Individual neurons are the basic units of computation in the brain. While some neuron classes have straightforward functional roles, many others exhibit complex multi-functional roles. First, neurons can exhibit multiple modes of dynamics. For example, in the stomatogastric ganglion of crustaceans, individual neuron classes can be co-active with multiple oscillatory networks, each with their own rhythm, and influence activity in each of these networks 1,2 .
Neuron classes whose activities are associated with multiple motor programs have also been identified in worms, flies, and mice [3][4][5][6] . Second, neurons can co-release multiple transmitters 7 .
For example, in C. elegans, RIM neurons control escape responses via release of glutamate, acetylcholine, tyramine, and neuropeptides [8][9][10] . In mammals, midbrain dopamine neurons corelease GABA to modulate basal ganglia circuits 11 . Third, neurons can contain multiple electrotonically isolated compartments 12,13 . In vivo studies that relate these multi-functional properties of neurons to animal behavior remain limited.
The roundworm C. elegans provides a tractable system for dissecting the functional properties of individual neurons in the context of behavior. There are 302 neurons in the C. elegans nervous system and the connectivity among these neurons is known [14][15][16] . C. elegans exhibit a well-characterized set of motor programs 17 that are extensively coordinated with one another [18][19][20][21] and can be flexibly generated depending on the context 22 . These include locomotion, egg-laying, head and body posture, defecation, and feeding. Mechanisms that underlie the coordination of these motor programs are still mostly unknown.
The egg-laying behavior of C. elegans is controlled by a neural circuit that innervates the vulval muscles whose contraction causes egg-laying 23 . A key neuron in the egg-laying circuit is HSN, which synapses onto VC neurons and vulval muscles and has a command-like role in controlling egg-laying. HSN also extends a neurite to the nerve ring in the head where it synapses with neurons in other circuits (Fig. 1A). However, the function of this projection to the nerve ring is not yet understood. HSN releases acetylcholine, serotonin, and several neuropeptides 23,24 and receives diverse modulatory inputs 25,26 . In freely-moving animals, HSN exhibits calcium peaks that are sometimes accompanied by egg-laying 19,27 ; its release of serotonin and the NLP-3 neuropeptide is required for egg-laying 28 . HSN's effects on egg-laying appear to be coordinated with changes in locomotion. Egg-laying is correlated with high-speed locomotion 18,20 , which depends on the AVF neuron and dopamine signaling 18,20 . Somewhat paradoxically, ablation of HSN or reduced serotonin production in HSN impairs low-speed dwelling states, which are stable periods of slow locomotion on food 29,30 . The cellular mechanisms that explain the diverse impacts of HSN on these distinct behaviors are still unclear.
Here, we examine the cellular mechanisms that allow HSN to exert multiple influences on behavior. We show that HSN plays a causal role in driving egg-laying and high-speed locomotion during egg-laying while also eliciting low-speed dwelling over minutes-long timescales. HSN increases egg-laying and acute locomotion via different transmitters and distinct subcellular compartments. In addition, serotonin released from HSN acts over minutes to induce low-speed dwelling. This effect is due in part to serotonin being taken up and re-released by the serotonergic neuron NSM in the head that directly evokes dwelling. Our results reveal how a single neuron can influence a broad suite of behaviors over multiple timescales and show that neurons can 'borrow' serotonin from one another to control behavior.
To examine the causal effect of HSN activity on behavior (Fig. 1A), we performed gainand loss-of-function perturbations to HSN. To activate HSN, we generated a strain with HSNspecific expression of the blue light-sensitive opsin CoChR 31 . There were no promoters known to drive expression uniquely in HSN, so we developed an intersectional approach for cell-specific expression based on the cat-4 and egl-6 promoters (Fig. 1B; Fig. S1A shows specificity; we refer to this strain as HSN::CoChR). We first examined the effects of activating HSN::CoChR with blue light in animals that were dwelling on a food source. HSN activation evoked an immediate increase in egg-laying, as previously reported 25 (Fig. 1C). In addition, we observed an immediate increase in forward locomotion speed (Fig. 1C; Fig. S1B). Since HSN activation had not been shown to induce speeding before, we performed additional controls to ensure that this effect was specifically due to HSN activation. We examined the effect of activating HSN::CoChR in egl-1(gf) mutants, where HSN dies from programmed cell death 32 . This abolished the light-induced increase in egg-laying and locomotion speed, indicating that both effects are due to HSN activation (Fig. 1D). We also examined whether the increase in locomotion speed was a consequence of egg-laying or, alternatively, a parallel output. To test this, we activated HSN in animals sterilized with FUDR. In these animals, there were no eggs, but HSN stimulation still increased locomotion speed (Fig. 1E). Taken together, these experiments indicate that HSN activity acutely increases egg-laying and forward locomotion and that the effect on locomotion is separable from HSN's effect on egg-laying.
We also examined the effects of HSN activation under conditions where animals were travelling at high speed due to recent transfer to a low-density food plate. HSN activation still increased locomotion during light stimulation, but this was followed by an acute reduction in locomotion that lasted for minutes (Fig. 1F). This slowing effect remained intact in animals sterilized with FUDR, suggesting that it was not dependent on egg-laying (Fig. S1C). Together with the above results, this suggest that HSN activation promotes acute egg-laying and highspeed locomotion, followed by a reduction in locomotion speed.
We next examined the impact of endogenous HSN activity on locomotion. For these experiments, we utilized egl-1(gf) mutants in which HSN is specifically killed. egl-1(gf) mutants are known to exhibit reduced egg-laying 32 . We examined whether the eggs that are laid in the mutant are accompanied by increased locomotion, as is the case in wild-type. To do so, we recorded egg-laying and locomotion simultaneously and examined animal speed surrounding egg-laying. Consistent with previous data 20 , wild-type animals displayed increased forward movement before and during egg-laying, but this was disrupted in egl-1(gf) mutants (Fig. 1G). This indicates that HSN is required for speeding prior to egg-laying and matches our finding that HSN activation increases speed and egg-laying. egl-1(gf) animals actually displayed a speed reduction prior to egg-laying, which could be due to other components of the egg-laying circuit that remain intact in egl-1(gf). We also examined dwelling behavior in egl-1(gf) mutants using an assay that quantifies animal exploration across a bacterial food lawn. As described previously, egl-1(gf) mutants displayed reduced dwelling behavior (or, equivalently, increased roaming; Fig. 1H) 30 . This indicates that HSN is necessary for proper dwelling and matches our findings that HSN activation can increase dwelling for minutes after optogenetic stimulation. Taken together, these data indicate that HSN is required for increased locomotion speed surrounding egg-laying events and low-speed dwelling on bacterial food.
The above experiments suggest that HSN activity induces acute egg-laying and speeding, as well as long-term slowing. We next examined how endogenous HSN activity was coupled to these behaviors. We generated a strain expressing GCaMP5A and mScarlett in HSN and performed ratiometric imaging as animals freely moved at low speed on bacterial food lawns.
HSN activity was organized into discrete calcium peaks that commonly occurred in bursts (Fig. 2A), as previously described 19,27,33,34 . Egg-laying was invariably coupled to HSN calcium peaks (Fig. 2B andS2A). However, there were many calcium peaks that were not accompanied by egglaying. We also examined animal speed surrounding HSN calcium peaks and found that the peaks were on average time-locked to transient increases in speed (Fig. 2C andS2B). This effect was just as robust when excluding the HSN peaks associated with egg-laying (Fig. S2C). This suggests that HSN calcium peaks are often accompanied by egg-laying and increased animal speed.
HSN calcium peaks often occur in bursts 19,27 . Based on an analysis of inter-peak intervals, HSN peaks within one minute of one another appeared likely to be part of the same burst (Fig. S2D). We examined whether HSN peaks differed in their correlation with behavior depending on whether they occurred earlier or later in bursts. HSN calcium peaks that were associated with egg-laying most frequently occurred later in bursts, shortly after several other peaks (Fig. 2D). In contrast, HSN calcium peaks were most strongly correlated with locomotion speed when there were fewer previous HSN peaks (Fig. 2E; see also Fig. S2E). Thus, when several HSN calcium peaks occur in close succession, earlier peaks are associated with increased locomotion and later peaks are associated with egg-laying. These data provide evidence that endogenous HSN activity is acutely correlated with speeding and egg-laying.
To examine the intrinsic coupling of HSN activity to motor networks, we also examined HSN activity in immobilized animals. During immobilization, the C. elegans nervous system exhibits fictive locomotion dynamics where neurons that encode forward and reverse movement switch between high and low activity states [35][36][37] . RIB is one of the forward-active neurons and its bi-stable activity reports the fictive locomotion state of the network. Therefore, we recorded HSN and RIB simultaneously. HSN calcium peaks were significantly more likely to occur when RIB activity was high (Fig. S2F-G), suggesting that HSN activity is coupled to the forward locomotion network even in the absence of actual movement.
We next sought to determine which HSN transmitter(s) mediate these behavioral effects.
To determine the transmitters that underlie HSN-induced speeding, we performed optogenetic HSN stimulation in mutant backgrounds lacking specific HSN transmitters. HSN evokes egglaying via serotonin and NLP-3 neuropeptides 28 . However, animals lacking serotonin (tph-1 mutants) or both serotonin and NLP-3 still displayed increased speed upon HSN::CoChR activation (Fig. 3A-B). HSN also releases acetylcholine 38 , so we examined its impact on HSNinduced speeding. Animals fully deficient in acetylcholine release are uncoordinated 39 .
Therefore, we engineered a conditional knockout allele of unc-17, which encodes the vesicular acetylcholine transporter (VAChT) 40 required for acetylcholine release (Fig. 3C; see Methods for validation). However, animals with an HSN-specific deletion of unc-17 still displayed robust HSN-induced speeding (Fig. 3D). This suggests that serotonin, NLP-3, and HSN-produced acetylcholine are not essential for HSN-induced speeding.
Several neuropeptides are expressed in HSN, according to previous gene expression studies 24,41,42 . Thus, we next asked whether HSN neuropeptide production is required for HSNinduced speeding. We examined HSN-induced speeding in animals with a null mutation in egl-21, which encodes a carboxypeptidase required for the production of many neuropeptides 43 . HSN-induced speeding was abolished in these mutants (Fig. 3E). However, egl-21 is expressed very broadly, so it was unclear whether this effect was due to loss of neuropeptide production in HSN or other neurons. Therefore, we examined HSN-induced speeding in a strain harboring an HSN-specific deletion of egl-21 (Fig. 3F). We used egl-6::nCre to inactivate egl-21 in HSN (along with ~5 other neurons that express egl-6) 26 and found that this abolished HSN-induced speeding (Fig. 3G; Fig. S3A shows no effect on baseline velocity off food). This suggests that HSN neuropeptide production is required for HSN-induced speeding.
We attempted to determine which HSN neuropeptide(s) mediate HSN-induced speeding.
We obtained a panel of 12 mutants lacking neuropeptides reported to be expressed in HSN 24,41,42 .
We examined speeding surrounding native egg-laying events in each of these single mutants. Of these, flp-26 and flp-28 impacted animal speed surrounding egg-laying (Fig. S3B-D; other mutants had no effect). We also crossed neuropeptide mutants into HSN::CoChR and found that loss of flp-2 and flp-26 partially attenuated HSN-induced speeding (Fig. 3H-J). This suggests that flp-2, flp-26, and flp-28 may be involved in egg-laying-or HSN-induced speeding. flp-2 is known to impact locomotion 44 , but flp-26 and flp-28 have not been closely examined. We made compound mutants lacking multiple neuropeptides and found that both HSN-induced speeding and the native coupling of speeding to egg-laying were attenuated in double mutants lacking flp-2 and flp-28 (Fig. 3K-L). HSN-specific expression of flp-2 and flp-28 in the double mutant restored normal HSN-induced speeding (Fig. 3M). Taken together, these experiments suggest that HSN acutely increases locomotion via its release of neuropeptides, including flp-2 and flp-28.
We next examined which HSN transmitter(s) drive the decrease in locomotion speed over longer time scales. Here, we analyzed baseline dwelling behavior and HSN-stimulated slowing (see Fig. 1F). We first examined serotonin and NLP-3, the HSN transmitters that control egglaying 28 . We found that tph-1 mutants (lacking serotonin) displayed decreased baseline dwelling (Fig. 4A) and a reduction in HSN-stimulated slowing, particularly several minutes after stimulation (Fig. 4B andD). Animals lacking nlp-3 had normal baseline dwelling (Fig. 4A) and normal HSN-stimulated slowing (Fig. 4C-D; see also Fig. S4A). We previously showed that cellspecific tph-1 deletion in HSN also causes a deficit in dwelling 30 . Here, we found that HSNspecific tph-1 expression partially rescued the tph-1 mutant deficit in dwelling, suggesting that HSN-produced serotonin is sufficient to drive dwelling (Fig. 4E). Thus, multiple lines of evidence suggest that HSN serotonin promotes dwelling behavior. We characterized the timescale over which HSN influences low-speed dwelling on food.
As shown above, HSN activation induces low-speed dwelling with a short latency and these effects last for minutes. To determine the timescale over which native HSN activity controls dwelling, we examined how long it takes for chemogenetic HSN silencing (with a histaminegated chloride channel, HisCl 45 ) to alter dwelling on food. We first used egg-laying assays to characterize how quickly HSN is inactivated after HSN::HisCl animals are transferred to histamine-containing plates. Egg-laying was inhibited within 5min of transfer, suggesting that HSN is silenced within minutes after histamine exposure (Fig. S4B). We then determined the time course of the effect of HSN silencing on low-speed dwelling behavior. Animals were recorded right after being transferred onto histamine-containing plates. HSN-silenced animals showed significantly higher speed only after 30-40min of histamine exposure (Fig. 4F). This suggests that HSN needs to be inactivated for tens of minutes for there to be an increase in speed, revealing a long-lasting effect.
Given this slow time scale, we hypothesized that HSN-released serotonin influences dwelling by contributing to the tonic pool of serotonin. In addition to directly interacting with downstream serotonin receptors, extracellular serotonin might be taken up by other serotonergic neurons via MOD-5 46,47 , the serotonin transporter (SERT), and re-released to influence behavior (Fig. 4G andS4G). The serotonergic neuron NSM in particular expresses high levels of mod-5. 46,47 NSM is activated by feeding; its activity is correlated with dwelling and NSM activation drives dwelling via serotonin release 6,[48][49][50] (Fig. 4G). It can produce its own serotonin, but it was unclear whether NSM serotonin can also be supplied by other neurons. Thus, we next examined whether HSN-produced serotonin could be taken up and re-released by NSM to influence dwelling.
A key prediction of this hypothesis is that NSM's ability to induce slow locomotion via serotonin release should depend on HSN serotonin production. To test this, we used an assay where we optogenetically activated NSM. Consistent with previous work 50 , we found that optogenetic NSM activation evoked slowing in a tph-1-dependent manner, indicating that NSM serotonin drives slowing (Fig. S4C). We next examined whether animals that only have tph-1 expressed in HSN displayed a rescue in NSM-induced slowing. Indeed, restoring HSN serotonin production via HSN::tph-1 expression partially rescued the ability of optogenetic NSM stimulation to induce slowing (Fig. 4H). This suggests that HSN serotonin production can partially rescue NSM-induced slowing.
We next performed a complementary experiment where we tested whether the loss of HSN (via egl-1 mutation) in a wild-type background impairs the ability of NSM to induce slow locomotion. Indeed, there was a significant reduction in NSM-induced slowing in egl-1(gf) mutants (Fig. 4I). This suggests that HSN is required for normal NSM-induced slowing.
These results are consistent with two possible interpretations. First, HSN serotonin might be taken up by NSM via MOD-5/SERT, such that altering HSN serotonin production alters NSM's ability to evoke slow locomotion via serotonin re-release. Alternatively, HSN serotonin may be required downstream of NSM activation for slowing. To distinguish between these possibilities, we tested whether HSN could still influence NSM-stimulated slowing in a mutant background where mod-5/SERT was deleted. If the first interpretation is correct, there should be no additional effect of HSN ablation on NSM-stimulated slowing if mod-5 is already deleted, since the mod-5 mutation would already prevent extracellular serotonin from being taken up by NSM. Alternatively, if NSM-stimulated slowing requires a downstream function of HSN, then HSN ablation should still affect NSM-induced slowing even when mod-5 is absent. Thus, we compared the effects of optogenetic NSM stimulation on locomotion in mod-5 mutants versus mod-5;egl-1(gf) double mutants. We performed this experiment using a range of different light intensities to ensure that we were examining NSM-induced slowing under conditions where slowing was non-saturating (Fig. 4J; see Fig. S4D-F for multiple light intensities). This was necessary because the loss of mod-5 leads to hyper-enhanced slowing, since released serotonin cannot be rapidly cleared by MOD-5/SERT 47,51 . At all light levels tested, there was no difference between the two strains (Fig. 4I-J). This indicates that HSN only impacts NSM-induced slowing when mod-5/SERT-dependent serotonin re-uptake is intact. Taken together, this set of experiments suggests that HSN serotonin is taken up and re-released by NSM to evoke slow locomotion.
HSN causally influences behavior in at least three ways: (1) it drives acute egg-laying through serotonin and NLP-3 release; (2) it drives acute speeding through release of neuropeptides; and (3) it drives dwelling over longer time scales via serotonin release. We next asked how these distinct functions of HSN map onto its unique anatomy. The HSN soma is in the mid-body, posterior to the vulva (Fig. 5A), and its neurite projects anteriorly towards the head 15 .
In the region where it passes over the vulva, HSN extends short branches where it synapses with vulval muscles and VC neurons (the 'vulval presynaptic region'). The main HSN axon then enters the ventral cord, projects anteriorly, and enters the nerve ring in the head, where it makes synapses with other neurons.
How does HSN activity propagate across the neuron? Previous work showed that HSN calcium peaks in the soma occur normally even when the HSN axon is severed before reaching the vulval presynaptic region 33 . This suggests that the HSN activity peaks can originate in the soma. To determine whether HSN soma calcium peaks are accompanied by calcium peaks along the entire axon, we performed calcium imaging of the whole HSN neuron in immobilized animals and examined signals in the soma, vulval presynaptic region, and distal axon near the nerve ring (Fig. 5B). HSN calcium peaks in each compartment were accompanied by calcium peaks in the other compartments (Fig. 5C), suggesting that calcium peaks are rapidly propagated across the entire neuron.
Next, we mapped out sites of transmitter release in HSN, focusing on serotonin. cat-1 encodes the C. elegans vesicular monoamine transporter (VMAT) that loads serotonin into synaptic vesicles. Its subcellular localization can be used as an indicator of serotonin-containing vesicle release sites 52 . Because cat-1 is broadly expressed, we generated a strain to visualize endogenous CAT-1 localization in individual neurons, like HSN. We engineered a strain with three tandem repeats of the split-GFP fragment GFP11 inserted before the native cat-1 stop codon, creating an in-frame fusion protein (Fig. 5D). We then expressed the other split-GFP fragment GFP1-10 under the egl-6 promoter that drives expression in HSN, but no other cat-1expressing neurons. When GFP1-10 interacts with CAT-1-3xGFP11 in HSN it reconstitutes a functional GFP fluorophore. Reconstituted CAT-1::GFP in HSN displayed punctate localization, suggestive of presynaptic release sites, matching previous non-cell-specific CAT-1 immunostaining 52 . HSN CAT-1::GFP puncta were brightest in the vulval presynaptic region, with much weaker fluorescence in the head and no expression along the neurite (Fig. 5E; Fig. S5A). EM studies indicated that HSN synapses in the mid-body are larger than those in the head, which may be related to the difference we observed here 15 . This suggests that HSN primarily releases serotonin in the vulval presynaptic region.
We also tested which behavioral functions of HSN require its axonal projection to the nerve ring. To do so, we examined animals where we axotomized the HSNL/R neurites just anterior to the vulval presynaptic region (Fig. 5F; Fig. S5B shows exact cut site). This leaves the HSN soma connected to the vulval presynaptic region but not to the remainder of the neurite that projects into the nerve ring. Egg-laying rates were unaffected by laser axotomy at this position (Fig. 5G), consistent with the notion that the HSN vulval presynaptic region controls egg-laying.
However, egg-laying in the axotomized animals displayed much weaker coupling to increased locomotion speed (Fig. 5H; laser controls in Fig. S5C). This suggests that HSN signal propagation to the head is required for proper speeding during egg-laying. In contrast, the baseline speed of animals on food was not significantly affected by HSN axotomy (Fig. 5I).
Thus, disrupting the HSN axonal projection to the nerve ring does not bias the animals towards roaming, even though HSN cell ablation or HSN silencing does (Fig. 5I; and above). Together, these experiments suggest that HSN axonal projections to the head are required for acute speeding during egg-laying but not baseline egg-laying or baseline on-food locomotion.
The above results provide information about how the functional outputs of HSN map onto the different anatomical and molecular features of this neuron. We also wanted to examine how sensory inputs are transmitted to HSN, given its unique anatomy and function. Egg-laying behavior is impacted by many aspects of the sensory environment, including food availability, aversive cues, and more [53][54][55][56] . Here, we focused on the effects of osmolarity, as it has been shown that a mild upshift in osmolarity (< 1 Osm) triggers reduced egg-laying 33,57 and reduced HSN activity 33 , suggesting that this may be a good system for studying sensory control over HSN activity and behavior.
To examine how high osmolarity inhibits HSN and egg-laying, we used an assay where animals were transferred to a high osmolarity plate for one hour and the number of eggs laid was counted (Fig. 6A). Osmolarity was increased beyond the normal 150 mOsm level by addition of sorbitol. This revealed a dose-dependent effect where higher levels of osmolarity in the plate inhibited egg-laying (Fig. 6B; baseline egg-laying rates in Fig. S6A). To determine the sensory mechanisms that link osmolarity to egg-laying, we examined behavioral responses of animals lacking either tax-2 or ocr-2, which encode ion channels required for sensory transduction in different sensory neurons 58 . While ocr-2 mutants still responded to the osmotic stimulus, tax-2 mutants had an attenuated response at 300 mOsm (Fig. 6B-C). Higher concentrations still inhibited egg-laying in tax-2, suggesting that additional mechanisms may inhibit egg-laying under those conditions. We observed the same behavioral phenotype in animals lacking tax-4, which encodes ion channel subunits that function together with TAX-2 (Fig. 6C) 58 .
We asked whether these mild 300 mOsm conditions that reduced egg-laying were sufficient to impact HSN activity using GCaMP imaging in moving animals. Indeed, HSN calcium peak frequency was significantly reduced in the animals on 300 mOsm agar, compared to standard 150 mOsm conditions (Fig. S6B).
To map out which exact sensory neurons are required for osmolarity-induced egg-laying inhibition, we performed two sets of experiments. First, we examined behavioral responses in a panel of transgenic and mutant strains with specific sensory neurons ablated. Second, we examined behavioral responses in animals that had tax-4 rescued in different sensory neurons to recover their sensory transduction. We examined the effects of cell ablation for >10 sensory cell types and only observed an attenuated behavioral response in animals with the sensory neuron BAG ablated (Fig. 6D; Fig. S6C). We examined the effects of restoring tax-4 expression in seven neuron classes and observed the most robust rescue when tax-4 was rescued in BAG (Fig. 6E; Fig. S6D). Consistent with the tax-2 mutant results above, BAG ablation did not prevent the reduction in egg-laying caused by higher osmolarity levels (Fig. S6E). This suggest that sensory transduction in BAG sensory neurons is important for osmolarity-induced inhibition of egglaying.
BAG senses gases through its ciliated sensory ending 59,60 . To test whether BAG responds to changes in osmolarity, we performed BAG GCaMP imaging in freely-moving animals as they moved from a 150 mOsm agar environment to a 300 mOsm environment (generated by fusing agar; see Methods). Indeed, BAG calcium levels increased as animals moved to the higher osmolarity environment (Fig. 6F). This suggests that BAG either directly senses osmolarity upshifts or receives inputs from other osmo-sensitive cells.
We attempted to characterize the BAG signal that is required for osmolarity-induced inhibition of egg-laying behavior. BAG is the main source of FLP-17 and one of the sources of FLP-10 61 . Both of these neuropeptides inhibit HSN, so we examined osmolarity-induced behavioral responses in mutants lacking these neuropeptides 26 . flp-17 mutants, as well as animals lacking the FLP-17 receptor egl-6, showed an attenuated egg-laying response to osmolarity upshift, matching BAG-ablated animals (Fig. 6G; Fig. S6F). This suggests that FLP-17/EGL-6 signaling is important for osmolarity-induced egg-laying inhibition. Related to this, we also found that HSN axotomy (in the position described above) did not attenuate the effects of osmolarity on egg-laying (Fig. 6H). This suggests that a humoral signal is relayed from sensory neurons to HSN to inhibit egg-laying, rather than local synaptic signaling in the head. Overall, these results suggest that increased osmolarity activates a BAG-FLP-17 signal to inhibit egg-laying. Given the known role of FLP-17/EGL-6 signaling in inhibiting HSN and our observation that exposure to 300 mOsm conditions activates BAG and inhibits HSN, this effect may be mediated by HSN inhibition. In addition, we have not ruled out the presence of other osmolarityinduced changes that may impact egg-laying in parallel.
Animals coordinate many distinct motor outputs as they execute purposeful behaviors.
Here, we show how specific multi-functional properties of the HSN neuron endow it with the ability to orchestrate a suite of behavioral changes. HSN promotes an acute increase in egglaying and locomotion, followed by low-speed dwelling behavior. The acute effects on egglaying and speeding are mediated by distinct sets of HSN transmitters and different subcellular compartments. The longer lasting effect on dwelling is mediated by HSN release of serotonin, which is taken up and re-released by serotonergic NSM neurons that directly evoke dwelling.
Our results illustrate how cellular morphology, multiple transmitter systems, and non-canonical modes of transmission like neurotransmitter "borrowing" endow a single neuron with the ability to orchestrate multiple features of a behavioral program.
While the role of HSN in egg-laying is well-established 23,56 , we used optical tools to reveal additional behavioral functions of HSN. We found that HSN releases multiple neuropeptides to acutely increase locomotion speed prior to egg-laying. Our axotomy data suggest that the speed-evoking effect of HSN requires its axon in the head. Given that the locomotion circuit is located in the head and that HSN calcium peaks are reliably transmitted along the HSN axon, these peptides may be well positioned to exert fast, direct action on locomotion circuits during HSN calcium peaks. We also found that HSN serotonin can induce dwelling behavior that lasts for minutes after HSN activity ends. Interestingly, HSN-released serotonin is taken up and re-released by NSM in the head, which directly drives dwelling. After serotonin is released, it can be absorbed by different cell types via the serotonin transporter, MOD-5/SERT 46,62,63 . In C. elegans, the neurons NSM, AIM, and RIH neurons have been shown to absorb endogenous and exogenous serotonin in a mod-5-dependent manner 46,47,64 . However, it has remained unclear whether this absorption is for serotonin turnover/degradation or, alternatively, whether this serotonin is re-released to impact behavior. Our work here provides evidence that serotonin can be transferred between serotonergic neurons and re-released to control behavior. Recent expression studies suggest that similar mechanisms could potentially occur for GABA as well 65 . This mechanism of neurotransmitter "borrowing" may be functionally important. NSM is activated by feeding and its release of serotonin evokes dwelling 49,66 . Our results here suggest that serotonin levels in NSM are influenced by HSN activity. This might allow HSN activity to have a priming-like effect, where its recent activity could increase NSM serotonin levels so that subsequent food-driven NSM activation could lead to more robust slowing. Future studies could make use of fluorescent serotonin sensors 67,68 to define the spatiotemporal dynamics of this extra-synaptic serotonin more precisely.
The behavioral coordination that HSN facilitates during egg-laying could be evolutionarily adaptive. One possible reason that animals may increase locomotion during egglaying may be to depolarize muscle cells adjacent to vulval muscles, facilitating egg-laying. In addition, high-speed movement during egg-laying may allow animals to distribute their eggs rather than depositing them all in one location. Coupled with HSN's ability to bias animals towards dwelling, this could allow animals to distribute their eggs within a high-quality environment and dwell in the overall vicinity as well. C. elegans egg-laying is impacted by many aspects of the sensory environment -food, aversive cues, and more 23 . We found that high osmolarity, an aversive stimulus, signals through a humoral factor to inhibit egg-laying.
Together, HSN's sensory inputs and its outputs that couple egg-laying to locomotion may allow animals to distribute their eggs across favorable sensory environments. For all panels, statistics were comparing the indicated day-matched groups.
Data are shown as means ± SEM.
See also Figure S6.
Further information and requests for resources and reagents should be directed to and will be fulfilled by the lead contact, Steven Flavell (flavell@mit.edu).
All plasmids, strains, and other reagents generated in this study are freely available upon request.
Data: Behavioral and neural data related to Figures 2 and Figure 6 Any additional information required to reanalyze the data reported in this paper is available from the Lead Contact upon request.
C. elegans Bristol strain N2 was used as wild-type. All wild-type, mutant and transgenic strains used are listed in the key resources table. Animals were maintained on NGM agar plates seeded with E. coli OP50 bacteria strain and kept at 22°C, 40% humidity. One day-old adults were used for all experiments. For genetic crosses, genotypes were confirmed by PCR and/or sequencing.
Transgenic animals were generated either by injecting DNA with fluorescent co-injection markers into the gonads of young adult hermaphrodites, CRISPR/Cas9 genome editing or mosSCI insertion.
Plasmid backbones: C. elegans codon-optimized GCaMP7b and GCaMP7f open reading frames were synthesized and inserted into the pSM vector. The intersectional promoters, consisting of the inverted/floxed vector and the Cre vector were previously described (Flavell et al., 2013).
The HisCl1 plasmid was previously described (Pokala et al., 2014). For tax-4 rescue, we used the previously described tax-4 cDNA 69 , but moved it into a pSM-t2a-GFP vector backbone for expression. For tph-1 rescue, we used the KZ1290.21.1 cDNA, which we subcloned into the pSM vector.
Promoters used in this study: egl-6 (Flavell et al., 2013), cat-4 (full length, 4kB immediately upstream of cat-4 start codon), cat-4prom68 70 , sto-3 (Ji et al., 2021). The NSM-specific promoter was a 158 bp fragment of tph-1 promoter, validated to be NSM-specific in our previous work through GFP expression and Ribotagging analysis (Rhoades et al., 2019). In addition, we showed that the resulting NSM::Chrimson line used here has no light-induced egg-laying even at maximum light intensities tested (Dag et al., 2023), further confirming that it confers no HSNspecific expression.
The egl-21 cell-specific deletion strain was constructed in an egl-21(n476) mutant background.
For the mosSCI insertion, the egl-21 genomic region (spanning entire genomic region up to adjacent genes in both directions) was inserted into the mosSCI insertion site on chromosome IV 71 . LoxP sites that we inserted into the egl-21 single-copy rescue allow for Cre-dependent deletion of exons 2 through 5 of the egl-21 gene, which is the majority of the coding sequence.
The conditional knockout allele of unc-17 was constructed via iterative rounds of CRISPR/Cas9 genome editing. One loxP site was inserted ~250bp after the end of the unc-17 3'UTR. Another loxP site was inserted in the intron before the last coding exon (which encodes the majority of the UNC-17 protein). We found that pan-neuronal Cre expression in this strain gave rise to animals with an Unc phenotype, validating that the loxP sites work effectively.
The cell-specific fluorescent labeling strain for cat-1 was generated via CRISPR/Cas9 genome editing. Three tandem repeats of the GFP11 sequence separated by short linker sequences (glygly-ser-gly-gly) were inserted immediately before the cat-1 stop codon.
Multi-animal recordings of C. elegans locomotion were conducted as previously described (Rhoades et al., 2019). One day old adult animals of the indicated genotypes were transferred to NGM plates with or without OP50 bacteria. For animals that were fasted, animals were transferred to NGM plates without OP50 for three hours prior to recording. All animals were recorded using Streampix software at 3 fps. JAI SP-20000M-USB3 CMOS cameras (5120x3840, mono) with Nikon Micro-NIKKOR 55mm f/2.8 were used. White-panel LEDs (Metaphase) provided backlighting. Videos were analyzed using previously-described custom MATLAB scripts (Rhoades et al., 2019). For optogenetic stimulation, light was supplied from a 470nm (for CoChR; 0.5 mW/mm2) or 625nm (for Chrimson; 0.6 mW/mm2 unless otherwise specified)
Mightex LED at defined times in the video.
For joint recordings of egg-laying and locomotion, we used previously described custom-built single worm tracking microscopes (Cermak et al., 2020). These custom microscopes have a livetracking function that permits long-term recording of single moving animals. L4s animals were picked 16-20 hours before the recording day. On the day of the recording, animals were transferred to NGM plates seeded with OP50 (1:20 dilution of saturated culture) the day before the recording (thin bacterial lawns are a requirement for the live-tracking function on the microscopes). LabView software controlled the microscope and acquired the images. Data were then analyzed in R Studio and MATLAB. For optogenetic stimulation, 532nm laser light was supplied at defined times at an intensity of 250 uW/mm 2 .
Laser axotomy was performed as previously described 72 with a few modifications. L4 stage transgenic animals were transferred to a 3% agarose pad and immobilized with 2.5mM levamisole in M9 buffer. Animals were visualized with a Nikon Eclipse 80i microscope, 100x
Plan Apo VC lens (1.4 NA), Andor Zyla sCMOS camera and a Leica EL6000 light source. HSN axons were severed anterior to the vulval presynaptic region and before they extend to the ventral nerve cord using a 435nm nitrogen pulsed MicroPoint laser fired at 20 Hz. Both HSNL and HSNR axons were sequentially severed by gently rolling the animal from one side to the other after the first axotomy. To facilitate rolling, grooved agarose pads were stamped with a portion of a vinyl record 73 . 'mock' control animals underwent the same immobilization and rolling protocol but were not axotomized. Both mock and axotomized animals were immediately recovered in M9 buffer and transferred to OP50 seeded NGM plates for behavioral analyses 20 hours later. For control axotomies, the laser was fired using the same settings that were used for the real axotomy, but it was targeted adjacent to the HSN neurite and confirmed to not visibly damage HSN, based on HSN fluorescence.
HSN calcium imaging in freely-moving animals was conducted as previously described (Rhoades et al., 2019) with a few small modifications. Animals were mounted on flat agar, with freshly seeded OP50 bacterial food. They were enclosed in a small chamber with a rubber gasket and cover glass and GCaMP/mCherry data was recorded at 10 fps (with 10ms exposure times).
Agar was either normal 150 mOsm or 300 mOsm (due to addition of sorbitol), as described in For BAG calcium imaging, animals were placed on agar that was fused. This was done by first pouring a normal osmolarity NGM pad on a microscope slide, then slicing it to make a flat edge on one side. 300 mOsm agar (adjusted with sorbitol) was then poured next to the first NGM pad and allowed to dry. Animals were picked to this flat agar surface and imaged as described above.
GCaMP data was recorded at 10 fps (with 10ms exposure times). Custom ImageJ tracking scripts were used to track and quantify BAG GCaMP fluorescence.
Calcium imaging of HSN and RIB, as well as different HSN compartments, in immobilized animals was conducted on a previously described spinning disk confocal microscope (Ji et al., 2021). Animals were mounted on 5% agar with 0.05um beads for immobilization 74 . They were imaged using a 20X/0.95 objective coupled to a 5000 rpm Yokogawa CSU-X1 spinning disk unit with a Borealis upgrade. Z-stacks were collected with NIS Elements software. For data analysis, data were converted to maximum intensity projections (RIB and HSN were typically in different z-planes), ROIs were manually drawn around the somas of RIB and HSN, and then backgroundsubtracted intensities within the ROIs were calculated.
CAT-1::GFP puncta in HSN were imaged on a spinning disk confocal microscope that has been previously described 6 . Imaging of the head and vulval regions was conducted using identical camera settings and entire z-stacks were collected spanning the depth of the animal's body, with exposure times set such that there were no saturated pixels. In addition, a longer exposure z-stack was collected for the head region to provide a higher SNR image of dim signals in the head.
NGM plates with varying osmolarities were used for egg-laying assays. Sorbitol was used to increase the osmolarity in the NGM plates to desired osmolarity (300-450 mOSm), and added together with CaCl2 and KPO4 buffer to the NGM agar. The day before the assay, assay plates were seeded with 200 ul OP50 per plate, and lids were left open for 20 minutes in the biosafety hood to allow them to dry. L4s animals were picked to OP50-seeded NGM plates the day before the assays and grown for 20-24 hours to become gravid adults. On the day of the experiment, adult animals were transferred onto seeded control or high osmolarity plates (10 animals per plate), and were left to lay eggs for an hour. After an hour, animals were removed from assay plates and the number of eggs laid was counted. The percentage of egg laid was calculated by the number of eggs laid on high osmolarity plates divided by the number of eggs laid on control plates.
Exploration assays that provide a reliable measure of roaming versus dwelling behavior were performed as previously described (Flavell et al., 2013) with minor modifications. Single L4 animals were picked to NGM plates with OP50 bacteria seeded 1 day prior. They were then left to explore the plates for 16 hours, after which the plate was superimposed on a transparent grid and the number grid squares that the animal tracks traversed was quantified. In some experiments (as indicated in the text), animals were mounted as adults on plates and allowed to explore for 3 hours, and then the number of squares was counted.
Descriptions of statistical tests and group sizes are provided in figure legends. In addition, definitions of center and dispersion and precision measures are also in the figure legends. Here, we provide an overview of statistical methods that were used in multiple places in the paper, providing additional information not in the figure legends.
For all HSN optogenetics experiments conducted using single-worm recordings (i.e.
those that include both velocity and egg-laying quantification), we used the following statistical procedure. First, for each individual optogenetic trial we determined the increase in velocity during optogenetic stimulation, compared to pre-light baseline (-2 to 0 min before each stimulus). We then averaged this for all trials in a given animal to get that animal's average increase in velocity caused by optogenetic stimulation. This n (# of animals) was the n used in statistical analyses. To compare across genotypes, we performed a Mann-Whitney U test comparing the velocity increase in the group of control animals to the velocity increase in the treatment (e.g. mutant) group. For statistical tests on the egg-laying effects, we performed a similar analysis, but just on the mean egg-laying rate during optogenetic stimulation (since the baseline is essentially zero). In many analyses, the appropriate control group was HSN::CoChR +ATR animals. The single worm tracker is low throughput (one animal is recorded over 3 or 6 hours), so we tested whether it was appropriate to pool animals of this control group across days.
We did this by taking all of our recordings and performing a one-factor ANOVA where 'recording date' was the factor (i.e. testing whether animals' HSN-induced velocity increase was dependent on recording date). We found that there was no significant effect of recording date (Fig. S1B; F=1.162; p=0.3738), so we pooled these animals and used them as the comparison group for several treatment groups recorded over the same overall time period.
For HSN optogenetics experiments conducted using multi-worm tracking, we only analyzed data over one single optogenetics trial per animal. We computed the change in speed (compared to baseline) for each animal and then compared the magnitude of the speed change in mutant animals to day-matched controls (using an 'n' that is equal to the number of animals assayed) with a Mann-Whitney U test.
For analyses of velocity surrounding spontaneous egg-laying events, data were recorded on single-worm trackers. Analysis was conducted very similarly to the HSN optogenetics single worm tracker analysis. For each egg-laying event, we determined the velocity increase right before the egg-laying event by computing the difference between velocity during egg-laying (from -15s to 0s before egg-laying event) and a preceding baseline period (-5 to -3 min before egg-laying event). We then averaged this value across all observations in each animal to obtain per-animal averages. To compare treatment groups, we performed a Mann-Whitney U test to compare the two respective groups of animals. In many cases, the appropriate control group was wild-type (N2). Again based on the low-throughput nature of this single-worm recording assay, we asked whether it was possible to pool wild-type animals' data across days. We again used a one-factor ANOVA where the factor was 'recording date' and found no effect. This suggested that it was valid to pool wild-type animals and compare them to treatment groups recorded over the same overall time period.
For egg-laying assays where we examined the influence of osmolarity on egg-laying, statistics were performed as follows: To obtain a single biological replicate (shown as individual dots in Fig. 6), animals were staged as L4s the day before the assay on a single OP50 growth plate. Then, the next day animals from a single growth plate were split onto one normal osmolarity plate and one high osmolarity plate (10 animals per plate). Egg-laying over one hour was counted for each plate. Therefore, from each staged L4 plate, we had paired measurements of egg-laying at each osmolarity level. We then took the ratio of these values (this is what is shown as dots in Fig. 6). We obtained distributions of these ratio values for each genotype based on multiple replicates and ran a Mann-Whitney U test, comparing day-matched wild-type controls to the mutant/transgenic of interest. For Fig. 6B, which was an initial trial that tested multiple osmolarity concentrations, we instead ran a two-factor ANOVA (with genotype and osmolarity level as the two factors) on raw egg-laying rates in each condition, as well as Bonferroni-corrected Mann-Whitney U tests comparing WT versus mutants at each of the four osmolarity levels. For the genotype of interest from Fig. 6B (tax-2), additional experiments were run and statistics for those experiments were performed using the above approach (Fig. 6C shows independent data for tax-2, not overlapping with the data in Fig. 6B).
When running parametric tests, including the ANOVAs described in this section and all ttests in the manuscript, the D'Agostino-Pearson test for normality was used to test that the distributions were normally distributed. In addition, multiple comparison corrections were conducted where appropriate, as stated in figure legends. Statistics were computed using MATLAB and Graphpad Prism. G. tax-4 rescue: (A) Event-triggered averages showing average HSN GCaMP signal surrounding randomly chosen timepoints, as a control for Fig. 2B. Each line is the mean of 16 timepoints (matching the n in Fig. 2B), and this control was run 100 times, resulting in 100 lines.
(B) Event-triggered averages showing average speed (black) and HSN GCaMP (green) surrounding randomly chosen timepoints, as a control for Fig. 2C. Each line is the mean of 104 timepoints (matching the n in Fig. 2C), and this control was run 100 times, resulting in 100 lines.
(C) Event-triggered average showing average animal speed surrounding HSN calcium peaks. This plot only includes HSN calcium peaks that were not accompanied by egg-laying. Note that speeding still occurs during these peaks. (B) Event-triggered averages for time periods surrounding native egg-laying events in flp-2(gk1039) animals. Data are shown as in Fig. 1G and statistics were performed as in Fig. 1G. n = 21 animals for wild-type (518 egg events) (same data as Fig. 1G) and n = 6 animals for flp-2 (194 egg-laying events).
(C) Event-triggered averages for time periods surrounding native egg-laying events in flp-26(gk3015) animals. Data are shown as in Fig. 1G and statistics were performed as in Fig. 1G. n = 21 animals for wild-type (518 egg events) (same data as Fig. 1G) and n = 6 animals for flp-26 (162 egg-laying events). *p<0.05, Mann-Whitney U test, Bonferroni-corrected for flp-2, flp-26 and flp-28 single mutants.
(D) Event-triggered averages for time periods surrounding native egg-laying events in flp-28(flv11) animals. Data are shown as in Fig. 1G and statistics were performed as in Fig. 1G. n = 21 animals for wild-type (518 egg events) (same data as Fig. 1G) and n = 6 animals (159 egglaying events) for flp-28. *p<0.05, Mann-Whitney U test, Bonferroni-corrected for flp-2, flp-26 and flp-28 single mutants. (B) Egg-laying of HSN::HisCl animals either exposed to histamine or not. Animals for this experiment were transferred to +his or -his plates immediately before this assay (i.e. at t = 0 min), and eggs laid were counted at different time points, up to 40min after transfer. Note that egg-laying is reduced even at the first time point, indicating the HSN::HisCl inhibits egg-laying within minutes of first exposure to histamine. n = 35 plates for HSN::HisCl with histamine and 20 plates for no histamine control group, with 3 animals on each plate. (B) Fluorescent image of the vulval region of an example animal from the transgenic line that was used for laser axotomy. This strain has HSN::GFP and pan-neural::mScarlett, as indicated.
The image is being used to illustrate the site of HSN laser axotomy (note the small dot at the tip of the white arrow for axotomy site, which is the exact cut site). Scale bar, 20 um (C) Event-triggered average showing average animal velocity surrounding native, spontaneous egg-laying events. Data are shown for mock and laser control animals, in which a laser was fired adjacent to the HSN axon, at the same settings used for the actual axotomy in Fig. 5H. Data are shown as in Fig. 1G and statistics were designed in the same manner as in Fig. 1G. Lines show means and error shading shows SEM. n = 17 animals for mock (118 egg events); n = 17 animals for laser control (108 egg events). Note that the mock group was specifically paired to these laser controls (i.e. run side-by-side in same experiment) and are different animals from the mock animals in the actual axotomy (Fig. 5H).