Influence of Opa1 Mutation on Survival and Function of Retinal Ganglion Cells

Abstract

Purpose: Mutations in the OPA1 gene cause autosomal dominant optic atrophy (ADOA), a visual disorder associated with degeneration of retinal ganglion cells (RGCs). Here, we characterized the disease progression in a homologous mouse model B6;C3-Opa1^329-355del and asked whether the pronounced cell death affects certain RGC types more than others.

Methods: The influence of the Opa1 mutation was assessed by morphologic (retina and optic nerve histology) and functional (multielectrode array) methods.

Results: The RGC loss of approximately 50% within 18 months was significantly more pronounced in RGCs with small-caliber axons. Small-caliber axon RGCs comprise a variety of functional RGC types. Accordingly, electrophysiological analyses of RGCs did not show a dropout of distinct functional RGC subgroups. However, the response properties of RGCs were affected significantly by the mutation. Surprisingly, these functional changes were different under different luminance conditions (scotopic, mesopic, and photopic). Finally, melanopsin cells are known to be less susceptible to retinal insults. We found that these cells are also spared in the Opa1 mouse model, and demonstrated for the first time that this resistance persisted even when the melanopsin gene had been knocked-out.

Conclusions: Small-caliber axons show a higher vulnerability to the Opa1 mutation in our mouse model for ADOA. Luminance-dependent functional changes suggest an influence of the Opa1 mutation on the retinal circuitry upstream of RGCs. Photoresponsive RGCs are protected against cell death due to the Opa1 mutation, but not by melanopsin expression itself.

Autosomal dominant optic atrophy (ADOA) is the most frequent hereditary neuropathy, besides Leber's hereditary optic neuropathy, with an estimated prevalence of 1:30,000 worldwide.¹ Autosomal dominant optic atrophy patients suffer from a slow but progressive bilateral vision loss characterized by caecocentral scotoma, pallor of the optic disk together with reduced thickness of the nerve fiber layer, and specific color vision disturbances in the blue-yellow axis. Further, extensive cell death has been observed in retinal ganglion cells (RGCs), but no other retinal cell type.^2,3 However, there is high variability in onset, visual impairment, and penetrance, indicating that the phenotype might be influenced by the genetic background, genetic modifiers or the environment.¹ The Optic Atrophy 1 gene (OPA1) is by far the most common gene mutated in ADOA and accounts for 65% to 90% of cases in well-defined patient cohorts.^4,5 The nuclear OPA1 gene encodes a ubiquitously expressed mitochondria-targeted, dynamin-related GTPase, which is a key factor in the mitochondrial fusion dynamics.⁶ OPA1 also participates in the formation of protein complexes that bridge and tighten the cristae junction leading to retention of cytochrome C in the lumen of the cristae,^7,8 and it interacts with the oxidative phosphorylation complexes I, II, and III.^9–11 

Two mouse models for nonsyndromic ADOA with mutation in the OPA1 gene have been developed: B6;C3-Opa1^Q285STOP and B6;C3-Opa1^329-355del.^12,13 Both models reflect the phenotype (optic nerve atrophy, altered mitochondria organization) observed in patients with similar mutations. However, the RGC loss in the B6;C3-Opa1^Q285STOP mouse is limited, while the B6;C3-Opa1^329-355del model shows a significant reduction in RGC counts.¹⁴ Retinal ganglion cells are a heterogeneous cell class, which can be divided into several clusters depending on their morphologic¹⁵ or functional¹⁶ properties. The various subpopulations might be affected differently by pathophysiological events. In fact, Williams et al.¹⁷ identified in the B6;C3-Opa1^Q285STOP mouse model an altered dendritic morphology together with changes in the synaptic profile in the On-center RGCs. The Off-center RGCs remained unaffected. So far, there has been no detailed study on how the Opa1 mutations affect the function of different RGC subpopulations. Here, we aimed at identifying RGC types that might be specifically affected in the B6;C3-Opa1^329-355del model, by morphologic analysis of the optic nerve and functional evaluation of light responses of RGCs. 

Further, we studied the melanopsin system and the potential protective role of melanopsin in the B6;C3-Opa1^329-355del mouse model. It has been shown that melanopsin-expressing RGCs¹⁸ might be preserved in ADOA patients as well as in the B6;C3-Opa1^Q285STOP mouse model.^3,19 This led to the hypothesis that melanopsin expression per se may protect these cells from light damage,³ whereas similar light exposure might be harmful to nonmelanopsin RGCs, especially in the presence of abnormal cellular function.²⁰ 

Materials and Methods

Animals and Experimental Design

Male and female B6;C3-Opa1^329-355del and Opn4^tauLacZ mice were used in this study. Animals were fed with standard food and tap water ad libitum, and maintained under constant temperature conditions (22°C ± 2°C) and daily cycles of 12-hour light/darkness. All experiments were performed in accordance to the ARVO Statement for the Use of Animals in Ophthalmic and Vision Research as well as the Tübingen Institutional Animal Care and Use Committee (Tübingen, Germany), and have been approved by the local authorities. 

Opa1 Mouse Line

The B6;C3-Opa1^329-355del mouse strain was used in this study. In the homozygous state this mutation is lethal. We used heterozygous mutants (Opa1^enu/+) and their wild-type littermates (Opa1^+/+, in the following text called WT) to determine the effects of the Opa1 mutation on the number and functionality of the RGCs and the melanopsin system. Animals were killed at 1, 6, 12, and 18 months of age. The day prior to the experiments, mice were kept in darkness (necessary to perform multielectrode array [MEA] recording under scotopic conditions) and killed under dim red light. All animals are listed in Supplementary Table S1. 

Opn4^tau-Lacz Mouse Line

The Opn4^tau-lacZ mouse (kindly provided by King-Wai Yau, Johns Hopkins School of Medicine, Baltimore, MD, USA), has the Opn4 gene (encoding melanopsin) replaced with sequences encoding for a tau-lacZ fusion protein. This mouse model has no expression of melanopsin, but “naturally” melanopsin-expressing RGCs can be labeled by β-galactosidase staining. This allows the identification of these cells when melanopsin is removed, and to determine the potentially protective role of the melanopsin protein itself. Therefore, we crossbred the B6;C3-Opa1^329-355del with the Opn4^tau-lacZ mice and analyzed 18- to 23-month-old animals. These animals are listed in Supplementary Table S2. 

Optic Nerve Analysis

Tissue Sampling.

Optic nerve samples were placed for 2 hours in 2% glutaraldehyde/2% paraformaldehyde in 0.1 M cacodylate (CACO) buffer (pH 7.4), then washed three times in 0.1 M CACO buffer and postfixed in 1% osmium tetroxide (OsO₄) in 0.1 M phosphate buffer (PB) for 80 minutes. After washing three times in 0.1 M CACO buffer, samples were dehydrated in ethanol: 50%, 70%, 70% ethanol with uranyl acetate overnight at 4°, 80%, 90%, 95%, 100%, 100% ethanol with molecular sieve and finished by immersing the samples twice in propylene oxide (Sigma-Aldrich Corp., St. Louis, MO, USA). The samples were placed in EPON resin (Sigma-Aldrich Corp.) with propylene oxide (ratio 1:1) for 1 hour and then embedded in EPON resin overnight. The resin was polymerized after placing the samples into a flat mold, covered with fresh 100% EPON resin with 1.5% Starter (N-Benzyldimethylamin 98%; Sigma-Aldrich Corp.) and incubated at 55°C for 96 hours. All steps were performed at room temperature (RT) and for 10 minutes, unless otherwise noted. 

Sectioning.

Using a glass knife on an ultramicrotome (Leica, Solms, Germany), 0.5-μm thick sections were cut and placed on distilled water on a poly-L-lysine coated slide, flattened by heating (40°C for ∼30 minutes), and stained with Richardson solution.²¹ The slides were then coverslipped and sealed with DPX mounting medium (Sigma-Aldrich Corp.). 

Acquisition and Processing of the Images.

Three random nonoverlapping 5809-μm² squares including peripheral and central regions were imaged with a digital camera (Olympus V-CMAC3; Olympus, Tokyo, Japan) mounted on an Olympus AX70 microscope. Images were processed with a modification of the method described by Tolivia et al.^22,23 (example images shown in Fig. 1A were contrast enhanced in Adobe Photoshop [Adobe Systems, Inc., San Jose, CA, USA] for clarity). Briefly, in Adobe Photoshop the specific outlines of axons were selected (mask). Being based on light-microscopic images, this analysis misses nonmyelinated axons, which however represent only 1% to 2% of the total number of axons.²⁴ Nevertheless, even small-caliber axons, if myelinated, are reliably detected, and consequently our results are fully consistent with previous observations based on electron microscopy analysis.¹³ The nonneural elements were then removed from the mask by hand, and this “clean” mask was analyzed with the “Analyze particles” routine (ImageJ 1.37c²⁵). By this process, we generated an image in which each axon is labeled with a number allowing its identification as well as its morphometric characteristics (area, perimeter; Fig. 1A). Additionally, the whole section of the optic nerve was imaged in order to measure the cross-sectional area. Data is represented as estimated number of axons per optic nerve, which was calculated based on the overall cross-sectional area of each optic nerve and the number of axons observed in the three studied areas. 

Figure 1

 

The number of small caliber axons is decreased in aged Opa1^enu/+ mice. (A) Representative light-microscopic images of Richardson-stained optic nerve cross sections. Top row: For axonal counting, a mask was automatically calculated and manually cleaned up to remove nonaxonal parts such as blood vessels (top middle); the resulting axon layouts were analyzed with respect to their number, area, and perimeter (top right, magnified view of the area indicated in the middle). In each optic nerve cross-section, three areas were analyzed (one central, two peripheral). Representative areas are shown for 1-month-old WT (top left) and Opa1^enu/+ mice (bottom left), and for 18-month-old WT (bottom middle) and Opa1^enu/+ mice (bottom right). Scale bar: 10 μm. (B) Axon numbers across age and genotype. The overall axonal number is significantly decreased in the Opa1^enu/+ 18-month group (∼50% decrease when compared with the WT littermates) with an influence of the variables “age” and “genotype” (ANOVA test, P < 0.0001). No influence of the age was detected within the WT group (ANOVA test, P > 0.05) or of the genotype at the age of 1 month (ANOVA test, P > 0.05; inset). The number of small axons (<0.4 μm²) is significantly higher than that of the big ones, without influence of the age or genotype (ANOVA test, P > 0.05). The estimated number of the small axons (<0.4 μm²) is significantly reduced in the Opa1^enu/+18-month group when compared with WT 18 month or Opa1^enu/+ 1 month. Scheffe post hoc test revealed a decrease in the number of axons that belong to the 0.1 to 0.2 μm² bin in the Opa1^enu/+ 18-month animals compared with WT 18 month and with Opa1^enu/+ 1 month (P < 0.01). Similar result was obtained related to the 0.2- to 0.3-μm² bin (Opa1^enu/+ 18 month versus WT 18 month, P < 0.01; Opa1^enu/+ 18 month versus Opa1^enu/+ 1 month, P < 0.05). Axonal loss was also observed in the 0.01- to 0.1- and 0.3- to 0.4-μm² bin in the Opa1^enu/+ 18-month group when compared with the WT 18 month (P < 0.05 for both comparisons). No significant changes were detected in bins with areas > 0.4 μm² (ANOVA test, P > 0.05). Opa1^enu/+ 18 month: n = 7, WT 18 month: n = 5; Opa1^enu/+ 1 month: n = 3, WT 1 month: n = 4). *P < 0.05, **P < 0.01, ***P < 0.001, Scheffe post hoc test.

Multielectrode Array Recordings

Eye cups were placed in Ringer solution (in mM: 110 NaCl, 2.5 KCl, 1 CaCl₂, 1.6 MgCl₂, 10 D-Glucose, and 22 NaHCO₃ bubbled with 5% CO₂/95% O₂). All recordings were performed with a perforated 60-electrode MEA (60pMEA200/30iR-Ti-gr; Multichannel Systems, Reutlingen, Germany) as described previously.²⁶ Briefly, the retina was placed ganglion cell-side down in the recording chamber and perfused with Ringer solution at 34°C. Stimulation was performed from the bottom with a Digital Light Processing projector (Sharp PG-F212X-L; Sharp Corporation, Osaka, Japan). Data was recorded at 25 kHz with a USB or MC-Card based MEA-system (USB-MEA1060 or MEA1060; Multichannel Systems). 

Visual Stimulation.

The same set of stimuli (see below) was focused through the condenser onto the retina at scotopic (mean intensity: 8 rhodopsin isomerizations (R*·s⁻¹ per rod), mesopic (8·10² R*·s⁻¹), and photopic light levels (8·10⁴ R*·s⁻¹) for 1 to 1.5 hours each. Each stimulus was repeated 4 to 15 times at each light level. 

Spike Sorting and Correlation of Stimuli and Responses.

Spike waveforms and spike-times were extracted from raw data with Matlab (MathWorks, Natick, MA, USA). Different features of the action potentials (amplitude, width, principal components) were calculated and projected onto 2-dimensional space either with Offline Sorter (Plexon, Inc., Dallas, TX, USA) or with an in-house written Matlab routine in order to semimanually assign spikes to individual units (presumably RGCs). We extracted 654 cells (Opa1^enu/+: n = 85 from 3 retinas (18 months), n = 119 from 4 retinas (12 months), n = 108 from 4 retinas (6 months), n = 62 from 4 retinas (1 month); WT: n = 80 from 5 retinas (18 months), n = 74 from 3 retinas (12 months), n = 52 from 4 retinas (6 months), n = 74 from 3 retinas (1 month), see also Supplementary Table S1. To calculate the firing rate, the spike train of each cell was convolved with a Gaussian (sigma 40 ms), and the resulting firing rate trace was plotted against the time course of the presented stimuli. For each brightness level (scotopic, mesopic, photopic), we analyzed the parameters described below. 

1.  
   Spatial tuning: Responses to 24 different drifting sine-wave gratings (combination of different temporal frequencies: 1, 2, 4, 8 Hz; and spatial periods: 100, 200, 500, 1000, 2000, 4000 μm) were used for spatial tuning calculations. For each stimulus we calculated the Fourier transform of the response. The amplitude of the Fourier transform at the frequency of the grating stimulus was taken as the response strength of the cell. To estimate the spatial tuning, we calculated the median response to all grating stimuli with the same spatial period (Figs. 2A, 2B);
2.  
   Temporal frequency tuning: The same analysis as for the spatial tuning was performed, but response median to the same temporal frequency were calculated (Figs. 2A, 2B);
3.  
   Sustainedness: Peak responses to full-field contrast steps (gray → black → gray → white → gray, step duration: 2 seconds) were extracted. For the step which caused the maximal response, the fraction of this maximal response and the remaining average response strength 1700 to 2000 ms after stimulus onset was defined as “sustainedness” (Fig. 3A₁). Cells whose firing rate dropped back to or below their background firing rate within these 2 seconds (i.e., transient cells) were assigned a sustainedness parameter of 0;
4.  
   Latency: The same full-field contrast step as above was taken for latency measurements. Latency was defined as the time between stimulus onset and the peak response (Fig. 3A₁); and
5.  
   Speed tuning: Speed tuning was calculated from a single bar moving with various speeds (bar width: 1000 μm; speeds: 0.2, 1, 2, 4, 8, 16 mm/s; black and white bars used). The peak firing rate to each speed was measured and summed starting from the slowest speed (cumulative sum). Further parameter calculations were restricted to the stimulus to which the cell responded better (black or white bar) based on the summed peak firing rates. As speed tuning parameter we defined the speed at which the cumulative sum exceeded 50% (Fig. 3A₂).

Figure 2

 

Spatial and temporal tuning parameter calculations and results. (A₁) Response of a single RGCs to a drifting sine wave grating. The cell responded throughout the presentation of a sine wave grating with a period of 200 μm moving with 2 Hz. (A₂) Fourier transformation of response in (A₁). The peak value around the stimulus frequency was taken as the response strength of the cell for further analysis (gray arrow). (B) Calculation of temporal frequency tuning and spatial tuning. Peaks from Fourier transforms (A₂) are color-coded with white being the maximal peak measured for a single cell. For spatial tuning parameters, the median for each spatial period over various temporal frequencies was calculated (below). Similarly, temporal frequency tuning parameters were calculated by averaging over various spatial periods (right). Gray arrow: data point from (A₁) and (A₂). (C) Distribution of spatial tuning parameters from 18-month-old animals. Left: median (circle), 95% confidence interval of median (thick line), and SD (thin line), for data obtained under scotopic conditions. Black arrow: data recorded in response to sine wave gratings with 200-μm periods, also indicated in (D). Right: bar-whisker plots for data recorded under photopic conditions. (D) Statistical differences between spatial tuning in WT and Opa1^enu/+ mutants. Wilcoxon rank sum tests were applied to detect significant differences between the spatial tuning parameter distribution of WT and Opa1^enu/+ RGCs at all ages and luminance conditions. Each square represents one P value which is color-coded according to the legend on the right. Black arrow indicates the same data as in (C). Underlying raw data is given in Supplementary Figures S1 through S4. (E) Statistical comparison of temporal frequency tuning in Opa1^enu/+ and WT retinas. Wilcoxon rank sum tests were applied to detect statistical differences between WT and Opa1^enu/+ RGCs as described for spatial tuning in (D). Underlying raw data is given in Supplementary Figures S5 through S8. *P < 0.05, **P < 0.01, ***P < 0.001, Wilcoxon rank sum test.

Figure 3

 

Response sustainedness and latency, and speed tuning in WT and Opa1^enu/+ RGCs. (A₁) Calculation of sustainedness and latency. Sustainedness was calculated from responses to full-field contrast steps. The remaining activity 1700 to 2000 ms after onset of the stimulus that elicited the maximal response was considered. Sustainedness was expressed as the fraction of this remaining activity from the maximal response (both baseline-subtracted). Latency was defined as the time from stimulus onset to the peak response. (A₂) Calculation of speed tuning parameter. Speed tuning was calculated from peak responses to a bar moving with six different speeds. Peak responses were summed starting from the slowest speed (cumulative sum), and speed tuning was defined as the speed for which 50% of the summed responses was reached. (B) Statistical differences between sustainedness, latency, and speed tuning of Opa1^enu/+ and WT RGCs calculated by Wilcoxon rank sum tests. P value of the difference is color-coded as described in Figure 2. Transient cells were excluded before applying statistical tests for the sustainedness parameter. Sustainedness differed between Opa1^enu/+ and WT retinas mostly at old age, but also under scotopic conditions at the age of 6 months (left). Retinal ganglion cells response latency to full-field flashes is similar in both genotypes under all conditions and at all ages (middle). Only under mesopic conditions, WT RGCs have a borderline significantly lower latency at the age of 12 months. Speed tuning is similar in both genotypes under all conditions (right). Raw data is depicted in Supplementary Figures S9 through S11.

Parameters described above were calculated automatically. Cells that did not respond at all to a certain stimulus were removed manually (for single parameters) after inspection of the raw response as well as of the analyzed parameters where applicable (e.g., Fourier transforms). 

Immunohistochemistry Analysis

Eye cups were fixed in 4% paraformaldehyde either for 24 hours at 4°C (melanopsin immunohistochemistry [IHC]) or for 2 hours at RT (β-galactosidase IHC). Fluorescence IHC was performed on whole-mounted retinae using an anti-mouse melanopsin antibody (UF006; Advanced Targeting Systems, San Diego, CA, USA) 1:5000 diluted in PBS with 0.4% Triton X-100 (PBS-T) for 72 hours at 4°C, and an anti-mouse β-galactosidase antibody (ab9361; Abcam, Cambridge, UK) 1:1500 diluted in PBS-T for 72 hours at 4°C as described previously.²⁷ Whole-mounted retinas were imaged with a fluorescence microscope (Olympus AX70; Olympus). The somata of the immunopositive cells were counted across the whole retina. 

Statistical Analysis

Optic Nerves Analysis and Immunohistochemistry.

SPSS 15 software (SPSS, Chicago, IL, USA) was used for statistical analysis. Kolmogorov-Smirnov and Levene's test were used to assess the normal distribution and homocedasticity of the data. One-way ANOVA tests were performed to analyze the effect of the genotype and age on the number of axons, number of melanopsin- and β-galactosidase-immunopositive cells. Scheffe post hoc tests were performed to detect differences between specific genotypes or ages. All data are presented as mean ± SEM. P less than 0.05 was considered statistically significant. Data was plotted in Matlab. 

MEA Recordings.

For each age group, brightness level, and response parameter, we performed a Wilcoxon rank sum test on the measurements of Opa1^enu/+ and WT RGC response parameters, as described in the results section. Where applicable, data is presented as median ± SD and ±95% confidence interval. Data was plotted in Matlab. 

Results

Optic Nerve Analysis

A large variability in visual impairment and corresponding RGC loss has been reported both in ADOA patients²⁸ and in the B6;C3-Opa1^329-355del mouse model.^13,29 In order to assess the progression of RGC degeneration over time, we have quantified the number of axons in the optic nerve in 1- and 18-month-old WT and Opa1^enu/+ animals (Fig. 1A). No significant differences in the cross-sectional area of the optic nerve was observed between the different groups (ANOVA test, P > 0.05; data not shown). However, one-way ANOVA test showed an influence of the variables genotype and age on the estimated number of axons (P < 0.0001). Scheffe post hoc test revealed significant reduction in axon number in Opa1^enu/+ 18 months when compared both with WT 18 months (P < 0.001) and with Opa1^enu/+ 1 month (P < 0.05; inset Fig. 1B), with similar axonal loss in the central and peripheral optic nerve (ANOVA test, P > 0.05; data not shown). However, the phenotype was quite variable: the axonal loss ranged from 23% to 77% (average: 47%; Opa1^enu/+ 18 versus WT 18 months). We detected no differences between WT and Opa1^enu/+ at the age of 1 month (Scheffe post hoc test, P > 0.05), indicating that no axonal loss occurred at earlier ages. Moreover, no significant influence of age was detected in the WT (Scheffe post hoc test, P > 0.05). These results suggest that the decrease in the number of axons observed in the Opa1^enu/+ 18-month animals is time- and mutation-dependent. 

We classified the axons into different groups depending on their morphometric parameters (area, perimeter). The axons were classified into 11 bins depending on their cross-sectional area (with 0.1-μm² step size up to 1 μm², and one additional bin with axons >1 μm², Fig. 1B). Small axons (smaller than 0.4 μm²) were more abundant than big ones (ANOVA test, P < 0.0001 in all groups, Fig. 1B). We observed a significant effect of the variables genotype and age on the number of axons with an area up to 0.4 μm² (ANOVA test, bins 0–0.1 μm² and 0.1–0.2 μm²: P < 0.001; bin 0.2–0.3 μm²: P < 0.01; bin 0.3–0.4 μm²: P < 0.05, Fig. 1B). No significant changes were detected for the other bins (ANOVA test, P > 0.05 in all bins with areas > 0.4 μm²). Similar results were observed when the axons were classified depending on their perimeter (data not shown). It is noteworthy that approximately 95% of the axons lost in the Opa1^enu/+ 18-month animals are smaller than 0.4 μm². We observed no influence of the age within the WT group or of the genotype at the age of 1 month (ANOVA test, P > 0.05). 

Ganglion Cell Responses Differ Between Old WT and Old Opa1^enu/+ Retinas

Our morphologic data suggest that ganglion cells with small-caliber axons are especially vulnerable in Opa1^enu/+ retinas. We wondered if this would be reflected in a selective loss of functional subtypes of RGCs. We thus recorded visual responses from RGCs in isolated whole-mount retina at three different luminance levels (scotopic, mesopic, photopic) using MEAs. We characterized each recorded RGC with several functional parameters (see Methods). 

Spatial and Temporal Tuning Differs Between Opa1^enu/+ and WT RGCs.

Spatial and temporal response properties of individual RGCs were calculated from 24 different drifting sine-wave grating stimuli (6 different spatial scales × 4 different temporal frequencies, see Methods and Figs. 2A, 2B; population data are presented in Supplementary Figs. S1–S8). We measured the response strength of individual RGCs for each of the 24 different grating stimuli. For example, Figure 2A₁ shows the raw response of a single RGC (from an 18-month Opa1^enu/+ retina recorded in scotopic conditions) to the 200-μm grating drifting at 2 Hz. Figure 2A₂ shows the corresponding Fourier transform, from which we extracted the response strength to this stimulus (arrow). The raster in Figure 2B shows the color-coded response strengths of the same cell to all 24-grating stimuli. Spatial and temporal tuning for this cell was then calculated by taking the median of the response strengths along the space and time dimensions of this response raster, as indicated in Figure 2B. Such measurements were performed for all RGCs and the results were grouped separately for the four age groups, three brightness levels, and two genotypes. Examples of the resulting distributions of spatial tuning are shown in Figure 2C where we compare the response strengths of 18-month-old WT (blue) and Opa1^enu/+ (orange) RGCs measured in scotopic (left) and photopic (right) conditions. We applied Wilcoxon rank sum tests to these distributions to assess if they differed between Opa1^enu/+ and WT retinas. A summary of these tests for each age group and brightness condition is shown in Figure 2D as color-coded P values (red if a parameter was enhanced in Opa1^enu/+ compared with WT retinas, blue if it was decreased in Opa1^enu/+). Similarly, Figure 2E shows the summary for the temporal tuning differences between Opa1^enu/+ and WT RGCs. The underlying raw data for the P values depicted in Figures 2D and 2E are shown in Supplementary Figures S1 through S8. It is striking that most differences between WT and Opa1^enu/+ response properties only emerge in the oldest age group (18 months). Secondly, the specific difference in spatial and temporal tuning was not fixed, but depended on the stimulus conditions. Under scotopic conditions, responses to gratings with small spatial periods were enhanced in old Opa1^enu/+ retinas compared with WT (Fig. 2C left, red squares in Fig. 2D), while the responses did not differ at the other spatial scales. Under photopic conditions, on the other hand, the responses of WT RGCs were stronger than in Opa1^enu/+ retinas at medium spatial scales (Fig. 2C right, blue squares in Fig. 2D). Similarly to spatial tuning, we found opposite effects under scotopic and photopic stimulus conditions for temporal frequency tuning: While old Opa1^enu/+ RGCs tended to respond more strongly to 4-Hz stimuli under scotopic luminance conditions, RGCs in WT retinas had stronger responses under photopic conditions (Fig. 2E). 

Sustainedness Is Increased in Old Opa1^enu/+ RGCs.

We next measured the sustainedness of RGC responses during full-field contrast steps stimuli (Fig. 3A₁). In general, 12- and 18-month Opa1^enu/+ RGCs had more sustained responses than WT cells, independent of the luminance level (Fig. 3B, left). Under scotopic conditions, we found differences already at the age of 6 months and with opposing tendency: There, WT RGCs displayed more sustained responses (population raw data in Supplementary Fig. S9). 

Latency and Speed Tuning Are the Same in WT and Opa1^enu/+ RGCs.

Latency and speed tuning did not differ between WT and Opa1^enu/+ RGCs, independently of age and luminance (Fig. 3B middle and right; population raw data in Supplementary Figs. S10, S11). As an exception we found a slightly higher latency in Opa1^enu/+ RGCs at the age of 12 months under mesopic conditions (P = 0.049). 

Aging Affects Responses of WT and Opa1^enu/+ Retinas Differently

In the previous section, we focused on differences between the two genotypes. In the following, we address the age-dependent development of RGC responses within each genotype. Synaptic input measurements in mouse RGCs suggest that 1-month retinas might not yet be fully developed functionally.³⁰ We thus decided to compare data from 6- (“young”) and 18-month (“old”) animals within each genotype group. 

Latency and Speed Tuning Are Only Affected by Age, But Not by the Opa1 Mutation.

In accordance with the finding that latency and speed tuning did not differ significantly between WT and Opa1^enu/+ retinas, we found similar aging effects in both genotypes for these parameters. Latency increased in old animals under scotopic conditions and, in Opa1^enu/+ retinas, also under mesopic conditions (P = 0.045). Speed tuning tended to drop in old retinas under scotopic conditions, significantly in old Opa1^enu/+ retinas (WT: P = 0.072; Fig. 4). 

Figure 4

 

Aging effects in WT and Opa1^enu/+ retinas. (A) Aging effects in Opa1^enu/+ retinas. P values obtained with Wilcoxon rank sum tests on parameters obtained at the age of 6 and 18 months. Latency increased with age under scotopic and mesopic conditions. Old Opa1^enu/+ RGCs were tuned to slower speeds under scotopic conditions. Further, in Opa1^enu/+ retinas, aging lead to increased activity modulation in response to sine wave gratings with small periods under scotopic and mesopic conditions, while responsiveness was decreased to large periods. Temporal frequency tuning decreased with age under scotopic (8 Hz) and photopic (1 Hz) conditions, and sustainedness increased in old Opa1^enu/+ retinas under all conditions. (B) Aging effects in WT retinas. Response latency increased and sustainedness decreased in old WT RGCs under scotopic conditions. Moreover, in old WT retinas, modulation of RGCs activity tended to decrease for all spatial periods under scotopic conditions (significant for 500 μm). Similarly to Opa1^enu/+ RGCs, responsiveness to higher frequency stimuli was decreased in old WT retinas under scotopic conditions, while at photopic luminance levels responsiveness to 8-Hz stimuli was increased.

Spatial and Temporal Tuning Is Affected by Age in Opa1^enu/+ Retinas.

Under scotopic conditions, the responses of Opa1^enu/+ RGCs became weaker with age when exposed to drifting gratings with large spatial periods, but response strength increased for gratings with small spatial periods (Fig. 4A). This increase was more pronounced under mesopic conditions. On the other hand, age had very little effect on WT RGCs (Fig. 4B). These different age-dependent response characters can explain the differences between the genotypes observed at 18 months (Fig. 2D). Temporal frequency was affected similarly by age in Opa1^enu/+ and WT retinas under scotopic conditions. However, at photopic levels, aging induced a drop in response strength to 1-Hz stimuli in Opa1^enu/+ retinas (Fig. 4A), while response strength for 8-Hz stimuli increased in WT retinas (Fig. 4B). 

Sustainedness Increases With Age in Opa1^Enu/+ Retinas.

Sustainedness of RGC responses increased with age in Opa1^enu/+ retinas under all luminance conditions (Fig. 4A). In WT retinas we detected a drop in sustainedness under scotopic conditions (Fig. 4B). 

Melanopsin Is Not Per Se Neuroprotective

Melanopsin-expressing RGCs were found to be spared from degeneration provoked by different insults,^31–33 including ADOA.^3,19 We therefore examined if melanopsin-expressing RGCs are also spared in our Opa1 mouse model and investigated a potential neuroprotective role of the melanopsin protein (representative immunohistochemistry for melanopsin is shown in Fig. 5A). 

Figure 5

 

Stable number of melanopsin-expressing and β-galactosidase reporter expressing cells per retina in Opa1^enu/+ mutants. (A) Melanopsin immunofluorescence in whole-mount retinas. Representative examples (contrast-enhanced in Adobe Photoshop for clarity) are shown for 18-month Opa1^enu/+ (left) and WT retinas (right). Scale bar: 20 μm. (B) Melanopsin-expressing cells. The number of immunopositive cells does not change with the variable “age” or “genotype” with an average of 1427 ± 16 immunopositive cells per retina (ANOVA test, P > 0.05). (C) β-galactosidase positive cells. The number of β-galactosidase–positive cells per retina does not change within the different genotypes (ANOVA test, P > 0.05). Scheffe post hoc test does not show significant differences between Opa1^enu/+ × Opn4^{taulacZ/taulacZ} and WT × Opn4^{taulacZ/taulacZ} groups (Scheffe post hoc test, P > 0.05). Additionally, no differences were identified when Opa1^enu/+ × Opn4^+/taulacZ was compared with the previous groups (Scheffe post hoc test, P > 0.05).

As shown in Figure 1, there is a significant decrease in the number of axons in the 18-month Opa1^enu/+ mice. However, we found the number of melanopsin-expressing cells in retinal whole-mounts to remain stable (Fig. 5B). One-way ANOVA test demonstrated no influence of the genotype or age on the number of immunopositive cells (P > 0.05). To determine whether the melanopsin-expressing cells are protected against Opa1 mutation–induced degeneration due to the presence of the melanopsin protein per se, we compared the number of β-galactosidase immunopositive cells in retinal whole-mounts. No significant difference in the number of immunolabelled cells was observed between the different genotypes, including the double heterozygote in which both melanopsin and β-galactosidase are expressed (ANOVA test, P > 0.5; Fig. 5C). Thus, these results suggest that the resistance to neurodegenerative insults of these cells is not dependent on the presence of melanopsin. 

Discussion

Loss of Optic Nerve Axons in the B6;C3-Opa1^329-355del Model for ADOA

In the present work we have used the B6;C3-Opa1^329-355del mouse model, which has a comparable phenotype with that observed in patients with OPA1-associated nonsyndromic ADOA.^13,29 By counting axons in the optic nerve, we observed an approximately 50% age-related axonal loss in the Opa1^enu/+ 18-month animals when compared with WT 18 month, which is in agreement with previous studies.^13,34 However, the axonal loss was quite variable from mouse to mouse, ranging from 20% to 75%. This finding corresponds with the pronounced clinical variability observed in ADOA patients.^4,35 

Optic nerve axons can be classified depending on their morphometric characteristics. We have observed that axonal loss tends to be biased to axons with small cross-sectional areas (up to 0.4 μm²) in the aged Opa1^enu/+ animals, representing 95% of the total axonal loss. Such loss of small caliber fibers has also been observed previously in this mouse model, using electron microscopy to classify and count axons in optic nerve cross sections,¹³ and in other optic neuropathies,³⁶ indicating that the axon size may be highly correlated with its vulnerability. 

No differences between Opa1^enu/+ and WT animals were observed at 1 month of age, indicating that no early onset of axonal loss occurs in this mouse model; however, this does not exclude potential subcellular changes. Additionally, we have not observed any significant changes in the number or morphology of optic nerve axons between 1- and 18-month-old WT animals. This is in agreement with previous studies,¹³ while other publications have shown an age-related decrease in the number of RGCs.^37,38 These differences might be due to a background effect as well as environmental factors. 

Aging Affects Functional Properties of RGCs

Based on MEA-recordings, we found two major functional changes in aging retinas, which were similar in WT and Opa1^enu/+ retinas. First, old RGCs seem to react more slowly (i.e., they had a longer latency), most pronounced under scotopic conditions. Further, in both genotypes we could detect a tendency to slower speed tunings. On the other hand, spatial and temporal frequency tuning as well as sustainedness were affected differently in ganglion cells of Opa1^enu/+ and WT retinas. 

Opa1 Mutation Changes Ganglion Cell Output: Are Upstream Neurons Affected?

We wondered whether the pronounced loss of RGC axons in older Opa1^enu/+ mice represents a general loss of all RGC types, or if we could find a hint for specific loss of certain functional cell types. If certain RGC types were more affected than others by the Opa1 mutation, this may be reflected in a disappearance of cells in a certain region in the functional parameter space (e.g., no or few transient cells in old Opa1^enu/+ retinas). It has been shown that cells with small axon diameter comprise a variety of morphologically, and presumably functionally, different RGC types.¹⁵ Consequently, we could not detect a clear drop-out of a specific functional cell type although we found differences in spatial and temporal frequency tuning, and sustainedness between the two genotypes, mostly in old (12- and 18-month) animals. This does not come into conflict with our histologic findings of a specific loss of small-axon RGCs. 

Instead of a clear drop-out of a functional type, we detected various shifts in the measured physiological parameters. This may be explained by two mechanisms, which are not mutually exclusive. First, although cell death is not limited to a specific cell type, it might still affect certain functional types more than others. Second, the output of the surviving RGCs might be altered by the Opa1 mutation and/or secondary to the loss of RGCs. While we can neither confirm nor refute these possibilities, our data actually hints at yet a third explanation, namely that there are alterations not only of the surviving RGCs but also of the remaining upstream retinal circuitry. If differences were only due to RGC death or changes in their properties, we would expect the differences to be similar under different luminance conditions. However, we found very different tendencies under scotopic, mesopic, and photopic conditions (e.g., spatial tuning of 18-month-old RGCs under scotopic versus photopic conditions, Fig. 2D). Such luminance-dependent changes suggest additional alterations in the retinal processing (e.g., in homeostasis, synaptic connections, neurotransmitter release, etc. of upstream neurons [bipolar cells, amacrine cells, photoreceptors]). So far, RGCs have been thought to be the only retinal cells affected in OPA1 deficiency.³⁹ However, previous publications support possible involvement of the outer retina: Heiduschka and colleagues²⁹ found that the amplitudes of scotopic but not of photopic visually evoked potentials in old Opa1^enu/+ mice were reduced significantly, which may indicate functional impairment of the rod driven signaling circuitry. Moreover, Reis and colleagues⁴⁰ recently reported significantly reduced multifocal ERG amplitudes in a cohort of ADOA patients with defined OPA1 mutation. 

Effect of Opa1 Deficiency on the Melanopsin System and Melanopsin's Potential Protective Role

Within the population of RGCs there is an outstanding cell subtype, the intrinsically photoresponsive RGCs (ipRGCs). These cells express the photopigment melanopsin and are responsible for nonimage forming vision.¹⁸ In the last years, melanopsin cells have been shown to be resistant to different retinal insults such as axotomy³² and glaucoma,³³ while the nonmelanopsin cells are massively affected. Furthermore, previous publications pointed out that melanopsin-expressing RGCs are also spared in ADOA in human and in the B6;C3-Opa1^Q285STOP mouse model.^3,19 However, just one patient was available for the study in humans, and the B6;C3-Opa1^Q285STOP mouse model presents with a limited loss of RGCs.¹² With our B6;C3-Opa1^329-355del mouse model, in which there is an extensive loss of RGCs (∼50% reduced number of RGC at 18 months), we could confirm that no loss of melanopsin-expressing RGCs occurs, indicating that they are resistant. Additionally, no influence of age was observed on the survival of melanopsin-expressing RGCs consistent with previous reports.^41,42 Why are melanopsin-expressing cells more resistant than other RGCs? A main difference between the melanopsin-expressing cells and the other RGCs is the presence of the melanopsin protein per se. It has been suggested that blue light might be damaging to RGCs due to the generation of reactive oxidative species, especially in case of mitochondria dysfunction, as in ADOA.²⁰ Since melanopsin absorbs blue light this might exert a protective effect against this light insult.⁴³ To test this hypothesis we studied the Opn4^tauLacZ mouse mutant in which the melanopsin gene is replaced by sequences coding for a tau-lacZ fusion protein.⁴⁴ If melanopsin itself would have a protective effect one would expect a reduced number of β-galactosidase positive cells in Opa1^enu/+ × Opn4^{taulacZ/taulacZ} mice. However, we found similar numbers of immunopositive cells in WT × Opn4^{taulacZ/taulacZ} and Opa1^enu/+ × Opn4^{taulacZ/taulacZ} animals, indicating that melanopsin does not have a protective role at least in the context of Opa1 deficiency. It should be noted that β-galactosidase expression can only be identified in the M1 cells.⁴⁵ Therefore, we cannot exclude a potentially protective role of melanopsin in the other melanopsin-expressing RGC subtypes (M2–M5).^45–47 Studies using the Opn4-EGFP mouse line,⁴⁸ in which the M1, M2, and M3 melanopsin-expressing cells can be identified, would allow a more detailed view of the potentially protective function of the melanopsin protein. 

As a further difference, melanopsin-expressing cells, but no other RGC group, express pituitary adenylate cyclase-activating peptide (PACAP).⁴⁹ Pituitary adenylate cyclase-activating peptide has been shown to have neuroprotective potency.^50–52 Additionally, melanopsin-expressing cells may have different mitochondrial dynamics due to their intrinsic photosensitivity, which may protect them against cell stress. Furthermore, no details on the axonal morphology of the melanopsin-expressing cells are known, so that we cannot relate melanopsin-expressing cells to our axonal diameter measurements. 

In conclusion, our study demonstrates that the axonal loss observed in the B6;C3-Opa1^329-355del mouse model for human ADOA is mutation- and age-dependent, with 50% of axons lost by 18 months. Moreover, we have shown that the RGC loss is biased toward RGCs with small axonal cross-sections. Our data reveal significant differences in the visual response properties between old Opa1^enu/+ and WT RGCs, but not a drop-out of a certain functionally defined RGC population. However, we found that differences between the two genotypes have different trends under scotopic and photopic conditions, suggesting that there are also alterations of the pre-RGC retinal circuitry. We thus suggest for future studies to more closely investigate presynaptic retinal neurons. Functional evaluation of RGCs can only be conclusive if changes in upstream neurons are either excluded or, if present, integrated with RGC measurements. 

Our study also verified that the melanopsin system is neither affected by ageing nor in Opa1-induced RGC degeneration. Moreover, our results show that the melanopsin protein is not the factor that imposes the resistance of these intrinsically photoresponsive cells. Several other factors may be implicated in the survival of these cells and future studies are required to elucidate this question, which, if solved, may offer some therapeutic approaches. 

Acknowledgments

Supported by grants from the Deutsche Forschungsgemeinschaft (DFG; Bonn, Germany) to the Werner Reichardt Centre for Integrative Neuroscience (DFG EXC 307), by the Bundesministerium für Bildung and Forschung (“Bernstein Center for Computational Neuroscience,” FKZ 01GQ1002 [TAM], and E-Rare “ERMION,” FZ 01GM1006 [BW], and a Stipend of the Pro Retina Foundation [KR]; Frankfurt, Germany). 

Disclosure: I. Gonzalez-Menendez, None; K. Reinhard, None; J. Tolivia, None; B. Wissinger, None; T.A. Münch, None