An Optical Coherence Tomography Study on Degeneration of Retinal Nerve Fiber Layer in Rats with Autoimmune Optic Neuritis

Abstract

Purpose.: The aim of the present study was to evaluate the ability and accuracy of spectral domain optical coherence tomography (OCT) for in vivo monitoring of retinal ganglion cell degeneration in a rat model of myelin oligodendrocyte glycoprotein-induced optic neuritis.

Methods.: First, OCT imaging was established for imaging of all retinal layers in Brown Norway rats. Second, thickness measurements of retinal nerve fiber layer (RNFL) were performed by periodically imaging during the development and progression of autoimmune optic neuritis. Third, the reproducibility of OCT measurements was determined by comparing RNFL measurements of two independent investigators. Fourth, OCT data were correlated with histopathology obtained ex vivo after the final imaging session.

Results.: Results showed that RNFL thickness declined significantly before clinical manifestation of the disease and decline progresses continuously during the disease course. RNFL thickness measured by OCT had good repeatability and also corresponded with histomorphometric measurements. The reproducibility was limited because of the post-processing analyses performed by manual measurements.

Conclusions.: In summary, it is shown here for the first time that OCT can reliably monitor neurodegeneration in an experimental model of autoimmune optic neuritis in rodents. Moreover, in comparing RNFL thickness decline with histopathological analyses of the optic nerve, these results suggest an early, and in part, inflammation-independent process of RNFL degeneration in autoimmune optic neuritis.

Neuronal cell death and axonal degeneration are now well-recognized pathologic features of multiple sclerosis (MS) and are likely the main reason for permanent disability. ^1,2 As neurodegeneration in the early stages of MS is difficult to detect and brain atrophy assessed by conventional magnetic resonance imaging (MRI), which is the preferential end point in clinical trials, is confounded by demyelination and astrogliosis, novel imaging parameters are needed to assess axonal injury. Optical coherence tomography (OCT) is a technology that uses reflexion patterns of infrared light to quantify retinal nerve fiber layer (RNFL) in real time with extremely high temporal accuracy. ³ Originally developed for retinal diseases and glaucoma, OCT demonstrated a significant reduction of mean RNFL thickness in patients with optic neuritis. ^{4 –6} In the past years, OCT has been improved by dramatic technological advances and enables monitoring of retinal degeneration in rodent eyes. ^{7 –9} However, no studies on evaluation of RNFL thickness by OCT have been performed in experimental models of optic neuritis so far. 

In the present study we used experimental autoimmune encephalomyelitis (EAE) induced by myelin oligodendrocyte glycoprotein (MOG). Previously, we demonstrated that MOG-EAE induced in female Brown Norway rats strongly reflects the neurodegenerative aspects of MS. In this model, 80%–90% of animals develop severe optic neuritis, which leads to acute axonal degeneration of the optic nerve and apoptosis of retinal ganglion cells (RGCs), the neurons that form its axons. ^10,11 Therefore, this model provides a unique opportunity to investigate the morphologic changes within the retina as a consequence of acute optic neuritis. The purpose of the present study was to ascertain the ability and accuracy of OCT to detect degeneration of the RNFL in autoimmune optic neuritis. 

Material and Methods

Rats

Female BN rats 8 to 10 weeks of age were used in all experiments. They were obtained from Charles River (Sulzfeld, Germany) and kept under environmentally controlled conditions without the presence of pathogens. All experiments that involve animal use were performed in compliance with the relevant laws and institutional guidelines. These experiments have been approved by the local authorities of Braunschweig, Germany, and all animals were treated in accordance with the ARVO Statement for the Use of Animals in Ophthalmic and Vision Research. 

Induction and Evaluation of EAE

The rats were anesthetized by inhalation anesthesia with methoxyflurane (Metofane; Pitman-Moore, Mundelein, IL) and injected intradermally at the base of the tail with a total volume of 100 μL inoculum, containing 50 μg rrMOG^Igd kindly provided by Doron Merkler (Department of Neuropathology and Immunology, Geneva, Switzerland) in saline emulsified (1:1) with complete Freund's adjuvant (CFA) (Sigma, St. Louis, MO) containing 200 μg heat-inactivated mycobacterium tuberculosis (strain H 37 RA; Difco Laboratories, Detroit, MI). Rats were scored for clinical signs of EAE and weighed daily until Day 14 of EAE (end of the study). Day 1 of EAE was defined to be the day when the first motor symptoms were detected. The signs were scored as described previously. ¹² This score reflects the amount of spinal cord lesions and does not include visual symptoms or correlate with the severity of optic neuritis. ^13,14  

OCT Technology and Imaging Procedure of RNFL

Animals were anesthetized with intraperitoneal injection of ketamine 10% (0.65 mL/kg; Inresa, Freiburg, Germany) and xylazine 2% (0.35 mL/kg; Albrecht; Aulendorf, Germany). The pupils were dilated with tropicamide (5 mg/mL; Mydriaticum Stulln; Pharma Stulln, Stulln, Germany) and animals were placed on a custom-made platform, positioned below the fixed OCT probe, set in a 30° position, which ensured that the incident OCT beam was perpendicular to the cornea. Additionally, the visualization of the eye was performed through a color digital camera (Thorlabs HL-AG; Lübeck, Germany) during alignment and centering. The image acquisition was controlled with custom-made software (based on Labview software and developed by Thorlabs HL-AG, Lübeck, Germany), which displays the OCT image in real time and the color camera image of the eye. As a coupling fluid, ultrasound gel was applied both to the cornea and to the OCT probe, which prevented cornea dehydration and cataract formation and reduced friction between OCT lens and the eye. Cross-sectional imaging of the retina was performed using a spectral radar OCT instrument (OCT900SR-HR; Thorlabs Inc., Newton, NJ) based on the detection of optical path differences, incorporating a broadband light source with a high-speed spectrometer. Optical properties of the retinal sample were determined by analyzing the back-reflected and scattered light from an illuminated retinal area. The light of a broadband low-coherence light source (center wavelength 930 nm) was guided to the sample with a small focus to get an illuminated volume with a small lateral dimension (2 or 4 mm). The OCT imaging engine consisted of 1000 lines/mm transmission grating and a 16-bit line-scan charge-coupled device camera (Thorlabs Inc.). By the imaging speed of 1100 axial scans per second (1.1 kHz) the complete raster scan, consisting of 50 scanning steps, took approximately 45 seconds. At this operating condition, the measured sensitivity was approximately 95 dB. The calibrated axial resolution was 6.36 μm in the air and approximately 4.7 μm in the tissue. The interference spectrum was converted from wavelength to frequency and axial scan data were obtained by fast Fourier transformation. The whole procedure, including the initial alignment of the animal, took approximately 5 to 10 minutes per eye. During this time, ultrasound gel was reapplied occasionally. None of the animals developed cataracts due to corneal dehydration. 

To monitor retinal degeneration during optic neuritis, animals were separated into four groups containing five animals each and OCT scan was performed periodically at Day 7 post immunization (p.i.), at the day of disease onset (EAEd1) and at days 8 (EAEd8) and 14 of EAE (EAEd14). 

RNFL Thickness Measurements

The raw OCT images were first saved as bitmap files and then quantitatively analyzed by software developed by Thorlabs (Lübeck, Germany) on a commercially-available software package (Labview). Thickness measurements of the RNFL were performed manually in the region of interest, which was set to be 500 μm temporal from the center of the optic disc (approximately 1 optic disc diameter from the edge of the optic disc) ^{15 –17} (Fig. 1). On the proximal surface of the retina the nerve fiber layer (NFL) and ganglion cell layer (GCL) appear together as a thin band with strong light scattering and cannot be distinguished from each other. Thicknesses were calculated using a group refractive index of 1.4. In areas where a major blood vessel interrupted RNFL tissue, the RNFL thickness was determined as the average of the RNFL thickness measurements from the two adjacent points as described previously. ⁷  

Figure 1.

 

Measurements of RNFL thickness were performed manually in the region of interest (ROI) (500 μm temporal or nasal from the center of the optic disc). One cursor (orange line, arrow 1) was placed at the center of the optic disc, while the second cursor (orange line, arrow 2) was set at a lateral distance of 500 μm from the first cursor. RNFL thickness was determined as the distance (double arrow 3) between the two A-scan graphs (green and magenta lines) which correspond to the upper and lower boundary of the RNFL, respectively. Positioning of the two A-scan lines was confirmed by the position of the intensity peak in the amplitude panel which shows the dynamic range of displayed OCT image in dB, because the intensity peak corresponds to the RNFL which has strong light scattering.

Histopathological Processing of the Eyes and Optic Nerves

Rats were euthanized at different time points (Day 7 p.i., EAEd1, EAEd8, and EAEd14). Both eyes as well as both optic nerves of each animal were included in the analysis. Eyes used for histologic investigation were marked with sutures, one in the superior rectus muscle and two at the medial rectus, to facilitate orientation for histologic processing. After enucleation, the eyes were stored in 4% phosphate-buffered glutaraldehyde for 1 hour and transferred to 10% phosphate-buffered paraformaldehyde (PFA) until paraffin-embedding. Serial 4-μm thick sections were cut vertically through the center to the peripheral retina. Sections were then stained for hematoxylin-eosin (HE), and ED1-staining (Serotec; Düsseldorf, Germany; diluted 1:500). For comparison of histopathology images and OCT images, HE stained retinal sections were observed under an optical microscope (Axioplan2; Zeiss, Jena, Germany). For each eye, RNFL thickness was measured with image analysis software (Axiovision 4.2, Zeiss) at 500 μm distant from the center of the optic disc. 

Histologic evaluation of the optic nerves was performed on paraformaldehyde-fixed, paraffin-embedded longitudinal sections of optic nerve obtained at Day 7 p.i., EAEd1, EAEd8, and EAEd14 as well. To assess demyelination, axonal pathology and inflammation, 0.5 μm sections were stained with Luxol-fast blue, Bielschowsky silver impregnation, and ED1 staining as described earlier. ¹⁸ The density of axons in each optic nerve (ON) was measured in at least 3 standardized microscopic fields of 2500 μm² close to the lamina cribrosa, in the middle and in the distal part of the ON. Mean axon density was calculated for each ON and compared with the density of healthy controls. The independent investigator who performed neuropathological examinations was blinded to the OCT data. 

Statistical Analyses

Data are presented as mean ± SEM. One-way ANOVA followed by Bonferroni correction was used for multiple group comparison to asses inflammatory infiltration, demyelination, axonal densities within the ONs, and differences in RNFL thickness at different time points in different groups. Histomorphometric measurements were compared with OCT measurements of RNFL thickness, detected at the final OCT session using Pearson′s correlation coefficient. The correlation between OCT-determined RNFL thickness and axonal density was assessed by a regression model. To account for a possible correlation between the two measurements (eyes) on the same animal a generalized estimating equation–approach was used. A P value < 0.05 was considered to be statistically significant. 

The intra-reader variability was determined by calculating the RNFL thickness difference in images acquired from three scans in the same eye during the same OCT session. The interreader reproducibility was determined by using RNFL thickness difference evaluated by two independent investigators following a blind protocol. Therefore, two-way ANOVA (SAS 9.2, SAS Institute Inc., Cary, NC) was used. 

Results

Clinical Symptoms

The different groups (n = 5 each) were clinically monitored until Day 7 after immunization as well as until Days 1, 8, and day 14 of the disease. The clinical manifestation of MOG-EAE was at Day 13.6 ± 1.4 (mean ± SEM) after immunization. The symptoms ranged from a mild paresis of the tail to a hind limb paralysis. In the disease group running until Day 14 of EAE four out of five animals remained stable over the disease period, and one of five animals showed disease progression from EAEd5 onward. 

Intra- and Interreader Reproducibility of RNFL Thickness Measurements by OCT

Intrareader Variability.

The intra-reader variability was analyzed by descriptive methods. The mean ± SD for the variability between the three measurements was 0.71 ± 0.42 μm. This amount of variability is relatively small compared with the total SD (5.53 μm). Hence only a negligible proportion of variation can be traced back to intrareader variability. 

Interreader Variability.

To assess the interreader variability, a second reader performed an independent evaluation of the RNFL thickness on the saved scans of all the different animal groups in a blinded manner. Independent from the disease stage investigated, reader 2 observed significantly smaller results than reader 1 (2.92 μm ± 0.42; P < 0.0001; Fig. 2). However, the decline of RNFL thickness during the whole disease course was consistently detected by both readers: no significant interaction between time point and reader could be detected (P = 0.47). Hence the two readers worked on different levels but showed qualitatively comparable results. 

Figure 2.

 

Interreader variability of RNFL thickness measurements performed on the saved scans by two independent investigators in a blinded manner. The two readers observed significantly different results (P < 0.0001). However, the decline of RNFL thickness over time was consistently assessed independent from investigator or investigation time point.

Monitoring of RNFL Thickness Changes in Autoimmune Optic Neuritis

To monitor RNFL degeneration during the disease course of MOG-induced optic neuritis, we performed repeated imaging of the same animal (before immunization, at Day 7 p.i., on EAEd1, EAEd8, and EAEd14). These time points were selected because of the rapid kinetics of neurodegeneration in our model with first signs of RGC degeneration detectable already during the induction phase. ¹¹ Analyzing the cross-sectional OCT images, mean RNFL thickness was 37.72 ± 0.43μm (for reader 1, and 34.72 ± 0.53 for reader 2) in animals at baseline. Reduction of RNFL thickness was found already in the early stage of the disease development: at Day 7 p.i., the mean layer thickness was reduced by 4.75 ± 0.58 μm (P < 0.0001 compared with the baseline value). At the day of clinical manifestation and during the further disease course, the reduction of the mean RNFL layer thickness continuously progressed (by 7.27 ± 0.61 μm, 9.97 ± 0.53 μm, and 13.14 ± 0.72 μm at EAEd1, EAEd8, and EAEd14, respectively; P < 0.0001; for all investigated time points; Fig. 2). During the OCT procedure no changes in the anterior segment of the eye indicating concomitant ocular disease were observed. 

RNFL Thickness Determined by OCT versus Histomorphometric Analyses

To determine whether RNFL thickness obtained by OCT is consistent with the actual RNFL thickness, RNFL thickness in areas corresponding with those assessed by OCT was measured on retinal HE-stained cross sections (Fig. 3). On average, data obtained from histomorphometric analysis showed smaller RNFL thickness in all eyes compared with the OCT measurements, which can be explained by the tissue shrinkage occurring during the fixation procedure. However, in accordance with OCT data, the histomorphometric measurements showed a reduction of RNFL thickness during the course of the disease. The mean values were 28.71 μm ± 0.46; 26.75 μm ± 0.32; 24.48 μm ± 0.37; and 20.68 μm ± 0.52; for Day 7 p.i., EAEd1, EAEd8, and EAEd14 respectively. Also, RNFL thickness measured by OCT correlated significantly with the thickness obtained from retinal sections at different investigation time points (r = 0.74; P < 0.0001; Fig. 4). 

Figure 3.

 

Comparison of OCT images at baseline and respective histologic sections (same retina each). The arrows in (A) and (B) indicate area of interest, which was set at 500 μm from the center of the optic disc. The retinal layers can be sufficiently distinguished to allow reliable quantitative analysis of each individual layer. Higher magnification of the OCT image (C) and respective histologic section (D). INL, inner nuclear layer; IPL, inner plexiform layer; IS, inner segment of photoreceptor layer; ONL, outer nuclear layer; OPL, outer plexiform layer; OS, outer segment of photoreceptor layer. Scale bars, 200 μm.

Figure 4.

 

Correlation of RNFL thickness determined by OCT and histology. Pearson′s correlation coefficient analysis shows a significantly positive relationship between RNFL thickness detected by OCT and histologic analysis (r = 0.74, P < 0.0001).

Histopathologic Analysis of ONs during Development of MOG-EAE

Serial ON longitudinal sections were evaluated at the same time points at which OCT measurements were performed. ONs were stained with Luxol fast blue (Figs. 5A–C), ED1 (Figs. 5D–F) and Bielschowsky silver impregnation (Figs. 5G–I) to assess demyelination, inflammation, and axonal pathology, respectively. Evaluating Luxol-fast blue staining, we calculated the percentage of the demyelinated area with respect to the whole area of every ON longitudinal section and found no abnormalities at Day 7 p.i. Also inflammatory infiltration could not be detected at this time point. The degree of demyelination (4.39 ± 1.03%) and the mean score of inflammation (0.67 ± 0.12; mean ± SEM) at Day 7 p.i. was not significantly different when compared with the optic nerves of sham-immunized animals at corresponding time points (data not shown). During the disease course, at Days 1, 8, and 14 of MOG-EAE, the extent of demyelination increased substantially (39.74 ± 10.38%, 43.38 ± 9.5%, and 45.64 ± 10.49%, respectively) and were consistent with inflammation (scores of 1.8 ± 0.37, 2.00 ± 0.45, and 2.67 ± 0.17 at Days 1, 8, and 14 of MOG-EAE, respectively). Analyzing axonal damage, we found a tendency toward lower axon counts at Day 7 p.i. when compared with healthy animals (67.26 ± 4.0%; mean ± SEM), indicating that axonal damage within the ONs occurs in part independent of apparent inflammation and demyelination. After manifestation of the disease, we observed significant axonal damage within the ONs (45.4 ± 7.08%, mean ± SEM). Until day 14 of EAE, a further reduction of axonal density occurred (34.17 ± 7.04% for EAEd8, and 30.99 ± 5.01% for EAEd14). Axonal density showed significant correlation with OCT-determined RNFL thickness (P < 0.0001; data not shown). 

Figure 5.

 

Histopathological analysis of representative longitudinal ON sections to assess demyelination, inflammatory infiltration, and axonal damage during development of optic neuritis. (A–C) Demyelination within ON longitudinal sections detected by Luxol fast blue staining at Day 7 after immunization (A), at Day 1 (B), and at Day 14 of the disease (C). At the latest time point, the ON was nearly completely demyelinated (C); scale bars, 50 μm. (D–F) ED1 staining to determine inflammation at Day 7 p.i. (D), Day 1 (E), and Day 14 of the disease (F). At the early time point, no significant inflammatory infiltration was found (D), whereas the optic nerve was heavily infiltrated at Day 14 of MOG-EAE (F); scale bars, 50 μm. (G–I) Bielschowsky silver impregnation to assess axonal pathology. Normal axonal morphology was observed within the ONs at Day 7 after immunization (G). At Day 14 of MOG-EAE (I), only a few remaining injured axons between inflammatory cells were detected; magnification, ×63.

Discussion

In a rat model of autoimmune optic neuritis, we used OCT to monitor the kinetics of RNFL thickness reduction during disease development and progression. The gradual decrease in RNFL thickness we found reflects atrophy of the proximal, nonmyelinated parts of RGC axons. To correlate in vivo imaging results with tissue changes ex vivo, we combined OCT measurements with histomorphometric analysis of the retina. Additionally, the severity of optic neuritis was assessed by histopathologic evaluation of the ONs at different stages of the disease. The results of this study demonstrate that OCT is a sensitive and reliable in vivo imaging tool which can be used in rodent models of optic neuritis to assess changes in RNFL thickness already before clinical manifestation of the disease and during the further disease course. 

Whereas OCT became a well established technique for monitoring neurodegeneration in patients with MS or autoimmune optic neuritis, ¹⁹ OCT imaging in rodent models is challenging because of the small size and thin retina compared with humans. In previous studies, conventional time-domain OCT of rodent eyes has been used to assess retinal degeneration. ^20,21 However, the axial resolution of the used OCT was not sufficient for evaluating RNFL thickness in rodent eyes. Recently, spectral domain OCT became available, enabling improvement in image acquisition speed and image resolution. This new technology allowed visualization of all major intraretinal layers in rat and mouse models. ^8,22 However, to our best knowledge, no in vivo OCT imaging has been performed in a rodent model of optic neuritis. 

In our recent study, we used spectral domain OCT to monitor changes in RNFL in MOG-induced optic neuritis. MOG-EAE is the principal model of MS mimicking the whole histopathological spectrum of the human disease. ²³ The characteristic feature of this disease model is that around 14 days after immunization animals develop optic neuritis followed by acute axonal damage within the optic nerve and apoptosis of RGCs, the neurons that form its axons. ¹⁰ Our previous observation showed that apoptotic neuronal cell death occurs early and progresses rapidly during the disease course of EAE and is in part independent from axonal damage. ¹¹ In our present study, no histopathological abnormalities were found in the longitudinal sections of the ON at day 7 after immunization. However, at this time point, a significant decline in RNFL thickness was detected by OCT. During the further course of EAE, BN rats developed severe autoimmune optic neuritis as observed by histopathological analysis of the ON sections. Also a gradual decrease of RNFL thickness occurred at different investigation time points as detected by OCT. However, the previously described RGC count reduction observed in the enucleated retina lies in a higher range compared with RNFL thickness decline detected in our OCT measurements. In our previous studies, animals had already lost 54% of their RGCs at the onset of clinical symptoms and no further substantial reduction of RGC counts was observed beyond Day 8 of the disease. ^11,14,18 In contrast, as detected by OCT, RNFL thickness declined by 21% at the time point of first clinical symptoms in our recent study. Moreover, we observed a further decline of RNFL thickness on Day 14 compared with Day 8 of EAE. Therefore, our data show that RGC loss has more rapid temporal kinetics and appears to precede the reduction of RNFL thickness suggesting that the process of neurodegeneration starts in RGC bodies and secondly involves the proximal parts of their axons. Our results are in accordance with findings obtained from studies in optic nerve trauma models, ^7,24 which show a similar kinetics of RGC loss occurring in MOG-EAE. However, the mechanism of this intriguing dissociation of time courses between loss of neuronal cell body and intraretinal axons during autoimmune inflammation remains unclear. 

Several studies exist describing the assessment of RNFL thickness by different OCT technologies in human diseases as well as in experimental disease models. In humans, although spectral domain OCT showed good agreement with time domain OCT, differences in the results of RNFL thickness measurements were observed. ²⁵ Compared with the results from ON trauma models in rodent and primates studies ^7,26 our data show a higher range of RNFL thickness values at baseline. These differences can be explained by different rat strain and OCT set-ups used in those studies. Previous studies in humans reported that a variability of RNFL thickness measurements lies within 5 to 6 μm depending on the OCT technology used (stratus versus fourier domain). ^27,28 In our study, RNFL thickness measurements by OCT had good repeatability (<1 μm) at all the different investigation time points. Moreover, RNFL thickness detected in vivo by OCT showed significant correlation with histology. The decline of RNFL thickness during the disease course as well as their extent was confirmed by two independent investigators. However, the interreader data showed high variability indicating that postprocessing analysis by manual measurements involves subjective bias and therefore is the major limitation of this study. 

Based on our previous studies evaluating various techniques for in vivo monitoring of disease processes in the retina and the optic nerve (e.g., electrophysiology, MRI, and calcium imaging) the shortcomings of each method are evident. ^17,29,30 MRI and electrophysiological techniques cannot reliably distinguish between demyelination and axonal damage. Because the RNFL is composed largely of nonmyelinated axons of RGCs, OCT allows for axonal loss to be measured independently of other confounding disease factors. As autoimmune optic neuritis can be associated with concomitant manifestation of several forms of ocular disease, ^31,32 we have demonstrated that OCT is a sensitive parameter to monitor different disease stages of experimental autoimmune uveoretinitis as well. ⁹  

In conclusion, our data demonstrate for the first time that OCT is a valuable tool for studying neurodegeneration in the retina in a rat model of autoimmune optic neuritis. Whereas histologic analysis and quantification of neuronal loss can only be done one time and terminates the in vivo phase of the experiment, OCT can be performed repeatedly. Therefore, it represents a noninvasive in vivo imaging technique which allows quantitative measurements of the rodent retina and provides full follow-up insight in disease progression. Moreover, OCT can be used for monitoring treatment effects in trials of neuroprotective strategies. Combining this technique with electrophysiological studies and MRI will allow a more accurate view of structure-function relations and might improve our understanding of disease processes of MS. 

Footnotes

 Supported by the European Union (FP6 program LSHM-CT-2005-018637, Neuropromise), and the Ministry of Education and Science, Republic of Serbia Grants ON174028 and III41007 (IG).

Footnotes

 Disclosure: K. Hein, None; I. Gadjanski, None; B. Kretzschmar, None; K. Lange, None; R. Diem, None; M.B. Sättler, None; M. Bähr, None