Locusts employ neuronal sensory prioritization to reconcile two conflicting olfactory signals while aggregating

Significance

The ability to decode complex olfactory signals and display appropriate responses is crucial for insect survival and adaptation. Gregarious locusts benefit from the integration of two pheromones, 4-vinylanisole (4VA) and phenylacetonitrile (PAN), which respectively facilitate population aggregation and predator resistance. Our findings reveal that 4VA and PAN are emitted sequentially and synergistically. Behavioral experiments involving blends of 4VA and PAN consistently elicited an attractive effect. In the antennae, 4VA inhibited the neural responses to PAN. The faster conduction velocities of 4VA compared to PAN in the antennal lobe ensured an attractive behavioral outcome. These multilayered interactions underscore the predominant role of 4VA in conspecific interactions and highlight an adaptive strategy that migratory locusts have evolved in response to complex stimuli.

Abstract

Migratory locusts (Locusta migratoria) emit two key odorants during aggregation: 4-vinylanisole (4VA), which serves as an aggregation pheromone attracting conspecifics to form swarms, and phenylacetonitrile (PAN), which acts as an aposematic signal and a precursor of a defense toxin, deterring conspecifics from cannibalism and protecting against predators. However, how locusts reconcile these two conflicting olfactory signals while aggregating is not yet understood. Our study addresses this by examining the release dynamics of the two signals, their behavioral effects, and the neural mechanisms underlying their perception. 4VA is released earlier and at lower locust densities than PAN, with PAN’s release increasing as aggregation progresses. Although PAN’s emission levels eventually exceed those of 4VA, locusts consistently exhibit a preference for the emitted blend, regardless of variations in proportions and concentrations. Notably, increasing amounts of 4VA added to PAN can counteract PAN’s repellent effects, but this is not the case when PAN is added to 4VA. Mechanistically, we found that antennal neurons responsive to 4VA suppress the activity of neurons responsive to PAN. In the antennal lobe, it is the conduction velocities of projection neurons, rather than other neural properties, that are responsible for the observed behavioral pattern, leading to an overall attractive response. Collectively, our findings imply that insects are capable of harmonizing the effects of two distinct pheromones to optimize both social cohesion and chemical defense.

For animals that live in groups, it is important to balance the benefits and risks of aggregation. Gregarious locusts can release two pheromones, 4-vinylanisole (4VA) and phenylacetonitrile (PAN), which are essential for swarm formation but perform entirely different biological functions (1–3). 4VA, as an aggregation pheromone, attracts conspecifics of the migratory locust (2), accelerates the acquisition of gregarious behavior (4) and promotes the synchronization of sexual maturation in female locusts (5). In contrast, PAN repels conspecifics and has been identified as an olfactory aposematic signal and a precursor of a toxin that protects against natural predators (3) and cannibalism (1). 4VA and PAN are both released by gregarious locusts, but PAN is emitted at a level much higher than 4VA (2, 6). These variations raise intriguing questions about how these insects interpret the chemical signals with opposing valences and how their nervous systems process mixed signals that convey conflicting information, ultimately enabling locusts to make adaptive behavioral decisions.

Different odorants can elicit a variety of physiological and behavioral effects within a species, often with odorants of opposing valence counteracting each other’s effects. In the nematode C. elegans, a repulsive odorant representing a threat can strongly block the attraction of other odorants representing food. The gene osm-5 acts in AWB and ASH, which are two sensory neurons known to mediate 2-nonanone-evoked avoidance, to regulate chemotaxis and olfactory integration of repellent and attractive odorants (7). In the antennal lobes (ALs) of the fruit fly Drosophila melanogaster, γ-aminobutyric acid–ergic (GABAergic) local neurons mediate the inhibitory crosstalk between attractant-responsive and repellent-responsive glomeruli (8). Activation of the glutamatergic mushroom body output neurons triggers the innate CO₂ avoidance behavior in fruit flies, and the innate avoidance can be reduced by the activation of dopaminergic neurons (9). In mice, attractive and aversive odorants can neutralize each other’s behavioral effects, and the blocking effects require sensory input from the trace amine-associated receptors (10). These studies on odorants with opposing valence decisions have revealed that the capacity to evaluate and integrate sensory cues with different valences is both evolutionarily conserved and functionally essential across animals. However, it is poorly understood how different odorant combinations are precisely regulated from the periphery to the central nervous system, nor is it clear what the adaptive significance of behavioral decisions might be. Locust perception of 4VA and PAN offers an opportunity to explore the decision-making mechanisms that underlie the integration of two distinct olfactory signals.

In the present study, we conducted a comprehensive investigation into the dynamics of pheromone release and assessed individual and combined effects of specific pheromones on the behavioral and neural responses of locusts. We examined the release patterns of 4VA and PAN under various crowding conditions to pinpoint the critical population density thresholds that trigger pheromone emissions and to understand the dynamics of this emission process. We found that 4VA can effectively counteract the repellent effects of PAN across a range of concentrations and ratios, thereby inducing attractive behaviors in locusts. This response is believed to be due to the synergistic interaction between 4VA and PAN in the antenna and AL. Our findings indicate that 4VA plays a dominant role in aggregation, with locusts integrating the complex combined effects of 4VA and PAN to bolster swarm cohesion and ensure that predators are kept at bay.

Results

4VA and PAN Are Sequentially Released during Aggregation.

We first assessed the population density threshold for the release of 4VA and PAN during the initial stage of aggregation. Solitary locusts were exposed to crowding treatments with varying population densities for 72 h, and the emissions of 4VA and PAN were quantified. Based on the fitting curve and the threshold calculation, we found that the presence of 16 to 17 solitary locusts in a cage initiated the release of PAN, while just 4 or 5 locusts in an identical cage could trigger the release of 4VA (Fig. 1A). Thus, a lower-density threshold is required for 4VA release compared to PAN.

Fig. 1.

4VA and PAN release patterns of locusts in response to changes of population density. (A) The emissions of 4VA and PAN from solitary locusts after 72 h of exposure to different population densities. n = 7, 4, 5, 8, 7, 10, and 8 biological replicates. (B) The relative release levels of PAN to 4VA in solitary locusts after crowding for 0 d, 1 d, 2 d, 3 d, 4 d, 5 d, and 6 d. n = 4, 5, 5, 5, 6, 6, and 5 biological replicates. (C) The release contents of PAN are approximately 10-fold high than 4VA contents in the gregarious locusts. Data are presented as mean ± SEM. Data are analyzed using two-tailed unpaired t test (B and C). *P < 0.05; **P < 0.01; ***P < 0.001; NS, not significant.

To determine the release dynamics of 4VA and PAN, the solitary locusts were crowded for different durations. The solitary locusts started to release 4VA after 1 d of crowding, while PAN was released later, after 2 d of crowding (Fig. 1B). From the third day of crowding, the treated solitary locusts exhibited higher emission levels of PAN than 4VA, and the PAN-to-4VA ratio increased significantly (Fig. 1B). The relative contents between 4VA and PAN of solitary locusts after crowding for 6 d approached approximately 1:10, and this release pattern was similar to that of typical gregarious locusts, which have 1:10.4 the relative amounts of 4VA and PAN (Fig. 1 B and C). Therefore, 4VA can be viewed as a pioneer compound, while PAN gradually becomes the main component during the aggregation process.

4VA Dominates the Attractive Behavioral Response to the Mixture of 4VA and PAN.

As the locusts always smell the mixture of 4VA and PAN simultaneously, we evaluated the behavioral responses of gregarious locusts to 4VA, PAN, and the mixture of 4VA and PAN in a well-established dual-choice arena system (2) (Fig. 2A). For individual 4VA and PAN, the locusts exhibited significant attraction to 4VA and repulsion to PAN at different concentrations, respectively (Fig. 2 B and C and SI Appendix, Fig. S1 A and B). Based on the physiological ratio of 4VA and PAN (1:10), we detected the responses of locusts to their mixtures. The mixtures of 4VA and PAN consistently elicited significant attraction responses in gregarious locusts (Fig. 2D and SI Appendix, Fig. S1C). Then, we decreased and increased the ratio of 4VA and PAN to 1:100 and 1:1, respectively. Surprisingly, the locusts still exhibited significant attraction responses to these mixtures (Fig. 2 E and F and SI Appendix, Fig. S1 D and E). Thus, as long as 4VA is present, the mixture can attract locusts regardless of the concentration and ratio of PAN.

Fig. 2.

4VA dominates the synergistic effects of the mixture on gregarious locusts. (A) Setup for behavioral tracing in each zone of a locust in a dual-choice arena system. (B) The behavioral responses of gregarious locusts to 4VA at different concentrations (0.5, 5, 50 ng/μL). n = 25, 30, and 14 locusts, respectively. (C) The behavioral responses of gregarious locusts to PAN at different concentrations (5, 50, 500 ng/μL). n = 24, 30, and 28 locusts, respectively. (D–F) The behavioral responses of gregarious locusts to the mixture of 4VA and PAN with different ratios (4VA: PAN, 1:10, D), (4VA: PAN, 1:100, E), (4VA: PAN, 1:1, F). n = 28, 68, 43 (D), 50, 26, 43 (E), 26, 26, 43 (F) locusts, respectively. (G) The behavioral responses of gregarious locusts to 4VA (50 ng/μL) and PAN with five different concentrations, versus 4VA (50 ng/μL). n = 26, 26, 29, 29, and 28 locusts, respectively. (H) The behavioral responses of gregarious locusts to 4VA (5 ng/μL) and PAN with four different concentrations, versus 4VA (5 ng/μL). n = 23, 24, 24, and 24 locusts, respectively. (I) The behavioral responses of gregarious locusts to 4VA (0.5 ng/μL) and PAN with four different concentrations, versus 4VA (0.5 ng/μL). n = 28, 29, 25, and 27 locusts, respectively. (J) The behavioral responses of gregarious locusts to PAN (500 ng/μL) and 4VA with six different concentrations, versus PAN (500 ng/μL). n = 30, 27, 23, 25, 28, and 27 locusts, respectively. Behavioral data are presented as mean ± SEM. P values in behavioral assays are determined by Wilcoxon signed rank test. *P < 0.05; **P < 0.01; ***P < 0.001; NS, not significant.

To determine whether 4VA displays a predominant role in the mixture, we designed a series of mixtures versus 4VA or PAN and conducted behavioral assays in the behavioral paradigm. First, the natural concentration of 4VA (50 ng/μL) was applied as the control, while varying amounts of PAN were added to 4VA on the other side (Fig. 2G). The gregarious locusts exhibited neutral responses to 4VA versus 4VA with PAN added at 0 to 5,000 ng/μL (Fig. 2G and SI Appendix, Fig. S1F). Furthermore, to determine the concentration threshold of 4VA inducing the neutral behavioral responses of gregarious locusts to both 4VA and PAN, two concentrations of 4VA (5 ng/μL, 0.5 ng/μL) were applied as the control, while varying amounts of PAN were added to 4VA on the other side. When 4VA at 5 ng/μL was used as the control, gregarious locusts exhibited neutral responses to 4VA versus 4VA with PAN at 0 to 500 ng/μL (Fig. 2H and SI Appendix, Fig. S1G). When 4VA at 0.5 ng/μL was used as the control, gregarious locusts showed neutral responses to 4VA versus 4VA with PAN (0 and 0.5 ng/μL), but exhibited repellent responses to 4VA with PAN (5 to 50 ng/μL) (Fig. 2I and SI Appendix, Fig. S1H). Then, we evaluated behavioral responses of gregarious locusts to PAN at the natural concentration (500 ng/μL) versus PAN mixed with different amounts of 4VA (Fig. 2J). The locusts preferred the side treated with PAN mixed with 4VA (5 to 5,000 ng/μL), while the gregarious locusts showed neutral responses to PAN versus PAN with 4VA (Fig. 2J and SI Appendix, Fig. S1I). Thus, 4VA dominates the attractive behavioral response to the mixture of 4VA and PAN at different ratios and concentrations, and 4VA at 5 ng/μL serves as the threshold concentration inducing the neutral behavioral responses of gregarious locusts to both 4VA and PAN.

4VA Suppresses PAN Responses in Antennae.

To explore why 4VA dominates the behavioral responses of locusts to the mixture, we first measured the electrophysiological responses of whole antennae using the electroantennogram (EAG) (Fig. 3A). The antennae displayed stronger overall potential to PAN than to 4VA at the same concentration (Fig. 3B). Then, we tested the antennal responses to the mixture (4VA:PAN, 1:100, 1:10, and 1:1) compared to PAN or 4VA alone. We found that the mixtures elicited a significantly reduced response compared to PAN alone (Fig. 3D), while antennal responses to the mixture and 4VA alone showed no significant differences (Fig. 3C).

Fig. 3.

4VA suppresses the responses of antennae to PAN in locusts. (A) Schematic diagram of the EAG system. (B) The EAG responses of locust antennae to individual PAN and 4VA at different concentrations (5 ng/μL, 50 ng/μL, 500 ng/μL). n = 30 antennae. (C) The EAG responses of locust antennae to individual 4VA and the mixture of 4VA and PAN at different ratios (4VA: PAN, 1:100, 1:10, 1:1), and in the mixture, the concentration of 4VA is always 50 ng/μL. n = 30 antennae. (D) The EAG responses of locust antennae to individual PAN (500 ng/μL) and the mixture of 4VA and PAN at different ratios (4VA: PAN, 1:100, 1:10, 1:1), and in the mixture, the concentration of 4VA is always 50 ng/μL. n = 30 antennae. (E) Schematic diagram of the single sensillum recording (SSR) system. (F) Representative spike traces of basiconic sensilla in response to individual PAN and 4VA at different concentrations (5, 50, 500 ng/μL). And the responses of basiconic sensilla to individual PAN and 4VA at different concentrations (5 ng/μL, 50 ng/μL, 500 ng/μL). n = 38 sensilla. (G) Representative spike traces and the responses of basiconic sensilla in response to individual 4VA, and the mixture of 4VA and PAN at different ratios (4VA: PAN, 1:100, 1:10, 1:1), and in the mixture, the concentration of 4VA is always 50 ng/μL. n = 38 sensilla. (H) Representative spike traces and the responses of basiconic sensilla to individual PAN and the mixture of 4VA and PAN at different ratios (4VA: PAN, 1:100, 1:10, 1:1), and in the mixture, the concentration of 4VA is always 50 ng/μL, n = 38 sensilla. Neural data are presented as mean ± SEM. P values are determined by two-tailed unpaired t test. *P < 0.05; **P < 0.01; ***P < 0.001; NS, not significant.

Because only basiconic sensilla respond to 4VA (2) and PAN (SI Appendix, Fig. S2) in four types of sensilla, we carried out single sensillum recording (SSR) to test the potentials of basiconic sensilla to 4VA, PAN and the mixture of 4VA and PAN (Fig. 3E). The responses of basiconic sensilla to PAN remained stronger than those to 4VA (Fig. 3F). The basiconic sensilla did not show significant differences in responses to 4VA and their mixture (Fig. 3G), but showed significantly lower responses to the mixture than those to PAN (Fig. 3H). Therefore, 4VA’s presence significantly reduces the responses of locusts to PAN, eliminating the neural strength differences between these two olfactory signals.

4VA Neurons Antagonize the Responses of PAN Neurons.

To determine how 4VA interacts with PAN in the antennae, we applied RNAi knockdown of Or35 (the specific odorant receptor of 4VA) using dsRNA (SI Appendix, Fig. S3A and Table S1) to block the activation of 4VA responsive neurons. In the dsOr35 injection group, antennal and basiconic sensilla responses to 4VA were significantly reduced compared to those in dsGFP-injected controls, whereas the responses to PAN did not show any significant difference in both antennae and basiconic sensilla (Fig. 4 A and B). Furthermore, after the stimulation of the mixtures at different ratios (4VA: PAN 1:100, 1:10, and 1:1), the electrophysiological responses of antennae and basiconic sensilla exhibited a significant increase in the Or35 RNAi knockdown group, compared to the dsGFP-injected group (Fig. 4 A and B). Subsequently, we tested the behavioral responses of gregarious locusts to the mixtures at different ratios after dsOr35 injection (Fig. 4C). The locusts injected with dsOr35 showed repulsive responses to the mixture (Fig. 4C and SI Appendix, Fig. S4B), while the dsGFP-injected group as control showed significant attraction response to the mixture (Fig. 4C and SI Appendix, Fig. S4A). Thus, 4VA perception antagonizes the responses of PAN in the mixture.

Fig. 4.

OR35 mediates the inhibition effects of 4VA to PAN in locust antennae. (A and B) The responses of antennae (A) and basiconic sensilla (B) to 4VA, PAN, and the mixture of 4VA and PAN at different ratios (4VA: PAN, 1:100, 1:10, 1:1) in locusts injected with dsRNA of GFP and Or35. n = 19 (dsGFP) and 18 (dsOr35) antennae (A), and 16 (dsGFP) and 24 (dsOr35) sensilla (B), respectively. (C) The behavioral responses of gregarious locusts to oil and the mixture of 4VA and PAN at different ratios (4VA: PAN, 1:100, 1:10, 1:1) in dsGFP-injected and dsOr35-injected locusts. n = 27, 36, 21 (dsGFP), 15, 22, and 15 (dsOr35) locusts, respectively. (D and E) The responses of antennae (D) and basiconic sensilla (E) to 4VA, PAN, and the mixture of 4VA and PAN at different ratios (4VA: PAN, 1:100, 1:10, 1:1) in locusts injected with dsRNA of GFP and Or70a. n = 27 (dsGFP), 23 (dsOr70a) antennae (D), and 12 (dsGFP), 24 (dsOr70a) sensilla (E), respectively. (F) The behavioral responses of gregarious locusts to oil and the mixture of 4VA and PAN at different ratios (4VA: PAN, 1:100, 1:10, 1:1) in dsGFP-injected and dsOr70a-injected locusts. n = 22, 25, 20 (dsGFP), 20, 30, and 27 (dsOr70a) locusts, respectively. Data are presented as mean ± SEM. P values in neural assays are determined by two-tailed unpaired t test (A, B, D, and E). P values in behavioral assays are determined by Wilcoxon signed rank test (C and F). *P < 0.05; **P < 0.01; ***P < 0.001; NS, not significant.

As OR70a was identified as the odorant receptor of PAN (1) (SI Appendix, Table S1), we detected the responses of locusts to the mixtures in both EAG and SSR after Or70a knockdown. After dsOr70a injection, Or70a expression level were significantly reduced compared to locusts injected with dsGFP (SI Appendix, Fig. S3B). The responses of antennae and basiconic sensilla to PAN were significantly reduced in the Or70a knockdown group, compared to those of the control groups injected with dsGFP (Fig. 4 D and E). After stimulation of the mixtures at different ratios, the electrophysiological responses of antennae and basiconic sensilla exhibited no significant differences between the locusts injected with dsOr70a and the control groups injected with dsGFP (Fig. 4 D and E). In addition, the locusts injected with dsOr70a still showed attraction responses to the mixture (Fig. 4F and SI Appendix, Fig. S4D), showing the consistency with that of the control group injected with dsGFP (Fig. 4F and SI Appendix, Fig. S4C). Therefore, 4VA can affect PAN perception for the mixture in the peripheral nervous system, but not vice versa.

Neural Conduction Velocity Differentiates PAN and 4VA in the AL.

The AL of the insect brain is the olfactory center, in which projection neurons (PNs) process and convey olfactory information to the higher brain center (11). We first recorded the electrophysiological responses of individual PNs in the AL of locusts using intracellular recording. The PNs that showed strong responses to 4VA or PAN were labeled with a fluorescent dye (SI Appendix, Fig. S5 A and B). The labeled PNs, respectively, responding to 4VA and PAN exhibited typical morphological features of PNs, displaying axonal innervation of the ipsilateral mushroom body calyx and lateral horn, and dendritic branches in numerous microglomeruli in the AL (Fig. 5 A and B).

Fig. 5.

PNs respond similarly to mixtures and 4VA in locust AL. (A and B) The responsive PNs to 4VA (A) and PAN (B) labeled using intracellular recording, exhibit typical morphological features of PNs in locusts, including axonal innervation of the ipsilateral calyx of mushroom body (MB) and lateral horn, and dendritic branches in many microglomeruli in the AL. (C) Schematic drawing of the multichannel recording of AL of locust. (D) Spike rasters of five representative PNs in response to 4VA, PAN, and the mixture of 4VA and PAN in AL. (E–G) PN population trajectories for 4VA (blue trace), PAN (red trace), and the mixture of 4VA and PAN with different ratios (4VA: PAN, 1:10, brown trace, E), (4VA: PAN,1:100, yellow trace, F), (4VA: PAN,1:1, green trace, G) are shown in PCA space (computed using data for 10 neurons). (H–J) Visualization of ensemble PN responses using the first three PCs (PC1, PC2, and PC3). Each dot represents a specific odor stimulus, ED1: Euclidean distance between 4VA and the mixture with different ratios (4VA: PAN, 1:10, H), (4VA: PAN, 1:100, I), (4VA: PAN, 1:10, J). ED2: Euclidean distance between PAN and the mixture with different ratios (4VA: PAN, 1:10, H), (4VA: PAN, 1:100, I), (4VA: PAN, 1:10, J).

High-throughput electrophysiological responses of PNs in the AL were recorded using multichannel recording (Fig. 5C). Among the total recorded 99 units, 43 units responded to both 4VA and PAN, while only 7 and 3 units specifically responded to 4VA and PAN, respectively (SI Appendix, Fig. S5C). Then, we calculated the firing rate of these PNs after applying the stimuli of the mixtures (Fig. 5D and SI Appendix, Fig. S5D). PN population dynamics, which refer to the trajectories of PN populations in response to 4VA, PAN, and their mixtures, were analyzed after dimension reduction via principal component analysis (PCA). The results indicate that the PN population dynamics in response to the mixture at different ratios (4VA: PAN 1:100, 1:10, and 1:1) are similar to those of 4VA, but are significantly different from those in response to PAN alone (Fig. 5 E–G). Furthermore, we constructed projections of these three stimuli represented each odor stimulus by the three-dimensional centroid of its trajectory, plotted these centroids in a three-dimensional space, and calculated the Euclidean distances from the mixture to PAN (ED1) and to 4VA (ED2) based on their centers of gravity (Fig. 5 H–J). The mixtures exhibited much shorter Euclidean distances to 4VA than those to PAN. Thus, the spatiotemporal patterns of PNs in response to mixture show similarity to that of 4VA rather than those of PAN.

To further explore neural mechanisms underlying the responding similarity between 4VA and their mixtures at different ratios (4VA: PAN 1:100, 1:10, and 1:1), the histograms (mean ± SEM) of all responsive PNs were calculated using instantaneous frequencies and cross-correlograms. No significant differences of response frequency were detected among 4VA, PAN, and their mixtures (Fig. 6 A–C). We then calculated the spike timings of 4VA, PAN, and their mixtures by analyzing peristimulus time scatter diagram using instantaneous frequencies of all PNs. The peak time of 4VA, PAN, and their natural mixture (4VA: PAN = 1:10) occurred at time 397 ms, 433 ms, and 389 ms, respectively. The peak time of 4VA and the mixture were very close, while the peak time of PAN was significantly slower than the other two (Fig. 6D). Additionally, the peak response times for the mixtures with ratios of 4VA: PAN at 1:100 and 1:1 were also similar to that of 4VA, but were faster than that of PAN (Fig. 6 E and F). Thus, the peak response times of 4VA and the mixture are significantly different from that of PAN.

Fig. 6.

Neural conduction velocity differentiates PAN and 4VA in locust AL. (A–C) Cross-correlogram of all PNs to 4VA, PAN, and the mixture at different ratios (4VA: PAN, 1:10, A), (4VA: PAN, 1:100, B), (4VA: PAN, 1:10, C). (D–F) Average responses of PNs to different stimuli, 4VA (blue trace), PAN (red trace), and the mixture with different ratios (4VA: PAN, 1:10, D), (4VA: PAN, 1:100, E), (4VA: PAN, 1:1, F) (brown trace). (G) Advanced the response window of variable PAN forward by 40 ms to align its response peak with that of variables 4VA. (H) PN population trajectories for 4VA (blue trace), PAN (red trace), and the forward-moved PAN (blank trace) are shown in PCA space. (I) Visualization of ensemble PN responses using the first three PCs (PC1, PC2, and PC3). Each dot represents a specific odor stimulus, ED1: Euclidean distance between the 4VA and PAN. ED2: Euclidean distance between 4VA and the forward-moved PAN.

To verify that the differences were caused by the properties of the two compounds themselves, we calculated the peak response times for different concentrations of 4VA and PAN. As the concentration increased, the peak response times of both compounds speeded up, but at each concentration, 4VA was consistently faster than PAN SI Appendix, Fig. S5 E-J). When shifted the response window of variable PAN forward by 40 ms (Fig. 6G), we recalculated the odor-specific trajectories in the spatiotemporal patterns using PCA. The results confirmed that the response patterns and the Euclidean distance of PAN gradually converge toward 4VA (Fig. 6 H and I). Thus, the conduction velocity to the mixture and 4VA contribute to the spatiotemporal differences from that of PAN in AL, which potentially lead to the final behavioral decision for attraction in locusts.

Discussion

Our research has uncovered the intricate mechanism by which locusts integrate conflicting olfactory signals to optimize individual benefits during aggregation. 4VA is released more quickly and requires a lower population density threshold to trigger a response to changes in population density. This mechanism ensures that aggregation is initiated prior to the activation of group defense responses. Both the peripheral and central nervous systems play a role in processing 4VA and PAN, but they are specialized for different neural response features. Antennal neurons exhibit a reduced response to PAN when 4VA is present in the mixture, ensuring that the neural signal intensities of both compounds transmitted to the AL remain consistent, regardless of the ratio of 4VA to PAN perceived by the antennae. The AL PNs have varying conduction velocities for different odors, enabling them to distinguish between 4VA and PAN. The faster conduction velocity for 4VA outpaces that for PAN, which leads to an attractive behavioral response when both compounds are detected simultaneously (Fig. 7).

Fig. 7.

The synergistic interactions between 4VA and PAN across multiple levels facilitate locust attraction and aggregation. In the antennae, the interaction between 4VA and PAN is observed in terms of reaction intensity, characterized by a reaction strength to the mixed odor that closely resembles that of 4VA, but is lower than that of PAN. Within the central nervous system (AL), this interaction is manifested in the context of signal conduction velocity, as both the mixed odor and 4VA exhibit faster transmission of signals, than that of PAN. Finally, locusts make decision for conspecific attraction and aggregation.

4VA Plays Predominant Roles in Locust Aggregation.

Our study reveals that the influence of 4VA consistently overrides that of PAN, suggesting that the recruitment of individuals into groups takes precedence over defending against predators for the establishment and sustenance of swarms. In the initial stage of aggregation, 4VA is triggered quickly and starts to function. As the aggregation progresses, PAN triggers pronounced defense against predators, particularly under the conditions of high population density where they inevitably encounter the risk of predation (3) and intraspecific cannibalism (1). In the groups without social hierarchy and labor division, locust individuals exhibit an equal risk of predation. As the number of individuals within a group grows, the risk borne by each individual diminishes, reflecting the benefits of group living (12–14). Thus, the locusts must devise an approach to synergistically exert the functions of 4VA and PAN, while ensuring the growth of their population.

Multilevel Processing of Conflicting Olfactory Information.

The neural processing of the mixture of 4VA and PAN involves a complex, multilevel process that coordinates activities from the peripheral to the central nervous system. Typically, the central nervous system is responsible for making the final decision, but for rapid evaluation, a stepwise refinement of the input signal is more efficient. In the peripheral system, 4VA can inhibit the responses of PAN, helping to neutralize the impact of concentration differences in mixed odors and potential variations in receptor expression levels. In previous studies, it was found that treating human arm skin with the mosquito repellent N,N-diethyl-meta-toluamide (DEET) significantly reduced the release of volatile compounds from the skin (15). Therefore, we first considered whether there might be a physical interaction between 4VA and PAN, which could indirectly affect locusts’ avoidance behavior toward PAN, thereby influencing the final attraction response to the mixture of the two compounds. We conducted corresponding tests (SI Appendix, Fig. S6A), however, the results demonstrated no detectable difference in airborne concentrations between the two compounds (SI Appendix, Fig. S6B), proving that the release of the mixed odor of 4VA and PAN revealed no physical interaction between them.

Previous studies in insects have proposed two potential mechanisms for mixture inhibition. The first is competitive binding, where two odorant molecules compete for the same receptor binding site, allowing one component to inhibit the response to another (16, 17). The second is ephaptic interactions, where neighboring olfactory receptor neurons (ORNs) within the same sensillum can inhibit each other through local electrical interactions (17, 18). Within the peripheral olfactory system, each ORN typically expresses only a single olfactory receptor (OR) gene (19, 20). In the migratory locust, 4VA and PAN bind to their specific receptors, OR35 and OR70a, respectively. 4VA and PAN do not competitively bind to each other’s receptor (1, 2), implicating that the interaction between these two pheromones is not due to a mixture inhibition. The inhibitory effect of 4VA on PAN is more likely due to nonsynaptic electrical coupling interactions, however, the underlying mechanisms need more investigations. Within the central nervous system (AL), the transmission rate of 4VA is faster than that of PAN. This is probably due to the unique sensory afferents of 4VA after long-term adaptation, enabling locusts to make efficient behavioral choices when faced with attractive stimuli. These neural strategies culminate in the final behavioral outcome of attraction when of 4VA and PAN coexist.

Behavioral Trade-Offs and Neural Interactions in Processing Complex Signals.

When organisms encounter various pheromones with important biological functions, their ultimate behavioral choices become critical for survival and reproductive advantage. The survival threat signal in Caenorhabditis elegans, 2-nonanone, can strongly block signals representing food compounds such as isoamyl alcohol or benzaldehyde (7). In Drosophila, fermentation odor component ethyl acetate can be suppressed by benzaldehyde, which represents a toxic odor (8), whereas the stress odor CO₂ is inhibited by acetic acid, representing a food odor in a different context (9). Mechanistically, these behavioral outcomes are predominantly linked to the reciprocal interactions among neurons. Certain fruit odors can inhibit CO₂-sensitive ORNs, when flies feed on fermenting fruits that release both food odors and the naturally aversive odor CO₂ (9). GABAergic local neurons mediate the inhibitory crosstalk between the attractant- and repellent-responsive glomeruli (8). The activation of food odor-responsive dopaminergic neurons reduces innate avoidance mediated by CO₂-responsive mushroom body output neurons (9). Given the responses of locust antennal neurons and PNs in processing 4VA and PAN, we speculate that animals share universal rules for processing complex signals, although the current research focuses on various aspects of neural systems.

Materials and Methods

Insects.

The gregarious locusts (Locusta migratoria) were raised in a well-ventilated, temperature-controlled (30° ± 2 °C) and humidity-controlled (60 ± 5%) greenhouse with a stable photoperiod (L14:D10) located at the Institute of Zoology, Chinese Academy of Sciences, Beijing, China. Then, 800 to 1,000 first-instar nymphs were reared in each cage measuring 30 cm × 30 cm × 30 cm. The food consisted of fresh wheat seedlings grown in the greenhouse and wheat bran (21).

Chemicals.

Mineral oil, 4VA, and PAN were purchased from Sigma-Aldrich (St. Louis, MO, USA).

Volatile Collection of Locusts.

Locust volatiles were gathered using static solid-phase microextraction (SPME) for 30 min, following our previous methodology (6). Briefly, a Polydimethylsiloxane/Divinylbenzene 65 μm fiber was placed in a glass jar (10.5 cm height × 8.5 cm internal diameter), with the tip inserted approximately 1 cm from the stainless steel lid (9 cm diameter with 2 mm diameter holes spaced 2 mm apart), which was used to confine a group of three fifth instar nymphs from various locust colonies. The nymphs were positioned at the bottom of the jar to prevent direct contact with the fiber. A control was performed by collecting SPME volatiles from an empty jar for 30 min. The fibers with absorbed odors were then analyzed chemically.

To assess the PAN production threshold in solitary locusts under crowding conditions, first, we individually subjected 1, 5, 10, 15, 20, 25, and 30 solitary fifth instar nymphs to crowding in 10 cm × 10 cm × 10 cm cages for 72 h, and then measured the odor release levels of the locusts under different density treatments. Then, we crowded 30 fourth instar solitary nymphs in a 10 cm × 10 cm × 10 cm cage, with the first day of treatment designated as 0 d. Starting from 4 d (by which time they had molted into the fifth instar), we measured their odor release daily until 6 d. In parallel, we reared solitary nymphs under the same conditions and crowded 30 fifth instar nymphs in an identical 10 cm × 10 cm × 10 cm cage. For this group, we began daily odor release from 0 d of treatment and continued until 3 d.

To address whether the physical interactions between 4VA and PAN could retard or enhance the release of one or both compounds, we detected the release of 4VA and PAN from their mixture. The 1:1 mixture of PAN and 4VA was applied to filter paper positioned at the bottom of sample vials. Volatiles were collected via static SPME and subsequently subjected to chemical analysis (SI Appendix, Fig. S6A).

Locust Behavioral Responses and Video-Tracking System.

We performed behavioral assays with fifth-instar nymphs to determine the behavioral responses of locusts to 4VA, PAN, and mixtures of 4VA and PAN. We used the same vertical airflow olfactometer as described in previous studies (2, 3) (Fig. 2A). Briefly, underneath a stainless steel surface measuring 60 cm × 30 cm and covered with small holes, there were two square conical stainless steel funnels connected to each other, each with a length of 28 cm. Each funnel had a square stainless steel sheet (2.8 cm × 2.8 cm) with small holes. The diameter of the small holes on the surface and steel sheets was 2 mm with an interval of 5 mm. A 60 cm × 30 cm × 30 cm acrylic glass (poly(methyl methacrylate)) frame was placed above the stainless steel surface, and the above setup was placed in a 150 cm × 100 cm × 180 cm observation room with ventilation and a ceiling-mounted camera. Then, 22 W energy efficient light tubes were installed on the side walls to provide uniform illumination. Air conditioning was used to maintain a constant interior temperature of 30° ± 2 °C. Each funnel was connected to the air purification system through a tube (Tygon, [ID, 0.7 cm]). The behavioral setup offered two choices for the test locusts: a control zone with clean vertical airflow and a test zone with adjacent vertical airflow containing synthetic chemicals. The locusts were introduced through the small door at the bottom center of the acrylic glass frame and allowed to remain for 10 min under different odors in mineral oil (4VA, PAN, and the mixture) for behavioral testing. The diluted odorant was applied onto a piece of filter paper (3 cm × 3 cm; Whatman No. 1), and solvent mineral oil was used as control in other funnel. After testing with 10 to 15 locusts, the positions of the two funnels were swapped to prevent any positional bias. At least 30 individuals were used for each treatment dosage or ratio to complete the dual-choice test. After completing a test with a group of insects, the funnels were heated to 180 °C for at least 2 h to eliminate any residual odor.

We captured the behavioral activities of locusts over a period of 10 min by using a camera (Panasonic, Osaka, Japan) connected to VCR software (Version 2, Noldus Information Technology) at a rate of 25 frames per second (frame/second). By utilizing EthoVision XT software (11.5 version, Noldus Information Technology) to analyze the video, we objectively computed the total movement time (unit:seconds) and distance (unit: centimeters) of the locusts on each side.

EAGs and SSRs.

The antennae of locusts were cut off at the base of the scape and fixed onto two electrodes with conductive gel Spectra 360 (Parker, Orange, USA). The electrical signal was transmitted to the computer via an amplifier (IDAC4, Syntech, Netherlands). The neural responses were recorded using Syntech EAG software v2.6c. The stimulation controller (Syntech CS-05) was capable of producing a steady airflow of 30 milliliters per second. The duration of stimulation was 1 s. Blank stimuli (clean air) were applied at both the beginning and end of the stimulation series. We subtracted the average blank EAG amplitude from the stimulus EAG amplitude.

The recording and analysis of SSRs have been described in the previous study (2). The locust was placed inside a 1 cm diameter, thick-walled plastic tube, with its head protruding outside of the tube, and its antennae fixed in place with dental wax. A tungsten wire electrode was sharpened electrolytically using a 10% solution of NaNO₂. Using a micromanipulator (Narishige, Japan), the recording electrode was inserted into the base of the sensilla. The reference electrode was inserted into the eye. The recording electrodes were connected to the amplifier (IDAC4, Syntech, Netherlands). Automatic frequency meter software was used to calculate the frequency variation of each pulse within 0.2 s.

Mineral oil was used as both EAG and SSR blanks, as well as for diluting all compounds. Then, 10 μL of volatile solution was added onto a filter paper (Whatman, UK) placed in a 15 cm Pasteur glass tube.

RNA Preparation and qPCR Assay for Genes.

TRIzol reagent (Life Technologies, Carlsbad, USA) was used to homogenize and extract total RNA from eight antennal samples of locusts (3 to 5 individuals per sample), following the manufacturer’s instructions. DNase was used to remove any DNA contamination from the RNA samples. One μg of total RNA was reverse transcribed in every sample using MMLV reverse transcriptase (Promega, Madison, USA) according to the manufacturer’s protocol to analyze gene expression levels, respectively. A Real Master-Mix (SYBR Green) kit (Tiangen, Beijing, China) was used for PCR amplification. The endogenous control for mRNA was rp49. The amplification program followed the manufacturer’s instructions of Kits, and the melting curve was examined to verify the amplification specificity of the target gene. All PCR amplifications were sequenced to confirm the primer specificity. The qPCR primers used are listed in SI Appendix, Table S1.

Intracellular Recording and Staining.

The locust head was isolated and immobilized, and labial palps and mouthparts were removed. Recordings were always performed on the hemisphere ipsilateral to the antenna to be stimulated. The membrane potentials were recorded with Axoclamp amplifier 900A (Molecular Devices) with a sampling rate at 30 kHz. Neurons responding to 4VA, PAN, or both were stained with 1 mg/mL Alexa Fluor 647 (Life Technologies, Carlsbad, USA) and scanned with a laser confocal microscope (Zeiss, Oberkochen, Germany).

Multichannel Recording of AL.

Three- to four-day-old fifth-instar nymphs were used for electrophysiological recordings. During multichannel recording, the whole head of the locust was cut off and fixed with two needles. All labial palps and mouthparts were removed, and the upper epidermis and fat body were carefully removed to expose the brain. The AL was exposed at the top of the preparation, and the surface film was manually removed with fine tweezers to promote the penetration of the microelectrode. The ipsilateral cerebral hemisphere of the stimulated antenna was penetrated. The exposed brain was continuously perfused with locust saline (147 mM NaCl, 10 mM KCl, 4 mM CaCl₂, 3 mM NaOH, 10 mM (4-(2-hydroxyethyl)-1-piperazineethanesulfonic acid), pH 7.2) to prevent drying.

PNs activity were recorded using a silicon tetrode (NeuroNexus, USA, A4 × 4 to 3 mm-50-125-177). In each experiment, silicon probes were placed in different positions of AL to maximize activity. A custom algorithm was used to sort the spikes from tetrode signals (22). A constant stream of humidified air, filtered with activated carbon, was blown at the antennae at a speed of 0.5 m per second. The airflow was transmitted through a Pasteur glass tube (8 mm in diameter) mounted 10 mm in front of the antenna. A tube containing mineral oil was used as control. The stimuli appeared in the order of low concentration to high concentration. The stimulant glass tube was inserted into the glass tube with constant air flow. A series of puffs of the stimuli tube air was injected into the airflow through the odor stimulus controller (3-Way Solenoid Valves, Parker) at the flow rate of 300 mL/min. Pulse stimulation was 3.3 Hz, each pulse lasted 200 ms and repeated 10 times.

For quantifying neural responses, the tetrode waveforms were transferred from the Tucker-Davis Technologies acquisition software to the Offline Sorter V4 program (Plexon® Inc., Dallas, TX, USA). This program facilitates both automatic and manual sorting of the waveforms. Within Offline Sorter, each set of tetrode waveforms, comprising 32 (A/D points) × 4 (recording sites), thus totaling 128 dimensions, underwent dimensionality reduction to three dimensions via PCA. Subsequently, each combined waveform was projected onto a three-dimensional space formed by the three principal components for visualization. The accuracy of spike sorting was statistically validated using Offline Sorter. Subsequently, the timestamps of all waveforms were exported to either NeuroExplorer5 (Nex Technologies, Dallas, TX, USA) or Matlab (Mathworks, Sherborn, MA, USA) for additional analysis.

Statistical Analysis.

The fitting curves for calculating the threshold of PAN and 4VA production and the peak time of different odors were conducted using Software Origin 2018 (OriginLab). The logistic fit was chosen for the fit curve based on related index. The emission of PAN and 4VA after crowding in solitary locusts, as well as the EAG and SSR response data, were analyzed by two-tailed unpaired t test. The data of olfactory preference were analyzed using the Wilcoxon signed rank test (mean ± SEM). Differences were considered significant at P < 0.05. All statistics were analyzed using SPSS 18.0 (SPSS Inc., Chicago, IL, USA). Segmented the spike trains of PNs for each odor into nonoverlapping time bins (5 ms) of 3 s. Linear PCA was applied to reduce the dimensionality of the data from the first 1 s of reaction. Then, 10 neurons were connected from low-dimensional points to visualize the neural response trajectories to different stimuli. Based on the PCA results, we calculated the centroids of the low-dimensional representations for different odors, visualized them in the same three-dimensional space, and computed the Euclidean distances. All analyses of PNs were performed using Matlab (Mathworks, Sherborn, MA, USA). All the statistical results are listed in SI Appendix, Table S2.

Supplementary Material

Appendix 01 (PDF)

Data, Materials, and Software Availability

All study data are included in the article and/or SI Appendix.