1. Neuroscience
Download icon

Spatially displaced excitation contributes to the encoding of interrupted motion by a retinal direction-selective circuit

  1. Jennifer Ding
  2. Albert Chen
  3. Janet Chung
  4. Hector Acaron Ledesma
  5. Mofei Wu
  6. David M Berson
  7. Stephanie E Palmer  Is a corresponding author
  8. Wei Wei  Is a corresponding author
  1. Committee on Neurobiology Graduate Program, The University of Chicago, United States
  2. Department of Neurobiology, The University of Chicago, United States
  3. Department of Organismal Biology, The University of Chicago, United States
  4. Graduate Program in Biophysical Sciences, The University of Chicago, United States
  5. Department of Neuroscience and Carney Institute for Brain Science, Brown University, United States
  6. Grossman Institute for Neuroscience, Quantitative Biology and Human Behavior, The University of Chicago, United States
Research Article
  • Cited 1
  • Views 512
  • Annotations
Cite this article as: eLife 2021;10:e68181 doi: 10.7554/eLife.68181

Abstract

Spatially distributed excitation and inhibition collectively shape a visual neuron’s receptive field (RF) properties. In the direction-selective circuit of the mammalian retina, the role of strong null-direction inhibition of On-Off direction-selective ganglion cells (On-Off DSGCs) on their direction selectivity is well-studied. However, how excitatory inputs influence the On-Off DSGC’s visual response is underexplored. Here, we report that On-Off DSGCs have a spatially displaced glutamatergic receptive field along their horizontal preferred-null motion axes. This displaced receptive field contributes to DSGC null-direction spiking during interrupted motion trajectories. Theoretical analyses indicate that population responses during interrupted motion may help populations of On-Off DSGCs signal the spatial location of moving objects in complex, naturalistic visual environments. Our study highlights that the direction-selective circuit exploits separate sets of mechanisms under different stimulus conditions, and these mechanisms may help encode multiple visual features.

Introduction

How do sensory systems convert sensory inputs into behaviorally relevant neural signals? This question has been extensively investigated in the early visual system, where a neuron’s responses to a set of parameterizable visual stimuli can be systematically probed to reveal a cell’s receptive field (RF) properties in space and time. It has been increasingly appreciated that different visual stimuli can engage different mechanisms to shape the neuronal RF even at the earliest stage of visual processing in the retina. The transformation from the visual input to a retinal ganglion cell’s spiking output is influenced by spatiotemporal patterns of stimuli in both the RF center and surround regions, highlighting the necessity of using diverse, ethologically relevant visual stimuli for delineating RF properties and for ultimately understanding neural coding of the animal’s natural environment (Chiao and Masland, 2003; Demb et al., 1999; Deny et al., 2017; Huang et al., 2019; Olveczky et al., 2003; Takeshita and Gollisch, 2014; Yao et al., 2018).

Direction-selective ganglion cells (DSGCs) in the mammalian retina are well-studied for their motion direction selectivity. A DSGC fires maximally to visual stimuli moving across its RF in its preferred direction and is inhibited from firing by stimuli moving in the opposite, null direction (Barlow and Hill, 1963; Barlow and Levick, 1965Oyster, 1968). The direction-selective spiking is largely attributed to the GABAergic input from the starburst amacrine cell (SAC). SAC dendrites are inherently direction-selective, as they are activated by centrifugal motion, or motion from soma to dendritic tip (Euler et al., 2002). Additionally, only SAC dendrites that extend along the null direction of the DSGC selectively make GABAergic synapses with the DSGC (Briggman et al., 2011; Fried et al., 2002; Lee et al., 2010; Wei et al., 2011; Yonehara et al., 2011). Both the intrinsic properties of the SAC and the ‘antiparallel’ wiring patterns between the SAC and the DSGC are necessary for a strong null-direction inhibition onto the DSGC. The asymmetry of inhibition evoked by motion in the preferred and null directions is important for the DSGC’s direction selectivity (Pei et al., 2015; Taylor and Vaney, 2002). The most well-studied DSGC type, the On-Off DSGC, prefers motion in one of the four cardinal directions (Oyster and Barlow, 1967; Sabbah et al., 2017). They have bistratified dendritic arbors in the On and Off sublamina of the inner plexiform layer (IPL) to extract motion directions of bright and dark signals, respectively (Figure 1A; Famiglietti, 1983; He and Masland, 1998; Kittila and Massey, 1997).

Figure 1 with 1 supplement see all
Drd4-GFP-labeled pDSGCs have spatially asymmetric glutamatergic receptive fields.

(A) Schematic showing types of presynaptic neurons to an On-Off DSGC and the neurotransmitters they use. (B) Top: Example On spiking responses of a pDSGC to four different peripheral spots presented around dendritic span. Bottom: Individual (gray) and mean (black) pDSGC On spike counts evoked by spots presented at different locations (25 cells). (C) Top left: Pairwise comparison of mean spike counts in regions evoking the maximum number of On spikes (Max) and the opposite region (Opp) in the control condition (25 cells). Top right: Polar histogram of Max region locations aligned to the preferred-null motion axis. Radius indicates number of cells. Bottom: Same as in top but experiments performed in DHβE + Gabazine (18 cells). (D) Left: Example On EPSC responses to spots shown in the regions evoking the strongest EPSCs (Max) and the opposite region (Opp) in control (top) and in DHβE + Gabazine (bottom). Middle: Pairwise comparisons of EPSC amplitude and charge transfer to spots presented in the Max region and Opp region (Control: 15 cells; DHβE + Gabazine: 24 cells). Right: Polar histograms of Max region locations determined by EPSC charge transfer aligned to the preferred-null motion axis. Radius indicates number of cells. Summary statistics are mean ± SEM, ***p<0.001.

Figure 1—source data 1

Drd4-GFP-labeled pDSGCs have spatially asymmetric glutamatergic receptive fields.

https://cdn.elifesciences.org/articles/68181/elife-68181-fig1-data1-v2.xlsx

The RF of On-Off DSGCs has been studied with conventional visual stimuli such as stationary and moving spots, bars, and gratings. For motion stimuli that traverse across the entire RF, On-Off DSGC responses remain direction-selective over a broad range of contrast, luminance, speed, and background noise levels (Barlow and Levick, 1965; Chen et al., 2016; Grzywacz and Amthor, 2007; Lipin et al., 2015; Sethuramanujam et al., 2016; Sivyer et al., 2010Wyatt and Daw, 1975). However, motion stimuli restricted to the distal RF subregion on the preferred side (defined as the side from which the preferred-direction moving stimulus approaches the RF) can elicit non-directional firing (He et al., 1999; He and Masland, 1998; Rivlin-Etzion et al., 2011; Trenholm et al., 2011). Based on the responses to moving stimuli presented to different subregions of the DSGC RF, the cell’s RF structure can be viewed as consisting of multiple ‘DS subunits’ and a ‘non-DS zone’ at the edge of the preferred side. However, the neural mechanisms underlying the modular and heterogenous RF subunits of On-Off DSGCs have not been elucidated. Furthermore, the functional significance of this fine RF structure is not clear.

In this study, we investigated the spatial RF structure of the mouse On-Off DSGC subtype that prefers motion in the posterior direction of the visual field (pDSGC). We found that the pDSGC spiking RF is skewed toward the preferred side of the cell for both stationary and moving stimuli, even in the absence of SAC-mediated inhibition. Combining anatomical and functional analyses, we found a spatially non-uniform glutamatergic excitatory conductance that contributes to this spatial displacement. As a result of the displaced RF, moving stimuli that only activate the preferred side of the pDSGC RF trigger robust firing during both preferred and null direction motion. Theoretical analyses of the On-Off DSGC population response allow us to speculate about the ethological relevance of the displaced RF in processing complex natural scenes, suggesting that it can allow for better estimation of object location when a moving object emerges from behind an occluder. We term this type of motion ‘interrupted motion’ to distinguish it from more standard smooth motion stimuli. This phenomenon might also allow synchronous firing from different subtypes of DSGCs to serve as a useful alarm signal in complex scenes.

Results

Glutamatergic excitation of the On-Off pDSGC is spatially asymmetric relative to the soma

To investigate the spatial distribution of excitatory inputs to On-Off pDSGCs, we targeted pDSGCs in the Drd4-GFP transgenic mouse line. Bright stationary spots were presented to the periphery of the DSGC RF and the spiking responses were recorded. These spots were 110 μm in diameter and centered 165 μm from the soma of the recorded DSGC (Figure 1B, schematic). Measurement of the average radius of pDSGC On and Off dendrites (On: 88.1, STDEV 13.9 μm, Off: 75.9, STDEV 13.7 μm, from 25 cells, see Materials and methods) indicates that the spot mostly covered areas beyond the dendritic span of these cells. For each cell tested, we also presented a moving bar stimulus to confirm its directional tuning to posterior-direction motion in the visual field.

The spot stimulus delivered to the pDSGC RF periphery uncovered an asymmetric RF organization where some spots evoked maximal spiking (Max) and other spots presented to the opposite regions (Opp) across the dendritic span evoked minimal responses (Figure 1B–C, Figure 1—figure supplement 1A-B). The Max-Opp axis largely corresponds to the preferred-null motion axis, which is shown by the polar plots in Figure 1 where the preferred side is aligned to the top. The preferred side is the side of the receptive field that a bar moving in the cell’s preferred direction enters first. The majority of the maximum responses to the peripheral spots occurs within 67.5 degrees from the preferred direction of the DSGC and half occur within 22.5 degrees of the preferred direction (Figure 1B–C, ‘Control’, Figure 1—figure supplement 1A).

Because a well-documented asymmetry in the direction-selective circuit is the asymmetric inhibition from SACs to DSGCs, we next tested whether the displacement of the pDSGC spiking RF is eliminated by blocking SAC inputs. We perfused the retina with the nicotinic antagonist DHβE and the GABAA receptor antagonist gabazine to block these inputs. Under this condition, we still observed a spatial asymmetry in pDSGC spiking activity evoked by the flashing spots (Figure 1C, Figure 1—figure supplement 1A-B, ‘DHβE and Gabazine’), indicating a spatial displacement of glutamatergic excitation of the pDSGC in the absence of SAC influence.

To directly measure the strength of excitatory inputs to pDSGCs at different stimulus locations, we performed whole-cell voltage clamp recordings of excitatory postsynaptic currents (EPSCs) evoked by peripheral flashing spots. Consistent with the pattern of pDSGC spiking activity, spot-evoked EPSCs show a spatial bias toward the preferred side (Figure 1D, Figure 1—figure supplement 1C and F, ‘Control’). Isolation of the glutamatergic component of the EPSC by the addition of DHβE and gabazine confirms the persistence of the spatial asymmetry, indicating that the glutamatergic excitation of the pDSGC is not isotropic but is spatially displaced relative to the soma (Figure 1D, Figure 1—figure supplement 1C-G, ‘DHβE and Gabazine’). We noted that gabazine and DHβE reduced the rise and decay time of the EPSC waveform compared to that in the control condition (Figure 1—figure supplement 1E), presumably due to the removal of the cholinergic component of the EPSC. Ablating the GABAergic contribution pharmacologically also rules out the possibility that a strong null-direction GABAergic inhibition is contaminating and artificially reducing the EPSCs on the null side, which is a potential confound during voltage clamp recordings due to imperfect control of membrane potential in distal dendrites (Poleg-Polsky and Diamond, 2011).

Non-uniform excitatory conductance across the preferred-null motion axis contributes to the asymmetric glutamatergic RF

An asymmetric excitatory RF in a retinal neuron may result from asymmetric dendritic arbors and/or an asymmetric distribution of excitatory synaptic inputs across its dendritic field. We performed two-photon imaging of dye-filled dendritic arbors after recording glutamatergic EPSCs evoked by the peripheral flashing spot stimulus described above (Figure 2A–B). Consistent with previous studies, dendritic arbors of pDSGCs do not exhibit a salient or consistent bias relative to the cell’s preferred motion direction. The total dendritic length or the number of dendritic branching points does not significantly differ between the preferred and null sides of the pDSGC dendritic field (Figure 2C–D, Figure 2—figure supplement 1A-C). This apparent randomness of dendritic arbor distribution relative to the cell’s preferred motion direction contrasts with the previously reported mouse On-Off DSGC subtype preferring motion in the superior direction that have dendritic arbors strongly displaced to the null side of the soma (El-Quessny et al., 2020; Kay et al., 2011; Trenholm et al., 2011).

Figure 2 with 1 supplement see all
pDSGC dendritic morphology does not show a spatial bias toward the preferred side.

(A) Traced On layer dendritic morphologies of pDSGCs aligned to their preferred-direction motion. (B) Example On morphology of a pDSGC cell divided into eight sectors for calculating normalized dendritic length vector and pairwise comparisons. (C) Left: Normalized vector sum of On dendritic length aligned to pDSGCs’ preferred direction motion. Right: Pairwise comparison of dendritic length on the preferred vs null sides of each cell (26 cells, p=0.70). Blue represents example cell in B. (D) Left: Normalized vector sum of On branch points aligned to pDSGCs’ preferred direction motion. Right: Pairwise comparison of branch points on the preferred vs null side (26 cells, p=0.68). Blue represents example cell in B. (E) Left: Normalized vector sum plot of On dendritic length in eight sectors aligned to the region evoking maximal glutamatergic EPSC (upward arrow). Middle: Total dendritic length in the region evoking maximal glutamatergic EPSC as determined by peripheral spot stimulus (Max) and opposite region (Opp) (31 cells, ***p<0.001). Right: Total number of branch points in the Max and Opp regions (31 cells, ***p<0.001). (F) On dendritic length RSI versus On EPSC charge transfer RSI (31 cells). RSI is based on regions evoking maximal glutamatergic charge transfer responses and opposite regions. Summary statistics are mean ± SEM.

Figure 2—source data 1

pDSGC dendritic morphology does not show a spatial bias toward the preferred side.

https://cdn.elifesciences.org/articles/68181/elife-68181-fig2-data1-v2.xlsx

Presentations of peripheral spots revealed that pDSGC glutamatergic RFs are consistently skewed toward the preferred side (Figure 1C–D), while pDSGC dendritic fields are not (Figure 2—figure supplement 1A-C). This indicates that the functional glutamatergic RF of a pDSGC cannot be solely explained by its dendritic morphology. However, since pDSGC dendrites define the physical locations of the cell’s excitatory postsynaptic sites, the spatial distribution of dendritic arbors may partially influence the position of the cell’s excitatory RF. Indeed, we found a non-random skew of the pDSGC dendritic length and number of branch points toward the regions corresponding to light spots evoking the strongest glutamatergic EPSCs (Figure 2E). This indicates that the pDSGC’s excitatory RF is partially shaped by its dendritic morphology, rather than being completely independent of the cell’s dendritic distribution.

However, dendritic morphology cannot fully account for the pattern of glutamatergic excitatory RF displacement. When we examined the relationship between EPSC bias and dendritic arbor distribution along the axis of maximal glutamatergic displacement, we did not find a positive correlation between the extent of EPSC bias and the extent of dendritic bias (Figure 2F), suggesting that mechanisms other than dendritic arbor density also contribute to the displacement of the pDSGC glutamatergic RF.

To further explore the relationship between the pDSGC glutamatergic RF and its dendritic distribution, we performed another set of experiments to obtain a more complete RF map. First, the preferred direction of each cell’s spiking activity was determined by loose cell-attached recordings using a moving bar stimulus with no synaptic blockers. Next, peripheral flashing spots were used to estimate the spatial displacement of the pDSGC glutamatergic RF in the presence of nicotinic and GABAergic receptor antagonists (DHβE + Gabazine) as described above. Then, a smaller, 20 μm diameter bright stationary spot was repeatedly flashed at random locations within a 11-by-11 220 μm square grid centered on the pDSGC soma. A heatmap of glutamatergic EPSC charge transfer evoked by the small flashing spot was generated and overlaid with the reconstructed dendritic arbors for each pDSGC (Figure 3A).

Figure 3 with 1 supplement see all
Non-uniform glutamatergic synaptic excitation across pDSGC dendritic span contributes to skewed excitatory receptive field.

(A) Left: Schematic for small spot RF mapping experiment in DHβE + Gabazine. Middle: Example heat map. Right: Preferred motion direction of the example cell in relation to the preferred side and the null side of the cell’s RF. (B) Vector sum plot of On EPSC charge transfer center of mass determined by small spot RF mapping aligned to the region where a peripheral spot evokes the maximum glutamatergic EPSC (upward arrow). (C) Spatial locations showing centers of mass of glutamatergic excitatory On charge transfer obtained from experiments illustrated in A aligned to each cell’s preferred motion direction (15 cells). (D) Example On EPSC responses to spots presented to a pDSGC along the maximum-opposite axis of glutamatergic RF displacement. (E) Left: On ‘EPSC current density’ (i.e. ratio of charge transfer per dendritic length) versus distance from soma along the maximum-opposite axis of glutamatergic RF displacement (16 cells, ***p<0.001) as well as along the orthogonal axis. Right: Summary plot of glutamatergic charge transfer as a function of total dendritic length centered around the flashing spot for spots shown more than 50 μm away from the soma (16 cells, ***p<0.001). (F) Same as E but along the preferred-null motion axis (15 cells, charge transfer/dendritic length vs distance from soma **p=0.0026, charge transfer vs dendritic length ***p<0.001). Summary statistics are mean ± SEM.

Figure 3—source data 1

Non-uniform glutamatergic synaptic excitation across pDSGC dendritic span contributes to skewed excitatory receptive field.

https://cdn.elifesciences.org/articles/68181/elife-68181-fig3-data1-v2.xlsx

RF mapping with small flashing spots revealed a spatial displacement of glutamatergic EPSC distribution that aligns well with the displacement determined by larger spots presented to the pDSGC periphery (Figure 3B). Similar to the displacement pattern revealed by stimulation of the RF periphery, the strongest glutamatergic EPSCs are preferentially located on the preferred side of pDSGC somas (Figure 3C). Notably, for most cells, the center of the glutamatergic EPSC RF is displaced from the center of the dendritic field (Figure 3—figure supplement 1A-B), indicating additional mechanisms underlying the glutamatergic RF displacement apart from the dendritic arbor distribution.

We further examined the strength of the glutamatergic input across the pDSGC dendritic field (Figure 3D). We normalized the EPSC charge transfer by the total dendritic length in a circle with a 60 μm diameter centered on each small flashing spot. This provides an estimate of the strength of glutamatergic inputs per unit dendritic length, or ‘EPSC density’, at each stimulus location. When we calculated the density along the axis of maximal glutamatergic RF displacement, we found that the EPSC density on the displaced side of the dendritic field is larger than the corresponding region on the opposite side (Figure 3E, left, Figure 3—figure supplement 1E). From 50 μm away from the soma, spots on the side corresponding to the maximum displacement (Maximum) yield stronger EPSCs than spots on the opposite side (Opposite), while controlling for the same dendritic length (Figure 3E, right, Figure 3—figure supplement 1E). This heterogeneity in EPSC density across the dendritic span persists when comparing EPSC responses along the preferred-direction motion axis of the cell (Figure 3F, Figure 3—figure supplement 1F). Analysis of the peak amplitudes showed similar results as charge transfer (Figure 3—figure supplement 1C-D).

Next, we asked what mechanisms could underlie the non-uniform excitatory conductance across the dendritic span. Because bipolar cells provide the dominant excitatory input to DSGCs, the enhanced excitation on the preferred side of the receptive field could stem from increased density or strength of bipolar synapses. Using a published serial block-face scanning electron microscopic (SBEM) dataset of the adult mouse retina (Ding et al., 2016), we reconstructed large numbers of On starburst amacrine cells and On-Off DSGCs (Figure 4A). The preferred direction of the traced DSGCs was inferred from the mean orientation of starburst dendrites making ‘wraparound’ synapses onto the DSGCs, following the previous work of Briggman et al., 2011 (Figure 4C). Among all DSGCs, four preferred directions were seen roughly 90 degrees apart, as expected for most retinal locations (Sabbah et al., 2017). We selected three On-Off DSGCs for which the inferred preferred direction of motion fell on the horizontal axis like the cells we studied electrophysiologically, two with one motion preference, and the other with the opposite preference. Technical constraints prevented us from saying which cells preferred posterior motion and which cells preferred anterior motion (see Materials and methods). We then mapped the distribution of bipolar ribbon synapses onto the dendrites of these three cells. We found no marked gradient in ribbon density across the preferred-null motion axis of the three cells (Figure 4D–F, Figure 4—figure supplement 1A-B), suggesting that the gradient in EPSC density across this axis is not determined by the density of bipolar inputs.

Figure 4 with 1 supplement see all
Neither the density of bipolar ribbon inputs nor the kinetics of glutamatergic EPSCs varies systematically along the axis of RF displacement.

(A) Top: Distribution of ribbon synaptic input to one of the three reconstructed On-Off DSGCs. (B) Ultrastructure of the synapse indicated by the red arrow in A, as visualized by serial block-face electron microscopy. The presynaptic bipolar cell (green) belongs to Type 5t. Red arrow marks the ribbon. The magenta profile belongs to the On-Off DSGC. The other postsynaptic partner at this dyad synapse was another ganglion cell (yellow). Although only a fragment was included in the volume, it was presumably a DSGC, since it costratified entirely with the inner dendrites of the On-Off DSGC (A; black dendrites). (C) SAC inputs onto example On-Off DSGC. Red arrows indicate the direction and location of the SAC inputs on the DSGC dendrites. (D) Density map of bipolar ribbon synapses for three example On-Off DSGCs with estimated preferred directions along the posterior-anterior axis (see Methods). The soma location is indicated by the white spot in the center. (E) Bipolar ribbon heat map for the three example cells, respectively. The number of ribbons in each square are indicated. (F) Quantification of ribbon density across the preferred-null axis, respectively. The soma location is at 0. (N.S. = null side, P.S. = preferred side). (G) Left: Summary of latency of glutamatergic EPSC responses along the maximum-opposite axis (Max = maximum glutamatergic EPSC region, Opp = opposite region) (16 cells, p=0.75). Right: Summary of latency along the preferred-null motion axis (N.S. = null side, P.S. = preferred side)(15 cells, p=0.88). (H) Same as in G, except for rise time (10%–90%). (Left: p=0.12, right: p=0.21). (I) Same as in G, except for decay time (90%–30%). (Left: p=0.25, right: p=0.26). Summary statistics are mean ± SEM.

Figure 4—source data 1

Neither the density of bipolar ribbon inputs nor the kinetics of glutamatergic EPSCs varies systematically along the axis of RF displacement.

https://cdn.elifesciences.org/articles/68181/elife-68181-fig4-data1-v2.xlsx

In addition to ribbon density, we also examined the time course of glutamatergic EPSCs in the presence of DHβE and gabazine along the axis of maximal glutamatergic RF displacement and along the preferred-null motion axis. We found no significant differences in the latency, rise and decay times in EPSC waveforms across the axis of maximal glutamatergic RF displacement in response to the small flashing spots (Figure 4G–I, left panels). In addition, we did not detect a monotonic change in glutamatergic EPSC kinetics along the preferred-null motion axis in DHβE and gabazine (Figure 4G–I, right panels) or in DHβE only (Figure 4—figure supplement 1C-D). That we found no differences in the kinetics of EPSC waveforms across the pDSGC dendritic field contrasts with a previous study of On DSGCs reporting a gradient of EPSC kinetics from slow/sustained to fast/transient along the preferred-null motion axis (Matsumoto et al., 2019). This gradient in On DSGC EPSC kinetics is thought to arise from different bipolar cell subtypes, and may implement a Hassenstein-Reichardt-Detector-like mechanism for the On DSGCs’ direction selectivity. Thus, our data suggest that the kinetics of bipolar cell signals onto On-Off pDSGCs do not differ substantially along the preferred-null motion axis.

Null-direction response emerges during partial activation of the displaced pDSGC RF

Because On-Off DSGCs are thought to be dedicated to encoding object motion, we next asked how the displaced excitatory RF of the pDSGC revealed by the stationary flashing stimuli contributes to motion processing. First, we considered whether a displaced excitatory RF could benefit the direction-selective mechanisms during full-field smooth motion. By comparing the onset times of EPSCs and IPSCs to preferred-direction motion, we found that cells with more spatially separated excitatory and inhibitory receptive fields were more direction-selective (Figure 7—figure supplement 1L).

When motion trajectories are more complex, as often occurs in the natural environment, the displaced excitatory pDSGC RF may confer additional characteristics to the cell’s motion encoding. We reasoned that because full-field smooth motion stimuli traverse the entire RF of the pDSGC, SAC-mediated inhibition exerts a dominant influence on pDSGC spiking by strongly suppressing the pDSGC null-direction response. However, when moving objects pass behind occluders or start moving from behind other objects, the RFs of certain DSGCs are partially activated, changing the interaction between SAC-mediated inhibition and displaced excitation. Therefore, we next investigated how the direction-selective circuit would process interrupted motion.

To examine if the displaced excitatory RF of the pDSGC plays a role in encoding interrupted motion, we created an occluded motion stimulus where a moving bar disappeared behind a central occluder 220 μm in diameter (Figure 5A). The occluder covered the dendritic span of the pDSGC and a substantial portion of its RF center. In contrast to the full-field motion stimulus, the occluded motion stimulus caused both preferred and null-direction spiking responses (Figure 5B, Figure 5—figure supplement 1A-C). These responses are only evoked when the bar travels in the preferred side, which corresponds to the displaced side of the pDSGC RF (Figure 5B, middle row, and 5C). In contrast, no spiking response is evoked when the bar moves across to the null side beyond the occluder (Figure 5B, bottom row, Figure 5C). To test whether this regional difference is due to the asymmetric wiring between SACs and DSGCs, we blocked the cholinergic and GABAergic transmission with DHβE and gabazine and saw that the regional difference persisted (Figure 5C, Figure 5—figure supplement 1B). Consistent with the spiking pattern, the EPSC responses also reflected this regional asymmetry (Figure 5D). In contrast, IPSC responses to the occluded bar stimulus are displaced to the opposite side compared to EPSC responses (Figure 5—figure supplement 1D), consistent with the asymmetric wiring pattern from SACs from the null side to the DSGC (Briggman et al., 2011; Fried et al., 2002; Lee et al., 2010; Wei et al., 2011; Yonehara et al., 2011).

Figure 5 with 2 supplements see all
Displaced excitatory receptive field contributes to null-direction responses in the preferred region.

(A) Full-field moving bar and occluded bar stimuli. (B) Example On spiking responses of a pDSGC to full-field bar (top) and occluded bar stimulus (middle, bottom) moving in the preferred (red) and null (black) directions. (C) Top: Mean spike counts of pDSGCs to the full-field moving bar (full-field) and the occluded moving bar on the preferred side and null side (48 cells). Bottom: Mean spike counts in DHβE + Gabazine to the occluded bar stimulus in the region evoking the maximum spiking (Max) and the opposite region (Opp) (14 cells). (D) Mean pDSGC EPSC peak amplitude and charge transfer in the region evoking the maximum response (Max) and the opposite region (Opp) in the control (top, 19 cells) and DHβE + Gabazine (bottom, 13 cells) conditions. (In DHβE + Gabazine: Pref Dir – Max vs Pref Dir – Opp amplitude **p=0.0028, Pref Dir – Max vs Pref Dir – Opp charge transfer **p=0.0012). (E) Mean On and Off spiking responses to occluded bar stimuli at different contrast configurations. Top left: On: 11 cells, Max null dir. Vs Opp null dir. **p=0.008. Off: 10 cells. Bottom left: On: 10 cells. Off: 8 cells. Top right: On: 9 cells, Max pref. dir. Vs. Opp pref. dir. **p=0.003. Off: 8 cells, Max pref. dir. Vs Opp. Pref. dir. *p=0.034, Max null dir. Vs Opp null dir. *p=0.02. Bottom right: On: 8 cells, Max null dir vs Opp null dir. *p=0.014. Off: 9 cells, Max pref. dir. Vs Opp pref. dir. **p=0.004, Max null dir. Vs Opp null dir. **p=0.002. (F) Top: Example directional tuning curve of GCaMP6 signal of a pDSGC. Bottom: Example tuning curve of an aDSGC. (G) Top left: Example GCaMP6 fluorescence traces of a pDSGC for the full-field moving bar in the preferred (red) and null (black) directions. Bottom left: Example GCaMP6 traces of the cell for the occluded bar stimulus. Shaded areas represent SEM. Top right: Normalized amplitude of the posterior-preferring cell null-direction response during the full-field bar and the occluded bar (12 cells). Bottom right: Normalized area of the null-direction response during the full-field bar and the occluded bar (12 cells, **p=0.0035). (H) Same as in G, except for aDSGCs (12 cells, normalized amplitude *p=0.013, normalized area *p=0.019). Summary statistics are mean ± SEM, ***p<0.001 except where specified otherwise.

Figure 5—source data 1

Displaced excitatory receptive field contributes to null-direction responses in the preferred region.

https://cdn.elifesciences.org/articles/68181/elife-68181-fig5-data1-v2.xlsx

Moreover, we individually blocked each of several types of signaling, including GABAC receptor activity, glycine receptor activity, and gap-junction coupling with TPMPA, strychnine, MFA, and Carbenoxolone, respectively, and found that the spatial asymmetry of the spike response remains during the occluded motion stimulus (Figure 5—figure supplement 1H-K). Based on these pharmacology results, we conclude that the asymmetric glutamatergic RF of the pDSGC contributes to robust spiking in both null and preferred directions when the preferred side of the RF is activated by the occluded motion stimulus.

Since there are two On-Off DSGC subtypes that are tuned to opposite directions along the posterior-anterior axis, we investigated if the anterior-direction-selective DSGC (aDSGC) exhibits the same null-direction response pattern to the occluded motion stimulus as the pDSGC. To identify both DSGC subtypes, we performed calcium imaging of GCaMP6-expressing RGCs in another transgenic mouse line carrying Vglut2-IRES-Cre and floxed GCaMP6f (Figure 5—figure supplement 2A) during the full-field moving bar stimulus. We then used an online analysis to identify aDSGCs and pDSGCs based on the directional tuning of their calcium signals to the anterior and posterior motion directions (Figure 5F, Figure 5—figure supplement 2B-C). Next, we centered the occluded motion stimulus on individual aDSGC and pDSGC somas and performed calcium imaging during the occluded motion stimulus. Consistent with pDSGC spiking activity, a significant null-direction calcium response of the pDSGCs was evoked when the occluded moving bar traveled across the receptive field to the preferred side of the RF (Figure 5G, Figure 5—figure supplement 2D). Notably, aDSGCs also exhibited a null-direction response to the occluded motion stimulus that is similar to that of pDSGCs (Figure 5H, Figure 5—figure supplement 2D), indicating that aDSGC RFs are also displaced to the preferred side.

The above results show that continuous motion interrupted by a stationary occluder in the center of the pDSGC’s RF causes unexpected null-direction spiking as the bar emerges from behind the occluder into the preferred side of the pDSGC RF. However, this occluded motion stimulus cannot distinguish whether the interruption itself or only the start position of the emerging bar was necessary for the null-direction response. Therefore, we created a stimulus where a bar emerges at different locations along the preferred-null motion axis of the cell. The start motion of the bar activates different parts of the DSGC’s receptive field (Figure 6A). For the null-direction moving bar, there is an increase in both the spike number and the firing rate as the starting position of the moving bar is located past the soma on the preferred side of the pDSGC RF (Figure 6A–C). Thus, the emergent growing edge caused null-direction spiking of pDSGCs in a similar pattern as the moving bar emerging behind the central occluder. These results illustrate that the null-direction response of the pDSGC during the occluded motion stimulus is dependent on the position in the receptive field from which the moving edge emerges, not the previous motion approaching the occluder.

Null-direction response is dependent on start position of emerging bar.

(A) Schematic of moving bars emerging from different locations along the DSGC’s preferred-null motion axis and example spiking responses. The soma location is at 0. Vertical dashed lines on the schematic indicate the positions of the emerging leading edge of the moving bar. (B) Mean spike counts (null direction: −330 vs 55 *p=0.012, –330 vs 110 **p=0.0088) and (C) firing rates (null direction: −330 vs 0 *p=0.014) to bars emerging from different locations along the preferred-null motion axis (10 cells). Summary statistics are mean ± SEM, ***p<0.001 except where specified otherwise.

Figure 6—source data 1

Null-direction response is dependent on start position of emerging bar.

https://cdn.elifesciences.org/articles/68181/elife-68181-fig6-data1-v2.xlsx

Null-direction responses of DSGCs during partial activation of their RFs can be useful for decoding object location

We explored how the displaced RFs could functionally benefit the encoding of interrupted motion by DSGCs. Our experimental results show that pDSGCs generate robust spiking activity in both the preferred and null directions when a moving object appears and starts moving on the preferred side of the cell’s RF, as when an object emerges from behind an occluder positioned over the cell’s soma. We asked whether this prima facie aberrant signaling can have a functional role relevant to the behavioral goals of the organism.

We hypothesize that partial, non-directional RF activation of a DSGC may provide precise information about the spatial position of motion initiation. In particular, if a moving object emerging from behind an occluder activates the preferred side of an On-Off DSGC’s RF, this DSGC would generate a null-direction spiking response together with the preferred-direction response of a nearby On-Off DSGC subtype preferring the opposite motion direction (Figure 7A). Such synchronous spiking activity between DSGCs of opposite preferred directions could yield a stronger and more localized spatial signal at the population level at the location of motion interruption.

Figure 7 with 1 supplement see all
Null-direction spiking of On-Off DSGCs during the occluded motion stimulus improves position estimation.

(A) Schematic of the DSGC population model along with example spike trains for two pDSGCs (red trace showing their preferred-direction responses) and an aDSGC (black trace showing its null-direction response). An occluder 220 µm wide (shaded circle) was placed near the middle of the population. (B) Example root-mean-square error in estimating the position of the light bar’s leading edge for full-field (black) and occluded (blue) motion across different bar speeds. Shaded area indicates the time window during which the moving edge is behind the occluder. (C) Same as in B, but for direction estimation error. (D) Left: Position error percent decrease across speeds. Right: Absolute error in visual angle decrease across bar speeds. (E) Same as in D, but for direction estimation error. Right: Absolute error in motion direction angle. (F) Temporal filter for detecting occlusion response. (G) Convolution of firing rate from the temporal filter. Dashed line represents example threshold. (H) ROC curve with different levels of baseline firing for a bar of speed 330 μm/s. Legend shows performance level of decoder for each level of baseline firing.

To test how the null-direction response during interrupted motion might enhance downstream estimation of stimulus position, we simulated the spiking responses of pDSGCs in a population model consisting of On-Off DSGC subtypes that prefer opposite motion directions on the posterior-anterior axis (aDSGCs and pDSGCs, Figure 7A). One thousand cells of each subtype were arranged in a two-dimensional array with biologically realistic positional jitter, such that the spatial positions of the two subtypes were uncorrelated. We simulated a bright edge moving along the posterior-anterior axis in a single direction at a constant velocity under both uninterrupted and interrupted motion conditions and analyzed the On response to the edge. The mean spiking response of each DSGC was modeled as a rectified sine wave with parameters obtained from our experimental data, and sub-Poisson trial-to-trial variability was introduced to the mean spiking response on each trial (Figure 7—figure supplement 1). The spatial positions of the DSGCs were shuffled in each simulation block to allow us to sample different spatial arrangements of the RFs.

We used a labeled-line decoder (equivalent to a ‘population vector’ decoder) that estimated the spatial position of the moving bar edge as a weighted average of the RF center positions where the weights were determined by the firing rate (response strength) and RF width (response precision) of each cell (see Materials and methods). Using this decoder, we compared the scenario in which none of the DSGCs were occluded (full-field) with the scenario in which a single occluder was placed near the center of the population. We chose to implement the model with different low baseline firing rates to more faithfully represent biological noise. Previous reports show baseline firing rates up to 0.1 spike/s (Yao et al., 2018) for On-Off DSGCs. Our experiments yielded background noise levels more on the scale of 0–0.025 spike/s. Thus, we evaluated the computational model at three different low noise levels (Figure 7D–E).

Our modeling results show that when the moving bar emerges from behind the occluder, null-direction responses from the pDSGCs undergoing partial RF activation on the preferred side transiently degrades the population estimate of the motion direction as would be expected (Figure 7C and E). However, the coincidence of the null-direction responses from pDSGCs and the preferred-direction response of neighboring aDSGCs when the bar exits the occlusion substantially improves the estimation of bar position at this time (Figure 7B and D). For a bar traveling at the speed of 330 μm/s (the lowest speed in our experiments and simulations), the synchronous firing between pDSGCs and the neighboring aDSGCs reduces the error in the population estimate of the bar edge’s spatial position by over 80 percent. This reduction in error is present, albeit smaller, even at higher speeds. At the highest speed of 2640 μm/s, there was still around a 40% decrease in position estimation error during the occlusion trials. The absolute position error decrease was around 3 degrees of visual angle, or around half the receptive field size of an On-Off DSGC, across bar speeds in models with low levels of background noise. Therefore, at the site of motion interruption, the DSGC population response transiently prioritizes the encoding of the emerging moving object position over its motion direction.

Because the displaced excitatory receptive field also induced differential firing responses to stationary spots presented outside of the cell’s dendritic field (Figure 1), we asked whether the displacement was useful for detecting non-directional contrast changes for a small stationary spot. Receptive field mapping experiments showed that the spiking receptive field is displaced toward the preferred side. However, the total diameter of the asymmetric receptive field is around 220 μm, which is larger than the average dendritic span of pDSGCs (Figure 1—figure supplement 1H). We analyzed whether a displaced receptive field would benefit the position estimation using non-directional contrast change signals. We found that the accuracy of estimating position from contrast changes is the same for populations with asymmetric receptive fields and populations with symmetric receptive fields. The detection of the contrast change would only improve if the receptive field size decreased (data not shown). Directed motion is required for an improvement in position estimation because a null-direction signal can only occur in a small region on the preferred side. Therefore, the synchrony of the null-direction and preferred-direction signals within the population yields a more spatially constrained signal necessary for fine spatial discrimination of moving stimuli.

The synchronous firing may, itself, also be a useful alarm signal that triggers processing downstream of the retina, independent of the precise position information it might additionally convey (Ishikane et al., 2005). During smooth motion, only one subtype of DS cell would respond to a bar moving across the visual scene. However, synchronous firing of two oppositely tuned DS cells would occur to represent interrupted or emergent motion. The synchronous firing of two DSGC subtypes can be a unique signature of encoding interrupted motion, which can differ from the encoding of non-motion contrast signals where the recruitment of all four DSGC subtypes would be expected. To investigate whether the synchronous firing during occlusion trials can be read out from a population response, we utilized a coincidence decoder to determine whether full-field motion trials can be distinguished from occluded motion trials. A filter was fit to the null-direction spiking response from the occlusion trials (Figure 7F), and when the convolution of the firing rate with the filter rose above a threshold level (Figure 7G), the detector would identify an occlusion trial.

Evaluating the decoder performance showed that the decoder is highly successful at identifying occlusion trials in conditions with low background firing rates. In conditions with background firing rates of 10 spikes/s or less, the decoder correctly identified occlusion trials with more than 99% accuracy (Figure 7H). This coincidence detection model suggests that the synchrony of preferred-direction responses from one DSGC subtype and the null-direction responses from the opposite DSGC subtype during interrupted motion conditions can be easily detected. This salient synchrony signal can potentially inform downstream visual areas of the type of motion that is occurring or provide an alarm signal to the animal that there is an unexpected change in their environment.

Discussion

Our study reveals a new form of asymmetry in the direction-selective circuit: a spatial displacement of glutamatergic inputs to the preferred side of the On-Off pDSGCs due to a non-uniform distribution of synaptic conductances across the pDSGC dendritic span. The impact of this displaced excitation on DSGC spiking is demonstrated by using moving stimuli with interrupted trajectories, a feature abundant in natural scenes. In contrast to full-field continuous motion which maximizes the contribution of SAC-mediated null-direction inhibition, occluded motion stimuli reduce the contribution of SAC-mediated inputs to allow the excitatory receptive field to dominate the spiking response in a non-directional manner when only part of the RF is stimulated. Therefore, an On-Off DSGC’s response to occluded motion stimuli is determined both by how much of the receptive field is activated and where that activation occurs.

The non-isotropic excitation of the pDSGC alludes to a more sophisticated set of signaling mechanisms from the bipolar cell population to pDSGCs. A detailed explanation of the glutamatergic RF displacement awaits future studies. The gradient of glutamatergic current density across the pDSGC dendritic span may arise from a number of possible scenarios including varying strengths of individual glutamatergic synapses, heterogeneous membrane properties across pDSGC dendritic field, and contribution from Vglut3 +amacrine cells (Franke et al., 2017; Kim et al., 2015; Lee et al., 2016; ).

It is intriguing to speculate how the displaced excitatory receptive field properties of On-Off DSGCs can influence downstream computation. Previous theoretical analyses have addressed the encoding of motion direction by DSGCs at the population level (Fiscella et al., 2015; Zylberberg et al., 2016). In this study, we explored a hypothesis that On-Off DSGC population activity contains information about both the direction and the location of a moving object. When motion trajectories are not continuous, the null-direction responses from cells near the occlusion edge transiently improves the encoding of location at the expense of direction encoding. We speculate that the trade-off between positional and directional encoding, which occurs when an object emerges from behind another object, may reflect the animal’s greater need for positional information of the emerging object than the direction in which it is moving. Additionally, the synchronous response of null-direction and preferred-direction spiking can potentially provide a salient alarm signal for discontinuous motion, which may help the animal quickly attend to the site of the change. Considering the population activity of multiple subtypes of On-Off DSGCs after an interruption in the motion allows for the encoding of more information than motion direction. Our findings that the population activity across cell types can help resolve ambiguities in single-cell responses share a common theme with previous modeling experiments from Kühn and Gollisch, 2019, which show that multiple DSGC subtypes with different motion direction tuning in the salamander retina are needed to isolate motion-related information from confounding contrast signals under complex texture motion. Our current investigation adds to the accumulating evidence that retinal population activity across multiple subtypes enhances decoding of visual features from ambiguous, multiplexed signals.

Elements of the visual processing scheme implicated in our study parallel those in other visual areas and species. For example, specific stimulus patterns, such as occluded motion in our study and the ‘reverse-phi’ illusory motion with alternating contrast polarities studied in flies, trigger null-direction responses in direction-selective neurons to serve context-specific encoding tasks that may be beneficial for extracting visual information from more complex natural scenes (Salazar-Gatzimas et al., 2018; Agrochao et al., 2020). As another example, both previous studies in flies and our study indicate that local motion detection can be carried out by neuronal populations with wider receptive fields (Fisher et al., 2015). In our model, partial RF activation on the preferred side of the pDSGC generates a non-directional, local contrast response that may contribute to a salient population signal from the retina to the brain to alert the animal about emerging motion. Interestingly, local contrast response properties of direction-selective neurons in the fly visual system have been shown to profoundly modulate their motion computations and contribute to visually guided behavior (Clark et al., 2014; Drews et al., 2020; Matulis et al., 2020).

In contrast to the mammalian On DSGC that encodes global motion during optic flow and participates in the optokinetic reflex, On-Off DSGCs are considered encoders of local motion, and project to the superficial layer of the superior colliculus (SC) and the shell region of the dorsal lateral geniculate nucleus (dLGN) (Cruz-Martín et al., 2014; Huberman et al., 2009Kay et al., 2011Rivlin-Etzion et al., 2011). In the SC, On-Off DSGC inputs give rise to the direction selectivity of postsynaptic collicular neurons (Shi et al., 2017), indicating that these collicular neurons do not receive retinal inputs from a broad range of RGC types, but specifically from On-Off DSGCs. Since the superficial layer of the SC is well recognized for its roles in encoding spatial locations and instructing stimulus-directed defensive and prey behaviors (Basso et al., 2021Ito and Feldheim, 2018), the encoding of the spatial location of an emerging moving object by On-Off DSGCs may benefit rapid sensorimotor decisions that involve collicular circuitry.

It is likely that other ganglion cell types also participate in fine spatial discrimination. For example, certain types of small receptive field RGCs such as W3 RGCs and HD-RGCs may also be well-suited to encode object location (Jacoby and Schwartz, 2017; Kim et al., 2010; Zhang et al., 2012). It is worth noting that W3 RGCs are activated only in specific instances where the background is completely uniform, whereas On-Off DSGCs can be activated in a wide range of visual environments, including environments with noisy backgrounds (Chen et al., 2016; Chen et al., 2020). HD-RGCs also have small receptive fields, and computational modeling experiments have shown that the errors in object location between On-Off DSGCs after motion interruption and HD-RGCs are very similar in scale (Jacoby and Schwartz, 2017). However, given the divergent and type-specific central projection patterns of mouse ganglion cell types (Dhande et al., 2015; Ellis et al., 2016Martersteck et al., 2017), the position information encoded by other position encoders such as W3 and HD-RGCs may not be available to the specific downstream circuits that receive On-Off DSGC inputs. Our modeling study suggests that the population response of On-Off DSGCs after a motion interruption helps On-Off RGCs achieve the same performance as other ganglion cell populations implicated in fine spatial discrimination.

Given the many outstanding questions on the retinorecipient circuitry, On-Off DSGCs may participate in different visual processing tasks compared to other ganglion cell types by projecting to different areas for implementing different responses (Kay et al., 2011; Sanes and Masland, 2015). Or, they could provide complementary information about the spatial location of moving objects when considered with other RGC populations. In either scenario, our theoretical analysis indicates that the information of spatial location is contained within the On-Off DSGC population response, and that it is possible that higher visual centers stand to benefit from this information. Ultimately, elucidating the roles of diverse RGC types in motion encoding requires a thorough understanding of visual signal transformations along the processing pathways and how the visual system instructs visually guided behavior.

Materials and methods

Key resources table
Reagent type
(species) or
resource
DesignationSource or
reference
IdentifiersAdditional
information
Gene (M. musculus)129S6-Chattm2(cre)Lowl/JThe Jackson LaboratoryRRID:IMSR_JAX:006410
Gene (M. musculus)129S6-Gt(ROSA)26Sortm9(CAG-tdTomato)Hze/JThe Jackson LaboratoryRRID:IMSR_JAX:007909
Gene (M. musculus)Slc17a6tm2(cre)Lowl/JThe Jackson LaboratoryRRID:IMSR_JAX:012898
Gene (M. musculus)B6J.Cg-Gt(ROSA)26Sortm95.1(CAG-GCaMP6f)Hze/MwarJThe Jackson LaboratoryRRID:IMSR_JAX:028865
Chemical compound, drugDihydro-b-erythroidine hydrobromideTocrisCat#2349
Chemical compound, drugSR 9551 hydrobromideTocrisCat#1262

Animals

129S6-Chattm2(cre)Lowl/J mice (Stock No: 006410) and 129S6-Gt(ROSA)26Sortm9(CAG-tdTomato)Hze/J mice (Stock No: 007909) were acquired from the Jackson Laboratory. Drd4GFP mice were originally developed by MMRRC (http://www.mmrrc.org/strains/231/0231.html) and were backcrossed to the C57BL/6 background. These lines were crossed so that both pDSGCs and SACs were labeled. Slc17a6tm2(cre)Lowl/J mice (Stock No: 016963) and B6J.Cg-Gt(ROSA)26Sortm95.1(CAG-GCaMP6f)Hze/MwarJ mice (Stock No: 028865) were acquired from Jackson Laboratory and were crossed and used for calcium imaging experiments. Mice of ages P21 – P60 of either sex were used. All procedures regarding the use of mice were in accordance with the University of Chicago Institutional Animal Care and Use Committee (ACUP# 72247), Institutional Biosafety Committee (IBC# 1214), and with the NIH Guide for the Care and Use of Laboratory Animals and the Public Health Service Policy.

Whole-mount retina preparation

Request a detailed protocol

Mice were dark adapted for 1 hr and then were anesthetized with isoflurane and euthanized by decapitation. Retinas were isolated at room temperature in oxygenated Ames’ medium (Sigma-Aldrich, St. Louis, MO) under infrared illumination. The retinas were separated into dorsal and ventral halves and were mounted ganglion-cell-layer up on top of a ~ 1.5 mm2 hole in a small piece of filter paper (Millipore, Billerica, MA). The orientation of the posterior, anterior, inferior, and superior directions were noted for each piece. During the experimental day, the mounted retinas were kept in darkness at room temperature in Ames’ medium bubbled with 95% O2/5% CO2 until use (0–8 hr).

Visual stimulation

Request a detailed protocol

A white organic light-emitting display (OLEDXL, eMagin, Bellevue, WA; 800 × 600 pixel resolution, 60 Hz refresh rate) was controlled by an Intel Core Duo computer with a Windows seven operating system and was presented to the retina at a resolution of 1.1 μm/pixel. Visual stimuli were generated using MATLAB and the Psychophysics Toolbox, and were projected through the condenser lens of the two-photon microscope focused on the photoreceptor layer.

For the peripheral spot experiments, stationary spots of size 110 μm diameter were presented 165 μm away from the soma of the recorded pDGSC for 2 s. At this distance, these spots were presented outside of the dendritic span of the majority of recorded pDSGCs. These spots were presented in eight different locations pseudorandomized around the recorded pDSGC, with three repetitions each.

For moving bar experiments, a bright moving bar with dimensions 110 μm (width) X 880 μm (length) and speed 330 μm/s moved in eight different pseudorandomized directions across the receptive field of the pDSGC, with three to six repetitions each. The occluded bar stimulus contains a moving bar of the same dimensions which moves into and out of a central occluder with a diameter size of 220 μm. The size of the occluder was large enough to obscure the entire dendritic span for the majority of recorded pDSGCs. Occluded bar stimuli of different contrasts were used. The population vector model in Figure 7 required the use of bar and occluded bar stimuli at speeds of 330 μm/s, 660 μm/s, 1320 μm/s, 1980 μm/s, and 2640 μm/s.

For the emergent moving bar experiments where the bar emerged from different locations of the receptive field, a bright moving bar with dimensions 110 μm (width) X 440 μm (length) and speed 330 μm/s could start moving pseudorandomly from 13 different locations across the receptive field (Figure 5). The bar could move in the preferred or null direction of the recorded cell.

For the receptive field mapping experiments measuring spiking activity, bright spots with a 60 μm diameter were presented across the receptive field of the recorded pDSGC. They were shown for three to four repetitions with a duration of 400 ms in a random sequence within an 11 × 11 grid, covering a total area of 660 × 660 μm2. For the receptive field mapping experiments measuring EPSC activity, bright spots with a 20 μm diameter were presented across the pDSGC and were shown for three to four repetitions with a duration of 400 ms in a random sequence within an 11 × 11 grid, covering a total area of 220 × 220 μm2. To examine the kinetics of EPSCs, bright spots with either a 20 μm diameter (for DHβE and gabazine experiments) or a 60 μm diameter (for DHβE experiments) were shown across the receptive field in a random sequence within an 11 × 11 grid for three to four repetitions. Spots were shown with a duration of 400 ms.

The light intensity for bright peripheral spot experiments, the moving bar and occlusion experiments, and the receptive field mapping experiments was ~1.6×104 R*/rod/s.

Targeting cells for electrophysiology

Request a detailed protocol

Cells were visualized with infrared light (>900 nm) and an IR-sensitive video camera (Watec). DSGCs were targeted with the aid of two-photon microscopy in Drd4-GFP mice. Cell identity was confirmed physiologically by extracellular recordings of responses to moving bars and/or by filling the cell with 25 μM Alexa 594 (Life Technologies) to show bistratified dendritic morphology. Because the tissue was aligned on the filter paper, we confirmed that the preferred direction of the spiking responses to moving bar aligned well with the anatomical posterior (visual field coordinate) or nasal (retinal coordinate) direction.

Quantification of dendritic morphology

Request a detailed protocol

The On and Off layers of each cell was isolated from a z-stacked image in ImageJ. To calculate the radius, a contour of the dendritic span was made by drawing a boundary around the dendritic tips, and the area within the contour and resulting equivalent radius was calculated. For calculating dendritic length, the dendrites from the z-stack projection was traced in NeuronStudio. The two-dimensional trace was then exported to MATLAB. Dendritic length was calculated in MATLAB using a custom-written code. We subdivided the cell into eight sectors for analysis. Sectors extended out from the soma and were aligned to the spot stimulus such that each spot was at the center between each sector’s radial boundaries. When computing normalized vector sums (Figure 2B–C, Figure 2—figure supplement 1A-B), total dendritic length in each sector was normalized by the total dendritic length of the cell.

Serial electron-microscopic analysis

Request a detailed protocol

We reconstructed DSGCs and starburst amacrine cells from a previously published SBEM dataset (Ding et al., 2016). These cell types were easily recognizable from their characteristic dendritic arbors and patterns of stratification. We inferred the preferred direction of motion of the DSGCs, relative to the boundaries of the volume from the orientation of SAC dendritic inputs onto the DSGC dendrites, as in Briggman et al., 2011. SAC dendrites preferentially form GABAergic synapses with DSGCs if their orientation, from soma to dendritic tip, corresponds to the null direction of the DSGC. Neither the eye of origin nor the retinal location of the sample was recorded when this sample was acquired, but it was possible to infer the ventral direction within the volume from reconstructions of multiple members of several other RGC types with strong dendritic asymmetries along the dorsoventral axis, including Jam-B RGCs (Kay et al., 2011) and F-RGCs (Rousso et al., 2016). One of the four On-Off DSGC types had an inferred preference for the ventral direction. Thus, we infer that the two types with preferred directions 90 degrees away from this ventral-motion-preferring DSGC type must therefore have been tuned to the horizontal motion axis. We could think of no way to distinguish which of these types prefers anterior motion and which prefers posterior motion.

Ribbon synapses from On bipolar cells onto the dendrites of these three DSGCs were mapped manually. The branching patterns of all cells were reconstructed using skeletons and synapses were marked with single nodes using the Knossos software package (https://knossostool.org/).

To measure local ribbon density across the DSGC dendritic field, the dendritic span was divided into individual squares with 10 × 10 μm (Figure 4—figure supplement 1) or 20 × 20 μm (Figure 4) dimensions. Total dendritic length in each square was calculated in Matlab with a custom-written code. The total number of ribbon synapses in each square was also quantified. Ribbon density was calculated by dividing the ribbon synapse number by the total dendritic length in the appropriate squares.

Electrophysiology recordings

Request a detailed protocol

Recording electrodes of 3–5 MΩ were filled with a cesium-based internal solution containing 110 mM CsMeSO4, 2.8 mM NaCl, 4 mM EGTA, 5 mM TEA-Cl, 4 mM adenosine 5′-triphosphate (magnesium salt), 0.3 mM guanosine 5′-triphosphate (trisodium salt), 20 mM HEPES, 10 mM phosphocreatine (disodium salt), 5 mM N-Ethyllidocaine chloride (QX314),filled with a cesium-based internal solution containing 110 mM CsMeSO4, 2.8 mM NaCl, 4 mM EGTA, 5 mM TEA-Cl, 4 mM adenosine 5′-triphosphate (magnesium salt), 0.3 mM guanosine 5′-triphosphate (trisodium salt), 20 mM HEPES, 10 mM phosphocreatine (disodium salt), 5 mM N-Ethyllidocaine chloride (QX314), and 0.025 mM Alexa 594, pH 7.25. Retinas were kept in oxygenated Ames’ medium with a bath temperature of 32–34°C.

Data were acquired using PCLAMP 10 and a Multiclamp 700B amplifier (Molecular Devices, Sunnyvale, CA), low-pass filtered at 4 kHz and digitized at 10 kHz. Light-evoked EPSCs were isolated by holding cells at −60 mV after correction for the liquid junction potential (~10 mV).

To isolate the contribution of synaptic inputs, a host of pharmacological agents were perfused in the bath during electrophysiology recordings. 8 μM Dihydro-b-erythroidine hydrobromide (DHβE; Tocris, Cat#2349); 10 μM SR 9551 hydrobromide (gabazine; Tocris, Cat #1262); 100 μM Meclofenamic acid sodium salt (MFA; Sigma-Aldrich. Cat#M4531); 1 μM Strychnine (Sigma-Aldrich, Cat#S0532); 50 μM TPMPA (Tocris, Cat#1040); 50 μM Carbenoxolone disodium (Tocris, Cat#3096).

Data analysis of electrophysiological recordings

Request a detailed protocol

Spiking data from loose-patch recordings were analyzed using custom protocols in MATLAB. The number of spikes evoked by the response to the peripheral spots were quantified in MATLAB and averaged across three repetitions in eight spatial locations. Light-evoked EPSC responses to peripheral spots were obtained as well, and three repetitions of EPSC traces were averaged to obtain the mean peak amplitude and charge transfer response to each condition. The RSI is determined by (Max - Opp)/(Max + Opp) where Max is the number of spikes, peak amplitude or charge transfer in the region of maximal activation using the peripheral spot stimulus, and Opp is the response in the region directly opposite to the Max region. EPSC parameters of latency, rise time (10–90%), and decay time (90–10%) were quantified to compare the whole cell kinetics after administration of DHβE and gabazine to the control condition (Figure 1—figure supplement 1E).

For the receptive field mapping experiments, the light-evoked EPSCs in response to spots sized either 20 µm or 60 µm in diameter were obtained and averaged across three to four repetitions. The displacement of the spiking RF was determined by using 60 µm spots. The distance from the soma to the edges of the receptive field on the preferred side versus the null side were determined (Figure 1—figure supplement 1H).

For the receptive field mapping EPSC experiments using 20 µm spots, the center of mass of the receptive field was determined using the EPSC charge transfer or amplitude at each square in the 11 × 11 grid. The dendritic length was calculated at each point on the RF stimulus grid by summing the dendritic length in a circle 60 μm in diameter centered on the square of interest. For maximum vs. opposite region analysis and the preferred side vs null side analysis (Figure 3E–F), we considered the EPSCs in two 5 × 5 subsets of the total grid located opposite of each other. The preferred-null side of each cell was determined by spiking responses to the full-field moving bar, and the maximum-opposite axis was determined by EPSC charge transfer responses to the peripheral spot experiment performed in DHβE and gabazine. The distance from soma at each grid location was defined as the distance between the center of the square and the soma center.

To estimate the spatiotemporal profile across the maximum-opposite or the preferred-null axis for cells in DHβE and gabazine, the light-evoked EPSCs in response to small spots sized 20 µm in diameter were obtained and averaged across three to four repetitions (Figure 4F–H). To estimate the spatiotemporal profile across the preferred-null axis for cells in DHβE only, the light-evoked EPSCs in response to small spots sized 60 µm in diameter were obtained and averaged across three repetitions (Figure 4—figure supplement 1C-D). For both experiments, EPSC parameters of latency, rise time (10–90%), and decay time (90–30%) were quantified. The preferred-null axis of each cell was determined by spiking responses to the full-field moving bar, and the maximum-opposite axis was determined by EPSC charge transfer responses to the peripheral spot experiment performed in DHβE and gabazine. The parameters of latency, rise time, and decay time were averaged across the squares at equal distances along the cell’s preferred-null axis or maximum-opposite axis.

Spiking data evoked by the moving bar and occlusion stimuli were quantified in MATLAB using three to six repetitions in eight different directions. Spiking data evoked by the bar starting in different positions were quantified in MATLAB across three repetitions in the preferred and null directions.

Calcium imaging in posterior- and anterior-preferring On-Off DSGCs

Request a detailed protocol

Genetically encoded calcium indicator GCaMP6f was expressed in all RGCs by crossing Slc17a6tm2(cre)Lowl mice and B6J.Cg-Gt(ROSA)26Sortm95.1(CAG-GCaMP6f)Hze/MwarJ mice. GCaMP6f fluorescence from isolated retinas was imaged in a customized two-photon laser scanning fluorescence microscope (Bruker Nano Surfaces Division). GCaMP6 was excited by a Ti:sapphire laser (Coherent, Chameleon Ultra II, Santa Clara, CA) tuned to 920 nm, and the laser power was adjusted to avoid saturation of the fluorescent signal. Onset of laser scanning induces a transient response in RGCs that adapts to the baseline in ~3 s. Therefore, to ensure the complete adaptation of this laser-induced response and a stable baseline, visual stimuli were given after 10 s of continuous laser scanning. To separate the visual stimulus from GCaMP6 fluorescence, a band-pass filter (Semrock, Rochester, MA) was placed on the OLED to pass blue light peaked at 470 nm, while two notched filters (Bruker Nano Surfaces Division) were placed before the photomultiplier tubes to block light of the same wavelength. The objective was a water immersion objective (60x, Olympus LUMPlanFl/IR). Time series of fluorescence were collected at 15–30 Hz.

We performed an initial direction selectivity test to identify posterior- and anterior-preferring On-Off DSGCs. We recorded GCaMP6f fluorescence from RGC somas within a 75 µm X 75 µm field of view while presenting a full-field moving bar visual stimulus (a bright moving bar 110 μm (width) X 880 μm (length) moving at 330 µm/s across a 660 µm circular mask diameter along eight different directions). At the onset of each moving bar sweep, a TTL pulse was triggered by the visual stimulus computer and recorded by the imaging software to correlate GCaMP6f signals with the direction of each moving bar. Immediately following acquisition of each time series stack, custom-written MATLAB scripts were used to extract fluorescence over time data from time-series images and sort calcium transient by direction of the moving bar. For each RGC soma, raw GCaMP6f fluorescence traces and tuning curves were plotted. On-Off DSGCs were identified by their characteristic singular-lobe directional tuning curves, DSI values ≥ 0.3, and two fluorescence peaks time-locked to the leading (On) and trailing edge (Off) of the moving bar. On-Off DSGCs with preferred directions along the posterior-anterior axis were then selected for further imaging.

Once an On-Off DSGC of interest was identified, the visual stimulus was centered to the soma of that cell and a new field of view was drawn to enclose this cell and some background with no GCaMP6f fluorescence. Full-field and occlusion moving bar visual stimulus were presented to the cells as described above (eight directions, three to four repetitions). Time series data was collected and subjected to offline analysis.

Imaging analysis for calcium imaging

Request a detailed protocol

Analysis was performed using ImageJ and MATLAB. Regions of interest (ROIs) corresponding to DSGC soma and background were manually selected in ImageJ. The fluorescent time course of each ROI was determined by averaging all pixels within the ROI for each frame. The fluorescence of the background region was subtracted from the raw fluorescent signal of the soma ROIs at each time frame. The visual stimulus included a 3–4 s intersweep interval between the end of one sweep and the start of another. Fluorescence intensities during these intersweep intervals were used to create a baseline (F0) trace for each ROI by fitting either a single- or two-term exponential decay function. Fluorescence measurements were then converted to ΔF/F0 values by calculating ΔF=(F−F0)/F0 for every datapoint. The transformed traces were then smoothed using an average sliding window of 4 datapoints. ΔF/F0 traces were clipped, sorted by visual stimulus direction (0, 45, 90, 135, 180, 225, 270, and 315 degrees), and averaged over three to four trials. Prior to further analysis, ROIs were subjected to a response quality test QI = Var[Avg. Resp]/Avg(Var[R(t)])≥0.45 to ensure consistency across trials. Responses to the full-field and occlusion moving bars were broken up into On and Off components according to the circular mask entrance and exit times of the leading and trailing edge, respectively. Peak, area ΔF/F0, and time of peak values were calculated for On, Off, and the full trace along all eight directions. Direction selectivity index (DSI), vector sum, and preferred direction were calculated for both On and Off components.

Statistical analysis

Request a detailed protocol

Grouped data are presented as mean ± SEM. The Kolmogorov-Smirnov test was used to test data for normality. Student’s t-test was used for statistical comparisons of paired samples in Figures 1 and 2. One-way analysis of variance was performed on grouped data in Figures 5 and 6 and subjected to Bonferroni correction.

For the EPSC/dendritic length vs distance from soma experiments, we performed linear regression analysis using an additional categorical predictor variable indicating the maximum-opposite or the preferred-null side. The p-value associated with interaction term (distance*region) in the resulting model was used to determine whether the slopes are significantly different between the two regions. For the EPSC vs dendritic length experiments, we again performed linear regression with an additional categorical predictor variable indicating the maximum-opposite or preferred-null region. The p-value associated with the categorical predictor in the resulting model was used to determine whether the y-intercepts were significantly different between the two regions. The p-value associated with the categorical predictor was used to determine whether the y-intercepts were significantly different between regions. The number of branches in each square of the grid was determined by a custom MATLAB code.

For the kinetic analyses of EPSC parameters in the receptive field mapping experiments of latency, rise time, and decay time, we performed linear regression analysis to determine whether a statistically significant linear relationship exists between the distance from the soma and each EPSC parameter.

For all data sets, p<0.05 was considered significant. *p<0.05; **p<0.01; ***p<0.001.

Experimental parameters for population model

Request a detailed protocol

Experimental data for the spiking response to full-field (73 cells) and occlusion (69 cells) stimuli moving with a constant speed of 330 μm/second were obtained. Three to six repetitions were obtained for each full-field or occlusion protocol. The baseline firing rate for each repetition was obtained by binning the spiking response in 25 ms time bins and taking the maximum firing rate during a silent period where no stimulus was displayed, and the baseline firing rate was averaged across all repetitions.

To model the spiking response of the pDSGCs, we binned the spikes evoked by the On response to the motion stimulus in 25 ms time bins and plotted the PSTHs for all eight motion directions. Then, we fit a rectified sine wave to the PSTH of each pDSGC. We defined the threshold for above-baseline firing to be 4 SD above the baseline firing rate. The onset of the spiking response was determined by the time bin at which the firing rate exceeded the threshold and was immediately followed by a second above-threshold bin (Rate change method; Levakova et al., 2015). We inspected the spiking response onset times returned by our detection algorithm and manually adjusted the spiking response onset times for 3 out of 91 pDSGCs to match the experimental data.

Onset times of spiking responses to the full-field bar moving in the preferred direction ±45 degrees were similar to those of the preferred-direction response (Figure 7—figure supplement 1A, left). Likewise, onset times of spiking responses to the occlusion stimulus moving in the null direction ±45 degrees were similar to those of the null-direction response (Figure 7—figure supplement 1A, right). Therefore, we included onset times of spiking responses to motions in the directions ± 45 degrees from the preferred-null motion axis in our analysis. Four-parameter beta distributions were fit to histograms of spiking response onset times for preferred- and null-direction motions (Figure 7—figure supplement 1F). The four parameters included two shape parameters and two parameters that specify the minimum and maximum of the distribution’s range.

To determine the offset of the spiking response, we used the algorithm for identifying spiking response onset. Unlike the protocol for determining spiking response onset, however, the detection algorithm started from the most recent time bin and traversed backwards in time. The offset of the spiking response was defined to be the time bin at which the firing rate exceeded the threshold in two out of three consecutive time bins. For each pDSGC, we calculated the spiking response duration by finding the difference between the spiking response onset and offset times. Furthermore, we calculated the linear correlation between spiking response duration and onset (Figure 7—figure supplement 1B). Figure 7—figure supplement 1B shows that using only spiking responses to motion along the preferred-null motion axis and using spiking responses to motion along the preferred-null motion axis as well as motion in the directions 45 degrees away from the preferred-null motion axis yielded consistent results. We also calculated the linear correlation between peak firing rate and spiking response onset, but the correlation was not significant for the full-field protocol (Figure 7—figure supplement 1C).

Motion direction tuning curves

Request a detailed protocol

We computed motion direction tuning curves for all pDSGCs exposed to the full-field moving bar stimulus (73 cells) using the CircStat toolbox in MATLAB developed by Berens, 2009. The height of each tuning curve was given by the total spike count evoked during the presentation of the full-field moving bar stimulus. We fit a Gaussian function to the histogram of the tuning curve widths (Figure 7—figure supplement 1G). Figure 7—figure supplement 1H shows all the normalized motion direction tuning curves. The heights of the tuning curves were rescaled by dividing by their peaks and the widths of the tuning curves were rescaled by dividing by their angular deviation, which is the square-root of twice the circular variance. The normalized tuning curves were fit to a one-term Gaussian model (Figure 7—figure supplement 1H).

Speed tuning analysis

Request a detailed protocol

To investigate whether the null-direction response remains robust at higher stimulus speeds, experimental data for the spiking response to full-field and occlusion stimuli moving at 660, 1320, 1980, and 2640 µm/s were further collected. We analyzed how the spike count, onset, and duration of the null-direction response changed across bar speeds (Figure 7—figure supplement 1I-K). Spiking response onset and offset were calculated using the same method as before. Linear fits were performed on spiking response onset data across speeds for both full-field and occlusion protocols (Figure 7—figure supplement 1J). Spiking response durations were first normalized by the response duration when the bar speed was at 330 µm/s and then fit to power-law functions (Figure 7—figure supplement 1K). Simulation parameters for stimulus speeds higher than 330 µm/s were adjusted according to the fit functions in Figure 7—figure supplement 1J and K.

Two-dimensional population model

Request a detailed protocol

In our computational model, we arranged two populations (left motion-preferring and right motion-preferring) of DSGCs in a two-dimensional array. Each population had 1,000 cells. Within each population, horizontal and vertical distances between nearest neighbors were Gaussian distributed, with a mean of 39 μm and a SD of 16 μm (Huberman et al., 2009). The spatial positions of the DSGCs between the two populations were uncorrelated.

Each DSGC’s mean spiking response to the moving edge was modeled as a rectified sine wave. The amplitude of the sine wave was given by the peak firing rate, while the period and phase were determined by the spiking response duration and onset time, respectively. For each spatial arrangement of DSGCs, we sampled peak firing rates directly from our experimental data. We sampled spiking response onset times from our four-parameter beta distributions (Figure 7—figure supplement 1F) and determined the spiking response durations by finding the linear correlation between the two (Figure 7—figure supplement 1B). For simulations with bar stimuli moving at speeds higher than 330 µm/s, spiking response onset times and durations were modified according to the speed of the bar (Figure 7—figure supplement 1J and K).

Noise was introduced into the preferred directions of the DSGCs so that they were not all perfectly aligned with the left/right motion axis. To determine the degree of jitter in a DSGC’s preferred direction, we sampled from a uniform distribution ranging from −14.1 degrees to +14.1 degrees, where 0 degrees represented a preference for motion directly along the left/right motion axis (Fiscella et al., 2015).

To determine the motion direction tuning width of each simulated DSGC, we sampled circular variances from a Gaussian distribution fit to the histogram of circular variance of the tuning curves from our experimental data (Figure 7—figure supplement 1G). We scaled the collapsed tuning curve (Figure 7—figure supplement 1H) by the sampled tuning width and peak firing rate to obtain the tuning curve. We used the motion direction tuning curve to adjust the peak firing rate of the spiking response according to the jitter in the preferred direction alignment.

We simulated the DSGC population response to a moving edge traveling from left to right at a constant speed. To simulate the occlusion protocol, we introduced an occlusion 220 µm in diameter whose position in space was fixed at 1800 µm along the horizontal axis and 800 µm along the vertical axis (approximately in the center of the two-dimensional array). For each spatial arrangement of DSGCs, the simulation was repeated 10 times. The spiking response of each DSGC was discretized in time.

At each 10 ms time bin, the firing rate given by the rectified sine wave fit was converted to a mean spike count. The number of spikes generated by a DSGC was obtained by sampling from a Gaussian distribution with this mean and a sub-Poisson, constant variance of 0.4. The sub-Poisson noise was determined from our experimental data by analyzing the variance of the spiking responses to 6 repetitions of the full-field moving bar (10 cells) and the occlusion stimulus (9 cells) (Figure 7—figure supplement 1D). The spatial positions of the DSGCs were shuffled in each simulation block for a total of 100 blocks with 10 repetitions in each block.

Position and direction decoding

Request a detailed protocol

The spatial position and motion direction of the moving bar’s leading edge were estimated via a labeled-line decoder (Dayan and Abbott, 2001). At each time point, the position estimate or direction x^ was given by the weighted average of the DSGCs’ RF center positions or preferred directions

x^=irix~i/wi2iri/wi2

where ri is the firing rate of the ith cell and x~i𝒩(xi,wi2) where xi is the RF center position or the preferred direction and wi is the RF width (radius) or the motion direction tuning width of the ith cell. For position decoding, the RF width was taken to be the 1 SD boundary of the Gaussian center profile (Chichilnisky and Kalmar, 2002). RF widths were obtained by scaling the dendritic field radii by 1.25. Dendritic field radii were obtained by sampling from Gaussian distribution with μ = 88 μm and σ= 14.8 μm. Position labels of the cells were determined by considering the extent of spatial displacement on the preferred side. Errors are reported as the root mean-square-error in the position or direction estimate (Figure 7F-E).

Coincidence detection

Request a detailed protocol

To assess the salience of the synchronous firing between two oppositely tuned subtypes of DSGCs, we constructed a coincidence decoder that consisted of a convolution with a temporal filter and a threshold operation, similar to Schwartz et al., 2007. At each time point, a difference-of-Gaussian temporal filter (Figure 7F) was convolved with the simulated DSGC population firing-rate activity. When the output of the convolution exceeded the threshold, the decoder determined that a coincidence of spiking activity between DSGC subtypes has occurred. For correct detections, the output must exceed the threshold during a time window of 125 ms around the occlusion event when the bar emerges from behind the occluder (Figure 7G). All above-threshold outputs outside of the time window were marked as false alarms. By varying the threshold across multiple simulations, we computed the receiver operator characteristic (ROC) (Green and Swets, 1966). We quantified the performance of the decoder under different levels of background firing noise using the area under the ROC curve (Figure 7H).

Data availability

Data available on Dryad Digital Repository (http://doi.org/10.5061/dryad.vq83bk3s8). Source data files have been provided for all main text and supplementary figures.Code for model available on Github: https://github.com/jnnfr-ding/Occlusion-model, copy archived at swh:1:rev:123261cdb72251e03cac9654713d17c4537d23a7.

The following data sets were generated
    1. Ding J
    2. Chen A
    3. Chung J
    4. Wu HM
    5. Berson DM
    6. Palmer SE
    7. Wei W
    (2021) Dryad Digital Repository
    Spatially displaced excitation contributes to the encoding of interrupted motion by the retinal direction-selective circuit.
    https://doi.org/10.5061/dryad.vq83bk3s8

References

  1. Book
    1. Dayan P
    2. Abbott L
    (2001)
    Theoretical Neuroscience - Computational and Mathematical Modeling of Neural Systems
    MIT Press.
  2. Book
    1. Green DM
    2. Swets JA
    (1966)
    Signal Detection Theory and Psychophysics
    John Wiley.

Decision letter

  1. Fred Rieke
    Reviewing Editor; University of Washington, United States
  2. Ronald L Calabrese
    Senior Editor; Emory University, United States
  3. Fred Rieke
    Reviewer; University of Washington, United States

In the interests of transparency, eLife publishes the most substantive revision requests and the accompanying author responses.

Acceptance summary:

This paper studies a non-directional signal generated in On-Off directionally selective ganglion cells. Through a combination of experiment and modeling, the paper supports a picture in which this non-directional signal helps signal the location of a moving object, particular as an object emerges from behind an occluder. This provides a nice example of how selectivity for multiple stimulus features can support an interesting circuit function.

Decision letter after peer review:

[Editors’ note: the authors submitted for reconsideration following the decision after peer review. What follows is the decision letter after the first round of review.]

Thank you for submitting your work entitled "Spatially displaced excitation contributes to the encoding of interrupted motion by retinal direction-selective circuit" for consideration by eLife. Your article has been reviewed by 3 peer reviewers, including Fred Rieke as the Reviewing Editor and Reviewer #1, and the evaluation has been overseen by a Senior Editor.

Our decision has been reached after consultation between the reviewers. Based on these discussions and the individual reviews below, we regret to inform you that your work will not be considered further for publication in eLife, at least in present form. We would be willing to consider a revised manuscript if you are able to strengthen the work along the lines detailed below; we recognize though that this would require considerable effort.

The reviewers were in broad agreement that the findings were interesting and that the experiments were well executed and clear. The main concern is that the paper does not provide either a definitive mechanistic insight into why excitatory input is asymmetric, or a definitive functional argument about the importance of this asymmetry. This concern is detailed in the individual reviews, and was a focus of the consultation among the reviews. To be considered further, the paper would need to be strengthened considerably in one of these directions.

Reviewer #1:

This paper describes a new finding about stimulus encoding in On-Off directionally selective ganglion cells. It is well established that these cells have spatially displaced inhibitory input from starburst amacrine cells, and that the spatial offset of inhibitory input contributes to the cells' selectivity for direction of motion. The work in this paper shows that the cells also have spatially offset excitatory input, and that this input can give rise to a non-directional response. Several functional roles are suggested for the non-directional response. I felt that the evidence for the non-directional response was strong, but that the connection to visual function was too preliminary.

Functional importance

The paper emphasizes the possible functional importance of the non-directional motion signal; this is a focus of the discussion, and is highlighted in both the abstract and introduction. I found this part of the paper less complete and convincing than the experimentally-driven results. Several issues contribute to this. One is that the contribution to identifying the position of a moving object is fairly modest. Another is that the impact of the non-directional component on other stimulus properties – e.g. the accuracy with which motion direction is encoded – is not explored. A third is that the position of a moving object is almost certainly encoded by multiple ganglion cell types, and hence the modest improvement in position encoding in the DS cell population may make even less contribution when the entire ganglion cell population is considered. A complete investigation of coding in the ganglion cell population is clearly too much, but a more balanced and complete consideration of the benefits and drawbacks of the mechanism described would strengthen the paper considerably.

Reviewer #2:

In this research, Ding and colleagues present evidence that the excitatory input to OO DS RGCs from bipolar cells is strongly asymmetric, with strong inputs occurring on the side opposite from the SAC inhibition. They performed careful studies to show that this was not due to spatial asymmetry in the DSGC morphology nor to ribbon synapse density. Using 'interrupted motion' stimuli, which are effectively local directional stimuli, they show that this asymmetry leads to a non-directional response on one side of the cell's RF. Last, they create a model to show that such firing patterns could be used to improve localization of edge position under the specific conditions of an edge emerging from behind an occlusion.

The work showing the asymmetry appeared careful, thorough, and well-done. The second half of the paper dealing with the functional consequences of this asymmetry left me with a few questions:

1) Throughout the paper, several experiments showed no changes when a mix of receptor antagonists was added to exclude SAC inhibition as the origin of these effects. But I did not find a positive control, showing that these antagonists had the desired effect. Later, in Figures 5CD, the remaining effect after application of these antagonists was cited as evidence that the excitational asymmetry was responsible for the effect; that interpretation is only valid if the drugs truly kill all SAC input to the DSGC. What if the drugs were not 100% effective? Relatedly, in the experiments in 5CD, the measured responses all decrease with the antagonists, an effect that seems surprising and is not explained. Connecting the asymmetry in excitation to the interrupted motion is central to this paper, so it should have strong support.

2) The measured functional results appear quite similar to results in Kuhn and Gollisch 2019, which is not cited in that context. That paper found that DSGCs responded to local contrast, not just motion, much like the results here, and suggested that oppositely tuned cells could be subtracted to eliminate this contaminating contrast signal or added to isolate the contrast signal. Here, the authors suggest a very similar use for these signals, albeit with a decoder of position and a focus on motion rather than contrast changes. (See line 528, where the authors suggest that this position-direction hypothesis is new. See also line 537: or could not be salient, if there's any kind of downstream opponent subtraction, as in primate MT.)

3) The interrupted motion stimuli are more complex than standard motion stimuli, but it's not clear how ethological or naturalistic they really are. In particular, the occluder was the same contrast as the rest of the background, which seems like a very specific kind of occluded motion, and it's not clear how this would generalize when the occlude is the same or opposite contrast of the moving edge. Moreover, the existence of directed motion in these stimuli lead the authors to emphasize the motion on the 'preferred side', rather than just non-directional contrast changes, which seem as though they would also induce responses.

4) The modeling/decoding aspect of this paper seems pretty speculative. It doesn't seem as though these cells are known to be involved in any kind of position encoding. The fact that they transmit information about contrast changes means they can enhance position-decoding, but many other RGCs could also (better?) serve this purpose. The optic-flow-field arrangement of these cells in the retina suggests just the opposite – that they appear likely to be used for optic flow detection, in which positional information is less relevant than the field structure.

5) Last, I kept wondering how this offset excitatory input made the DSGCs look very similar to a classical Barlow-Levick model (though with DS inhibition). I believe a classical BL model would have many of the properties shown here, including the sensitivity to occluded ND motion on its 'preferred side'. Is there an advantage in the BL model formulation to having disjoint excitatory and inhibitory spatial inputs, rather than a broad excitatory field that overlaps with the delayed inhibition? If so, would such an advantage explain why this asymmetry might exist in these DSGCs, even with DS inhibition from the SACs? I guess I'm asking whether there is an advantage for general motion detection, rather than proposing a new role for these cells in localizing specific types of motion stimuli.

Reviewer #3:

This very interesting manuscript further describes the receptive field structure of ON-OFF retinal direction selective ganglion cells. The authors demonstrate that spot light stimuli flashed at positions that do not correspond with dendritic processes of the recorded DSGC evoke strong excitatory responses that are most powerful on the preferred side of the (moving bar determined) receptive field. The authors go onto show that small light stimuli flashed in the dendritically sampled area of visual space are also non-uniform, and maximal on the preferred side. The authors data are in line with previous reports of a nondirectional zone at the periphery of the dendritic tree of DSGCs. The experimental approaches taken by the authors seem sound. I was concerned by the obviously different kinetics of the flash response recorded under control and GABAA/nAChR antagonists in Figure 1 D, is this a consistent finding, what are the authors thoughts on the unusual shape of the current in Figure 1 D (lower, red trace)? As indicated in the discussion the authors have not investigated the mechanisms underlying this asymmetry, other than dismissing structural determinants (dendritic tree asymmetry, investigation of existing EM volume). This to my mind is a vital component missing from the manuscript. The authors however do go onto describe using elegant light stimulus patterns and modelling some of the potential emergent properties of this behaviour. In this reviewers mind, I am left puzzled and wanting to understand the cellular basis of the behaviour the authors have identified.

[Editors’ note: further revisions were suggested prior to acceptance, as described below.]

Thank you for resubmitting your work entitled "Spatially displaced excitation contributes to the encoding of interrupted motion by the retinal direction-selective circuit" for further consideration by eLife. Your revised article has been reviewed by 3 reviewers, one of whom is a member of our Board of Reviewing Editors, and the evaluation has been overseen by Ronald Calabrese as the Senior Editor.

The reviewers appreciate the revisions to the modeling work, and all agreed that those strengthened the connection to function. Several issues remain that could be further clarified or expanded. (1) The impact of the non-DS region of the RF on encoding of motion direction (Figure 7D) is mentioned briefly in the text, but deserves a more complete treatment, especially as the effects seem substantial. (2) It is not clear why the lack of asymmetry of the dendrites along the preferred-null direction is emphasized and the bias in the direction of maximum excitatory input is relegated to the supplemental figure. The latter result appears to provide at least a partial mechanistic explanation for the bias in excitatory input, and it is not clear why it is not more prominent.

Reviewer #1:

This paper investigates the encoding of motion stimuli by On-Off directionally-selective ganglion cells. The paper introduces a new aspect of such motion sensitivity: the lack of directional selectivity in one part of the receptive field. Corresponding to this functional asymmetry, the glutamatergic excitatory input that the cells receive is not centered on the some, but instead is displaced towards the non-directional part of the receptive field. Modeling shows that this non-directional signal could help identify the position of a moving object, particularly as it emerges from occlusion. The work in the paper is new and interesting. The paper has improved considerable in revision – particularly the new modeling sections strengthen the conclusions about function quite a bit. I do not have any comments on the science itself. There are several places in which the writing could be clearer, as detailed below.

Paragraph starting on line 210. The text here I find confusing. First, you show that excitatory input shows an asymmetry that is in decent alignment with the preferred-null axis. Then you show that the dendrites do not exhibit an asymmetry along the preferred/null axis (at least as measured by branch points and dendritic length). But then Figure S2 and the text in this paragraph argues that there is a dendritic asymmetry in the direction of the largest EPSCs. I know these are not contradictory, but if you first distill each to a simple message then they appear to be. If I am representing these results correctly, some extra text to explain how they are consistent with each other would be helpful.

Lines 128-129: this is still a bit confusing – can you say something like "the preferred side is the side of the receptive field that a bar moving in the preferred direction enters first"

Lines 338-340: Another possibility here is that the kinetics of the bipolar signals do not differ substantially.

Lines 347-350: I would move "We reasoned.…" one sentence later; at present it comes in a sentence that summarizes previous findings.

Lines 406-407: Clarify that these drugs were used individually, not as a cocktail.

Lines 462-467: This paragraph seems out of place – can it go elsewhere?

Line 501: Can you add a sentence to define how the labeled-line decoder works? It would be helpful to have a little intuition here without having to go to the Methods.

Lines 513-515: some reference for what 3 degrees of visual angle means would help (e.g. give the RF sizes in degrees).

Figure 7: can you explain in the text a bit more about why you look at different baseline rates, and why the specific rates were chosen?

Figure 7: What is the "occlusion edge" trace in B? Also the text in this figure is very small and hard to read.

Lines 533-535: suggest emphasizing that simultaneous firing of two oppositely tuned DS cells is not expected for smooth motion. And simultaneous firing of two but not all four is not expected for non-motion contrast signals.

Reviewer #2:

In this research, Ding and colleagues present strong evidence that the excitatory input to OO DS RGCs from bipolar cells is strongly asymmetric, with strongest inputs occurring on the side opposite from the SAC inhibition. They performed careful studies to show that this was not due to spatial asymmetry in the DSGC morphology nor to ribbon synapse density. Using 'interrupted motion' stimuli, which are effectively local directional stimuli, they show that this asymmetry leads to a non-directional response on one side of the cell's RF. Last, they created a model to show that such firing patterns could be used to improve localization of edge position under the specific conditions of an edge emerging from behind an occlusion. They also show that such a set up could lead to an 'alarm response' indicating something interesting going on when local contrast changes, and may also improve the direction-selectivity of DS RGCs.

The strength of this work lies in its careful characterization of this property of ooDRGCs, and its tests of the mechanistic origins of this property. Some similar properties have been observed before, but not characterized with the detail here. The main challenge for this paper is to figure out what the functional significance of this finding is for the animal. This paper proposes a few grounded options, but they remain quite speculative.

I have a few comments, but I'd categorize them as perhaps important but not major. The authors addressed the critical points from the first review in their revised paper. I would have liked to see more about how this property affects direction-selectivity (the known property of these cells), since I suspect it's a large and important effect, but I'm satisfied with the panel presented.

The title refers to "the retinal ds circuit", but this paper only focuses on one of two DS RGC types, and primarily on only 1 (sometimes 2) of the four directions of that type. It seems as though these findings apply to "a retinal DS circuit", though this depends a bit on how one defines "the retinal DS circuit".

The authors have better grounded the modeling of reversed motion and 'alarm signals' in this revision. Since it is still difficult to connect mouse retina experiments to behavior, I wonder if the paper could benefit from a broader comparison in the discussion, since I believe the non-direction-selective responsiveness of motion detectors characterized here has also been found in other well-studied systems. In particular, I believe that DS V1 cells respond non-direction-selectively to local contrast changes, so that the logic of the modeling would apply there. Similarly, studies in flies have highlighted that their local motion detectors respond to local changes in contrast, and connected it to behavior in some cases (see Fisher et al., 2015; Haag et al., 2016; Gruntman et al., 2018; Wienecke et al., 2019; Salazar-Gatzimas et al., 2018; Agrochao et al., 2020), so that the modeling logic should apply there, too. I think a discussion of how this modeling and proposed function might relate to similar features in other systems would strengthen the paper.

The authors chose to focus on a null result in Figure 2, in which dendritic arbors don't correlate strongly with direction-selectivity, but did not focus on the positive result shown in Supp 2, in which dendritic arbors correlated with the glutamatergic excitation field (S2D and E). I'm a little puzzled about why this positive result didn't warrant more discussion in the main text. Supp 2F shows no correlation between the dendritic and response RSIs, but crucially, this is only in the maximal direction for the response. If one computed the RSI for each of the 4 orientations in S2B, then the correlation would surely be significant (as it is in the other analyses in D and E). In that sense, this lack of correlation specifically in the max-opp orientation seems slightly second order. I agree that this plot can be interpreted to mean that it's not all due to dendritic arbor asymmetries, but it seems clear that the dendritic arbor has some influence on the location of the offset excitation in these neurons, so it's a little surprising that doesn't merit more attention.

L 598 typo "anther"

Note: I may have missed it, but I did not see any statement about the rawest data and analysis code being made available to readers, using Dryad and Github or another method. My impression was that this was mandatory.

Reviewer #3:

The revised manuscript is improved, and features more sophisticated modelling analysis. My first review emphasised a missing mechanistic explanation for the asymmetrical excitatory input received by this particular class of ON-OFF-DSGC. The authors have highlighted through analysis of dendritic morphology and a previously published EM volume that a simple structural basis for this asymmetry does not exists. They have not however investigated what the origin of the asymmetrical excitatory input is, which to this reviewers mind is a central ingredient of the work. The authors have however significantly improved their modelling work. My view of the manuscript remains unchanged, and I am left with a simple question – what is the basis of the asymmetyrical excitatory input received by this class of DSGC?

My major concern remains from the first round of review and has not been addressed in resubmission.

https://doi.org/10.7554/eLife.68181.sa1

Author response

[Editors’ note: the authors resubmitted a revised version of the paper for consideration. What follows is the authors’ response to the first round of review.]

Reviewer #1:

This paper describes a new finding about stimulus encoding in On-Off directionally selective ganglion cells. It is well established that these cells have spatially displaced inhibitory input from starburst amacrine cells, and that the spatial offset of inhibitory input contributes to the cells' selectivity for direction of motion. The work in this paper shows that the cells also have spatially offset excitatory input, and that this input can give rise to a non-directional response. Several functional roles are suggested for the non-directional response. I felt that the evidence for the non-directional response was strong, but that the connection to visual function was too preliminary.

Functional importance

The paper emphasizes the possible functional importance of the non-directional motion signal; this is a focus of the discussion, and is highlighted in both the abstract and introduction. I found this part of the paper less complete and convincing than the experimentally-driven results. Several issues contribute to this. One is that the contribution to identifying the position of a moving object is fairly modest.

We thank the reviewer for raising this point. In the original manuscript, the receptive field of the ganglion cell is modeled only in one dimension. In the revised manuscript, we improved our model by incorporating two-dimensional receptive fields of oppositely tuned ganglion cell populations and including moving stimuli at different speeds (Figures 7A-C, lines 486-515, lines 557-562). We found a more pronounced contribution of the displaced receptive field to position estimation during the occlusion stimulus. At the lowest bar speed, the reduction in position error when accounting for the occlusion-associated null response is over 80%, and there is around a 3 degree visual angle reduction in error across speeds. These results provide stronger support that the null-direction occlusion response benefits the localization of a moving object emerging from behind an occluder.

Another is that the impact of the non-directional component on other stimulus properties – e.g. the accuracy with which motion direction is encoded – is not explored.

Thank you for your suggestion. Because interrupted motion can evoke null-direction responses from cells near the occlusion edge, there is indeed an ambiguous signal about motion direction that can be sent to higher order visual areas. We have included an additional plot illustrating the degradation of the direction estimation during occlusion trials (Figure 7D, lines 503-505, line 563). The direction estimation error increase at the time of occlusion is quite significant across speeds, around 50 degrees.

Ultimately, the non-directional response during occlusion trials presents a trade-off between directional and positional decoding. However, this trade-off occurs only during specific visual conditions when an object emerges from behind another object. We speculate that when a moving object first appears out of an occluder, the positional information of the emerging object at that moment might be more important than the direction in which it is moving.

A third is that the position of a moving object is almost certainly encoded by multiple ganglion cell types, and hence the modest improvement in position encoding in the DS cell population may make even less contribution when the entire ganglion cell population is considered.

Thank you for raising this interesting question. We agree that multiple ganglion cell types can certainly encode object position. However, given the divergent central projection patterns of mouse ganglion cell types (over 50 retinorecipient regions in the mouse brain, Martersteck et al., 2017), the position encoding by the entire ganglion cell population may not be available to a given circuit in a retinorecipient structure. On-Off DSGCs project to the superficial layer of the superior colliculus (SC) and the shell of dorsal lateral geniculate nucleus (dLGN) (Huberman et al., 2009). In the SC, directionally tuned On-Off DSGC inputs give rise to the direction selectivity of postsynaptic collicular neurons (Shi et al., 2017), indicating that these collicular neurons do not receive retinal inputs from a broad range of RGC types, but specifically from On-Off DSGCs. While the detailed synaptic organization of the SC is unclear, increasing evidences suggest that inputs from different RGC types are further processed by parallel SC channels (Reinhard et al., 2019). Therefore, the encoding of both motion direction and location by On-Off DSGCs may benefit motion processing in their specific collicular targets. Interestingly, cells in the superficial layer of SC have been implicated in initiating both defensive and prey capture behaviors (Basso et al., 2021). Therefore, the direction and spatial location of a moving object encoded by On-Off DSGCs may help the specific collicular circuit evaluate threats or inform other behaviors.

Additionally, the synchronous response of null-direction and preferred-direction spiking can potentially provide a salient alarm signal. We have created a decoder which can distinguish whether a population of DSGCs has been presented with smooth or interrupted motion (Figures 7E-G, lines 533-551, lines 563-566). The decoder performs extremely well at low levels of baseline firing, which is a characteristic of DSGCs. This indicates that the synchronous response after discontinuous motion is extremely detectable. This detectability can potentially help indicate a change in the visual environment to higher order visual areas, which can help the animal quickly attend to the site of the change. We have revised the Discussion section to incorporate the above responses (lines 613-654).

A complete investigation of coding in the ganglion cell population is clearly too much, but a more balanced and complete consideration of the benefits and drawbacks of the mechanism described would strengthen the paper considerably.

Thank you for your comment. We have further investigated the mechanism by performing multiple new experiments as well as developed a more biologically plausible model with realistic noise levels to explore the functional significance of the receptive field properties we found.

On the experimental side, we have completed several positive controls to test the efficacy of Gabazine (Figures S5E-F, Supplemental material lines 179-183), DHbE (Figure S5G, Supplemental material lines 183-186), and gap junction blockers (Figures S5J-K, Supplemental material lines 192-194). We have also shown that the null-direction response is robust for occluded motion stimuli of different contrast configurations (Figure 5E, lines 390-396). Moreover, we have included additional analyses such as a kinetic analysis of EPSCs after the administration of DHbE and gabazine (Figure S1E, Supplemental material lines 13-15), an analysis of the inhibitory input under the occluded bar stimulus (Figure S5D, Supplemental material lines 178-180), a receptive field mapping experiment to determine the extent of the receptive field displacement on the preferred side (Figure S1H, lines 520-531, Supplemental material 22-25), and an analyses of how the separation of excitatory and inhibitory receptive fields correlates with a cell’s direction selectivity (Figure S7M, lines 462-467, Supplemental material lines 262-263), and Off dendritic analyses(Figures S2C, E, F, Supplemental material lines 76-84) and Off RF property analyses (Figures S3E-F, Supplemental material lines 142-144). These results will be further described below in the point-by-point responses to reviewers’ comments.

For the computational section, we have strengthened the model with two-dimensional receptive field placement and multiple motion speeds to show how the null-direction response may benefit position estimation of an object emerging from behind an occluder. We completed new experiments to measure the spiking activity during the fullfield and occlusion stimulus (Figures S7J-L, Supplemental material lines 259-262) and used these data for the model. We show a robust decrease in the position estimation error across speeds (Figures. 7A-C, lines 486-515, lines 557-562). Additionally, we have implemented a decoder to illustrate that the null-direction response is very detectable, which could indicate that it may be a useful signal for downstream visual areas (Figures. 7E-G, lines 533-551, lines 563-566).

Reviewer #2:

[…] The work showing the asymmetry appeared careful, thorough, and well-done. The second half of the paper dealing with the functional consequences of this asymmetry left me with a few questions:

1) Throughout the paper, several experiments showed no changes when a mix of receptor antagonists was added to exclude SAC inhibition as the origin of these effects. But I did not find a positive control, showing that these antagonists had the desired effect. Later, in Figures 5CD, the remaining effect after application of these antagonists was cited as evidence that the excitational asymmetry was responsible for the effect; that interpretation is only valid if the drugs truly kill all SAC input to the DSGC. What if the drugs were not 100% effective? Relatedly, in the experiments in 5CD, the measured responses all decrease with the antagonists, an effect that seems surprising and is not explained. Connecting the asymmetry in excitation to the interrupted motion is central to this paper, so it should have strong support.

Thank you for your comments, and we agree that there needs to be strong support of the efficacy of the antagonists. We have performed additional experiments and analyses as positive controls to address this concern.

The efficacy of gabazine on blocking inhibitory signaling is shown by the following results: 1. The addition of gabazine caused the spiking of DSGCs to become non-DS in response to a full-field moving bar (Figure S5E, Supplemental material lines 179182). 2. We have added new positive control experiments showing that the use of gabazine completely ablated IPSCs of DSGCs (Figure S5F, Supplemental material lines 182-183) in response to a full-field moving bar. These experiments demonstrate the efficacy of gabazine in ablating SAC inhibition and support our findings of an inhibition-independent mechanism for the displaced excitatory receptive field.

The efficacy of DHbE on blocking nicotinic excitation is verified by a new positive control experiment showing diminished EPSCs of DSGCs after DHbB application during the presentation of a full-field moving bar (Figure S5G, Supplemental material lines 183186). This experiment shows that nicotinic inputs from SACs significantly contribute to the excitatory drive of DSGCs, a finding consistent with previous reports (Lee and Zhou, 2010; Sethuramanujam et al., 2016).

Regarding the decrease of responses in Figures 5C and 5D with both gabazine and

DHbE, the decrease in EPSC is due to DHbE blocking excitatory cholinergic inputs

(Figure 5D, 386-390). For decreased spiking activity during occluded bar stimulus (Figure 5C, 383-386), we reasoned that since nicotinic excitation is a major excitatory drive of DSGC firing during visual stimulation at the RF periphery, blocking inhibition of DSGCs by gabazine is not sufficient to offset the effect of DHbE on spike reduction.

2) The measured functional results appear quite similar to results in Kuhn and Gollisch 2019, which is not cited in that context. That paper found that DSGCs responded to local contrast, not just motion, much like the results here, and suggested that oppositely tuned cells could be subtracted to eliminate this contaminating contrast signal or added to isolate the contrast signal. Here, the authors suggest a very similar use for these signals, albeit with a decoder of position and a focus on motion rather than contrast changes. (See line 528, where the authors suggest that this position-direction hypothesis is new. See also line 537: or could not be salient, if there's any kind of downstream opponent subtraction, as in primate MT.)

Thank you for this comment. We apologize for the oversight and we have added more details about the similarity between the findings of the Kuhn and Gollisch paper and our results. We have amended the discussion to discuss how both studies find that multiple subtypes of DSGCs are necessary to disambiguate multiplexed signals within the DSGC population (lines 604-611).

3) The interrupted motion stimuli are more complex than standard motion stimuli, but it's not clear how ethological or naturalistic they really are. In particular, the occluder was the same contrast as the rest of the background, which seems like a very specific kind of occluded motion, and it's not clear how this would generalize when the occlude is the same or opposite contrast of the moving edge.

We agree that showing a generalization with stimuli of different contrasts was necessary. In response to this comment, we have conducted experiments using 4 new stimuli with different contrast configurations among the occluder, the moving bar and the background (Figure 5E, lines 390-396). We show that both the On and Off responses to the new stimuli also show the occlusion null-direction response, which supports a generalization of this phenomenon across contrasts.

Moreover, the existence of directed motion in these stimuli lead the authors to emphasize the motion on the 'preferred side', rather than just non-directional contrast changes, which seem as though they would also induce responses.

Indeed, stationary flashing spots presented to the preferred side evoke responses (Figure 1). We have also added a receptive field mapping experiment to show the extent of the receptive field displacement to the preferred side (Figure S1H, Supplemental material lines 22-25). The diameter of the receptive field is 220 microns, which is larger than the average dendritic span of a DSGC.

To follow up the reviewer’s comment, we investigated whether the displaced receptive field would benefit the detection of non-directional contrast changes like a small flashing spot. We found that there is no difference in the position estimation of local contrast changes in populations with displaced receptive fields compared to populations with symmetric receptive fields in relation to the soma. Only when we decrease the size of the overall receptive field do we get an improvement in the position estimation of the local contrast change (lines 522-531). Therefore, the location information provided by synchronous preferred-direction and null-direction responses, which localizes the stimulus to a much narrower region, is necessary for improving the location estimation of the moving stimulus.

4) The modeling/decoding aspect of this paper seems pretty speculative. It doesn't seem as though these cells are known to be involved in any kind of position encoding. The fact that they transmit information about contrast changes means they can enhance position-decoding, but many other RGCs could also (better?) serve this purpose. The optic-flow-field arrangement of these cells in the retina suggests just the opposite – that they appear likely to be used for optic flow detection, in which positional information is less relevant than the field structure.

We agree that many other RGCs can also encode object position. For example, certain types of small receptive field RGCs such as W3 RGCs and HD-RGCs may also be well suited to encode object location (Jacoby et al., 2017; Kim et al., 2010; Zhang et al., 2012). However, W3 RGCs are activated only in specific instances where the background is completely uniform, such as when a bird is swooping down from an open sky, whereas On-Off DSGCs can be activated in a wide range of visual environments, including environments with noisy backgrounds. HD-RGCs also have small receptive fields, and computational modeling experiments have shown that the errors in object location between On-Off DSGCs after motion interruption and HD-RGCs are very similar (Jacoby et al., 2017). However, the position information encoded by W3 and HDRGCs may be not available to the specific downstream circuits that receive On-Off DSGC inputs. Our modeling study suggests that the population response of On-Off DSGCs after a motion interruption helps On-Off RGCs achieve the same performance as other ganglion cell populations implicated in fine spatial discrimination.

In addition to position encoding, the synchronous response of null-direction and preferred-direction spiking at the site of motion interruption can potentially provide a salient alarm signal. We have created a decoder which can distinguish whether a population of DSGCs has been presented with smooth or interrupted motion (Figures 7E-G). The decoder performs extremely well at low levels of baseline firing, which is a characteristic of DSGCs. This indicates that the synchronous response after discontinuous motion is extremely detectable. This detectability may help indicate a change in the visual environment to higher order visual areas, which can help the animal quickly attend to the site of the change.

The reviewer also mentioned optic flow detection. In this regard, another type of DSGCs, the On DSGC, which has a larger receptive field and is sensitive to global motion at slower speed, is critical for optic flow detection and the optokinetic reflex. They project to the accessory optic system in the brain stem, and provide the information of the optic flow to instruct the compensatory eye movements for stabilizing the retinal image during the animal’s self-movement. In contrast, On-Off DSGCs are considered local motion detectors since they have smaller RFs than On DSGCs, and are more sensitive to local motion compared to global motion. On-Off DSGCs project to the superficial layer of the superior colliculus (SC) and confer direction selectivity to their postsynaptic collicular neurons. SC is “a system integral for encoding spatial locations and transforming them into stimulus-directed orienting and approach behaviors” (Basso et al., 2021). The encoding of the spatial location of an emerging moving object by On-Off DSGCs may benefit rapid sensorimotor decisions that involve collicular circuitry.

As stated in the above discussions and in our revised manuscript, our conclusions about the functional significance of the null-response after motion interruption is indeed speculative, as we do not know what types of information from the retina are used by each retinorecipient brain circuit. Therefore, we resort to the modeling/theoretical approach and show that it is possible, and perhaps beneficial, that higher visual areas receive information about object position from multiple, independent channels. We wanted to show that the information of spatial location is contained within the On-Off DSGC population response, and that it is possible that higher visual centers stand to benefit from this information. We have expanded the Discussion section in the revised manuscript to include the above discussion.

5) Last, I kept wondering how this offset excitatory input made the DSGCs look very similar to a classical Barlow-Levick model (though with DS inhibition). I believe a classical BL model would have many of the properties shown here, including the sensitivity to occluded ND motion on its 'preferred side'. Is there an advantage in the BL model formulation to having disjoint excitatory and inhibitory spatial inputs, rather than a broad excitatory field that overlaps with the delayed inhibition? If so, would such an advantage explain why this asymmetry might exist in these DSGCs, even with DS inhibition from the SACs? I guess I'm asking whether there is an advantage for general motion detection, rather than proposing a new role for these cells in localizing specific types of motion stimuli.

Thank you for your comment. We have investigated whether the displacement of the excitatory and inhibitory receptive fields contributes to the directional tuning of the On-Off DSGC. Our hypothesis is that a more displaced excitatory RF to the preferred side would further minimize the impact of the weak and delayed inhibition during preferred direction motion and thereby improve direction selectivity. We checked for any correlation between the degree of displacement and the cell’s direction selectivity index (DSI) (Figure S7M, Supplemental material lines 262-263). To quantify the displacement between excitation and inhibition, we determined the onset times of the EPSC and the IPSC in the preferred direction and subtracted the EPSC onset time from the IPSC onset time. If the difference is positive, that means that the IPSC occurs later than the EPSC. We have shown that the cells with an earlier onset in EPSC compared with IPSC in the preferred direction have higher DSIs. This means that the cells with more displaced excitatory RFs compared with inhibitory RFs are generally more direction selective, indicating a potential benefit of the displacement on the cell’s directional tuning (lines 462-467). We have included this analysis in the revised manuscript.

Reviewer #3:

This very interesting manuscript further describes the receptive field structure of ON-OFF retinal direction selective ganglion cells. The authors demonstrate that spot light stimuli flashed at positions that do not correspond with dendritic processes of the recorded DSGC evoke strong excitatory responses that are most powerful on the preferred side of the (moving bar determined) receptive field. The authors go onto show that small light stimuli flashed in the dendritically sampled area of visual space are also non-uniform, and maximal on the preferred side. The authors data are in line with previous reports of a nondirectional zone at the periphery of the dendritic tree of DSGCs. The experimental approaches taken by the authors seem sound. I was concerned by the obviously different kinetics of the flash response recorded under control and GABAA/nAChR antagonists in Figure 1 D, is this a consistent finding, what are the authors thoughts on the unusual shape of the current in Figure 1 D (lower, red trace)?

Thank you for this comment about kinetics. We have added a new analysis of the kinetics of the flash response (Figure S1E, lines 171-173, Supplemental material lines 13-15). We have found that the addition of the GABAA/nAChR antagonists decrease the rise time and the decay time of the response. We reason that the thinner and faster current is due to the blockage of the cholinergic component of the EPSC by the nAChR antagonist. When bipolar cells are activated, they activate both the DSGC and the starburst amacrine cell at the same time. Therefore, the total EPSC of the DSGC is the integration of the monosynaptic glutamatergic input from bipolar cells and the slower disynaptic cholinergic input from the starburst amacrine cells. Removing the cholinergic component of the EPSC by DHbE thereby isolates the more transient waveform of the glutamatergic component.

As indicated in the discussion the authors have not investigated the mechanisms underlying this asymmetry, other than dismissing structural determinants (dendritic tree asymmetry, investigation of existing EM volume). This to my mind is a vital component missing from the manuscript. The authors however do go onto describe using elegant light stimulus patterns and modelling some of the potential emergent properties of this behaviour. In this reviewers mind, I am left puzzled and wanting to understand the cellular basis of the behaviour the authors have identified.

Thank you for your comment. We share the same interest as the reviewer on identifying underlying cellular mechanisms of the excitatory RF displacement. We have been able to exclude several mechanisms and pinpoint what kind of input must be creating the asymmetry. While we have not narrowed this down completely, we feel that we have made several significant advances not only in characterizing this phenomenon, but also in guiding future work towards its mechanistic origin.

In this paper, we have identified a spatially displaced excitatory receptive field that emerges upon partial receptive field stimulation. We have shown that the excitatory receptive field is independent of inhibitory input from starburst amacrine cells, and have identified that the excitatory asymmetry is due to glutamatergic bipolar input. In our revised manuscript, we have added additional positive controls to verify the efficacy of the GABAergic and cholinergic antagonists. We have also performed new experiments at different stimulus contrasts to show that the displaced receptive field is evident under different visual conditions.

Using EM microscopy and dendritic reconstruction, we have ruled out bipolar input distribution and gross dendritic morphology for the mechanism underlying the displaced excitatory receptive field. With a receptive field mapping experiment, we have shown that there is a non-uniform glutamatergic conductance across the dendritic span of On-Off DSGCs. We agree that it would be fascinating to uncover a detailed mechanism of the displaced excitatory receptive field. We are planning to investigate dendritic and synaptic mechanisms in a continuation of this project. In the future, we plan on evaluating the density of glutamate receptors along the dendritic span as well as the dendritic integration properties of the On-Off DSGCs.

Having identified an interesting receptive field property, we were also wondering what potential functional significance the property may play in visual processing. Using a population vector decoder, we show that the null-direction response arising from the displaced RF during partial RF activation can help the population distinguish object location. This suggests that On-Off DSGCs can encode multiple visual features, of motion direction and of object location. Though we have not completely elucidated the mechanism, we hope to share our current results with our colleagues in a timely manner. We believe that our study will be of interest to retinal physiologists as well as computational neuroscientists by investigating how a previously overlooked cellular property can benefit population coding during visual processing.

[Editors’ note: what follows is the authors’ response to the second round of review.]

The reviewers appreciate the revisions to the modeling work, and all agreed that those strengthened the connection to function. Several issues remain that could be further clarified or expanded.

(1) The impact of the non-DS region of the RF on encoding of motion direction (Figure 7D) is mentioned briefly in the text, but deserves a more complete treatment, especially as the effects seem substantial.

We thank the reviewers for the comment and have now expanded on the impact of the non-DS RF region on the encoding of motion direction. We include a new figure panel (revised Figure 7 C) to plot the direction encoding error over time. Together with Figure 7B, they demonstrate that when an object emerges from an occluder, the DSGC population response transiently prioritizes the encoding of the emerging moving object position over its motion direction. We have also revised the text accordingly to provide more discussion about the relationship between position and direction encoding during the occluded motion stimulus.

(2) It is not clear why the lack of asymmetry of the dendrites along the preferred-null direction is emphasized and the bias in the direction of maximum excitatory input is relegated to the supplemental figure. The latter result appears to provide at least a partial mechanistic explanation for the bias in excitatory input, and it is not clear why it is not more prominent.

Thank you for raising this point. We agree with the reviewers that the dendritic morphology partially contributes to the shape of the excitatory RF. Previously we put this point in the supplemental figure since we thought this is not very surprising given that the dendrites provide the physical substrate for the excitatory synapses. We have now followed the reviewers’ suggestions to move this information to the main figure (Figure 2E).

We also add a new main figure panel (Figure 2F) from the previous supplemental figure to illustrate that the spatial skew of the excitatory RF map cannot be solely explained by the dendritic skew, because there is no positive correlation between the extent of the dendritic skew and the extent of the EPSC skew (Figure 2F). Together, these results show that the spatial organization of excitatory RF can be influenced, but cannot be solely explained, by dendritic arbor distribution. We have also revised the text to clarify this conclusion.

Reviewer #1:

[…] The work in the paper is new and interesting. The paper has improved considerable in revision – particularly the new modeling sections strengthen the conclusions about function quite a bit. I do not have any comments on the science itself. There are several places in which the writing could be clearer, as detailed below.

Paragraph starting on line 210. The text here I find confusing. First, you show that excitatory input shows an asymmetry that is in decent alignment with the preferred-null axis. Then you show that the dendrites do not exhibit an asymmetry along the preferred/null axis (at least as measured by branch points and dendritic length). But then Figure S2 and the text in this paragraph argues that there is a dendritic asymmetry in the direction of the largest EPSCs. I know these are not contradictory, but if you first distill each to a simple message then they appear to be. If I am representing these results correctly, some extra text to explain how they are consistent with each other would be helpful.

We appreciate this feedback and have revised this part of the text and Figure 2 to better explain the relationship between dendritic morphology and EPSC displacement. In the revised manuscript, we use the following order in the text (lines 178-196):

1. EPSCs are skewed toward the preferred side (Figure 1).

2. Dendritic fields are not consistently skewed toward the preferred side (Figures 2A-2D).

3. Dendritic fields partially influence ESPC displacement (Figure 2E)

4. Dendritic fields cannot solely explain EPSC displacement (Figure 2F), suggesting additional mechanisms also playing a role.

Lines 128-129: this is still a bit confusing – can you say something like "the preferred side is the side of the receptive field that a bar moving in the preferred direction enters first".

Thank you for your suggestion. We have amended the sentence to include your suggestion (lines 128-130).

Lines 338-340: Another possibility here is that the kinetics of the bipolar signals do not differ substantially.

We have amended the sentence to include this possibility (lines 271-273).

Lines 347-350: I would move "We reasoned.…" one sentence later; at present it comes in a sentence that summarizes previous findings.

Thank you for your suggestion. We have moved the sentence later (lines 289-292).

Lines 406-407: Clarify that these drugs were used individually, not as a cocktail.

We have clarified that these drugs are used individually (lines 317-319).

Lines 462-467: This paragraph seems out of place – can it go elsewhere?

Thank you for pointing this out. We have moved the paragraph earlier to lines (278-285)

Line 501: Can you add a sentence to define how the labeled-line decoder works? It would be helpful to have a little intuition here without having to go to the Methods.

We have included a brief description of the labeled-line decoder in the Results section (line 396-399):

“We used a labeled-line decoder (equivalent to a “population vector” decoder) that estimated the spatial position of the moving bar edge as a weighted average of the RF center positions where the weights were determined by the firing rate (response strength) and RF width (response precision) of each cell (see Methods).”

Lines 513-515: some reference for what 3 degrees of visual angle means would help (e.g. give the RF sizes in degrees).

We have amended the text according to the suggestion:

“The absolute position error decrease was around 3 degrees of visual angle, or around half the receptive field size of an On-OFF DSGC, across bar speeds in models with low levels of background noise.” – amended sentence (line 421)

Figure 7: can you explain in the text a bit more about why you look at different baseline rates, and why the specific rates were chosen?

Thank you for your suggestion. We have added more information about why we chose to look at different baseline rates (lines 402-406). We wanted to match the range of baseline rates reported in the literature.

“We chose to implement the model with different low baseline firing rates to more faithfully represent biological noise. Previous reports show baseline firing rates up to 0.1 Hz (Yao et al., 2018) for On-Off DSGCs. Our experiments yielded background noise levels more on the scale of 0 – 0.025 Hz. Thus, we evaluated the computational model at three different low noise levels.” – amended sentence

Figure 7: What is the "occlusion edge" trace in B? Also the text in this figure is very small and hard to read.

Thank you for noting these issues. We have revised Figure 7B. Instead of using a line, we use a shaded area indicating the position of the occluder. We have also increased the font size of this figure for easier reading.

Lines 533-535: suggest emphasizing that simultaneous firing of two oppositely tuned DS cells is not expected for smooth motion. And simultaneous firing of two but not all four is not expected for non-motion contrast signals.

Thank you for this excellent suggestion. We have included the following text (lines 445-450):

“During smooth motion, only one subtype of DS cell would respond to a bar moving across the visual scene. However, synchronous firing of two oppositely tuned DS cells would occur to represent interrupted or emergent motion. The synchronous firing of two DSGC subtypes can be a unique signature of encoding interrupted motion, which can differ from the encoding of non-motion contrast signals where the recruitment of all four DSGC subtypes would be expected.”

Reviewer #2:

[…] The strength of this work lies in its careful characterization of this property of ooDRGCs, and its tests of the mechanistic origins of this property. Some similar properties have been observed before, but not characterized with the detail here. The main challenge for this paper is to figure out what the functional significance of this finding is for the animal. This paper proposes a few grounded options, but they remain quite speculative.

I have a few comments, but I'd categorize them as perhaps important but not major. The authors addressed the critical points from the first review in their revised paper. I would have liked to see more about how this property affects direction-selectivity (the known property of these cells), since I suspect it's a large and important effect, but I'm satisfied with the panel presented.

The title refers to "the retinal ds circuit", but this paper only focuses on one of two DS RGC types, and primarily on only 1 (sometimes 2) of the four directions of that type. It seems as though these findings apply to "a retinal DS circuit", though this depends a bit on how one defines "the retinal DS circuit".

Thank you for pointing this out. We have changed the title to “a retinal ds circuit”. We also modified the abstract to indicate that we are referring to On-Off DSGCs tuned to motion in the horizontal motion axes.

The authors have better grounded the modeling of reversed motion and 'alarm signals' in this revision. Since it is still difficult to connect mouse retina experiments to behavior, I wonder if the paper could benefit from a broader comparison in the discussion, since I believe the non-direction-selective responsiveness of motion detectors characterized here has also been found in other well-studied systems. In particular, I believe that DS V1 cells respond non-direction-selectively to local contrast changes, so that the logic of the modeling would apply there. Similarly, studies in flies have highlighted that their local motion detectors respond to local changes in contrast, and connected it to behavior in some cases (see Fisher et al., 2015; Haag et al., 2016; Gruntman et al., 2018; Wienecke et al., 2019; Salazar-Gatzimas et al., 2018; Agrochao et al., 2020), so that the modeling logic should apply there, too. I think a discussion of how this modeling and proposed function might relate to similar features in other systems would strengthen the paper.

Thank you for the helpful suggestion. We have included an additional paragraph discussing the interesting parallels of local contrast responses in vertebrate and fly motion detectors (lines 516-530).

“Elements of the visual processing scheme implicated in our study parallel those in other visual areas and species. […] Interestingly, local contrast response properties of direction-selective neurons in the fly visual system have been shown to profoundly modulate their motion computations and contributes to visually guided behavior (Clark et al., 2014; Drews et al., 2020; Matulis et al., 2020).”

The authors chose to focus on a null result in Figure 2, in which dendritic arbors don't correlate strongly with direction-selectivity, but did not focus on the positive result shown in Supp 2, in which dendritic arbors correlated with the glutamatergic excitation field (S2D and E). I'm a little puzzled about why this positive result didn't warrant more discussion in the main text. Supp 2F shows no correlation between the dendritic and response RSIs, but crucially, this is only in the maximal direction for the response. If one computed the RSI for each of the 4 orientations in S2B, then the correlation would surely be significant (as it is in the other analyses in D and E). In that sense, this lack of correlation specifically in the max-opp orientation seems slightly second order. I agree that this plot can be interpreted to mean that it's not all due to dendritic arbor asymmetries, but it seems clear that the dendritic arbor has some influence on the location of the offset excitation in these neurons, so it's a little surprising that doesn't merit more attention.

Thank you for your comments. A related question is raised by reviewer 1 and in the summary statement. We agree with the reviewers that the dendritic morphology partially contributes to the shape of the excitatory RF. Previously we put this point in the supplemental figure since we thought this is not the most important finding given that the dendrites provide the physical substrate for the excitatory synapses. We have now followed the reviewers’ suggestions to move this information to the main figure (Figure 2E). We also add a new main figure panel (Figure 2F) from the previous supplemental figure to illustrate that the spatial skew of the excitatory RF map cannot be solely explained by the dendritic skew, because there is no positive correlation between the extent of the dendritic skew and the extent of the EPSC skew (Figure 2F). Together, these results show that the spatial organization of excitatory RF can be influenced, but cannot be solely explained, by dendritic arbor distribution. We have revised the text to clarify this conclusion. (lines 178-196).

L 598 typo "anther".

Thank you for this correction. We have fixed this typo.

Note: I may have missed it, but I did not see any statement about the rawest data and analysis code being made available to readers, using Dryad and Github or another method. My impression was that this was mandatory.

We have uploaded a statement on the transparency form that the raw data and analysis code will be available on the eLife website, Github (jnnfr-ding/Occulsion-model) and Dryad.

Reviewer #3:

The revised manuscript is improved, and features more sophisticated modelling analysis. My first review emphasised a missing mechanistic explanation for the asymmetrical excitatory input received by this particular class of ON-OFF-DSGC. The authors have highlighted through analysis of dendritic morphology and a previously published EM volume that a simple structural basis for this asymmetry does not exists. They have not however investigated what the origin of the asymmetrical excitatory input is, which to this reviewers mind is a central ingredient of the work. The authors have however significantly improved their modelling work. My view of the manuscript remains unchanged, and I am left with a simple question – what is the basis of the asymmetrical excitatory input received by this class of DSGC?

My major concern remains from the first round of review and has not been addressed in resubmission.

We understand the reviewer’s concern and are currently working on additional experiments regarding the mechanism. We share our preliminary plots with the reviewers in the attachment. However, we think that this result is too preliminary to include in this manuscript, and plan to corroborate it with additional functional and anatomical characterizations for more definitive and detailed mechanistic insights.

https://doi.org/10.7554/eLife.68181.sa2

Article and author information

Author details

  1. Jennifer Ding

    1. Committee on Neurobiology Graduate Program, The University of Chicago, Chicago, United States
    2. Department of Neurobiology, The University of Chicago, Chicago, United States
    Contribution
    Conceptualization, Formal analysis, Validation, Investigation, Visualization, Methodology, Writing - original draft, Writing - review and editing
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0003-2282-6615
  2. Albert Chen

    Department of Organismal Biology, The University of Chicago, Chicago, United States
    Contribution
    Conceptualization, Software, Formal analysis, Investigation, Visualization, Methodology, Writing - original draft, Writing - review and editing
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0002-9306-8703
  3. Janet Chung

    Department of Neurobiology, The University of Chicago, Chicago, United States
    Contribution
    Investigation, Methodology
    Competing interests
    No competing interests declared
  4. Hector Acaron Ledesma

    Graduate Program in Biophysical Sciences, The University of Chicago, Chicago, United States
    Contribution
    Investigation, Methodology
    Competing interests
    No competing interests declared
  5. Mofei Wu

    Department of Neurobiology, The University of Chicago, Chicago, United States
    Contribution
    Investigation, Methodology
    Competing interests
    No competing interests declared
  6. David M Berson

    Department of Neuroscience and Carney Institute for Brain Science, Brown University, Providence, United States
    Contribution
    Methodology
    Competing interests
    No competing interests declared
  7. Stephanie E Palmer

    1. Committee on Neurobiology Graduate Program, The University of Chicago, Chicago, United States
    2. Department of Organismal Biology, The University of Chicago, Chicago, United States
    3. Grossman Institute for Neuroscience, Quantitative Biology and Human Behavior, The University of Chicago, Chicago, United States
    Contribution
    Conceptualization, Software, Supervision, Funding acquisition, Visualization, Writing - original draft, Writing - review and editing
    For correspondence
    sepalmer@uchicago.edu
    Competing interests
    Reviewing editor, eLife
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0001-6211-6293
  8. Wei Wei

    1. Committee on Neurobiology Graduate Program, The University of Chicago, Chicago, United States
    2. Department of Neurobiology, The University of Chicago, Chicago, United States
    3. Grossman Institute for Neuroscience, Quantitative Biology and Human Behavior, The University of Chicago, Chicago, United States
    Contribution
    Conceptualization, Resources, Supervision, Funding acquisition, Investigation, Methodology, Writing - original draft, Project administration, Writing - review and editing
    For correspondence
    weiw@uchicago.edu
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0002-7771-5974

Funding

NIH (R01 NS109990)

  • Wei Wei

McKnight Endowment Fund for Neuroscience (McKnight Scholarship Award)

  • Wei Wei

NSF (GRFP DGE-1746045)

  • Jennifer Ding

NIH (F31 EY029156)

  • Hector Acaron Ledesma

NSF (Career Award 1652617)

  • Stephanie E Palmer

National Science Foundation (PHY-1734030)

  • Stephanie E Palmer

NIH (RO1 EY012793)

  • David M Berson

The funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication.

Acknowledgements

We thank Chen Zhang for managing the mouse colony and Dr. John Maunsell, Dr. Siwei Wang, Mathew Summers, Malak El-Quessny, and Benjamin Hoshal for advice on the manuscript. This work was supported by NIH R01 NS109990 and the McKnight Scholarship Award to WW, NSF GRFP DGE-1746045 to JD, NIH F31 EY029156 to HEA, NSF Career Award 1652617 and the Physics of Biological Function PHY-1734030 to SEP, and NIH RO1 EY012793 to DB.

Ethics

Animal experimentation: All procedures regarding the use of mice were in accordance with the University of Chicago Institutional Animal Care and Use Committee (IACUC) (ACUP protocol 72247) and with the NIH Guide for the Care and Use of Laboratory Animals and the Public Health Service Policy.

Senior Editor

  1. Ronald L Calabrese, Emory University, United States

Reviewing Editor

  1. Fred Rieke, University of Washington, United States

Reviewer

  1. Fred Rieke, University of Washington, United States

Publication history

  1. Received: March 8, 2021
  2. Accepted: June 6, 2021
  3. Accepted Manuscript published: June 7, 2021 (version 1)
  4. Version of Record published: June 17, 2021 (version 2)

Copyright

© 2021, Ding et al.

This article is distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use and redistribution provided that the original author and source are credited.

Metrics

  • 512
    Page views
  • 83
    Downloads
  • 1
    Citations

Article citation count generated by polling the highest count across the following sources: PubMed Central, Crossref, Scopus.

Download links

A two-part list of links to download the article, or parts of the article, in various formats.

Downloads (link to download the article as PDF)

Download citations (links to download the citations from this article in formats compatible with various reference manager tools)

Open citations (links to open the citations from this article in various online reference manager services)

Further reading

    1. Neuroscience
    Debora Fusca, Peter Kloppenburg
    Research Article

    Local interneurons (LNs) mediate complex interactions within the antennal lobe, the primary olfactory system of insects, and the functional analog of the vertebrate olfactory bulb. In the cockroach Periplaneta americana, as in other insects, several types of LNs with distinctive physiological and morphological properties can be defined. Here, we combined whole-cell patch-clamp recordings and Ca2+ imaging of individual LNs to analyze the role of spiking and nonspiking LNs in inter- and intraglomerular signaling during olfactory information processing. Spiking GABAergic LNs reacted to odorant stimulation with a uniform rise in [Ca2+]i in the ramifications of all innervated glomeruli. In contrast, in nonspiking LNs, glomerular Ca2+ signals were odorant specific and varied between glomeruli, resulting in distinct, glomerulus-specific tuning curves. The cell type-specific differences in Ca2+ dynamics support the idea that spiking LNs play a primary role in interglomerular signaling, while they assign nonspiking LNs an essential role in intraglomerular signaling.

    1. Neuroscience
    Wanhui Sheng et al.
    Research Article Updated

    Hypothalamic oxytocinergic magnocellular neurons have a fascinating ability to release peptide from both their axon terminals and from their dendrites. Existing data indicates that the relationship between somatic activity and dendritic release is not constant, but the mechanisms through which this relationship can be modulated are not completely understood. Here, we use a combination of electrical and optical recording techniques to quantify activity-induced calcium influx in proximal vs. distal dendrites of oxytocinergic magnocellular neurons located in the paraventricular nucleus of the hypothalamus (OT-MCNs). Results reveal that the dendrites of OT-MCNs are weak conductors of somatic voltage changes; however, activity-induced dendritic calcium influx can be robustly regulated by both osmosensitive and non-osmosensitive ion channels located along the dendritic membrane. Overall, this study reveals that dendritic conductivity is a dynamic and endogenously regulated feature of OT-MCNs that is likely to have substantial functional impact on central oxytocin release.