Main

Natural selection favours behavioural traits that enhance an organism’s fitness. This process results in a wide range of behaviours adapted to specific ecological niches and facilitates the evolution of behavioural diversity across species1. Certain behaviours such as foraging, mating, predator avoidance and aggression are subject to strong environmental pressures and are therefore particularly susceptible to evolutionary change. Crucially, our nascent understanding of these complex processes stems from interspecies comparative studies. These include comparisons of Peromyscus mice which have evolved distinct burrowing behaviours for predator avoidance and parental care3,4, as well as behavioural differences between diverse Drosophila species which influence courtship signals and foraging preferences5,6,7,8. Despite this progress, the precise genetic, molecular and neural bases of behavioural adaptations are poorly understood. Nematodes with their small nervous systems and well-developed genetic and molecular tools are potent systems for understanding evolutionary behavioural processes in detail. Compared with Caenorhabditis elegans, the predatory nematode Pristionchus pacificus has evolved striking behavioural differences. This includes diversification in odorant sensitivity9, social preference10,11 and aggressive behaviours used to establish territory and remove competitors from their environment12,13,14,15. Consequently, we have leveraged the behavioural diversity between these nematode species to investigate the evolutionary adaptations moulding P. pacificus aggressive traits.

P. pacificus is an omnivorous nematode species which, in addition to feeding on bacteria, also preys on the larvae of other nematodes (Fig. 1a and Supplementary Video 1). Importantly, although predation can be used by P. pacificus to generate an extra nutrient source, these interactions also show hallmarks of aggression. More specifically, alongside the killing of other species and the cannibalism of conspecifics for nutrients14,15,16, killing is often not followed by consumption and instead can be used as an aggressive behaviour to remove rivals13. Furthermore, aggressive biting without killing also serves to displace competitors from a shared food source12. These behaviours are accomplished through the presence of teeth-like armaments and distinct pharyngeal dynamics that facilitate the puncturing of the prey cuticle and feeding on their innards16,17,18. However, it is evident that not every contact with potential prey results in an attack, indicating that P. pacificus predatory aggression is not indiscriminate (Supplementary Videos 1 and 2). Therefore, to investigate the mechanism underlying the evolution of these behaviours and their regulation in P. pacificus, we developed a high-throughput analysis tool to automatically characterize P. pacificus behaviour.

Fig. 1: Machine learning model predicts behavioural states from high-throughput tracking data of a predatory nematode.

a, Scanning electron microscopy image of P. pacificus (background nematode) with C. elegans larvae (foreground nematode). b, myo-2p::YFP expression in C. elegans compared with myo-2p::RFP expression in P. pacificus. c, Predatory P. pacificus animal surrounded by larval C. elegans prey. d, Schematic of the machine learning pipeline used to classify behavioural states. ML, machine learning; UMAP, uniform manifold approximation and projection. e, UMAP embedding of behavioural features. Colours indicate the six behavioural states identified by hierarchical clustering. f, Performance metrics of the behavioural state classifier on new, unseen data using weighted metrics. g, Distribution of key behavioural features in each state. Each point in the box plots corresponds to the mean value per state and per tracked animal. Box plots follow Tukey’s rule with the box from first to third quartiles, and a line at the median. The whiskers denote 1.5× interquartile range. h, Feature importance (Shapley additive explanations (SHAP) values) for each of the basic behavioural features involved in the model. Features derived from the same base feature are averaged. Colours correspond to the behavioural states represented in e. cwt, continuous wavelet transform. See the Methods for statistics. Scale bars, 20 μm (a), 50 μm (b), 500 μm (c).

Behavioural tracking and state prediction

Nematodes feed through their pharynx, a neuromuscular organ in the animal head. In C. elegans, a fluorophore label targeted to the pharyngeal muscle has facilitated the dissection of its feeding and locomotion behaviours across development and under different environmental conditions19. To track these behaviours in P. pacificus, we exploited a similar pharyngeal fluorophore label using the Ppa-myo-2 promoter (Fig. 1b). This plasmid was integrated into the P. pacificus genome with no adverse behavioural effects (Extended Data Fig. 1). We observed much lower levels of Ppa-myo-2 expression in the terminal bulb of the P. pacificus pharynx than in C. elegans because of the enlarged gland cells found in this space along with the absence of a hardened grinder structure in Diplogastridae20. Despite this difference, we were able to use this method to successfully track feeding and locomotion in many animals simultaneously on standard assay conditions (Fig. 1c). These consist of Ppa-myo-2p::RFP-expressing P. pacificus predators placed onto an assay arena that contains either C. elegans larvae in abundance as potential prey or a bacterial lawn as a food source. During behavioural tracking, we extracted multiple features including speed, reversals and feeding events, as well as posture-related measures using the image analysis tool PharaGlow19. To automatically identify predation-associated states, we employed a machine learning pipeline combining low-dimensional embedding and hierarchical clustering (Fig. 1d). Similar pipelines have been used for unsupervised state classification of behavioural tracking data in flies, mice, fish and C. elegans21,22,23,24,25,26,27. Using this approach, we found six distinct behavioural states (Fig. 1e and Extended Data Fig. 2a,b). We subsequently identified the behaviours in each state on the basis of the overlap with annotations of an expert human annotator (Extended Data Fig. 2c). Several of these seem to correspond to canonical states also observed in C. elegans, including ‘roaming’ and ‘dwelling’ states; however, we also identified three new behavioural states that correlate to a predatory environment. These states are labelled ‘predatory search’, ‘predatory biting’ and ‘predatory feeding’. To further validate these states, we applied the same pipeline to an independent dataset, using the same set of parameters for dimensionality reduction. We again found the same six behavioural states as in the original dataset, indicating these clusters are consistent and robust across independent experiments (Extended Data Fig. 2d–f). Next, to extend the model to unseen data, we fit an XGBoost multi-class classifier28 to the clustered data, and analysed its performance on a test set of held-out recordings. The model captures the cluster labels with greater than 95% in both accuracy and recall (Fig. 1f and Extended Data Fig. 3), with velocity, pumping rate and head swings contributing most to state identity (Fig. 1g,h). Consequently, this pipeline allowed us to predict the behavioural states for unrestrained animals (Supplementary Video 3), and provides a mechanism to investigate the molecular determinants driving the evolution of predatory aggression in P. pacificus.

Context modulates predatory drive

To determine the specificity of our identified behavioural states, we investigated the influence of the environmental and sensory context on P. pacificus state occupancy by analysing approximately 10.5 animal hours of tracking data on either a bacterial food source or larval prey. As the most descriptive features for each state are velocity and pumping rate of the animals (Fig. 1h and Extended Data Fig. 3d), the joint distribution of these two features showed distinct clusters when animals were placed on prey larvae or on bacterial food (Fig. 2a,b). By comparing these density plots with the identified behavioural states, we observed that animals preferentially occupy predation-related states when placed on larvae (Fig. 2b), whereas they spend more time in low-pumping-rate states with higher speeds when exposed to bacteria alone (Fig. 2b, right). We identified two types of search states, which we label as ‘search’ and ‘predatory search’. Both are characterized by higher speeds; however, on prey, they showed exaggerated head swing amplitudes during ‘predatory search’ making this state distinct from the ‘search’ state, which is more prominent on bacterial foods (Fig. 2c,d). These findings suggest that sensory context dictates state occupancy such that predatory states are nearly unique to prey-rich environments, further validating our previous annotations (Fig. 2c–f and Extended Data Fig. 2c). Interestingly, we found that although transitions between states and the total time spent in each state are dependent on sensory contexts, the average duration of a behavioural state remained similar for most states (Fig. 2e,f and Extended Data Fig. 4a–c). We therefore focused on the total time spent in each state as a measure of the propensity for predatory and aggressive behaviour versus non-predatory behaviour.

Fig. 2: Automatic classification of behavioural data reveals context-dependent predation drive.

a, Probability density map of velocity and pumping rate for animals on larval prey or OP50 bacteria. Scatter plots indicate the corresponding state assignments. b, Ethograms showing predicted behavioural states for animals on prey (left) or bacteria (right). c, Ethograms, velocity, pumping rate and head swings for a representative animal on larval prey (top) or bacterial food (bottom). d, Pharyngeal centreline as coloured by the assigned behavioural state. The tracks shown correspond to the grey regions in c. e, Average transition rates between behavioural states for animals on larval prey and bacterial food, respectively. Numbers in circles indicate the fraction of time per state. Arrow thickness indicates the transition rate normalized to outgoing transitions. f, Mean fraction of time spent in each behavioural state per animal. Box plots follow Tukey’s rule with the box from first to third quartiles, and a line at the median. The whiskers denote 1.5× interquartile range. Statistics were performed with Mann–Whitney U-test using Bonferroni correction for multiple tests against WT on larvae. ****P < 0.0001; n = 131 for WT on larvae, n = 92 for WT on OP50; NS, not significant. Minimum of three biological replicates. Statistics, sample size and P values are available in Supplementary Table 3. Scale bars, 250 μm (d).

Given that predation relies on nose contact with prey, we provided direct evidence for predatory-associated states by observing predator–prey interactions using a dual-colour tracking epifluorescence microscope29. We tracked individual P. pacificus predators placed in an environment containing C. elegans prey that express GFP in their body wall muscles (Extended Data Fig. 5a). Using this method, we were able to predict behavioural states of the predator from its behavioural features (Extended Data Fig. 5b,c and Supplementary Video 4) and simultaneously observe the location of the prey without considering prey information in the model. By collating and aligning all predicted biting events from multiple datasets, we found that the prey signal rose shortly before the onset of biting, confirming that animals were in contact with prey during ‘predatory biting’ states (Extended Data Fig. 5d,e). Furthermore, as parts of the prey animal were labelled, we could also observe ingestion by following the fluorescent signal of food that moved from the mouth towards the intestine (Extended Data Fig. 5f and Supplementary Video 5). These events correlated with the predicted ‘predatory feeding’ state, reinforcing our model prediction.

As predatory biting fundamentally serves two functions, nutrient-driven behaviour and aggression, we wanted to further dissect the impact of aggression on the ‘predatory biting’ state which we use as a proxy for predatory aggression throughout. Nutrition-driven biting should be followed by ‘predatory feeding’, which is ultimately the state in which material is ingested (Extended Data Fig. 5f). By contrast, when ‘predatory biting’ occurs without consecutive feeding, we expect these bites to have primarily an aggressive function. To differentiate distinct function in our data, we compared behavioural state changes in P. pacificus animals surrounded by larvae with those surrounded by both larvae and a bacterial food source.

Here, although we found that the time spent in the ‘predatory biting’ state remained relatively consistent between these conditions, animals exposed to both bacteria and larvae spent less time in the ‘predatory feeding’ state (Extended Data Fig. 5g). In addition, transitions from ‘predatory biting’ to ‘predatory feeding’ compared with larvae alone were also reduced (Extended Data Fig. 5h,i). Taken together, these data confirm that a greater proportion of contacts are aggressive in nature and non-nutritionally associated when both food choices are available, supporting the interpretation that predatory biting also serves an aggressive drive. Thus, we observe a greatly expanded state complexity in P. pacificus, going beyond the canonical foraging switch between roaming and dwelling found in C. elegans. This complexity may be a feature of the P. pacificus omnivorous lifestyle and dietary switching, or instead a characteristic of the evolution of predation.

Noradrenergic modulation of aggression

Persistent behavioural states are often stabilized by distinct neuromodulators. Frequently, these act antagonistically to set mutually exclusive patterns of behaviour, for example, by regulating the duration of opposing states such as sleep and wakefulness, or hunger and satiety30,31,32,33. Therefore, to explore whether similar circuits are involved in establishing and maintaining the aggressive predatory states versus the non-predatory states, we screened through mutants of the four major neuromodulators using strains generated by means of CRISPR–Cas9 (Supplementary Table 1)34,35. These have one-to-one orthology with their C. elegans counterparts (Extended Data Fig. 6a). Mutations generated in Ppa-tph-1, Ppa-cat-2 and Ppa-tbh-1 affect serotonin, dopamine and octopamine production, respectively, whereas Ppa-tdc-1 affects the production of both tyramine and octopamine, as these neuromodulators are in the same biochemical synthesis pathway (Fig. 3a). All mutants were crossed into the Ppa-myo-2p::RFP background and assessed using the high-throughput tracking and machine learning pipeline previously established.

Fig. 3: Noradrenergic systems modulate predatory aggression.

a, Synthesis pathway of tyramine and octopamine from the precursor tyrosine. The enzymes involved in tyramine (TDC-1) and octopamine (TBH-1) synthesis act in the same pathway. b, Example animal tracks for WT and tph-1, cat-2, tdc-1 and tbh-1 mutants. c, Probability density map of velocity and pumping rate for animals corresponding to the genotypes in b. d, Time spent in each behavioural state normalized to the total track duration. Box plots follow Tukey’s rule with the box from first to third quartiles, and a line at the median. The whiskers denote 1.5× interquartile range. Statistics were performed with Mann–Whitney U-test using Bonferroni correction for multiple tests against WT. *P < 0.05, ***P < 0.001, ****P < 0.0001; n = 91 for WT, n = 56 for Ppa-tph-1, n = 97 for Ppa-cat-2, n = 167 for Ppa-tdc-1, n = 141 for Ppa-tbh-1. Minimum of three biological replicates. e, Average transition rates between behavioural states for WT and tdc-1 and tbh-1 mutants on larval prey. The number in circles indicates the average state duration as in d and the arrow size indicates the transition rate normalized to outgoing transitions. f, tdc-1p::GFP expression in P. pacificus. Arrows indicate a putative pair of RIM neurons whereas * indicates a putative pair of RIC neurons. g, Messenger RNA of tdc-1 (red) and tbh-1 (cyan) and colocalization visualized using HCR. Arrows indicate a putative pair of RIM neurons whereas * indicates a putative pair of RIC neurons. See the Methods for statistics. Scale bars, 1 mm (b), 50 μm (f), 25 μm (g).

In C. elegans, serotonin plays a multifaceted role in its behaviour including regulating feeding, locomotion, foraging, egg laying, stress response, and learning and memory36,37,38. In P. pacificus, previous work identified a further predation-specific role for this neurotransmitter in synchronizing the action of the tooth and pharyngeal pumping, proving essential for efficient predation35. Correspondingly, Ppa-tph-1 mutants show increased roaming behaviours similar to observations in C. elegans and a strong decrease in predation consistent with previous findings (Fig. 3b–d and Extended Data Fig. 6b). Similar to Ppa-tph-1, the Ppa-cat-2 dopamine-synthesis-deficient animals also show a change in body posture and movement speed (Fig. 3b,c and Extended Data Fig. 6b); however, they do not exhibit the decrease in ‘predatory biting’ observed in Ppa-tph-1 serotonin mutants. Instead, we see a reduction in the ‘predatory search’ and ‘predatory feeding’ states (Fig. 3d). In C. elegans dopamine is critical for efficient foraging when food is present36,39, and we predict that in P. pacificus, dopamine may be required for initiating feeding behaviour post successful kill which may relate to a food reward signal.

Although the effects of serotonin and dopamine are well described in C. elegans, much less is known regarding the function of the noradrenergic neuromodulators tyramine and octopamine. Tyramine has been implicated in modulating a rapid ‘flight’ escape response by linking head movements with locomotion and also plays a role in other long-term stress responses40,41,42,43. By contrast, octopamine initiates a fasting signal facilitating exploration and optimizing foraging strategies during nutrient scarcity44,45,46. In P. pacificus, we detect significantly reduced levels of the ‘predatory biting’ state in Ppa-tbh-1 mutants and fewer transitions into this state, indicating octopamine promotes predatory aggression in P. pacificus (Fig. 3b–e). Strikingly, although Ppa-tdc-1 is required for the biosynthesis of both tyramine and octopamine, in Ppa-tdc-1 mutants predatory-associated states are maintained at wild-type (WT) levels. This indicates the further loss of tyramine suppresses the octopamine-induced predation defect (Fig. 3b–e and Extended Data Fig. 6c,d). A similar finding is observed in Ppa-tdc-1; Ppa-tbh-1 double mutants which phenocopy the Ppa-tdc-1 mutants (Extended Data Fig. 6e–h). In line with these data, applying exogenous tyramine to WT animals also induces non-predatory bouts whereas the addition of exogenous octopamine maintains high levels of predation in WT animals and rescues the low predatory aggression phenotype observed in Ppa-tbh-1 mutants (Extended Data Fig. 7a–f). Thus, in P. pacificus, these neuromodulators regulate a new behaviour absent in C. elegans. Octopamine enables robust and prolonged predatory behaviours associated with an aggressive drive whereas tyramine acts antagonistically to establish the docile, non-predatory state.

Evolution rewires noradrenergic networks

Having established the functional differences associated with these neuromodulators in P. pacificus, we next explored the neural circuits associated with their biosynthesis. In C. elegans, the interneurons RIM and RIC represent the only tyraminergic neurons, and they express Cel-tdc-1, although the RIC neurons additionally express Cel-tbh-1 and are therefore also octopaminergic41. To investigate whether the expression of these enzymes and the potential circuits involved in P. pacificus are conserved, we generated a transgenic strain expressing *tdc-1p::*GFP (Fig. 3f). Similar to C. elegans, we detected two pairs of neurons that we putatively identified as the P. pacificus RIM (anterior) and RIC (posterior) neurons. To clarify the identity of these neurons further, we also compared the soma shape, position and neurite projections of these cells with the known morphology of the P. pacificus RIM and RIC neurons acquired from the recently published head connectome47. All of these features aligned closely with the neurons in our dataset. In addition, we used in situ hybridization chain reaction (HCR) to investigate the presence of tdc-1 and tbh-1 transcripts in P. pacificus. As in C. elegans, we detected Ppa-tdc-1 transcripts in two pairs of neurons that owing to soma position and neurotransmitter expression we identified as the P. pacificus equivalent of RIM and RIC. Additionally, we detected Ppa-tbh-1 transcripts in a single neuron pair that co-localized with the posterior Ppa-tdc-1-positive neuron that we identified as RIC (Fig. 3g). Importantly, our identification of these cells is consistent with recent findings describing many of the monoaminergic neurons in P. pacificus48. Subsequently, we also attempted to confirm that these cells are the relevant neurons producing the bioamines involved in predatory aggression by genetically silencing them. In P. pacificus, expression of a histamine-gated chloride channel (HisCl) enables the inducible inhibition of targeted neurons upon the addition of exogenous histamine. Accordingly, we used the tdc-1 promoter to drive expression of HisCl in both RIM and RIC and attempted to recapitulate the tdc-1 mutant phenotype. Silencing of both neurons resulted in a partial rescue of the reduced predatory aggression phenotype observed in tbh-1 mutants. This is similar to the tdc-1 mutant rescue, confirming these cells are the probable functional origin of octopamine and tyramine (Extended Data Fig. 8). Thus, the production of tyramine and octopamine in the RIM and RIC neurons is probably conserved between C. elegans and P. pacificus although the neuromodulatory function has diverged.

Next, we attempted to elucidate the receptor circuits to further understand the molecular mechanisms involved in generating the aggressive predatory and non-predatory states. In C. elegans, three octopamine receptors have been identified, Cel-ser-3, Cel-ser-6 and Cel-octr-1 (refs. 49,50,51). We identified 1:1 orthologues in P. pacificus for all three receptors and generated CRISPR–Cas9 mutants in these genes in the Ppa-myo-2p::RFP strain (Extended Data Figs. 9 and 10a,b and Supplementary Tables 1 and 2). We then assessed the predatory aggressive drives of these animals using our behavioural state model pipeline. Although mutations in Ppa-octr-1 maintained WT predatory biting, mutations in both Ppa-ser-3 and Ppa-ser-6 phenocopied the ‘reduced predatory biting’ state observed in Ppa-tbh-1 mutant (Fig. 4a–c and Extended Data Fig. 10c). Therefore, both Ppa-ser-3 and Ppa-ser-6 are required to establish efficient P. pacificus predatory aggressive bouts through octopamine signalling. Similarly, there are four known tyramine receptors in C. elegans, Cel-tyra-2, Cel-tyra-3, Cel-ser-2 and Cel-lgc-55 (refs. 50,52,53,54). To identify potential tyramine receptors involved in this pathway, we also identified 1:1 orthologues in P. pacificus for all four of these receptors and generated corresponding CRISPR–Cas9 mutants in the Ppa-tbh-1; Ppa-myo-2p::RFP strain to determine whether any rescued the Ppa-tbh-1 reduced-killing phenotype (Extended Data Figs. 9 and 10b and Supplementary Tables 1 and 2). Mutations in Ppa-tyra-2, Ppa-tyra-3 and Ppa-ser-2 as well as the corresponding triple mutant maintained low levels of predatory aggression similar to the Ppa-tbh-1. However, in Ppa-lgc-55 mutants, predation was restored to similar levels as seen in WT and Ppa-tdc-1 mutants (Fig. 4d–f and Extended Data Fig. 10d–g). Additionally, we observe a substantial increase in the ‘predatory search’ behaviour in Ppa-tyra-3 mutants indicating a potential role for Ppa-tyra-3 in regulating this behavioural state (Fig. 4d–f and Extended Data Fig. 10d–g). Thus, two octopamine receptors and a single tyramine receptor are required to mediate the predatory states and associated aggressive transitions.

Fig. 4: Octopamine and tyramine receptors in sensory neurons gate aggressive state entry and exit.

a, Probability density map of velocity and pumping rate for WT, tbh-1 and the octopamine receptors ser-3, ser-6 and octr-1. ser-3 and ser-6 phenocopy tbh-1. b, Relative time in each behavioural state for all genotypes in a. Statistics were performed with Mann–Whitney U-test using Bonferroni correction for multiple tests against Ppa-tbh-1. n = 170 for WT, n = 197 for Ppa-tbh-1, n = 137 for Ppa-ser-3, n = 104 for Ppa-ser-6, n = 215 for Ppa-octr-1. Minimum of three biological replicates. c, Average transition rates between behavioural states for ser-3 and ser-6. The number in circles indicates the average state duration as in b, and the arrow size indicates the transition rate normalized to outgoing transitions. d, Probability density map of velocity and pumping rate for tdc-1 and the tyramine receptors tyra-2, ser-2, tyra-3 and lgc-55, all in the tbh-1 background. e, Relative time in each behavioural state for all genotypes in d. Box plots follow Tukey’s rule with the box from first to third quartiles, and a line at the median. The whiskers denote 1.5 × interquartile range. Statistics were performed with Mann–Whitney U-test using Bonferroni correction for multiple tests against Ppa-tdc-1. n = 116 for WT, n = 197 for Ppa-tbh-1, n = 235 for Ppa-tdc-1, n = 184 for Ppa-tbh-1-tyra-2, n = 298 for Ppa-tbh-1-ser-2, n = 60 for Ppa-tbh-1-tyra-3, n = 175 for Ppa-tbh-1-lgc-55. Minimum of three biological replicates. f, Average transition rates between behavioural states for tbh-1; lgc-55. The number in circles indicates the average state duration as in e and the arrow size indicates the transition rate normalized to outgoing transitions. g, Comparative expression pattern analysis for the octopamine receptors ser-3 and ser-6, as well as the tyramine receptor lgc-55, in P. pacificus (top) and C. elegans (bottom). Arrow indicates putative IL2 neurons in P. pacificus. Arrow with * indicates putative OL cell neurites in P. pacificus. Scale bar, 50 μm. See the Methods for statistics. *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001.

To identify the neurons governing behavioural state-switching, we first used HCR to detect the transcripts of the Ppa-ser-3, Ppa-ser-6 and Ppa-lgc-55 receptors associated with the predatory states. Many head cells were positive for these transcripts making it difficult to determine cellular identification from transcript position alone (Extended Data Fig. 10h). Consequently, we also generated transcriptional reporter lines for all three receptors and compared these with reporter lines of the same receptors in C. elegans. The octopamine receptor Ppa-ser-3 is expressed in the neck muscles, several head neurons and strikingly the IL2 and IL1 head sensory neurons with their distinctive neurites projecting to the worm’s nose47. Ppa-ser-6 is expressed in a non-overlapping set of head neurons which also includes a neuron pair with anterior processes (Fig. 4g). For the tyraminergic receptor Ppa-lgc-55, we again observe robust expression in the neck muscles which is similar to observations in C. elegans52, but we also detect strong expression in a distinct set of head sensory neurons separate from the octopamine receptor expressing IL2s and IL1s (Fig. 4g). These are putatively identified as the OL neurons on the basis of neurite morphology and soma placement matching data from electron microscopy47. Using the CeNGEN dataset55 and transcriptional reporter lines, we found these three receptors in C. elegans are expressed throughout a large subset of head neurons including sensory, inter and motor neurons (Extended Data Fig. 10i). However, there is minimal overlap with any of the head sensory neurons we observe in P. pacificus (Fig. 4g). Therefore, in P. pacificus the octopamine and tyramine receptors regulating aggressive behavioural states are localized to head sensory neurons which are distinct from their localization in C. elegans.

Silencing IL2 neurons disrupts predation

Some of the most distinctive neurons receiving octopamine signals in P. pacificus are the Ppa-ser-3-expressing IL2 sensory neurons. In C. elegans, these neurons project sensory endings into the environment and are associated with nictation behaviour and sensory modulation56 but they do not express any octopamine receptors[55](https://www.nature.com/a