A dedicated circuit links direction-selective retinal ganglion cells to the primary visual cortex

Abstract

How specific features in the environment are represented within the brain is an important unanswered question in neuroscience. A subset of retinal neurons, called direction-selective ganglion cells (DSGCs), are specialized for detecting motion along specific axes of the visual field. Despite extensive study of the retinal circuitry that endows DSGCs with their unique tuning properties, their downstream circuitry in the brain and thus their contribution to visual processing has remained unclear. In mice, several different types of DSGCs connect to the dorsal lateral geniculate nucleus (dLGN), the visual thalamic structure that harbours cortical relay neurons. Whether direction-selective information computed at the level of the retina is routed to cortical circuits and integrated with other visual channels, however, is unknown. Here we show that there is a di-synaptic circuit linking DSGCs with the superficial layers of the primary visual cortex (V1) by using viral trans-synaptic circuit mapping and functional imaging of visually driven calcium signals in thalamocortical axons. This circuit pools information from several types of DSGCs, converges in a specialized subdivision of the dLGN, and delivers direction-tuned and orientation-tuned signals to superficial V1. Notably, this circuit is anatomically segregated from the retino-geniculo-cortical pathway carrying non-direction-tuned visual information to deeper layers of V1, such as layer 4. Thus, the mouse harbours several functionally specialized, parallel retino-geniculo-cortical pathways, one of which originates with retinal DSGCs and delivers direction- and orientation-tuned information specifically to the superficial layers of the primary visual cortex. These data provide evidence that direction and orientation selectivity of some V1 neurons may be influenced by the activation of DSGCs.

Extended Data Figure 1. The retino-geniculo-cortical pathway links retinal cells and circuits, to the brain

a, Diagram of retina, dorsal lateral geniculate nucleus (dLGN) and primary visual cortex (V1). The optic tract which carries retinal ganglion cell (RGC) axons and thalamocortical (dLGN to V1) pathway also shown. b, Diagram of retinal layers: PRL, photoreceptor layer; opl, outer plexiform layer; INL, inner nuclear layer; ipl, inner plexiform layer; GCL, ganglion cell layer; nfl, nerve fiber layer. c, Retina diagram with cells shown (labels same as in b).

Extended Data Figure 2. Approach for assessing laminar specificity of mouse geniculocortical projections

a, Focal retrograde tracer injection to V1. Scale, 3 mm. b, Diagram of the three different injection depths used to generate data in Fig. 2. c, Percentage of fluorescence in V1 from superficial (black line) versus deep (gray line) injections. Superficial: peak intensity occurs at 25 μm from pial surface (4 mice). Deep: peak intensity occurs at 350 μm from pial surface. Gray shaded regions: s.e.m. (superficial vs. deep= ***P < 0.0001; two-way ANOVA). d, Assessment of retrogradely labeled cells across the width of the dLGN. 0% is at optic tract, 100% is at medial border (see Fig. 2g–i).

Extended Data Figure 3. Retrograde tracers to superficial V1 label cells in the DSGC-RZ

a–c, Same dLGN as in main Fig. 2f but with GFP⁺ On-Off DSGC⁶ axons shown. a, most of the retrogradely labeled cells (magenta/dashed circles) reside in the DSGC-RZ (green terminals). Asterisk: labeled cell outside the DSGC-RZ. Scale, 200 μm b,c, High magnification views of retrogradely labeled dLGN neuron cell bodies with potential contact from GFP⁺ DSGC axons (arrow in b); c, this cell is in vicinity of DGSC axonal boutons (arrowheads). b,c, Scale, 15 μm. d, Diagram of laminar specific connections between DSGC-RZ and superficial V1 and dLGN core and deeper V1 layers 4 and 6.

Extended Data Figure 4. Analysis of dLGN neurons retrogradely infected from superficial V1

a–f, Example serial sections of anterior, middle and posterior portions of dLGN in a mouse with GFP expressing On-Off DSGC axons that was injected with ΔG-RABV-mCherry in superficial layers of V1. a, dapi to show cytoarchitectural landmarks and dLGN borders. b, GFP⁺ DSGC axons and AAV-Glyco-hGFP infected cell bodies (see main Fig. 4 and text). c, Mask of GFP⁺ DSGC axons (Methods). d, ΔG-RABV-mCherry⁺ dLGN relay neurons. e, GFP⁺ DSGC axon mask superimposed with mCherry signal; this was used to determine colocalization. f, mCherry and hGFP signals merged. Scale, 200 μm.

Extended Data Figure 5. Putative sites of contact between DSGC axons and a dLGN neuron retrogradely infected from superficial V1

a–i, GFP⁺ On-Off DSGC axons (green in all panels except black in b) and mCherry⁺ dLGN relay neuron (magenta in all panels except white in c) infected by injection to superficial V1. Framed region in a is shown at higher magnification in b–d. Arrowhead (a): thalamocortical axon of mCherry⁺ dLGN cell. Scale in a, 50 µm. Yellow boxed region in c,d, is shown at higher magnification in e–i. Scale in d, 15 µm. e–i, Some DSGC axon-dendrite contacts contain VGLUT2 (blue). f–i, arrowhead: site of GFP/mCherry co-localization that does not contain VGLUT2. Arrow: GFP/mCherry/VGLUT2⁺ contact.

Extended Data Figure 6. The axons of GFP⁺ On-Off DSGCs and dLGN neurons infected with AAV2-Glyco-hGFP can be distinguished on the basis of their cellular localization

High magnification view of DSGC-RZ in mouse with GFP ⁺ posterior-tuned On-Off DSGCs that was injected 14 days prior with AAV2-Glyco-hGFP. Glyco-hGFP⁺ neurons have nuclear GFP labeling (arrows) whereas DSGCs have GFP in axon terminals (arrowheads). Scale, 50 μm.

Extended Data Figure 7. Signature anatomical and physiological characteristics of GFP tagged On-Off DSGCs

Flat mount retina with a, GFP⁺ On-Off DSGCs. b, co-stained with dapi. c, Positions of GFP⁺ RGCs. Scale in c, 150 µm. d–f, High magnification views. Scale, 12 µm. g, Targeted fill of a GFP⁺ DSGC. Scale, 50 µm. h, Schematic of On-Off DSGC stratification and starburst amacrine cells (magenta). Labeling as in Extended Fig. 1. i,j, Higher magnification of framed region in g stained for VAChT (starburst amacrine processes). Asterisk: ‘looping arborizations’. Dashed line: GFP arbor, which matches VAChT plexus. Scale, 10 µm. k,l, Side (x-z plane) views of cell in (g). GFP⁺ dendrites co-stratify with both the On and Off sublayers. Scale, 5 µm. m, Direction-tuned response of a GFP⁺ On-Off DSGC targeted for recording and receptive field characterization. The spike count is highest for bars moving toward ~270° in the cardinal axes.

Extended Data Figure 8. Injections of ΔG-RABV-mCherry into both superficial and deep V1 combined with AAV-Glyco-hGFP infection of dLGN core

a, mCherry⁺ neurons in the DSGC-RZ and the core of the dLGN. b, AAV2-Glyco-hGFP: many neurons throughout the dLGN, but mostly along the medial border and not in the shell/DSGC-RZ express Glyco-hGFP. DSGC-RZ marked by axons of GFP⁺ On-Off DSGCs. c, Merged of a,b. Scale in a, 100 µm. Boxed regions with arrows: two dLGN neurons; both RABV-mCherry⁺ and AAV2-Glyco-hGFP⁺. One or both of these cells infected their presynaptic partner, the RGC shown in Figure 4 (panels cc-ee) of the main text. Scale, 15µm.

Figure 1. The layer of the dLGN that receives input from DSGCs projects to V1

a, GFP⁺ DSGC axons in dLGN. OT: Optic tract. Inset: GFP⁺ boutons. d: dorsal, m: medial. b, merged GFP/CTβ−594 (all RGC axons) and c, with dapi (cell nuclei). a–c, Scale, 125 µm. d, Summary. e, AAV2-tdTomato injection to dLGN. Asterisk: tdTomato⁺ neurons in core. Scale, 100 µm. f, tdTomato⁺ thalamocortical axons in V1. (L1-6, layers 1–6; wm: white matter). Scale, 150 µm. Higher magnification of (g) L4, (h) L5 and L6 and (i) L1 and L2. Arrows: axons in L1. Scale, g, h, 50 µm. Scale, i, 100 µm. 16 mice.

Figure 2. Parallel, layer-specific thalamocortical circuits in the mouse

Tracer injections to a, all V1 layers b, layer 4 and c, superficial V1. d–f, dLGN neurons labeled after (d) full depth injection (asterisk: axons of L6 neurons; yellow box, neurons in shell; green box, middle; blue box, core) e, deep V1 injection and f, superficial V1 injection (arrows: labeled cells; arrowhead: labeled cell outside shell.) g–i, Position of retrogradely labeled cells along dLGN width: g, Full depth: 35.54 ± 1.62% (4 mice; n= 232 cells); h, deep : 68.53 ± 1.22 % (4 mice; n= 69 cells); i, superficial: 14.68 ± 1.57 % (4 mice; n= 78 cells). Superficial vs. deep= ***P<0.0001; Tukey’s multiple comparisons. c, Scale, 200 µm. f, Scale, 100 µm.

Figure 3. DSGC axons contact thalamic relay neurons projecting to superficial V1

a, ΔG-RABV-mCherry injection to superficial V1, to infect axons of DSGC-RZ neurons. b, V1 injection. Asterisk: infected L5/6 neurons. Scale, 200 µm. c,d, mCherry⁺ dLGN neurons. Arrowhead, axon. Scale, 75 µm. d, Arrows: proximal and arrowheads: distal, dendrites. Scale, 25 µm. e, % mCherry somas within GFP⁺ DSGC-RZ (74.77 ± 0.12%; 8 mice, n= 83 cells) (Extended Data Figure 4). f–h, DSGC axons and dLGN somas and dendrites. g, magnified view of frame in f. Arrowheads, putative contact sites¹³f, Scale, 125 µm. g, 20 µm. h–k. GFP, mCherry, VGLUT2 (% mCherry signal contacted by GFP⁺/VGLUT2⁺ profiles= 5.23 ± 1.39; 4 mice; n= 4 cells). k, Scale, 2 µm.

Figure 4. Synaptic circuit linking DSGCs to superficial V1, and non-DSGCs to L4

a, Trans-synaptic tracing. b, Infected dLGN neurons. Arrow, arrowhead: double-infected cells. Scale, 100 µm. c,d, Cell from b. Scale, 15 µm. Dashed line: lateral border. e, Distribution of double-infected dLGN cells = 9.29 ± 1.82% (8 mice, n= 21 cells). f, On-Off DSGC trans-synaptically labeled from superficial V1. g, On (red) and Off (black) dendrites. Arrowhead: axon. Scale, 50 µm. h–j, Cell (f) is GFP⁺ On-Off DSGC⁶. Scale, 10 µm. k–m’, Trans-synaptically labeled GFP⁺ and Cart⁺ DSGC⁵m, Scale, 75 µm. m’, 10 µm. n, Trans-synaptically labeled J-RGC⁴o, Off dendrites (black). Scale, 50 µm. p, Same as a, but layer 4 injection. q–s, Infected neurons in core; q, arrow, arrowhead: double-infected cells. Scale, 100 µm. r,s, Cell from q (arrow). s, Scale, 15 µm. t, Distribution of double-infected dLGN cells= 70 ± 2.65% (7 mice, n= 53 cells) (P< 0.0001 versus e; two-tailed t-test). u,v, Alpha RGC labeled from V1 layer 4. u, Scale, 100 µm; sideview, 50 µm. w–y, cell (u,v), SMI-32⁺. Scale, 20 µm. z–bb, SMI-32 and Cart. Scale, 25 µm. cc–ee, ΔG-RABV-mCherry⁺ alpha RGC; lacks GFP⁶ and Cart. Scale, 150 µm.

Figure 5. In vivo imaging of visually-evoked Ca⁺⁺ signals in thalamocortical axons

a, AAV2-GCaMP6 injection to dLGN shell b, GCaMP6⁺ neurons (arrows). Scale, 50 µm; inset, 10 µm. c, GCaMP6⁺ dLGN axons, superficial V1. Arrows: varicosities. Scale, 50 µm. d, GCaMP6⁺ axons, superficial V1. Circles, square: polar plots l,o,p. Scale, 5 µm, e, In vivo imaging/visual stimulation. f, Visually-evoked Ca⁺⁺ signal in thalamocortical axon (top trace: photodiode signal; bottom trace: ΔF/F. g, Directional stimuli (0°, 45°, 90°, 135°, 180°, 225°, 270°, 315°). h–j, Direction- (h,i) and orientation-tuned (j) varicosities. 5–8 trial average. k–s, Polar plots of F1 (red) or F2 (black) magnitude responses (Methods). Inner solid ring: average response to mean grey stimulus. Shaded: 3 standard deviations greater than the mean response to grey stimuli. Lower right of each plot: OSI/DSI . Upper right: Fourier amplitudes. t, DSI/OSI, all varicosities (5 mice, n= 58 varicosities). Mean ± s.e.m. u, Cumulative distributions: OSI (circles), DSI (squares).