Transsynaptic Degeneration of Retinal Ganglion Cells Following Lesions to Primary Visual Cortex in Marmosets

Abstract

Purpose: A lesion to primary visual cortex (V1) in primates can produce retrograde transneuronal degeneration in the dorsal lateral geniculate nucleus (LGN) and retina. We investigated the effect of age at time of lesion on LGN volume and retinal ganglion cell (RGC) density in marmoset monkeys.

Methods: Retinas and LGNs were obtained about 2 years after a unilateral left-sided V1 lesion as infants (n = 7) or young adult (n = 1). Antibodies against RBPMS were used to label all RGCs, and antibodies against CaMKII or GABA_A receptors were used to label nonmidget RGCs. Cell densities were compared in the left and right hemiretina of each eye. The LGNs were stained with the nuclear marker NeuN or for Nissl substance.

Results: In three animals lesioned within the first 2 postnatal weeks, the proportion of RGCs lost within 5 mm of the fovea was ∼twofold higher than after lesions at 4 or 6 weeks. There was negligible loss in the animal lesioned at 2 years of age. A positive correlation between RGC loss and LGN volume reduction was evident. No loss of CaMKII-positive or GABA_A receptor-positive RGCs was apparent within 2 mm of the fovea in any of the retinas investigated.

Conclusions: Susceptibility of marmoset RGCs to transneuronal degeneration is high at birth and declines over the first 6 postnatal weeks. High survival rates of CaMKII and GABA_A receptor-positive RGCs implies that widefield and parasol cells are less affected by neonatal cortical lesions than are midget-pathway cells.

It has been known for over a century that visual cells in the eye and brain depend on each other for survival. Van Buren¹ reviewed the earlier literature on primates and other mammals and reported his own results of a longstanding lesion to primary visual cortex (V1) in an adolescent macaque monkey. He showed that 4 years after right-sided removal of V1, the right dorsal lateral geniculate nucleus (LGN) was almost completely degenerated, and the right hemiretinas showed cystic degeneration of bipolar and ganglion cells. Cowey² later estimated there was 80% loss of retinal ganglion cells near the fovea in one macaque, 7 years after bilateral removal of V1 in adolescence. Subsequent studies have manipulated the age at time of cortical lesions, size and location of lesion, and survival time and demonstrated that all of these factors play a role in survival of LGN and ganglion cells in macaque monkeys. 

Weller et al.³ and Cowey et al.⁴ described selective degeneration in the parvocellular layers of LGN and in small ganglion cells (presumed midget class) 4 or more years after lesions to striate cortex in infant³ or young adult⁴ macaques. Ptito et al.⁵ showed, however, all LGN layers appeared degenerated 2 years following hemidecortication in infant (4 months postnatal) macaques. Dineen and Hendrickson⁶ showed that macaques lesioned in infancy had lost up to 80% of ganglion cells 1 year after V1 removal, whereas adult-lesioned animals surviving for 1 year showed only a 15% loss. A broadly consistent result was obtained by Weller and Kaas,⁷ who found 80% loss of parvocellular-projecting ganglion cells following lesions in infant macaque and ∼20% loss following lesions in adolescent or adult animals. Consistently, in marmoset monkeys, Hendrickson et al.⁸ showed substantial (∼70%) loss of ganglion cells following V1 lesions in infants but only a 20% loss following lesions in adults. In sum, there is broad agreement that the parvocellular division of the cortical afferent visual pathway is susceptible to cortical lesions in early life, but the age at time of lesion and, possibly, species-dependent differences in susceptibility play a role (for discussion, see Weller and Kaas⁷). 

In the present study, we compare ganglion cell loss following cortical lesions in infant marmosets, at a range of postnatal ages including that tested by Hendrickson et al.⁸ (2 weeks). We confirm their findings and further show that lesions at later ages (between 2 and 7 weeks) do cause ganglion cell loss but to a lesser extent. Using selective ganglion cell markers, we provide direct evidence that parasol-magnocellular pathway ganglion cells and widefield (koniocellular- and collicular-projecting) ganglion cells are unaffected by neonatal lesions to V1. 

Materials and Methods

V1 Lesions

All procedures were conducted according to the ARVO Animal Statement and provisions of the National Health and Medical Research Council (NHMRC) code of practice for the care and use of animals and were approved by the Monash University Animal Ethics Committee. The V1 lesion surgery was based on the technique introduced by Rosa et al.⁹ The procedure involves an occipital lobectomy along a vertical plane along the border between V1 and the second visual area, resulting in a complete loss of the representation of the visual field up to 10° eccentricity along the vertical meridian and 20° to 30° along the horizontal meridian. Reconstructions of lesions and visual field defects created with this procedure can be found in earlier publications.^10,11 Briefly, the animal was anesthetized with isoflurane (1%–2%), and then a craniotomy was made over the occipital pole of the left hemisphere and cortical tissue was excised caudal to a plane extending from the dorsal surface of the occipital lobe to the cerebellar tentorium. The size of the excision was adjusted to encompass the maximum possible extent of V1 without encroachment onto extrastriate visual areas, as estimated during surgery by reference to the characteristic pattern of capillary vessels at the V1/V2 border¹² and confirmed for each case postmortem by anatomic reconstruction, as illustrated in earlier publications.^10,13 The skull and skin were reconstructed and the animal was placed in a humidicrib until recovery of movement, then returned to parental care. Throughout the postlesion period, the animal was housed with family groups, having access to both indoor and outdoor environments. 

Tissue Collection and Preparation

Eyes or retinal tissue were obtained from animals that had received V1 lesions as described above.^13,14 With exceptions of case W6L and case W6F, tissue was available from only one eye (normally the left eye) of lesioned marmosets. The other eye tissue of the lesioned animals was used for other experiments. In addition, two retinas from nonlesioned marmosets were analyzed (Table). Animals were overdosed with pentobarbitone sodium (100 mg/kg) following unrelated electrophysiologic recording experiments, then perfused with saline followed by 4% paraformaldehyde (PFA) in 0.1 M phosphate buffer (PB). Subsequently, the eyes and brain were removed, the cornea and lens were dissected out, and the posterior eyecup was postfixed in the same fixative for 1 hour at room temperature. After rinses in PB, the retina was dissected, immersed in 30% sucrose in PB overnight, frozen in liquid nitrogen, and kept in the freezer until further use. Vertical 100-µm-thick vibratome sections through foveal retina were obtained and processed as described previously.¹⁵ 

Table.

 

Animals

 

The brains were postfixed in the buffered PFA for 24 hours, followed by immersion in the same fixative containing increasing concentrations of sucrose (10%, 20%, and 30%). Frozen 40-µm coronal sections were obtained in five series; one series was devoted to the NeuN and one was devoted to Nissl staining. The other series were used for other purposes. 

Immunofluorescence

A guinea pig antibody against the RNA binding protein with multiple splicing (RBPMS, #1832-RBPMS, RRID: AB_2492226; PhosphoSolutions, Aurora, CO, USA) was used to label all retinal ganglion cells.¹⁶ A mouse monoclonal antibody against calmodulin-dependent protein kinase II (CaMKII, #ab22609, clone 6G9, RRID: AB_447192; Abcam, Cambridge, UK) was used to label CaMKII-expressing ganglion cells.¹⁷ Two monoclonal antibodies against GABA_A receptor subunits (bd17 alpha chain; bd24 beta 2,3 chain; Boehringer Mannheim, Mannheim, Germany) were used to label foveal parasol cells.^18,19 Retinal flat-mount preparations and vibratome sections through the central retina were prepared and processed as described in detail previously.^17,19 

DAB NeuN Immunostaining

For NeuN immunostaining of LGN,¹⁴ brain sections were incubated in blocking solution (10% normal horse serum and 0.3% Triton-X100 in 0.1 M PB) for 1 hour at room temperature, then incubated with the primary antibody for NeuN (1:800, MAB377; Merck Millipore, Bayswater VIC, Australia) at 4°C for 42 to 46 hours. Secondary antibody (1:200, PK-6102, Vectastain elite ABC HRP kit; Vector Laboratories, Newark, CA, USA) was applied for 30 minutes at room temperature, followed by avidin/biotin interaction and DAB staining (DAB Peroxidase Substrate Kit SK-4100; Vector Laboratories, Newark, CA, USA). 

Microscopy

Vibratome sections through foveal retina were imaged using a confocal scanning microscope (LSM700; Zeiss, Jena, Germany) equipped with 405-, 488-, 555-, and 635-nm lasers using a 20× air objective (Plan Apochromat no. 420650-9901, Zeiss, Germany) at a resolution of 2048 × 2048 or 1024 × 1024 pixels and a z-axis step size of 0.87 to 1.15 µm for each optical section. Tiled stacks of each vibratome section were stitched together using Zeiss ZEN Black Software. Cells in flat-mounted regions of the retina were imaged en face. The contrast and brightness of the images were adjusted using Zen Blue (Zeiss) or Adobe Photoshop (Adobe, San Jose, CA, USA) software. 

Analysis

Density Measurements

In vibratome sections, the cell densities were determined from tiled image stacks using Zen Blue software as described in our previous publications.^20,21 Cells were counted across the entire length and depth of at least three vibratome sections, except where the retinal layers were mechanically distorted or the immunohistochemical staining was too weak to obtain accurate cell counts. Individual vibratome sections were separated into bins (normally 100 µm width), and cell nuclear profiles were counted at various eccentricities along the temporal horizontal meridian at a single z-plane position. For at least three reference bins in each section, cells were also counted across a minimum depth of 10 µm in the z-plane. The resulting volumetric densities (cells/mm²/µm depth) were used to convert the single z-plane counts to areal densities (cells/mm² of retinal surface area). Density measurements were pooled in 0.25-mm bins for vibratome sections and in 0.25- to 1-mm bins for wholemount preparations. Outlier data points (>2 SD away from the local pooled bin mean) were discarded. As noted above, all animals received lesions to V1 on the left side of the brain. In the following, we therefore refer to the temporal retina of the left eye and the nasal retina of the right eye as ipsilesion hemiretinas. Likewise, we refer to the nasal retina of the left eye and the temporal retina of the right eye as contralesion hemiretinas. For each eye, densities at matched eccentricities in the ipsilesion and contralesion hemiretinas were measured in 0.25-mm bins. Ganglion cell loss was estimated as L = 100  ×  [(D_c −  D_i)/D_c], where D_c is ganglion cell density in the contralesion hemiretina, D_i is ganglion cell density in the ipsilesion hemiretina, and L is cell loss (%). Pairwise comparisons were made using the Wilcoxon nonparametric rank-sum test for paired samples. 

For LGN volume measurement, sections containing the LGN were scanned using Aperio Image Scope software (Leica Biosystems, Wetzlar, Germany). Estimates of the volume of the LGN were obtained using the Cavalieri estimator^22,23 based on the area measured in 16 to 22 equally spaced sections through the LGN. The total volume of the LGN was calculated based on the sum of the volumes obtained for each section (LGN area × thickness). In this calculation, the thickness estimates took into account the separation between sections (except for the last one, which was 40 µm). All measurements were corrected with a shrinkage factor of 0.801, which was obtained based on measurements of the known distance between electrode tracks in the contralateral cortex and their separation in the sections.²⁴ 

Results

Spatial Density of Ganglion Cells in Central Retina

As reported previously, antibodies against RBPMS can be used to label ganglion cells in marmoset retina.^17,25 Figure 1 shows RBPMS labeling in a section through the fovea of a normal (nonlesioned) adult marmoset (case M2018). As expected, labeled cells densely occupy the ganglion cell layer with symmetric distribution around the center of the foveal pit (Fig. 1A). The ganglion cell layer comprises five to six layers of labeled ganglion cell somas close to 0.5 mm (4 degrees)²⁶ eccentricity (Figs. 1C, 1D) and about three layers at 1 mm (8 degrees) eccentricity (Figs. 1B, 1E); these data are consistent with previous studies of marmoset retina.^8,27,28 

Figure 1.

 

Retinal ganglion cells in normal marmoset. Confocal images of a vertical section through the fovea of an adult marmoset. The layers of the retina are visible with Nomarski optics and DAPI nuclear stain (blue). The section was processed with antibodies against RBPMS to reveal retinal ganglion cells (green). (A) Overview image of the entire section. The scale shown at the bottom of the image indicates retinal eccentricities in millimeters. (B–E) Regions of interest taken at different eccentricities (indicated at the lower right corner). GCL, ganglion cell layer; INL, inner nuclear layer; N, nasal; ONL, outer nuclear layer; IPL, inner plexiform layer; T, temporal. Scale bar shown in B = 50 µm, applies to B–E.

In animals subject to left-sided V1 lesioned in infancy, the ganglion cell layer shows thinning in the temporal retina of the left eye (i.e., the ipsilesion hemiretina). Figures 2A–E show RBPMS labeling in a section through the left fovea of case W2E (lesion at 2 postnatal weeks). The thickness of the ganglion cell layer in nasal retina (Figs. 2B, 2D) is comparable to normal (Figs. 1B–E), but thickness is reduced in the temporal retina with only three to four layers of labeled ganglion cell somas at 0.5 mm (4 degrees) eccentricity (Fig. 2C) and two layers at 1.0 mm (8 degrees) eccentricity (Fig. 2E). This result is consistent with that reported by Hendrickson et al.⁸ following V1 lesions at 2 postnatal weeks. 

Figure 2.

 

Retinal ganglion cell density is reduced in temporal retina ipsilateral to V1 lesions. Confocal images are shown of sections that were processed with antibodies against RBPMS to reveal retinal ganglion cells (green). The layers of the retina are visible with DAPI nuclear stain (white), revealing the nuclei of cell bodies. (A–E) Confocal images of a vertical vibratome section through the left central retina of an adult marmoset (case W2E, 2 years age), which received a left-sided lesion to V1 at 2 weeks of age. (A) Overview image of the entire section. (B–E) Regions of interest taken at different eccentricities (in millimeters) as indicated in the lower right corner. (F, G) Confocal images of sections through the temporal hemiretinas of an adult marmoset (case W6L), which received a left-sided lesion to V1 at 6 weeks of age. (F) Normal cell density in the right (contralesion) temporal retina. (G) Thinning of the ganglion cell layer in the left (ipsilesion) temporal retina. Scale bar shown in A = 500 µm, scale bar shown in D = 50 µm, applies to B–E; scale bar shown in F = 50 µm, applies to F and G.

Apart from the overt thinning in the ganglion cell layer of the ipsilesion hemiretinas, there were no obvious signs of long-term pathology in the retinas of lesioned animals, and there were no obvious changes in the (sparse) distribution or numbers of displaced ganglion cells in the inner nuclear layer in regions with loss of ganglion cells. The morphology of Müller cells was not overtly changed in regions of ganglion cell loss, but we did not specifically search for changes in glial fibrillary acid protein expression in Müller cells or astrocytes in the present study. 

Comparison of the ganglion cell layer thickness in the retina of a normal animal (Fig. 3A) and the left temporal retina of lesioned animals demonstrates that the reduction is greatest in an animal that was lesioned at 1 week postnatal (Fig. 3B, two layers), followed by an animal lesioned at 2 weeks postnatal (Fig. 3C, three layers). Density appears only slightly reduced in an animal that received a lesion at 6 postnatal weeks (Fig. 3D, five layers). These results provide qualitative evidence that the age at lesion is correlated with retinal ganglion cell loss. Quantitative data to support this assertion are presented in a later section. It is nevertheless important for the reader to remember that our analysis is based on a small number of animals, with most of the lesion time points represented by a single case (Table). This limitation reflects the nature of experimental research on primates, where large-scale population comparisons of different age groups are difficult to justify on ethical or financial grounds. 

Figure 3.

 

Retinal ganglion cell density is reduced in temporal retina of marmosets that received a left-sided lesion to V1 at different postnatal ages. Confocal images show vertical sections through left retina of animals that received V1 lesions at different postnatal ages (indicated in the upper right corner) in comparison to the retina of a normal animal. All images are taken at 0.5 mm (4 degrees) eccentricity (temporal; i.e., ipsilesion hemiretina). The sections were processed with antibodies against RBPMS to reveal retinal ganglion cells (green). The layers of the retina are visible with DAPI nuclear stain (white), revealing the nuclei of cell bodies. Scale bar shown in C = 50 µm, applies to all.

Degeneration in the LGN

As previously shown,^8,14,29 removal of most of the primary visual cortex (V1) in marmosets leads to substantial retrograde degeneration in the region of dorsal LGN, which normally innervates the lesioned part of V1. Figures 4A, 4B show a 40-µm coronal section through the left and right LGNs, respectively, of an adult (2 years postnatal age) marmoset that had received a left-sided V1 lesion at 2 weeks postnatal age (case W2D). A wedge-shaped zone of degeneration extending through all LGN layers and corresponding to the visuotopic location of excised cortex is apparent in the left LGN (Fig. 4A).^14,30,31 By contrast, the normal lamination pattern^30,32 is preserved in the right LGN (Fig. 4B). Figures 4C, 4D show these images schematically with the borders of the lesion projection zone (lpz) and the remnants of the parvocellular (P) and magnocellular (M) layers in the left LGN. Anatomic reconstruction of the cortical lesion in this animal, as well as the extent of the visual field scotoma, appears on page 12481 of Yu et al.¹⁰ Quantitative comparison of LGN volume with ganglion cell degeneration is given in a later section. 

Figure 4.

 

Degeneration in the left LGN following a lesion of the left V1 at 2 weeks postnatal and 2-year survival period. Case W2D. (A) Nissl-stained 40-µm coronal section through left LGN. The lesion projection zone (lpz) is visible as a wedge-shaped zone of degeneration extending through all LGN layers. (B) Section through the right LGN showing that the normal layering pattern is preserved on the nonlesioned side of the brain. (C, D) Schematic drawings of these sections showing, in the left LGN, the borders of the lpz and the remnants of the parvocellular (P) and magnocellular (M) layers.

Ganglion Cell Degeneration in Central Retina

Figure 5A shows ganglion cell densities in the retina of a normal (nonlesioned) adult marmoset (case M2018; these data were previously published in Nasir-Ahmad et al.³³). In normal marmosets, there is up to twofold variation in peak ganglion cell density,^8,28,34 but ganglion cells within 5 mm eccentricity of the foveal center are nevertheless distributed with a very high degree of radial symmetry.^8,28,34 Therefore, as expected, the ganglion cell densities in the normal retina are almost identical at equivalent eccentricities in nasal and temporal retina (Fig. 5A), falling in parallel from a peak near 60,000 cells/mm² at 0.5 mm to ∼10,000 cells/mm² at 2 mm. Figures 5B–D show data from marmosets subjected to left-sided V1 lesions in the first (Fig. 5B), second (Fig. 5C), or sixth (Fig. 5D) postnatal week. Data from the left eye of each lesioned animal are shown, meaning that transsynaptic degeneration, if present, should be evident in the temporal (ipsilesion) hemiretina. 

Figure 5.

 

Spatial density of RBPMS-positive ganglion cells in the central marmoset retina. Density is plotted against eccentricity. (A) Normal (nonlesioned) animal, case M2018. Note near-to-identical distribution of ganglion cells in nasal versus temporal retina. (B) Left retina following left-sided V1 lesion at 1 week postnatal, case W1A. Note asymmetric ganglion cell distribution indicating degeneration in the temporal (left) hemiretina. (C) Left retina following left-sided V1 lesion at 2 weeks postnatal, case W2E; note similar degree of degeneration as in B. (D) Left retina following left-sided V1 lesion at 6 weeks postnatal; note reduced degree of degeneration compared to animals lesioned at earlier postnatal ages. Points show means of pooled data taken in 0.25-mm eccentricity bins from at least three vibratome sections from one retina in each case. Error bars (in some cases smaller than the data symbols) show standard deviations. (E) Cell loss estimated at matching eccentricities within the first 2 mm. *P < 0.05, Wilcoxon nonparametric rank test for paired samples; A, animal lesioned at 2.3 years postnatal; N, case M2018, normal (nonlesioned) animal. (F) Cell loss estimates for all available eccentricities for these cases. Cumulative cell loss between the fovea and indicated eccentricity is shown. Negative values in cases W6F and M2018 are set to zero.

Previous studies of normal human retina^21,35 and normal marmoset retina (as noted above) showed there can be up to twofold interindividual variation in peak ganglion cell density. This fact makes it important to compare densities at equivalent eccentricities in the ipsilesion and contralesion hemiretina of the same eye. The effect of V1 lesions on the retina is manifest as a change from a symmetrical pattern in normal retina (Fig. 5A) to marked asymmetry following cortical lesions at 1 (Fig. 5B) or 2 (Fig. 5C) postnatal weeks. This density difference following a lesion at 1 postnatal week corresponds to a mean (SD) 51.4% (11.7%; n = 7) loss of ganglion cells at matched eccentricities within 2 mm of the ipsilesion fovea (P < 0.001, Wilcoxon nonparametric rank-sum test for paired samples [henceforth, PWIL]). The mean (SD) loss following a lesion at 2 postnatal weeks is 51.8% (18.3%; n = 7; P = 0.016, PWIL). A milder asymmetry is evident after a V1 lesion at 6 postnatal weeks (Fig. 5D), corresponding to mean (SD) loss of ganglion cells of 31.2% (10.9%; n = 7; P = 0.016, PWIL). 

Figure 5E summarizes foveal ganglion cell loss across each eye analyzed. The mean cell loss for matched eccentricity samples in ipsilesion versus contralesion hemiretina, taken in 0.25-mm bins across the central-most 2 mm, is shown. For case W6F, the density of ganglion cells in the nasal (ipsilesion) hemiretina of the right eye was marginally, but significantly, higher than that in the temporal hemiretina (P = 0.02, PWIL). In all other cases, the density in the ipsilesion hemiretina was lower than that in the contralesion hemiretina. Adequate retina tissue from case W6H (Table) was not suitable for quantitative analysis; therefore, this case is not shown on this plot. Figure 5F shows cumulative cell loss between the foveal center and indicated eccentricities between 1 and 6 mm. The reader should note that these percentage plots do not take into account the increase in retinal surface area for equal eccentricity increments as distance from the fovea increases. The plots show mild increases in cumulative loss at eccentricities above 1 mm, which is consistent with evidence presented below that both central and peripheral retina are affected by cortical lesions. 

Ganglion Cell Degeneration in Peripheral Retina

In four animals subjected to V1 lesions as infants (case W1A, case W2D, case W2E, case W4D) and the single animal lesioned at 2 years postnatal (case WA13), we were able to measure equivalent eccentricities in nasal and temporal peripheral retina at close to the horizontal meridian (other pieces of these retinas and peripheral retinas from other animals were used for other purposes). Figure 6 compares values from a nonlesioned animal (case M2018, Fig. 6A) with animals lesioned at 1 week (case W1A, Fig. 6B) or 4 weeks (case W4D, Fig. 6C): here, it is evident that there is cell loss in the temporal (ipsilesion) hemiretina, extending to at least to 5 mm (40 degrees) eccentricity in both cases. At 5 mm (40 degrees) eccentricity, this density difference is equivalent to, respectively (case, postnatal age at lesion, cell loss): W1A, 1 week, 64%; W2D, 2 weeks, 17%; W2E, 2 weeks, 41%; W4D, 4 weeks, 44%; and WA13, 2 years, 37%. These values must be treated with caution, in light of established nasotemporal density asymmetries in ganglion cell size and density (cells have smaller dendritic fields and are more tightly packed in nasal retina above ∼4 mm [∼32 degrees] eccentricity).^28,36–38 The nasotemporal density difference in the normal (nonlesioned) adult marmoset (Fig. 6A), for instance, would be equivalent to 19% cell loss at 5 mm (40 degrees). It is nevertheless clear that, at least for the animals lesioned in infancy, the nasotemporal asymmetry is greater than normal, indicating that transneuronal degeneration has occurred in peripheral as well as central retina. 

Figure 6.

 

Spatial density of RBPMS-positive ganglion cells in one normal and two lesioned animals. Density data are plotted on a logarithmic scale to facilitate comparison of peripheral eccentricities. (A) Case M2018, normal (nonlesioned) animal. Note symmetric distribution of cell density at eccentricities below 5 mm (∼40 degrees). (B) Case W1A, lesioned at 1 week postnatal. Note extensive cell loss in temporal (ipsilesion) hemiretina. (C) Case W4D, lesioned at 4 weeks postnatal. Note moderate cell loss at eccentricities below 5 mm (∼40 degrees) in temporal (ipsilesion) hemiretina. Error bars (in some cases smaller than the data symbols) show SEM.

In sum, across the animals we studied, there is substantial variation but a clear indication that (1) cells are lost from the ipsilesion hemiretinas, (2) more cells are lost following V1 lesions at 1 or 2 weeks than following lesions at later postnatal ages, and (3) both central and peripheral retina are subject to transneuronal degeneration following V1 lesions. 

Spatial Density of Parasol Ganglion Cells in Central Retina

Previous studies of macaque retina have shown that parasol ganglion cells in central retina express GABA_A receptor subunits.^18,19 Here we first present evidence that in marmoset retina, antibodies against GABA_A receptor subunits produce comparable staining patterns to those obtained in macaque, then show that there is little sign of parasol cell loss following V1 lesions in infant marmosets. 

In normal marmoset fovea, GABA_A receptor labeling is strongly expressed in the inner plexiform layer as well as in small somas in the inner nuclear layer and large somas in the ganglion cell layer (Figs. 7A, 8A). Double labeling with antibodies against RBPMS shows that the large GABA_A receptor-positive cells in the ganglion cell layer are double labeled and thus are ganglion cells (Figs. 7B, 8A). Based on their soma size and morphology, we suggest that these cells are parasol cells. Intracellular DiI injection into GABA_A receptor-positive cells in peripheral retina of a normal animal confirms that these cells include parasol cells (Fig. 7F). Taken together. we conclude that GABA_A receptor subunits in central marmoset retina are expressed by parasol cells. 

Figure 7.

 

Parasol ganglion cell density is comparable in temporal and nasal retina of lesioned animals. (A–E) Confocal images of vertical sections through the fovea of marmosets. The sections were processed with antibodies against RBPMS to reveal retinal ganglion cells (green) and antibodies against GABA_A receptors (magenta) to label parasol cells. DAPI labeling revealing nuclei is shown in white. (A, B) Images of sections from a normal animal. In A, the GABA_A receptor labeling is shown together with differential contrast optics to reveal retina layers. (C–E) Images of sections from a marmoset that received a lesion of the left V1 at the age of 2 weeks. (C) Overview image of the foveal region. The rectangles indicate the regions shown in D and E taken at 0.5 mm in contralesion nasal (N) and ipsilesion temporal (T) retina, respectively. There is a reduction in the thickness of the GCL in temporal retina, but the density of presumed parasol cells is comparable to that in nasal retina. (F) DiI-injected GABA_A receptor-positive inner stratifying parasol ganglion cell in a nonlesioned animal at 5.35 mm eccentricity. The inset on the right shows the GABA_A receptor-positive soma (green); DiI labeling in the soma is shown in magenta. Scale bar shown in C = 100 µm, applies for A and C; scale bar shown in D = 50 µm, applies to B–D; scale bar shown in F = 50 µm.

Figure 8.

 

Parasol ganglion cell density in normal and lesioned animals is comparable. Confocal images of sections through temporal (T) left retina of a normal (A) and two lesioned animals (B, C) taken at comparable eccentricities (indicated in millimeters in the upper right corner). The sections were processed with antibodies against RBPMS to reveal retinal ganglion cells (green) and antibodies against GABA_A receptors (magenta) to label parasol cells. Scale bar shown in A = 50 µm, applies to all.

We thus used antibodies against GABA_A receptor subunits to identify parasol cells in the retina of animals lesioned at 1 week (Fig. 8B) and 2 weeks of age (Figs. 7C–E, 8C). At 0.5 mm (4 degrees) eccentricity, the number of GABA_A receptor subunit–expressing parasol cells in both nasal (Fig. 7D) and temporal retina (Fig. 7E) appears comparable to that in normal retina (Fig. 7B). Similarly, no obvious differences in the expression of GABA_A receptor subunits were seen at 0.75 mm (6 degrees) eccentricity in normal (Fig. 8A) and lesioned animals (Figs. 8B, 8C; see also Fig. 9 and the following text section for related results). A small number of spatial density measurements of GABA_A receptor-positive parasol cells were also made, in central retina of animals lesioned at 1 week (case W1A), 2 weeks (case W2E), and 6 weeks (cases W6L, W6F). Here, as in the experiments described above (in the section “Ganglion Cell Degeneration in Central Retina”), as far as possible, we measured parasol density at matched eccentricities in ipsilesion and contralesion hemiretinas. These measurements showed closely matched densities of parasol cells in ipsilesion and contralesion hemiretinas (Figs. 10A, 10C). These data are consistent with other evidence that parasol cells are less susceptible to retrograde transsynaptic degeneration than are midget cells.^3,4,7,8 

Figure 9.

 

Widefield ganglion cell density in normal and lesioned animals is comparable. Confocal images of sections through temporal (T) left retina of a normal (A) and three lesioned animals (B–D) taken at comparable eccentricities (indicated in millimeters in the upper right corner). The sections were processed with antibodies against RBPMS to reveal retinal ganglion cells (green) and antibodies against CaMKII to label widefield ganglion cells. Scale bar shown in B = 50 µm, applies to all.

Figure 10.

 

Preservation of parasol and widefield cells in foveal retina following V1 lesions. (A) Case W2E. Density of parasol cells labeled with GABA_A receptor antibodies in left eye following a left-sided V1 lesion at 2 weeks postnatal age. Note comparable density in nasal (contralesion) and temporal (ipsilesion) retina. (B) Case W2E. Density of widefield cells labeled with CaMKII antibodies in left eye following a left-sided V1 lesion at 2 weeks postnatal age. Note comparable density in nasal (contralesion) and temporal (ipsilesion) retina. (C) Density of widefield cells labeled with CaMKII antibodies in the right eye of a normal (nonlesioned) animal. (D) Comparison of overall ganglion cell loss (blue bars, data from Figure 5) with loss of parasol cells labeled with GABA_A receptor antibodies (red bars) and widefield cells labeled with CaMKII antibodies (yellow bars) following lesions at 1 week postnatal age (case W1A), 2 weeks postnatal age (case W2E), and 6 weeks postnatal age (cases W6F, W6L; mean values are shown). Error bars (in some cases smaller than the data symbols) in A–C show SEM.

Spatial Density of Widefield Ganglion Cells in Central Retina

We have previously shown that CaMKII in marmoset retina is expressed by a variety of widefield ganglion cell types.¹⁷ Consistently, double-label immunofluorescence with antibodies against RBPMS shows that in a nonlesioned animal, CaMKII is expressed by a small population of retinal ganglion cells in central retina in addition to some amacrine and displaced amacrine cells (Fig. 9A). Similar staining pattern is observed in animals lesioned at 1, 2, and 4 weeks (Figs. 9B–D). These qualitative data, as well as a small number of cell counts made in sections from the same pieces of retina as used to count parasol cells (case W2E; see above), showed that the spatial density of CaMKII-expressing ganglion cells is comparable at matched eccentricities in ipsilesion and contralesion hemiretinas (Fig. 10B) and is also close to the spatial density of CaMKII-expressing ganglion cells in central retina of a normal (nonlesioned) animal (Fig. 10C; data reanalyzed from Baldicano et al.¹⁷). We can therefore conclude that widefield ganglion cells are not substantially susceptible to retrograde transsynaptic degeneration following lesions to V1. 

Relation to LGN Volume Loss

Finally, we asked whether and how the transneuronal degeneration evident in LGN of infant marmosets subject to V1 lesions is related to loss of ganglion cells. Figure 11A shows, in a scatterplot, the same data as shown in Figure 5E, together with linear regression (solid line) and 95% confidence intervals (dashed lines). For purpose of comparison, this scatterplot includes case WA13, which was lesioned at 2 years of age. The age at lesion of this case was arbitrarily set at 8.1 weeks. The age-related decline in ganglion cell susceptibility is thus somewhat overestimated by this plot, but it is nevertheless clear that ganglion cell loss is greater in animals lesioned in the first 2 postnatal weeks than in animals lesioned at greater postnatal ages. Figure 11B shows that the age at lesion is, likewise, inversely related to the (well-established) loss of LGN volume following striate cortex lesions in monkeys.^1,2,14,39,40 Accordingly, there is a mild positive correlation between ganglion cell loss and LGN volume loss (Fig. 11C). In other words, these data suggest that both the retina and LGN become less vulnerable to cortical lesions over the first 2 postnatal months. We consider the source and relevance of these correlations below. 

Figure 11.

 

Relation of ganglion cell loss to geniculate volume loss following cortical area V1 lesions. Solid lines show linear regression. Dashed lines show 95% confidence intervals. (A) Scatterplot of age at lesion against retinal ganglion cell (RGC) loss. (B) Scatterplot of age at lesion against dorsal LGN volume loss. (C) Scatterplot of RGC loss against LGN volume loss. A, case WA13, animal lesioned at 2.3 years postnatal; N, nonlesioned (normal) animal.

Discussion

We show here that in marmosets, susceptibility of retinal ganglion cells to cortical lesions is high at 1 week after birth but decreases in the first postnatal months. Our results also indicate that the parvocellular projecting midget ganglion cells appear to be most affected. It is important to note that all of these animals survived more than 1 year postlesion. It is likely that LGN degeneration has completed by this time, as our previous study in marmoset LGN¹⁴ showed that volume loss in LGN of adult animals is complete by 6 to 7 months after a lesion to V1. In the following, we first consider the time course of susceptibility in relation to other developmental events in marmoset visual system, then weigh up the evidence for differential involvement of parallel afferent pathways. Finally, we ask how LGN and retinal susceptibility to cortical lesions could be related. 

Relation to Other Developmental Events

Marmoset gestation (152 days) is only slightly shorter than macaque (165 days), and the LGN volume in both species reaches maximum at 1 to 3 months after birth.^41,42 Thus, we can safely assume that in the animals we studied, the pattern of retinogeniculate connections at time of lesion is already adult-like.⁴³ In the retina, the centripetal migration of cones, as well as centrifugal migration of ganglion cells, is not complete until 8 months after birth,⁸ but the pattern of synaptic marker expression at birth is consistent with adult-like synaptic connectivity.⁴⁴ 

There are two main temporal correlates of increasing ganglion cell resilience to cortical lesions during the first postnatal month. The first correlate is to the overall volume of LGN, which doubles over the first postnatal month.⁴² This increase is accompanied by rapid production of synapses and spines in the LGN,⁴⁵ which is followed by regression across the next few postnatal months. These changes may be associated with stabilization of geniculostriate and other connections. The second correlate is a reorganization of thalamocortical pathways, whereby initially widespread inputs from LGN and pulvinar to association area MT become dominated by input from V1.^29,46 These dorsal thalamic inputs to MT can be stabilized by neonatal lesions to V1,²⁹ indicating competitive interaction between transcortical and subcortical pathways during early postnatal development. As described below, it is possible that this stabilization can increase the capacity of LGN cells to produce some trophic factor that keeps ganglion cells alive. 

Given the transneuronal nature of retinal degeneration, similar studies dealing with LGN degeneration in neonatal animals would be very informative, but currently, such data are not available. Nevertheless, our study in marmoset adds to previous data^6,8 by looking at changes in retinal degeneration in the neonatal phase with much more resolution, showing a sharp decline in the first 2 postnatal weeks. 

Susceptibility of the Midget-Parvocellular Pathway

Our results add to previous evidence that in macaque and marmoset monkeys, the midget-parvocellular pathway is especially susceptible to early life damage to the primary visual cortex.^3,4,7,8 We show quantitatively, using specific cell markers, that both parasol-magnocellular pathway ganglion cells and a variety of widefield ganglion cells in marmoset central retina are immune to the effects of cortical lesions to at least the age of postnatal 6 weeks (Figs. 8–10). These results are consistent with the hypothesis that midget-parvocellular pathway cells are vulnerable because they lack a sustaining collateral to other subcortical centers^3,6 (but see also Weller and Kaas⁷). Relatedly, we have shown previously that the midget and small bistratified (“blue-on/yellow-off”) ganglion cell populations project exclusively to LGN, whereas all other ganglion cells types project additionally to the pretectum, superior colliculus, or pulvinar nucleus.^47–49 

It would be interesting to establish whether the different LGN divisions (parvocellular, magnocellular, and koniocellular) are equally affected by cortical lesions, but this question could not be addressed directly in the cohort of animals studied here. Previous work⁵⁰ indicated that some koniocellular cells do survive cortical lesions and maintain their projections to extrastriate cortex. But precise quantification of the relative sparing is difficult because the normal lamination pattern is disrupted, making it difficult to assign cells to layers based on their position. Further, cortical lesions change the LGN neurochemistry, whereby calbindin (an established marker for K cells^51–53) is expressed in magnocellular and parvocellular cells.⁵⁰ The most straightforward explanation of our results is nevertheless that midget-parvocellular pathway cells are selectively lost: midget cells make up ∼80% of all ganglion cells in central marmoset retina, where our measurements are concentrated.⁵⁴ What remains unexplained is why, despite the lack of a sustaining collateral, the resilience of the midget pathway to cortical lesions should increase over the first 2 postnatal months. The increased resilience may be related to the stabilization of dorsal thalamic inputs to extrastriate cortex following V1 lesions, as noted above. The increased resilience may also be linked to developmental critical periods⁵⁵ that are associated with a multitude of changes in visual circuitry, including the presence of specific growth factors⁵⁶ and/or changes in neuronal activity, leading to better cell survival.⁵⁷ It is also possible that changes in neurochemistry comparable to those reported in LGN^13,50 can lead to increased ganglion cell resilience across the first 2 postnatal months. 

Relation of LGN and Retinal Susceptibility

Following a V1 lesion in primates, there is rapid and largely complete retrograde degeneration of the LGN neurons that normally would project to the excised cortex. Transneuronal retrograde degeneration of ganglion cells following V1 lesions is, however, considered a much more variable affair, depending (as described in the “Introduction” section) on factors including age at time of lesion, survival time, and, possibly, species. We found that in marmosets lesioned within the first 6 postnatal weeks, there is a positive correlation between LGN volume loss and ganglion cell loss (Fig. 11). At first glance, this correlation is unsurprising and consistent with the idea that withdrawal of LGN-derived trophic factors causes ganglion cell degeneration. But, on the other hand, all of the V1 lesions removed the entire foveal representation. Therefore, the variation in LGN volume is likely attributable to varying amounts of damage to the peripheral visual field representation in V1. Such variation would not be expected to change the degree of cell loss in central retina. A larger sample involving systematic changes in cortical lesion sizes might shed light on this unresolved puzzle. In any event, our data point to multiple mechanisms at play in retinal transneuronal degeneration. These mechanisms likely include geniculocortical connectivity patterns and the intrinsic developmental profile of structures involved at different stages of the afferent visual pathway. 

Acknowledgments

The authors thank Arzu Demir, Sammy Lee, Subha Nasir-Ahmad, Richa Verma, and Katrina Worthy for assistance with experiments. 

Supported by Australian Research Council Centre of Excellence for Integrative Brain Function Grant CE140100007 and Australian National Health and Medical Research Council Grants 1194206 and 2019011. 

Disclosure: T. Sepehrisadr, None; N. Atapour, None; A.K. Baldicano, None; M.G.P. Rosa, None; U. Grünert, None; P.R. Martin, None