Retinal ganglion cells in the eastern newt Notophthalmus viridescens: Topography, morphology, and diversity

Abstract

The topography and morphology of retinal ganglion cells (RGCs) in the eastern newt were studied. Cells were retrogradely labeled with tetramethylrhodamine-conjugated dextran amines or horseradish peroxidase and examined in retinal wholemounts. Their total number was 18,025 +/- 3,602 (mean +/- SEM). The spatial density of RGCs varied from 2,100 cells/mm(2) in the retinal periphery to 4,500 cells/mm(2) in the dorsotemporal retina. No prominent retinal specializations were found. The spatial resolution estimated from the spatial density of RGCs varied from 1.4 cycles per degree in the periphery to 1.95 cycles per degree in the region of the peak RGC density. A sample of 68 cells was camera lucida drawn and subjected to quantitative analysis. A total of 21 parameters related to RGC morphology and stratification in the retina were estimated. Partitionings obtained by using different clustering algorithms combined with automatic variable weighting and dimensionality reduction techniques were compared, and an effective solution was found by using silhouette analysis. A total of seven clusters were identified and associated with potential cell types. Kruskal-Wallis ANOVA-on-Ranks with post hoc Mann-Whitney U tests showed significant pairwise between-cluster differences in one or more of the clustering variables. The average silhouette values of the clusters were reasonably high, ranging from 0.52 to 0.79. Cells assigned to the same cluster displayed similar morphology and stratification in the retina. The advantages and limitations of the methodology adopted are discussed. The present classification is compared with known morphological and physiological RGC classifications in other salamanders.

Structure and diversity of retinal ganglion cells in steller's sculpin Myoxocephalus stelleri tilesius, 1811

“…All three types are arranged as regular mosaics spatially independent from each other, providing strong evidence for their being natural GC types (Cook et al, ; Cook, ). Cell types similar to α a , α c , and biplexiform cells are also found in nonmammalian species (e.g., Cook and Noden, ; Pushchin and Karetin, ). Their presence in the Steller's sculpin supports the hypothesis of the symplesiomorphy and potential homology of large GCs in nonmammals (Cook et al, ).…”

“…Our previous research showed that parameters related to soma size proved to be least discriminating ones, and their inclusion into analysis often had masking effects, deteriorating the clustering quality (Pushchin and Karetin, ). For these reasons, these parameters were not used in the present study.…”

Classification of retinal ganglion cells in the southern hemisphere lamprey Geotria australis (Cyclostomata)

“…Analysis of morphological characteristics (Rodieck and Brening, 1983;Cook, 1998) can be used to provide an indication of the retinal ganglion cell types present. Recent methods using an "unsupervised" quantitative approach, such as cluster analysis, have become increasingly popular for objectively segregating cells into types (Costa and Velte, 1999;Badea and Nathans, 2004;Kong et al, 2005;Coombs et al, 2006Coombs et al, , 2007Pushchin and Karetin, 2009). These methods, combined with improved technology for digital reconstruction, provide greater objectivity and validity to morphological classification of neurons than previous methods.…”

“…Multimodality is important for variables used in cluster analysis. If a variable is unimodally distributed, then it is less likely to be helpful for cluster analysis (Schweitzer and Renehan, 1997;Pushchin and Karetin, 2009). The parameter with the largest MI from each category of parameters was chosen.…”

Retinal ganglion cells in the Pacific redfin, Tribolodon brandtii dybowski, 1872: Morphology and diversity

“…Parameters with the lowest MIs and least discriminating power, as revealed by discriminant function analysis, were successively excluded from the clustering models as previously described (Pushchin and Karetin, 2009). We also used the automatic variable weighting (OVW) algorithm to improve the clustering quality (De Soete, 1986Soete, , 1998Makarenkov and Legendre, 2001).…”