Limits...
Automatic quantitative analysis of experimental primary and secondary retinal neurodegeneration: implications for optic neuropathies

View Article: PubMed Central - PubMed

ABSTRACT

Secondary neurodegeneration is thought to play an important role in the pathology of neurodegenerative disease, which potential therapies may target. However, the quantitative assessment of the degree of secondary neurodegeneration is difficult. The present study describes a novel algorithm from which estimates of primary and secondary degeneration are computed using well-established rodent models of partial optic nerve transection (pONT) and ocular hypertension (OHT). Brn3-labelled retinal ganglion cells (RGCs) were identified in whole-retinal mounts from which RGC density, nearest neighbour distances and regularity indices were determined. The spatial distribution and rate of RGC loss were assessed and the percentage of primary and secondary degeneration in each non-overlapping segment was calculated. Mean RGC number (82 592±681) and RGC density (1695±23.3 RGC/mm2) in naïve eyes were comparable with previous studies, with an average decline in RGC density of 71±17 and 23±5% over the time course of pONT and OHT models, respectively. Spatial analysis revealed greatest RGC loss in the superior and central retina in pONT, but significant RGC loss in the inferior retina from 3 days post model induction. In comparison, there was no significant difference between superior and inferior retina after OHT induction, and RGC loss occurred mainly along the superior/inferior axis (~30%) versus the nasal–temporal axis (~15%). Intriguingly, a significant loss of RGCs was also observed in contralateral eyes in experimental OHT. In conclusion, a novel algorithm to automatically segment Brn3a-labelled retinal whole-mounts into non-overlapping segments is described, which enables automated spatial and temporal segmentation of RGCs, revealing heterogeneity in the spatial distribution of primary and secondary degenerative processes. This method provides an attractive means to rapidly determine the efficacy of neuroprotective therapies with implications for any neurodegenerative disorder affecting the retina.

No MeSH data available.


Diagram of RGC segmentation and counting used in this study. (a) Identification of RGC number, area and centroid from Brn3a-labelled retinal whole-mounts was completed using previously described methods.10 (b) Retinal area was determined by first creating a binary mask of the whole-mount area before applying a customised ImageJ script to measure the retinal area (black pixels) in each retinal segment (ring and quadrants). (c) On defining the centre of the ONH, a series of 15 concentric rings with 0.3 mm radii were drawn and the position of each RGC determined by calculating the Euclidean distance [D] of each RGC relative to the ONH using ring boundaries as thresholds. RGCs were further segmented into quadrants (d and e) (where m=b/a) defined using Cartesian (d) or polar coordinates (e) which gave identical results. For each segment, in each case RGC density, average nearest neighbour distance and regularity index were calculated as described in the text.
© Copyright Policy - open-access
Related In: Results  -  Collection

License
getmorefigures.php?uid=PMC4979431&req=5

fig1: Diagram of RGC segmentation and counting used in this study. (a) Identification of RGC number, area and centroid from Brn3a-labelled retinal whole-mounts was completed using previously described methods.10 (b) Retinal area was determined by first creating a binary mask of the whole-mount area before applying a customised ImageJ script to measure the retinal area (black pixels) in each retinal segment (ring and quadrants). (c) On defining the centre of the ONH, a series of 15 concentric rings with 0.3 mm radii were drawn and the position of each RGC determined by calculating the Euclidean distance [D] of each RGC relative to the ONH using ring boundaries as thresholds. RGCs were further segmented into quadrants (d and e) (where m=b/a) defined using Cartesian (d) or polar coordinates (e) which gave identical results. For each segment, in each case RGC density, average nearest neighbour distance and regularity index were calculated as described in the text.

Mentions: Brn3a is a POU-domain transcription factor considered a good histological RGC marker, due to its high RGC specificity and good agreement with retrograde axonal tracer Fluoro-Gold (97% RGC co-labelling in rats).29 As Brn3a is localised to RGC nuclei, immunohistochemistry of retinal whole-mounts yields images from which RGCs can be readily segmented (Figure 1a). Quantification of Brn3a RGCs is frequently obtained by manually counting retinal segments;7,30–33 yet a limitation of this approach is the high variability in RGC density between central and peripheral rodent retina, which can complicate RGC quantification by sampling.34 To address these concerns, whole retinas need to be assessed and several groups, including our own, have developed tools to automatically count Brn3a-labelled RGCs, results of which are most frequently presented quantitatively as average whole-mount RGC density (RGC/mm2) or qualitatively as isodensity maps.10,29,35,36


Automatic quantitative analysis of experimental primary and secondary retinal neurodegeneration: implications for optic neuropathies
Diagram of RGC segmentation and counting used in this study. (a) Identification of RGC number, area and centroid from Brn3a-labelled retinal whole-mounts was completed using previously described methods.10 (b) Retinal area was determined by first creating a binary mask of the whole-mount area before applying a customised ImageJ script to measure the retinal area (black pixels) in each retinal segment (ring and quadrants). (c) On defining the centre of the ONH, a series of 15 concentric rings with 0.3 mm radii were drawn and the position of each RGC determined by calculating the Euclidean distance [D] of each RGC relative to the ONH using ring boundaries as thresholds. RGCs were further segmented into quadrants (d and e) (where m=b/a) defined using Cartesian (d) or polar coordinates (e) which gave identical results. For each segment, in each case RGC density, average nearest neighbour distance and regularity index were calculated as described in the text.
© Copyright Policy - open-access
Related In: Results  -  Collection

License
Show All Figures
getmorefigures.php?uid=PMC4979431&req=5

fig1: Diagram of RGC segmentation and counting used in this study. (a) Identification of RGC number, area and centroid from Brn3a-labelled retinal whole-mounts was completed using previously described methods.10 (b) Retinal area was determined by first creating a binary mask of the whole-mount area before applying a customised ImageJ script to measure the retinal area (black pixels) in each retinal segment (ring and quadrants). (c) On defining the centre of the ONH, a series of 15 concentric rings with 0.3 mm radii were drawn and the position of each RGC determined by calculating the Euclidean distance [D] of each RGC relative to the ONH using ring boundaries as thresholds. RGCs were further segmented into quadrants (d and e) (where m=b/a) defined using Cartesian (d) or polar coordinates (e) which gave identical results. For each segment, in each case RGC density, average nearest neighbour distance and regularity index were calculated as described in the text.
Mentions: Brn3a is a POU-domain transcription factor considered a good histological RGC marker, due to its high RGC specificity and good agreement with retrograde axonal tracer Fluoro-Gold (97% RGC co-labelling in rats).29 As Brn3a is localised to RGC nuclei, immunohistochemistry of retinal whole-mounts yields images from which RGCs can be readily segmented (Figure 1a). Quantification of Brn3a RGCs is frequently obtained by manually counting retinal segments;7,30–33 yet a limitation of this approach is the high variability in RGC density between central and peripheral rodent retina, which can complicate RGC quantification by sampling.34 To address these concerns, whole retinas need to be assessed and several groups, including our own, have developed tools to automatically count Brn3a-labelled RGCs, results of which are most frequently presented quantitatively as average whole-mount RGC density (RGC/mm2) or qualitatively as isodensity maps.10,29,35,36

View Article: PubMed Central - PubMed

ABSTRACT

Secondary neurodegeneration is thought to play an important role in the pathology of neurodegenerative disease, which potential therapies may target. However, the quantitative assessment of the degree of secondary neurodegeneration is difficult. The present study describes a novel algorithm from which estimates of primary and secondary degeneration are computed using well-established rodent models of partial optic nerve transection (pONT) and ocular hypertension (OHT). Brn3-labelled retinal ganglion cells (RGCs) were identified in whole-retinal mounts from which RGC density, nearest neighbour distances and regularity indices were determined. The spatial distribution and rate of RGC loss were assessed and the percentage of primary and secondary degeneration in each non-overlapping segment was calculated. Mean RGC number (82 592±681) and RGC density (1695±23.3 RGC/mm2) in naïve eyes were comparable with previous studies, with an average decline in RGC density of 71±17 and 23±5% over the time course of pONT and OHT models, respectively. Spatial analysis revealed greatest RGC loss in the superior and central retina in pONT, but significant RGC loss in the inferior retina from 3 days post model induction. In comparison, there was no significant difference between superior and inferior retina after OHT induction, and RGC loss occurred mainly along the superior/inferior axis (~30%) versus the nasal–temporal axis (~15%). Intriguingly, a significant loss of RGCs was also observed in contralateral eyes in experimental OHT. In conclusion, a novel algorithm to automatically segment Brn3a-labelled retinal whole-mounts into non-overlapping segments is described, which enables automated spatial and temporal segmentation of RGCs, revealing heterogeneity in the spatial distribution of primary and secondary degenerative processes. This method provides an attractive means to rapidly determine the efficacy of neuroprotective therapies with implications for any neurodegenerative disorder affecting the retina.

No MeSH data available.