Figure 4. Anatomy and function of input from VG3-ACs to W3-RGCs.

(A) Orthogonal projections of a confocal image stack through a representative W3-retinal ganglion cell (RGC) labeled biolistically with cyan fluorescent protein (CFP). W3-RGCs were identified by their characteristic morphology (Kim et al., 2010; Zhang et al., 2012) with small dendritic fields (territory size: 10,783 ± 409 μm², n = 25) filled by densely branched neurites stratifying in the center of the IPL with a secondary arborization near the border between the inner plexiform and inner nuclear layers (INLs). The fluorescent signal is colored to represent depth in the IPL. Inset bar graph shows the mean ± SEM territory size of W3-RGCs (n = 15) measured as the areas of the smallest convex polygons to encompass their arbors in a z-projection. (B, C) Overview projections (B) and single plane excerpts (C) of a W3-RGC biolistically labeled with CFP (red) and PSD95-YFP (green) in a VG3-Cre Ai9 mouse (tdTomato shown in blue). (D) Summary data indicating the fraction of PSD95-YFP puncta apposed by VG3-AC boutons (‘Materials and methods’) in the obtained images (real) or when positions of PSD95-YFP puncta were randomized within the synaptic layer (random) in Monte Carlo simulations (n = 9 cells, p < 10⁻⁶). Gray lines indicate data from individual cells; circles (error bars) show the mean (±SEM) of the population. (E–H) Representative EPSC (E) and spike response (G) traces to the texture motion stimulus illustrated in Figure 2A recorded from W3-RGCs (wild-type [WT] black, vesicular glutamate transporter 3 [VGluT3^−/−] blue), and bar plots summarizing differences in excitatory conductance (F) and spike rates (H) during different segments of the stimulus in WT (left, black) and VGluT3^−/− mice (right, blue). Bars (error bars) indicate the mean (±SEM) of the respective data sets. W3-RGC EPSCs in VGluT3^−/− mice were significantly reduced compared to WT littermates during differential center motion (WT n = 8, VGluT3^−/− n = 9, p < 0.02), but not global or differential surround motion (p > 0.1 for both). A similar pattern was observed in the spike responses of W3-RGCs, which were decreased for differential center motion (WT n = 13, VGluT3^−/− n = 9, p < 0.001), but not altered during global image motion (p > 0.9). (I–L) Representative EPSC (I) and spike response (K) traces, and summary data (excitation in J, spikes in L, mean ± SEM) recorded in W3-RGCs during stimulation with dark bars of different heights (indicated by color saturation) moving at 400 μm/s in WT (left, black) and VGluT3^−/− mice (right, blue). Whereas excitatory inputs and spike responses were reduced for narrow bars (excitation at 100 μm, WT n = 6, VGluT3^−/− n = 6, p < 0.03, spikes at 100 μm, WT n = 14, VGluT3^−/− n = 8, p < 0.01), they did not differ significantly for bars of greater heights (excitation at 600 μm, WT n = 6, VGluT3^−/− n = 6, p > 0.6, spikes at 600 μm, WT n = 14, VGluT3^−/− n = 8, p < 0.2). See also Figure 2—figure supplement 1, Figure 4—figure supplement 1 and Figure 4—figure supplement 2.

DOI: http://dx.doi.org/10.7554/eLife.08025.010

Figure 4—figure supplement 1. Detection of object motion by W3-RGCs.

(A) Representative spike rate (black), EPSC (red), and IPSC (blue) traces recorded from W3-RGCs during presentation of the texture motion stimuli (illustrated in Figure 2A). W3-RGCs were either recorded under conventional infrared illumination and identified by characteristic responses in cell-attached recordings or targeted under 2-photon guidance in Isl2-GFP transgenic mice (Triplett et al., 2014). In both cases, correct targeting was confirmed by intracellular dye filling and reconstruction of dendritic arborizations at the end of the recordings. (B–D) Summary data of spike rate (B, black), excitatory (C, red), and inhibitory (D, blue) response amplitudes to global, differential center (diff_Ce), and differential surround (diff_Su) motion segments. Dots show data from individual cells and circles (error bars) indicate mean (± SEM) of the respective population. W3-RGCs remain mostly silent during global and differential surround motion stimulation, but show robust spike responses to grating movements restricted to their receptive field center (n = 13, p < 10⁻⁷ for diff_Ce vs global and vs diff_Su). Differential motion sensitivity of WG3-RGCs appears to be inherited from their excitatory input, which during center-only motion exceeds that observed during global motion nearly fivefold, and which is suppressed from tonic levels during isolated surround stimulation (n = 8, p < 0.003 for diff_Ce vs global and vs diff_Su). In addition, W3-RGCs receive stronger direct inhibition when grating movements include the surround (n = 7, p < 0.001 for diff_Ce vs global and vs diff_Su) (E–J) Representative spike rate (E, black), EPSC (G, red), and IPSC (I, blue) traces and summary data of spike (F, black), excitatory (H, red), and inhibitory (J, blue) response amplitudes to dark bars (height: 200 μm) moving at different speeds indicated by matching color saturation of example traces and summary data. Circles (error bars) represent the mean (± SEM) of these data sets (spikes n = 32, excitation n = 8, inhibition n = 5). (K–P) Representative spike rate (K, black), EPSC (N, red), and IPSC (O, blue) responses elicited by bars of different heights moving at 400 μm/s and summary data of spike rate (L, black), excitatory (N, red), and inhibitory (P, blue) response amplitudes. Bar heights are encoded by matching color saturation of responses traces and summary data. Circles (error bars) represent the mean (± SEM) of these data sets. Responses of W3-RGCs are progressively suppressed when bar heights increase above 100 μm (n = 14, p < 10⁻⁴ for 200 μm vs 600 μm bar heights). The spatial tuning of the excitatory input to W3-RGCs is similar to that observed in spike responses (n = 6, p < 0.02 for 200 μm vs 600 μm bar heights), whereas inhibition rises monotonically and saturates with increasing bar heights (n = 6, p > 0.4 for 200 μm vs 600 μm bar heights).

DOI: http://dx.doi.org/10.7554/eLife.08025.011

Figure 4—figure supplement 2. Lamination patterns of cells and neurites are preserved in VGluT3^−/− mice.

(A–P) Maximum intensity projections of confocal image stacks of representative vibratome sections of WT (A, C, E, G, I, K, M, O) and VGluT3^−/− (B, D, F, H, J, L, N, P) retinas. Sections were stained for choline acetyltransferase (ChAT, A, B), protein kinase C alpha (PKCα, C, D), synaptotagmin II (Syt II, E, F), hyperpolarization activated cyclic nucleotide gated potassium channel 4 (HCN4, G, H), Tyrosine hydroxylase (TH, I, J), Recoverin (K, L), Melanopsin (M, N), and Protein Kinase A regulatory subunit II beta (PKARIIβ, O, P).

DOI: http://dx.doi.org/10.7554/eLife.08025.012

Figure 4—figure supplement 3. Dendritic morphology of W3-RGCs is unchanged in VGluT3^−/− mice.

(A, B) Maximum intensity projections through 2-photon image stacks of representative W3-RGCs recorded in WT (A) and VGluT3^−/− (B) mice. (C) Summary data of dendritic territories covered by W3-RGCs labeled biolistically or filled during patch-clamp recordings in WT (black) or VGluT3^−/− (blue) retinas.

DOI: http://dx.doi.org/10.7554/eLife.08025.013

Figure 4—figure supplement 4. Schematic of object motion detection circuit.

(A) Overview illustration of the object motion detection circuit. Somata of ON and OFF BCs are shown as open and filled ovals, respectively. Axons of these neurons converge onto the other components of the circuit: wide-field ACs (wACs), VG3-ACs, and W3-RGCs. (B) Schematic of the inferred connectivity motif, repeated in ON and OFF layers of the object motion detection circuit. Excitatory and inhibitory synaptic output is shown by circles and triangles, respectively, and use of spikes (wAC, ss) or graded potentials (BC, VG3) is indicated by different waveforms.

DOI: http://dx.doi.org/10.7554/eLife.08025.014