Contrast polarity-specific mapping improves efficiency of neuronal computation for collision detection

  1. Richard Burkett Dewell
  2. Ying Zhu
  3. Margaret Eisenbrandt
  4. Richard Morse
  5. Fabrizio Gabbiani

  • Curated by eLife

    eLife logo

    eLife assessment

    This paper will be of interest to neuroscientists who study visual processing or are interested in dendritic integration. The authors used calcium imaging, pharmacology, and electrophysiology to investigate how a large, loom-sensitive neuron in grasshoppers integrates visual input to respond to both light and dark looming objects. These experiments support the finding that the integration is done by two distinct arbors of the neuronal dendritic tree, one of which loses retinotopic information. The authors suggest potential advantages of this dendritic arrangement.

Reviewed by

Read the full article See related articles

Listed in

Log in to save this article

Abstract

Neurons receive information through their synaptic inputs, but the functional significance of how those inputs are mapped on to a cell’s dendrites remains unclear. We studied this question in a grasshopper visual neuron that tracks approaching objects and triggers escape behavior before an impending collision. In response to black approaching objects, the neuron receives OFF excitatory inputs that form a retinotopic map of the visual field onto compartmentalized, distal dendrites. Subsequent processing of these OFF inputs by active membrane conductances allows the neuron to discriminate the spatial coherence of such stimuli. In contrast, we show that ON excitatory synaptic inputs activated by white approaching objects map in a random manner onto a more proximal dendritic field of the same neuron. The lack of retinotopic synaptic arrangement results in the neuron’s inability to discriminate the coherence of white approaching stimuli. Yet, the neuron retains the ability to discriminate stimulus coherence for checkered stimuli of mixed ON/OFF polarity. The coarser mapping and processing of ON stimuli thus has a minimal impact, while reducing the total energetic cost of the circuit. Further, we show that these differences in ON/OFF neuronal processing are behaviorally relevant, being tightly correlated with the animal’s escape behavior to light and dark stimuli of variable coherence. Our results show that the synaptic mapping of excitatory inputs affects the fine stimulus discrimination ability of single neurons and document the resulting functional impact on behavior.

Article activity feed

  1. Version published to 10.7554/elife.79772 on eLife
  2. eLife

    Author Response

    Reviewer #3 (Public Review):

    1. This work focuses exclusively on excitatory input. However, as the authors mention, LGMD neurons also receive inhibitory inputs, and these inputs also appear to segregate to different areas of the dendritic tree depending on the pathway. The contribution of inhibition is mostly ignored throughout the manuscript, but I think that it would be beneficial to discuss how inhibitory inputs fit into the story. For example, if OFF inhibition maps onto the C field, then presumably when there is mixed ON/OFF stimulation there is inhibition of the ON excitation onto the C field? If so, how much excitation of the C field is left? How much does the retainment of spatial coherence sensitivity with mixed stimuli arise from the fact that OFF excitation might dominate because it inhibits the C field? I don't think that additional experiments are needed, but a discussion would be useful. Related, does the model include inhibitory synapses?

    We have not elaborated more specifically on inhibition, as the experimental characterization of its interaction with excitation has not yet been investigated experimentally. We agree that the interaction between excitation and inhibition for mixed ON/OFF stimuli in field C is an interesting topic, but it is unlikely to affect substantially responses to ON stimuli alone. We added a paragraph on E-I integration to the discussion (lines 461-473). The model does include inhibitory synapses which are now more clearly described.

    1. The argument that the cellular organization found here is good because it allows grasshoppers to be sensitive to white approaching stimuli while disregarding spatial coherence and saving energy seems plausible. But it's not clear to me why this is 'optimal' (from the title - 'optimizes neuronal computation'). What exactly is being optimized here? And why is it good that grasshoppers can't discriminate the spatial coherence of ON looming stimuli? Is everything that approaches a grasshopper fast and white always a bad thing, but not the case if the approaching thing is black? Some further placement of these findings into an ecological setting might be helpful here.

    Our thinking is not that there is an advantage to responding to incoherent white looms (on the contrary), but that white looming stimuli in nature are likely less frequent than black/white mixtures or than all dark stimuli. Thus, the inability to discriminate white spatial coherence might have been sacrificed to decrease energy expenditure. We agree that ‘optimal’ might be too strong a wording and we have modified the title and text accordingly. Hopefully the text is now clearer on this point.

  3. eLife

    eLife assessment

    This paper will be of interest to neuroscientists who study visual processing or are interested in dendritic integration. The authors used calcium imaging, pharmacology, and electrophysiology to investigate how a large, loom-sensitive neuron in grasshoppers integrates visual input to respond to both light and dark looming objects. These experiments support the finding that the integration is done by two distinct arbors of the neuronal dendritic tree, one of which loses retinotopic information. The authors suggest potential advantages of this dendritic arrangement.

  4. eLife

    Reviewer #1 (Public Review):

    In this paper, the authors investigate how a large loom-sensitive neuron in grasshoppers becomes sensitive to looming light objects (ON looms) and looming dark objects (OFF looms). They use different visual stimuli, calcium imaging, electrophysiology, and pharmacology to identify how ON and OFF looms each elicit responses in this large neuron.

    This topic is important because the segregation of visual signals into ON and OFF channels is fundamental to visual processing, yet these signals must typically be recombined to yield useful visual signals. How and where this happens remains of interest across visual systems. This study finds that, interestingly, ON looms are integrated into the neural response via a pathway that does not retain retinotopic information. The authors suggest potential energetic and functional advantages for the observed arrangement of dendritic integration.

    The strength of this paper is in its dissection of the mechanisms of dendritic integration and in its surprising findings. The major weakness in this paper is that when the authors perform detailed modeling of the neural response, they do not provide enough information to evaluate their results. They make some strong arguments about energetic favorability of different synaptic arrangements, which are also not explained in enough detail.

  5. eLife

    Reviewer #2 (Public Review):

    The LGMD for well over 40 years has served as a model for understanding neural computations, and its mechanisms for integrating visual stimuli are thought to be well established (including past work from the authors and other labs). The LGMD has one large dendrite field that renders it selective to dark expanding objects through a combination of retinotopically distributed off inputs and intrinsic conductances. The LGMD has two smaller dendrite fields that receive on (luminance increments) or off (luminance decrements) inhibition. Surprisingly, Dewell et al. find one of the small dendrite fields, previously found to process off inhibition, also responds robustly to expanding white objects (on excitation). Interestingly, its integration strategy differs from how the larger dendrite processes off excitation. Ca2+ activity within this smaller dendrite field shows minimal to no retinotopic arrangement of inputs. Ca2+ responses to white looming stimuli are also maintained as the coherence of the stimulus decreases, suggesting the change in luminance, but not the spatial pattern of change in luminance, underlies the LGMD's response to white expanding objects. Interestingly, the grasshopper takeoff behavior, for which the LGMD is involved, also follows a similar trend. The probability a dark looming stimulus elicits an escape strongly depends on stimulus coherence, while the probability a white looming stimulus elicits an escape does not. Overall, these findings shed light on how feature inputs can be differentially computed within the same neuron and how these computations shape behavioral responses.

    Claims:

    1. ON excitation occurs on the LGMD dendrite field previously thought to receive only OFF inhibition.
    a. The authors provide calcium imaging and local delivery of cholinergic antagonist data to support this claim.

    2. ON inputs do not have retinotopic mapping across the dendrite field, unlike OFF inputs dedicated to a different dendritic field
    a. Analyzed calcium imaging data support this claim, but analysis methods need to be clarified and relevant anatomy need to be discussed in relation to the columnar structure of the lobula.

    3. Lack of retinotopy of ON inputs makes the LGMD insensitive to ON looming stimuli coherence
    a. The authors provide calcium imaging data supporting the response within the dendrite receiving ON inputs does not have a strong dependency on the coherence within the looming stimulus.

    4. Behavior follows witnessed dendrite integration, with decreasing coherence affecting escapes to dark but not white looms.
    a. The provided behavior data support these claims.

    5. Limited coherence reduces energetic cost
    a. The rationale for this claim and the methods for the modeling experiments that support these claims need to be included/expanded.

  6. eLife

    Reviewer #3 (Public Review):

    This work investigates how looming stimuli that increase in luminance are processed by the lobula giant movement detector (LGMD) neuron in grasshoppers. The manuscript starts by arguing that real life approaching predators are likely to generate a mixture of looming stimuli that increase (ON) and decrease (OFF) in luminance. Previous work has characterised well the behavioural and neurophysiological responses to OFF looms, showing that they efficiently evoke escape responses in grasshoppers and that they are mapped in a retinotopic manner to the A dendritic field for LGMD, a property important for computing that spatial coherence of the stimulus. In this manuscript, behavioural experiments show that ON looms are as efficient as OFF stimuli in eliciting escape, but that surprisingly the behaviour is independent of spatial coherence. Calcium imaging experiments show that in ON stimuli activate the C field of the LGMD neuron, suggesting a strong segregation at the cellular level between the ON and OFF pathways. Further analysis of these data show that in contrast with the OFF pathways, there appears to be no retinotopic organization of the inputs onto the dendritic tree and instead, the distribution is random. Electrophysiological recordings then reveal a progressive increase in firing rate as the ON looming stimulus approaches, with a profile that is independent of the spatial coherence of the stimulus, in agreement with the behaviour. The manuscript ends by demonstrating that mixed ON and OFF looming stimuli activate both the C and A dendritic fields and retain sensitivity to spatial coherence, and a biophysical model is shown to reproduce the experimental findings.

    The overall conclusion from this work is that the visual system of the grasshopper is sensitive to ON approaching stimuli, but it is unable to discriminate their spatial coherence because of the random distribution of ON inputs onto the LGMD dendritic tree. The authors further argue that this organization allows grasshoppers to be sensitive to these stimuli while reducing the number of synapses require to reach AP threshold, thereby conserving energy. I think that the experiments are very nicely done, well designed, the data are of great quality and support the main arguments. The greatest strength of this work, and indeed of the model system, is the ability to link behaviour, sensory processing, and cellular physiology with biophysical detail in a single piece of work. I believe that this is a valuable contribution to all these fields. I have a couple of main comments for the authors to consider.

    1 - This work focuses exclusively on excitatory input. However, as the authors mention, LGMD neurons also receive inhibitory inputs, and these inputs also appear to segregate to different areas of the dendritic tree depending on the pathway. The contribution of inhibition is mostly ignored throughout the manuscript, but I think that it would be beneficial to discuss how inhibitory inputs fit into the story. For example, if OFF inhibition maps onto the C field, then presumably when there is mixed ON/OFF stimulation there is inhibition of the ON excitation onto the C field? If so, how much excitation of the C field is left? How much does the retainment of spatial coherence sensitivity with mixed stimuli arise from the fact that OFF excitation might dominate because it inhibits the C field? I don't think that additional experiments are needed, but a discussion would be useful. Related, does the model include inhibitory synapses?

    2 - The argument that the cellular organization found here is good because it allows grasshoppers to be sensitive to white approaching stimuli while disregarding spatial coherence and saving energy seems plausible. But it's not clear to me why this is 'optimal' (from the title - 'optimizes neuronal computation'). What exactly is being optimized here? And why is it good that grasshoppers can't discriminate the spatial coherence of ON looming stimuli? Is everything that approaches a grasshopper fast and white always a bad thing, but not the case if the approaching thing is black? Some further placement of these findings into an ecological setting might be helpful here.

  7. Version published to 10.1101/2022.05.10.491323v1 on bioRxiv