OCT Biomarker Testing for Early Dysregulated Photoreceptor Metabolism in Dark-Reared Pde6b^rd10 Mice

Abstract

Purpose

The purpose of this study was to noninvasively test the hypothesis that dysregulated photoreceptor metabolism occurs with early retinal degeneration in dark-reared Pde6b^rd10 mice.

Methods

Postnatal day 23 (P23) phosphodiesterase (PDE) mutant (Pde6b^rd10) mice, P23 wild-type (WT) mice were continuously dark-reared, and 2-month-old dark-adapted (DA) C57BL/6J WT mice were studied following acute injection of the carbonic anhydrase inhibitor acetazolamide (ACZ) or saline control. Two optical coherence tomography (OCT) metabolic biomarkers were measured: (i) external limiting membrane–retinal pigment epithelium (ELM-RPE) thickness, a proxy for subretinal space acidosis, and (ii) the profile shape aspect ratio of the hyper-reflective band just distal to the ELM (band 2), a proxy for mitochondrial configuration within photoreceptor inner segments. Retinal laminar thicknesses were also assessed.

Results

P23 Pde6b^rd10 mice responded to ACZ with ELM-RPE contraction, but no change in band 2 profile shape aspect ratio; outer nuclear layer thickness was subnormal. In contrast, P23 WT mice showed no change in either OCT metabolic biomarker following ACZ administration. DA 2-month-old WT mice showed ELM-RPE contraction with ACZ; band 2 profile shape aspect ratio increased in male mice, with no change in female mice.

Conclusions

P23 Pde6b^rd10 mice with limited rod atrophy had biomarker values suggestive of dysregulation of RPE fluid/ion transport but with stable rod mitochondrial configuration following ACZ. DA 2-month-old male WT mice biomarkers were responsive to ACZ in a manner similar to those in light-adapted 2-month-old male WT mice; the difference in biomarker response pattern in female WT mice suggests a possible hormonal influence.

Retinitis pigmentosa (RP) is an umbrella term for a variety of mutations in rod photoreceptor genes that cause progressive vision loss and blindness. One form of RP involves an autosomal recessive missense mutation in the Pde6b gene that reduces PDE6 catalytic activity.^1–5 The Pde6b^rd10 mouse is a widely used RP model that develops a degeneration phenotype resembling that in humans.^5,6 The photoreceptors of Pde6b^rd10 mice in standard light conditions undergo rapid degeneration substantially limiting the time window available for testing for pathogenic mechanisms and neuroprotective therapies.^7–9 However, Pde6b^rd10 mice continuously raised in darkness show a markedly slower rate of photoreceptor degeneration but also outer retinal oxidative stress.^8,10 Further, ex vivo studies suggest impaired outer retinal metabolism prior to rod death in continuously dark-reared Pde6b^rd10 mice.^1,10–14

In this study, we test in vivo for metabolic problems in continuously dark-reared Pde6b^rd10 mice following systemically administered acetazolamide (ACZ) at an early stage of photoreceptor degeneration. ACZ inhibits carbonic anhydrase, an extracellular membrane-bound enzyme that facilitates the reversible hydration of carbon dioxide to form bicarbonate and protons on the basolateral surface of the retinal pigment epithelium (RPE).^15–17 When carbonic anhydrase is inhibited there is a buildup of protons in the subretinal space by reduction in H⁺/HCO₃^– transport across the RPE.^17–21 The resulting extracellular acidification upregulates a co-transporter on apical RPE that is involved with removing the low pH fluid; studies with diffusion magnetic resonance imaging (MRI), microelectrodes in vivo, and cell cultures ex vivo all support the notion that external limiting membrane (ELM)-RPE contraction result from subretinal space acidification.^16–23 Indeed, ACZ induces contraction of the ELM-RPE thickness in light-adapted 2-month-old male wild-type (WT) mice.¹⁹ Here, we further test this association in 2-month-old dark-adapted (DA) WT mice and their already-acidified subretinal space compared with light conditions.^19,20

In addition, we also measured the profile shape aspect ratio of the hyper-reflective band immediate posterior to the ELM (band 2), a proxy for mitochondrial configuration within photoreceptors as measured with its aspect ratio (MCP/AR).^{17–21,24–27} The acidosis accompanying an ACZ challenge also tests outer retinal metabolic regulation as measured by the rescaled and rotated MCP (rrMCP)/AR response.^{1,8,10–14,17–21,27,28} For example, mice with a nicotinamide nucleotide transhydrogenase null mutation (e.g. C57BL/6J) have a reduced respiratory reserve and reduced ability to handle metabolic stress, and their rrMCP/AR increases following an ACZ challenge.^{19,20,24–28} However, mice without this mutation (e.g. C57BL/6Ntac and 129S6/SvEv) show a stable or decreased rrMCP/AR response to ACZ.^{19,20,24–28}

Methods

All mice were treated in accordance with the National Institutes of Health Guide for the Care and Use of Laboratory Animals, the Association for Research in Vision and Ophthalmology Statement for the Use of Animals in Ophthalmic and Vision Research, and with specific authorization by the Wayne State University Division of Laboratory Animal Resources Institutional Animal and Care Use Committee (IACUC). The following mice groups were studied by optical coherence tomography (OCT) following an acute injection of the carbonic anhydrase inhibitor ACZ or phosphate buffered saline (PBS, control): continuously dark reared postnatal day 23 (P23) phosphodiesterase (PDE) mutant (Pde6b^rd10) mice (male mice: PBS, n = 5, ACZ, n = 5; female mice: PBS, n = 5, ACZ, n = 6), P23 WT mice (male mice: PBS, n = 5, ACZ, n = 4; female mice: PBS, n = 8, ACZ, n = 6); similarly studied were 2-month-old DA C57BL/6J WT mice (male mice: PBS, n = 5, ACZ, n = 4; female mice: PBS, n = 6, ACZ, n = 6). Breeder mice (Jackson Laboratories, Bar Harbor, ME, USA) were housed and maintained in complete darkness as were their pups (P23 Pde6b^rd10 and P23 WT); adult mice were housed and maintained in 12-hour/12-hour light-dark cycle laboratory lighting in the animal facility (2-month-old C57BL/6J WT). After scanning, mice were humanely euthanized by an overdose of ketamine/xylazine followed by cervical dislocation, per our Wayne State University Institutional Animal and Care Use Committee–approved protocol. Up to three OCT data sets were collected from the left eye to obtain the best orientation before mice woke from anesthesia, precluding acquisition from the right eye.

Acetazolamide Dosing

All mice were DA overnight and given an intraperitoneal injection of either 250 mg/kg ACZ (pH 9; Sigma product A6011) or PBS (vehicle, equal volume, pH 9) approximately 3 hours before OCT examination.

Spectral Domain Optical Coherence Tomography

Data were collected with a cross-sectional design. All mice examined by OCT (Envisu UHR2200; Bioptigen, Durham, NC, USA) were anesthetized with 100 mg/kg ketamine (Covetrus, Portland, ME, USA) and 6 mg/kg xylazine (MWI Animal Health, Boise, ID, USA) in the morning (i.e. before noon). Iris dilation was performed with 1% atropine sulfate, and Systane Ultra (Alcon, Geneva, Switzerland) was used to lubricate the eyes.

From the central retina, we collected radial volume scans with following parameters: A-scans/B-scans = 1000 lines; B-scans/volume = 1000 scans; and frames/B-scan = 1 frame. To minimize the effects of noise, and boost the signal, 100 images are registered and averaged before being analyzed. These 100 images were extracted from B-scan numbers 450 to 549 (representing inferior–superior retina) were registered (in-house script for R software; R Foundation for Statistical Computing, Vienna, Austria). Briefly, rigid body registration with RNiftyReg (a function in R software) was used to rotate the image and interpolate signal at each pixel. Next, columns of pixels are translated up or down to maximize alignment with the rest of the retina – a piecewise rigid-body registration effectively flattening the image to align retinal layers for further analysis. From these 100 inferior-superior oriented images, a Z-stack average intensity projection was obtained. Representative data are shown in Figure 1.

Figure 1.

Illustration of representative OCT images with ELM-RPE, EZ, and RPE bands noted for (A) continuously dark-reared (CDR) P23 male WT, (B) CDR P23 male Pde6brd10 (rd10; figure conventions are as in A), and (C) DA 2-month-old male WT groups (figure conventions are as in A); the DA 2-month-old female groups are not shown to save space. The EZ profile shape is used to generate rrMCP/AR as the ratio of minor to major aspect ratios (indicated in bottom ellipses in each panel). A black arrow indicates the direction of A-line data plotted in A, B, and C; the black arrow provides the x-axis scale reference.

Image Analysis

Layer segmentation boundaries were estimated using a machine learning model-based computer program, as described previously.¹¹ In short, the machine learning model was a U-net convolutional neural network trained using the “dice loss” function and the Adam optimizer (learning rate = 0.001), with 665 previously labeled images for training and 166 and 356 images for validation and testing, respectively. To improve the performance of the model, its predictions were postprocessed by applying a shortest-path algorithm. Based on these model-based estimates for laminar borders, final segmentation refinement was performed with an R script, as described previously.¹¹ The intensity values used to generate the ellipsoid zone (EZ) reflectivity profile shape are measured from a log-based image with a 16-bit depth (default in the Bioptigen system). We have previously compared images before and after converting them from log to linear values using a simple, empirically derived equation and confirmed that EZ reflectivity profile shape differences would be detected with either output.¹¹

Once segmented, inferior and superior retinas (i.e, 350–624 µm from the optic nerve head on the inferior and superior sides) were each analyzed; we started at 350 µm to ensure that images were analyzed without influence from the optic nerve head, and the outer retina is relatively uniform in all OCT data. Thus, the retinal length examined by the B-scan was 1248 mm. We measured the outer nuclear layer (ONL) and ELM-RPE thicknesses from in-house R software scripts that objectively extracted layer boundaries obtained after searching the space provided by machine-learning estimates (“seed boundaries”) as above. The ELM-RPE thickness was calculated by the distance from the ELM to the basal side of the RPE at the level of Bruch's membrane; ONL thickness was measured by the distance from the outer plexiform layer to the ELM.^29–31 The output image had the basal aspect of the RPE held in a fixed position without stretching the image; this non-stretched image was used to generate the A-line reflectivity profiles shown in Figure 1.

Our second biomarker is the rrMCP/AR.^19,24,32,33 The rrMCP/AR was measured from 10 inferior retina windows and 10 superior retina windows; each window was sized to be 19 (width) × 49 (height) pixels, or 23 (width) × 87 (height) mm. In each window, the hyper-reflective band immediately posterior to the ELM was displayed and its intersection with the baseline manually marked (MATLAB script; MathWorks, Natick, MA, USA). The program then found the local minimum near this mark in all 20 windows in an unbiased fashion. Connecting the two local minimums on either side of the hyper-reflective band defined its baseline. Note that the middle of each window was used as the distance for each window so distances for the rrMCP/AR are slightly different from those for ELM-RPE and ONL. Next, we measured the profile shape for the averaged A-line in each window. Using ImageJ software, under Analyze/Set measurements, we check the “Fit ellipse,” “Shape descriptors,” and “Invert Y coordinates.” The Fit ellipse measurement performs its “fitting” by obtaining the profile shape second moment, orientation, and centroid, and then generating an ellipse with the same second moment, orientation, and centroid. The rationale for generating an ellipse as a descriptor of an arbitrary shape based on the second moment is described formally in the literature.³⁴ This process in ImageJ software is described at https://github.com/imagej/ImageJ/blob/master/ij/process/EllipseFitter.java. The advantage of reframing the hyper-reflective band in terms of an ellipse is that an ellipse's shape can simply be described as an unbiased minor/major aspect ratio; note that in the ImageJ software output, the minor/major calculation is presented under the “Round” column. More work is needed to determine if the ellipse aspect ratio is an optimal shape description for these types of studies.^24,26,33

To reduce variability and thus better characterize the profile shape, here, we have refined MCP/AR with rescaling and rotating processing (i.e. rrMCP/AR) of the A-lines. The rescaling minimized the impact of reflectivity differences within and between mice. The rrMCP/AR thus represents a measure of profile shape which can include a baseline contribution. Each A-line is rotated so the slope of its baseline is zero to determine the location of max point of the profile and then rotated back for ellipse fitting. Whereas in most cases rotating the baseline to identify the maximum point on the A-line profile is a benign part of the processing pipeline, there are situations of extreme baseline tilt when it is beneficial (Supplementary Fig. S1) and thus, is performed on all images. The identified maximum point is rescaled to 100 and left-most location is set to minimum to 0 for the unrotated profile (MATLAB script). The rrMCP/AR does not appear to be a function of ELM-RPE thickness.¹⁹

Statistical Analysis

Data are presented as means with 95% confidence intervals. All statistical tests used a two-sided significance level of 0.05. Initial exploratory plots of spectral-domain OCT (SD-OCT) measurements (ELM-RPE thickness, rrMCP/AR, and ONL thickness) indicated that differences in means did not vary systematically with distance from the optic nerve head. Accordingly, we analyzed each measurement averaged by side within each mouse for all analyses. Data were analyzed separately for the different mouse ages (P23, 2-months-old). All outcomes were analyzed using linear mixed models that included the fixed effects of ACZ treatment (P23 mice only), genotype, sex, and side, as well as all interactions for initial models. The initial model also included a random intercept for mouse nested within sex, genotype, and treatment to account for within-mouse correlation. Model selection proceeded by comparing alternative random-effects and variance structures using Akaike's Information Criterion (AIC) and Bayesian Information Criterion (BIC). Interaction terms were evaluated using a hierarchical backward elimination approach, with nonsignificant higher-order interactions removed sequentially. Exploratory plots suggested the variances might depend on the various groups. As such, we evaluated models where the residual or random intercept variances were allowed to vary by sex, genotype, and/or treatment. For both rrMCP/AR and ELM-RPE thickness, the best-fitting models included the residual variance differing by genotype. For ONL thickness, the optimal model included a random intercept variance differing by genotype and the residual variance differing by sex, genotype, and treatment. Degrees of freedom for evaluating all fixed effects were calculated using the Kenward-Roger method. Nonsignificant higher-order interactions were removed to obtain the final model. All statistical modeling was done using the Proc Mixed procedure in SAS 9.4 (SAS software, Cary, NC, USA), and figures were generated using R software.

Results

P23 Pde6b^rd10 vs. P23 WT Mice ± ACZ

ELM-RPE Thickness

Following ACZ challenge, ELM-RPE contracted in P23 Pde6b^rd10 mice but was not evident in P23 WT mice regardless of sex or side (Fig. 2).

Figure 2.

Summary of ELM-RPE thicknesses ± ACZ in P23 WT (left graph) and Pde6b^rd10 (rd10, right graph) mice; male and female mice in each group were combined because the Treatment*Sex interaction was not significant (P = 0.98). Graphs show predicted means, raw data (closed circles), and 95% confidence intervals from the same linear mixed effects model fit to ELM-RPE. The solid horizontal line shows P < 0.05 (linear contrasts and hierarchical testing approach). These data represent information averaged from ± 350 to 624 µm from the optic nerve head on the inferior and superior sides.

Rescaled and Rotated Mitochondrial Configuration Photoreceptors Aspect Ratio

In neither group did rrMCP/AR change in response to ACZ, regardless of sex or side (Fig. 3).

Figure 3.

Summary of rrMCP/AR ± ACZ in P23 WT (left graph) and Pde6b^rd10 (rd10, right graph) mice; male and female mice in each group were combined because the Treatment*Sex interaction was not significant (P = 0.64). All figure conventions are as in Figure 2.

Outer Nuclear Layer

ONL thickness remained stable in both groups following ACZ regardless of sex or side (Fig. 4).

Figure 4.

Summary of ONL thickness ± ACZ in P23 WT (left graph) and Pde6b^rd10 (rd10, right graph) mice; male and female mice in each group were combined because the Treatment*Sex interaction was not significant (P = 0.56). All figure conventions are as in Figure 2.

DA 2-Month-Old WT Mice ± ACZ

ELM-RPE Thickness

ELM-RPE contracted in DA 2-month-old WT mice treated with ACZ compared with vehicle controls, regardless of sex or side (Fig. 5).

Figure 5.

Summary of ELM-RPE thickness ± ACZ in DA 2-month-old WT mice; male and female mice in each group were combined because the Treatment*Sex interaction was not significant (P = 0.35). All figure conventions are as in Figure 2.

Rescaled and Rotated Mitochondrial Configuration Photoreceptors Aspect Ratio

In male mice but not female DA 2-month-old mice, rrMCP/AR increased following ACZ compared with vehicle controls (Fig. 6).

Figure 6.

Summary of rrMCP/AR ± ACZ in DA 2-month-old male (left graph) and female (right graph) WT mice; male and female mice in each group were not combined because the Treatment*Sex interaction was significant (P = 0.04; female ACZ effect = 0.05 [95% CI = –0.02 to 0.13]; male ACZ effect = 0.17 [95% CI = 0.09 to 0.26]). All figure conventions are as in Figure 2.

Outer Nuclear Layer

Regardless of sex or side, ONL thickness decreased following ACZ (Fig. 7).

Figure 7.

Summary of ONL thickness ± ACZ in DA 2-month-old WT mice; male and female mice in each group were combined because the Treatment*Sex interaction was not significant (P = 0.49). All figure conventions are as in Figure 2.

Discussion

In this study, we tested in vivo the hypothesis of an underlying early metabolic dysregulation in the outer retina of continuously dark-reared Pde6b^rd10 mice. This hypothesis is based on previous studies showing early disruption of retinal nucleotide and acylcarnitine metabolism measured with metabolomics, and outer retinal oxidative stress in dark-reared Pde6b^rd10 mice.^10,11,13 We first needed to address the problem of how to challenge the retina without the confounding effects of light-induced acceleration of rod photoreceptor degeneration in the Pde6b^rd10 group.^5,8,9,13 To this end, we chose ACZ to test the regulation of outer retinal metabolism.^8–10,13

In this study, we also asked if ACZ-challenged DA 2-month-old male WT mice show similar responses as light-adapted 2-month-old male WT mice with regard to OCT metabolic biomarker patterns. Indeed, both light- and DA C57BL/6J mice challenged with ACZ produce similar metabolic biomarker patterns and thus a similar conclusion of inefficient photoreceptor mitochondria.^{1,8,10–14,17–21,27,28} An unexpected secondary finding was that the biomarker response pattern in DA+ACZ-challenged 2-month-old female WT mice was different from that in male mice; further experiments are needed in, for example, light-adapted, ACZ challenged 2-month-old female WT mice to evaluate a possible hormonal influence.

ELM-RPE Thickness and its Response to ACZ

Somewhat surprisingly, P23 WT mice did not show ELM-RPE contraction following ACZ, in contrast to the contractile response observed in 2-month-old DA WT male C57BL/6J mice.^7,8,19,20 We considered two possibilities for why ELM-RPE is unresponsive to ACZ in P23 WT mice: their younger age and/or their dark-rearing. However, our previous data did not support younger age as a factor underlying the lack of response to ACZ in P23 WT male mice.¹³ In that previous study, P20 WT mice showed proof-of-concept evidence that ELM-RPE contraction occurs in response to subretinal space acidosis resulting from mitochondria activity maintaining the dark current during dark adaptation.¹³ Thus, dark-rearing and its associated changes in photoreceptor metabolic regulation seem to be a reasonable explanation for the ACZ-unresponsive ELM-RPE results in P23 WT mice.^35,36 In support of this hypothesis, we note recent evidence that dark-rearing in WT mice impairs retinal energy metabolism.³⁷ Additional work is needed to further test this hypothesis.

On the other hand, P23 Pde6b^rd10 mice showed an ELM-RPE contraction in response to ACZ – unlike the unresponsive ELM-RPE in P23 WT mice. Previous studies find evidence for light-dependent alterations in fluid transport across the RPE in 12:12 dark:light-reared Pde6b^rd10 mice, but the present study is the first to suggest such changes under dark-rearing conditions.^7,38 Together, the above considerations provide evidence supporting abnormal RPE fluid/ion transport regulation and/or pH buffering in Pde6b^rd10 mice.

rrMCP/AR and its Response to ACZ

This increase in rrMCP/AR was also noted in the present study in 2-month-old DA WT male C57BL/6J mice. In this study, rrMCP/AR was unresponsive following ACZ in both P23 WT and Pde6b^rd10 mice, suggesting preserved metabolic control.^7,8,19,20

Secondary Finding of ONL Thickness and its Response to ACZ

Previous histology reports suggested that the retina does not thin in dark-reared male or female Pde6b^rd10 mice by P23.^8,39,40 However, our OCT measurements in this study showed statistically significant thinning of the ONL in continuously dark-reared P23 Pde6b^rd10 mice relative to WT. In addition, neither P23 WT nor Pde6b^rd10 mice showed ONL thickness responses to ACZ in this study. In contrast, DA 2-month-old WT mice showed an ACZ-evoked thinning of ONL.¹⁹ This thinning response was also previously noted in 2 strains of light-adapted 2-month-old WT mice.¹⁹ As considered earlier, one intriguing possibility may involve ACZ-sensitive carbonic anhydrase in Müller cells, an important route of water removal from nuclear layers.⁴¹ Notably, Müller cell migration toward the ONL has been reported in Pde6b^rd10 mice, which could both reduce the ONL thickness and blunt an ACZ response.⁴⁰ Additional studies are needed to follow up on these possibilities.

Conclusions

In summary, we provide first-in-kind noninvasive biomarker evidence for dysregulation of outer retinal ion transport/pH buffering in continuously dark-reared Pde6b^rd10 mice, as measured by an abnormal contraction of ELM-RPE thickness following ACZ.^7,38 ELM-RPE contraction can arise from oxidative stress-induced subretinal space acidosis, and oxidative stress has been reported in continuously dark-reared Pde6b^rd10 mice.^10,13,42 However, the importance of rod oxidative stress is unclear, because treatment of continuously dark-reared Pde6b^rd10 mice with antioxidants that suppress rod oxidative stress in this model in vivo did not confer neuroprotection; published studies from other groups find mixed results on rods and cones following anti-oxidant treatment.^13,43,44

On the other hand, intact regulation of metabolism was observed following ACZ, as measured by rrMCP/AR, in male and female Pde6b^rd10 mice.^10,11,13 These results are somewhat surprising because perturbed steady-state metabolism has been reported in continuously dark-reared Pde6b^rd10 mice.^10,11,13 Taken together, the novel, noninvasive results herein do not support strategies that only target mitochondria to slow or prevent the progression of photoreceptor degeneration in continuously dark-reared Pde6b^rd10 mice.¹³