A Fair Assessment of Evaluation Tools for the Murine Microbead Occlusion Model of Glaucoma

Despite being one of the most studied eye diseases, clinical translation of glaucoma research is hampered, at least in part, by the lack of validated preclinical models and readouts. The most popular experimental glaucoma model is the murine microbead occlusion model, yet the observed mild phenotype, mixed success rate, and weak reproducibility urge for an expansion of available readout tools. For this purpose, we evaluated various measures that reflect early onset glaucomatous changes in the murine microbead occlusion model. Anterior chamber depth measurements and scotopic threshold response recordings were identified as an outstanding set of tools to assess the model’s success rate and to chart glaucomatous damage (or neuroprotection in future studies), respectively. Both are easy-to-measure, in vivo tools with a fast acquisition time and high translatability to the clinic and can be used, whenever judged beneficial, in combination with the more conventional measures in present-day glaucoma research (i.e., intraocular pressure measurements and post-mortem histological analyses). Furthermore, we highlighted the use of dendritic arbor analysis as an alternative histological readout for retinal ganglion cell density counts.


Introduction
As a leading cause of irreversible blindness, glaucoma continues to place a heavy burden on millions of people worldwide. Apart from the inability to restore vision loss, current therapeutic options are incapable of halting disease progression in all glaucoma patients. Despite a myriad of research efforts, novel treatment routes do not reach or pass clinical trial. This inefficient clinical translation can be partially explained by the lack of validated preclinical models and readouts [1].
A multitude of animal species has been used to study the physiopathology of glaucoma [2], yet rodents have become the preferred model organism due to the broad array of available genetic and experimentally induced models and readout approaches, all as recently reviewed by Pang and Clark [1]. In general, genetic and experimentally induced glaucoma models are looked upon as chronic vs. acute, respectively. Chronic genetic models, e.g., the DBA/2J mouse strain, are popular due to a closer resemblance to human progressive pathology. Unfortunately, the asynchronous and variable disease onset, progression, and severity observed in these models often results in constraints regarding experimental set-up in the search towards novel glaucoma therapies, hence explaining the need for experimentally induced models that offer a higher degree of investigational control [1]. The only chronic experimental glaucoma model is the microbead occlusion model. An intracameral injection of microbeads is used to occlude the conventional aqueous humor outflow pathway, thereby provoking a mild and chronic glaucomatous pathology that phenocopies the key glaucomatous features (i.e., elevated intraocular pressure (IOP), retinal ganglion cell (RGC) degeneration, and optic nerve damage), along with a slow disease progression. Yet, the mild phenotype (reports show RGC loss as low as 7-12% at 4 to 12 weeks after glaucoma onset [3][4][5][6][7]) also represents one of its weaknesses, as it leaves a rather small window to chart glaucomatous damage or neuroprotection.
Other significant drawbacks of the microbead model include mixed success rates and weak reproducibility between different experiments and laboratories, despite abundant protocol optimizations (including the need of repeated injections; co-injection of viscoelastic agents; and a wide range of optimized microbead characteristics such as type, size, or concentration). Hence, glaucoma research would benefit from robust readouts to assess the successful induction of the model. In ocular hypertension-dependent glaucoma models such as the microbead occlusion model, IOP elevation is the gold standard to define the success rate. Nonetheless, measuring IOP in mice is challenging and influenced by various factors (e.g., lighting, diurnal fluctuations, anesthesia), often resulting in imprecise and inconsistent data. Given the modest IOP elevation in the microbead occlusion model, the rise in IOP could be obscured by the variable nature of the technique. For this purpose, we explored the use of anterior segment optical coherence tomography (OCT) imaging as an easier-to-measure parameter to evaluate the success rate of ocular hypertension induction in microbead-injected mice.
Confronted with the abovementioned drawbacks of the microbead occlusion model (i.e., mixed success rate, weak reproducibility, and limited window of opportunity for charting RGC neurodegeneration or survival), we aimed to explore whether focusing on readouts for early stage pathological changes could overcome these issues. This focus is especially relevant given the irrevocable nature of glaucomatous neurodegeneration where therapies should interfere as early as possible in the course of the disease. Indeed, the neurodegeneration field (including Alzheimer's and Parkinson's diseases) has shifted its research focus towards the presymptomatic disease stages, acknowledging that neuronal loss is preceded by a (often long) phase of neuronal dysfunction and that central nervous system function can only be preserved by tackling these early disease stages. In the glaucoma field, pinpointing the primary cause and the chronological order of subsequent cellular processes is a long-standing debate. Existing research, however, recognizes a differential vulnerability in the sub-compartments of RGCs (i.e., dendrites vs. soma vs. axon vs. synapse) [8]. Previous reports have disclosed that RGC neurites display one of the earliest signs of glaucomatous damage [9][10][11]. Moreover, as dendrites and axons are critical for RGC function, functional deficits presumably precede actual RGC loss [9]. Within the chronic microbead occlusion model, the glaucomatous stress imposed on the RGCs during the first weeks might not (yet) be severe enough to fatally damage them. On the basis of current insights into glaucoma pathogenesis, we evaluated various readouts that enable early stage disease characterization and subsequent therapy intervention by looking at rapid-onset structural (axon and dendrite morphology) and functional (axonal transport and electrophysiology) changes. Our ambition was to select in vivo approaches that allow for longitudinal disease monitoring, together with a high translatability to the clinic, in order to potentially narrow the gap between preclinical and clinical research in glaucoma research.

Gold Standard Evaluation Tools: Intraocular Pressure (IOP) and Retinal Ganglion Cell (RGC) Loss
The protocol of the microbead occlusion model was adapted from Yoko and Belforte et al. [12]. Briefly, 2 µL non-transparent, paramagnetic microbeads (4.5 µm diameter) were injected with a single intracameral injection and re-positioned towards the trabecular meshwork with a handhold magnet. Successful induction of ocular hypertension and glaucomatous damage were evaluated via two old-school parameters: IOP measurements and histological RGC soma counts with the pan-RGC marker RBPMS. The mean IOP was elevated at two weeks post-injection (15.24 ± 1.28 mmHg) as compared to the contralateral controls (11.25 ± 0.52 mmHg; p = 0.0097; Hedges' g = 1.24 with CI 95% of 0.47 to 2.04). This mild ocular hypertension led to a moderate RGC loss: the mean RBPMS density of microbead-injected eyes (2978 ± 55 RGCs/mm 2 ) was lower than in the contralateral eyes (3269 ± 44 RGC/mm 2 ; p = 0.0006, Hedges' g = −1.77 with CI 95% of −2.91 to −0.69) at five weeks post-injection (Figure 1b,c). Of note, no difference in RBPMS density was observed between vehicle-injected eyes (3131 ± 41 RGCs/mm 2 ) as compared to the contralateral control group (Hedges' g = −0.98 with CI 95% of −2.02 to −0.05), and the contralateral eye was used as reference value during the following experiments. Overall, these data indicated a mild glaucomatous phenotype (22 ± 6% IOP elevation and 9 ± 2% RGC loss) yet left us with a need of more convincing readouts to properly assess glaucomatous damage and create a larger window of opportunity to measure a positive outcome on RGC survival in neuroprotective studies.

Gold Standard Evaluation Tools: Intraocular Pressure (IOP) and Retinal Ganglion Cell (RGC) Loss
The protocol of the microbead occlusion model was adapted from Yoko and Belforte et al. [12]. Briefly, 2 µl non-transparent, paramagnetic microbeads (4.5 µm diameter) were injected with a single intracameral injection and re-positioned towards the trabecular meshwork with a handhold magnet. Successful induction of ocular hypertension and glaucomatous damage were evaluated via two old-school parameters: IOP measurements and histological RGC soma counts with the pan-RGC marker RBPMS. The mean IOP was elevated at two weeks post-injection (15.24 ± 1.28 mmHg) as compared to the contralateral controls (11.25 ± 0.52 mmHg; p = 0.0097; Hedges' g = 1.24 with CI95% of 0.47 to 2.04). This mild ocular hypertension led to a moderate RGC loss: the mean RBPMS density of microbead-injected eyes (2978 ± 55 RGCs/mm 2 ) was lower than in the contralateral eyes (3269 ± 44 RGC/mm²; p = 0.0006, Hedges' g = −1.77 with CI95% of −2.91 to −0.69) at five weeks post-injection (Figure 1b,c). Of note, no difference in RBPMS density was observed between vehicle-injected eyes (3131 ± 41 RGCs/mm 2 ) as compared to the contralateral control group (Hedges' g = −0.98 with CI95% of −2.02 to −0.05), and the contralateral eye was used as reference value during the following experiments. Overall, these data indicated a mild glaucomatous phenotype (22 ± 6% IOP elevation and 9 ± 2% RGC loss) yet left us with a need of more convincing readouts to properly assess glaucomatous damage and create a larger window of opportunity to measure a positive outcome on RGC survival in neuroprotective studies.  , measured with a rebound tonometer, was increased in microbead-injected eyes at two weeks post-injection, in comparison to the contralateral eyes. Of the 10 mice, 4 did not show elevated IOP (defined as an IOP lower than the mean IOP value of the contralateral eye + one standard deviation) (shown in red). Unpaired two-tailed t-test, ** p ≤ 0.01, n = 10. (b) RGC density of entire RBPMS-stained flatmounts was calculated by an automated deep learning tool (RGCode) [13] and revealed a mild RGC loss at five weeks post-injection, as compared to contralateral eyes. Unpaired two-tailed t-test, *** p ≤ 0.001, n = 10. (c) Representative photomicrographs of mid-peripheral regions of RBPMS-stained flatmounts revealed no evidently visible changes in RGC densities across contralateral control and microbead-injected eyes. Scale bar = 50 µm. CL = contralateral eyes and MB = microbead-injected eyes.

Anterior Chamber Depth as an Easy-to-Measure Parameter to Evaluate Model Induction
Intracameral microbead injection and the resulting ocular hypertension may lead to changes in the structure of the anterior segment that could be useful to evaluate the effectiveness of ocular hypertension induction and serve as a complementary or alternative measure for IOP measurements [14]. OCT imaging of the anterior segment revealed an enlarged anterior chamber depth and more concave, flattened irises after bead injection in comparison to the contralateral controls (Figure 2a). At endpoint measurements (five weeks post-injection), a substantial enlargement of the anterior chamber depth was observed in microbead-injected eyes (452 ± 13 µm) as compared to contralateral controls (400 ± 4 µm; p = 0.0012; Hedges' g = 1.65 with CI 95% of 0.71 to 2.62) (Figure 2b). Longitudinal follow-up of anterior chamber depth measurements disclosed baseline values up to and including two weeks post-injection, after which the anterior chamber depth gradually increased (Figure 2c,d). Importantly, a strong, positive correlation (R 2 = 0.84) was found between IOP (two weeks post-injection) and anterior chamber depth (five weeks post-injection) measurements. Hence, anterior chamber depth measurements could provide additional confirmation regarding the success rate of the microbead occlusion model.

Anterior Chamber Depth as an Easy-to-Measure Parameter to Evaluate Model Induction
Intracameral microbead injection and the resulting ocular hypertension may lead to changes in the structure of the anterior segment that could be useful to evaluate the effectiveness of ocular hypertension induction and serve as a complementary or alternative measure for IOP measurements [14]. OCT imaging of the anterior segment revealed an enlarged anterior chamber depth and more concave, flattened irises after bead injection in comparison to the contralateral controls (Figure 2a). At endpoint measurements (five weeks post-injection), a substantial enlargement of the anterior chamber depth was observed in microbead-injected eyes (452 ± 13 µm) as compared to contralateral controls (400 ± 4 µm; p = 0.0012; Hedges' g = 1.65 with CI95% of 0.71 to 2.62) (Figure 2b). Longitudinal follow-up of anterior chamber depth measurements disclosed baseline values up to and including two weeks post-injection, after which the anterior chamber depth gradually increased (Figure 2c,d). Importantly, a strong, positive correlation (R 2 = 0.84) was found between IOP (two weeks post-injection) and anterior chamber depth (five weeks post-injection) measurements. Hence, anterior chamber depth measurements could provide additional confirmation regarding the success rate of the microbead occlusion model.   depth (i.e., the distance between the corneal endothelium and the anterior surface of the lens) are indicated, together with a clear example of a convex iris in a naïve eye vs. a concave, flattened iris after microbead injection. (b) Quantification of the anterior chamber depth at five weeks postinjection revealed a substantial enlargement (13 ± 3%) in microbead-injected eyes as compared to the control eyes. Of the 10 mice, 3 did not show elevated anterior chamber depth (defined as anterior chamber depth smaller than the mean anterior chamber depth of the contralateral eye + one standard deviation) (shown in red). Unpaired two-tailed t-test, ** p ≤ 0.01, n = 10. (c,d) Representative graphs of longitudinal anterior chamber depth follow-up showing the percentage of anterior chamber depth enlargement of mice with a successful (c) and failed (d) microbead surgery, both in comparison to the contralateral eyes. CL = contralateral eyes and MB = microbead-injected eyes.

Axonal Transport Deficits yet Normal Signal Conduction after Ocular Hypertension
Multiple studies have shown that RGC axons are among the first RGC sub-compartments to be affected by glaucomatous injury. Hence, we assessed the effects of mild ocular hypertension on axonal integrity at five weeks after microbead injection. In our study, no changes in mean axonal density were observed after microbead injection (404,486 ± 45,063 axons/mm 2 ) compared to contralateral optic nerves (372,772 ± 32,556 axons/mm 2 ; Hedges' g = 0.33 with CI 95% of −1.13 to 1.86) (Figure 3a,b). Qualitative assessment revealed densely packed axons surrounded by intact myelin sheaths after ocular hypertension, identical to the contralateral controls ( Figure 3c). Since no overt axon degeneration was observed at five weeks post-injection, we opted to explore axonal functionality via two studies: (i) evaluating retrograde axonal transport deficits via tracing from the superior colliculus with hydroxystilbamidine metanesulfonate (OHSt) and (ii) assessing anterograde signal propagation via measuring the scotopic full-field visual evoked potential (VEP) at the level of the visual cortex, in response to brief light flashes. The tracing experiment revealed a considerably lower number of OHSt-labeled (OHSt+) RGCs in the microbead-injected eyes as compared to the vehicle-injected controls at six days after tracer introduction (2083 ± 84 vs. 2587 ± 69 OHSt+ cells/mm 2 ; p = 0.0002; Hedges' g = −1.99 with CI 95% of −2.82 to −1.17) (Figure 3d [15] and revealed an unaltered density in the distal optic nerve (3 mm after optic nerve head) upon microbead injection. Unpaired two-tailed t-tests, ns = non-significant, n = 5. (b) Representative images of semi-thin optic nerve cross-sections (contralateral vs. microbead-injected optic nerves) revealed no clear signs of axon degeneration. Scale bar = 5 µm. (c) OHSt-labeled (OHSt+) RGC density after retrograde tracing from the superior colliculus was calculated by an automated deep learning tool (RGCode) [13] on entire flatmounts and revealed that axonal transport was notably disrupted in microbead-injected eyes as compared to vehicle-injected controls. Unpaired two-tailed t-test, *** p ≤ 0.001, n = 10. (d) Representative images of mid-peripheral retinal regions after retrograde OHSt tracing revealed a diminished number of OHSt+ RGCs after microbead injection. Scale bar = 50 µm. (e) Average (± CI95%) full-field flash visually evoked potential (VEP) responses indicated a non-significant but slightly decreased amplitude upon microbead injection as compared to contralateral controls. (f,g) The amplitude and latency of the VEP response remained unaltered upon microbead injection. Unpaired two-tailed ttest, ns = non-significant, n = 8. CL = contralateral eyes, VH = vehicle-injected eyes, and MB = microbead-injected eyes.  [15] and revealed an unaltered density in the distal optic nerve (3 mm after optic nerve head) upon microbead injection. Unpaired two-tailed t-tests, ns = nonsignificant, n = 5. (b) Representative images of semi-thin optic nerve cross-sections (contralateral vs. microbead-injected optic nerves) revealed no clear signs of axon degeneration. Scale bar = 5 µm. (c) OHSt-labeled (OHSt+) RGC density after retrograde tracing from the superior colliculus was calculated by an automated deep learning tool (RGCode) [13] on entire flatmounts and revealed that axonal transport was notably disrupted in microbead-injected eyes as compared to vehicleinjected controls. Unpaired two-tailed t-test, *** p ≤ 0.001, n = 10. (d) Representative images of mid-peripheral retinal regions after retrograde OHSt tracing revealed a diminished number of OHSt+ RGCs after microbead injection. Scale bar = 50 µm. (e) Average (± CI 95% ) full-field flash visually evoked potential (VEP) responses indicated a non-significant but slightly decreased amplitude upon microbead injection as compared to contralateral controls. (f,g) The amplitude and latency of the VEP response remained unaltered upon microbead injection. Unpaired two-tailed t-test, ns = non-significant, n = 8. CL = contralateral eyes, VH = vehicle-injected eyes, and MB = microbeadinjected eyes.
We next interrogated the thickness of the inner plexiform layer (IPL) via OCT imaging to evaluate dendritic retraction via an in vivo tool (Figure 5a). Unfortunately, quantification of the mean IPL thickness revealed no thinning after microbead injection compared to the contralateral eyes: 48.82 ± 0.62 vs. 49.06 ± 0.71 µm, respectively (Hedges' g = −0.11 with CI95%% of −1.04 to 0.78) (Figure 5b). We also quantified the combined thickness of the nerve fiber and ganglion cell layer (NFL + GCL) to interrogate RGC axonal and (d-f) Dendritic arbor analysis revealed a markedly lower dendritic length (d) and tree surface (c), whilst a trend towards a lower number of branches was observed (e). Unpaired two-tailed t-tests, ns = non-significant, * p ≤ 0.05, n = 5. CL = contralateral eyes and MB = microbead-injected eyes.

Changes in Retinal Ganglion Cell (RGC) Electrophysiology as a Precursor for RGC Degeneration
To further explore rapid-onset glaucomatous changes, we extended our study with scotopic threshold response (STR) recordings, a noninvasive electrophysiology readout for selective interrogation of RGC functionality. Interestingly, five weeks following microbead injection, the STR amplitude was prominently reduced in microbead-injected eyes as compared to their contralateral eyes (103 ± 12 vs. 149 ± 7 µV, respectively; p = 0.0031; Hedges' g = −1.39 with CI95% of −2.29 to −0.44) (Figure 6a,c), whilst the latency time remained unaltered (142 ± 4 vs. 141 ± 3 ms, respectively; Hedges' g = 0.05 with CI95% of −0.91 to −0.94) (Figure 6b,c). To avoid having the non-transparent, paramagnetic microbeads as a confounding factor, we implemented two control studies: (i) the addition of a microbead-injected control in which an equivalent concentration of microbeads was injected, without relocating the magnetic microbeads towards the iridocorneal angle and hence without inducing ocular hypertension, and (ii) the evaluation of scotopic full-field flash ERGs after microbead injection. Microbead-injected control eyes did not show a decreased STR amplitude (149 ± 6 µV; Hedges' g = 0.02 with CI95% of −0.90 to −0.90) ( Figure  6a,c) or latency time (132 ± 2 ms; Hedges' g = −1.21 with CI95% of −2.05 to −0.25) (Figure  6b,c) compared to the microbead-injected eyes. Amplitudes and latency times of the aand b-waves of the full-field flash ERG, reflecting the functionality of other retinal cell types (primarily photoreceptors and Müller/ON-bipolar cells, respectively) that are typically not affected in glaucomatous models, were found unaltered at five weeks after ocular hypertension induction as compared to contralateral controls (Figures 6d,e and Figure S1, Table S1). Likewise, the amplitude and latency time of the oscillatory potentials (OPs), which primarily mirror amacrine cell functioning, remained unchanged at five weeks after microbead injection (Figures 6f and Figure S1, Table S1).

Changes in Retinal Ganglion Cell (RGC) Electrophysiology as a Precursor for RGC Degeneration
To further explore rapid-onset glaucomatous changes, we extended our study with scotopic threshold response (STR) recordings, a noninvasive electrophysiology readout for selective interrogation of RGC functionality. Interestingly, five weeks following microbead injection, the STR amplitude was prominently reduced in microbead-injected eyes as compared to their contralateral eyes (103 ± 12 vs. 149 ± 7 µV, respectively; p = 0.0031; Hedges' g = −1.39 with CI 95% of −2.29 to −0.44) (Figure 6a,c), whilst the latency time remained unaltered (142 ± 4 vs. 141 ± 3 ms, respectively; Hedges' g = 0.05 with CI 95% of −0.91 to −0.94) (Figure 6b,c). To avoid having the non-transparent, paramagnetic microbeads as a confounding factor, we implemented two control studies: (i) the addition of a microbead-injected control in which an equivalent concentration of microbeads was injected, without relocating the magnetic microbeads towards the iridocorneal angle and hence without inducing ocular hypertension, and (ii) the evaluation of scotopic full-field flash ERGs after microbead injection. Microbead-injected control eyes did not show a decreased STR amplitude (149 ± 6 µV; Hedges' g = 0.02 with CI 95% of −0.90 to −0.90) (Figure 6a,c) or latency time (132 ± 2 ms; Hedges' g = −1.21 with CI 95% of −2.05 to −0.25) (Figure 6b,c) compared to the microbead-injected eyes. Amplitudes and latency times of the a-and b-waves of the full-field flash ERG, reflecting the functionality of other retinal cell types (primarily photoreceptors and Müller/ON-bipolar cells, respectively) that are typically not affected in glaucomatous models, were found unaltered at five weeks after ocular hypertension induction as compared to contralateral controls (Figure 6d,e and Figure S1, Table S1). Likewise, the amplitude and latency time of the oscillatory potentials (OPs), which primarily mirror amacrine cell functioning, remained unchanged at five weeks after microbead injection (Figure 6f and Figure S1, Table S1).

Anterior Chamber Depth and Scotopic Threshold Response (STR) as Easy, Alternative Measures to Assess Model Induction and Disease Progression in vivo
Integrating the results described above, we conclude that the most outstanding alternatives to the conventional RGC soma counts are the post-mortem evaluation of dendritic arbor changes and in vivo STR recordings of RGC functionality. Second, anterior chamber depth measurements via in vivo OCT imaging represent an easier, less variable tool to assess successful induction of the model in comparison to IOP recordings. Analysis of three independent studies confirmed the reproducibility and low experiment-to-experiment variability of these readouts ( Figure S2a,b). The standard deviation and effect sizes of the anterior chamber depth (SD = 12, 13, and 14 µm; Hedges' g = 1.37, 1.55, and 1.96) and STR measures (SD = 11, 8, and 12 µV; Hedges' g = −0.72, −1.12, and −1.21) were comparable in the three experiments. One-way ANOVA tests revealed no differences between the anterior chamber depth or STR measurements of different experiments.
Next, the predictive value of the newly proposed and conventional readouts was assessed. Logistic regression with a single input variable revealed a positive association between disease status and IOP or anterior chamber depth, whereas RBPMS+ RGC counts and STR were found to be negatively related with disease status. In other words, the higher the IOP and anterior chamber depth measurements, the more likely it is that model induction was successful, while lower RBPMS+ RGC counts and STR values indicate a higher likelihood of glaucomatous neurodegeneration (Table 1). Moreover, IOP, anterior chamber depth, and STR recordings were all found to be significant predictors of disease

Anterior Chamber Depth and Scotopic Threshold Response (STR) as Easy, Alternative Measures to Assess Model Induction and Disease Progression In Vivo
Integrating the results described above, we conclude that the most outstanding alternatives to the conventional RGC soma counts are the post-mortem evaluation of dendritic arbor changes and in vivo STR recordings of RGC functionality. Second, anterior chamber depth measurements via in vivo OCT imaging represent an easier, less variable tool to assess successful induction of the model in comparison to IOP recordings. Analysis of three independent studies confirmed the reproducibility and low experiment-to-experiment variability of these readouts ( Figure S2a,b). The standard deviation and effect sizes of the anterior chamber depth (SD = 12, 13, and 14 µm; Hedges' g = 1.37, 1.55, and 1.96) and STR measures (SD = 11, 8, and 12 µV; Hedges' g = −0.72, −1.12, and −1.21) were comparable in the three experiments. One-way ANOVA tests revealed no differences between the anterior chamber depth or STR measurements of different experiments.
Next, the predictive value of the newly proposed and conventional readouts was assessed. Logistic regression with a single input variable revealed a positive association between disease status and IOP or anterior chamber depth, whereas RBPMS+ RGC counts and STR were found to be negatively related with disease status. In other words, the higher the IOP and anterior chamber depth measurements, the more likely it is that model induction was successful, while lower RBPMS+ RGC counts and STR values indicate a higher likelihood of glaucomatous neurodegeneration (Table 1). Moreover, IOP, anterior chamber depth, and STR recordings were all found to be significant predictors of disease state (i.e., normal or glaucomatous eye), yet RBBPMS+ RGC counts were not (Table 1). Multilogistic regression analysis including both STR recordings and RBPMS+ RGC counts suggested no added value of RBPMS+ RGC counts alongside STR measurements (Table 2). Hence, these analyses reinforced the use of anterior chamber depth as an equivalent predictor of model induction to IOP measurements and evidenced that STR measures are preferred over RBPMS+ RGC counts to evaluate glaucomatous damage.

Discussion
Historically, glaucoma has been classified as an optic neuropathy, with the optic nerve as the initial site of damage. In our study, axonal density counts were unaltered and optic nerve histology looked normal. These observations are in line with several reports in literature, establishing that axonal loss is only visible in advanced disease stages [20]. In fact, various studies have shown that axonal and RGC malfunctioning are typically detected prior to the occurrence of axonal loss [20]. Indeed, retrograde tracing from the superior colliculus revealed clear transport deficits in our study: 20% less RGC somas were filled up with tracer at eight weeks after microbead injection. The biggest drawback of this approach, however, is the need of invasive surgical procedures to introduce the tracer either at the optic nerve or in the superior colliculus. Alternative approaches to document abnormal conduction within post-retinal visual pathways are ex vivo compound action potential recordings in the optic nerve [21] or in vivo VEP recordings from the visual cortex [22,23]. In our study, full-field VEP measurements showed a trend towards attenuation of the VEP signal upon ocular hypertension, yet did not reach statistical significance, perhaps due to limitations in detection. As full-field VEPs are known to lack sensitivity and display highly variable outcomes, it is possible that minor differences in conductivity were not picked up in our study. Akopian et al. [24], however, found reduced VEP amplitudes in their microbead occlusion model, albeit looking at a different endpoint (eight weeks post-injection) and using a protocol with repeated injections that led to a higher RGC loss (≈36%) [24,25]. To probe for subtle, early changes, it is advisable to opt for pattern VEP recordings [22]. Yet, although more sensitive, pattern VEP recordings come with a high installation cost and are not standard lab equipment. In summary, studying axonal dysfunction or degeneration at five weeks after ocular hypertension via either histological analysis of axon density, retrograde tracing, or VEP recordings does not meet our predefined goal: finding an accessible and preferably noninvasive in vivo tool to evaluate early glaucomatous pathology in the microbead occlusion model. Nevertheless, measuring axonal dysfunction is a valuable option for post-mortem evaluations, and VEP recordings may be a suitable in vivo measure at later disease stages.
There is a consensus that glaucoma manifests as early synaptic alterations within the retina, as reviewed by Agostinone and Di Polo [17]. Dendritic retraction is believed to be one of the earliest responses of the RGCs to stress, and can be detected prior to axonal loss [17,[26][27][28]. Multiple studies show dendritic changes of αRGCs and other RGC types in acute experimental [29][30][31][32][33][34][35] as well in genetic [36] murine glaucoma models. Likewise, our observations in the chronic microbead model disclosed an aberrant dendritic retraction of αRGCs and a lower area under the Sholl curve at five weeks after microbead injection. These observations are in accordance with the study of El-Danaf and Huberman [37], who showed a shift towards lower intersections in the Sholl plot, as well as a considerably lower dendritic length and field area in murine transient OFF αRGCs as early as one week after microbead injection, as well as the work of Bhandari et al. [38], who revealed a slight reduction in Sholl crossings and a decrease in total dendritic length at five weeks after microbead injection in ON αRGCs of microbead-injected mice. Additionally, Della Santina et al. [26] showed a reduction in dendritic area, length, and number, together with a change in arborization pattern in a Sholl analysis at four weeks after microbead injection in OFF-transient αRGCs. Hence, in our study and in accordance with literature reports, changes in dendritic arbor architecture are early signs of glaucoma. The downside, however, is the substantial time investment that is required for such a study, most often resulting in small sample sizes that reduce the power of the study. Moreover, to avoid overlapping dendritic trees, one requires the use of a transgenic mouse line or a viral vector approach, again imposing limitations on study design. Of note, the sparsely labeled Thy1-YFP-H mice used in this study show a labeling bias for αRGCs [18]. The evaluation of glaucomatous stress on different RGC subtypes would thus demand the use of distinct transgenic mouse lines and/or the use of the few identified molecular markers for RGC subtypes [17]. An alternative, fast, and in vivo readout for dendritic retraction is the evaluation of IPL thinning on OCT images. Unfortunately, no IPL thinning was observed in our study, probably due to the lack of sensitivity of the technique. This observation contrasts the findings of Bhandari et al. [38], who reported an increase in the thickness of the ganglion cell complex (i.e., the three innermost layers of the retina: NFL + GCL + IPL) at five weeks after bilateral microbead injection, presumably as a result of inhibited axonal transport and concurrent axonal swelling. A more sensitive alternative to measure IPL thickness is morphometric analysis on transverse retinal sections, although the advantage of in vivo measurements is lost again. To conclude, if the abovementioned drawbacks of dendritic arbor analysis do not interfere with the envisioned experimental setup, we recommend dendritic arbor analysis as a substitute histological readout for the traditional RGC soma labeling and counts.
From all measures assessed in this study, STR recording ticks all the boxes (i.e., an early onset, easy accessible, and in vivo parameter) and is, in our opinion, the best readout for early stage glaucomatous damage in the microbead occlusion model. At five weeks after microbead injection, the amplitude of the STR peak was reduced by 31%. This observation is in line with the observations of Akopian et al. [25], who observed a 34% decline in STR amplitude at four weeks after single microbead injection. The alterations in STR amplitude in the microbead model are thus of a higher order of magnitude compared to the loss of RBPMS+ RGCs (here 9%) and potentially offer a larger window for therapeutic intervention in neuroprotection studies.
Being one of the major risk factors and currently still the only modifiable factor for glaucoma, elevated IOP typifies the majority of human glaucoma cases. Detecting modest changes in IOP, as seen in the chronic microbead occlusion model, is challenging given the variable nature of the tonometry technique in rodents, to say the least. We opted to explore the use of anterior segment OCT imaging as an easier-to-measure parameter to evaluate the model's success rate. Upon occlusion of the trabecular meshwork by the microbeads, the conventional outflow of aqueous humor is obstructed and the resulting fluid build-up in the anterior chamber presumably leads to iris flattening, altered iridocorneal angles, and an anterior chamber depth increment. While quantitative assessment of the iridocorneal angle in mice is unfeasible with standard lab equipment, including OCT, the anterior chamber depth is easy measurable. Substantial enlargement of the anterior chamber depth was noted from three weeks after microbead injection in our study, and anterior chamber depth measurements at five weeks after microbead injection showed a positive correlation with IOP values (measured at two weeks post-injection). These findings are in line with the spontaneous increase in anterior chamber depth in aging DBA/2J mice, a genetic glaucoma model [39]. Yet, this is the first report introducing the use of anterior chamber depth measurements as an alternative tool to evaluate the effectiveness of the microbead model.
We proved that changes in anterior chamber depth and STR are significantly altered at five weeks after ocular hypertension and reproducible in independent experiments. Logistic regression analysis was performed to judge the use of these measures as evaluation tools. This analysis revealed that both IOP and anterior chamber depth are good predictors of the success rate of the microbead occlusion model, yet the acquisition of the anterior chamber depth is easier and more robust. Moreover, STR recording, but not RBPMS+ RGC counts, is a significant predictor of disease status (i.e., normal or glaucomatous eyes) at five weeks after induction of the model, and has the added value of being noninvasive and highly translational. Of note, both anterior chamber depth and STR measurements necessitate corneal transparency. As such, we wish to stress the importance of correct microbead positioning and we recommend the use of paramagnetic microbeads that can be manually re-positioned around the circumference of the anterior chamber, as proposed by Ito and Belforte et al. [12]. During the intracameral injection, contact between the microbeads and the inner surface of the cornea should be avoided, as the microbeads have the tendency to perpetually stick to the cornea and could disable OCT and STR measurements [12]. This, together with the avoidance of inflammation after injection, requires an experienced operator. Another remark regarding this study is the lack of determination of an exact quantitative threshold for anterior chamber depth enlargement and STR decline, which was only defined as being higher/lower than the mean of the contralateral eyes ± one standard deviation. We recommend that each laboratory defines their own threshold, tailored to their experimental design and equipment. However, future work could include the development of a linear regression model that predicts STR values for a given IOP or anterior chamber depth measurement. This model would allow for an evaluation of the efficacy of neuroprotective treatments by assessing whether the STR is higher than the predicted value.

Microbead Occlusion Model
For the intracameral injections, mice were anesthetized with isoflurane (4% for induction, 1.5% for maintenance, Iso-Vet 1000 mg/g, Dechra, Northwich, UK), along with topical eye drops (0.4% oxybuprocaine hydrochloride, Unicaïne, Théa Pharma, Wetteren, Belgium). After pupil dilation (1% tropicamide, Mydriacyl, Novartis Pharma, Basel, Switzerland), 2 µL of 4.5 µm paramagnetic microbeads (Dynabeads™ M-450 Epoxy, ThermoFisher Scientific, Waltham, MA, USA), dissolved in sterile balanced salt solution (BSS plus, Alcon, Fort Worth, TX, USA), was unilaterally injected in the anterior chamber of the right eye, as previously described [13]. Intracameral injection was performed with in-house-made, beveled glass capillaries (1.0/0.75 mm OD/ID with filament, World Precision Instruments, Sarasota, FL, USA) and a MicroSyringe Pump Controller (100 nl per second, Micro4, World Precision Instruments). After injection, the microbeads were re-positioned towards the iridocorneal angle via a handheld magnet. Afterwards, an antibiotic ointment was applied to both eyes (0.3% tobramyxine, Tobrex, Alcon). As vehicle control, an equivalent volume of BSS was injected in the right eye, whilst the microbead-injected control refers to an identical microbead injection without the re-positioning of the microbeads towards the trabecular meshwork. The contralateral eye was always left untouched. Enucleations and optic nerve dissection were performed at five weeks post-injection.

Intraocular Pressure (IOP) Measurements
Using a calibrated rebound tonometer (TONOLAB, type TV02 for rat/mouse measurements, iCare, Vantaa, Finland), we measured the IOP two weeks after microbead injection in awake mice by gently grabbing the fur in the neck whilst still allowing free movement of the animal. The tip of the probe was kept at a distance of 1-3 mm from the center of the cornea. IOP measurements were performed in the morning between 9:00 and 11:00 a.m. under similar lightning settings to minimize the effect of diurnal IOP fluctuations, and were always recorded by the same experienced operator. Six single measurements were recorded, whereafter the four middle measurements were averaged. The same procedure was repeated three times, and the final IOP value was defined as an average of the three previous averages. The contralateral eye was used as an untreated control.

Retrograde Tracing
Unilateral retrograde tracing from the superior colliculus was performed as described by Nadal-Nicolás et al. [40]. Mice were anesthetized by an intraperitoneal injection of a mixture of 75 mg/kg body weight ketamine (Anesketin; Eurovet, Bladel, The Netherlands) and 1 mg/kg medetomidine (Domitor; Pfizer, New York City, NY, USA). The mouse head was fixed in a stereotaxic frame and a 2 × 2 mm 2 cranial window was made above the superior colliculus. The visual cortex was aspirated, and a gelatin sponge soaked in 0.9% saline containing 4% hydroxystilbamidine metanesulfonate (OHSt) and 10% dimethylsulfoxide (Life Technologies, Carlsbad, CA, USA) was applied on the left superior colliculus. The craniotomy was sealed with elastomer (Kwik-cast, World Precision Instruments), and the skin was sutured. After the procedure, mice were given meloxicam (5 mg/kg, Metacam, Boehringer-Ingelheim, Ingelheim am Rhein, Germany) for post-operative analgesia. Anesthesia was reversed with an intraperitoneal injection of 1 mg/kg atipamezole hydrochloride (Antisedan, Pfizer). A lubricant (Vidisic, Bausch + Lomb, Bridgewater, NJ, USA) was used to prevent dry eyes. Mice were sacrificed six days after tracing. Given the unilateral tracing, we used retinas of different, vehicle-injected mice as controls.

Spectral Domain Optical Coherence Tomography (OCT)
Upon general anesthesia with ketamine/medetomidine (cfr. above) and pupil dilation (0.5% tropicamide, Tropicol, Théa), two different scans (retina and anterior chamber segment) were acquired via spectral domain optical coherence tomography (OCT, Envisu R2210, Bioptigen, Morrisville, NC, USA). Both scans consisted out of 100 consecutive B-scans, each composed of 1000 A-scans, imaged in a rectangular 1.4 × 1.4 mm 2 (retina) or 3.0 × 3.0 mm 2 (anterior segment) field. During the imaging, a lubricant (GenTeal, Alcon) was used to improve tear film optics and to hydrate the cornea. Retinal layer thickness was measured using InVivoVue Diver 3.0.8 software (Bioptigen) at 16 locations equally spaced around the optic nerve head and averaged per mouse. Anterior chamber depth was defined as the distance between the corneal endothelium and the anterior surface of the lens, which was calculated using Fiji (ImageJ, version 1.52p) [15] on five, 150 µm spaced B-scans per mouse, positioned around the horizontal plane in the middle of the eye.

Electroretinography (ERG)
Prior to the electroretinogram recordings (Celeris, Diagnosys, Lowell, MA, USA), mice were dark-adapted overnight. Mice were anesthetized with ketamine/medetomidine (cfr. above), and pupils were dilated (0.5% tropicamide, Tropicol, Théa, and 15% phenylephrine hydrochloride, Phenylephrine, Théa) under dim red light. In both protocols-flash electroretinogram (ERG)/visual evoked potential (VEP) vs. scotopic threshold response (STR)-the eyes were alternately stimulated under fully dark-adapted conditions. To measure the ERG and VEP response, we used three active electrodes: two lens electrodes (with integrated touch stimulators) and one needle scalp electrode. The latter was inserted subdermally, above the midline near the visual cortex. The ground and reference electrode were placed in the tail base and cheek, respectively. Full field VEP/ERG responses were recorded at a single flash intensity of 0.05 cd*s/m 2 , averaging 300 brief flashes with an inter-sweep delay of 690 ms. To measure the STR, we solely used two lens electrodes, and the electrode on the contralateral eye was used as a reference. Full field STRs were averaged from 50 white light (6500 K) flashes with a single-flash intensity of 0.0001 cd·s/m 2 and an inter-sweep delay of 1 s under fully dark-adapted conditions. Espion v6.59.9 software (Diagnosys) was used to define the amplitudes and latency times of each response. The amplitudes and latencies of the ERG components were computed as previously described [41]. VEP and STR amplitudes were defined as the amplitude from the baseline to the trough of the negative VEP response, and from the baseline to the peak of the positive STR. Latencies were determined from flash onset to minimal and maximal amplitudes of the VEPs and STRs, respectively.

Tissue Sampling and Processing
Animals were euthanized by an overdose of sodium pentobarbital (60 mg/kg, Dolethal, Vetoquinol, Aartselaar, Belgium) and transcardial perfusion with 0.9% saline followed by 4% paraformaldehyde (PFA) in phosphate-buffered saline (PBS). Eyes and optic nerves were harvested. Eyes were fixated for one hour in 4% PFA at room temperature and flatmounted after 3 rinsing steps in PBS. Retinal flatmounts were fixated once more for one hour in 4% PFA. The optic nerves were fixated overnight in cold 2% glutaraldehyde buffered at pH 7.3 with 0.05 M sodium cacodylate and 0.05 M saccharose and post-fixated in cold 2% osmium tetroxide in the same buffer. Next, they were dehydrated in a graded acetone series and embedded in Araldite. A series of semi-thin sections (1 µm) was made with a EM UC6 ultramicrotome (Leica, Wetzlar Germany) at 3 mm distance from the optic nerve head, as axon degeneration is believed to progress in a distal-to-proximal manner [42].

Analysis of Axonal Integrity on Semi-Thin Sections
The semi-thin sections were stained with 1% toluidine blue (Sigma-Aldrich, Saint Louis, MO, USA) and 2% sodium borohydrite (Sigma-Aldrich) in distilled water for three minutes at 60 • C. After rinsing with distilled water, they were mounted with Depex (VWR, Radnor, PA, USA). Brightfield microscopy (100 × oil objective, DM6, Leica) was used to image the optic nerves. Axon density was calculated on three distal (3 mm post optic nerve head) cross-sections per optic nerve with the Fiji (ImageJ, version 1.52p) [43] plug-in AxoNet [15], which automatically outlines the optic nerve boundary, counts the RGC axons within this outline, and calculates the corresponding axon density (axons/mm 2 ).

Statistical Analysis
Eyes with inflammation or cataract were excluded from analysis. Logistic regression was performed in R (version 4.0.0), after scaling and centering the data automatically with the scale function [46]. All logistic models were exponentiated to report odds ratios with corresponding 95% confidence intervals. Effect size calculations (Hedges' g) were performed with DABEST [47]. Other calculations (e.g., sample mean and deviation) and statistical tests were performed using Prism 9.0.2 (GraphPad Software, version 9.0.2). Normality was tested with Shapiro-Wilk tests, while other statistical tests were specified in the figure legends, along with the sample size. All data are described as mean ± SEM in the text, whilst figures show box plots with the inter-quartile range (box) and total range (whiskers) of data, unless otherwise indicated. Statistical significance was set as p < 0.05 for all analyses.

Conclusions
We introduced the use of anterior chamber depth measurements via OCT imaging and STR recordings as easy-to-measure, in vivo tools to probe model induction and disease progression during the initial stages of the glaucomatous pathophysiology in the microbead occlusion model. First, the ease of use and fast acquisition time make anterior chamber depth measurements an attractive alternative compared to IOP readings that suffer from high intra-animal, inter-animal, and inter-observer variability, making this measure suitable for routine use in diverse studies. Second, measuring RGC functionality via STR is a better predictor for glaucomatous damage than quantification of RBPMS-immunopositive RGCs. Moreover, STR recordings allow for early detection of glaucomatous damage and yields effects that are large and consistent enough to be rescued by neuroprotectants. The noninvasive STR recordings can be assessed longitudinally during disease progression and can thus allow for frequent and long-term assessment of therapeutic efficacy. Given that OCT and ERG are common methods of early diagnosis of glaucoma in patients [2,48], the use of these tools might, at least in part, help to bridge the gap between preclinical and clinical glaucoma research. Finally, if feasible and opportune, we suggest using dendritic arbor analysis as a more sensitive alternative to RGC density counts in order to assess early stage glaucomatous damage in more in-depth experimental research.