Adaptation in cone photoreceptors contributes to an unexpected insensitivity of primate On parasol retinal ganglion cells to spatial structure in natural images

  1. Zhou Yu
  2. Maxwell H Turner
  3. Jacob Baudin
  4. Fred Rieke

  • Curated by eLife

    eLife logo

    Evaluation Summary:

    This manuscript provides strong evidence that adaptation in cone photoreceptors of the primate retina can subtly change the balance of excitatory and inhibitory inputs to On parasol ganglion cells and thereby fundamentally affect how these cells integrate visual information. This study provides important mechanistic insight into the previous observation that On parasol cells display nonlinear spatial stimulus integration under standard reversing gratings but linearly integrate signals in the context of natural scenes. The findings will be of great interest to visual neuroscientists.

    (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 #1 and Reviewer #2 agreed to share their name with the authors.)

Reviewed by

Read the full article See related articles

Listed in

Log in to save this article

Abstract

Neural circuits are constructed from nonlinear building blocks, and not surprisingly overall circuit behavior is often strongly nonlinear. But neural circuits can also behave near linearly, and some circuits shift from linear to nonlinear behavior depending on stimulus conditions. Such control of nonlinear circuit behavior is fundamental to neural computation. Here, we study a surprising stimulus dependence of the responses of macaque On (but not Off) parasol retinal ganglion cells: these cells respond nonlinearly to spatial structure in some stimuli but near linearly to spatial structure in others, including natural inputs. We show that these differences in the linearity of the integration of spatial inputs can be explained by a shift in the balance of excitatory and inhibitory synaptic inputs that originates at least partially from adaptation in the cone photoreceptors. More generally, this highlights how subtle asymmetries in signaling – here in the cone signals – can qualitatively alter circuit computation.

Article activity feed

  1. Version published to 10.7554/elife.70611 on eLife
  2. eLife

    Author Response

    Reviewer #1 (Public Review):

    Strengths: The manuscript is very thorough and convincingly homes in on key circuit elements and mechanisms that likely underpin this unexpected linearity in the On-parasol circuit.

    Weaknesses: perhaps this is just me, but I found the MS is quite "hard work" to parse. To some extent I suspect this just is what it is, given the complexity of the circuit, but perhaps the occasional explanatory statement and further streamlining of the figures might help get the key points across a little more easily. I was also a little unsure about what exactly each figure and statistical measure depicted, something that presumably is easily fixed by expanding the legends accordingly. I offer some specific suggestions in the author section.

    Thank you - this is helpful. We have revised the text throughout with this comment in mind. This included being clearer about what is plotted in each figure, both in the main text and in the figure legends. We have also expanded (and in some cases added) the final paragraphs of each section of the Results to make sure that the main points of each section are emphasized.

    Finally, the manuscript rather glances over "the middle" of the circuit, i.e. the bipolar cells, and their computations. The circuit insights gained from the study do not require spatially nonlinear computations in bipolar cells (e.g. via amacrine cells), rather, they seem to require the bipolar cells to be linear in this regard. However bipolar cells, like cones or ganglion cells, also have many sources of nonlinearities, some well understood and probably several others less well understood. Is it then not surprising that they should not play a role in the circuit as well? It would be good to see some discussion/acknowledgement of this topic.

    We have revised the text to emphasize the importance of nonlinearities in the bipolar cells, starting with our description of Figure 1A. We previously discussed this issue somewhat abstractly - referring to “rectifying nonlinearities.” We have now clarified that those nonlinearities, which are essential for nonlinear spatial integration, largely originate at the bipolar output synapse. These changes include clarifying that what we describe in this paper is a previously unappreciated mechanism, initiated from adaptation in the cones, that controls the impact of the bipolar nonlinearities on RGC responses. We find this interplay between different nonlinear components very interesting (and have emphasized that now in the Discussion). This comment helped us describe that interaction in much more concrete terms. See page 11 (next to last paragraph), page 14 (next to last paragraph), and page 17 (last paragraph).

    Reviewer #3 (Public Review):

    This manuscript reports an interesting result regarding retinal ganglion cells in primates, although the presentation of the main findings was slightly confusing. There is a classic distinction between ganglion cells that integrate linearly vs. nonlinearly over space. The primary test has been the presentation of a high spatial frequency contrast-reversing grating, which generates no response in a linear cell - because the responses to bright and dark bars cancel each other. A nonlinear cell would instead respond at twice the frequency of reversal, because (e.g., for an ON cell) the increased response to the bright bars on each phase of the grating cannot be canceled by the decreased response to the dark bars. The explanation has been a nonlinearity at the output of bipolar cells, such that increases in glutamate release to the preferred contrast cannot be canceled by decreases in release to the non-preferred contrast, either because the release rate at baseline is very low (i.e., rectified) or else there is some difference in the dynamics to increases vs. decreases in release even when the output is not strongly rectified. This latter idea is illustrated in Borghuis et al. (2013; Fig. 7) - which the authors might consider citing.

    Thank you. We have added a citation to that paper (along with the citations to two earlier papers from Jon Demb that implicate bipolar cells (page 2, top; page 11, second to last paragraph; page 17, last paragraph). We have also emphasized throughout the paper the important role that the nonlinearity at the bipolar cell output synapse plays in nonlinear spatial integration. This is highlighted in Figure 1A, which we now refer back to several times. We previously referred to the key nonlinearity abstractly as a “rectifying nonlinearity” and now more directly indicate the role of the bipolar synapse. See page 11 (next to last paragraph), page 14 (next to last paragraph), and page 17 (last paragraph).

    The authors report that On parasol cells in primate show the nonlinear behavior to a high contrast grating. Although - and this an important detail - there is no such response at the first half-cycle of the grating (Fig. 5C). Likewise, there is no response to a grating flashed briefly from a gray background (Fig. 1C). There could be an advantage to presenting these results together at the front of the paper - indeed they seem to show results from the same cell (at least, Fig. 1C and 5A spike traces look identical).

    Thank you - this is a helpful suggestion. We have moved full responses to contrast-reversing gratings into Figure 1, and revised the text considerably to point out the differences in spatial integration at the onset of the grating vs during later grating cycles (gray boxes in Figure 1B; see page 2, bottom and page 3, second paragraph). This included modifying the text to emphasize that our results highlight that spatial integration depends (surprisingly) on stimulus time course. Natural images fit into this stimulus dependence, but other stimuli also elicit near-linear spatial integration. Throughout the paper, staarting in the abstract, we now emphasize that spatial integration is stimulus dependent. We hope that this shifts the focus away from specific stimuli towards the more general issue of how this stimulus dependence originates.

    The main section of the paper shows that natural images presented either in sequence or flashed briefly from a gray background lack nonlinear response behavior - that is, a linear receptive field seems to be encoding the image, because structure in the image can be replaced by the mean luminance across the receptive field center. This makes natural images and gratings seem like they bring out truly unique behaviors in the ganglion cell. But what could be so special about a grating?

    It seems, simply, that the natural image dataset did not match the high contrast grating in the detailed properties of the stimulus, and the proposed nonlinearity in the cone is also obviously important. For example, when the grating is first presented in the contrast-reversing stimulus sequence (i.e, the first half-cycle) or the grating is simply flashed briefly, the On parasol cell is excited by the presynaptic On bipolar cells that see a gray-to-white transition. But the release is relatively weak, because the cones are partly adapted to the gray background. The inhibition driven by 'crossover' circuitry (i.e., driven by the Off pathway) is relatively strong at the gray-to-black transition, and the responses cancel.

    The lack of response to spatial structure cannot be explained by the range of contrasts alone. To help make that point, we now show the contrast dependence of the grating responses in Figure 1. The image patches in our dataset often had contrasts exceeding 0.5. Similarly, the onset of a high contrast grating elicited little or no response in On parasol cells (Figures 1B and C). We have emphasized these points in the text on page 4, and we have added a few sentences about the image statistics to the methods (page 20, second to last paragraph). Instead, our analysis indicates that it is the periodicity of the contrast-reversing grating that is important - and specifically that this periodicity leads to periodic changes in the gain of the cone response. We emphasize this in the revised paper in several places (see page 10, next to last paragraph; page 13, first and second paragraph; page 18, second paragraph).

    What is less obvious is that for subsequent cycles of the grating, the On bipolar cells that see black-to-white transitions are now releasing very strongly (because the cone response is relatively strong coming from an un-adapted state), which can no longer be canceled by the Off-pathway inhibition. The grating is the optimal stimulus to reveal this nonlinearity because there are simultaneous full-contrast changes in luminance, in opposite directions, at different points across the image. There is no reason why a similar sequence in a natural movie would not drive a similarly strong nonlinear response. It just happens that the natural images and movies (i.e., image sequences) used here apparently never contain these patterns of light changes across the receptive field center.

    Indeed, we do not mean to imply that there is anything special about natural images, and we have revised the text so as not to draw a specific distinction between natural stimuli and others. This includes emphasizing that the periodicity of contrast-reversing gratings appears central to the strong responses that they elicit. This periodicity links the gain of the cone responses to the stimulus that the cones encounter: cones encountering an increase in light level have a high signaling gain because they start from a low light level, and conversely cones encountering a decrease in light level have a low signaling gain. This periodic change in intensity is not commonly encountered in natural stimuli (though we agree that natural stimuli exhibiting such a periodic change would likely elicit nonlinear spatial integration).

    So there is a fundamental feature of the circuitry that determines the degree of nonlinear spatial summation in On parasol cells (in an interesting way that differs from Off parasol cells). My main recommendation is that the authors present the findings with gratings in a more compact fashion rather than alternating between gratings and natural images. For example, the results in FIg. 1B and C are puzzling until one learns of the result in Fig. 5A and realizes that Fig. 1B is showing the average across many cycles, which does not capture the unusual response in the first half cycle that is completely consistent with the result in Fig. 1C.

    Good suggestion. We have included a more complete summary of the grating responses in Figure 1 - including removing the cycle average responses to contrast-reversing gratings in favor of traces that show the responses to the grating onset as well as responses to later grating cycles (Figure 1B). We have emphasized these features of the grating response in the text as well.

    There is also a clear difference in the contrast of the grating vs. the contrasts within the natural scenes (which are more difficult to define, perhaps). It could be interesting to explore the responses to lower contrast gratings to see if a single model can explain both grating and natural image responses over a range of contrasts in a satisfying way.

    See response above. We now include summary data showing F2 responses across a range of grating contrasts to demonstrate that high contrast values are not necessary to elicit nonlinear spatial integration in either On or Off cells. We note this in the text on page 4, first paragraph. We have also added a sentence about the contrasts of the natural image patches used here to the Methods (see page 20, next to last paragraph).

  3. eLife

    Evaluation Summary:

    This manuscript provides strong evidence that adaptation in cone photoreceptors of the primate retina can subtly change the balance of excitatory and inhibitory inputs to On parasol ganglion cells and thereby fundamentally affect how these cells integrate visual information. This study provides important mechanistic insight into the previous observation that On parasol cells display nonlinear spatial stimulus integration under standard reversing gratings but linearly integrate signals in the context of natural scenes. The findings will be of great interest to visual neuroscientists.

    (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 #1 and Reviewer #2 agreed to share their name with the authors.)

  4. eLife

    Reviewer #1 (Public Review):

    Strengths: The manuscript is very thorough and convincingly homes in on key circuit elements and mechanisms that likely underpin this unexpected linearity in the On-parasol circuit.

    Weaknesses: perhaps this is just me, but I found the MS is quite "hard work" to parse. To some extent I suspect this just is what it is, given the complexity of the circuit, but perhaps the occasional explanatory statement and further streamlining of the figures might help get the key points across a little more easily. I was also a little unsure about what exactly each figure and statistical measure depicted, something that presumably is easily fixed by expanding the legends accordingly. I offer some specific suggestions in the author section.

    Finally, the manuscript rather glances over "the middle" of the circuit, i.e. the bipolar cells, and their computations. The circuit insights gained from the study do not require spatially nonlinear computations in bipolar cells (e.g. via amacrine cells), rather, they seem to require the bipolar cells to be linear in this regard. However bipolar cells, like cones or ganglion cells, also have many sources of nonlinearities, some well understood and probably several others less well understood. Is it then not surprising that they should not play a role in the circuit as well? It would be good to see some discussion/acknowledgement of this topic.

  5. eLife

    Reviewer #2 (Public Review):

    The manuscript by Yu, Turner, Baudin, and Rieke investigates spatial integration of visual information by On parasol cells in the primate (macaque) retina. Having previously shown that these cells integrate natural stimuli linearly over space, but show characteristics of nonlinear spatial integration when probed with standard reversing gratings, the authors here study the mechanisms behind this surprising stimulus-dependent difference of spatial integration characteristics. Using patch-clamp recordings of synaptic inputs into On parasol cells as well as computational modeling, they reveal that crossover inhibitory inputs, which are elicited by Off-type stimulation, linearizes spatial integration and that this crossover inhibition is strongly triggered by natural stimuli as well as at the onset of a spatial grating stimulus, but is reduced during ongoing stimulation with reversing gratings, leading to nonlinear integration. The authors further show that this shift in crossover inhibition may follow from adaptation dynamics in the cone photoreceptors, which shift the balance between excitation triggered by brightening and inhibition triggered by darkening. Experiments with modified grating stimuli, designed to mimic photoreceptor signals without adaptation, confirm that subtle asymmetries in the signaling of bright and dark stimulus parts can fundamentally alter the integration characteristics of these retinal ganglion cells.

    This is an excellent manuscript, which combines state-of-the-art experiments with sophisticated modeling and data analysis to tackle an important question about signal processing in the retina. It provides a convincing line of reasoning, from clear characterizations of linear and nonlinear stimulus integration under different stimulus contexts via a clever evaluation of the relative contributions of feedforward and crossover inhibition to the model-based evaluation of adaptation dynamics in cone photoreceptors. The idea that photoreceptor adaptation evokes a systematic shift in balance between excitation and crossover inhibition under reversing gratings, which represent a widely used standard stimulus, is thought-provoking, and the clever use of a computational model to transform the visual stimulus for probing retinal processing without photoreceptor adaptation is compelling. My only major comment is that the manuscript provides relatively little information about the photoreceptor adaptation model. Explaining how the adaptation dynamics relate to the asymmetry in the responses and how the non-adapting, linear version of the model compares to the full model, would help understand why the linear cones create a scenario that's more similar to the steady-state response under reversing gratings than to the onset response.

  6. eLife

    Reviewer #3 (Public Review):

    This manuscript reports an interesting result regarding retinal ganglion cells in primates, although the presentation of the main findings was slightly confusing. There is a classic distinction between ganglion cells that integrate linearly vs. nonlinearly over space. The primary test has been the presentation of a high spatial frequency contrast-reversing grating, which generates no response in a linear cell - because the responses to bright and dark bars cancel each other. A nonlinear cell would instead respond at twice the frequency of reversal, because (e.g., for an ON cell) the increased response to the bright bars on each phase of the grating cannot be canceled by the decreased response to the dark bars. The explanation has been a nonlinearity at the output of bipolar cells, such that increases in glutamate release to the preferred contrast cannot be canceled by decreases in release to the non-preferred contrast, either because the release rate at baseline is very low (i.e., rectified) or else there is some difference in the dynamics to increases vs. decreases in release even when the output is not strongly rectified. This latter idea is illustrated in Borghuis et al. (2013; Fig. 7) - which the authors might consider citing.

    The authors report that On parasol cells in primate show the nonlinear behavior to a high contrast grating. Although - and this an important detail - there is no such response at the first half-cycle of the grating (Fig. 5C). Likewise, there is no response to a grating flashed briefly from a gray background (Fig. 1C). There could be an advantage to presenting these results together at the front of the paper - indeed they seem to show results from the same cell (at least, Fig. 1C and 5A spike traces look identical).

    The main section of the paper shows that natural images presented either in sequence or flashed briefly from a gray background lack nonlinear response behavior - that is, a linear receptive field seems to be encoding the image, because structure in the image can be replaced by the mean luminance across the receptive field center. This makes natural images and gratings seem like they bring out truly unique behaviors in the ganglion cell. But what could be so special about a grating?

    It seems, simply, that the natural image dataset did not match the high contrast grating in the detailed properties of the stimulus, and the proposed nonlinearity in the cone is also obviously important. For example, when the grating is first presented in the contrast-reversing stimulus sequence (i.e, the first half-cycle) or the grating is simply flashed briefly, the On parasol cell is excited by the presynaptic On bipolar cells that see a gray-to-white transition. But the release is relatively weak, because the cones are partly adapted to the gray background. The inhibition driven by 'crossover' circuitry (i.e., driven by the Off pathway) is relatively strong at the gray-to-black transition, and the responses cancel.

    What is less obvious is that for subsequent cycles of the grating, the On bipolar cells that see black-to-white transitions are now releasing very strongly (because the cone response is relatively strong coming from an un-adapted state), which can no longer be canceled by the Off-pathway inhibition. The grating is the optimal stimulus to reveal this nonlinearity because there are simultaneous full-contrast changes in luminance, in opposite directions, at different points across the image. There is no reason why a similar sequence in a natural movie would not drive a similarly strong nonlinear response. It just happens that the natural images and movies (i.e., image sequences) used here apparently never contain these patterns of light changes across the receptive field center.

    So there is a fundamental feature of the circuitry that determines the degree of nonlinear spatial summation in On parasol cells (in an interesting way that differs from Off parasol cells). My main recommendation is that the authors present the findings with gratings in a more compact fashion rather than alternating between gratings and natural images. For example, the results in FIg. 1B and C are puzzling until one learns of the result in Fig. 5A and realizes that Fig. 1B is showing the average across many cycles, which does not capture the unusual response in the first half cycle that is completely consistent with the result in Fig. 1C.

    There is also a clear difference in the contrast of the grating vs. the contrasts within the natural scenes (which are more difficult to define, perhaps). It could be interesting to explore the responses to lower contrast gratings to see if a single model can explain both grating and natural image responses over a range of contrasts in a satisfying way.

  7. Version published to 10.1101/2021.06.29.450295v1 on bioRxiv