1. Neuroscience

The cell adhesion molecule Sdk1 shapes assembly of a retinal circuit that detects localized edges

  1. Pierre-Luc Rochon
  2. Catherine Theriault
  3. Aline Giselle Rangel Olguin
  4. Arjun Krishnaswamy  Is a corresponding author
  1. Department of Physiology, McGill University, Canada
https://doi.org/10.7554/eLife.70870
Share this article
Cite this article
  1. Pierre-Luc Rochon
  2. Catherine Theriault
  3. Aline Giselle Rangel Olguin
  4. Arjun Krishnaswamy
(2021)
The cell adhesion molecule Sdk1 shapes assembly of a retinal circuit that detects localized edges
eLife 10:e70870.
https://doi.org/10.7554/eLife.70870
  2. Download BibTeX
  3. Download .RIS

Abstract

Nearly 50 different mouse retinal ganglion cell (RGC) types sample the visual scene for distinct features. RGC feature selectivity arises from their synapses with a specific subset of amacrine (AC) and bipolar cell (BC) types, but how RGC dendrites arborize and collect input from these specific subsets remains poorly understood. Here we examine the hypothesis that RGCs employ molecular recognition systems to meet this challenge. By combining calcium imaging and type-specific histological stains, we define a family of circuits that express the recognition molecule Sidekick-1 (Sdk1), which include a novel RGC type (S1-RGC) that responds to local edges. Genetic and physiological studies revealed that Sdk1 loss selectively disrupts S1-RGC visual responses, which result from a loss of excitatory and inhibitory inputs and selective dendritic deficits on this neuron. We conclude that Sdk1 shapes dendrite growth and wiring to help S1-RGCs become feature selective.

Introduction

In the retina, each of the ~46 types of retinal ganglion cell (RGC) synapses with a unique subset of amacrine (AC) and bipolar cell (BC) types to create circuits that detect a unique aspect of the visual scene (Clark and Demb, 2016; Gollisch and Meister, 2010; Jadzinsky and Baccus, 2013; Sanes and Masland, 2015; Shekhar et al., 2016; Tran et al., 2019; Yan et al., 2020). A growing body of work suggests that recognition molecules guide the neurites of newborn retinal neurons to grow into sublayers of a specialized neuropil, the inner plexiform layer (IPL), where they contact hundreds of potential synaptic targets (Rangel Olguin et al., 2020; Sanes and Yamagata, 2009; Sanes and Zipursky, 2020; Zhang et al., 2017).

The factors that guide developing arbors to synapse with appropriate targets within a layer are not well understood. An initial idea, called Peter’s principle, posited that developing neurons synapse with nearby cells according to how often they make contact (Binzegger et al., 2004; Peters and Feldman, 1976; Shepherd et al., 2005). However, recent connectomic studies of the IPL demonstrate no obvious relationship between contact frequency and synapse number (Briggman et al., 2011; Helmstaedter et al., 2013). Instead, these connectomic data support a model in which neurons recognize targets in their immediate vicinity and synapse specifically with them. Key molecules in this recognition process are thought to be members of the immunoglobulin superfamily (IgSFs).

Briefly, IgSFs are adhesion molecules that bind to themselves (homophilic) or compatible IgSFs (heterophilic) across cell-cell junctions. It has been proposed that selective IgSF expression within synaptic partners allows their neurites to adhere and synapse when they encounter each other (Sanes and Zipursky, 2020). A recent study in mouse retina provides support for this view (Krishnaswamy et al., 2015). In this study, the IgSF Sidekick-2 (Sdk2) in VG3-ACs and W3B RGCs drives these neurons to connect with each other far more than they connect with the Sdk2-negative neurons they contact. Loss of Sdk2 ablates the enhanced VG3-W3B connectivity but does not alter the gross structure or overlap of their arbors, suggesting that IgSFs increase the probability of synapses between this pair (Krishnaswamy et al., 2015). On the other hand, direction-selective RGCs require the IgSF contactin 5 (Cntn5) to grow dendritic branches in IPL layers bearing axons of their AC/BC partners. Loss of Cntn5 from these RGCs decreases their dendritic branches and reduces synaptic input (Peng et al., 2017), suggesting that IgSFs influence connectivity by regulating intralaminar dendritic growth. These studies suggest a common role for these IgSFs to stabilize/promote the growth of small dendrites that lead to synapses or suggest differing roles for IgSFs in synaptic specificity and neurite growth. Too few IgSFs have been studied in the context of mammalian circuit assembly to draw a firm conclusion. To gain more insight, we investigated the closest IgSF relative of Cntn5 and Sdk2, called Sidekick-1 (Sdk1), in retinal circuit assembly.

We show that Sdk1 is expressed by a family of five RGCs whose dendrites target IPL layers bearing the processes of five Sdk1-expressing interneurons (ACs and BCs). We uncover molecular markers for each Sdk1 RGC and applied these markers following calcium imaging to investigate their visual responses, discovering that the Sdk1 family includes two ON-direction-selective RGCs and an RGC (S1-RGC) sensitive to local edges. Loss of Sdk1 disrupted responses to visual stimuli on S1-RGCs without affecting other members of the Sdk1-RGC family. Using electrophysiological recordings, we show that Sdk1 loss reduced excitatory and inhibitory synaptic inputs to S1-RGCs, decreasing its firing to visual stimuli. These synaptic deficits were specific to S1-RGCs as stimulus-evoked responses on Sdk1+ ON-alpha RGCs were unaffected by Sdk1 loss. Finally, we show that the loss of Sdk1 does not alter the IPL layer targeting of S1-RGC dendrites but selectively reduces their intralaminar complexity and size. From our results, we conclude that Sdk1 is required for S1-RGCs to develop dendritic arbors and receive AC and BC synapses.

Results

Sdk1 labels a family of retinal circuits

To investigate the expression of Sdk1 across the retina, we used mice in which the Sdk1 gene is disrupted by the presence of cDNA encoding a Cre-GFP fusion protein (Sdk1CG); as heterozygotes these lines allow selective access to Sdk1 neurons, as homozygotes they are Sdk1 nulls (Krishnaswamy et al., 2015; Yamagata and Sanes, 2018). Cross-sections and whole mounts from these animals showed GFP+ nuclei within subsets of neurons expressing the BC marker Chx10, the AC marker Ap2-α, and the RGC marker RBPMS (Figure 1A–F); no labeling was found in horizontal cells and photoreceptors. Counting double-labeled cells showed ~7795 Sdk1-BCs//mm2, ~350 Sdk1-ACs /mm2, and ~390/mm2 Sdk1-RGCs (Figure 1G), suggesting that Sdk1 labels several retinal circuits. To address this, we set out to match GFP+ neurons to retinal subtypes.

Figure 1 with 4 supplements see all
Download asset Open asset
Sidekick-1 (Sdk1) labels a family of retinal ganglion cell (RGC) types.

(A–F) Sample cross-sections (A–C) and whole mounts (D–F) from Sdk1CG mice stained with antibodies to GFP and the bipolar cell (BC) marker Chx10 (A, D), amacrine (AC) marker AP2-α (B,E), and RGC marker RBPMS (C, F). Scale = 25 μm. (G) Bar graph showing density of BCs, ACs, displaced amacrine cells (dACs), and RGCs expressing Sdk1 computed from experiments like those shown in (D–F) (n = 14 fields from three animals for each experiment). (H–K) Top: GFP-stained sample Sdk1CG retinal whole mounts co-stained with antibodies to osteopontin (Ost) and neurofilament heavy chain (Nefh) (H), Brn3c (I), or Nr2f2 and calbindin (Calb) (J, K). Dotted circles indicate co-labeled neurons and scale = 50 μm. Bottom: normalized density recovery profiles and average density of co-labeled RGCs measured from corresponding experiments shown in the top row. Colored traces indicate density recovery profiles for co-labeled neurons; gray traces indicate density recovery profiles for all GFP+ cells in the GCL (n = 13–18 fields from six animals for each experiment). (L–O) Images showing the dendritic morphology of individually labeled Ost+/Nefh+ RGCs (L), Brn3c+ RGCs (M), Nr2f2+/Calb+ RGCs (N), and Nr2f2+/Calb- RGCs (O) in Sdk1CG/+ retinal whole mounts. Inset shows marker expression in the soma. (P–S) Top: images showing the laminar morphology of individually labeled Ost+/Nefh+ RGCs (P), Brn3c+ RGCs (Q), Nr2f2+/Calb+ RGCs (R), and Nr2f2+/Calb- RGCs (S) in Sdk1CG/+ retinal cross-sections. Red staining indicates VAcht, a marker for sublaminae 2 (S2) and 4 (S4) (scale = 50 μm). Bottom: inner plexiform layer (IPL) linescans measured from corresponding experiments shown in the top row. Red traces show VAChT intensity, and colored traces show reporter intensity measured from experiments like those shown in the top row (n = 6–12 fields per RGC type from more than 16 animals). (T) Summary cartoon of the Sdk1 RGC family showing Ost+/Nefh+ Onα-RGCs, Brn3c+ S1 RGCs, Ost+/Nefh- M2-RGCs, and Nr2f2+/Calb+ and Calb- ON-DSGCs.

Sdk1 defines a family of five RGCs

Publicly available single RGC sequencing atlases indicate that Sdk1 is highly expressed in approximately six RGC clusters (Rheaume et al., 2018; Tran et al., 2019). Two clusters correspond to alpha RGCs with sustained responses to light onset (ONα-RGCs) and type 2 melanopsin-positive RGCs (M2-RGCs). Four others predicted novel RGCs types (Figure 1—figure supplement 1A). To map these clusters onto GFP+ RGCs, we compared Sdk1 clusters to all RGCs and identified five genes that combinatorially label each Sdk1-RGC type: (1) Onα-RGCs should strongly express neurofilament heavy chain isoform (Nefh), osteopontin (Ost), and the intracellular calcium buffer calbindin (Calb); (2) M2-RGCs should express high levels of Ost+, low levels of Nefh, and low levels of Calb; (3) two predicted novel RGCs should express the steroid hormone receptor Nr2f2 (Nr2f2), (4) with one of these also expressing Calb; (5) another novel RGC should express the Pou-domain containing transcription factor (Brn3c); and (6) a final RGC should not express any of these genes (Figure 1—figure supplement 1B).

We took advantage of the mosaic arrangement of retinal types as an initial test of these predictions. Briefly, retinal neurons of the same type are often spaced apart at a characteristic distance, whereas neurons of different types are often spaced randomly (Keeley et al., 2020; Sanes and Masland, 2015). The degree of regularity varies, but for many types, the density of cells labeled with a candidate marker will drop at short distances from a reference cell if the candidate labels a single type (Keeley et al., 2020; Sanes and Masland, 2015). Sdk1CG whole mounts stained with antibodies to Nefh and Ost showed a pair of Nefh+/Ost+ and Nefh-/Ost+ mosaics spaced at distances expected for the ONα-RGC (Figure 1H, Figure 1—figure supplement 2A) and M2-RGC (Figure 1—figure supplement 1C), respectively. A high-density mosaic was labeled by Brn3c (Figure 1I, Figure 1—figure supplement 2C), and a final pair of mosaics was found to be Nr2f2+/Calb- and Nr2f2+/Calb1+ (Figure 1J and K, Figure 1—figure supplement 2B). We were unable to find an RGC that corresponded to the sixth Sdk1 cluster, as a stain with all these RGC markers and Ap2-α labeled all GFP+ neurons in the GCL (Figure 1—figure supplement 1E–G). Thus, we provide molecular definition for five Sdk1-RGCs, which include three predicted novel types.

We used two approaches to define the anatomy of Sdk1-RGCs. In one approach, retrogradely infecting AAVs bearing Cre-dependent reporters were delivered to the lateral geniculate nucleus (LGN) or superior colliculus (SC) of Sdk1CG mice. In the other, tamoxifen was used to drive reporter expression in a related strain (Sdk1CreER) whose Sdk1 gene is disrupted with cDNA encoding a Cre-human estrogen receptor (CreER) fusion protein (Krishnaswamy et al., 2015). Staining retinae from these mice with our panel of molecular markers revealed the morphology of two Sdk1-RGCs: (1) Ost+/Nefh+ RGCs bore large somas and wide dendritic arbors confined to IPL sublamina 5 and matched previous descriptions of ONα-RGCs (Figure 1L and P; Krieger et al., 2017; Krishnaswamy et al., 2015); and (2) Brn3c+ RGCs had small somas and grew dendritic arbors confined to the center of sublamina 3 (Figure 1M and Q). We refer to these Brn3c+ neurons as S1-RGCs. Ost+/Nefh- and Nr2f2+ RGCs were rarely observed using this method. These results suggest that only a subset of Sdk1-RGCs project strongly to the LGN and SC (Figure 1—figure supplement 3A and D).

One possibility for our relatively poor labeling of Ost+/Nefh- and Nr2f2+ RGCs is that these RGCs innervate ‘non-imaging-forming’ brain regions more strongly than they innervate the LGN and SC (Martersteck et al., 2017). Consistent with this idea, brain sections taken from intraocularly infected Sdk1CG mice showed reporter-labeled RGC axons in the non-image-forming medial terminal nucleus (MTN) and olivary pretectal nuclei (OPN; Figure 1—figure supplement 3B and C). To sparsely label these RGCs, we tamoxifen-treated Sdk1CreER/+ mice crossed to Cre-dependent reporters and stained their retinae with Ost, Nefh, Nr2f2, and Calb. These experiments revealed that both Nr2f2+/Calb+ and Nr2f2+/Calb- RGCs grew oblong dendritic arbors confined to IPL sublamina 4 (Figure 1N, R, O and S). We observed a few Ost+/Nefh- RGC somas in fields labeled densely by reporter (Figure 1—figure supplement 1H), but we were unable to mark them individually, likely due to their low density (~20/mm2) (Sanes and Masland, 2015). Thus, Sdk1 labels a family of five RGCs, which include image- and non-image-forming types.

Sdk1 defines at least five interneuron types

We next asked if Sdk1-ACs and BCs target the same layers as Sdk1-RGCs. Staining reporter-labeled retinae from Sdk1CreER/+ mice with Ap2-α revealed three kinds of Sdk1-ACs (Figure 1—figure supplement 4A–C). One of these had very large (>1 mm) diameter dendritic arbors located in sublamina 5 (Figure 1—figure supplement 4A), another had a similarly large arbor located in sublamina 3 and resembled descriptions of type 2 catacholaminergic ACs (Figure 1—figure supplement 4B; Knop et al., 2014; Krishnaswamy et al., 2015). A third Sdk1-AC grew comparatively narrower dendritic arbors that exhibited the characteristic waterfall appearance of the A17 AC (Figure 1—figure supplement 4C; Masland, 2012). Cross-sections from the same experiments also showed the presence of many rod BCs as judged by their anatomy (Figure 1—figure supplement 4D), as well as many BCs that targeted the interface between sublamina 3 and 4, matching reports of type 7 BCs (Figure 1—figure supplement 4E; Wassle et al., 2009). The IPL lamination profiles of Sdk1-ACs, BCs, and RGCs show that these cells target a common set of IPL laminae. Thus, we conclude that Sdk1 defines a family of neurons whose circuits are contained within the inner three layers of the IPL.

Sdk1-RGCs include direction-selective and edges-detecting subtypes

Registering calcium imaged fields to post hoc stains assigns molecular identity to RGC response

To characterize the functional properties of the Sdk1 RGC family, we devised a procedure to relate marker-gene expression to neural response (Figure 2A, Figure 2—figure supplement 1). Briefly, retinal neurons in Sdk1CG mice were intraocularly infected with AAVs bearing Cre-dependent genetically encoded calcium indicators (GCaMP6f), and 2 weeks later, we imaged their responses to full-field flashes and bars moving at high or low velocity in eight different directions using two-photon microscopy. To register each two-photon field to the optic nerve head, we labeled blood vessels fluorescently labeled with sulphorhodamine 101 and matched the vascular pattern (Figure 2—figure supplement 1A and B). Next, we fixed and stained retinae for GFP, Sdk1-RGC markers, and vessels, and reimaged these retinae with a confocal microscope (Figure 2—figure supplement 1C). Vascular patterns let us register confocal and two-photon fields (Figure 2—figure supplement 1D–K) and Ost distribution let us orient each retina along the dorsoventral and nasotemporal axes (Bleckert et al., 2014). Finally, regions-of-interest (ROIs) were drawn using the confocal image and applied to channels containing marker stains and to the two-photon images to extract responses (Figure 2—figure supplement 1L–O).

Figure 2 with 2 supplements see all
Download asset Open asset
Function of Sidekick-1 (Sdk1) circuits.

(A) Sample whole mount from a Sdk1CG retina infected with AAV-GCaMP6f labeled with the positions of a typical two-photon field (scale = 500 μm). Inset magnifies the boxed field after immunostaining and shows GCaMP-labeled Sdk1-RGCs (green) stained with Brn3c (red) and osteopontin (Ost, blue). (B–E) Magnified somata image, average full-field response, average bar response, sample polar plot, population direction-selective indices (DSI), and ON-OFF index for Ost+ (B, n = 107 cells from 18 retinae), Nr2f2+/Calb+ (C, n = 41 cells from 6 retinae), Nr2f2+/Calb- (D, n = 34 cells from 6 retinae), and Brn3c+ RGCs (E, n = 95 cells from 6 retinae). Raster above each averaged trace shows every response within each retinal ganglion cell (RGC) group. Each RGC shows a characteristic pattern of responses to stimuli. ON-OFF index distribution from a mouse line in which all RGCs express GCaMP6f (all RGCs; n = 1426 cells from five retinae) is shown for comparison. Vertical scales = z-score of 2. (F) Average DSI computed from bar stimuli moving at ~1000 μm/s (fast) or ~200 mm/s (slow) for each Sdk1-RGC group. Significance: **p<0.01; ***p<0.001. (G) Average orientation-selective indices computed from bar stimuli for each Sdk1-RGC group. (H) Average response magnitude to bars moving at ~1000 μm/s (fast) or ~200 mm/s (slow) for each Sdk1-RGC group. (I) Whole-mount retina from the Pcdh9-Cre line crossed to Cre-dependent reporter stained for Nr2f2 and Calb. Triple-labeled cells are encircled (scale = 50 μm).

A balance between antibody species restrictions and throughput led us to group the low-abundance M2-RGC (<20/mm2) with ONα-RGCs using the marker Ost. This registration procedure divided our GCaMP6f dataset into four marker-defined groups, with each bearing a characteristic response to visual stimuli and enrichment of RGC markers (Figure 2, Figure 2—figure supplement 2A–D). To check the consistency of these marker-defined groups, we compared the bar-evoked responses within each group to the average response of all four groups and asked if grouped traces were most similar to their group mean. To do this, we normalized each trace and computed its cosine similarity to the mean response of each group. Ost+ and Brn3c+ traces showed highest similarity to their own mean group responses. Nr2f2+ and Nr2f2+/Calb+ traces also showed high similarity to their mean group response but also showed similarity to each other’s mean response, consistent with the strong resemblance in their visual responses (Figure 2—figure supplement 2A–D). These results indicate that our registration procedure can group molecularly and functionally similar RGCs.

Ost+ RGCs show sustained responses to light onset

Ost+ RGCs (M2- and Onα- RGCs) showed sustained responses to the onset of a full-field flash and leading edge of the moving bar. Converting these bar responses to an ON-OFF index (see Materials and methods) and comparing the distribution of these indices to those computed from a dataset in which all RGCs express GCaMP6f (Figure 2B–E, ON-OFF Index All RGCs; Slc17a6-Cre::Ai95D) emphasize this observation (Figure 2B). Plotting bar direction versus evoked response on a polar plot showed no obvious preference for motion direction (Figure 2B). Converting each polar plot to a direction-selective index (DSI) and preferred angle and viewing the entire Ost+ population on polar axes confirmed this picture, showing weak directional tuning with no systematic preference for stimuli moving along any cardinal axis (Figure 2B). Comparing the average DSI within this group for fast- or slow-moving bars was similar, indicating a lack of motion selectivity within the Ost+ RGC group (Figure 2F). These results are consistent with previous reports of ONα- and M2-RGCs (Berson et al., 2010; Krieger et al., 2017).

Nr2f2+ RGCs are ON-direction selective

On the other hand, both Nr2f2+ RGC types gave transient responses to the leading edge of the moving bar that varied systematically with bar direction. Aligning these responses to retinal orientation revealed polar plots that pointed ventrally for Nr2f2+/Calb+ RGCs (Figure 2C) or moving dorsally for Nr2+/Calb- RGCs (Figure 2D). DSI polar plots for these two populations showed the same respective biases for ventral or dorsal motion and reducing bar velocity caused the average DSI of both RGC types to increase (Figure 2F), indicating that Nr2f2+ RGCs encode the direction of slow-moving bright stimuli. Taken together with our anatomical results, Nr2f2+ RGCs strongly resemble ON-direction-selective RGCs (ON-DSGCs), which grow dendrites in sublamina 4 and comprise three subtypes attuned to motion toward either dorsal, ventral, and nasal poles of the retina (Dhande et al., 2013; Sanes and Masland, 2015; Yonehara et al., 2009). Staining reporter-labeled retinae from the Pcdh9-Cre line, which marks ventrally tuned ON-DSGCs (Lilley et al., 2019; Matsumoto and Yonehara, 2018), showed overlap with ventral motion-selective Nr2f2+/Calb+ RGCs (~3% of reporter-labeled cells; 1035 Pcdh9-Cre cells/mm2; Figure 2I). Thus, we conclude that Nr2f2+ RGCs likely correspond to a pair of ON-DSGCs.

Brn3c+ S1-RGCs are ON-OFF RGCs that respond to bars traveling along the same axis

S1-RGCs (Brn3c+) showed ON responses to a full-field flash but responded to both the leading and trailing edge of the bright moving bar (Figure 2E). These results indicate that S1-RGCs can respond to both bright and dark stimuli and suggest that large full-field stimuli may recruit surround mechanisms that attenuate S1-RGC responses. Plotting S1-RGC bar responses against bar direction showed that many cells respond best to bars traveling slowly (200 μm/s) along the same axis, suggesting that these neurons detect stimulus orientation (Figure 2E and H). Computing an orientation selectivity index (OSI) for these cells showed a higher orientational preference in this neuron as compared with the other Sdk1-RGCs (Figure 2G); however, these values are significantly weaker than the OSI found in recently described orientation-selective RGCs (Nath and Schwartz, 2016; Nath and Schwartz, 2017). Thus, we conclude that S1-RGCs are ON-OFF cells that respond best to edges that fall within their receptive fields.

Sdk1 loss selectively impairs S1-RGC visual responses

Sdk1 loss impairs a subset of types in the Sdk1 RGC family

IgSFs like Sdk1 are thought to guide RGCs to synapse with specific AC and BC types to create feature-selective neural circuits (Sanes and Zipursky, 2020). Given that Sdk1 labels a family of five RGCs and interneurons whose processes overlap in a common set of lamina, we next asked if Sdk1 is important for the development of functional properties in Sdk1 RGCs. To test this idea, we labeled Sdk1 neurons with GCaMP using intraocular AAV injections in Sdk1CG/CG mice, imaged their responses, and grouped these responses using type-specific markers. The responses of Sdk1-null Ost+ RGCs, which include Onα- and M2-RGCs, resembled their heterozygote counterparts (Figure 3A). Nr2f2+ RGCs, which include two kinds of ON-DSGCs, showed subtle changes in their average responses to full-field and moving bar stimuli (Figure 3B and C) with altered directional preference and selectivity, particularly in the NR2f2+/Calb+ group (Figure 3B). S1-RGCs were most affected by Sdk1 loss and showed a significant decrease in response magnitude to bar stimuli (Figure 3D). These results indicate that Sdk1 loss strongly impairs S1-RGCs. Interestingly, single-cell sequencing data showed high levels of Sdk2 in other Sdk1-RGCs relative to S1-RGCs, suggesting that there could be functional redundancy in Onα-, M2-, and ON-DSGCs (Figure 3—figure supplement 1A). We partially confirmed this expression pattern by staining retinae from a Sdk2-CreER knock-in line crossed to reporters with Ost, Nefh, Nr2f2, and Calb antibodies and found many Onα- and Nr2f2+/Calb+ RGCs (Figure 3—figure supplement 1B). However, whole-retina quantitative PCR measurement of Sdk2 transcript levels did not show a significant upregulation in Sdk1 nulls as compared to heterozygotes (Figure 3—figure supplement 1C), although an upward trend was detected. Based on these results, we focused on the S1-RGC for the rest of this study.

Figure 3 with 1 supplement see all
Download asset Open asset
Sidekick-1 (Sdk1) loss causes selective deficits in S1-RGCs.

(A–D) Magnified somata image, average full-field response, average bar response, average peak response to bars, population direction-selective indices (DSI), ON-OFF index, mean orientation-selective index (OSI), and mean DSI corresponding to Ost+ (A, n = 97 cells from 10 retinae), Nr2f2+/Calb+ (B, n = 59 cells from 6 retinae), Nr2f2+/Calb- (C, n = 77 cells from 6 retinae), and Brn3c+ RGCs (D, n = 147 cells from 7 retinae) in Sdk1null retinae. Grayed traces and bars show the same measurements from Sdk1 heterozygotes (Het) (vertical scales = z-score of 2). Sdk1 loss alters Brn3c retinal ganglion cell (RGC) visual responses. Significance: *p<0.05.

Sdk1 loss selectively disrupts S1-RGC visual responses

Retrograde viral injections in the LGN/SC of Sdk1CG/+ offered us a way to study the role of Sdk1 in S1-RGCs and use Sdk1+/Sdk2+ ONα-RGCs as an internal comparison (Figure 4A and B). Recordings from Sdk1CG/+ mice showed many RGCs whose anatomical and functional properties matched one of these two groups. S1-RGCs were easily targeted for loose-patch recordings by their small soma and showed ON and OFF responses to a moving bar that were strongest for motion along a single axis as judged by their elongated profiles on a polar plot (Figure 4A–C). Similar recordings from nearby large-soma ONα-RGCs showed sustained responses to light onset with little tuning for motion direction (Figure 4C).

Figure 4 with 2 supplements see all
Download asset Open asset
Selective loss of visual responses on S1-RGCs in the absence of Sidekick-1 (Sdk1).

(A) Cartoon of S1-RGCs and ONα-RGCs labeled by delivering retrogradely infecting AAVs bearing Cre-dependent reporters in the lateral geniculate nucleus (LGN) or superior colliculus (SC) of Sdk1CG mice. Targets of other Sdk1-RGCs that project to olivary pretectal nuclei (OPN) and medial terminal nucleus (MTN) are also shown. (B) Sample two-photon image of a Sdk1CG/+ retina labeled as described in (A) showing a large-soma ONα-RGC and a small-soma S1-RGC. Sulphorhodamine 101 labels vessels in the GCL. Scale = 25 μm. (C) Polar plots of spike responses to a bar moving in eight different directions recorded from example ONα- and S1-RGCs in experiments like those shown in (B). (D, E) Raster of spike responses to an expanding flashing spot recorded from example S1-RGCs (D) and Onα-RGCs (E) in Sdk1CG/+ (Het) or Sdk1CG/CG retinae (KO). (F, G) Average S1-RGCs firing rates versus bright (F) or dark (G) spot diameter measured from experiments like those shown in (D). (H) Average Onα-RGCs firing rate versus bright (ON) or dark (OFF) spot diameter measured from experiments like those shown in (E). (I, J) Raster of spike responses to centered dark or bright bar rotating through eight orientations recorded from S1-RGCs (I) and Onα-RGCs (J) in Sdk1CG/+ (Het) or Sdk1CG/CG retinae (KO). (K, L) Average firing rate versus bar orientation for S1-RGCs measured from experiments like those shown in (I). (M) Average firing rate versus bar orientation for ONα-RGCs measured from experiments like those shown in (J). (N) Average orientation selectivity indices computed for S1-RGC and ONα-RGC responses to the oriented bar stimulus in control (Het) and Sdk1-null (KO) retinae (n = 7 for Sdk1CG/+ ONα RGCs, n = 12 for Sdk1CG/+ Brn3c RGCs, n = 14 for Sdk1CG/CG Brn3c RGCs, n = 6 for Sdk1CG/CG ONα RGCs; *p<0.05). RGC: retinal ganglion cell.

Our calcium imaging experiments showed that S1-RGCs have strong surround suppression. To relate these signals to spiking behavior, we recorded responses from control S1-RGCs to an expanding spot centered over their receptive field that flashed ON and OFF. S1-RGCs showed ON and OFF responses to this stimulus that were strongly suppressed by spot size, with OFF responses nearly silenced by spots exceeding ~300 μm (Figure 4D, F, G, Figure 4—figure supplement 1). Both ON and OFF components were most strongly activated by spot sizes close to the diameter of S1-RGC dendritic arbors, indicating that the receptive field center on these neurons is ~120–150 μm (Figure 4F and G). Recordings from Sdk1 heterozygote ONα-RGCs displayed higher baseline spiking than S1-RGCs and sustained ON responses that were poorly suppressed by large stimuli (Figure 4E), consistent with their full-field and moving bar responses in our calcium imaging experiments. Similar responses were found in wild-type Onα-RGCs, bearing two copies of Sdk1, labeled with the Kcng4-Cre line crossed to Cre-dependent reporters (Figure 4—figure supplement 2A and B). In contrast, the same recordings from S1-RGCs in Sdk1CG/CG retinae showed a dramatic loss of responsivity to dark stimuli and significantly weaker responses to ON stimuli (Figure 4D, F, G). Recordings from nearby Sdk1-null ONα-RGCs showed comparable responses to their control counterparts (Figure 4E and H), indicating that Sdk1 loss selectively affects S1-RGCs.

We found that most S1-RGCs showed responses to axial bar motion, similar to what we saw in our calcium imaging experiments. We suspected the inability to align bars with S1-RGC-receptive fields in the imaging studies could have activated their strong surround and attenuated their responses to this stimulus. We revisited the idea that these neurons might exhibit sensitivity to stimulus orientation, presenting stationary bars whose width matched the receptive field size of S1-RGCs and rotated through eight different orientations. As expected from their responses to moving bars, S1-RGCs responded preferentially to both bright and dark bars oriented along a single axis (Figure 4I), whereas Onα-RGCs respond to bright bars alone with little response variation to bar orientation (Figure 4J). The same stimulus evoked poor responses from S1-RGCs in Sdk1CG/CG retinae (Figure 4I,K,L), with significantly weaker responses to dark and bright bars, confirming our results using bright and dark expanding spot stimuli. The same recordings from nearby Sdk1-null Onα-RGCs showed similar responses to those in controls (Figure 4J and M). Computing OSI values for these cells and comparing the mean OSI for S1-RGCs and ONα-RGCs in controls and knockout retina showed a selective reduction of orientation selectivity for S1-RGCs in the absence of Sdk1 (Figure 4N). Thus, we conclude that Sdk1 is required for S1-RGCs to develop normal responses to visual stimuli.

Sdk1 loss impairs excitatory and inhibitory synaptic inputs to S1-RGCs

The deficits we observed on S1-RGCs in Sdk1CG/CG retinae might result from a loss of excitatory inputs, a change in inhibitory inputs, or both. To investigate this idea, we recorded synaptic currents from S1-RGCs and ONα-RGCs in control and Sdk1 null retinae and compared their responses to visual stimuli. We began with expanding spots, isolating excitatory and inhibitory inputs by holding neurons at –60 mV and 0 mV, respectively (Figure 5A and B). S1-RGCs in controls showed inward (excitatory) currents at –60 mV to both bright and dark spots that grew in strength until the spot reached ~150 μm in diameter and then steadily weakened (Figure 5C and D). Outward (inhibitory) currents at 0 mV were also found to both bright and dark spots but grew steadily with spot diameter (Figure 5E and F). Recordings from S1-RGCs in SdkCG/CG retinae showed significantly weaker inward and outward currents to bright and dark spots at all spot diameters (Figure 5A–F), indicating a loss of functional excitatory and inhibitory synapses on S1-RGCs in the absence of Sdk1. Whole-cell recordings from nearby ONα-RGCs showed inward and outward currents that were similar between controls and knockouts (Figure 5—figure supplement 1A and B), indicating that Sdk1 loss impairs synaptic inputs on S1-RGCs rather than generally affecting excitatory and inhibitory synaptic strength.

Figure 5 with 3 supplements see all
Download asset Open asset
Sidekick-1 (Sdk1) loss causes S1-RGCs to lose synaptic inputs and dendritic arbor complexity.

(A–B) Whole-cell recordings from S1-RGCs held at potentials to isolate excitatory (~–60 mV) or inhibitory (~0 mV) currents to an expanding flashing spot in Sdk1CG/+ (Het) or Sdk1CG/CG retinae (KO). (C, D) Average peak current versus expanding bright (C) or dark (D) spot diameter measured from control or knockout S1-RGCs held at –60 mV in experiments like those shown in (A, B). (E, F) Average peak current versus expanding bright (E) or dark (F) spot diameter measured from control or knockout S1-RGCs held at –0 mV taken experiments like those shown in (A, B) (n = 8 for Sdk1-Het, n = 9 for Sdk1-KO). (G) Retinal cross-sections showing S1-RGCs in control (Het) Sdk1 knockout (KO) retinae. Scale = 50 μm. (H) Linescans through S1-RGC arbors in control in Sdk1 null retinae taken from experiments like those shown in (G). (I) Skeletonized S1-RGC dendrites from control (black) and Sdk1 null (red) retinae (scale = 50 μm). (J–L) Average branch number (J), cumulative branch length (K), and dendritic area (L) measured from control and Sdk1 null S1-RGC dendritic arbors. (M) Sholl analysis of dendritic arbors measured from Het and KO S1-RGCs (n = 8 for both Sdk1-Het and Sdk1-KO; *p<0.05; **p<0.01). RGC: retinal ganglion cell.

Finally, we examined the synaptic currents evoked on S1-RGCs to our oriented bar stimuli. Controls showed inward and outward currents to both bright and dark oriented bars, but the magnitudes of these currents varied with bar orientation (Figure 5—figure supplement 2A and B). Integrating these currents across the presentation time of each bar and normalizing the responses showed a systematic change in stimulus-evoked charge with bar angle. Orientations that produced the strongest excitation were orthogonal to those producing the strongest inhibition (Figure 5—figure supplement 2E and F). The same stimulus evoked significantly weaker inward and outward currents from S1-RGCs in Sdk1CG/CG retinae, consistent with their reduced responses to the expanding spot stimulus (Figure 5—figure supplement 2C, D,G–I). Integrating these responses across stimulus presentation time and normalizing these responses to the average maximal response in controls showed that excitatory currents retained their preference for bar orientation, but the tuning of inhibitory inputs became less selective (Figure 5—figure supplement 2G–I). Thus, we conclude that Sdk1 is required for S1-RGCs to develop excitatory and inhibitory synaptic inputs.

Sdk1 loss impairs S1-RGC dendritic development

Finally, we asked whether there are structural correlates of the reduced synaptic input to S1-RGCs. One possibility is that Sdk1-null S1-RGCs target inappropriate IPL sublayers and therefore cannot receive input from their interneuron partners. However, comparing the laminar position of S1-RGC dendrites in Sdk1CG/+ or Sdk1CG/CG retinal cross-sections showed no obvious difference between controls and nulls (Figure 5G and H). We next asked if Sdk1 loss impacts the lateral anatomy of S1-RGC dendrites. We found that the loss of Sdk1 led S1-RGCs to grow dendritic arbors that were less complex than their control counterparts (Figure 5I). Sdk1-null S1-RGC arbors contained similar numbers of dendritic branches (Figure 5J), but they were approximately half as long on average, which led to fewer intersections across the entire dendritic arbor as assessed by Sholl analysis (Figure 5K–M). Reduced S1-RGC dendritic arbor size did not alter the spatial distribution of these neurons (Figure 5—figure supplement 1A and B). These deficits arose with little gross change in the overall structure of the IPL, as assessed by staining with a variety of markers that label AC and BC subsets targeting specific sublayers (Figure 5—figure supplement 3). Reconstructed ONα-RGCs in the same Sdk1CG/CG retina showed subtler deficits and more closely resembled their counterparts in heterozygote retinae (Figure 5—figure supplement 1C–H). Heterozygote Onα-RGCs in turn resembled wild-type Onα-RGCs labeled with the KCNG4-Cre line, supporting our functional observations, and suggesting that retinae bearing one or two copies of Sdk1 are similar (Figure 4—figure supplement 2). Taken together, these results show that Sdk1 loss impairs S1-RGCs dendritic arborization.

Discussion

Here, we investigated the role of Sdk1 in retinal circuit development. In the first section of this study, we molecularly, anatomically, and functionally defined five Sdk1 interneurons (ACs and BCs) and five Sdk1-RGCs that target an inner set of IPL lamina. This family of RGCs includes two ON-DSGC types and ON-OFF S1-RGCs; the latter shows selectivity for edges located in their receptive field. In the second section, we found that Sdk1 loss caused a significant loss of S1-RGC responsivity with smaller effects on the other Sdk1-RGC types. By comparing visual responses between S1-RGCs and Onα-RGCs in control and Sdk1 nulls, we show that the loss of Sdk1 impairs S1-RGCs’ responses to both bright and dark spot stimuli, as well as oriented bars. Finally, we show that these deficits arise from a selective loss of excitatory and inhibitory synaptic input on S1-RGCs and correlates with a selective loss of small branches in this neuron’s dendritic arbor. We conclude that Sdk1 is required for the dendritic and synaptic development of a local edge-detecting RGC type.

A family of Sdk1 circuits

Transcriptomic studies indicate that RGCs can be divided into at least 46 different clusters, which include several clusters that do not correspond to any known RGC and are predicted to be novel. In the process of studying Sdk1, we developed molecular signatures for three of these orphan clusters and characterized them using calcium imaging.

We show a pair of Nr2f2+ RGCs can be distinguished based on their expression of Calb and that exhibit selective responses for stimuli traveling ventrally (Nr2f2+/Calb+; C10, Tran et al., 2019) or dorsally (Nr2f2+/Calb-; C27, Tran et al., 2019) over the retina. Both cell types exhibit stronger responses to bars whose speed matches the preferred tuning of ON-DSGCs (200 μm/s) and elaborate medium-field dendritic arbors confined to sublamina 4 (Lilley et al., 2019; Matsumoto et al., 2019; Sanes and Masland, 2015). Moreover, reporter-labeled RGCs in the Pcdh9-Cre line, which marks ventral motion-selective ON-DSGCs (Lilley et al., 2019; Matsumoto et al., 2019), colocalize the ventral motion-selective Nr2f2+/Calb+ RGCs. Finally, we show strong RGC axon labeling in the MTN, a brain region known to be targeted by ON-DSGC axons (Dhande et al., 2013; Martersteck et al., 2017; Oyster and Barlow, 1967; Yonehara et al., 2008; Yonehara et al., 2009). These results strongly suggest that Nr2f2+ Sdk1 RGCs comprise two subtypes of ON-DSGCs.

S1-RGCs are a high-density, narrow dendritic field neuron, characterized by small (<200 μm) diameter dendritic arbors, strong surround suppression, and responses to both bright and dark stimuli. Similar properties have been found in several other RGCs that respond to stimuli falling in their receptive field center but are silenced when the same stimuli fall in their surround. One of these RGCs, called W3B, expresses Sdk2 and grows into sublamina 3 just like S1-RGCs, suggesting that the Sdk family plays unique roles in the development of two physically entangled, but functionally distinct local edge-detecting circuits. Four other edge-sensing types are described in a recent atlas of functionally and anatomically defined RGC types. Of these, S1-RGCs most closely resemble type 1 or type 2 high-definition RGCs (HD1 or HD2). Like S1-RGCs, HD1 and HD2 are ON-OFF cells and can respond to edges (Jacoby and Schwartz, 2017). However, S1-RGCs also show sensitivity to edge orientation, which could arise from their orthogonally tuned excitatory and inhibitory inputs. This input arrangement has been observed in recently described horizontally and vertically selective OS-RGCs; however, these neurons show significantly stronger orientation selectivity compared to S1-RGCs (Nath and Schwartz, 2016; Nath and Schwartz, 2017). We cautiously speculate that S1-RGCs are a kind of local edge detector with sensitivity to edge orientation. Such a local orientation-selective RGC was reported in a recent calcium imaging survey of RGCs (G14, Baden et al., 2016); however, a lack of molecular markers in both studies prevented us from drawing a clear correspondence. The molecular markers and genetic access provided by our work offer a way to resolve this issue in future.

IgSFs and circuit assembly

Several molecular recognition systems direct various steps of retinal circuit assembly, including IPL region selection (Deans et al., 2011; Matsuoka et al., 2011a; Matsuoka et al., 2011b; Matsuoka et al., 2013; Sun et al., 2013) and layer selection (Duan et al., 2014; Duan et al., 2018; Ray et al., 2018; Yamagata and Sanes, 2008; Yamagata et al., 2002). The factors that guide neurons to synapse with targets once they reach appropriate layers are the least characterized, but preliminary work in mammals and invertebrates suggests that IgSF members play a key role in this phenomenon (Carrillo et al., 2015; Cosmanescu et al., 2018; Krishnaswamy et al., 2015; Peng et al., 2017; Sanes and Zipursky, 2020; Tan et al., 2015; Yamagata and Sanes, 2018). Our work here adds to these findings.

We previously showed that an IgSF called Sdk2 enriches connections between VG3-ACs and W3B-RGCs, permitting this RGC to sense object motion (Kim et al., 2015; Krishnaswamy et al., 2015; Lee et al., 2014). In a separate study, we showed that a different IgSF, contactin 5 (Cntn5), was required for ON-OFF direction-selective RGCs to elaborate dendrites within sublamina 4 and collect input from ACs and BCs (Peng et al., 2017). Too few IgSFs have been studied in the context of mammalian circuit assembly to know whether the differing roles for Sdk2 and Cntn5 represent distinct roles for each IgSF or whether they represent two ends of a continuum in which IgSFs direct small-scale RGC-AC/BC interactions that lead to specific synapses. Here, we show that the closest common relative of Sdk2 and Cntn5, Sdk1, has roles in both phenomena as Sdk1 loss causes deficits in both dendritic branching and appearance of functional synapses on S1-RGCs. We hypothesize that both deficits arise because of a lack of homophilic adhesion between S1-RGCs and Sdk1+ interneurons. Our histological survey of Sdk1-interneurons suggests sublamina 3 projecting wide-field ACs, and possibly type 7 BCs, as potential interneuron targets of S1-RGCs. More characterization of these Sdk1-interneuron types, independent genetic access to these neurons, and Sdk1 misexpression methods is needed to directly test this idea.

Four RGCs express both Sdk1 and Sdk2. Effects of Sdk1 deletion on RGC types that express both Sdk1 and Sdk2 were modest (ON-DSGCs) or undetectable (ONα-RGCs). We considered the possibility that Sdk1 loss led to Sdk2 upregulation, but we did not observe a significant change in Sdk2 mRNA in Sdk1 null retinae. Simple alternatives are that Sdk1 plays a subsidiary role in these cells or that it is required for structural or functional features that we did not assay. Examining the morphology and function of Onα-RGCs and/or ON-DSGCs in Sdk2 nulls and Sdk1/Sdk2 double knockouts would be a starting point to investigate the role of Sdks in these neurons.

Exactly how Sdks exert their connectivity-enriching effects is not clear, but both proteins localize to synapses through an interaction with the MAGI family of PDZ-scaffolding proteins (Yamagata and Sanes, 2010). In the case of Sdk2, VG3-W3B connectivity was reduced with only minor changes to the dendritic arbors of these two neurons, suggesting that Sdks instruct synapses between interneuron-RGC pairs, which may in turn have positive effects on their dendritic arbors. Sdk1 loss causes similar losses in interneuron input to S1-RGCs but the morphological deficits on this cell are more severe, suggesting a more permissive role for Sdks in synapse formation through control of dendritic growth. The genetic and molecular definition of Sdk1 and Sdk2 expressing RGCs provided by our studies, paired with time-lapse imaging in explants, might offer a way to directly examine these alternatives.

Materials and methods

Key resources table
Reagent type (species) or resourceDesignationSource or referenceIdentifiersAdditional information
Strain, strain background (Mus musculus)Sdk1-CreGFP, Sdk1CGKrishnaswamy et al., 2015
(doi: 10.1038/nature14682)		 
Strain, strain background (M. musculus)Sdk2-CreER, Sdk1CreERKrishnaswamy et al., 2015
(doi: 10.1038/nature14682)		 
Strain, strain background (M. musculus)Sdk1-CreER, Sdk2CreERKrishnaswamy et al., 2015
(doi: 10.1038/nature14682)		 
Strain, strain background (M. musculus)KCNG-CreJackson LaboratoryRRID:IMSR_JAX:016963
Strain, strain background (M. musculus)Ai27D-ChR2-tdTomatoJackson LaboratoryRRID:IMSR_JAX:012567
Strain, strain background (M. musculus)Slc17a6-CreJackson LaboratoryRRID:IMSR_JAX 016963
Strain, strain background (M. musculus)Pcdh9-CreMutant Mouse
Research and
Resource Centers		RRID:MMRRC_036084-UCD
Tissue donated
by Yonehara K
Strain, strain background (M. musculus)GCaMP6fJackson
Laboratory		RRID:IMSR_JAX
024105
Sequence-based reagentSdk2-FIDTAssayID: Mm.PT.58.41577551GCTGTCCG
TAAAGAAC
TCCTT
Sequence-based reagentSdk2-RIDTAssayID: Mm.PT.58.41577551ATGAGGTCG
TTGTACTTGGTG
Sequence-based reagentGapdh-FKechad et al., 2012
(doi: 10.1523/JNEUROSCI.4127-12.2012)		Accession#:NM 008084.2TGCAGTGGCA
AAGTGGAGATdonated by
Cayouette M
Sequence-based reagentGapdh-RKechad et al., 2012
(doi: 10.1523/JNEUROSCI.4127-12.2012)		Accession#:NM 008084.2ACTGTGCCG
TTGAATTTGCCdonated by Cayouette M
Commercial assay or kitRNeasy Mini-kitQiagenCat#:74134
Commercial assay or kitEZ DNAseThermoFisherCat#:11766050
Commercial assay or kitSuperScriptIV VILO master mixThermoFisherCat#:11756050
Commercial assay or kitPowerUp SYBR Green Master MixThermoFisherCat#:A25741
Recombinant DNA reagentAAVrg CAG-flex-GCaMP6fNeurophotonics
Platform Canadian
Neurophotonics
Platform Viral Vector
Core Facility		RRID:SCR_016477
Recombinant DNA reagentAAV9 CAG-flex-GCaMP6fCanadian Neurophotonics
Platform Viral Vector
Core Facility		RRID:SCR_016477
Recombinant DNA reagentAAV9 ef1a-flex-TdtomatoCanadian Neurophotonics
Platform Viral Vector
Core Facility		RRID:SCR_016477
Recombinant DNA reagentAAVrg-flex-TdtomatoAddGeneCat#:28306-AAVrg
AntibodyAnti-DsRed (rabbit polyclonal)Clontech
Laboratories		RRID:AB_10013483IF(1/1000)
AntibodyAnti-GFP (chicken polyclonal)AbcamRRID:AB_300798IF(1/1000)
AntibodyAnti-Nr2f2 (mouse monoclonal)AbcamRRID:AB_742211IF(1/1000)
AntibodyAnti-Brn3c (mouse monoclonal)Santa Cruz
Biotechnology		RRID:AB_2167543IF(1/250)
AntibodyAnti-Nefh (mouse monoclonal)BioLegendRRID:AB_2314912IF(1/1000)
AntibodyAnti-calbindinSwantRRID:AB_2314070IF(1/10 000)
AntibodyGoat anti-osteopontinR&D SystemsRRID:AB_2194992IF(1/1000)
AntibodyGoat anti-VAChT (goat polyclonal)MilliporeSigmaRRID:AB_2630394IF(1/1000)
AntibodyAnti-calretinin (rabbit polyclonal)MilliporeSigmaRRID:AB_94259IF(1/2000)
AntibodyAnti-vGlut3 (guinea pig polyclonal)MilliporeSigmaRRID:AB_2819014IF(1/2000)
AntibodyPig anti-RBPMS (guinea pig polyclonal)PhosphosolutionsRRID:AB_2492226IF(1/100)
AntibodyAnti-Chx10 (goat polyclonal)Santa Cruz
Biotechnology		RRID:AB_2216006IF(1/300)
AntibodyAnti-Ap2-α (mouse monoclonal)Developmental
Studies
Hybridoma Bank		Clone:3b5IF(1/100)
AntibodyDonkey anti-rabbit Alexa Fluor 405AbcamRRID:AB_2715515IF(1/1000)
AntibodyDonkey anti-chicken Alexa Fluor 488CedarlaneRRID:AB_2340375IF(1/1000)
AntibodyDonkey anti-goat FITCMilliporeSigmaRRID:AB_92588IF(1/1000)
AntibodyDonkey anti-rabbit Cy3MilliporeSigmaRRID:AB_92588IF(1/1000)
AntibodyDonkey anti-guinea pig Cy3Jackson
ImmunoResearch		RRID:AB_2340460IF(1/500)
AntibodyDonkey anti-goat Cy3MilliporeSigmaRRID:AB_92570IF(1/1000)
AntibodyDonkey anti-mouse Alexa Fluor 647MilliporeSigmaRRID:AB_2687879IF(1/1000)
Chemical compound, drugIsolectinFisher ScientificRRID:SCR_014365IF(1/200)
Chemical compound, drugSulphorhodamine 101MilliporeSigmaCat#:S7635.2 mg/mL
Chemical compound, drugTamoxifenMillliporeSigmaCat#:T5648~1 g/50 g body
weight
Chemical compound, drugFluorescein 3000 MW dextranThermo ScientificCat#:D7156
Chemical compound, drugQX314 BromideTocrisCat#:10145 mM
Chemical compound, drugAmes MediumMilliporeSigmaCat#:A1420
Software, algorithmImageJFijihttps://imagej.net/Fiji
Software, algorithmMATLABSimulinkhttps://www.mathworks.com/products/matlab.html
Software, algorithmSimple Neurite Tracerhttps://imagej.net/plugins/snt/RRID:SCR_016566
Software, algorithmTrees Toolboxhttps://www.treestoolbox.org/RRID:SCR_010457
Software, algorithmRStudioRStudiohttps://rstudio.com
Software, algorithmggplot2https://ggplot2.tidyverse.org/RRID:SCR_014601

Animals

Animals were used in accordance with the rules and regulations established by the Canadian Council on Animal Care and protocols were approved by the Animal Care Committee at McGill University. Male and female Sdk1-CreGFP (Sdk1CG), Sdk1-CreER (Sdk1CreER), and Sdk2-CreER mice aged 35–100 days old were used in this study. Details about the generation of these lines can be found in previous studies (Krishnaswamy et al., 2015; Yamagata and Sanes, 2018). Rosa26-LSL-ChR2-tdTomato mice were obtained from the Jackson Laboratory (AID27, Jackson Labs, RRID:IMSR_JAX:012567) and crossed with Sdk1CreER mice for some anatomical experiments. Slc17a6-Cre mice were obtained from the Jackson Laboratory (Jackson Labs, RRID:IMSR_JAX 016963) and crossed with Cre-dependent GCaMP6f lines (AI95D, Jackson Labs, RRID:IMSR_JAX 024105). KCNG4-Cre mice were obtained from the Jackson Laboratory (KCNG4-Cre, Jackson Labs, RRID: IMSR_JAX: 029414) and crossed with Cre-dependent tdtomato lines (AID27).

Viruses

AAVrg CAG-flex-GCaMP6f, AAV9 CAG-flex-GCaMP6f, and AAV9 ef1a-flex-Tdtomato viral vectors were purchased from the Canadian Neurophotonics Platform Viral Vector Core Facility (RRID:SCR_016477) and AAVrg-flex-Tdtomato was purchased from Addgene (viral prep: 28306-AAVrg). Retrogradely infecting AAVs were used for brain injections while AAV9 was used for intravitreal injections.

Injections

Request a detailed protocol

AAVs were injected either intraocularly or intracranially to label Sdk1-RGCs. For intracranial injections, mice were anesthetized using isoflurane (2.5% in O2), given a combination of subcutaneous carprofen and local bupivacaine/lidocaine mix for analgesia, transferred to a stereotaxic apparatus, and a small craniotomy (<1 mm) made in the appropriate location on the skull using a dental drill. Next, a Neuros syringe (65460–03, Hamilton, Reno, NV) filled with virus was lowered into either the LGN (2.15 mm posterior from bregma, 2.27 mm lateral from the midline and 2.75 mm below the pia) or SC (3.85 mm posterior from bregma, 0.75 mm lateral from the midline and 1.5 mm below the pia) using a stereotaxic manipulator. A microsyringe pump (UMP3-4, World Precision Instrument, Sarasota, FL) was used to infuse 400 nL of virus (15 nL/s) bilaterally in dLGN or SC and the bolus allowed to equilibrate for 8 min before removing the needle. For intraocular injections, mice were anesthetized as above, given carprofen as analgesic, and 1 μL of virus injected via an incision posterior to the eye’s ora serrata using a bevelled Hamilton syringe (7803-05, 7634-01, Hamilton). Mice were given at least 2 weeks to recover before experimental use.

Tamoxifen

Request a detailed protocol

Tamoxifen (TMX) (MillliporeSigma, Cat#T5648) was dissolved in anhydrous ethanol at 200 mg/mL, diluted in sunflower oil to 10 mg/mL, sonicated at 40ºC until dissolved, and stored at –20°C. Prior to injection, TMX aliquots were heated to 37°C and delivered intraperitoneally at ~1 g/50 g body weight to Sdk1CreERxAi27D mice. The dose was repeated twice over 2 days and mice were used between 2 and 4 weeks following treatment.

Histology

Request a detailed protocol

Mice were euthanized by isoflurane overdose and transcardially perfused with chilled PBS followed by 4% (w/v) paraformaldehyde (PFA) in PBS and enucleated. Eye cups were fixed for an additional 45 min and brains fixed for an additional 2–5 hr in chilled 4% (w/v) PFA. For whole-mount stains, tissue was incubated in primary antibodies for 7 days at 4°C and incubated in secondary antibodies overnight at 4°C following a series of washes. Stains following two-photon calcium imaging were performed similarly but included lectin to stain blood vessels for registration. Identical procedures were used to stain thick brain sections obtained cutting 100-μm-thick slices from tissue embedded in 2.5% low EEO agarose (MilliporeSigma, A0576) using a compresstome (VF-200-0Z, Precisionary Instruments, Greenville, NC). For cross-sections, post-fixed retinae were sunk in 40% (w/v) sucrose/PBS, transferred to embedding agent (OCT, Tissue-Plus, Fischer Scientific), flash frozen in 2-methylbutane at –45°C, and then sectioned onto slides at 30 μm thickness on a cryostat. For immunostaining, slides were first washed in PBS, blocked with blocking solution (4% normal donkey serum, 0.4% Triton-X-100 in PBS) for 2 hr and incubated overnight with primary antibodies at 4°C. Tissue was then washed with PBS and incubated in secondary antibodies for 2 hr at room temperature prior to the final wash and tissue mounting.

Antibodies and blood vessel stains

Request a detailed protocol

Antibodies used were as follows: rabbit anti-DsRed (1:1000, Clontech Laboratories; RRID:AB_10013483); chicken anti-GFP (1:1000, Abcam, Cambridge, UK; RRID:AB_300798); mouse anti-Nr2f2 (1:1000, Abcam; RRID:AB_742211); mouse anti-Brn3c (1:250, Santa Cruz Biotechnology, Dallas, TX; RRID:AB_2167543); mouse anti-Nefh (SMI-32, 1:1000, BioLegend, San Diego, CA; RRID:AB_2314912); rabbit anti-calbindin (1:10,000, Swant, Marly, CH; RRID:AB_2314070); goat anti-osteopontin (1:1000, R&D Systems, Minneapolis, MN; RRID:AB_2194992); goat anti-VAChT (1:1000, MilliporeSigma, Darmstadt, DE; RRID:AB_2630394); rabbit anti-calretinin (1:2000, MilliporeSigma; RRID:AB_94259); guinea pig anti-vGlut3 (1:2000, MilliporeSigma; RRID:AB_2819014); guinea pig anti-RBPMS (1:100, Phosphosolutions, Aurora, CO; RRID:AB_2492226); mouse anti-Chx10 (1:300, Santa Cruz Biotechnology; RRID:AB_2216006); and mouse anti-Ap2-α (1:100, clone 3b5 from Developmental Studies Hybridoma Bank, Iowa City, IA). Secondary antibodies were conjugated to Alexa Fluor 405 (Abcam; RRID: AB_2715515), Alexa Fluor 488 (Cedarlane, Ontario, CA; RRID:AB_2340375), FITC (MilliporeSigma; RRID:AB_92588), Cy3 (MilliporeSigma; RRID:AB_92588, RRID:AB_92570, or Jackson ImmunoResearch, West Grove, PA; RRID:AB_2340460) or Alexa Fluor 647 (MilliporeSigma; RRID:AB_2687879). Isolectin (Fisher Scientific, Waltham, MA; RRID:SCR_014365) was incubated along with the secondary antibodies when applicable.

Confocal imaging and analysis

Request a detailed protocol

Images of stained tissue were acquired on an Olympus FV1000 confocal laser-scanning microscope or a Zeiss LSM-710 inverted confocal microscope (Advanced BioImaging Facility, McGill University) at a resolution of 512 × 512 or 1024 × 1024 pixels with step sizes ranging from 0.45 μm to 8 μm. Following image acquisition, minor processing of images including landmark correspondence transforms to match two-photon and stained retinal fields as well as stitching of large tilescans was performed using ImageJ.

For density recovery profile analysis (DRP), RGC soma centers were selected manually on ImageJ and their x–y coordinates used to generate DRPs using code modeled after Rodieck, 1991. Briefly, intercellular distances from each cell to every cell were computed using Pythagoras’ theorem and binned into 15-μm-wide annuli centered upon the reference cell; for Onα- and M2-RGCs, annulus width equaled 25 μm. Counts within each annulus were divided by annulus area, averaged across a given condition, fitted with a sigmoid curve for visual clarity, and plotted alongside DRPs obtained from the entire GFP+ population for comparison. The density of each Sdk1-RGC type was obtained by dividing marker-labeled cells by image area. These values were averaged for each type and compared to mean density of all GFP+ cells as a bar graph. The mean cell density obtained from this analysis was used to assess the proportion of cells labeled by Sdk1 among BCs, ACs, and RGCs.

We analyzed dendritic morphology as described previously (Krishnaswamy et al., 2015). Briefly, images of single S1-RGCs and Onα RGCs taken from P30-60 retinae were traced manually through the z-stack using simple neurite tracer (ImageJ). Path ROIs describing neuronal processes were converted to stacks and analyzed for morphological properties using the Trees Toolbox on MATLAB (Cuntz et al., 2010). The len_tree, B_tree and vhull functions in the Trees Toolbox were used to measure the cumulative length of all dendritic branches, branch length, and arbor area, respectively.

To obtain mean IPL projection depth of each Sdk1+ type, we took linescans of pixel intensity across the IPL from images stained with antibodies to VAChT (sublamina 2 and 4) or reporter, normalized these signals to the maximum intensity, and averaged these traces across each condition. IPL depth is expressed as a percent and sublaminae judged by the position of the peak intensities in the VAChT channel. We applied the straighten transform (ImageJ) to correct the VAChT bands on a few curved sections and applied the same transform to reporter channels prior to linescan procedure.

Calcium imaging

Request a detailed protocol

Calcium imaging was performed as previously described (Liu et al., 2018). Briefly, mice were dark adapted for at least 2 hr, euthanized, and then retinae rapidly dissected under infrared illumination into oxygenated (95% O2; 5% CO2) Ames solution (MilliporeSigma, A1420). Next, retinae were mounted onto a filter paper with the RGC layer facing up, placed in a recording chamber, mounted on the stage of a custom-built two-photon microscope, and perfused with oxygenated Ames solution warmed to 32–34°C. Responses of GCaMP6f+ RGCs to visual stimuli delivered through the objective were imaged at 920 nm and collected at 8 Hz. Each image plane (360 × 72 μm) of the movie contained GCaMP fluorescence, SR101 fluorescence, stage coordinates, and visual stimulus synchronization pulses to permit offline analysis. A few microliters of sulphorhodamine 101 (SR101, 2 mg/mL, MilliporeSigma, Cat#S7635) was added to the recording chamber to label blood vessels and a map of the main blood vessels emanating from the optic disk acquired for post hoc image registration.

Post hoc image registration and RGC ROI selection

Request a detailed protocol

Following each calcium imaging session, averages of each calcium imaged field were stitched into a single image using custom software written in MATLAB (Rochon et al., 2021; Dryad: https://doi.org/10.5061/dryad.4xgxd2593). These large images contained a map of the blood vessel pattern surrounding the retinal optic disk and imaged fields (Figure 2—figure supplement 1A and B). These images were then scaled into microns, overlayed on images of post hoc stained retinae, and rotated or translated until the blood vessel patterns between the two datasets aligned (Figure 2—figure supplement 1A–C). Fine blood vessel morphology was then used as inputs to the landmark correspondences function in Fiji to perform final alignment between these two fields (Figure 2—figure supplement 1D–K). ROIs of stained neurons were used to extract responses from two-photon movies (Figure 4—figure supplement 1L–O). Osteopontin distribution, which is enriched in the dorsal-temporal quadrant of the retina (Bleckert et al., 2014), permitted alignment of RGC calcium responses to the cardinal axes. For a subset of our studies, we were able to restain retinae with two additional markers, register these restains as above, and together with the trace similarity score (see below) categorize Sdk1-RGCs.

Calcium imaging analysis

Request a detailed protocol

RGC responses corresponding to stimulus synchronization pulses were extracted, aligned, and analyzed using custom software written in MATLAB (Simulink). Responses to stimuli were expressed as a z-score where the mean and standard deviation were obtained from an equal duration of the prestimulus baseline on each trial. We computed three higher descriptive statistics: the ON-OFF index was calculated as described previously (Baden et al., 2016), but briefly

On−OffIndex=ONresp−OffrespONresp+Offresp

where ONresp and OFFresp is the mean cell response over a bright and dark full-field stimulus, respectively. ON-OFF indexes from a dataset in which all RGCs express GCaMP6f (Slc17a6-Cre::AI27) were plotted alongside Sdk1-RGC ON-OFF indexes after filtering this dataset using a quality index computed as described in Baden et al., 2016. A direction selectivity index was calculated using the circular variance of the cell response for all eight moving bar directions (Nath and Schwartz, 2016):

DSI=|∑(Respθ⋅eiθ)|∑(Respθ)

where Respθ is the maximum cell response for each bar direction. A similar approach was done to compute an orientation selectivity index, where instead of the first complex exponential the second (e2iθ) is used. A similarity score was used to measure each functional RGC group’s similarity to the group mean. This score is the cosine between the group mean and each individual trace in our dataset. Both traces were Euclidean normalized prior to this measurement using the norm function in MATLAB.

Electrophysiology

Request a detailed protocol

Retinae for electrophysiological recordings were prepared and visualized as described for calcium imaging. For cell-attached recordings, the patch electrodes (4–5 MΩ) were filled with Ames solution. For whole-cell recordings, patch electrodes (5–7 MΩ) were filled with an internal solution containing 112 mM Cs methanosulfate, 10 mM NaAc, 0.2 mM CaCl2, 1 mM MgCl2, 10 mM EGTA, 8 mM CsCl, and 10 mM HEPES (pH 7.4). In both cell-attached and whole-cell recordings, fluorescein 3000 MW dextran (Thermo Scientific, D7156) was added to make the electrode visible under two-photon illumination. For whole-cell recordings, internal solution was supplemented with 5 mM QX314 Bromide. Excitatory currents and inhibitory currents were isolated by adjusting the holding potential to match reversal potentials for excitation (0 mV) and inhibition (ECl ~ –60 mV). Signals from loose-patch and whole-cell recordings were acquired with a MultiClamp 700B amplifier (Molecular Devices) and digitized at 20 kHz using custom software written in LabView. For spikes, the MultiClamp was put into I = 0 mode and Bessel filter set at 1 kHz. For currents, the MultiClamp was put in VC mode and Bessel filter set at 3 kHz. Analysis of electrophysiological signals was performed in MATLAB (Simulink) as follows. Briefly, action potentials were detected in loose patch recordings using the peakfinder function and binned (50 ms) over the entire length of the trial; firing rate histograms for each trial were then averaged and subjected to further processing based on each stimulus. Direction and orientation-selective indices were computed from mean firing rate histograms as described for calcium imaging data. For whole-cell currents, trials were averaged, peak amplitude measured, and integral were computed using the trapz function across each stimulus epoch. For stationary bar stimuli, integrated currents for each bar flash were normalized to average maximum in controls, plotted against bar orientation for each cell, and then peak responses aligned and averaged across all RGCs in heterozygote or knockout retinae.

Visual stimuli

Request a detailed protocol

A DLP light crafter (Texas Instruments, Dallas, TX) was used to project monochrome (410 nm) visual stimuli through a custom lens assembly that steered stimulus patterns into the back of a 20× objective (Euler et al., 2009). All visual stimuli were written in MATLAB using the psychophysics toolbox and displayed with a background intensity set to 1 × 104 R*/rod/s. Custom electronics were made to synchronize the projector LED to the scan retrace of the two-photon microscope.

For calcium imaging experiments, visual stimuli were centered around the microscope field of view. Moving bar stimuli consisted of a bright bar moving along its long axis in one of eight directions. The bar was 1200 μm wide, 3200 μm long moving at either 960 μm/s (fast) or 240 μm/s (slow). For electrophysiological experiments, the cell-receptive field center was identified using a grid of flashing spots and a small user-controlled probe and the location with the highest response assigned as the center for all subsequent stimuli. Moving bars were thinned for electrophysiological experiments to 200 μm; length and speed were the same as calcium imaging studies. Expanding spot stimuli began as a small circle that flashed at the receptive field center and grew after each ON-OFF flash. Stationary, rotating bars had the same dimensions as those for the moving bar but flashed and rotated through nine orientations. All stimuli were preceded by a gray background whose duration equaled the stimulus duration in each trial.

RNAseq data analysis

Request a detailed protocol

RNA sequencing data was taken from the publicly available Broad Institute Single Cell Portal for three previous studies (Shekhar et al., 2016; Tran et al., 2019; Yan et al., 2020). To identify type-specific marker for Sdk1-RGCs we first compared expression of markers between these neurons and all other RGCs to find a differentially expressed subset. Next, we analyzed this subset for the presence five-gene combinations that could segregate each Sdk1 type; Nr2f2 was identified from Rheaume et al., 2018. Sequencing data plots were generated with the ggplot plot package in R.

qPCR

Request a detailed protocol

For RT qPCR, standard procedures were followed (Kechad et al., 2012). Briefly, RNA was extracted from retinas of both Sdk1-Het and Sdk1-KO mice using the RNEasy Mini-kit (Qiagen, Hilden, Germany, Cat#74134) and cDNA synthesized using EZ DNAse (ThermoFisher, Cat#11766050) and SuperScriptTMIV VILO master mix (ThermoFisher, Cat#11656050). Quantitative PCR was done using the PowerUp SYBR Green Master Mix (ThermoFisher, Cat#A25741) on a ViiA 7 Real-Time PCR System. Primers used were as follows: Sdk2-F: GCTGTCCGTAAAGAACTCCTT; Sdk2-R: ATGAGGTCGTTGTACTTGGTG; GAPDH-F: TGCAGTGGCAAAGTGGAGAT; GAPDH-R: ACTGTGCCGTTGAATTTGCC. The Sdk2 levels were normalized to the abundance of the housekeeping gene Gapdh to obtain a relative expression level.

Statistics

No statistical method was used to predetermine sample size. Statistical comparisons between Sdk1-knockout and Sdk1-heterozygote electrophysiological, morphological, and calcium imaging data were performed using the anova1 function in MATLAB (Simulink) followed by the multcompare function for pairwise testing.

Data availability

Sample registration fields, registration code, and GCaMP6f datasets are available at Dryad (https://doi.org/10.5061/dryad.4xgxd2593).

The following data sets were generated
    1. RochonP-L TC
    2. Rangel Olguin AG
    3. Krishnaswamy A
    (2021) Dryad Digital Repository
    Data From: The cell adhesion molecule Sdk1 shapes assembly of a retinal circuit that detects localized edges.
    https://doi.org/10.5061/dryad.4xgxd2593
The following previously published data sets were used

References

Decision letter

  1. Marla B Feller
    Reviewing Editor; University of California, Berkeley, United States
  2. Gary L Westbrook
    Senior Editor; Oregon Health and Science University, United States

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

Acceptance summary:

This study represents a significance contribution to our understanding of molecular factors which govern development of neural circuits. The breadth of techniques used is extremely impressive and the authors are to be commended for such a thorough, interesting and important study.

Decision letter after peer review:

Thank you for submitting your article "The cell adhesion molecule Sdk1 shapes assembly of a retinal circuit that detects localized edges" for consideration by eLife. Your article has been reviewed by 3 peer reviewers, one of whom is a member of our Board of Reviewing Editors and the evaluation has been overseen by Gary Westbrook as the Senior Editor. The reviewers have opted to remain anonymous.

The reviewers have discussed their reviews with one another, and the Reviewing Editor has drafted this to help you prepare a revised submission. The full reviews are attached at the end. The summary of the Essential Revisions should be used to guide your revision.

Essential Revisions:

(1) Include some additional data to compare the heterzygot Sdk1+/- to WT. Reviewer #2, is quite specific about what would be acceptable.

(2) Further characterization of S1- RGCs would strengthen the finding in that it could confirm whether they are indeed edge detecting cells.

(3) If possible, it would be great to determine whether there is an upregulation of other Sdks in RGCs subtypes that are not impacted in Sdk1 ko. Along the same lines, all reviewers have asked for a larger discussion Sdks and related proteins and retinal wiring.

(4) All reviewers have commented on mosaic analysis – their issues should be addressed.

Reviewer #1 (Recommendations for the authors):

A major goal of developmental neurobiology is to understand the molecular cues that are involved in the wiring up of neural circuits. In this study, the research is carried out in the retina where there are several features that make it a strong candidate to address these questions-namely cell types are well defined and the locations of synapses between different cell types are clearly demarcated in plexiform layers within which axons and dendrites of the neurons that comprise distinct microcircuits co-stratify.

The authors take advantage of this system to explore the role that a candidate molecular marker from immunoglobulin superfamily of adhesion molecules called Sidekicks (Sdks). A prior study showed that Sdk2 was critical for matching of a particular interneuron with a particular subtype of RGC. Here the authors explore Sdk1.

The authors use a large array of anatomical markers (derived from lists of genes in recent RNA transcript libraries) to identify the 5-6 different RGC subtypes that express Sdk1. They use multiple methods to confirm this assignment to cell types including morphology, immunofluorescence, projection patterns to the brain and light responses. They go on to show that in the Sdk1-ko mouse, only one of these subtypes exhibits a large phenotype – a significant reduction in the light responses. Using whole cell voltage clamp recordings, they show that his reduction in light response is correlated with a is a reduction in excitatory and inhibitory synaptic inputs, consistent with a lack of synapse formation. This is a convincing set of data that they have identified a factor that is critical for the formation of a microcircuit within the retina. There is also a bigger picture, which is that this class of Sdk2 adhesion molecules may be part of a molecular code used by multiple circuits to wire up during development.

This is an extremely thorough study that identifies the role that Sidekick 1 – a member of the IgSF family of adhesion molecules – plays in the wiring up a particular microcircuits in the retina. First the authors use multiple methods to identify the cells that express Sdk1 – identifying 5-6 different RGC subtypes and a few amacrine and bipolar cell subtypes. Using information from RGC sequencing atlases, they identify candidate proteins that can be used to further classify these RGC subtypes and are able to positively identify 5 of them – 2 that were named subtypes and 3 that were not. Taking advantage of some genetic tricks (i.e. sparse labeling of SDK-1-Cre x with flex mouse, non-sparse labeling for anterograde projections, retrograde AAVs injected in L and SC) the authors were able to use immunofluorescence, morphology and projections patterns to further classify the cells. Finally, they characterize the light responses if these cells using calcium imaging. By combining all of these methods, they make a strong case that they have labeled ON-α RGCs, M2-iRGCs two different ON direction selective GCs and a difficult-to-classify cell – -the S1. Note the S1 cells send dendrites in S3 of the IPL, which the authors show also contains processes from excitatory and inhibitory interneurons that are also Sdk1 positive. Hence these are good candidate for Sdk1-dependent circuit-wiring.

They then look in the Sdk1 KO mouse. First they use 2p calcium imaging to assess the impact on all the different Sdk1+ RGCs. They identify the subtypes post imaging by finding the same field of view and using antibody staining (though this method has now been used by a few labs, they do an excellent job describing their method in a supplemental figure). They find deficits in actually quite a few cell types but this biggest impact is on the S1-RGCs, which have a greatly reduced light response.

Using the intersection between GFP expression under SDK1-Cre-GFP fusion mouse + retrograde labeling from SC/LGN, they are able to target electrodes to just two subtypes of Sdk1+ RGCs, the S1s and the On-alphas. Hence they can directly compare the impact of the knockout on spiking properties and synaptic inputs to these two cell types. They see dramatic reduction in light responses and EPSCs and IPSCs onto these cells in S1-RGCs. In contrast, they see no phenotype at all on On-α RGCs. This is strong evidence that there is reduction in synapse formation onto S1 cells but not all Sdk1+ RGC subtypes.

In some ways the impact of this study seems small – one adhesion molecule impacts one microcircuit within the retina. But I would argue that the methods used and the completeness of the study represent a broader impact. There were many important insights – for example multiple cells might have a particular adhesion molecule but it is required for only one microcircuit. Also, the authors have identified more markers for different RGC types and have done this classification in the context of development. These sorts of observations are critical for ultimately understanding the molecular code for wiring up the retina. For these reasons, I think this study represent a significant step forward in the field.

I do think in a revision that perhaps the authors can zoom out a bit and give the readers a bigger picture. How big is the Sdk family of proteins? Are they alternatively spliced? How numerous are they compared to say protocadherins, or semaphorins or other factors that have been implicated. Again – I think this will make it more relevant for broader audience.

Below are a few comments that will hopefully clarify the presentation.

1. The one major issue I have is with Figure 1 H-K. The authors are strongly recommended to read a recent review (Keely, Elgen and Reese,JCN, 2020) arguing that density recover profiles that do not take into account soma size are not revealing mosaic organization and therefore this really cannot be used to "verify" they have a cell type. That said. I don't think the lack of the mosaic analysis will not significantly weaken their results -- the antibody staining + physiology is sufficient for indicating that they have different cell types.

2. (Lines 108-114/Figure 1 T-U )- the authors want to argue that Ost+/Nefh- are M2-ipRGCs and so they look at projections to the MTN via anterograde labeling. However, it is unclear how they correlate anterograde projections with the Ost+/Nefh- RGCs in retina. This needs to be clarified.

3. Labels in Figure 1are very hard to read. Also, I strongly recommend that the authors refer to the different RGC subtypes in subsequent figures as their better known subtypes (M2, α, ON-DSGC) rather than the immunipositive proteins. It was quite confusing. – I had to print out Figure 1V and use it several times to figure out what was going on in several figures.

4. For Figure 2 – it was not clear when the authors were using GCaMP in all RGCs (expressed under vglut-Cre – line 163) or specific for Sdk1-Cre (line 138).

5. Figure 2 – is bar velocity 1000mm/sec (which I believe is the speed of an action potential!) or 1000 µm/sec?

6. Figure 2 – would be great if there were some quantification go figures H and I – are all Pcdh-Cre stained for Nr2f2? What percentage for Calb+?

7. Figure 3 – there seems like a substantial loss of DS in the ON-DSGCs that the authors should note. Again, this broadens the impact of the study because clearly Sdk1 is important for this circuit, but the impact is more subtle. It is not that they need to study it here but it is important to emphasis the multiple impacts it has. Hence line is 220-221 are not quite accurate.

Reviewer #2 (Recommendations for the authors):

In this study Rochon et al., elucidate the distribution of an adhesion molecule called Sidekick 1 (Sdk1) among retinal neurons. Sidekicks are a family of cell adhesion molecules, implicated in organization of dendritic arbors among retinal neurons, and also in regulating the organization of synapses across specific visual circuits. The authors show that Sdk1 is selectively expressed among 5 retinal output (ganglion cell) types, 3 inhibitory interneuron (amacrine) cells and 2 excitatory interneuron (bipolar) cell types, all in the inner retina. Using a combination of genetic tools and cell specific markers, the authors detail the identity of the cell types expressing Sdk1 in the retina. Using a loss of function approach, the authors show that lack of Sdk1 impacts the synaptic connectivity and visual response properties of a specific retinal output neuron (retinal ganglion cell or RGC) type that responds to local edges. This interesting work reveals a selective role of Sdk1 for the function of a specific visual circuit, and as such the authors succeeded in the task they set out to achieve i.e., determine the distribution and an insight into possible functional role(s) of Sdk1 in the retina. The strengths of this study are the combinatorial approach of genetic tools, electrophysiology and functional imaging and the data are of a high standard.

There are some weaknesses though, in the current version of the manuscript that if addressed would strengthen the work and its implications.

(i) The current version used Sdk1 HET animals (with one copy of Sdk1) as controls for all functional and morphological experiments, but it is unclear how cells in the HET retina compare in terms of visual responses and morphology to cells in a wildtype retina with two copies of Sdk1. It is thus difficult to extrapolate the magnitude of visual response deficits or dendritic mis-organization of cells in the Sdk1 deficient retinas.

(ii) The feature of the S1-RGC responding to local edges could be detailed further to include consideration of potential temporal sensitivity/tuning of the cells and asymmetry between their ON and OFF response profiles as is often observed for other RGCs that respond to local edges. Comparison between the ON vs OFF components in control retina are also of merit since the OFF responses are impacted more than the ON responses in the Sdk1 deficient retinas, underscoring a potential asymmetry in how Sdk1 could wire these RGCs.

(iii) The authors show that Sdk1 could be implicated in regulating the synaptic connectivity of a specific retinal ganglion cell (S1-RGC) type that responds to local edges but do not discuss whether these interactions could be homophilic or heterophilic and whether or not the 5 interneurons that express Sdk1 could serve as presynaptic partners to the S1-RGC (or the other RGC types expressing Sdk1). Such a discussion would be of benefit in consideration of the fact that some RGC types that express Sdk1, like the ON DS or ONalpha, have known presynaptic bipolar cell and amacrine cell partners that lack Sdk1 expression. Additionally, most of the amacrine and bipolar cell types shown to express Sdk1 do not co-laminate with the S1-RGC. Thus, a heterophilic mode of interaction for Sdk1's wiring of specific visual circuits seems to be likely which appears different from the homophilic interactions that underlie Sdk2 connections between VG3 amacrine cells and W3B RGCs that both express Sdk2 (authors' previous work). Because Sdk1 is a close relative of Sdk2 that appears to operate differently, a discussion would enable readers to better grasp the context and significance of the work.

(iv) Even though Sdk1 is expressed by 5 different ganglion cell types, its loss seems to selectively perturb the synaptic connectivity and functional responses of one specific (S1-RGC) ganglion type, indicating that perhaps other members of the Sdk family could compensate for lack of Sdk1 in the unaffected ganglion cell types. Inclusion of data indicating whether or not Sdk2 or other Sdk1 relatives are upregulated in these unaffected ganglion cell types in Sdk1 deficient retinas, would clarify the redundancy between these IgSF members during retinal circuit formation and also clarify why the loss of Sdk1 leaves certain retinal circuits unperturbed.

The following concerns should be addressed by the authors:

(1) How close is the Sdk1 HET retina to a wildtype retina with two copies of Sdk1? Could the authors compare RGC responses they obtain in the Sdk1 HET to responses in a wildtype retina to enable the reader to grasp the magnitude of deficits upon Sdk1 deficiency? As the authors have done an incredible job of findings cell specific markers for each of the RGC types that express Sdk1, they could use these markers to correlate responses in HET vs wildtype retina. At the minimum, responses from easily identifiable RGC types like the ONalpha or the ON DS which the authors have a specific transgenic line for, could be recorded in retinas with two copies of Sdk1 to evaluate how similar these are compared to the responses (using the same stimuli and lighting conditions) obtained from HET retinas with a single Sdk1 copy. This data would be helpful to validate the use of the Sdk1 HET as a control.

(2) Given that previous accounts of edge sensitive RGC types like the HD RGCs show that these cells are most responsive to specific speeds (Jacoby and Schwartz 2017), it would be worth determining if the S1-RGC type detailed in this work are also tuned to a specific speed of stimuli. This aspect of temporal tuning would aid the characterization of the response profile of the S1-RGC type. The authors would most likely already have this data as they have made a similar temporal tuning determination for the ON DS RGCs.

(3) Along the same lines, edge sensing RGCs can have ON and OFF responses that are not symmetrical (Jacoby and Schwartz 2017; Roska and Werblin 2001). As such it would add to the characterization of the S1-RGC response profile, if the authors could determine if the ON and OFF responses of the S1-RGC are symmetric/equivalent or not. From the examples shown in Figure 4 it appears as though the ON responses are more prominent than the OFF both for the expanding spot and orientation experiments i.e., more spikes allocated during light increments. Other edge sensing RGCs can have one component (ON or OFF) more dominant and thus it would be important for the reader to know whether the S1-RGC shows a similar asymmetry or not. Again, the authors already have this data so it should be straight forward to analyze and perhaps include plots where the ON and OFF responses (# of spikes) of the S1-RGCs are compared in the same plot.

(4) The authors indicate that Sdk2 might be compensating for Sdk1 in the RGC types unaffected by Sdk1 deficiency. However, they do not provide evidence for whether or not Sdk2 is upregulated in the unaffected RGCs upon Sdk1 deficiency. The authors have generated antibodies to Sdk2 in their previous work which could be used to determine whether or not Sdk2 expression is upregulated amongst select RGC types upon Sdk1 deficiency (this could be compared in HET vs knockout).

(5) The authors bring up the homophilic vs heterophilic interactions between IgSFs like Sdks in the Introduction section when discussing the role of IgSFs like Sdks for synapse formation but do not include a discussion of whether Sdk1 could be operating via homophilic or heterophilic interactions to set up synaptic connections in the inner retina. Given the Sdk1 expression profiles it appears that Sdk1 could operate via heterophilic interactions, which is different from how Sdk2 wires the VG3 amacrine cell-W3B RGC circuit and it would be of benefit to the reader to include a discussion of the mechanistic basis of Sdk1 interactions in the inner retina and speculation of which Sdk1 presynaptic partners (amacrine or bipolar) are likely to be engaged or not.

(6) The authors need to include more details as to how the mean cell density analysis was done from the density recovery profiles as the current version of the manuscript lacks sufficient detail for this analysis.

Reviewer #3 (Recommendations for the authors):

This paper evaluates the role of sidekick-1, a recognition molecule that is part of the immunoglobulin superfamily. Sidekick-1 is expressed in interneurons and 3 previously-defined ganglion cell types, including the ON α, 2 ON-DS, and a novel edge detector S1-RGC. The elimination of the sidekick 1 has the most severe effects on the S1-RGC, causing its dendritic arbors to be less extensive, to receive fewer excitatory and inhibitory inputs, and in losing its orientation tuning. The knockout of sidekick-1 has less severe effects on the other ganglion cell types. This work fits into the great challenge of identifying cell types by molecular, functional, and morphological parameters. The authors have focused on one type of ganglion cell that expresses sidekick-1 and use calcium imaging, single cell recordings, and morphological assessment to demonstrate what this cell might do functionally and how it is perturbed in the absence of sidekick-1. The strength of this paper is the demonstration that elimination of Sidekick-1 has effects on the inputs to these ganglion cells. Perhaps the main weaknesses of this paper are (1) the S1 ganglion cell's functional classification as weakly tuned orientation selective and an edge detector has not been shown convincingly, and (2) how and why these cells are changed in the absence of Sidekick-1 remains speculative, yet the mechanism is attributed to sidekick-1's role in synapse formation, which has not been demonstrated directly.

Specific comments:

The authors show that high levels of Sdk2 in On-α, M2, and ON-DSGCs compensate for the loss of Sdk1 and attenuates phenotypes. To elucidate whether intact phenotypes are truly due to compensatory increase in Sdk2 expression or the functional properties of these 3 RGC types are independent of Sdk1, it would be helpful to understand the relative expression levels of Sdk1 and Sdk2 in Sdk1-expressing RGCs. Without such data, it is difficult to distinguish between lack of a role for Sdk-1 vs a compensatory effect of Sdk2 expression in the other cell types (α, ON DS).

The authors call these ganglion cells local edge detectors without clear evidence in Figure 2.

This system used by the authors has the advantage of gaining mechanistic insight into how this recognition molecule works in the development of a well-defined circuit, yet the authors fall short of providing this insight.

The authors should confirm if there are any morphological and physiological changes in Sdk1 heterozygotes compared to control retina. (Note from Reviewing editor: experiments proposed by Reviewer 2 would address this).

Figure 1: If possible, the authors also show the ganglion cell axon terminals in the LGN and SC labeled from the Cre expression in the eye. This is important to establish that the AAV-based retrograde labeling in Figure 4 is working as the authors think it is.

Figure 1 H-K, Figure 2H: The multiple channels are difficult to evaluate. The authors should show in the supplementary material the immunolabeling in individual channels.

It is unclear what about Figure 2E demonstrates that S1-RGCs are responding to local edges since there is not information about receptive field sizes yet. Figure 2 shows that the S1-RGCs have weak OS, are not DS, and have ON-OFF light responses. Only the information from Figure 4 reveals that these cells may be local edge detectors. (Note from Reviewing editor: experiments proposed by Reviewer 2 would address this).

Figure 3. Sholl analysis is not showing a striking difference between the control and Sdk-1 KO. The ON α ganglion cells also have differences at 3 locations with significance. The authors' claim of the relative stability of ON α is not compelling and should soften these claims.

The authors speculate that the orthogonal tunings of excitatory and inhibitory inputs contribute to orientation-selective properties of S1-RGCs. While excitatory inputs retain orientation-tuning properties, inhibitory inputs completely lose the orientation selectivity. To elucidate how the loss of Sdk1 affects inhibitory inputs from amacrine cells, it would be helpful to investigate the morphology of 3 kinds of Sdk1-ACs that the author discovered in KO mice. How does the loss of inhibitory orientation-selectivity affect spike output? As shown in the supplementary material, the most striking effect of the Sdk1 KO is the obliteration of inhibition to orientation selectivity. Can the authors better elucidate the contributions of each of these inputs to orientation selectivity? Clarification of these issues is needed.

The authors demonstrate that the two mostly likely partners of the S1 RGC is the wide field amacrine cell and the type 7 ON cone bipolar cell. Have the authors examined the interaction of these interneurons with the S1 RGC in the Sidekick-1 KO? Additional data would make this compelling but if that does not exist, the authors should soften these claims.

If the major change in the Sidekick-1 KO is a change in the dendritic field sizes of S1 RGCs, then the question remains whether the mosaic of these cells is disrupted in the KO through lack of homophilic homotypic interactions. The authors may want to soften these claims.

Although the authors state that the overall structure of IPL remain unchanged in KO retina, it would be helpful to see the quantification of depth of the entire retina or IPL to see if the loss of Sdk1 affects retinal structures.

Supplementary Figure 5 lacks quantification and demonstrates some differences in RBPMS, VGlut3, Ap2-α staining.

Supplementary Figure 5 lacks a diagram of lamination.

The authors posit that "Sdk1 promotes synapse formation between these neurons and their interneuron targets." To elucidate the roles of Sdk1 in synapse formation, it requires morphological investigation of excitatory and inhibitory synapses onto S1 ganglion cells or physiological investigation of mEPSCs and mIPSCs in control and KO retina. Without such data, it would be difficult to conclude that the role of Sdk1 is in synapse formation.

Lines 322-328. The authors land on synapse formation has the deficit in the Sidekick-1 KO; however, this is not apparent from the data. If the authors would like to make the main point about synapse formation, then quantifying synaptic puncta would be necessary. The authors could consider either changing their claim or providing additional data.

https://doi.org/10.7554/eLife.70870.sa1

Author response

Essential revisions:

(1) Include some additional data to compare the heterzygot Sdk1+/- to WT. Reviewer #2, is quite specific about what would be acceptable.

We now provide new data on Sdk1+ Onα-RGCs in the KCNG4-Cre line which shows that the anatomy and visual responses of these wild-type RGCs are like their counterparts in Sdk1 heterozygotes.

(2) Further characterization of S1- RGCs would strengthen the finding in that it could confirm whether they are indeed edge detecting cells.

As requested, we now provide new analysis of S1-RGC ON and OFF responses with detail about their symmetry and speed preference. Briefly, S1-RGC ON-OFF balance varies with stimulus size, with stimuli larger than their receptive field center producing stronger ON versus OFF responses. S1-RGCs also prefer bars moving at ~200µm/sec to those moving at ~1000µm/sec.

(3) If possible, it would be great to determine whether there is an upregulation of other Sdks in RGCs subtypes that are not impacted in Sdk1 ko. Along the same lines, all reviewers have asked for a larger discussion Sdks and related proteins and retinal wiring.

We now provide measurements of Sdk2 transcript levels in whole retinae taken from Sdk1-heterozygotes and Sdk1 knockouts. At this resolution, Sdk1 loss does not result in statistically significant increase in Sdk2 mRNA. We have incorporated this result with a broader text on Sdks and retinal wiring in the discussion.

(4) All reviewers have commented on mosaic analysis – their issues should be addressed.

We now provide extra detail regarding our analysis of the spatial distribution of Sdk1 RGCs, have modified our description of this method to reflect the limitations described in Keeley et al., 2020, and have included new mosaic analysis in Sdk1-KOs as requested by reviewer #3.

Reviewer #1 (Recommendations for the authors):

Below are a few comments that will hopefully clarify the presentation.

1. The one major issue I have is with Figure 1 H-K. The authors are strongly recommended to read a recent review (Keely, Elgen and Reese,JCN, 2020) arguing that density recover profiles that do not take into account soma size are not revealing mosaic organization and therefore this really cannot be used to "verify" they have a cell type. That said. I don't think the lack of the mosaic analysis will not significantly weaken their results -- the antibody staining + physiology is sufficient for indicating that they have different cell types.

We thank the reviewer for recommending this recent paper on the limitations of the density recovery profile (DRP) in assessment of cell type. We did not intend to use the DRP as our sole assessment of cell-type and agree with the reviewer that the combination of approaches we used (histology, anatomy, physiology) are sufficient to show different RGC types. We have now softened our language regarding the use of this measurement and have referenced this review. We have also clarified our figures to indicate that the grey lines show the density of all GFP labelled RGCs in the Sdk1-CreGFP line.

2. (Lines 108-114/Figure 1 T-U )- the authors want to argue that Ost+/Nefh- are M2-ipRGCs and so they look at projections to the MTN via anterograde labeling. However, it is unclear how they correlate anterograde projections with the Ost+/Nefh- RGCs in retina. This needs to be clarified.

We appreciate that Reviewer 1 has thought carefully about our labelling studies. In short, we cannot directly assign staining from the MTN to specific Sdk1 RGCs in the retina using anterograde labelling. Our goal here was to illustrate why retrograde labelling from the lateral geniculate nucleus (LGN) and superior colliculus (SC) preferentially labelled the Onα− and S1-RGCs. The presence of Sdk1-RGC axon labeling in MTN and OPN are consistent with the idea that Sdk1 labels non-image forming types that might project more strongly to these areas than they do to LGN/SC. We have clarified this portion of the text.

3. Labels in Figure 1are very hard to read. Also, I strongly recommend that the authors refer to the different RGC subtypes in subsequent figures as their better known subtypes (M2, α, ON-DSGC) rather than the immunipositive proteins. It was quite confusing. – I had to print out Figure 1V and use it several times to figure out what was going on in several figures.

We apologize to the reviewer for the difficult figure labels and use of immunopositive proteins to refer to Sdk1 RGCs rather than their subtype nomenclature. We have increased the size of the figure labels and have replaced immunopostive protein labels with RGC type wherever possible.

4. For Figure 2 – it was not clear when the authors were using GCaMP in all RGCs (expressed under vglut-Cre – line 163) or specific for Sdk1-Cre (line 138).

We apologize for our lack of clarity in this figure. We wanted to give readers a way to compare the ON-OFF index for each Sdk1-RGC group to the full RGC population. The grey lines in the On-Off index graphs shown in Figures 2B-E were obtained by calcium imaging GCaMP6f-expressing RGCs in retinae taken from the Vglut2-Cre/Rosa-LSL-GCaMP6f line. All other RGC data in this figure comes from Sdk1-RGCs infected with AAVs bearing Cre-dependent GCaMP6f. We have further clarified this in the text and the legends.

5. Figure 2 – is bar velocity 1000mm/sec (which I believe is the speed of an action potential!) or 1000 µm/sec?

We apologize for the careless mistake. The reviewer is correct the bars moved at 1000µm/sec.

6. Figure 2 – would be great if there were some quantification go figures H and I – are all Pcdh-Cre stained for Nr2f2? What percentage for Calb+?

We obtained a few retinae as a gift from Dr. K. Yonehara and stained them with Nr2f2 and Calb. The Pcdh9-Cre retinae we analyzed show ~1035cells/mm2 in the GCL and we now include the proportions of Nr2f2+/Calb+/Pcdh9-Cre+ cells in the text.

7. Figure 3 – there seems like a substantial loss of DS in the ON-DSGCs that the authors should note. Again, this broadens the impact of the study because clearly Sdk1 is important for this circuit, but the impact is more subtle. It is not that they need to study it here but it is important to emphasis the multiple impacts it has. Hence line is 220-221 are not quite accurate.

This is a good point. In an effort to be cautious about our interpretation we may have inadvertently downplayed phenotypes on Sdk1+ ON-DSGC types. We have altered lines 220-221 to incorporate this point.

Reviewer #2 (Recommendations for the authors):

The following concerns should be addressed by the authors:

(1) How close is the Sdk1 HET retina to a wildtype retina with two copies of Sdk1? Could the authors compare RGC responses they obtain in the Sdk1 HET to responses in a wildtype retina to enable the reader to grasp the magnitude of deficits upon Sdk1 deficiency? As the authors have done an incredible job of findings cell specific markers for each of the RGC types that express Sdk1, they could use these markers to correlate responses in HET vs wildtype retina. At the minimum, responses from easily identifiable RGC types like the ONalpha or the ON DS which the authors have a specific transgenic line for, could be recorded in retinas with two copies of Sdk1 to evaluate how similar these are compared to the responses (using the same stimuli and lighting conditions) obtained from HET retinas with a single Sdk1 copy. This data would be helpful to validate the use of the Sdk1 HET as a control.

The reviewer asks whether Sdk1-RGCs in heterozygotes show an intermediate phenotype relative to the same RGCs bearing two functional copies of Sdk1. We acknowledge the reviewer’s concern and provide recordings and dendritic reconstructions from ONα-RGCs genetically targeted using the KCNG4-Cre mouse line. These data are shown in Figure 4 —figure supplement 2 and show that the morphology and physiology of Onα-RGCs bearing two copies or one copy of Sdk1 are similar.

(2) Given that previous accounts of edge sensitive RGC types like the HD RGCs show that these cells are most responsive to specific speeds (Jacoby and Schwartz 2017), it would be worth determining if the S1-RGC type detailed in this work are also tuned to a specific speed of stimuli. This aspect of temporal tuning would aid the characterization of the response profile of the S1-RGC type. The authors would most likely already have this data as they have made a similar temporal tuning determination for the ON DS RGCs.

This is a good point. We have now provided a comparison of speed tuning for the S1-RGC and ONα-RGC obtained from our calcium imaging experiments. These results show that S1-RGC responses are stronger for bars moving at 200µm/sec versus bars moving at 1000µm/sec.

(3) Along the same lines, edge sensing RGCs can have ON and OFF responses that are not symmetrical (Jacoby and Schwartz 2017; Roska and Werblin 2001). As such it would add to the characterization of the S1-RGC response profile, if the authors could determine if the ON and OFF responses of the S1-RGC are symmetric/equivalent or not. From the examples shown in Figure 4 it appears as though the ON responses are more prominent than the OFF both for the expanding spot and orientation experiments i.e., more spikes allocated during light increments. Other edge sensing RGCs can have one component (ON or OFF) more dominant and thus it would be important for the reader to know whether the S1-RGC shows a similar asymmetry or not. Again, the authors already have this data so it should be straight forward to analyze and perhaps include plots where the ON and OFF responses (# of spikes) of the S1-RGCs are compared in the same plot.

We appreciate that Reviewer 2 has thought deeply about the response properties of S1-RGCs in relation to other edge-sensing RGC types. We now provide example raster plots of S1-RGC firing to a centered spot and full field flash, as well as graphs comparing their ON and OFF firing rate versus spot diameter. These data are now presented as Figure 4 —figure supplement 1 and show that S1-RGCs are ON-dominated if stimuli are larger than their receptive field center.

(4) The authors indicate that Sdk2 might be compensating for Sdk1 in the RGC types unaffected by Sdk1 deficiency. However, they do not provide evidence for whether or not Sdk2 is upregulated in the unaffected RGCs upon Sdk1 deficiency. The authors have generated antibodies to Sdk2 in their previous work which could be used to determine whether or not Sdk2 expression is upregulated amongst select RGC types upon Sdk1 deficiency (this could be compared in HET vs knockout).

This is a great idea, but difficult for a couple of reasons. First, the Sdk2 antibody stain appears as small puncta with no detectable signal in somata. Second, Sdk1/2 double positive and Sdk2-exclusive RGCs, such as W3B-RGCs, all ramify in the same region of the IPL. It would be hard to assign these puncta to Sdk1/Sdk2 double-positive RGCs and compare their intensity/number between controls and Sdk1 nulls. Instead, we have performed whole-retina real-time qPCR for Sdk2 from Sdk1-Hets and Sdk1-KOs. These results are shown in Figure 3 —figure supplement 1C and no significant elevation of Sdk2 in the absence of Sdk1. We have incorporated this result into our results and discussion on Sdks and IgSFs.

(5) The authors bring up the homophilic vs heterophilic interactions between IgSFs like Sdks in the Introduction section when discussing the role of IgSFs like Sdks for synapse formation but do not include a discussion of whether Sdk1 could be operating via homophilic or heterophilic interactions to set up synaptic connections in the inner retina. Given the Sdk1 expression profiles it appears that Sdk1 could operate via heterophilic interactions, which is different from how Sdk2 wires the VG3 amacrine cell-W3B RGC circuit and it would be of benefit to the reader to include a discussion of the mechanistic basis of Sdk1 interactions in the inner retina and speculation of which Sdk1 presynaptic partners (amacrine or bipolar) are likely to be engaged or not.

We have now expanded our discussion of the differences between Sdk1 and Sdk2 which includes some discussion about potential synaptic partners.

(6) The authors need to include more details as to how the mean cell density analysis was done from the density recovery profiles as the current version of the manuscript lacks sufficient detail for this analysis.

We apologize for the lack of detail. We have now provided a more thorough discussion of how mean cell density analysis was performed in the methods and clarified their presentation in the manuscript.

Reviewer #3 (Recommendations for the authors):

The authors show that high levels of Sdk2 in On-α, M2, and ON-DSGCs compensate for the loss of Sdk1 and attenuates phenotypes. To elucidate whether intact phenotypes are truly due to compensatory increase in Sdk2 expression or the functional properties of these 3 RGC types are independent of Sdk1, it would be helpful to understand the relative expression levels of Sdk1 and Sdk2 in Sdk1-expressing RGCs. Without such data, it is difficult to distinguish between lack of a role for Sdk-1 vs a compensatory effect of Sdk2 expression in the other cell types (α, ON DS).

The authors call these ganglion cells local edge detectors without clear evidence in Figure 2.

We have altered the title of this segment to reflect the data shown up until Figure 2. It now reads:

“Brn3c+ RGCs are ON-OFF RGCs that respond to bars travelling along the same axis”

This system used by the authors has the advantage of gaining mechanistic insight into how this recognition molecule works in the development of a well-defined circuit, yet the authors fall short of providing this insight.

The authors should confirm if there are any morphological and physiological changes in Sdk1 heterozygotes compared to control retina. (Note from Reviewing editor: experiments proposed by Reviewer 2 would address this).

The reviewer asks whether Sdk1 heterozygotes show an intermediate phenotype relative to animals bearing two functional copies of Sdk1. We acknowledge the reviewer’s concern and provide recordings and dendritic reconstructions from ONα RGCs genetically targeted using the KCNG4-Cre mouse line. These data are shown in Figure 4 —figure supplement 2 and show that the morphology and physiology of Onα-RGCs bearing two copies or one copy of Sdk1 are similar.

Figure 1: If possible, the authors also show the ganglion cell axon terminals in the LGN and SC labeled from the Cre expression in the eye. This is important to establish that the AAV-based retrograde labeling in Figure 4 is working as the authors think it is.

We agree with the reviewer that LGN/SC sections showing Sdk1-RGC terminals would reinforce our method to retrogradely label Sdk1 RGCs. We have now grouped this data along with our other brain projection data in Figure 1 —figure supplement 3.

Figure 1 H-K, Figure 2H: The multiple channels are difficult to evaluate. The authors should show in the supplementary material the immunolabeling in individual channels.

We apologize for the difficult presentation of this data. We now show the immunolabelling in individual channels in a new supplemental figure (Figure 1 —figure supplement 2).

Agreed. This sentence was an undetected fragment from an earlier draft of this manuscript. Please see our response to point#5.

Figure 3. Sholl analysis is not showing a striking difference between the control and Sdk-1 KO. The ON α ganglion cells also have differences at 3 locations with significance. The authors' claim of the relative stability of ON α is not compelling and should soften these claims.

It is a good point. In our effort to be cautious about interpreting the effects of Sdk1 loss we may have inadvertently downplayed potential effects on the Onα-RGC. We have now softened our claims about the absence of phenotypes from the ONα-RGC and included new discussion of potential functional phenotypes on the ON-DSGCs in the Sdk1-KO as recommended by reviewer 1.

The authors speculate that the orthogonal tunings of excitatory and inhibitory inputs contribute to orientation-selective properties of S1-RGCs. While excitatory inputs retain orientation-tuning properties, inhibitory inputs completely lose the orientation selectivity. To elucidate how the loss of Sdk1 affects inhibitory inputs from amacrine cells, it would be helpful to investigate the morphology of 3 kinds of Sdk1-ACs that the author discovered in KO mice. How does the loss of inhibitory orientation-selectivity affect spike output? As shown in the supplementary material, the most striking effect of the Sdk1 KO is the obliteration of inhibition to orientation selectivity. Can the authors better elucidate the contributions of each of these inputs to orientation selectivity? Clarification of these issues is needed.

The reviewer wants to know how loss of inhibitory input from Sdk1-ACs causes the visual response deficits seen in S1-RGCs. This is a great idea, but a direct test is methodologically not possible since we cannot perturb the Sdk1-ACs independently of the Sdk1-RGCs and assess the consequences on S1-RGC function. We are in the process of generating intersectional reagents to specifically label the S1-RGC and Sdk1 interneurons to better characterize their visual responses and potential synaptic interactions. These experiments are still in early stages and are months away from completion. We now bring up this important idea in our revised discussion.

The authors demonstrate that the two mostly likely partners of the S1 RGC is the wide field amacrine cell and the type 7 ON cone bipolar cell. Have the authors examined the interaction of these interneurons with the S1 RGC in the Sidekick-1 KO? Additional data would make this compelling but if that does not exist, the authors should soften these claims.

We too are excited by the prospect that the likely partners of S1-RGCs would be the sublamina-3 targeting wide-field amacrine and type 7 bipolar cells. In short, we have not examined their interaction because we currently cannot access these neurons without also affecting the Sdk1-RGCs. An intersectional approach might be feasible but will require months of development and troubleshooting to deploy. We have softened these claims.

If the major change in the Sidekick-1 KO is a change in the dendritic field sizes of S1 RGCs, then the question remains whether the mosaic of these cells is disrupted in the KO through lack of homophilic homotypic interactions. The authors may want to soften these claims.

It is a good point, we now include mosaic analysis of Onα- and S1-RGCs in Sdk1-KOs and compare them to those found in Sdk1-Hets in Figure 5 —figure supplement 3. Density recovery profiles of both RGCs in Sdk1 nulls resemble those found in controls.

Although the authors state that the overall structure of IPL remain unchanged in KO retina, it would be helpful to see the quantification of depth of the entire retina or IPL to see if the loss of Sdk1 affects retinal structures.

Supplementary Figure 5 lacks quantification and demonstrates some differences in RBPMS, VGlut3, Ap2-α staining.

We apologize for this careless omission and poor choice of representative crossections. We have updated the panels in this figure showing that gross laminar morphology in Sdk1 nulls resembles that found in controls.

Supplementary Figure 5 lacks a diagram of lamination.

We apologize for this careless omission and indicate IPL and nuclear layer landmarks for these panels.

The authors posit that "Sdk1 promotes synapse formation between these neurons and their interneuron targets." To elucidate the roles of Sdk1 in synapse formation, it requires morphological investigation of excitatory and inhibitory synapses onto S1 ganglion cells or physiological investigation of mEPSCs and mIPSCs in control and KO retina. Without such data, it would be difficult to conclude that the role of Sdk1 is in synapse formation.

We agree. Our data show a deficit in functional synaptic input on S1-RGCs and a structural correlate in the form of reduced dendritic arborization. Sdk1 could be involved in promoting or stabilizing synapses, dendrites, or both. We have removed this phrase from the results and have better balanced these alternatives in the discussion.

Lines 322-328. The authors land on synapse formation has the deficit in the Sidekick-1 KO; however, this is not apparent from the data. If the authors would like to make the main point about synapse formation, then quantifying synaptic puncta would be necessary. The authors could consider either changing their claim or providing additional data.

https://doi.org/10.7554/eLife.70870.sa2

Article and author information

Author details

  1. Pierre-Luc Rochon

    Department of Physiology, McGill University, Montreal, Canada
    Contribution
    Conceptualization, Formal analysis, Investigation, Methodology, Visualization, Writing - original draft, Writing - review and editing
    Competing interests
    No competing interests declared

  2. Catherine Theriault

    Department of Physiology, McGill University, Montreal, Canada
    Contribution
    Investigation, Methodology, Writing - review and editing
    Competing interests
    No competing interests declared

  3. Aline Giselle Rangel Olguin

    Department of Physiology, McGill University, Montreal, Canada
    Contribution
    Formal analysis, Investigation, Methodology, Writing - review and editing
    Competing interests
    No competing interests declared

  4. Arjun Krishnaswamy

    Department of Physiology, McGill University, Montreal, Canada
    Contribution
    Conceptualization, Funding acquisition, Investigation, Methodology, Project administration, Supervision, Visualization, Writing - original draft, Writing - review and editing
    For correspondence
    arjun.krishnaswamy@mcgill.ca
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0002-7706-4657

Funding

Canadian Institutes of Health Research (project grant)

  • Arjun Krishnaswamy

Natural Sciences and Engineering Research Council of Canada (discovery grant)

  • Arjun Krishnaswamy

Canadian Institutes of Health Research (Graduate Fellowship)

  • Pierre-Luc Rochon

Fonds de Recherche du Québec - Santé (Graduate Fellowship)

  • Aline Giselle Rangel Olguin

Consejo Nacional de Ciencia y Tecnología (Graduate Fellowship)

  • Aline Giselle Rangel Olguin

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

Acknowledgements

We thank K Yonehara for gifting us fixed retinal tissue from PCDH9-Cre mice and M Cayouette and C Jolicoeur for advice, help, and reagents for Sdk2 RT qPCR studies. We also thank Drs J Sanes, E Cooper, S Trenholm, K Yonehara, E Cook, and E Feinberg for helpful comments and suggestions on our manuscript.

Ethics

Animals were used in accordance with the rules and regulations established by the Canadian Council on Animal Care and protocol (2017-7889) was approved by the Animal Care Committee at McGill University.

Senior Editor

  1. Gary L Westbrook, Oregon Health and Science University, United States

Reviewing Editor

  1. Marla B Feller, University of California, Berkeley, United States

Version history

  1. Received: June 1, 2021
  2. Preprint posted: June 9, 2021 (view preprint)
  3. Accepted: September 11, 2021
  4. Accepted Manuscript published: September 21, 2021 (version 1)
  5. Version of Record published: October 13, 2021 (version 2)

Copyright

© 2021, Rochon 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

  • 1,105
    Page views
  • 127
    Downloads
  • 3
    Citations

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

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

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

  1. Pierre-Luc Rochon
  2. Catherine Theriault
  3. Aline Giselle Rangel Olguin
  4. Arjun Krishnaswamy
(2021)
The cell adhesion molecule Sdk1 shapes assembly of a retinal circuit that detects localized edges
eLife 10:e70870.
https://doi.org/10.7554/eLife.70870

Categories and tags

Research organism

Further reading

    1. Neuroscience

    Two human brain systems micro-structurally associated with obesity

    Manfred G Kitzbichler, Daniel Martins ... Neil A Harrison
    Research Article Updated

    The relationship between obesity and human brain structure is incompletely understood. Using diffusion-weighted MRI from ∼30,000 UK Biobank participants, we test the hypothesis that obesity (waist-to-hip ratio, WHR) is associated with regional differences in two micro-structural MRI metrics: isotropic volume fraction (ISOVF), an index of free water, and intra-cellular volume fraction (ICVF), an index of neurite density. We observed significant associations with obesity in two coupled but distinct brain systems: a prefrontal/temporal/striatal system associated with ISOVF and a medial temporal/occipital/striatal system associated with ICVF. The ISOVF~WHR system colocated with expression of genes enriched for innate immune functions, decreased glial density, and high mu opioid (MOR) and other neurotransmitter receptor density. Conversely, the ICVF~WHR system co-located with expression of genes enriched for G-protein coupled receptors and decreased density of MOR and other receptors. To test whether these distinct brain phenotypes might differ in terms of their underlying shared genetics or relationship to maps of the inflammatory marker C-reactive Protein (CRP), we estimated the genetic correlations between WHR and ISOVF (rg = 0.026, P = 0.36) and ICVF (rg = 0.112, P < 9×10−4) as well as comparing correlations between WHR maps and equivalent CRP maps for ISOVF and ICVF (P<0.05). These correlational results are consistent with a two-way mechanistic model whereby genetically determined differences in neurite density in the medial temporal system may contribute to obesity, whereas water content in the prefrontal system could reflect a consequence of obesity mediated by innate immune system activation.

    1. Neuroscience

    Human attention during goal-directed reading comprehension relies on task optimization

    Jiajie Zou, Yuran Zhang ... Nai Ding
    Research Article

    The computational principles underlying attention allocation in complex goal-directed tasks remain elusive. Goal-directed reading, that is, reading a passage to answer a question in mind, is a common real-world task that strongly engages attention. Here, we investigate what computational models can explain attention distribution in this complex task. We show that the reading time on each word is predicted by the attention weights in transformer-based deep neural networks (DNNs) optimized to perform the same reading task. Eye tracking further reveals that readers separately attend to basic text features and question-relevant information during first-pass reading and rereading, respectively. Similarly, text features and question relevance separately modulate attention weights in shallow and deep DNN layers. Furthermore, when readers scan a passage without a question in mind, their reading time is predicted by DNNs optimized for a word prediction task. Therefore, we offer a computational account of how task optimization modulates attention distribution during real-world reading.

    1. Neuroscience

    Eelbrain, a Python toolkit for time-continuous analysis with temporal response functions

    Christian Brodbeck, Proloy Das ... Jonathan Z Simon
    Tools and Resources

    Even though human experience unfolds continuously in time, it is not strictly linear; instead, it entails cascading processes building hierarchical cognitive structures. For instance, during speech perception, humans transform a continuously varying acoustic signal into phonemes, words, and meaning, and these levels all have distinct but interdependent temporal structures. Time-lagged regression using temporal response functions (TRFs) has recently emerged as a promising tool for disentangling electrophysiological brain responses related to such complex models of perception. Here we introduce the Eelbrain Python toolkit, which makes this kind of analysis easy and accessible. We demonstrate its use, using continuous speech as a sample paradigm, with a freely available EEG dataset of audiobook listening. A companion GitHub repository provides the complete source code for the analysis, from raw data to group level statistics. More generally, we advocate a hypothesis-driven approach in which the experimenter specifies a hierarchy of time-continuous representations that are hypothesized to have contributed to brain responses, and uses those as predictor variables for the electrophysiological signal. This is analogous to a multiple regression problem, but with the addition of a time dimension. TRF analysis decomposes the brain signal into distinct responses associated with the different predictor variables by estimating a multivariate TRF (mTRF), quantifying the influence of each predictor on brain responses as a function of time(-lags). This allows asking two questions about the predictor variables: 1) Is there a significant neural representation corresponding to this predictor variable? And if so, 2) what are the temporal characteristics of the neural response associated with it? Thus, different predictor variables can be systematically combined and evaluated to jointly model neural processing at multiple hierarchical levels. We discuss applications of this approach, including the potential for linking algorithmic/representational theories at different cognitive levels to brain responses through computational models with appropriate linking hypotheses.