Topography of Neurons in the Rod Pathway of Human Retina

Abstract

Purpose: The objective of this study was to map the distribution and density of the three major components of the classical scotopic “night vision” pathway (rods, rod bipolar, and AII amacrine cells) in postmortem human retinas.

Methods: Four postmortem donor eyes (male and female, aged 44–56 years) were used to cut vertical sections through the temporal horizontal meridian. The sections were processed for immunohistochemistry and imaged using high-resolution multichannel confocal microscopy. Rods, rod bipolar, and AII amacrine cells were counted along the temporal horizontal meridian. Two additional retinas were used for intracellular injections.

Results: Rod peak density is close to 150,000 cells/mm² at 4 to 5 mm (15° to 20°) eccentricity, declining to below 70,000 cells/mm² in peripheral retina. Rod bipolar density is lower but follows a similar distribution with peak density near 10,000 cells/mm² between 2 and 4 mm (7° to 15°) eccentricity declining to below 4000 cells/mm² in peripheral retina. The peak density of AII amacrine cells (near 4000 cells/mm²) is located close to the fovea, at 0.5- to 2 mm-eccentricity (2° to 7°) and declines to below 1000 cells/mm² in the periphery. Thus, convergence between rods and AII cells increases from central to peripheral retina.

Conclusions: Comparison with human psychophysics and ganglion cell density indicates that the spatial resolution of scotopic vision is limited by the AII mosaic at eccentricities below 15° and by the midget ganglion cell mosaic at eccentricities above 15°.

In the classical (primary) rod pathway of the mammalian retina rod photoreceptors are hyperpolarized by light and transfer their signal to the dendrites of a single type of rod bipolar cell.^1,2 Rod bipolar cells contact AII amacrine cells in the inner plexiform layer. The AII cells transfer the rod signal into the ON cone pathway via gap junctions with the axons of ON cone bipolar cells, and into the OFF cone pathway via chemical synapses with the axons of OFF cone bipolar cells. Two additional rod pathways have been identified: in the second rod pathway, rods contact cones via gap junctions,³ and in the third rod pathway some OFF cone bipolar cell types contact rods directly.⁴ The primary rod pathway is the most sensitive pathway^2,5 and is the focus of the present study. 

Rods, rod bipolar and AII cells have been the topic of quantitative studies in macaque retina.^6–11 A conclusion of these previous studies is that scotopic acuity in central retina is limited by the AII amacrine mosaic whereas scotopic acuity in peripheral retina is limited by the ganglion cell array. This conclusion, however, depends on applying a mixture of anatomic data, from macaque retinas, to scotopic acuity measurements obtained from human subjects.¹² Further, these previous studies characterized each cell population in isolation (i.e., in separate preparations), making measures such as convergence (that is, the density ratio between two populations) susceptible to errors due to variations between individuals. Although technically challenging, double- and triple-label immunochemistry provides a means to simultaneously sample multiple cell populations in individual retinas, and this is the approach we have used in the present study. 

For human retina, to date, quantitative data on rod pathway neurons are only available for rods.^13–15 The aim of the present study was therefore to provide quantitative data for the downstream neurones, rod bipolar, and AII cells across the retina. Quantitative analysis of the rod pathways in human retina gives data of most direct relevance for understanding human visual performance under scotopic and mesopic conditions. Furthermore, baseline knowledge of the distribution of neurons in human retina is important for interpreting images obtained with optical coherence tomography, and for understanding the basis of the normal electroretinogram (ERG), which is the most important objective diagnostic test of visual function. Specifically, multifocal photopic and scotopic ERG responses provide information that can be related to the distribution of the receptors and postreceptor cells of the central retina (about 6 mm around the fovea).^16,17 Finally, clinical strategies for gene therapy to restore light sensitivity in photoreceptor diseases^18,19 can benefit from knowledge of how photoreceptors and postreceptoral neurons are distributed in human retinas. 

Materials & Methods

Tissue Collection and Preparation

Postmortem human donor eyes (with corneas removed) with no known history of posterior eye disease were received with informed consent from the Lions NSW Eye Bank and Australian Ocular Biobank and with ethical approval by The University of Sydney Human Research Ethics committee (HREC# 2012/2833). The protocols adhered to the tenets of the Declaration of Helsinki. Details about the eyes are summarized in Table 1. All eyes were fixed in 2% paraformaldehyde in phosphate buffer and processed as described previously.²⁰ Retinal pieces of defined eccentricities along the temporal side of the horizontal meridian were embedded in agarose and vertical vibratome sections were cut and processed for immunohistochemistry as reported previously.^20,21 Details of the primary antibodies used are summarized in Table 2. 

Table 1

 

Subjects

 

Table 2

 

Antibodies

 

Microscopy

A confocal microscope (LSM700, Carl Zeiss) equipped with 405, 488, 555, and 635 nm lasers was used for imaging. Tiled imaged stacks were taken with a ×20 air objective (Plan Apochromat, 20×/0.8, #420650-9901) of the entire vibratome section (approximately 3 mm in length and 100 μm in thickness) at a resolution of 2048 × 2048 pixels with z-axis step sizes from 0.89 to 1.15 μm for each optical plane. In addition, stacks of images were taken from some regions of interests using a ×40 water immersion objective (Plan Apochromat, 340/1.2, #421767- 9970). Depth readings taken using the air objective were corrected for the refractive index of the mounting medium (Vectashield 1.45; Vector Laboratories, Burlingame, CA, USA). The contrast and brightness of the images was adjusted using imaging software (Zen Black; Carl Zeiss Microscopy, Thornwood, NY, USA), image editing software (Adobe Photoshop; Adobe, Inc., Mountain View, CA, USA), or microscopy image analysis software (Imaris; Bitplane, Zurich, Switzerland). 

Definitions

Based on previous studies,^22–25 we will use the following terms. Foveal retina refers to 5.5° diameter of visual angle, which is equivalent to a diameter of 1.6 mm or a radius of 0.8 mm. Central retina (macula) refers to eccentricity within 10° (equivalent to a diameter of 6 mm or a 3.0 mm radius), midperiphery includes eccentricities between 3 and 6 mm and far periphery refers to eccentricities beyond 6 mm. For simplicity, we used the same magnification factor (0.29 mm/degree) for all eyes. 

Analysis

Reconstructions: Individual rod bipolar cells and AII cells (central retina) were reconstructed from stacks of confocal images using the surface tool in the microscopy image analysis software (Bitplane). Density measurements: Cell densities were determined from tiled image stacks through vibratome sections using microscope software (ZEN Blue; Carl Zeiss Microscopy). The full length (∼3 mm) of each section was divided into 100 μm bins and cells were counted for each eccentricity sample point. Counts were made from 4 to 12 sections (from a total of four retinas) for each antibody; a minimum of 20 cells was counted for each 100 μm bin. These values were chosen in pilot measurements to give the best practical compromise between sample statistical power at each point (i.e., ∼5% sample noise at each point) and estimate of cell density gradients (i.e., 10 density samples/linear mm). Rod bipolar cells and AII cells were counted for 40 μm depth within the image stack. Rod photoreceptors, being much smaller and denser than bipolar and amacrine cells, were counted for 10 μm depth. A set of cross-validation counts (n = 7 regions of interest) for rod photoreceptors, obtained using a water immersion ×40 objective but otherwise identical methods, returned 29% higher density values at equivalent sample positions. Density values reported here for rod photoreceptors counted with the ×20 air objective are corrected by this factor. 

Areal cell densities (cells/mm²) were calculated for each counted bin and averages were determined for eccentricities across different samples. Eccentricity sample points were calculated taking into account the triangulated distance (if any) of the source Vibratome section from the horizontal meridian. Density values were pooled across retinas in 0.5-mm bins for the central 2 mm and in 1 mm bins at greater eccentricities. Pooled data means and SD are plotted at the maximum eccentricity for each bin. Pooled density values (as a function of distance from the foveal center) were fit with two- or three-stage difference-of-exponentials functions of the form: Display Formula\(\def\upalpha{\unicode[Times]{x3B1}}\)\(\def\upbeta{\unicode[Times]{x3B2}}\)\(\def\upgamma{\unicode[Times]{x3B3}}\)\(\def\updelta{\unicode[Times]{x3B4}}\)\(\def\upvarepsilon{\unicode[Times]{x3B5}}\)\(\def\upzeta{\unicode[Times]{x3B6}}\)\(\def\upeta{\unicode[Times]{x3B7}}\)\(\def\uptheta{\unicode[Times]{x3B8}}\)\(\def\upiota{\unicode[Times]{x3B9}}\)\(\def\upkappa{\unicode[Times]{x3BA}}\)\(\def\uplambda{\unicode[Times]{x3BB}}\)\(\def\upmu{\unicode[Times]{x3BC}}\)\(\def\upnu{\unicode[Times]{x3BD}}\)\(\def\upxi{\unicode[Times]{x3BE}}\)\(\def\upomicron{\unicode[Times]{x3BF}}\)\(\def\uppi{\unicode[Times]{x3C0}}\)\(\def\uprho{\unicode[Times]{x3C1}}\)\(\def\upsigma{\unicode[Times]{x3C3}}\)\(\def\uptau{\unicode[Times]{x3C4}}\)\(\def\upupsilon{\unicode[Times]{x3C5}}\)\(\def\upphi{\unicode[Times]{x3C6}}\)\(\def\upchi{\unicode[Times]{x3C7}}\)\(\def\uppsy{\unicode[Times]{x3C8}}\)\(\def\upomega{\unicode[Times]{x3C9}}\)\(\def\bialpha{\boldsymbol{\alpha}}\)\(\def\bibeta{\boldsymbol{\beta}}\)\(\def\bigamma{\boldsymbol{\gamma}}\)\(\def\bidelta{\boldsymbol{\delta}}\)\(\def\bivarepsilon{\boldsymbol{\varepsilon}}\)\(\def\bizeta{\boldsymbol{\zeta}}\)\(\def\bieta{\boldsymbol{\eta}}\)\(\def\bitheta{\boldsymbol{\theta}}\)\(\def\biiota{\boldsymbol{\iota}}\)\(\def\bikappa{\boldsymbol{\kappa}}\)\(\def\bilambda{\boldsymbol{\lambda}}\)\(\def\bimu{\boldsymbol{\mu}}\)\(\def\binu{\boldsymbol{\nu}}\)\(\def\bixi{\boldsymbol{\xi}}\)\(\def\biomicron{\boldsymbol{\micron}}\)\(\def\bipi{\boldsymbol{\pi}}\)\(\def\birho{\boldsymbol{\rho}}\)\(\def\bisigma{\boldsymbol{\sigma}}\)\(\def\bitau{\boldsymbol{\tau}}\)\(\def\biupsilon{\boldsymbol{\upsilon}}\)\(\def\biphi{\boldsymbol{\phi}}\)\(\def\bichi{\boldsymbol{\chi}}\)\(\def\bipsy{\boldsymbol{\psy}}\)\(\def\biomega{\boldsymbol{\omega}}\)\(\def\bupalpha{\unicode[Times]{x1D6C2}}\)\(\def\bupbeta{\unicode[Times]{x1D6C3}}\)\(\def\bupgamma{\unicode[Times]{x1D6C4}}\)\(\def\bupdelta{\unicode[Times]{x1D6C5}}\)\(\def\bupepsilon{\unicode[Times]{x1D6C6}}\)\(\def\bupvarepsilon{\unicode[Times]{x1D6DC}}\)\(\def\bupzeta{\unicode[Times]{x1D6C7}}\)\(\def\bupeta{\unicode[Times]{x1D6C8}}\)\(\def\buptheta{\unicode[Times]{x1D6C9}}\)\(\def\bupiota{\unicode[Times]{x1D6CA}}\)\(\def\bupkappa{\unicode[Times]{x1D6CB}}\)\(\def\buplambda{\unicode[Times]{x1D6CC}}\)\(\def\bupmu{\unicode[Times]{x1D6CD}}\)\(\def\bupnu{\unicode[Times]{x1D6CE}}\)\(\def\bupxi{\unicode[Times]{x1D6CF}}\)\(\def\bupomicron{\unicode[Times]{x1D6D0}}\)\(\def\buppi{\unicode[Times]{x1D6D1}}\)\(\def\buprho{\unicode[Times]{x1D6D2}}\)\(\def\bupsigma{\unicode[Times]{x1D6D4}}\)\(\def\buptau{\unicode[Times]{x1D6D5}}\)\(\def\bupupsilon{\unicode[Times]{x1D6D6}}\)\(\def\bupphi{\unicode[Times]{x1D6D7}}\)\(\def\bupchi{\unicode[Times]{x1D6D8}}\)\(\def\buppsy{\unicode[Times]{x1D6D9}}\)\(\def\bupomega{\unicode[Times]{x1D6DA}}\)\(\def\bupvartheta{\unicode[Times]{x1D6DD}}\)\(\def\bGamma{\bf{\Gamma}}\)\(\def\bDelta{\bf{\Delta}}\)\(\def\bTheta{\bf{\Theta}}\)\(\def\bLambda{\bf{\Lambda}}\)\(\def\bXi{\bf{\Xi}}\)\(\def\bPi{\bf{\Pi}}\)\(\def\bSigma{\bf{\Sigma}}\)\(\def\bUpsilon{\bf{\Upsilon}}\)\(\def\bPhi{\bf{\Phi}}\)\(\def\bPsi{\bf{\Psi}}\)\(\def\bOmega{\bf{\Omega}}\)\(\def\iGamma{\unicode[Times]{x1D6E4}}\)\(\def\iDelta{\unicode[Times]{x1D6E5}}\)\(\def\iTheta{\unicode[Times]{x1D6E9}}\)\(\def\iLambda{\unicode[Times]{x1D6EC}}\)\(\def\iXi{\unicode[Times]{x1D6EF}}\)\(\def\iPi{\unicode[Times]{x1D6F1}}\)\(\def\iSigma{\unicode[Times]{x1D6F4}}\)\(\def\iUpsilon{\unicode[Times]{x1D6F6}}\)\(\def\iPhi{\unicode[Times]{x1D6F7}}\)\(\def\iPsi{\unicode[Times]{x1D6F9}}\)\(\def\iOmega{\unicode[Times]{x1D6FA}}\)\(\def\biGamma{\unicode[Times]{x1D71E}}\)\(\def\biDelta{\unicode[Times]{x1D71F}}\)\(\def\biTheta{\unicode[Times]{x1D723}}\)\(\def\biLambda{\unicode[Times]{x1D726}}\)\(\def\biXi{\unicode[Times]{x1D729}}\)\(\def\biPi{\unicode[Times]{x1D72B}}\)\(\def\biSigma{\unicode[Times]{x1D72E}}\)\(\def\biUpsilon{\unicode[Times]{x1D730}}\)\(\def\biPhi{\unicode[Times]{x1D731}}\)\(\def\biPsi{\unicode[Times]{x1D733}}\)\(\def\biOmega{\unicode[Times]{x1D734}}\)\(D = {c_1}\exp \left( {{\lambda _1}X} \right) + {c_2}\exp \left( {{\lambda _2}X} \right)[ + {c_3}\exp \left( {{\lambda _3}X} \right)]\), where Display Formula\(D\) is cell density (cells/mm²), Display Formula\({c_1}\), Display Formula\({c_2}\) and Display Formula\({c_3}\) are multiplicative coefficients, Display Formula\({\lambda _1}\), Display Formula\({\lambda _2}\), and Display Formula\({\lambda _3}\) are exponential coefficients, and Display Formula\(X\) is distance from the fovea (mm). Data were fit by constrained nonlinear error minimization (trust-region algorithm) in a computing environment (MATLAB version 9.2; MathWorks, Natick, MA, USA). Optimal fit parameters are given in Table 3. 

Table 3

 

Best Fit Parameters for Pooled Data

 

For purposes of comparison with human scotopic acuity measurements,^9,10,12 the inter-cell distance Display Formula\(a\) of an hexagonal mosaic with spatial density Display Formula\(D\) was calculated for each measured cell population using the formula Display Formula\( = \sqrt {1/\left( {{{\sqrt 3 } \over 2}D} \right)} \). The maximum resolvable spatial frequency (Nyquist limit) for this mosaic is given by Display Formula\(N = m/\left( {a\sqrt 3 } \right)\), where Display Formula\(N\) is the Nyquist limit (cycles per degree), Display Formula\(m\) is the retinal magnification factor, and Display Formula\(a\) is the intercell distance. For simplicity, we ignored the nonlinear reduction of retinal magnification with increasing eccentricity²⁶ and used a constant magnification factor of 0.29 mm/degree. The simplification introduces maximum ∼20% overestimate of retinal magnification at 60°; exact magnification values can be recovered by reference to Figure 5 in Ref. 26. 

Results

Postmortem human retinas such as shown in Figure 1A were dissected into small pieces (3 or 4 mm length and 5 mm width) across the temporal side of the horizontal meridian and used to cut and immunolabel vertical vibratome sections (Figs. 1B–E). Thus, we were able to count rods, rod bipolar and presumed AII cells in the same retinal preparation. 

Figure 1

 

Processing of a postmortem human donor retina. (A) Photomontage of preparation #14064 after dissection from the eye cup and applying relieving cuts. The arrow points to the fovea, the arrowhead points to the location of the optic disk. (B–E) Confocal micrographs of a vertical vibratome section taken through the fovea along the horizontal meridian of the retina shown in A. (B) Composite image of the section taken with Nomarski optics. C–E show the same section immunolabeled with antibodies against S-antigen, protein kinase Ca (PKCα), and calretinin (CaR), respectively. (F) Foveal region from the same section as shown in C labeled for S-antigen together with Nomarski optics to reveal the retinal layers. Note the absence of rods in the center of the fovea. (G) High resolution image of the section shown in F demonstrating that rods including the rod Henle fibers are stained with antibodies against S-antigen. (H) Confocal micrograph of a vibratome section through peripheral retina at about 10-mm eccentricity processed for S-antigen immunoreactivity. In addition to rods, a labeled S-cone can be seen. (I) Confocal image of a vibratome section through foveal retina processed with antibodies against the short wavelength sensitive (S) cone pigment. The image was taken at about 300 μm eccentricity and shows labeled S-cones. GCL, ganglion cell layer; HFL, Henle fiber layer; INL, inner nuclear layer; IPL, inner plexiform layer; IS, inner segments; ONL, outer nuclear layer; OS, outer segments. Scale bar shown in A: 5 mm. Divisions shown in B: 0.5 mm (valid for B–E). Scale bar shown in F: 200 μm. Scale bar shown in I: 50 μm (valid for G–I).

Identification of Rods and Rod Bipolar Cells

Rods were identified with antibodies against S-antigen or antibodies against rhodopsin (Figs. 1, 2). Immunoreactivity for S-antigen (later named rod arrestin²⁷) is expressed by the entire rod^28,29 (Figs. 1F, 1G), whereas rhodopsin is only found in the outer segments²⁹ (Figs. 2A, 2B). High resolution images show that S-antigen labeling for rods is absent from the center of the fovea (Fig. 1F). Antibodies against S-antigen have also been shown to label short-wavelength sensitive (S) cones in tree shrew and baboon.^28,30,31 Consistently, labeled S-cones were seen in our S-antigen labelled preparations (Fig. 1H). As illustrated in Figure 1H, I S-cones could be easily distinguished from rods, because they are much larger than rods and the S-cone soma is located very close to its inner segment, whereas rod nuclei are located further vitread. 

Figure 2

 

Labeling of rods and rod bipolar cells in human retina. (A) Confocal image of a vertical vibratome section from preparation #13699 at about 3-mm eccentricity. The section was processed with antibodies against antibodies against rhodopsin to identify rods and antibodies against protein kinase Cα (PKCα) to identify rod bipolar cells. The upper and lower rectangles indicate the regions shown in B and C. Nomarski optics is used to reveal the retina layers. (B) The outer segments of rods are labeled with an antibody against rhodopsin. (C) Antibodies against PKCα reveal the typical morphology of rod bipolar cells (open arrowhead) displaying a strong primary dendrite with many fine branches ending in distinct terminal knobs. (D–F) Confocal images of a section through the fovea of preparation #13587 that was processed with antibodies against PKCα. Rod bipolar cells are strongly PKCα positive whereas DB4 cells (open arrow) are only weakly labeled. Note the absence of rod bipolar cells in the center of the fovea (F). The arrows indicate the axon terminals of DB4 cells. Rod bipolar axons terminating close to the GCL are visible from about 1-mm eccentricity (arrowheads in D, E). IPL, inner plexiform layer; NFL, nerve fiber layer; OPL, outer plexiform layer. Numbers indicate eccentricity in mm. Scale bars: 100 μm. Scale bar shown in B also applies to C; scale bar shown in F also applies to D, E.

Rod bipolar cells were identified with antibodies against PKCα, which is a well-established marker for rod bipolar cells (Figs. 2, 3).^7,32,33 As reported previously^24,33 cells with typical rod bipolar morphology are absent from the human foveola but a thin band of PKCα immunoreactivity in stratum 4 of the inner plexiform layer extends toward the center of the fovea (arrows in Fig. 2F). This band originates from the axon terminals of the diffuse cone bipolar type DB4,³⁴ which has been shown to be PKCα immunoreactive in primates including humans.^8,10,35 The somas of PKCα immunoreactive rod bipolar cells are much more strongly labeled (arrowhead in Fig. 2C) than the somas of DB4 cells (open arrow in Fig. 2F). This fact and the fact that DB4 cell axons terminate close to the middle of the inner plexiform layer enables the distinction of the two bipolar types. 

Figure 3

 

Reconstructions of rod bipolar cells in central (A) and peripheral retina (B) from vertical sections that were processed for PKCα immunoreactivity. The eccentricity is given in the lower left for each cell. Scale bar: 20 μm applies to all.

We measured the dendritic tree diameter for samples of well-labeled rod bipolar cells in central, midperipheral and far peripheral retina. Figure 3 shows examples of reconstructed rod bipolar cells from 2 and 10-mm eccentricity. The mean diameter of rod bipolar dendritic fields at 1 mm (14.5 μm ± 2.4 SD, n = 12) was marginally smaller (i.e., statistically not significantly different from that at 4 mm; 17.1 μm ± 3.1 SD, n = 11, P = 0.03), and fields of both these samples were significantly smaller than the fields at 12 mm (33.5 μm ± 2.8 SD, n = 9, P < 0.01 for both comparisons, Wilcoxon nonparametric rank test). These values were used to estimate the convergence from rod photoreceptors to rod bipolar cells at these three locations. 

Identification of AII Amacrine Cells

Antibodies against the calcium binding protein calretinin are well established markers for AII amacrine cells in macaque^9–11 and human retinas^20,24,36 Here we double labeled vertical sections with antibodies against PKCα and calretinin to analyze the relationship between rod bipolar and AII cells (Fig. 4A). The calretinin positive cells have the morphology of AII cells and their distal dendrites are intermingled with the axon terminals of PKCα positive rod bipolar cells (inset). However, our previous studies also showed for selected eccentricities that a small proportion of calretinin positive cells is also positive for GAD,^20,24 suggesting that an additional amacrine cell type is calretinin positive. Here, we determined the density and proportion of glutamic acid decarboxylase (GAD) positive/calretinin positive cells within the population of calretinin-positive cells across the retina (Figs. 4B–G). This allowed us to calculate, by subtraction, the density of AII cells across the temporal retina (see below). 

Figure 4

 

Immunohistochemical labeling of AII amacrine cells in temporal retina. (A) Confocal image of a vertical vibratome section that was double labeled with antibodies against CaR (CaR, green) and antibodies against protein kinase C (PKCα, magenta, rod bipolar cells.). The inset shows a region of a different section taken at higher resolution. Calretinin-positive cells have AII morphology and their arboreal dendrites intermingle with rod bipolar axon terminals. The numbers indicate the distance from the fovea in mm. (B–G) Confocal images of a triple-labeled section from about 5-mm eccentricity (temporal) processed with antibodies against calretinin, glycine transporter 1 (GlyT1) and GAD. Most calretinin-positive cells are GlyT1 positive (white arrows in B, C, E) and most GAD-positive cells are calretinin negative (open arrows in D, F, G). Scale bar shown in the inset: 25 μm. Scale bar shown in G: 50 μm; applies to B–G.

Calretinin-positive cells were injected with DiI to reveal their morphology²⁰ (Fig. 5) and to measure their dendritic field diameter (n = 24 cells from flat mount preparations; n = 10 cells from vertical sections) at eccentricities between 9 and 15 mm. The diameters of the fine arboreal dendritic fields (i.e., the sites of the synaptic input from rod bipolar cells) were ∼2 times larger than the diameter of the lobular dendritic fields (i.e., the sites of the synaptic output to cone bipolar and ganglion cells). The average arboreal dendritic field diameter was 109.2 μm ± 68.4 SD and the average AII lobular dendritic field diameter was 47.3 μm ± 25.4 SD. 

Figure 5

 

Identification of AII amacrine cells by intracellular injection. (A) Confocal images of a flatmount preparation that was prelabeled with antibodies against (CaR, green) and subsequently injected with Dil (magenta). The focus in B is on the lobular appendages and the focus in C is on the arboreal dendrites. (D, E) Reconstructions of AII cells after prelabeling and DiI injection. The lobular appendages are drawn in magenta, arboreal dendrites are drawn in gray. Scale bar: 25 μm. Scale bar shown in C applies to A–C. Scale bar shown in D applies to D and E.

In central and midperipheral retina, the dendritic field diameters of AII cells were determined from vertical immunolabeled image stacks. At 1 mm, the mean diameter of arboreal dendritic fields (19.2 μm ± 4.9 SD, n = 3) was close to the diameter of the lobular dendritic fields (16.8 μm ± 1.6 SD, n = 3) and at 4 mm the mean diameter of arboreal dendritic fields (33.5 μm ± 2.8 SD, n = 5) was marginally larger but not statistically different than the diameter of the lobular dendritic fields (24.2 μm ± 5.6 SD, n = 5, P = 0.06, paired Wilcoxon nonparametric rank test). The values were used to estimate the number of rod bipolar cells within the reach of AII arboreal dendrites at these three eccentricities (see below). 

All retinas used in the present study had an excellent overall morphology with both rods and rod bipolar cells appearing normal. However, as reported previously,²⁴ signs of postmortem necrosis were visible in calretinin labeled amacrine cells of some retinas. The only retina free of this phenomenon was preparation #14064, which had the shortest delay between death and fixation. The AII density in this retina was nevertheless lower than that of two other retinas which showed signs of necrosis. Thus, it is unlikely that postmortem necrosis led us to underestimate the density of AII cells. 

Quantitative Analysis of Rod Pathways Cells

Figure 6 shows that in all preparations the peak density of rods is found at eccentricities between 2.5 and 5 mm with mean value about 150,000 cells per mm². The peak measured rod densities were at 3.01 mm (202,500 cells per mm²) for case 14064; 5.2 mm (250,800 cells per mm²) for case 13587, and 2.04 mm (201,400 cells per mm²) for case 13699. These values agree well with the mean peak density of 176,200 cells per mm² reported by Curcio et al.¹⁴ Rod density drops slowly to about 70,000 cells per mm² at 15-mm eccentricity. 

Figure 6

 

Spatial density of rod photoreceptors on temporal axis of four human retinas. (A) Individual values for four retinas (preparation details are given in Table 1). (B) Pooled values. Error bars show standard deviation. Smooth curve shows difference-of-exponents fit (fit parameters are given in Table 3). Peak rod density ∼150,000 cells/mm² is near 4–5 mm.

Rod bipolar densities were determined from PKCα labeled preparations (see above). Figure 7 shows the densities for rod bipolar cells in four preparations. In all cases, away from the foveal center the rod bipolar density increases rapidly to reach maximum at an eccentricity between 2–4 mm. The mean peak rod bipolar density was close to 10,000 cells/mm² and decreased gradually to less than 4000 cells/mm² in peripheral retina. The peak measured rod bipolar densities were at 2.81 mm (14,150 cells per mm²) for case 14064; 1.2 mm (18,580 cells per mm²) for case 13587, and 1.77 mm (16,600 cells per mm²) for case 13699. 

Figure 7

 

Spatial density of rod bipolar cells on temporal axis of four human retinas. (A) Individual values for four retinas (preparation details are given in Table 1). (B) Pooled values. Error bars show standard deviation. Smooth curve shows difference-of-exponents fit (fit parameters are given in Table 3). Peak rod bipolar density ∼10,000 cells/mm² is near 2 to 4 mm.

In order to quantify AII cells, we counted all calretinin positive amacrine cells in the inner nuclear layer in four preparations (Fig. 8A). In one preparation, we also determined the number of calretinin positive/GAD-positive cells (Fig. 4). The calretinin-/GAD-positive cells have a relative low density (Fig. 8B) when compared to the total number of calretinin-positive cells. There was a ∼2-fold variation in AII density between individual retinas (Fig. 8A). The density of calretinin-/GAD-positive cells was subtracted from the mean pooled density of all calretinin-positive amacrine cells to estimate the density of AII cells (Fig. 8C). The average peak density for AII cells was 4000 cells/mm² near 1 mm and decreased gradually to less than 1000 cells/mm² at 15-mm eccentricity. 

Figure 8

 

Spatial density of AII amacrine cells on temporal axis of four human retinas. (A) Density of calretinin-positive amacrine cells. Individual values for four retinas (preparation details are given in Table 1). (B) Density of cells double-labeled with CaR and GAD in one retina. (C) Estimate of AII amacrine cell density. Curve with data points shows pooled values for calretinin. Error bars show standard deviation. The curve from panel B is subtracted from this curve to give estimate of AII amacrine cell density. Smooth curves show difference-of-exponents fits (fit parameters are given in Table 3). Peak AII amacrine density ∼4,000 cells/mm² is near 1 mm.

Figure 9A shows the densities of rod, rod bipolar, and AII cells on a logarithmic scale. Equal displacements on the logarithmic scale represent equivalent density multiples allowing comparison of relative density in the different cell populations. The density of rods and rod bipolar cells falls at a similar and constant rate outside the fovea, implying relatively constant convergence between these populations. Accordingly, as shown in Figure 9B (dotted black curve), the density ratio (numerical convergence) between rods and rod bipolar cells rises minimally from 13.0 at 1 mm to 18.1 at 7 mm, then decreases only slightly to 13.9 at 15 mm. Figure 9A shows the density of AII cells drops more steeply than that of rods and rod bipolar cells between 1- and 5-mm eccentricity, whereas all three curves run in parallel at greater eccentricities. Thus, the numerical convergence between rod bipolar and AII cells (Fig. 9B) rises from 1.4 at 1 mm to 4.8 at 5 mm, then remains relatively constant thereafter. Finally, the ratio of rods to AII cells (Fig. 9C) rises rapidly from 18.1 at 1 mm to 96.0 at 8 mm, then shows a shallow decline to 74.0 at 15 mm. 

Figure 9

 

Convergence in the primary rod pathway. (A) Densities of rod, rod bipolar, and AII amacrine cells. Smooth curves show difference-of-exponents fits (fit parameters are given in Table 3). (B) Dotted black curve: numerical convergence (spatial density ratio) between rods and rod bipolar cells. Curve marked by orange circles shows numerical convergence between rod bipolar and AII cells. Symbols joined by straight line segments show ratios calculated from regions of double-labeled sections. (C) Numerical convergence between rods and AII cells. (D) Schematic view showing densities of rods at eccentricities near AII peak density (1 mm, left), rod peak (4 mm, center) and 12 mm (right). Rod matrix is represented as jittered hexagonal arrays at the relevant spatial density. Middle right: rod bipolar cells. Shaded pink ellipses represent dendritic field area of a single rod bipolar cell at each eccentricity. Lower right: AII amacrine cells. Shaded blue ellipses represent arboreal dendritic field area of a single AII cell at each eccentricity. Both rod bipolar and AII populations obey a “touch thy neighbor” rule yielding 50% dendritic field overlap and coverage factor ∼5. Bottom right: cone mosaic at these eccentricities is shown for reference, note that cone density is substantially lower than rod density at these key eccentricities.

We also calculated convergence by direct counts of both cell types from the same positions in double-labeled sections. This was possible only for regions of double-labeled sections where staining quality for both markers was uniformly high and tissue distortion was not present. Example data from three cases (#14064, #13587 and #13699) are superimposed on the data plots in Figures 9B and 9C. The convergence values agree well with the average data and show consistent changes across the range (∼2–6 mm). We conclude from these data that convergence in the scotopic pathway in human retina increases rapidly between the fovea and 5 mm and is relatively constant at greater eccentricities. 

Figure 9D shows schematically the relative densities of rods, rod bipolar cells, and AII cells, at eccentricities near the peak density of AII cells (1 mm, left column), near the peak density of rods (4 mm, center column) and at 12 mm in peripheral retina (right column). The cell populations are represented as jittered hexagonal arrays within a 50 × 50 μm field. The shaded pink ellipses in the middle row represent the mean dendritic field area of individual rod bipolar cells taken from our measurements from vertical sections. At each eccentricity the dendritic tree radius is close to the intercell distance, that is the dendrites of each cell reach to the soma of the neighboring cell (“touch-thy-neighbor” rule), yielding a dendritic coverage factor close to 5.³⁷ The shaded blue ellipses in the lower row represent the arboreal dendrites of individual AII cells. Values are taken from calretinin-stained AII cells (at 1 and 4 mm) and from DiI injected cells in peripheral retina (the ellipse at 12 mm is only partly drawn, because the average diameter of AII arboreal dendrites at this eccentricity is greater than 100 μm). In common with the rod bipolar dendritic fields, dendrites of neighboring AII arboreal dendrites obey the touch-thy-neighbor rule, yielding a coverage factor close to 5. These examples suggest that the coverage factors of rod bipolar and AII populations are roughly constant across eccentricity, in other words, the population spatial density is linked to the dendritic field size of individual cells across eccentricity. This arrangement in turn predicts that spatial pooling in these mosaic populations can be (to first approximation) estimated from population spatial density. 

Relation to Scotopic Vision

Previous studies^9,10 applied anatomic data from macaque monkeys to measurements of human scotopic acuity,¹² and concluded that the AII cell mosaic sets the spatial resolution limit or “bottleneck” for scotopic acuity. We therefore asked whether this conclusion is supported by our direct anatomic measures from human retina. Figure 10A compares the spatial resolution limit (Nyquist frequency) and Snellen acuity of the human AII mosaic to human scotopic acuity values from the study by Lennie and Fairchild¹² (figure 3 in Ref. 12). The curves agree well for eccentricities between 5° and 15°, but scotopic acuity falls below the theoretical limit of the AII mosaic at eccentricities above 15° (arrow, Fig. 10A). The data are redrawn in Figure 10B, together with an estimate based on measurements of ganglion cell density from vertical sections of retina #13587 (Masri RM, et al. IOVS 2018;59:ARVO E-Abstract 2997). For this calculation, we assumed that midget ganglion cells make up 90% of ganglion cells in the fovea, with proportion falling to 46% at 30°,³⁸ and that the sampling limit is based on a single (OFF or ON) midget cell mosaic.¹² Parasol cells are assumed to make up 10% of ganglion cells.³⁹ The midget cell acuity limit agrees well with scotopic acuity at eccentricities above 15°, providing direct support from human retina for previous conclusions based on anatomic data from macaque monkey.^9,10 As expected,¹² parasol cell mosaic resolution lies below the resolution of the AII mosaic throughout the eccentricity range. These data support the view that scotopic visual acuity is limited in central retina by the AII cell mosaic, and in peripheral retina by the midget ganglion cell mosaic. 

Figure 10

 

Comparison of scotopic acuity and retinal population density. (A) Comparison of scotopic acuity (Lennie and Fairchild¹²) with resolution limit (Nyquist limit) and Snellen acuity of the AII array from the current study. Note good correspondence for eccentricities between 5° and 15° (arrow), whereas scotopic acuity falls below AII acuity at greater eccentricities. (B) Same data as A with additional estimate of OFF-midget and OFF-parasol ganglion cell density for retina #13587 (Masri RM, et al. IOVS 2018;59: ARVO E-Abstract 2997). Note that OFF-midget ganglion cell acuity matches scotopic acuity for eccentricities above 15°.

Discussion

Our study provides the first comprehensive view of neuron populations in the primary rod pathway in human retina. Our discussion refers to the temporal horizontal meridian. The nasal horizontal meridian was not analyzed. In nasal retina, cell densities in general are higher,^14,40 but we do not expect the pattern of connectivity to be greatly different to what we report here for temporal retina. 

The findings allow estimates of the convergence between rods, rod bipolar, and AII cells across the human retina, and of the potential spatial resolving power (acuity) of these populations. The peak density of rods (150,000 cells per mm²) was found at eccentricities between 2.5 and 5 mm (or 9–17°, Fig. 6). At this eccentricity, rod bipolar cells have a density of about 10,000 cells per mm² (Fig. 7) and AII cells have a density of about 2000 cells per mm² (Fig. 8). Therefore, at the region where rods are most numerous, the numerical convergence (density ratios) between rods, rod bipolar, and AII cells is 75:4:1 (Fig. 9). In nonhuman primates, the overall rod to AII numerical convergence at the point of peak rod density is lower (i.e., 30:2.5:1 in macaque^9,10 and 50:4:1 in marmoset²¹). It should be noted that in macaque retina values from different retinas (and possibly different species) are compared and thus could be underestimates. 

Close to the fovea, at around 1 mm (or 3.5°) eccentricity the AII cells have a peak density of 4000 cells/mm² (Fig. 8), the rod density is 80,000 cells/mm² (Fig. 6) and the rod bipolar density is 9000 cells/mm² (Fig. 7). Therefore, at 1-mm eccentricity, the numerical convergence of rods to rod bipolar to AII in human retina is 20:2.3:1 (Fig. 9). In macaque, a numerical convergence of ∼10:2:1 was reported at comparable eccentricities.^9,10 

Our data show that the ratio of rods to rod bipolar cells increases only slightly from 13:1 at 1-mm eccentricity to about 18:1 at 9 mm (Fig. 9). These ratios are comparable to those reported for macaque and marmoset retina.^8,21 Taking into account the facts that the size of rod terminals (rod spherules) and the number of invaginating rod bipolar cell processes per rod terminal do not change with eccentricity,^41–43 our findings are consistent with previous studies showing that rod bipolar cells in peripheral retina contact more rods than in central retina.^7,44 

Calculation of the actual convergence in the rod pathway needs to take into account the size of the dendritic tree of rod bipolar cells (rod contacts), and the arboreal dendritic tree of AII cells (rod bipolar contacts). From our measurements we find at 1-mm eccentricity, the mean diameter of the rod bipolar cell dendritic field is 14.5 μm and the local rod density is 80,000 cells/mm² (Fig. 9, right panel). Thus, a rod bipolar cell at 1 mm could contact 13 rods. Parallel calculations give values of 35 rod contacts per rod bipolar cell at 4 mm and 49 at 12 mm. Consistently, Kolb and colleagues⁴⁴ estimated from Golgi preparations that rod bipolar cells in human retina contact 30 to 35 rods at 4.5-mm eccentricity, and 40 to 45 rods at 12-mm eccentricity. Next, from our measurements of calretinin-labeled AII cells, we estimate that at 1-mm eccentricity, an AII cell contacts four rod bipolar cells (Fig. 9D), meaning at 1-mm eccentricity an AII cell could receive input from ∼50 rods. At 4-mm eccentricity, the mean diameter of AII arboreal dendrites was 33.5 μm, and an AII cell could contact nine rod bipolar cells (Fig. 9D), which in turn receive input from ∼300 rods. In more peripheral retina at 9- to 15-mm eccentricity, the mean arboreal dendritic field diameter for AII cells was 109.2 μm. This means that in peripheral retina, an AII cell could receive input from 56 rod bipolar cells (Fig. 9D), which in turn receive input from ∼2500 rods. 

Taken together these data show that in human retina, the actual convergence from rods to AII cells increases from central to peripheral retina. Further, there is considerable similarity between humans, macaque monkeys, and marmosets; the main differences are attributable to a higher rod density in humans compared to monkeys. 

Relevance for Understanding Scotopic Visual Processing

Our analysis based on cell population density implicates the AII mosaic as the spatial bottleneck for scotopic visual acuity within the central 15° of vision, whereas the midget ganglion cell mosaic sets the limit to scotopic acuity at greater eccentricities.^9,10 It is not straightforward to go from cell number to acuity, because acuity also depends on physiologic factors such as neuron contrast sensitivity and receptive field size. Direct behavioral measures in macaque monkeys, however, revealed excellent correspondence between OFF-midget array acuity and grating detection acuity across the first 30° eccentricity.⁴⁵ Our calculations did not incorporate gap junction coupling, which would enlarge the functional receptive field size in the rod bipolar and AII amacrine arrays beyond the anatomical estimates indicated in Figure 9 (reviewed by Bloomfield and Dacheux²). At maximum, gap junction coupling should enlarge the receptive field diameter up to 3-fold with respect to the dendritic field diameter.⁴⁶ We note that this physiologic result predicts a (paradoxical) reduction in acuity between starlight and twilight conditions. 

Two further aspects of scotopic vision which can be related to our measurements are absolute threshold and spatial summation. Absolute scotopic thresholds remain relatively constant outside the central-most 15°.^47,48 This result is broadly consistent with our finding shown in Figure 9C, where we find a steep increase in rod to AII convergence within 5 mm (17°) of the fovea, but much shallower convergence at greater eccentricities, where increases in AII dendritic field diameter are balanced by decreasing rod density. This more or less reciprocal relation may serve to keep the combined effects of receptor and postreceptor noise relatively constant. The relation of our results to the spatial summation area (spatial pool size) of scotopic vision is less clear. A rough calculation based on arboreal dendritic field diameter of AII cells and their rod bipolar inputs gives collecting diameters of 34 μm (6.8 arc minutes) at 1-mm eccentricity (3.4°); 51 μm (12 arc minutes) at 4-mm eccentricity (14°) and 135 μm (28 arc minutes) at 12-mm eccentricity (41°). Scholtes and Bouman,⁴⁹ however, found that at scotopic threshold, the diameter of Ricco's region of complete spatial summation has an almost constant value of 100 minutes between 15° and 40° eccentricity, that is, substantially larger than these anatomic estimates. Presumably here as for scotopic grating acuity, ganglion cell sampling density rather than bipolar or amacrine cells are the limiting factor for spatial pooling in scotopic vision. 

Relevance for Vision Restoration

The AII cells have been proposed as a suitable target cell type to restore light sensitivity in human sufferers of photoreceptor dystrophies.^18,19 The logic behind this suggestion is that the AII cells feed signals into both ON and OFF branches of the cone bipolar system, thus providing access to preexisting retinal circuits feeding parallel retinal output streams. Our results and the current state of understanding of human retinal circuitry support these proposals in three ways. First, our results and the calculations described above predict the maximum spatial acuity afforded by the AII array in humans is close to the limit of scotopic visual acuity (∼7 cycles per degree near the location of peak AII density at 1–2 mm [∼3°–5°]), thus supporting visually guided functions at low but adequate spatial resolution. Second, there is convergent evidence that AII cells do feed signals into cone bipolar pathways within the fovea²⁴ and outside the fovea in human as well as nonhuman primates.^9,10 Thus, restoration of visual function could extend to foveal vision (albeit, as noted above, at low spatial resolution). Finally, we have shown here that the large majority (∼85%) of calretinin-expressing amacrine cells in human retina are AII cells (Figs. 4, 8). Thus, if expression of optogenetic vectors were targeted to intracellular pathways for calretinin expression, then light sensitivity would be conferred selectively onto the AII cell array. 

Acknowledgments

The authors thank Rhian J. Aghajani, Arzu Demir, Rania Masri, Siva Purushothuman, and Felix Weltzien for assistance with experiments and the Lions NSW Eye Bank and Australian Ocular Biobank for providing postmortem eye tissue. 

Supported by National Health & Medical Research Council (NHMRC) project grant (1123418; UG, PRM); Australian Research Council Centre of Excellence for Integrative Brain Function (ARC Centre Grant CE140100007; UG, PRM), Fellowship of the Sydney Medical School Foundation, University of Sydney (UG); Claffy Foundation (SCSL). 

Disclosure: S.C.S. Lee, None; P.R. Martin, None; U. Grünert, None