In order to determine the extent of glaucomatous damage in the posterior segment of donor eyes, we systematically evaluated histological features from 21 glaucomatous eyes (65–97 years old) and 31 age-matched non-glaucomatous control eyes. The histologic parameters evaluated were the thickness of the GCC, cupping of the optic disc, number of remaining RGC soma, immunoreactivity of NEFL in retinal and optic nerve sections, and glial activation (immunoreactivity for GFAP and IBA1). At this stage, since limited clinical and medical histories of the donors were available to us, the donors remained unstratified.

Gross anatomical changes in retina, macula, and optic nerve head in glaucoma

Macular damage in samples with a positive medical history of glaucoma was observed by a discernible reduction in RGC cell numbers and density within the temporal macular region from H&E in 11 of 21 donors (Fig. 1A). In some donors, the RGC layer was completely lost in both temporal and nasal macular regions (Supplemental Fig. 2A) accompanied by local RGC loss in the central macula for one donor (Supplemental Fig. 2B). Cupping of the optic disc was observed in 11 of 21 glaucomatous donor eyes (Fig. 1B). Remodeling of the lamina cribrosa (compression and disorganization of the prelaminar, lamina and post-lamina regions) was observed in 13 of 21 donor eyes. The extent of cupping and lamina cribrosa remodeling varied between donors and did not always correlate with RGC loss [25,26,27,28]. Given that sample fixation and processing for histology could confound accurate assessment of cupping, we did not consider cupping for additional sample classification or stratification. We did not observe consistent changes in other layers of the macula, including the retinal pigment epithelium, suggesting that the observed changes relate to glaucoma, and not other conditions such as age-related macular degeneration.

Fig. 2

RGC loss in glaucomatous donors observed by loss of RBPMS positive cells. A Loss of RGCs in peripheral retina and macula was observed in glaucomatous donors; arrows indicate RBPMS positive cells. B Scatter dot plot represents the individual values of RBPMS-positive RGCs (RBPMS + RGCs) for non-glaucomatous (n = 16) and glaucomatous donor tissues (n = 21). Quantitative analysis demonstrated a 54% reduction in RBPMS + RGCs in glaucomatous temporal peripheral retina. Horizontal bars indicate mean ± standard deviation. ****p < 0.0001, unpaired t-test

Next, we compared GCC thickness of non-glaucomatous and glaucomatous retinas from donors (Fig. 1C-E). Qualitative thinning of the peripheral retina (RGC loss, RFNL thinning) was observed by H&E staining (Fig. 1A) in 12 of 21 glaucomatous retinas. We next directly evaluated RGC soma loss in the peripheral retina comparing temporal to macular regions, and the nasal to central area (between macula and ONH). RGC loss was noted in both regions around the macula (the temporal and nasal) from 8 of 12 donors with retinal thinning, but only in the temporal retina of the remaining 4 donors with thinning. To confirm reduction of RNFL thickness in glaucoma, GCC thickness was measured in the temporal peripheral retina of all donor retinas (temporal to macula, 7 mm from BMO, Supplemental Fig. 1). On average, the GCC thickness of non-glaucomatous eyes was 56.3 ± 1.9 µm (mean ± SEM) while it was significantly lower by 15% (47.8. ± 2.30 µm) in glaucomatous eyes (Fig. 1C). Across donors, we observed no significant association of GCC thickness with age (R2 = 0.03301 for non-glaucoma, R2 = 0.1421 for glaucoma, Fig. 1D). When samples were stratified into 3 age groups (< 80y, 80-90y, > 90y) we did not observe any statistically significant differences in GCC thickness between glaucoma and non-glaucoma donors, when compared within or across groups (Fig. 1E). Together, the data demonstrate that while there is a statistically significant difference in temporal GCC thickness between groups in this cohort, we did not observe clear age-associated trends, suggesting that disease is the primary driver of the observed differential.

Loss of retinal ganglion cells in glaucoma

To confirm RGC loss observed by H&E staining, we immunolabelled a randomized subset of non-glaucomatous and all glaucoma donor tissues with the RGC specific marker, RBPMS. Similar to our prior assessments, we observed 5–7 layers of RBPMS + RGCs in non-glaucomatous donors. However, in glaucomatous donors, most sections exhibited 2–5 RBPMS + cell layers. In some cases, there were only 0–2 layers suggesting advanced severity (Fig. 2A). Further quantitative analysis was performed in the temporal peripheral retina (5 mm from the end of the macular outline, Supplemental Fig. 1). We observed a 54% decrease in RGC density in the temporal peripheral retina from glaucomatous donors compared with non-glaucomatous donors (Fig. 2B).

Axonal and dendritic loss in glaucoma

Loss of axonal integrity and dendrites in the neural retina was determined by measuring the immunoreactivity of neurofilament light chain protein (NEFL) in both the peripheral temporal retina and ONH. Qualitative evaluation revealed a reduction in NEFL + processes in the peripheral temporal neural retina accompanied by irregular distribution of NEFL in the glaucomatous ONH (Fig. 3A). Expression levels of NEFL were quantified within a 2 mm peripheral temporal region and 1 mm central region between the BMO and macula by normalizing the percentage positive immunolabel to total area measured. Normalized NEFL immunoreactivity in non-glaucomatous donor retina was 62.6 ± 11.5%, while it was significantly reduced to 37.1 ± 20% in glaucoma (Fig. 3B) demonstrating a net reduction of NEFL + axons by 40%. The reduction in NEFL immunoreactivity in glaucomatous ONH was similar (42.4%) (Fig. 3B).

Fig. 3

Axonal and dendritic loss of RGCs visualized by NEFL immunolabelling in the peripheral retina and ONH. A Decreased NEFL expression in temporal peripheral retina and ONH, and disorganization of NEFL immune staining was observed in glaucomatous tissues. Asterisks indicate representative areas with loss of NEFL. B Scatter dot plot represents the individual values of neurofilament light chain (NEFL) for non-glaucomatous (retina, n = 16, ONH, n = 16) and glaucomatous donor tissues (retina, n = 21; ONH, n = 18). Quantitation of NEFL staining showed a statistically significant reduction in the temporal peripheral retina (40%) and optic nerve head (42.4%) of glaucoma donors. Horizontal bars indicate mean ± standard deviation. ****p < 0.0001, unpaired t-test

Glial activation in glaucoma

Qualitative assessment revealed increased immunoreactivity of GFAP in the glaucomatous peripheral retina accompanied by disorganized GFAP + regions within the ONH, compared with non-glaucomatous donors. Specifically, in non-glaucomatous tissues, when GFAP + signals were observed, these were localized to inner layers of the retina (especially in RNFL and RGCL) and around blood vessels. However, in glaucomatous tissues, the intensity of GFAP labelling was significantly higher and extended from the inner retina to the inner nuclear layer (INL), outer plexiform layer (OPL), and in some cases, to the outer nuclear layer (ONL) (Fig. 4A). Also, increased GFAP intensity was observed in glaucomatous ONH throughout the peripapillary, prelaminar and lamina cribrosa regions. Disorganized GFAP staining patterns were also observed in post-laminar regions of glaucomatous ONH (Fig. 4A). Using a similar quantitative approach to NEFL measurement, we report a 64% increase in normalized GFAP-immunoreactivity in glaucomatous temporal peripheral retina compared with non-glaucomatous donor retinas (Fig. 4B). Similar increases in GFAP immunoreactivity were seen in the ONH (80.6%).

Fig. 4

Detection of activated astrocytes by GFAP immunoreactivity. A Compared with non-glaucomatous donors, upregulation of GFAP was detected in glaucomatous peripheral retina and ONH. Arrows indicate representative GFAP immunoreactivity. Asterisks on the bottom panel highlight the region of the ONH quantified in Fig. 4B. B Scatter dot plot represents the individual values of reactive astrocytes (GFAP + cells) for non-glaucomatous (retina, n = 16; ONH, n = 16) and glaucomatous donor tissues (retina, n = 21; ONH, n = 19) assessed by histology. Quantitative analysis showed that GFAP was significantly upregulated in the temporal peripheral retina (64%) and optic nerve head (80.6%) of glaucoma donors. Horizontal bars indicate mean ± standard deviation. **p < 0.01, unpaired t-test

Next, we evaluated the number of microglia present in the tissue by immunolabelling for IBA1 + cells. In glaucomatous retinas, IBA1 + microglia were mostly present in the RNFL, ganglion layer, IPL and INL (Fig. 5A). Dual staining of IBA1 and RBPMS detected IBA1 + cells to be closely associated with RBPMS + RGCs (Supplemental Fig. 3). A significantly greater cell density (% positive cells per total area measured) of IBA1 + microglia was observed in glaucomatous retinas (7.8 ± 3.7%) compared with non-glaucomatous tissues (4.1 ± 1.3%; Fig. 5B) reflecting a net increase of 90.9%. In the optic nerve head (within the first 1 mm below BMO, a significant increase in IBA1 positive cell density was observed in glaucomatous tissues compared with non-glaucomatous tissues (58.4%; Fig. 5B).

Fig. 5

Detection of microglia by IBA1 immunoreactivity. A IBA1 was upregulated in glaucomatous peripheral retina and ONH compared with non-glaucomatous donors. Arrows indicate IBA1 positive cells in the retina and asterisks increased IBA1 immunoreactivity in the ONH. B Scatter dot plot represents the individual values of IBA1 + cells for non-glaucomatous (retina, n = 12; ONH, n = 16) and glaucomatous donor tissues (retina, n = 18, ONH, n = 14) assessed by histology. Quantitative analysis of IBA1 positivity showed that IBA1 was statistically upregulated in the temporal peripheral retina (90.9%) and optic nerve head of (58.4%) glaucoma samples. Horizontal bars indicate mean ± standard deviation. **p < 0.01, unpaired t-test

Stratification of disease severity by multivariate analysis

After assessing individual markers, we sought to determine if there was any correlation between them. First, we asked if there was any correlation between protein marker staining between the retina and optic nerve head tissues. This analysis was performed only on sections from donors where both the retina and optic nerve were present. Spearman’s analysis demonstrated a statistically significant correlation for GFAP and NEFL immunolabelling between the retina and optic nerve head, but no statistical significance was observed for IBA1 labelling between the two tissues (Fig. 6). For all subsequent analyses, measurements from only the retinal tissue were used.

Fig. 6

Correlation analysis between retina and optic nerve head. Spearman’s correlation and linear regression was performed for NEFL, GFAP and IBA1 expression comparing retina and ONH measurements for glaucoma and non-glaucoma donors. NEFL and GFAP exhibit strong correlation between retina and ONH measures (R2 = 0.6165 and R2 = 0.5328, respectively), but this is not the case for IBA1 (R2 = 0.3077)

Since loss of RGCs is considered a hallmark of disease, we performed all our Spearman correlation analyses against quantitative RBPMS + immunoreactivity in the temporal retina. We observed a positive correlation between loss of NEFL + immunoreactivity and loss of RBPMS + RGC cell density (R2 = 0.8153; Spearman correlation, Fig. 7A) in glaucomatous tissues. This was a significant increase when comparing the same parameters in non-glaucomatous donors (R2 = 0.3648; Spearman correlation). No significant correlation was observed comparing RBPMS and GFAP or IBA1 levels in either glaucoma or non-glaucomatous donors. In the subset of donors where IBA1 positive immunoreactivity was observed, no relationship between GFAP and IBA1 immunostaining was observed in either glaucomatous or non-glaucomatous donors (Fig. 7B); although the variability in staining for both parameters was significantly greater for glaucomatous donors than non-glaucomatous donors.

Fig. 7

Spearman’s correlation analysis. A Spearman’s correlation demonstrated maximal correlation between RBPMS and NEFL in glaucomatous donors, with minimal correlation seen when comparing other markers in glaucomatous and non-glaucomatous donor tissues. B No statistically significant correlation was observed when comparing GFAP and IBA1 immunoreactivity in the retina for both non-glaucomatous and glaucomatous donors

Next, to determine the extent that donor variability contributed to the significance in quantitative measurements, and to compare sensitivity vs. specificity of the outcomes comparing glaucomatous and non-glaucomatous donors, we performed receiver operating characteristic (ROC) curve analysis (Supplemental Fig. 4). ROC analysis revealed that the area under the curve was the greatest for RBPMS (AUC = 0.9583) and those for NEFL (AUC = 0.8333), GFAP (AUC = 0.8274), and IBA1 (AUC = 0.8472) were significant, reinforcing the results that the quantitative changes observed in glaucoma were true positives.

Since variability was observed both from the scatter and ROC plots, we hypothesized that these changes in temporal retinal NEFL (axonal & dendritic loss), RBPMS (RGC loss), and GFAP (glial activation) levels may be reflective of disease stage. PCA and factor analysis of the data revealed that the first two principal components with 3 variables (NEFL, RBPMS, and GFAP) contributed to the maximal variance (> 85%) in the data (Fig. 8A, B). Using the quantitative measurements for these 3 variables, unbiased k-means clustering revealed 3 clusters for glaucoma samples and 2 clusters for non-glaucoma donors (Fig. 8C).

Fig. 8

Multivariate statistical analysis. A Principal component analysis for GFAP, RBPMS, NEFL and IBA1 immunoreactivity demonstrated that > 85% of the variance in the data was contributed by the first 2 principal components, as indicated in the Scree plot. B Score plot from factor analysis showing that maximal variance was weighted by 3 factors: RBPMS, NEFL and IBA1, with the former 2 representing neuronal components, and the latter representing a glial component. C Cluster analysis using the principal components. Severity of glaucoma was assessed by applying the Calinski-Harabasz criterion for unbiased k-means clustering using the expression values for RBPMS, NEFL, and GFAP positive staining. Non-glaucomatous donors clustered in 2 distinct groups while glaucomatous tissues clustered in 3 distinct groups (mild, moderate, and severe). Arrows as indicated in the figure identify tissues with qualitative cupping of the ONH observed through histology

The two non-glaucoma clusters were predominantly factored on possessing a high level of RBPMS + and NEFL + staining, and a low proportion of GFAP + glial cells. For glaucomatous tissues, the clustering was parameter dependent. Donor sections with the highest level of RBPMS + NEFL + staining and least glial proliferation/GFAP positivity were categorized as mild glaucoma. Donor samples with the greatest increase in GFAP + cells but moderate loss of NEFL + RBPMS + staining were classified as moderate glaucoma. Finally, donors with greatest loss of RBPMS + NEFL + staining but a persistence of some GFAP + cells were classified as severe glaucoma. These data suggest a phasic stratification with disease where maximal glial activation is seen in the moderate stage.

Changes in immunohistochemical markers stratified by glaucoma severity

After implementation of clustering and definition of the three distinct glaucoma groups, we reassessed our data based on this refined system. We observed a progressive loss in RBPMS + RGCs and NEFL staining in the temporal retinas with increasing glaucoma severity (Fig. 9A, B). GFAP + cell density was severity-dependent, demonstrating maximal glial activation with moderate severity (Fig. 9C). On the other hand, IBA1 + immunoreactivity increased in the moderate and severe glaucoma groups with large variability but did not drastically decrease in severe glaucoma (Fig. 9D).

Fig. 9

Post-hoc analysis after stratification of glaucomatous donors. Progressive RGC degeneration and increased glial reactivity were observed in the retina with disease severity. A Loss of RBPMS + RGC cells, B loss of axons (NEFL), C increased astrocytic activation (GFAP), and D microglia/macrophage activation (IBA1) was stratified by disease state. Horizontal bars indicate mean ± standard deviation. *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001, One-way Analysis of variance (ANOVA) followed by Tukey’s multiple comparison test