine Receptor 1) 41 , gar-2 42 , and gar-3 43 . Although we analyzed the behavior of triple gene mutants 44 , turn frequency was not significantly lower in these animals than in wild-type animals ( Fig. S5d) (P = 0.226, n = 9, 8, t-test). Consequently, we could not identify the AIA-AIB inhibitory receptors, but the possibility remains that an unknown G-protein coupled acetylcholine receptor participates. unc-130 upregulates the expression of a predicted cation channel. unc-130 mutants showed no stimulus-dependent OFF responses (calcium increases) (Fig. 2a). Therefore, we tested whether the turn behavior could be recovered by cation flux through ChR2(H134R) selectively expressed in AIBs. Contrary to our expectations, turn behavior induction was minor in unc-130 mutants (Fig. 3a). In optogenetics, cation flux is reported to be milder than that under natural stimulation 45 . Therefore, we proposed three hypotheses: 1. that there is dysfunction in the calcium signaling pathway in AIBs, 2. that there are reductions in presynaptic output from AIBs, and 3. AIBs are developmentally altered in unc-130 mutants.
First, we tested the former. There are eight of nine predicted voltage-gated calcium channel subunits in C. elegans 46 . We excluded three genes from our analysis. One of them, ccb-1 (Calcium Channel, Beta subunit 1, predicted to have high voltage-gated calcium channel activity), is not expressed in AIBs 46 . We performed a drop test for mutants of the other seven channels and found that only nca-2 (putative Nematode CAlcium channel 2, a homolog of the vertebrate cation leak channel NALCN) mutants showed a lower turn frequency in response to a 4 M sorbitol drop than wild-type animals ( Fig. 3b; Fig. S6a and b) (P < 0.001, n = 13, 11, t-test). The mutants for Figure 3. unc-130 upregulates the expression of putative cation channels in AIB neurons. (a) All-trans retinoic acid and blue light induce turning. The turning frequency was 77.9% in the wild type but only 15.3% in unc-130 mutants. *** indicates P < 0.001, * indicates P < 0.05 (n = 14, 14, 17, 15, Kruskal-Wallis test with Dunn's multiple comparisons test as a post hoc test). (b) nca-2 mutants showed significantly reduced turn frequencies in response to 4 M sorbitol than wild-type animals. *** indicates P < 0.001 (n = 13, 11, t-test). (c) Expression of nca-2 promoter-driven GFP in AIB cell bodies (arrowheads). inx-1 promoter-driven mCherry was coexpressed as a cell identification marker for AIBs. nca-2 showed weaker AIB expression in a typical unc-130 animal than in a typical wild-type animal. Scale bar = 5 µm. (d) The ratiometric analysis of the intensity of nca-2 promoterdriven GFP expression was significantly lower in AIBs of unc-130 mutants than in those of wild-type animals. *** indicates P < 0.001 (n = 10, 11, t-test). (e) AIB-selective expression of NCA-2a rescued the decreased turn frequency of nca-2 mutants. ** indicates P < 0.01 (n = 12, 12, 12, one-way ANOVA followed by Tukey's post hoc test). See Fig. S7a- 47 , showed no reductions in turn frequency (Fig. S6b). However, it should be noted that nca-1 (gk9) and ccb-2 (ok862) provide no evidence for null alleles only because they are predicted to have hypothetical reduced expression and partial loss of a domain, respectively (Fig. S13). So we could not completely exclude the possible roles of these channels.
To uncover whether unc-130 regulates nca-2 expression in AIBs, we quantified GFP intensity driven by the nca-2 promoter. The results showed that AIBs expressed nca-2, but the unc-130 mutants showed significantly reduced fluorescence intensity compared with wild-type animals ( Fig. 3c and d). The neural subsets expressed nca-2. To test whether reduced nca-2 expression in the AIBs correlates with reduced turn frequency, we performed a rescue experiment with the AIB-selective expression of the NCA-2a protein. The results revealed that the defect in nca-2 mutants was substantially alleviated such that the mutants recovered an optimization pattern relatively similar to that of wild-type animals ( Fig. 3e; Fig. S7). Thus, we conclude that nca-2 expression in AIBs is necessary for avoidance behavior optimization, suggesting a voltage-gated cation channel or a cation leak channel expression in the AIBs may upregulate synaptic transmission from AIBs to downstream inter-/motor neurons (Fig. 3f).
unc-130 upregulates the expression of functional molecules for glutamatergic synapses. Transcription factors frequently determine neural identity by regulating the expression of multiple genes 46 . To determine whether unc-130 contributes to the excitatory input from upstream neurons to the AIB, we observed glr-1 (GLutamate Receptor family (AMPA) 1) promoter-driven GFP expression. glr-1 is one of the primary excitatory receptors in AIB neurons 50 . GLR-1 is the primary glutamate receptor between ASHs and AIBs. Among the wildtype animals, 65.0% of individuals expressed GFP in two AIB cells, while 35.0% expressed it in one ( Fig. 4a and  b). Notably, transgenic C. elegans carrying extrachromosomal transgenes frequently display mosaic expression 47 . Among the unc-130 mutants, only 5.0% of individuals expressed GFP in two AIB cells, while 52.4% expressed it in one ( Fig. 4a and b); mutants showed a significant difference compared with the wild-type animals. This result indicates a reduction in excitatory synaptic input from the ASHs. We also tried to use the Pinx-1-driven markers, but the transgenic strains could not be maintained, so we used Podr-2 as the marker promoter in this experiment. AIBs are glutamatergic neurons 51 . To determine whether typical glutamatergic synaptic vesicles are formed in AIBs, we analyzed the expression level of eat-4 (a predicted L-glutamate transmembrane transporter), which fills vesicles with glutamate 37 . We detected a significant reduction in the fluorescence intensity of eat-4 promoter-driven mCherry in the AIBs of unc-130 mutants (P < 0.05, n = 20, 20, Mann-Whitney test) (Figs.4c and d), suggesting that unc-130 is required for the release of proper amounts of glutamate. A decrease in the expression of eat-4 might cause the decrease in turns upon AIB::ChR2 stimulation shown in Fig. 3a.
We suspected that the unc-130 mutant forms abnormal presynapses. Then, we observed the localization of mCherry-tagged synaptobrevin (SNB-1), a typical presynaptic marker, at synaptic sites. However, both the maximum size and the total fluorescence intensity of SNB-1::mCherry fusion puncta indicated no significant differences between the unc-130 mutants and wild-type animals ( Fig. 4e-g). RAB-3, a component of synaptic vesicles 52 , also showed no differences in accumulation size and fluorescence intensity between the two strains. These results indicate that unc-130 is required for both the input and output of glutamatergic synaptic transmission at AIB neurons but not for the fundamental function of synapse formation.

unc-130 does not alter the expression of electrical synapses in AIBs.
Previously, we reported that lin-32 upregulates a broad spectrum of genes in AIB neurons, including the dominant electric synapses inx-1, and that its mutation causes secondary hypoesthesia similar to that caused by unc-130 mutation 6 . Therefore, we analyzed the property of inx-1 positive cells, expression rate, cell morphology, and expression intensity, as one AIB indicator of functional overlap between unc-130 and lin-32. First, in terms of expression rate, GFP-positive cells were observed in all unc-130 mutants (n = 128) ( Fig. S8a and b). 116 (90.6%) animals had normal AIB location and morphology, while 12 animals (9.38%) had an ectopic dendrite extending to the tip of the nose (Fig. S8a  and b). However, the occurrence rates were not significantly different (P = 0.0714, n = 40, 128, Fisher's exact test),  Finally, we performed a drop test on the double mutants to confirm that unc-130 and lin-32 act in different molecular pathways. The double mutants exhibited more remarkable optimization defects than the single mutants (Fig. 5c), and the turns almost completely disappeared (Fig. 5d). Meanwhile, the reversal ability remained ( Fig. 5b; Fig. S12), suggesting that unc-130 and lin-32 regulate parallel molecular pathways for turning. We conclude that unc-130 impacts behavioral optimization by regulating the characteristic gene expression of calcium dynamics and glutamatergic synapse function, unlike lin-32. unc-130 starts to be expressed in early embryogenesis but not detected in neurons from the larvae stage onward 53 . In order to clarify whether unc-130 contributes to AIB development and determination decisions in embryogenesis or gene expression in the adult stage, we examined behavioral rescue in the adult driven unc-130 by the inx-1 or npr-9 promoters, which cause AIB-selective expression from larva onward. As a result, all three or four independent transgenic lines driven UNC-130 from the larva onward could not rescue (Figs. S9 and S10), suggesting that function during embryogenesis is essential. Behavioral optimization defects in unc-130 adults were not rescued by overexpression of NCA-2a or EAT-4 driven by the inx-1 promoter (Figs. S10 and S11). Co-transduction with multiple genes in AIB regulated by unc-130-at least nca-2, eat-4 and glr-1 -might be able to rescue behavioral defects.

Discussion
unc-130 regulates the restricted genes for specific neuronal phenotypes to optimize behavior. We have reported that the proneural gene lin-32 promotes the expression of a wide range of genes in AIBs, including gap junctions 6 . However, unc-130 seems to play a more limited role in AIB identity. The unc-130 mutants had reduced expression of glr-1, nca-2, and eat-4 in the AIBs but not the electrical synapse component inx-1 and standard chemical synaptic component SNB-1 in AIB neurons, suggesting that the unc-130 mutation causes behavioral optimization defects by disrupting selective glutamatergic synapse functions (Fig. 6a).
During intense osmotic stimulation, the AIBs receive excitatory inputs through a direct circuit between the ASHs and AIBs via glutamate and GLR-1 receptors as well as inhibitory inputs mediated by AIAs (Fig. 6a). www.nature.com/scientificreports/ Considering the AIB suppression that occurs during osmotic stimulation (Fig. 2a), inhibitory input may exceed GLR-1-mediated excitatory input during hyperosmolar stimulation. We could not identify the inhibitory receptors on AIBs in this study (Figs. S4 and S5). In addition to the unanalyzed inhibitory acetylcholine receptors, neuropeptide receptors might be involved in AIA-AIB suppression. The unc-130 mutation may regulate either inhibitory receptors or genes in the downstream calcium signaling cascade. The molecular basis of optimization after intense osmotic stimulation is also intriguing. After stimulation, OFF responses (calcium increases) in ASHs can induce glutamate release. GLR-1, a receptor for glutamate on AIBs, induces a change in conductance to open NCA-2, a predicted voltage-gated cation channel or a cation leak channel, in order to promote the intracellular influx of cations. This contributes to the increase in the OFFcalcium concentration response in AIBs. We showed that overexpression of NCA-2a in AIBs of unc-130 mutants rescued the optimization defect (Fig. 3e), and unc-130 partially downregulated nca-2 and glr-1 (Figs. 3d and 4b).
These results imply that excitatory input is reduced and suggest the amplification mechanism.
The vesicular glutamate transporter EAT-4 transports glutamate into synaptic vesicles. Calcium induction promotes the secretion of glutamate-releasing synaptic vesicles that transmit information to downstream inter-/ motor neurons, including neck motor neurons. We have shown that the contraction strength of the neck positively correlates with turn behavior 6 . Thus, the AIBs appear to integrate the two excitatory and inhibitory (E/I) pieces of information from the AIAs and ASHs to evaluate the intensity and exposure time of the stimulus and to output avoidance behavior with appropriate intensity and timing.
We demonstrated that unc-130 mutants show reduced expression of glr-1 and nca-2 (Figs. 3c, d, 4a, b), suggesting that insufficient excitatory input from ASHs and voltage-dependent calcium influx occur. Additionally, the reduced expression of EAT-4 may lead to insufficient glutamate loading in the synaptic vesicles, resulting in reduced transmission of information to the downstream neural circuit.
On the other hand, since AIB-specific driven unc-130 does not rescue behavioral abnormalities (Figs. S3 and S4), showing the importance of an act of unc-130 in a developmental stage. Often, with the loss of transcription factors during developmental stages, unexpected circuits may form or differentiate into different cells that are similar in lineage. It has also been reported that unc-130 is required to make a difference between AWA and ASG, chemosensory neurons generated from the ABp(l/r)aapap lineage from which AIB is derived 30 . unc-130 specifies two glial types that arise from the neighboring lineage (ABp(l/r)aapa) 54 . As for the AIBs that we are focusing on in this study, they express at least inx-1, which is selective for AIBs, and show a consistent number, location, and interneuron-like morphology. We consider that they are incomplete differentiated AIBs rather than sister cells of sensory neurons or glial cells. Now we conclude that unc-130 is involved in both the function and development of AIBs, and comprehensive discussion summarizing the importance of both sides is more appropriate for the role of unc-130. unc-130 is required for the establishment of some AIB identities (e.g., biphasic response, expression of nca-2, glr-1, and eat-4, and role as a behavioral optimization center), and unc-130 contributes to behavioral optimization through its role in regulating the expression of a group of genes that are necessary for AIB identities. unc-130 has the defect in the lineage determination, so the possibility remains that it reflects an unidentified reorganization of neurons in addition to AIB defects. It might explain the synthetic defect in the behavior of unc-130;lin-32 double mutants.
Biphasic neural response of AIB neurons. We provide the first evidence for a biphasic neural response of AIBs: calcium decreases during osmotic stimulation followed by a calcium increase after the stimulus (Fig. 2a  and Fig. S3a). First, we discuss the neural mechanisms of the calcium decrease in wild-type C. elegans during stimulation. Calcium imaging results showed that ASHs and AIAs were excited and that AIBs were inhibited during the stimulus (Fig. 2a, d, and e; Fig. S1a, e and f). Since ASHs and AIAs form synapses and since AIA-AIB connections are likely to be inhibitory (Fig. 2b) 7,32 , we speculate that ASHs may directly excite AIAs during stimulation and that excited AIAs inhibit AIBs (Figs. 2b and 6a), resulting in decreased intracellular calcium concentrations in AIBs (Fig. 6b). It has been implied that excitation of AWC chemosensory neurons might cause calcium decreases in AIBs during stimulus exposure 55,56 , but this has not been stringently verified. This study clearly showed such responses with the indicator IP2.0, which monitors the decrease in intracellular calcium concentration associated with osmotic stimulation received by ASHs.
Next, we considered the neural mechanisms of calcium increases in wild-type C. elegans after stimulation. Calcium imaging results showed that ASHs and AIBs, but not AIAs, were excited after the stimulus (Figs. 2a, d,  and e; Fig. S3a, e and f). Although removing the CO 2 stimulus evokes an AIA OFF response (calcium increase) 57 , it is likely due to neural circuits distinct from those involved in response to noxious osmolality. Anatomical analysis showed a direct synaptic connection between the ASHs and AIBs (Fig. 2b) 58 , but the details of ASH-AIB synapses are not well understood. Since ASHs show the biphasic (ON/OFF-increase) response (Fig. 2d) 46 , we speculate that the ASH OFF response (calcium increase) engages the disinhibition of AIB after stimulation 6 .
Physiologically, such a biphasic interneuron response has also been observed in mammals. In rodents, over 60% of suprachiasmatic nucleus neurons may use such "rebound responses" or "postinhibitory rebounds, " and the probability of the response is positively correlated with the duration of hyperpolarization 59 . One of the next questions will be to determine whether the duration of ON-hyperpolarization of AIBs is correlated with the probability of OFF-calcium responses in C. elegans, so that similarity with a property of suprachiasmatic nucleus neurons can be clarified. Rebound responses are also observed in striatal neurons and imply the existence of an essential mechanism for fear processing and decision-making 60 . C. elegans AIBs share similarities in that they are involved in escape behavior optimization and behavioral choice in response to harmful stimuli. The biphasic neural response with excitatory-inhibitory association seems likely to have been evolutionarily acquired for behavioral diversity. We consider that the biphasic neural response in C. elegans may be one of the neural mechanisms of primitive excitation-inhibition association. In addition, rebound firing is also associated with www.nature.com/scientificreports/ diseases; for example, drug addiction pathogenesis in rodents and thalamic neurons generate unusual rebound firing at the end of inhibition in Parkinson's disease in humans 60,61 . In the future, C. elegans AIB neurons may be used as a neuron model for drug addiction pathogenesis and neurodegenerative diseases. Furthermore, AIB biphasic responses may biologically contribute to temporal control of the initiation of turn behavior. Direct AIB excitation by optogenetics facilitates turning without reversal 6 , suggesting that AIA-AIB inhibition suppresses time-consuming turn behavior under conditions of toxic osmolality to expedite escape.

Involvement of FOXD3/4 in avoidance behavior optimization.
In this study, we demonstrated, for the first time, that a FOXD3/4 ortholog, unc-130, specifies avoidance behavior patterns using C. elegans. FOXD3 is widely conserved from invertebrates to vertebrates, including humans 10 . FOXD3 defines the early pluripotency of neural crest stem cells in vertebrates to differentiate into diverse cells, such as neurons and muscles [62][63][64] . Even in C. elegans, unc-130 has a similarity to FOXD3/4 with regard to its expression in both neural progenitors and adult muscle cells 30,65,66 . Future unc-130 studies may lead to universal molecular insights into primary neurogenesis. In addition, whole-human genome analysis has implied that FOXD4 is a risk factor for suicide and obsessive-compulsive disorder 67 . Such outcomes may be attributable to vulnerability to stress. unc-130 might be useful for illustrating a prototypical circuit for improved coping behavior under exposure to harmful stimuli.
During development, combinations of transcription factors determine neuron identities. Hobert et al. have comprehensively mapped the combination of transcription factors expressed in all neurons of C. elegans 37,68 . They speculate that combinations of homeobox (Hox) transcription factors can code for almost all neural identities 69 ; these findings will accelerate the understanding of individual neural characters. In addition, basic helix-loophelix (bHLH) transcription factors, including proneural genes, act in the initial phase and have a crucial role in neurogenesis 61 . However, forkhead box (Fox) transcription factors were grouped relatively recently in 2000 61 , and their functions in the nervous system are still less known than those of Hox and bHLH molecules.

Conclusion
Our findings suggest that a FOXD3/4 ortholog, unc-130 contributes to behavioral optimization mediated by preand postsynaptic function to mediate biphasic neural responses. In summary, reductions in the ON and OFF calcium responses required for integrating this information and producing behavioral outputs result in incorrect behavioral choices in response to stimuli of different intensities in unc-130 mutants.

Methods
Nematodes and maintenance. We cultured C. elegans strains using modified standard techniques 70 .

Drop test.
We performed a drop test as we have previously described 6 . In this study, we used 1-6 M sorbitol dissolved in S basal 6 . "n = x" in the figure legends indicates the number of plates analyzed. Responses were classified as omega turns, long reversals, or short reversals, as previously described 6 . Each score was calculated as the average percentage for 10 ± 3 animals. For a rescue experiment, we selected Pinx-1::gfp marker-positive animals as AIB-rescued animals since extrachromosomal transgenic animals exhibited a mosaic expression pattern. The experimenter was blinded to the nematode strains during the experiment.
Harsh tap assay with a platinum wire. We tapped the nematodes on the tip of their noses with a platinum wire (diameter: 0.25 mm) on NGM plates, respectively. The platinum wire has a smooth polished surface to avoid damaging the nematode. "n = x" in the figure legends indicates the number of plates analyzed. Responses were classified as omega turns and others, as previously described 6 . Each score was calculated as the average percentage for 10 animals. The experimenter was blinded to the nematode strains during the experiment.
Channelrhodopsin 2-induced avoidance assay. We performed a ChR2(H134R)-induced avoidance assay as previously described 6,71 . The animals were individually irradiated with 100% blue light (approximately 2.47 μW/cm 2 ) using a CFP filter (365 nm) at their heads for 2 s. Each score was calculated for 10 ± 3 animals. We performed the experiments on at least three different days and calculated the average percentage. "n = x" in the figure legends indicates the number of plates analyzed. The experimenter was blinded to the nematode strains and to whether all-trans retinoic acid (ATR) was added or not.
Calcium imaging of neurons. As previously described, we performed calcium imaging using an olfactory chip 6,13 . We used 2 M sorbitol dissolved in S basal as a stimulus. All optical recordings of neurons were performed on an IX71 microscope with a 40X immersion objective (Olympus Optical) and an ORCA-Flash2.8 CMOS camera (Hamamatsu Photonics) and analyzed with MetaMorph software (Molecular Devices). We captured time stacks of the fluorescence images at one frame per second. The images were analyzed as previously described 18 . We calculated the percent change in the fluorescence intensity relative to the average intensity during the 5 s before stimulation. In the IP2.0 analysis, the same baseline fluorescence was seen in both wild-type animals and unc-130 mutants, we shifted them to 0 for subtraction and normalization as corrected baselines. In the statistical analysis of AIA, we compared the maximum ΔF/F values from the 5 s before stimulation to 100 s after stimulation. We performed image tracking using a custom ImageJ (NIH, https:// imagej. nih. gov/ ij/) plugin. We drew a rectangular region of interest (ROI) surrounding the cell body, and the ROI was shifted according to the new position of the center for every frame. www.nature.com/scientificreports/ Microscopy. Nematodes were immobilized with M9 buffer containing 50 mM sodium azide on a 5% agarose pad containing 10 mM sodium azide. We obtained fluorescence images (Figs. 3c and 4a; Figs. S4a, b, S5b, and S8a) using a BX51 microscope equipped with a DP30BW CCD camera (Olympus Optical). Confocal microscopic images were captured with Zeiss LSM710 confocal microscopes with either 40X oil immersion objectives for single-plane projections (Fig. 4c) or Z-stacks spanning the focal depths (1 μm/step) of the AIB neurons (Figs. 4e and 5a) using ZEN 2011 software (Zeiss, https:// www. zeiss. co. jp/ micro scopy/ downl oads/ zen. html). We drew an ROI surrounding the cell body ( Figs. 3d and 4d; Fig. S2c) or neurite (Figs. 4g and 5b) and measured the total fluorescence intensity using a custom ImageJ (NIH, https:// imagej. nih. gov/ij/) plugin. We performed ratiometric analysis for an accurate discussion of expression levels in extrachromosomal arrays, comparing marker fluorescence to nca-2 (GFP) (Fig. 3d), SNB-1(mCherry) (Fig. 4e), and goa-1 (GFP) (Fig. S5c) fluorescence by drawing the same ROI position. We calculated the synapse dimensions (Fig. 4f) using a custom ImageJ plugin.