Genetic strategies target Bhlhe22-expressing amacrine cells and cells that co-express Bhlhe22 and KOR (B/K).

A. Intersectional genetic strategy. A KOR-Cre allele, acting on Loxp sites, combined with the Bhlhe22-Flpo allele, acting on Frt sites, removes two stop sequences and induces the expression of a reporter gene (either ReaChR-mCitrine fusion protein or tdTomato), driven by the CAG promoter. B. Bhlhe22 transcript readings in previously identified groups of amacrine cells (Yan et al., 2020). Data points show the median Bhlhe22 transcript reading and the inter-quartile range (IQR) of these readings within each group. Cell groups are identified as GABAergic, glycinergic, both or neither, according to the definitions in the original study, and groups that correspond to known cell types are labeled with the name of the type (e.g., AII or A17 amacrine cell) or a characteristic gene name (e.g., VIP+ or nNOS+ amacrine cells). C. Intersection between KOR-Cre and Bhlhe22-Flpo labels amacrine cell processes primarily in two layers (green rectangles, left) relative to calretinin bands (blue arrows; bracket marks the IPL). Top and bottom calretinin bands are identical to the OFF and ON ChAT bands, respectively. B/K amacrine cells were labeled with Bhlhe22 antibody (yellow arrows); the antibody also labeled amacrine cells without reporter expression (yellow arrowheads). D. Whole-mount retina showing B/K amacrine cells in the GCL with Bhlhe22 antibody labeling (arrows). E. Same as E. for the INL, which shows a higher density of labeled B/K somas. F. Density of B/K cells in different layers (n = 5 retinas) based on counts within 0.102 mm2 regions. Error bars indicate ± SEM across samples. G. Fluorescence profile of B/K intersectional genetic labeling relative to the ChAT bands in eight individual z-stacks (top). The average (bottom) shows fluorescence relative to the layers where ON and OFF bipolar cells stratify their axon terminals. H. Example images of long, straight B/K amacrine cell processes in ON and OFF layers of the IPL and the boundary between these layers.

B/K amacrine cells have long, unbranched dendrites and stratify at distinct levels of the inner plexiform layer.

A1. Confocal image (left) and tracing (right) of a B/K wide-field amacrine cell (WAC) in the ON layer. A2. Four additional ON B/K WACs. Cell at far right stratified between the ChAT bands. B1, B2. Same as A for OFF B/K WACs. Scale bar in A applies to all cells in A and B. Lower-case letters in A and B correspond to data in C. C. Fluorescence profiles of labeled dendrites of example ON and OFF WACs in A and B. Stratification is shown within each cell relative to ChAT bands labeled in the same tissue. Stratification is normalized to the inner/ON ChAT band (0, ChATON) and the outer/OFF ChAT band (1, ChATOFF). D. Population analysis (n = 55 cells) showing the number of dendrites relative to the stratification, normalized to the ChAT bands. Apparent groups of cells are shown with outlines. Soma position is indicated by the symbol. E. A subset of B/K WAC dendrites were visualized near the soma (within ∼150 µm) and at a distal location (∼500 µm from the soma). Dendrites remained in their original layer, as indicated by the points falling near the identity line. Symbols indicate soma location, as in D.

B/K amacrine cells comprise ON and OFF types and do not fire action potentials.

A. Membrane potential recording of response to a contrast-reversing spot stimulus (1.5-mm diameter) for three B/K WACs, including two OFF cells with somas in either the GCL or INL and an ON cell with a soma in the GCL. B. Responses to contrast-modulated spots of varying diameters. Response amplitudes to the dark (OFF) and bright (ON) phases of the spot distinguish ON and OFF B/K WACs. Error bars in G. and I. indicate ± SEM across cells. The primary response (depolarization for OFF cells, hyperpolarization for ON cells) asymptotes at ∼1 mm diameter. C. Resting membrane potential of B/K WACs in the presence of a mean luminance. OFF B/K WACs in the GCL were more hyperpolarized than OFF B/K WACs in the INL (t(12) = 2.66, p = 0.02) or ON B/K cells in the GCL (t(19) = 4.64, p < 0.001), whereas OFF B/K cells in the INL and ON B/K cells in the GCL were more similar (t(21) = 1.84, p = 0.08, two-sample t-tests). D. Peak depolarization to the preferred stimulus versus peak hyperpolarization to the non-preferred stimulus. Cells with equal modulation around the resting potential would fall along the dashed line. The combined group of OFF B/K WACs had larger depolarizing responses to preferred contrast compared to ON B/K WACs (t(27) = 5.29, p < 0.001), whereas ON B/K WACs had relatively larger hyperpolarizing responses to non-preferred contrast (t(27) = 4.63, p < 0.001).

Population dendritic imaging reveals orientation-selective B/K dendrites at different layers in the inner plexiform layer (IPL).

A. Example imaging planes for ON (top) and OFF (bottom) layers. Each plan shows a local correlation images over time. Example regions of interest (ROIs) are rectangles whose long axis aligns with the orientation of the dendrite. Corresponding baseline-subtracted fluorescence traces (in arbitrary units, a.u.) and polar plots of orientation tuning are shown on the right. B. Spatial footprints tuned to four orientations extracted by sparse non-negative matrix decomposition for the exemplary ON (top) and OFF (bottom) layers shown in A. C. Orientation selectivity index (OSI) of dendritic ROIs (n = 135) at different IPL depths. D. OSI versus Direction-selective index (DSI) of B/K dendrites at different depths of the IPL. The smoothed probability density/kernel density of ROI OSI and DSI are plotted at the margins. OSI (0.75 ± 0.02; n = 135) is significantly larger than DSI (0.22 ± 0.02; t(134) = 20.91, p < 0.001). E. Physical orientation versus preferred orientation of B/K dendrites with OSI > 0.4 at different depths of the IPL (n = 126 ROIs). The circular Pearson correlation between physical and preferred orientations is 0.91 (p < 0.001).

B/K amacrine cells make synapses with non-orientation tuned alpha/delta-type RGCs.

A. Dendritic tree of a recorded OFF Delta RGC (also known as a Sustained OFF Alpha RGC). Image shows average fluorescence in a confocal stack. Dashed square shows region in B. The RGC was filled with Lucifer Yellow (LY) during whole-cell recording, which was subsequently amplified with LY primary antibody and a red secondary antibody. B. Single confocal section showing OFF Delta RGC dendrites from A. relative to B/K WAC dendrites, labeled by the Cre/Flpo-dependent ReaChR-mCitrine reporter. C. Optogenetic stimulation (5.3 x 1017 Q s-1 cm-2) of ReaChR-expressing B/K WACs caused an inhibitory postsynaptic current (IPSC, black) in an OFF Delta RGC. The IPSC (133 ± 46 pA; n = 11) was mostly blocked by gabazine (50 µM SR95531; 46 ± 6 pA; n = 7; difference of 106 ± 25 pA, t(6) = 4.22, p = 0.0056) and completely blocked by subsequent addition of strychnine (1 µM; −0.33 ± 1.17 pA; n = 7; difference of 46 ± 5 pA, t(6) = 8.68, p < 0.001). Error bars in all panels indicate ± SEM across cells. Optogenetic experiments were performed in the presence of glutamate receptor blockers (see Methods) to prevent photoreceptor contributions to the light response. D. Increasing optogenetic stimulation caused increasing IPSC amplitude in OFF Delta RGCs. E-H. Same as A-D. for ON Alpha RGCs (also known as Sustained ON Alpha RGCs). The ON Alpha RGC IPSC (231 ± 32 pA; n = 19) was mostly blocked by gabazine (38 ± 12 pA; n = 12; difference of 214 ± 41 pA, t(11) = 5.23 p < 0.001) and completely blocked by strychnine (−2.9 ± 0.8 pA, n = 8; difference of 58 ± 14 pA, t(7) = 4.24; p = 0.0038).

Connectomic analysis uncovers synaptic connections between B/K WACs and bipolar cells.

A. SBEM sections from dataset k0725 show connectivity between a WAC (green), an OFF cone bipolar cell terminal (CB; red), and an OFF Delta RGC (yellow). Also shown are an unidentified amacrine cell (AC) and an AII amacrine cell (AC). Cell-type identification was determined from reconstructions in a previous study (Grimes et al., 2022). The WAC makes synapses (green arrows, middle and right panels) onto both an OFF bipolar cell terminal and the RGC dendrite that is postsynaptic to the OFF bipolar cell terminal, the signature of sign-inverted feedforward inhibition (SIFI). Additional BC synapses (red arrows) and non-WAC AC synapses (black arrows) are indicated. B. OFF Delta RGC with two amacrine cell types that also make SIFI synapses. Reconstructions (side view) are from a previous study (Grimes et al., 2022). Ci. Sample WAC neurites (top-down view) with SIFI synapses. Contacts are color-coded per the text key. Cii. Some WACs (black lines) exhibit dense synaptic connectivity; whereas others (red lines) have sparse connectivity. Di. Side view of the reconstruction of DAB+ processes in the OFF layer adjacent to the INL (S1) of a B/K dAPEX2 retina SBEM dataset. Somas (black circles) demarcate the INL/IPL boundary. Dii. Labeled somas descended beyond the volume towards the ON layer; none of the reconstructed dendrites in the OFF layer were connected to a soma within this small volume. Ei. Sample of reconstructed B/K WAC neurites (top-down view) with synapses color-coded per text key. Eii. Connections between two B/K WACs (green and blue dendrites) and an RGC (orange) from circled area in Ei. F. Partial reconstruction of a cone bipolar (CB) cell terminal with WAC dendrites that make inputs (green) and receive outputs (magenta) from the CB. Some dendrites are only presynaptic to the CB (†), whereas others are both pre- and postsynaptic (*). G. SBEM: left, OFF cone bipolar (CB, red) ribbon-type synapse (red arrows) to RGC (orange) and to DAB+ WAC (green). Center, DAB+ WAC inhibitory feedback synapse to CB (green arrow). Right, inhibitory synapses (green arrows) from DAB+ WAC to RGC and to another DAB+ WAC.

Proposed circuit function of B/K WACs

A. A mosaic of amacrine cells can provide coverage to a region of retina (square) with either tiled processes of seven narrow-field cells or straight dendrites projecting from the same number of WACs. B. Either selective or non-selective integration of OS-tuned WAC dendrites can result in either OS-tuned or non-OS-tuned postsynaptic neurons. C. Schematic diagrams for feed-forward inhibition and sign-inverted feed-forward inhibition (SIFI).

B/K amacrine cells make weak connections with certain retinal ganglion cell types

A. Dendritic tree of a recorded OFF Alpha RGC (also known as a Transient OFF Alpha RGC). Image shows average fluorescence in a confocal stack. Dashed square shows region in B. The RGC was filled with Lucifer Yellow (LY) during whole-cell recording, which was subsequently amplified with LY primary antibody and a red secondary antibody. B. Single confocal section showing OFF Alpha RGC dendritic tree relative to B/K WACs (Cre/Flpo-dependent ReaChR-mCitrine reporter). C. OFF Alpha RGCs had weaker IPSCs relative to OFF Delta and ON Alpha RGCs (Fig. 5). Responses were only weakly blocked by gabazine; in one cell, there was an inward current following addition of strychnine (inset). D. ON-OFF Direction-Selective RGCs showed weak IPSCs following optogenetic stimulation of B/K cells.

Studying labeled B/K amacrine cells with a scanning block face electron microscopy (SBEM) data set

A. Relative size and depth of the two SBEM datasets: an existing large dataset (k0725) and a smaller new data set with labeled B/K processes (B/K dApex2). The B/K dApex2 dataset focused on the OFF layer. B. There was an apparent bias to observe higher synapse density in short segments in the new dataset. We analyzed segments greater than 10 µm in length. C. Neurites in layer S1 had similar densities of input and output synapses in the new data set (B/K dApex2) and the subset of straight dendrites with dense synapses in the existing data set (k0725; boxed region).