Fig 1.
Morphology and light responses of mouse alpha retinal ganglion cells (αRGCs).
A-D, Sample neurons with whole-mount views (outer images) and responses to a flashing spot (inner plots) for the Off-sustained (A), Off-transient (B), On-sustained (C), and On-transient (D) types. Raster graph illustrates action potentials on repeated trials of a spot flashing on (white background) and off (gray) every 2 s. Continuous curve is average firing rate over 10 or more trials.
Fig 2.
Light response kinetics define four physiological types of αRGCs.
A: Time course of the firing rate during flashing spot experiments (as in Fig 1), normalized to the peak rate for each cell, and sorted by response type. Results from individual cells (faint lines) and their mean (bold). B: Scatter plot of response parameters for all αRGCs analyzed. For each cell the time course was approximated with an exponential decay (see inset). The abscissa shows the time constant of the decay, and the ordinate plots the ratio of final value to peak value of the firing rate. Among 91 alpha cells recorded by this targeting method we encountered 26% Off-t, 30% Off-s, 13% On-t, and 22% On-s.
Fig 3.
Confirmation that On-transient KCNG4-cre neurons are a separate αRGC type.
A-D: On-t cells express heavy neurofilament. A loose patch recording of a fluorescent neuron (A) revealed a transient response of the firing rate to light steps (B). After fixation and antibody staining one can identify the same cell based on YFP label (C) and confirm that it is strongly labeled with the neurofilament antibody SMI-32 (D). E-F: Two fluorescent neurons in close proximity (black and green arrowheads in E) showed sustained (black) and transient (green) response of the firing rate to a light step (F).
Fig 4.
The four αRGC types stratify their dendrites in distinct layers of the IPL.
A: Stratification of the 4 neurons from Fig 1. Each histogram indicates the depth distribution of fluorescent voxels in the dendrites of one cell relative to the two ChAT bands in the inner plexiform layer (at 0 and 12 μm depth). B: Stratification and dendritic diameter of many αRGCs, visualized either using the KCNG4-Cre line (all types) or the W7 transgenic line (Off types). For each cell, the stratification level is the mean of the histogram computed as in panel a. C-D: Total dendritic length (C) and soma diameter (D) for the four αRGC types; mean ± SEM (n = 7, 4, 6, 5 left to right in each bar graph).
Fig 5.
Spatio-temporal responses of the four αRGC types.
A: Responses of a sample On-s cell to flashing spots of increasing radius (see inset color scale). Firing rate averaged over 6 repeats. B: Peak firing rate as a function of spot size from experiments such as in panel a. Curves were normalized to the maximal rate for each cell, then averaged over all cells of the same type. The error bars on the last data point are representative for SEM throughout the curve. C: From the spot stimulus giving the strongest response in panel a, one derives the latency to peak firing (L), the peak firing rate (P), the final firing rate (F) and the exponential decay time (T), as indicated in this schematic. D: From the measurements of panel B, one defines the baseline firing rate (B), the spot size producing the maximal rate (C) and the response to large uniform stimuli (U), as indicated in this schematic. E-K: Response parameters of all 4 cell types. Mean ± SEM over all cells of the same type (n = 28, 32, 8, 22 left to right in all bar graphs). Based on the measures from panels c and d: Latency = L; Peak rate = P; Center diameter = C; Surround strength = 1-(U-B)/(P-B); Decay time = T; Final/Peak response = (F-B)/(P-B); Baseline rate = B.
Fig 6.
Nonlinear subunits and directional processing.
A-C: Tests for nonlinear summation within the receptive field. The stimulus was a 400 μm-diameter spot centered on the receptive field, filled with a stripe grating that contrast-reversed every 1 s. For stripes of width 200 μm or less, a stripe boundary passed through the spot center. A: A sample RGC fires a burst of spikes on every grating transition unless the stripe width drops below a threshold, here 25 μm. B: Peak firing rate as a function of stripe width, normalized to the response to the uniform spot (400 μm); mean ± SEM across cells of each type. C: Threshold stripe width, an estimate of subunit size; mean ± SEM across cells of each type (n = 22, 23, 8, 20 left to right in panels B and C). All alpha types except Off-s show nonlinear summation over subunits ~30 μm in size. D-E: Tests of direction selectivity. D: Firing of a sample RGC in response to a 250 μm diameter spot of the preferred polarity moving through the receptive field center at 700 μm/s in 8 directions spaced at 45° (different colors). E: Direction selectivity index computed from such responses as where φk is the direction of motion of the k-th stimulus, and Pk is the peak firing rate evoked by that stimulus. Mean ± SEM across cells of each type (n = 3, 3, 7, 3 left to right).
Fig 7.
All αRGC types share a distinctive spike shape.
A: The spike waveforms of seven ganglion cells, each averaged over many hundreds of spikes. These are representatives of the four αRGC types and three other identified types marked in transgenic lines: J-RGCs (J [31]), upward-coding On-Off DS cells (Hb9 [32]), and W3-RGCs (W3 [33]). All waveforms are aligned on the point with maximum time derivative. B: Waveform analysis of spikes from 50 RGCs of the types introduced in panel A. We computed the time derivative of each waveform, then subjected this set to a principal components analysis, and plotted the coefficients along the first two components (which accounted for 67% of the variance). Each point is one RGC’s waveform. Dotted line separates the αRGCs from all other RGCs, with just one exception. Corners of the plot are marked with the spike waveforms (width 2 ms) corresponding to those points in principal components space. C: Spike width—defined as the time between points of minimum and maximum slope of the action potential—for cells of the different types. Bars indicate Mean ± SEM for each cell type (n = 16, 12, 8, 18, 8, 8, 3 left to right). Two outliers (marked with open symbols in panels B and C) were excluded from this spike width analysis.
Fig 8.
Molecular distinctions among αRGCs.
A, B: Retinal whole mounts were stained with antibodies to Opn (green), plus antibodies to one of 3 POU-domain transcription factors (Brn3a, Brn3b or Brn3c; red in A) or one of 3 calcium binding proteins (parvalbumin [PV], calbindin or calretinin; red in B). Brn3b and PV mark most αRGCs whereas Brn3a, Brn3c and calbindin mark subsets; most αRGCs are calretinin-negative. C: Whole mounts of YFP-H retina were quadruply stained for YFP, Opn, vAChT and the indicated marker. Cells that were Opn, YFP, and marker triple-positive were identified (green, cyan and red, respectively in top panels) and imaged (YFP only, middle panels). Stratification of YFP-positive dendrites was then determined with reference to that of starburst amacrines (vAChT-positive, red in bottom panels). Arrows point to the same cell displayed in each panel. Results are representative of 7–10 cells per type from 5 mice. Scale bar = 50 μm.
Fig 9.
Existing transgenic lines label subsets of αRGCs.
A: Morphology and physiology of YFP-positive cells in the TYW7 line. Structure and function were assessed as in Fig 1. The cell shapes (top: whole mount view, bottom: vertical projection including ChAT label) and the light responses (firing rate under periodically flashing spot) identify these as Off-s (left) and Off-t (right) αRGCs. Of n = 10 cells recorded in this line, 5 were Off-s and 5 were Off-t. B: Use of the cre-off feature of the TYW3 and TYW7 lines: YFP is flanked by lox sites in these lines, so it is excised in cells that also express cre. C: YFP disappears with age from retinas in the TYW7; Kcng4-cre double transgenics at the indicated postnatal ages. Therefore the TYW7 line labels a subset of αRGCs. D: YFP persists in TYW3; Kcng4-cre double transgenics. Therefore the TYW3 line labels a set of non-αRGCs. E: Triple transgenic strategy to label both Kcng4-cre neurons (RFP) and CB2-GFP neurons (GFP). F: In these triple transgenics, the CB2-GFP-positive RGCs also express Kcng4-cre and osteopontin (Opn). Therefore the CB2 line labels a subset of αRGCs. Scale bars: (A) 20 μm, (C and D) 200 μm, (F) 20 μm.
Fig 10.
Mosaic organization of αRGC types.
Three types of αRGCs were distinguished in retinal whole mounts by the molecular markers indicated above the top panels (see text and Fig 8 for details; asterisks show cells of the indicated type). The TYW7 line was used instead of Brn3a to label all Off αRGCs, because of species incompatibilities of antibodies; the Off-s αRGCs are W7 positive but Brn3c negative. For the On-t cells, we lacked a combination of markers to label them with sufficient reliability. A density recovery profile (DRP) analysis was then performed on each αRGC type (bottom panels; n = 3–5 retinas per type from 3 mice). The prominent dip in density at short distances is characteristic of “repulsion” between cell bodies of the same type. Dashed line, normalized average density. Mean ± SEM across cells of each type. Scale bar = 50 μm.
Fig 11.
Morphological, physiological and molecular properties distinguish four αRGC types.
A summary of morphological, physiological and molecular features for the four types of αRGCs, as reported in Figs 2, 4 and 8–10. Dotted box indicates dim labeling. See Discussion.