Neural responses evoked by a stimulus reduce upon repetition. While this adaptation allows the sensory system to attend to novel cues, does information about the recurring stimulus particularly its intensity get compromised? We explored this issue in the locust olfactory system. We found that locusts’ innate behavioral response to odorants varied with repetition and stimulus intensity. Counter-intuitively, the stimulus-intensity dependent differences became significant only after adaptation had set in. Adaptation also altered responses of individual neurons in the antennal lobe (neural network downstream to insect antenna). These response variations to repetitions of the same stimulus were unpredictable and inconsistent across intensities. Although both adaptation and intensity decrements resulted in an overall reduction in spiking activities across neurons, these changes could be disentangled and information about stimulus intensity robustly maintained by ensemble neural responses. In sum, these results show how information about odor intensity can be preserved in an adaptation-invariant manner.
This study addresses an important question in sensory neuroscience: how the olfactory system distinguishes decreases in stimulus intensity from decreases in neural responses due to adaptation. Based on a combination of electrophysiological and behavioral analyses, solid evidence establishes that neural coding changes differently between intensity reductions and adaptation. Intriguingly, behavioral responses tend to increase as the neural responses decrease, suggesting that core features of the odor response persist through adaptation. While the experimental results are convincing overall, the conclusions would be strengthened by more refined statistical analysis and data quantification.
The ability to adapt to sensory cues is key to the survival of many living organisms1–8. The goal of this sensory adaptation is to condense neural responses to a persisting or recurring stimulus in order to focus on novel cues. Such condensed neural responses are thought to increase system efficiency by reducing the metabolic requirements of information encoding9–12. However, is sensory information lost when neural responses are condensed during adaptation? While this computational task appears relatively elementary, any solution should satisfy two important requirements. First, attenuation of stimulus-evoked responses due to adaptation should not alter the identity of the stimulus. Second, information regarding the intensity of the stimulus may have to be preserved when odor intensity conveys information of behavioral importance. The latter requirement is paramount when sensory stimuli have different behavioral significance at varying intensities. For example, some odorants are attractive at lower intensities but switch to being aversive beyond certain threshold values13–15. Alternately, sensitivity to intensity gradients has to be maintained for sustained periods to allow for navigation towards or away from the stimulus source16–18. Such biological contexts necessitate the preservation of odor intensity information across repeated encounters with the odorant. However, the mechanism by which information about stimulus intensity is robustly maintained is not fully understood.
Sensory adaptation occurs over both long- and short-time scales12,19–21. Long-term adaptation requires gene transcription and protein translation that occurs on the time scale of hours to days6,22–24, whereas short-term adaptation occurs on a millisecond to minute time scale2,11,25. Short-term adaptation results from a wide range of mechanisms with prominent models in the olfactory system being vesicle depletion and facilitation of lateral inhibition25–32. While both vesicle depletion and lateral inhibition are likely to occur in the olfactory system, these complementary mechanisms will manifest in two different,but testable, outcomes. Vesicle depletion will impact the most prominent neurons with elevated odorevoked responses, whereas facilitation of lateral inhibition would progressively suppress activity in the relatively weak responders. Irrespective of which mechanisms underlie olfactory adaptation, the neural responses evoked by a stimulus will change over time and across repeated encounters. This raises the question, does adaptation then corrupt or alter information regarding the odorant?
The ambiguity that neural adaptation introduces has been documented in visual33 and auditory systems34,35. However, prior studies have shown that odor identity is robustly maintained by the spatiotemporally patterned neural responses and can allow precise recognition across repeated encounters with the odorant30,36–38. But is intensity information corrupted by the changes in neural activity due to sensory adaptation? Both reductions in stimulus intensity and sensory adaptation lead to a reduction in the overall spiking responses in most neural circuits. Do these neural response variations due to neural adaptation and intensity decrements be disentangled? We sought to examine this in the relatively simple insect olfactory system.
Our results show that odor intensity changes result in subtle variations in the overall combination of activated neurons. While sensory adaptation results in an overall reduction in the population neural responses, the activated ensemble is robustly maintained to allow simultaneous and precise decoding of odor identity and intensity. Using an innate behavioral assay, we show that physiological adaptation coincided with enhanced behavioral differentiation between odorant intensities. Thus, our work reveals a simple but elegant approach to achieve adaptation-invariant intensity coding and recognition in the insect olfactory system.
Behavioral responses to an odorant vary with repetition and intensity
We began by examining how behavioral responses change with repetitions of an odorant. We used an innate behavioral assay where the presentation of certain odorants triggered an appetitive palp-opening response (POR) in starved locusts (Fig. 1a; Supp. Fig. 1). The probability of an odor eliciting a POR (P(POR)) was regarded as an indicator for stimulus perception (Fig. 1b). Note that the P(POR) was computed by averaging the behavioral responses across locusts for each trial or repetition number (Fig. 1c, d). Intriguingly, we found that the P(POR) systematically increased with repetitions of the same odorant and stabilized after about five stimulus presentations.
Previous results have shown that the POR is a function of both stimulus identity and intensity39–41. Therefore, we presented each locust with randomized blocks of the same odorant at either high (H, 1% v/v) or low (L, 0.1% v/v) intensities (Fig. 1e). Our results indicate that the P(POR) evoked by high and low stimulus intensities of an odorant were not significantly different in the earlier trials (Fig. 1f; p > 0.05 for hex, iaa, and bza; one-tailed t-test). However, the P(POR) became significantly different for the high and low-intensity exposures of the same odorant during the later trials (p ≤ 0.05 for hex, oct, and iaa; one-tailed t-test). In other words, high and low stimulus intensities were more behaviorally discriminable in the later trials. Hence, these results counterintuitively suggest that behavioral responses to an odorant at two different intensities were different after multiple encounters with the same stimulus.
Odor-evoked response adaptation leads to complex changes in individual neurons
How do neural responses change as a function of stimulus repetition? To understand this, we examined the odor-evoked responses of individual projection neurons (PNs) in the locust antennal lobe as different odorants at two different intensities were repeatedly presented (Fig. 2a). Odor-evoked responses in individual PNs varied subtly across trials (Fig. 2b). In general, most PN responses to the odorant were stronger in the first trial and reduced during later trials of the same stimulus. Notably, the changes in PN responses in two different blocks of trials, where the same odorant was presented, showed repeatable changes in odor-evoked spiking patterns and the overall response strength (Fig. 2c).
Next, we examined how individual PNs adapted during stimulus repetitions. Based on existing literature, two widely considered models of adaptation in the olfactory system are vesicular depletion and facilitation of lateral inhibition. While the vesicular depletion model predicts that the strongest firing neurons would adapt the most, in the lateral inhibition facilitation model the weakly activated neurons would progressively be suppressed more (Fig. 2d). To test these two opposing but complementary theories, we calculated the percentage change in the total spike counts observed during first and the 25th odor presentations (Fig. 2e, y-axis; positive values indicate weaker response in the later trial and hence response reduction, whereas negative values indicate facilitation of odor-evoked responses). We plotted these normalized response changes against the total spike counts observed in the very first trial (Fig. 2e, x-axis). The dotted lines represent the best linear model that fit the two datasets we recorded. If vesicular depletion was the predominant form of adaptation, the slope of the regression line would be positive. Whereas a negative slope would indicate that lateral inhibition was the predominant form of adaptation. However, as can be noted, across both datasets with two concentrations of the four odorants tested, neither model captured the adaptation dynamics in our datasets (R2 = 0.05, 0.01, respectively).
Are trial-to-trial changes in stimulus-evoked neural responses at two different concentrations of the same odorant correlated? If this is the case, then it would be reasonable to expect neurons that adapt more at the higher concentration to also change in a similar fashion at a lower concentration and vice versa. To assess this, we calculated and plotted the normalized change in spike counts for each PN at both high and low concentrations of the same odorant (Fig. 2f). Notably, the PNs with the greatest response change during repetitions of the higher-intensity odor exposures were not the ones that adapted heavily when encountering the same stimulus at a lower intensity (R2 = 0.034 (hex), 0.032 (oct), 0.067 (iaa), 0.0016 (bza)). In sum, these results indicate that change due to adaptation in individual neurons is not a simple function of its response strength. Further, how individual neurons adapted is a function of both odor identity and intensity.
Adaptation and intensity decrements both reduce ensemble neural response strength
What is the overall change in the spiking response across the neural ensemble? To examine this, we calculated the average spiking response across all PNs recorded and plotted the spike counts as a function of time for each trial (Fig. 3a). Expectedly, the overall response strength was stronger in the first trial and reduced when the same stimulus was repeatedly encountered. A similar reduction in overall response strengths was also observed when the odorant intensity was reduced (Fig. 3b)
Are the response reductions during stimulus repetition comparable to those observed with stimulus intensity decrements? To determine this, we plotted the total spike counts summed across all PNs over the entire stimulus presentation duration and plotted it as a function of the trial number (Fig. 3c). We compared the changes over trials with the total spike counts observed during the high (H) and low (L) intensities of the same odorant. Notably, the total PN ensemble spike counts decreased due to both stimulus intensity reduction and repetition of the stimulus. Do these results indicate that adaptation could potentially confound information regarding stimulus intensity?
Adaptation-invariant encoding of odor intensity
Could the information regarding stimulus intensity be robustly encoded in the population neural responses? As noted earlier, both spiking activities in individual neurons and the volume of spikes at the ensemble level vary across trials. If adaptation alters the overall spiking activity amplitude (i.e. vector length; Fig. 4a), while changes in stimulus intensity alter the combination of activated neurons (i.e. vector direction; Fig. 4b), then information about odor identity and intensity can be preserved in an adaptationinvariant fashion. To test whether the recorded ensemble of neurons in our datasets encoded information in such an adaption-invariant fashion, we first visualized the high-dimensional neural activities in each trial using a dimensionality reduction approach (see Methods). We plotted the ensemble responses in each 50 ms time bin during the odor presentation time window (4 s), and linked them based on the order of their occurrence, to generate trial-by-trial odor response trajectories (Fig. 4c, d). Note that each trial generated a single loop response trajectory after dimensionality reduction. Further, neural responses in different trials evoked closed-loop trajectories that systematically reduced in length with increasing repetition but maintained their direction indicating that the combination of neurons activated were mostly conserved. Different odorants evoked responses that were markedly different in the combination of activated neurons and therefore their response trajectories were oriented in different directions. Notably, different concentrations of the same odorant also generated response trajectories that subtly varied in direction. As a direct consequence, even though the overall spiking activities reduced over trials, the odor response trajectories reliably maintained their direction in all the trials. Hence, information about odor identity and intensity was robustly maintained across all trials.
To quantify these observations, we performed a correlation analysis using high-dimensional PN response vectors. If a similar combination of PNs (high-dimensional activity vectors) were activated in two different trials, then their correlation will be high. Correlations across different trials and concentrations of the same odorant are shown in Fig. 5a. Confirming results from the dimensionality reduction analysis (Fig. 4), we found odor-evoked responses across different trials of an odorant at a particular intensity were highly correlated (diagonal blocks on each panel). The similarity between odorevoked responses evoked by different concentrations of the same odorant was considerably less correlated (off-diagonal blocks). A hierarchical clustering analysis revealed that the odor-evoked responses nicely clustered based on odor identity and then by intensity (Fig. 5b)
In sum, our results reveal that a combinatorial code could encode information regarding odor identity and intensity in an adaptation-invariant fashion. Since the same information about a stimulus is represented with fewer spikes in the later trials, we conclude that adaptation refines the odor codes by making them more efficient.
Neural response suppression vs. behavioral output facilitation
Neural responses evoked by an odorant reduced with repetition (Fig. 3). However, as we noted earlier, the behavioral responses increased upon stimulus repetition (Fig. 1). This negative correlation is evident when the neural responses (total spiking activities evoked across neurons) was plotted against the behavioral POR response probability during a particular trial (Fig. 6). This negative correlation between neural suppression and behavioral enhancement was observed for all odorants and both concentrations examined. It is worth noting that the last trial of the high-intensity exposure evoked neural responses that were comparable in strength to the first trial of the low-intensity exposure of the same stimulus. Yet, the behavioral responses observed during these two trials were markedly different. Hence, our results also indicate that the probability of the POR response is not simply a function of the total number of spikes elicited.
Neural and behavioral changes due to adaptation are odor specific
Finally, we wondered if the neural response reduction and behavioral enhancements were a global, non-specific state change in the olfactory system brought about by the repetition of any odorant, or are the observed neural and behavioral response changes odor specific. To examine this, we used a ‘catch-trial paradigm.’ (Fig. 7a) In these sets of experiments, we repeatedly presented an odorant to induce adaptation. After a substantial reduction in the neural response or stabilization of the behavioral response, a deviant or catch stimulus was presented. Persistence of the observed reduction in neural responses and enhancement in behavioral responses when the catch stimulus is introduced would reveal that the adaptation-induced changes are not odor specific.
We used the catch-trial experiments with two different catch stimuli. Our neural recordings reveal that when iaa (ester) was used as the catch stimulus after repeated presentations of hex (alcohol), the neural responses strength recovered and increased to an overall higher-spike count (Fig. 7b). The observed catchiaa response strength matched the unadapted response level to iaa when it was separately presented following a 15-minute reset window (i.e. Fig. 7b, trial# 31). In contrast, when apple (a complex mixture) was used as the catch stimulus, the neural responses did not recover to higher levels indicating significant cross-adaptation (Fig. 7b). Nevertheless, in both cases, the ensemble responses evoked by the catch stimulus were highly similar to the unadapted responses it usually generates (Fig. 7c). Therefore, these results indicate that there is stimulus specificity to the adaptation-induced changes in neural responses and that cross-adaptation does not corrupt the identity of the deviant stimulus.
Behaviorally, we chose odor pairs where the component odorants had markedly different levels of POR responses (Fig. 7d). Repeated presentations of an odorant (bza or iaa) increased the P(POR) across the locusts as observed before (Fig. 7e). Notably when the catch stimulus was presented, the behavioral 1responses dropped close to the first-trial P(POR) elicited by that odorant (Fig. 7f; dotted line). These findings were also observed when the recurring and catch odorants were switched (Fig 7f).
In sum, these neural and behavioral results indicate that the observed neural and behavioral response changes brought about by the repeated encounters with an odorant are at least partially odorspecific and are not global non-specific changes in the olfactory system.
Sensory systems often have to deal with constraints that are of an opposing nature. For example, the ability to maintain stable representations is key to robust recognition of sensory stimuli42–44. Often variations in the external environment, such as humidity and temperature, may introduce changes in neural responses that have to be filtered out to preserve the identity of the stimuli. Additionally, the strength or intensity of the stimulus can also change widely. Compressing or even removing these variations might be necessary to recognize an odorant independent of its intensity42–47. Alternately, in certain cases, it might also become important to maintain some of these variations so they can be exploited for guiding or altering the behavioral response. As an example, it might be worth noting that many chemicals that are pleasant at lower intensities can become repelling or even harmful to the organism at extremely high intensities13–15. Hence, the behavioral responses should accordingly vary with intensity. Further, many organisms have shown that they can follow an odorant to their source (source localization). Such behaviors are of ecological importance as they allow the organisms to find the food source or potential mates. In all these cases, it becomes important to maintain variations with stimulus intensity to ensure the organisms are able to approach the odor source. In our earlier studies, we have focused on examining how robust odor recognition is achieved37,40,48. Here we examined the latter problem of how to maintain information about the stimulus intensity while the responses in the neural circuits are becoming weaker due to adaptation.
Consistent with earlier studies30,49,50, our data revealed that the response of individual projection neurons in the antennal lobe varied with the repetition of the same stimulus. While a reduction in odorevoked responses was observed in many neurons, an increase in spiking activity and changes in temporal firing patterns was also observed. Notably, these changes were not random but highly repeatable (Fig. 2c). Hence, these results indicate that the effect of adaptation in a neural circuit is highly consistent and reliable.
In addition to spike rate reduction, oscillatory synchronization, and inter-neuronal coherence build up in the locust antennal lobe over repeated encounters with an odorant30,51,52. This short-term memory was not observed in the antenna but endured in the antennal lobe for several minutes after the termination of the stimulus. Similar results were also reported in the mouse olfactory bulb43. The inputs from sensory neurons were shown to be consistent across repetition whereas the outputs of the olfactory bulb were systematically reduced with stimulus repetition. These results suggest that loci of short-term adaptation, at least when the stimulus is discontinuous and recurring, is in the antennal lobe/olfactory bulb circuitry. Although the precise mechanism for achieving the same is not fully understood49,53.
What is the computational significance of this neural adaptation? Sensory adaptation has been implicated in several important computations such as high-pass filtering, matching neural response to stimulus statistics, generating sparser codes, and optimal representations33,35,54–56. In the olfactory system, adaptation has also been suggested to stabilize neural representation such that recognition of odor identity from ensemble neural responses improved after repeated encounters with the stimulus30. However, adaptation also introduces a potential confound whereby the absolute neural response strength becomes an unreliable indicator of stimulus strength. Our results extend these earlier studies by revealing how this potential confound could be resolved in the olfactory system.
Increases in stimulus intensity have also been shown to result in higher response amplitudes in firing neurons and in the recruitment of additional activated neurons35,42,44,57. While our results show that this is indeed true, we also found that several neurons responded preferentially during the lower concentration exposures but did not respond when the higher intensity of the same stimulus was presented (Supp. Fig. 2). Hence, the activated combination of neurons differed both with stimulus identity and intensity. Furthermore, even though adaptation reduced the overall spiking activities across neurons, the combination of neurons activated was robustly maintained. In other words, the population response vector direction (determined by the combination of neurons activated) was consistent. Only the vector length (determined by the total spikes across neurons) is reduced with the repetition of the stimulus. This simple approach was sufficient to encode the information about the odor intensity in an adaptation-invariant fashion. Additionally, since equivalent information about stimulus identity and intensity was preserved with fewer spikes, neural adaptation can be considered as a method of optimizing the neural representation2,3,9,10,33,34.
Does the observed persistence of concentration information in the neural activities after adaptation translate to observable differences in behavioral outcomes? Using an innate palp-opening response (POR) we examined how behavioral responses varied with stimulus repetition. Unlike the neural responses, we found that the PORs increased with repetition and after about five trials settled into odor-specific levels of responses. Notably, the PORs in the later trials were significantly higher for higher stimulus intensities for most of the odorants used. In sum, our results showed that locusts are better able to behaviorally differentiate stimulus intensity after adaptation.
Alternately, adaptation due to the repetition of an odorant could alter the perceived intensity of the recurring odorant or other related test odorants. The compromise of stimulus intensity information due to adaptation has been observed in behavioral discrimination tasks in rats and even in psychophysical studies in humans58,59. Such a result can arise when the combinatorial response patterns evoked by the adapting and the test stimulus have significant overlap. However, our results indicate that when changes in odor intensities alter the combination of neurons activated, it is feasible to maintain information about odor intensity levels in a robust fashion.
Finally, we note that the anticorrelated link between neural suppression and behavioral enhancements (called ‘repetition priming’) is well established in other animal models60. How can weaker neural responses lead to enhanced behavioral outcomes? Our results indicate that the antennal lobe neural responses in later trials were not sharpened by pruning out weaker responses (Supp. Fig. 2). The weakening of the response strength could potentially lead to a sparser response in the downstream mushroom body. In addition, as we noted earlier, neural response synchrony increases with stimulus repetition. Whether weaker but synchronized activity in neurons drives sparser and more selective responses which then produce enhanced behavioral outcomes remains to be carefully examined.
We thank members of the Raman Lab (Washington University in St. Louis) for feedback on the manuscript. We thank Pearl Olsen for insect care. This research was supported by NSF (1707221, 1724218, 2021795) and ONR (N00014-19-1-2049, N00014-21-1-2343) grants to B.R. and an Imaging Sciences Fellowship to D.L.
For delivering odorants, we followed a protocol described in our earlier work. Briefly, odorants were diluted in mineral oil to either 1% or 0.1% concentration (v/v) and sealed in glass bottles (60 ml) with an air inlet and outlet. A pneumatic picopump (WPI Inc., PV-820) was used to displace a constant volume (0.1 L/min) of the static headspace above the diluted odor-mineral oil mixture into a desiccated carrier air stream (0.75 L/min) directed toward one of the locust’s antennae. A vacuum funnel placed behind the locust preparation continuously removed the delivered odors.
The first set of experiments included multiple blocks, with twenty-five trials each when one odorant at one intensity was repeatedly presented. Each trial in the block included a four-second stimulus presentation window. The inter-stimulus interval (between trials) was 60 seconds. Different odorants at different intensities were repeatedly presented in different blocks. A 15-minute window, when no stimulus was presented, separated two consecutive blocks of trials. This window was included to reset any short-term memory that may have formed due to repeated presentation of the same stimulus30.
The second set of experiments involved two blocks of trials. The first block of trials included 30 trials and the second block consisted of ten trials. In the first 25 trials of block 1, hexanol at 1% v/v was repeatedly presented. In the 26th trial, a puff of either isoamyl acetate 1% v/v or apple 1% v/v (a ‘deviant’ odorant) was presented. In trials 27-30, hexanol at 1% v/v was again presented. This was followed by a 15-minute reset window when no stimulus was presented. In the second block of ten trials (trials 31-40), the deviant stimulus was repeatedly presented. All inter-stimulus intervals (between trials) were 60 seconds and all odor presentations were four seconds long.
Post-fifth instar adult locusts (Schistocerca americana) were reared in a crowded colony with a 12-hour - 12 hour light-dark cycle. both males and females were used for electrophysiological experiments. First, the locusts were immobilized with both antennae intact. Then the primary olfactory region of their brain, the antennal lobes, were exposed, desheathed, and perfused with room-temperature locust saline. Extracellular multiunit recordings of projection neurons (PNs) were performed with a 16-channel, 4x4 silicon probe (NeuroNexus) that was superficially inserted in the antennal lobe (AL). Prior to each experiment, all probes were electroplated with gold to achieve impedances in the range of 200 to 300 kΩ. The recordings were acquired with a custom 16-channel amplifier (Biology Electronics Ship; Caltech, Pasadena, CA). The signals were amplified with a 10k gain, bandpass filtered (0.3 to 6 kHz), and sampled at 15 kHz using a LabView data acquisition system. A visual demonstration of this protocol is available online61.
PN spike sorting
To obtain single-unit PN responses, spike sorting was performed offline using the four best recording channels and conservative statistical principles62. Spikes belonging to single PNs were identified as described in earlier work37,63. The following criteria were used to identify single units: cluster separation >5x noise standard deviations, number of cluster spikes within 20ms < 6.5% of total spikes, and spike waveform variance < 6.5x noise standard deviations. In total, 161 PNs from 40 locusts were identified. Two datasets were collected. In the first dataset responses of 80 PNs to hex and iaa were recorded, and in the second dataset responses of 81 PNs to iaa and bza were monitored.
The PN spikes were binned in 50 ms non-overlapping time bins, and spike counts of different PNs were concatenated to obtain a population spike count vector. Pearson correlation coefficients between two PN ensemble spike count vectors were calculated using Equation 1.
Here, Xtrial i and Xtrial j are time-averaged high-dimensional activity in trials i and j, respectively. σtrial i and σtrial j are the standard deviations of X rial i and Xrial j, respectively.
Each pixel/matrix-element in the correlation plot shown in Fig. 5a indicates the similarity between PN spike count vectors observed in the ith and jth trials.
Tensor-based data decomposition
We first organized neural response data as a three-way array (Neuron × Time × Trials; the stimulus information was also blended into trial dimension) and then employed a direct 3-way tensor decomposition approach64,65. Here, the 3-d data cube was approximated using three loading matrices, A, B, and C with elements ai f(neuron dimension), bjf (time dimension), and ckf (trial dimension). ei j k was the residual element (see the equation below). The tri-linear model was found using alternating least squares.
where i, j, k denote the three different dimensions, and F indicates the total number of factors used for the analysis that was determined by the core consistency diagnostics64,65. In our case, when F = 3, the core consistency was above 50 %, while it dropped to below 40% when F = 4. Therefore, we used three factors for our data decomposition.
Trial-to-trial odor trajectory
For this analysis, we first reconstructed the dataset by computing the outer product of the loading matrices that were obtained by the tensor decomposition. The reconstructed 3-d tensor was then unfolded into a concatenated matrix (i.e. along the trial dimension). After unfolding, the ensemble projection neuron responses were arranged as time series data of n dimensions (where n is the number of neurons) and m steps (the number of 50 ms time bins × the number of trials). Note that only the projection neuron activities during the four-second stimulus presentation window in each trial were used for this analysis. The ensemble projection neuron response vectors (in a given 50 ms time bin) were projected onto the three eigenvectors of the response covariance matrix that accounted for the most variance in the dataset, using principal component analysis. Finally, the low-dimensional points were connected in a temporal order to visualize neural response trajectories to different stimuli on a trial-to-trial basis. All trajectory plots shown in Fig. 4 were generated after smoothing with a 3-point running average low-pass filter.
Neural response similarity and dendrogram analysis
First, we calculated the summed spike counts during the 4 s odor presentation window for each projection neuron. Then the correlation similarity between two spike count profiles across projection neurons was calculated using Equation 1. Similarly, in this analysis, xj and x j represent a n × 1 vector (n = 80 for hex-2oct; n = 81 for iaa-bza) for trial i and j, respectively. The dendrogram was generated by hierarchical clustering of all stimulus identities, intensities, and individual trials based on the correlation distance. The dendrogram was created in such a way that the furthest pairwise distance between any two samples assigned to an individual cluster was minimized.
Experiments were performed on post-fifth instar locusts of either sex that were starved for approximately 24 hours. All behavioral experiments occurred between 10 am and 2 pm. The protocols for the innate behavioral preference experiments and palp tracking used in this study were published in previous studies37,41,48,63. The odor delivery setup and stimulus sequences for the single odorants were similar to the ones described for the electrophysiological experiments (Figs. 1a, 2a). Any palp movement that occurred within 15 seconds of the odor stimulus onset was considered a palp-opening response (POR).
- 1.Neuronal Mechanisms of Habituation and Dishabituation of the Gill-Withdrawal Reflex in AplysiaScience 167:1745–1748https://doi.org/10.1126/science.167.3926.1745
- 2.Fast intensity adaptation enhances the encoding of sound in DrosophilaNat Commun 9https://doi.org/10.1038/s41467-017-02453-9
- 3.Olfactory Cortical Adaptation Facilitates Detection of Odors Against BackgroundJ Neurophysiol 95:1888–1896https://doi.org/10.1152/jn.00812.2005
- 4.Adaptation processes in insect olfactory receptors. Mechanisms and behavioral significanceAnn N Y Acad Sci 510:104–112https://doi.org/10.1111/j.1749-6632.1987.tb43475.x
- 5.Habituation Is More Than Learning to Ignore: Multiple Mechanisms Serve to Facilitate Shifts in Behavioral StrategyBioEssays 41https://doi.org/10.1002/bies.201900077
- 6.Sensory adaptation mutants of E. coliCell 15:1221–1230https://doi.org/10.1016/0092-8674(78)90048-X
- 7.Neural Adaptation in the Generation of Rhythmic BehaviorAnnu Rev Physiol 62:723–753https://doi.org/10.1146/annurev.physiol.62.1.723
- 8.Positive interaction between ASH and ASK sensory neurons accelerates nociception and inhibits behavioral adaptationiScience 25https://doi.org/10.1016/j.isci.2022.105287
- 9.Adaptive Rescaling Maximizes Information TransmissionNeuron 26:695–702https://doi.org/10.1016/S0896-6273(00)81205-2
- 10.A Simple Coding Procedure Enhances a Neuron’s Information CapacityZ Für Naturforschung C 36:910–912https://doi.org/10.1515/znc-1981-9-1040
- 11.Adaptation of retinal processing to image contrast and spatial scaleNature 386:69–73https://doi.org/10.1038/386069a0
- 12.Coding Principles in AdaptationAnnu Rev Vis Sci 5:427–449https://doi.org/10.1146/annurev-vision-091718-014818
- 13.Decoding of Context-Dependent Olfactory Behavior in DrosophilaNeuron 91:155–167https://doi.org/10.1016/j.neuron.2016.05.022
- 14.Odour concentration-dependent olfactory preference change in C. elegansNat Commun 3https://doi.org/10.1038/ncomms1750
- 15.Rong H, Das P, Lube A, et al. Early warning signals regarding environmental suitability in the Drosophila antenna. Published online June 7, 2017:146522. doi:10.1101/146522https://doi.org/10.1101/146522
- 16.Algorithms for Olfactory Search across SpeciesJ Neurosci 38:9383–9389https://doi.org/10.1523/JNEUROSCI.1668-18.2018
- 17.Elementary sensory-motor transformations underlying olfactory navigation in walking fruit-flieseLife 7https://doi.org/10.7554/eLife.37815
- 18.Smelling on the fly: sensory cues and strategies for olfactory navigation in DrosophilaCurr Opin Neurobiol 22:216–222https://doi.org/10.1016/j.conb.2011.12.010
- 19.Rapid Sensory Adaptation Redux: A Circuit PerspectiveNeuron 92:298–315https://doi.org/10.1016/j.neuron.2016.09.046
- 20.Early Olfactory Processing in Drosophila: Mechanisms and PrinciplesAnnu Rev Neurosci 36:217–241https://doi.org/10.1146/annurev-neuro-062111-150533
- 21.Distinct mechanism for long-term contrast adaptationProc Natl Acad Sci 109:5898–5903https://doi.org/10.1073/pnas.1113503109
- 22.Odorant-specific adaptation pathways generate olfactory plasticity in C. elegansNeuron 14:803–812https://doi.org/10.1016/0896-6273(95)90224-4
- 23.Mechanism of odorant adaptation in the olfactory receptor cellNature 385:725–729https://doi.org/10.1038/385725a0
- 24.Ca2+–Calmodulin Feedback Mediates Sensory Adaptation and Inhibits Pheromone-Sensitive Ion Channels in the Vomeronasal OrganJ Neurosci 29:2125–2135https://doi.org/10.1523/JNEUROSCI.5416-08.2009
- 25.Slow presynaptic mechanisms that mediate adaptation in the olfactory pathway of DrosophilaeLife 8https://doi.org/10.7554/eLife.43735
- 26.Asynchronous release of synaptic vesicles determines the time course of the AMPA receptor-mediated EPSCNeuron 15:1097–1107https://doi.org/10.1016/0896-6273(95)90098-5
- 27.Sensory Neuron Signaling to the Brain: Properties of Transmitter Release from Olfactory Nerve TerminalsJ Neurosci 24:3023–3030https://doi.org/10.1523/JNEUROSCI.5745-03.2004
- 28.A Quantitative Description of Short-Term Plasticity at Excitatory Synapses in Layer 2/3 of Rat Primary Visual CortexJ Neurosci 17:7926–7940https://doi.org/10.1523/JNEUROSCI.17-20-07926.1997
- 29.Crayfish escape behavior and central synapses. II. Physiological mechanisms underlying behavioral habituationJ Neurophysiol 35:621–637https://doi.org/10.1152/jn.1922.214.171.1241
- 30.Short-term memory in olfactory network dynamicsNature 402:664–668https://doi.org/10.1038/45244
- 31.Odor response adaptation in Drosophila—a continuous individualization processCell Tissue Res 383https://doi.org/10.1007/s00441-020-03384-6
- 32.The Cellular and Molecular Basis of Odor AdaptationChem Senses 25:473–481https://doi.org/10.1093/chemse/25.4.473
- 33.Efficiency and ambiguity in an adaptive neural codeNature 412:787–792https://doi.org/10.1038/35090500
- 34.A Neural Mechanism for Time-Window Separation Resolves Ambiguity of Adaptive CodingPLOS Biol 13https://doi.org/10.1371/journal.pbio.1002096
- 35.Neural adaptationCurr Biol 31:R110–R116https://doi.org/10.1016/j.cub.2020.11.054
- 36.Intensity versus Identity Coding in an Olfactory SystemNeuron 39:991–1004https://doi.org/10.1016/j.neuron.2003.08.011
- 37.A spatiotemporal coding mechanism for background-invariant odor recognitionNat Neurosci 16:1830–1839https://doi.org/10.1038/nn.3570
- 38.Encoding a temporally structured stimulus with a temporally structured neural representationNat Neurosci 8:1568–1576https://doi.org/10.1038/nn1559
- 39.Engaging and disengaging recurrent inhibition coincides with sensing and unsensing of a sensory stimulusNat Commun 8https://doi.org/10.1038/ncomms15413
- 40.Invariant odor recognition with ON–OFF neural ensemblesProc Natl Acad Sci 119https://doi.org/10.1073/pnas.2023340118
- 41.Chandak R, Raman B. Neural manifolds for odor-driven innate and acquired appetitive preferences. Published online August 6, 2021. doi:10.1101/2021.08.05.455310https://doi.org/10.1101/2021.08.05.455310
- 42.Intensity versus Identity Coding in an Olfactory SystemNeuron 39:991–1004https://doi.org/10.1016/j.neuron.2003.08.011
- 43.Measuring the olfactory bulb input-output transformation reveals a contribution to the perception of odorant concentration invarianceNat Commun 8https://doi.org/10.1038/s41467-017-00036-2
- 44.A circuit supporting concentration-invariant odor perception in DrosophilaJ Biol 8https://doi.org/10.1186/jbiol108
- 45.Odor concentration invariance by chemical ratio coding
- 46.A primacy code for odor identityNat Commun 8https://doi.org/10.1038/s41467-017-01432-4
- 47.Recurrent cortical circuits implement concentration-invariant odor codingScience 361https://doi.org/10.1126/science.aat6904
- 48.Dynamic contrast enhancement and flexible odor codesNat Commun 9https://doi.org/10.1038/s41467-018-05533-6
- 49.Chapter 1 - Olfactory Habituation in Drosophila—Odor Encoding and its Plasticity in the Antennal LobeOdor Memory and Perception :3–38https://doi.org/10.1016/B978-0-444-63350-7.00001-2
- 50.Impaired odour discrimination on desynchronization of odour-encoding neural assembliesNature 390:70–74https://doi.org/10.1038/36335
- 51.Encoding of Olfactory Information with Oscillating Neural AssembliesScience 265:1872–1875https://doi.org/10.1126/science.265.5180.1872
- 52.Distinct Mechanisms for Synchronization and Temporal Patterning of Odor-Encoding Neural AssembliesScience 274:976–979https://doi.org/10.1126/science.274.5289.976
- 53.Network Plasticity in Adaptive Filtering and Behavioral HabituationNeuron 82:1216–1229https://doi.org/10.1016/j.neuron.2014.04.035
- 54.Olfactory receptor neurons use gain control and complementary kinetics to encode intermittent odorant stimulieLife 6https://doi.org/10.7554/eLife.27670
- 55.Sparse Odor Coding in Awake Behaving MiceJ Neurosci 26:8857–8865https://doi.org/10.1523/JNEUROSCI.0884-06.2006
- 56.Speed-Accuracy Tradeoff in OlfactionNeuron 51:351–358https://doi.org/10.1016/j.neuron.2006.07.013
- 57.Processing of Odor Mixtures in the Drosophila Antennal Lobe Reveals both Global Inhibition and Glomerulus-Specific InteractionsJ Neurosci 27:11966–11977https://doi.org/10.1523/JNEUROSCI.3099-07.2007
- 58.Single Scale for Odor Intensity in Rat OlfactionCurr Biol 24:568–573https://doi.org/10.1016/j.cub.2014.01.059
- 59.Odor intensity after self-adaptation and cross-adaptationPercept Psychophys 7:271–275https://doi.org/10.3758/BF03210163
- 60.Repetition and the brain: neural models of stimulus-specific effectsTrends Cogn Sci 10:14–23https://doi.org/10.1016/j.tics.2005.11.006
- 61.Multi-unit Recording Methods to Characterize Neural Activity in the Locust (Schistocerca Americana) Olfactory CircuitsJ Vis Exp https://doi.org/10.3791/50139
- 62.Using noise signature to optimize spike-sorting and to assess neuronal classification qualityJ Neurosci Methods 122:43–57https://doi.org/10.1016/s0165-0270(02)00276-5
- 63.Behavioural correlates of combinatorial versus temporal features of odour codesNat Commun 6https://doi.org/10.1038/ncomms7953
- 64.PARAFAC. Tutorial and applicationsChemom Intell Lab Syst 38:149–171https://doi.org/10.1016/S0169-7439(97)00032-4
- 65.A new efficient method for determining the number of components in PARAFAC modelsJ Chemom 17:274–286https://doi.org/10.1002/cem.801