Mechanical vibration patterns elicit behavioral transitions and habituation in crawling Drosophila larvae

Article activity feed

1. 

   Version published to 10.7554/elife.69205 on eLife

   Oct 19, 2023
2. 

   eLife

   Sep 7, 2023

   Author Response

   Reviewer #1 Public Review:

   In this manuscript, Berne et al apply state-of-the-art methodology for quantifying animal behavior to identify distinct behavioral components associated with the repeated application of mechanical stimuli. A central strength of this manuscript is the development of a sophisticated system for precisely applying mechanical stimuli and measuring behavior. This is a significant advance over commonly used approaches and has the potential to broadly impact the field. I have some concerns about the methods used to define discrete behaviors and the interpretations drawn from them (see point 2), the opposing phenotypes of memory mutants, and the circuit modeling. However, the overall results provide strong evidence that a small set of behaviors reflect the intensity of response to stimuli, and these …

   Author Response

   Reviewer #1 Public Review:

   In this manuscript, Berne et al apply state-of-the-art methodology for quantifying animal behavior to identify distinct behavioral components associated with the repeated application of mechanical stimuli. A central strength of this manuscript is the development of a sophisticated system for precisely applying mechanical stimuli and measuring behavior. This is a significant advance over commonly used approaches and has the potential to broadly impact the field. I have some concerns about the methods used to define discrete behaviors and the interpretations drawn from them (see point 2), the opposing phenotypes of memory mutants, and the circuit modeling. However, the overall results provide strong evidence that a small set of behaviors reflect the intensity of response to stimuli, and these combine to reflect an overall complex behavioral response to mechanical stimuli. Overall the manuscript is well written, and clearly communicates results. The level of analysis has the potential to broadly impact many fields examining innate and learned responses to sensory stimuli.

   1. A central strength of this manuscript is the resolution of behavioral analysis. Implicit in this is the potential to use a wealth of genetic analysis and sophisticated genetic tools to dissect the neural basis of these behaviors. These implications would be clearer if the introduction provided more description of this literature.

   This is certainly true, where the findings from behavior experiments should lead to interesting investigations at the neural circuit level. This is especially true for Drosophila, which has a wealth of genetic tools readily available. We have added a new paragraph at the end of the Introduction section to discuss this, and provide citations to a number of commonly used tools that could be used to identify and characterize the circuit side of mechano-sensation and adaptation in flies.

   1. It is unclear how the 4 discrete behaviors were decided upon, and whether there are rarer behaviors, or subcategories within them (for example, sideways crawl).

   We do list a number of behaviors in the third paragraph of the Introduction, and describe some of these in more detail in the next paragraph, but agree that a clearer justification needs to be given for focusing on the four specific behaviors in the paper. The answer is that these are the only behaviors that larvae perform given the constraints we place on their movement (hard, flat agar gel), and because we avoid overly strong stimuli that would cause more drastic pain responses. This is now noted directly near the end of the 5th paragraph of the Introduction.

   1. From figure 1A it looks like the mechanical transducer remains in the center independently of where the larvae is. Could it be possible that subtle differences in mechanical force are detected across the arena and this impacts the response? Does the degree of turning matter?

   While the first paragraph of the Results section notes we use a “customized platform,” and the details and purpose of this are listed later in the second paragraph of Materials and Methods, I think it is warranted to include more details up front, as many readers will likely have the same question. We now clearly state what is customized about the platform and that its purpose is to achieve a spatially uniform vibration stimulus, and point the reader to Materials and Methods for further details.

   1. I am not clear about the application of statistics. For example, 2D states that as a general trend, increasing vibration also increases reversals. I can see this, clearly but is there reason not to run statistics on these data?

   We agree, it is not sufficient to simply state there is a general trend, when statistics can be readily applied (especially to binary/fractional data like this!). We have performed statistical comparison tests for reverse crawling response probabilities in the data in Figure 2C, which shows fractional behavior usage for a wide range of vibration frequency and acceleration. We show the statistics in two ways. (1) Adjacent graphs are connected with bridging lines that are black (p>0.05) or yellow (p<0.05) (Fisher’s exact test for both), which shows the onset of significant reverse crawling behavior when looking along gamma or f axes. (2) Each of the 29 graphs was tested against the baseline (zero vibration) reverse crawl fraction, and red dots indicate significant reverse crawl use. The graphs and captions for Figure 2C have been updated accordingly.

   We also did more serious statistics with the data in Figure 5 (habituation model compared to data) and Figure 7 (simple circuit model compared to data), and those are described below with their associated comments.

   1. The importance of vibration behavior in research is discussed but the ecological relevance of these behaviors is not described.

   A very good idea for setting the context better. We have added a new paragraph to the Introduction with 56 references for readers interested in learning more about this side of things. Vibration response is important in real larvae in nature too, it helps them communicate and avoid predators.

   1. The results of habituation times in mutants are not clear to me. One might predict dnc and rut would have the same phenotype but they have opposing phenotypes with rut being a super-habituate.

   The dnc and rut mutants both desensitize faster than the CS control larvae (comparing the traces in Fig6A to the gray wild type version), which would agree with this prediction, but the details are still finer details to sort out. For example that rut is faster than dnc, or that rut is faster at both desensitizing and re-sensitizing than wild type, but dnc is slow to re-sensitize. This would be interesting to piece together, but for now the mutant results highlight the importance of extracting the finer details (and multiple time constants) involved in vibration response, and explaining why the mutants (or other future strains tested) have the specific values is a bit beyond the scope of this paper.

   We have noted the comparisons with dnc and rut more directly in the text now, accompanying the descriptions of Fig. 6A and 6B in the Results section.

   1. I appreciate the application of circuit modeling, but it would seem that this would be strengthened by including what is already known about the biological circuit.

   We were not very clear about describing the purpose of the circuit model – we did not intend the circuit components of the model to directly match the actual neural circuit elements. It is primarily a visualization tool for what appears to be happening based on the empirical results (although the math behind the circuit might suggest some possible real mechanisms, noted in Discussion). In earlier drafts the visualization tool was a water bucket pouring into a second bucket with a hole in the bottom, with water volume analogous to habituation (the math was identical to the capacitor circuit). We have added a sentence at the beginning of the circuit model section to clarify its purpose better.

   That said, we agree it is important to discuss the context of the real neural circuit. This was in the Introduction already, but not emphasized or introduced very well. This section now has its own paragraph, which we have expanded and added additional references (paragraph starting with “Some aspects of the neural circuitry…”).

   We have also substantially edited the Results section about the circuit model in response to other comments below, and it should be more focused and clearer now.

   Reviewer #2 (Public Review):

   Berne et al. establish the responses of Drosophila larvae to mechanical vibrations as a novel paradigm to study habituation. The authors first comprehensively quantify the different types of locomotor responses to vibrations and find that larvae respond to faster and stronger vibrations with more avoidance-type behaviors, like pauses, turns, and reversals. The authors then combine genetic and computational methods to characterize the strong de-sensitization of avoidance responses to vibrations. De-sensitization of reversals follows a simple, exponential decay with a single time constant. By contrast, re-sensitization dynamics are more complex and strongly accelerate after repeated exposure to a vibration stimulus. The authors then test mutants for genes involved in learning and memory (rut, dnc, cam) and find altered desensitization and re-sensitization dynamics, suggesting that these genes mediate this behavior. Finally, a simple and intuitive electrical circuit model is used to explain these complex dynamics results. Overall, the results are interesting and they successfully combine behavioral characterization, genetic manipulations, and computational modeling to explain the behavior.

   The analyses are all sound and support most of the conclusions but additional control experiments and analyses are required.

   1. To convincingly show that the computational models capture the key aspects of the behavior and therefore provide insight into the underlying phenomenon, model predictions and behavioral data need to be compared systematically and quantitatively. This is not sufficiently done for the electrical circuit model, and the analyses shown in Fig. 7C need to be extended. The model should be fitted to the data and the match between model and data should be A) quantified using a suitable measure of goodness-of-fit and B) illustrated by overlaying behavioral data and model predictions.

   We agree, and thank the referee for pointing this out. The circuit model was intended as primarily a visualization tool, but it was not fair of us to say that it correctly predicts anything real without being more precise and quantitative, including using significance metrics. We also feel that Fig. 7C was not a very compelling demonstration and not very interesting. We have replaced 7C with a new panel that shows empirical reverse crawl probability overlayed with the circuit model’s prediction of reverse crawl behavior (where FREV ~ exp(-Q2). The peak values match very closely, although the overall shape does not, due to the simplicity of the model. This is discussed fully in the Results text and in a redone Fig. 7 caption.

   Moreover, the contribution of individual circuit elements should be quantified, for instance by removing key elements from the model like the second capacitor. If a good quantitative fit is for some reason hard to obtain, then more effort should be spent to demonstrate a good qualitative agreement between model and data.

   We have shown what we think is the bare minimum circuit model that can include the accumulation and decay of a substance (the charge Q2 standing in for “habituation”). We could have built a more complicated circuit and essentially forced it to have the same time constants as we extracted from data, but felt that would lose sight of its appeal as a visualization tool and qualitative idea. We could not remove C2, for example, because the “output” of the circuit model itself is the charge on that capacitor.

   In response to further comments below we have overhauled and simplified the section about the circuit model, and hope this also helps alleviate any concerns.

   The same goes for the phenomenological model in Fig. 5. Predictions of model variants with a constant re-sensitization time constant and a time constant that changes with pulse number should be shown and their fit to the data should be quantified.

   Absolutely. We have added two other versions of the model to Fig. 5E (one with only desensitization and the other that doesn’t have the time constant changing with pulse number) and performed significance tests on the peak values for each pulse response. The model with all three aspects of habituation performs the best. Fig. 5E has been made larger to better see the traces, we have added visual cues and a legend for the significance tests, and the caption has been expanded accordingly.

   1. The Markov model in Fig. 3 is used to state that habituation is a one-way process from reversals to other behaviors, with only rare transitions back to reversals. However, the low transition rates to reversals (Fig. 3) seem at odds with the fast re-sensitization after repeated stimulation (Fig. 5). This should be explained and both results should be linked.

   This is a really good observation, and fortunately does have an explanation. The assigned behaviors in Fig. 3 are what we observe during the first 3 seconds after vibration onset. Habituation sets in as the stimulus stays on, then re-sensitization (even if not complete) occurs while the stimulus is off. Then when the stimulus turns on again, we assign the next behavior. An individual with a strong (reversal) response will most often (85% of the time) reverse again the next time the stimulus turns on. We would not classify that as a transition back to reversal, but as a repeat of the reversal behavior following de-sensitization and resensitization. For the 15% of individuals that did not reverse the second time, they will only very rarely (< 2%) reverse the third time. The re-sensitization process in fact explains why strong response behaviors so often repeat for the next vibration pulse response.

   We have expanded a paragraph in the Results section to add text similar to what we have written here to clear up this point. It’s the last paragraph in the “Re-sensitization rates increase…” subsection.

   1. Based on altered de-sensitization and re-sensitization dynamics in mutants, the authors claim that three different genes - rut, dnc, cam - are involved in the molecular pathway that mediates habituation of larval locomotor responses to vibrations. This is interesting and deserves further study. However, it is unclear whether the observed effects are specific to the genes that were altered or whether the effects stem from differences in the genetic background across the mutants. This could be resolved in two ways: Ideally, with rescue experiments; if this is not feasible, then data from different wild-type strains could be used to show that the de-sensitization and re-sensitization dynamics are similar across wild types and somewhat robust to genetic background.

   Additional control data with other wild type strains was not doable due to personnel issues noted in our resubmission letter, and also time constraints (for example, each trace like the one in Fig. 5A requires 1000 animals to construct – we suspect that the required number of larva-hours to determine habituation parameters is a large part of why other researchers have not observed these habituation characteristics in larvae before). We do acknowledge this limitation directly in the manuscript now, and highlight why it would be important for further experiments like these to be carried out in the future. A new paragraph in the “Conclusions” subsection of Discussion discusses this. We now state directly that the mutant results are there to highlight the importance of characterizing multiple time constants and other dependencies when determining anything about habituation. The fact that habituation parameters are not the same as this particular CS wild type is suggestive, but given the lack of additional controls it would not be fair to make specific statements about any of the mutants at this stage.

   Read the original source
3. 

   eLife

   Jul 21, 2021

   Evaluation Summary:

   This is a strong manuscript due to its sophisticated behavioral analysis and modeling of behavioral output. The system and results provide a framework for future genetic analysis examining the biological basis of sensory behaviors.

   (This preprint has been reviewed by eLife. We include the public reviews from the reviewers here; the authors also receive private feedback with suggested changes to the manuscript. Reviewer #2 agreed to share their name with the authors.)

   Read the original source
4. 

   eLife

   Jul 21, 2021

   Reviewer #1 (Public Review):

   In this manuscript, Berne et al apply state-of-the-art methodology for quantifying animal behavior to identify distinct behavioral components associated with the repeated application of mechanical stimuli. A central strength of this manuscript is the development of a sophisticated system for precisely applying mechanical stimuli and measuring behavior. This is a significant advance over commonly used approaches and has the potential to broadly impact the field. I have some concerns about the methods used to define discrete behaviors and the interpretations drawn from them (see point 2), the opposing phenotypes of memory mutants, and the circuit modeling. However, the overall results provide strong evidence that a small set of behaviors reflect the intensity of response to stimuli, and these combine to …

   Reviewer #1 (Public Review):

   In this manuscript, Berne et al apply state-of-the-art methodology for quantifying animal behavior to identify distinct behavioral components associated with the repeated application of mechanical stimuli. A central strength of this manuscript is the development of a sophisticated system for precisely applying mechanical stimuli and measuring behavior. This is a significant advance over commonly used approaches and has the potential to broadly impact the field. I have some concerns about the methods used to define discrete behaviors and the interpretations drawn from them (see point 2), the opposing phenotypes of memory mutants, and the circuit modeling. However, the overall results provide strong evidence that a small set of behaviors reflect the intensity of response to stimuli, and these combine to reflect an overall complex behavioral response to mechanical stimuli. Overall the manuscript is well written, and clearly communicates results. The level of analysis has the potential to broadly impact many fields examining innate and learned responses to sensory stimuli.

   1. A central strength of this manuscript is the resolution of behavioral analysis. Implicit in this is the potential to use a wealth of genetic analysis and sophisticated genetic tools to dissect the neural basis of these behaviors. These implications would be clearer if the introduction provided more description of this literature.

   2. It is unclear how the 4 discrete behaviors were decided upon, and whether there are rarer behaviors, or subcategories within them (for example, sideways crawl).

   3. From figure 1A it looks like the mechanical transducer remains in the center independently of where the larvae is. Could it be possible that subtle differences in mechanical force are detected across the arena and this impacts the response? Does the degree of turning matter?

   4. I am not clear about the application of statistics. For example, 2D states that as a general trend, increasing vibration also increases reversals. I can see this, clearly but is there reason not to run statistics on these data?

   5. The importance of vibration behavior in research is discussed but the ecological relevance of these behaviors is not described.

   6. The results of habituation times in mutants are not clear to me. One might predict dnc and rut would have the same phenotype but they have opposing phenotypes with rut being a super-habituate.

   7. I appreciate the application of circuit modeling, but it would seem that this would be strengthened by including what is already known about the biological circuit.

   Read the original source
5. 

   eLife

   Jul 21, 2021

   Reviewer #2 (Public Review):

   Berne et al. establish the responses of Drosophila larvae to mechanical vibrations as a novel paradigm to study habituation. The authors first comprehensively quantify the different types of locomotor responses to vibrations and find that larvae respond to faster and stronger vibrations with more avoidance-type behaviors, like pauses, turns, and reversals. The authors then combine genetic and computational methods to characterize the strong de-sensitization of avoidance responses to vibrations. De-sensitization of reversals follows a simple, exponential decay with a single time constant. By contrast, re-sensitization dynamics are more complex and strongly accelerate after repeated exposure to a vibration stimulus. The authors then test mutants for genes involved in learning and memory (rut, dnc, cam) and …

   Reviewer #2 (Public Review):

   Berne et al. establish the responses of Drosophila larvae to mechanical vibrations as a novel paradigm to study habituation. The authors first comprehensively quantify the different types of locomotor responses to vibrations and find that larvae respond to faster and stronger vibrations with more avoidance-type behaviors, like pauses, turns, and reversals. The authors then combine genetic and computational methods to characterize the strong de-sensitization of avoidance responses to vibrations. De-sensitization of reversals follows a simple, exponential decay with a single time constant. By contrast, re-sensitization dynamics are more complex and strongly accelerate after repeated exposure to a vibration stimulus. The authors then test mutants for genes involved in learning and memory (rut, dnc, cam) and find altered de-sensitization and re-sensitization dynamics, suggesting that these genes mediate this behavior. Finally, a simple and intuitive electrical circuit model is used to explain these complex dynamics results. Overall, the results are interesting and they successfully combine behavioral characterization, genetic manipulations, and computational modeling to explain the behavior.

   The analyses are all sound and support most of the conclusions but additional control experiments and analyses are required.

   1. To convincingly show that the computational models capture the key aspects of the behavior and therefore provide insight into the underlying phenomenon, model predictions and behavioral data need to be compared systematically and quantitatively. This is not sufficiently done for the electrical circuit model, and the analyses shown in Fig. 7C need to be extended. The model should be fitted to the data and the match between model and data should be A) quantified using a suitable measure of goodness-of-fit and B) illustrated by overlaying behavioral data and model predictions. Moreover, the contribution of individual circuit elements should be quantified, for instance by removing key elements from the model like the second capacitor. If a good quantitative fit is for some reason hard to obtain, then more effort should be spent to demonstrate a good qualitative agreement between model and data.

   The same goes for the phenomenological model in Fig. 5. Predictions of model variants with a constant re-sensitization time constant and a time constant that changes with pulse number should be shown and their fit to the data should be quantified.

   1. The Markov model in Fig. 3 is used to state that habituation is a one-way process from reversals to other behaviors, with only rare transitions back to reversals. However, the low transition rates to reversals (Fig. 3) seem at odds with the fast re-sensitization after repeated stimulation (Fig. 5). This should be explained and both results should be linked.

   2. Based on altered de-sensitization and re-sensitization dynamics in mutants, the authors claim that three different genes - rut, dnc, cam - are involved in the molecular pathway that mediates habituation of larval locomotor responses to vibrations. This is interesting and deserves further study. However, it is unclear whether the observed effects are specific to the genes that were altered or whether the effects stem from differences in the genetic background across the mutants. This could be resolved in two ways: Ideally, with rescue experiments; if this is not feasible, then data from different wild-type strains could be used to show that the de-sensitization and re-sensitization dynamics are similar across wild types and somewhat robust to genetic background.

   Read the original source
6. 

   eLife

   Jul 21, 2021

   Reviewer #3 (Public Review):

   In this work the authors seek to characterize in detail how Drosophila larvae are de-sensitized and re-sensitized to aversive stimuli. They perform technically sophisticated high-throughput measurements of the animal's behavior in response to mechanosensory stimuli, and they set out to provide a simple model that captures the key attributes of de-sensitization and re-sensitization. The authors succeed in their effort. Their characterization is, to my knowledge, the most detailed to-date. They also provide a simple mathematical description, and show that the same math suggests an elegant analog electrical circuit. The circuit provides a useful conceptual model, and, as the authors hint, may even map onto certain biological mechanisms. Moreover, the authors show that their method could be useful for an …

   Reviewer #3 (Public Review):

   In this work the authors seek to characterize in detail how Drosophila larvae are de-sensitized and re-sensitized to aversive stimuli. They perform technically sophisticated high-throughput measurements of the animal's behavior in response to mechanosensory stimuli, and they set out to provide a simple model that captures the key attributes of de-sensitization and re-sensitization. The authors succeed in their effort. Their characterization is, to my knowledge, the most detailed to-date. They also provide a simple mathematical description, and show that the same math suggests an elegant analog electrical circuit. The circuit provides a useful conceptual model, and, as the authors hint, may even map onto certain biological mechanisms. Moreover, the authors show that their method could be useful for an investigation into biological mechanisms by characterizing mutants that are defective for certain genes thought to be involved in de-sensitization. If there were to be a weakness of the work, it would be that the authors leave to the future the challenging task of chasing down details of the underlying mechanisms of de-sensitization and re-sensitization.

   Read the original source
7. 

   Version published to 10.1101/2021.04.26.441415v1 on bioRxiv

   Apr 27, 2021