Axonal architecture of the mouse inner retina revealed by second harmonic generation

Abstract

We describe a novel method for visualizing the network of axons in the unlabeled fresh wholemount retina. The intrinsic radiation of second harmonic generation (SHG) was utilized to visualize single axons of all major retinal neurons, i.e., photoreceptors, horizontal cells, bipolar cells, amacrine cells, and the retinal ganglion cells. The cell types of SHG+ axons were determined using transgenic GFP/YFP mice. New findings were obtained with retinal SHG imaging: Müller cells do not maintain uniformly polarized microtubules in the processes; SHG+ axons of bipolar cells terminate in the inner plexiform layer (IPL) in a subtype-specific manner; a subset of amacrine cells, presumably the axon-bearing types, emits SHG; and the axon-like neurites of amacrine cells provide a cytoskeletal scaffolding for the IPL stratification. To demonstrate the utility, retinal SHG imaging was applied to testing whether the inner retina is preserved in glaucoma, using DBA/2 mice as a model of glaucoma and DBA/2-Gpnmb+ as the nonglaucomatous control. It was found that the morphology of the inner retina was largely intact in glaucoma and the presynaptic compartments to the retinal ganglion cells were uncompromised. It proves retinal SHG imaging as a promising technology for studying the physiological and diseased retinas in 3D.

Keywords: amacrine cell; axon; bipolar cell; glaucoma; second harmonic generation.

Fig. 1.

Visualizing the fresh wholemount retina by SHG. (a) Experimental configuration for detecting transmitted SHG. (b) Multiphoton imaging of the CAG-H2B-EGFP retina (10) (an axial projection of z-stack images). The plexiform and nuclear layers are segregated by SHG (gray scale) and EGFP (green), respectively. Vertical SHG processes are also visible (arrows). (c) Volumetric rendering (371 × 371 × 186 µm³). (d) The segmented layers (lateral projections): ONL, the outer nuclear layer; OPL, the outer plexiform layer; INL, the inner nuclear layer; and IPL, the inner plexiform layer. Scale bars, 30 µm.

Fig. 2.

Bipolar cell axons visualized by SHG. (a) Co-registration of SHG and GFP/YFP in GUS-, Thy1-, and GFAP-GFP/YFP mice. Overlaps in GUS and Thy1 (arrows), but not in GFAP (arrowheads). (b and c) Vertical slices of the GUS and Thy1 retinas, respectively. Scale bars, 10 µm.

Fig. 3.

Quantification of the bipolar cell axons. (a) The density of bipolar cell axons. (b) A lateral section of the IPL. Right panels are fly-through sections corresponding to the dashed box with bipolar cell axons converging into the interstitial space of the INL (arrowheads). Scale bar, 10 µm. (c) The number of bipolar cell axon terminals, i.e., the derivative of (a), exhibits an internal structure of five layers (arrowheads). (d) 3D rendering of the GUS+ subpopulation of single SHG axons (colored and overlaid with grayscale SHG), identified by the overlap with GFP at zero IPL (green square). (e) The distribution of GUS+ SHG axon terminals (n = 77, magenta) and the mean GFP profile (green). Right, the depths of three GUS+ species (CBC4, 7, and RBC) for comparison. (f and g) GFP+ axon traces overlaid with GFP images, confirming the IPL depth and the characteristic morphology of RBC and CBC7 axon terminals. Scale bars, 5 µm.

Fig. 4.

Variable SHG intensity of individual bipolar cell axons. (a) The mean SHG intensity increases toward the INL. (b) Histograms of the SHG intensity at zero IPL depth of total (gray) and GUS+ (green) bipolar cell axons. The red line shows the maximum likelihood fit to the Gaussian mixture model, which has three components. (c) The SHG intensities along the GUS+ axons (n = 77).

Fig. 5.

Cytoskeletal substrate underlying the IPL stratification. (a) Two intra-IPL gaps divide SHG strata approximately at the depth of ChAT bands. (b) A region, corresponding to the dashed box in (a), is averaged laterally over 6, 20, and 60 µm. (c), (d) The profiles of SHG versus GFP/YFP in the GUS-GFP and ChAT-EYFP retinas, respectively. The positions of five SHG strata are shown with arrowheads. Scale bars, 10 µm.

Fig. 6.

The origin of the IPL neuropil signal. (a) SHG intensity after the treatment with nocodazole. (b) Co-registration of SHG and GFP/YFP in the GUS, Thy1-YFP-H, and -16 retinas. Overlaps only in Thy1-YFP-16 (arrowheads). (c) The stratification of mid- (blue), long-range (red), and displaced amacrine cell (green) SHG+ neurites. (d and e) The identity of SHG+ displaced cell. SHG+ neurites are Thy1+ but not ChAT+ (arrowheads). Conversely, the neurites of ChAT cells are SHG- (arrow). Scale bars, 10 µm.

Fig. 7.

Retinal SHG imaging is tunable to longer wavelengths for safety. (a) SHG imaging at = 800 and 1150 nm. Scale bars, 20 µm. (b and c) Light-induced changes evaluated by autofluorescence and SHG, respectively. The postirradiation images are the 10^th of z-stacks acquired every 5 min (i.e., t = 50 min). (d) SHG kymographs of the retina under continuous illumination. Axial swelling at 800 nm (arrowheads) but not at 950 nm.

Fig. 8.

Morphological changes in the glaucomatous inner retina. (a) Representative SHG images of DBA and DBA-Gpnmb+. The IPL sublaminae appear normal in both strains (arrowheads). Scale bars, 20 µm. (b) The thicknesses of the RGC axon bundles vs. the IPL. (c) The average SHG intensity of the IPL neuropil. (d) The density of axon terminals of bipolar cells.