Objective Trans-synaptic axonal degeneration is a mechanism by which neurodegeneration can spread from a sick to a healthy neuron in the central nervous system. This study investigated to what extent trans-synaptic axonal degeneration takes place within the visual pathway in multiple sclerosis (MS).
Methods A single-centre study, including patients with long-standing MS and healthy controls. Structural imaging of the brain (MRI) and retina (spectral-domain optical coherence tomography) were used to quantify the extent of atrophy of individual retinal layers and the primary and secondary visual cortex. Generalised estimation equations and multivariable regression analyses were used for comparisons.
Results Following rigorous quality control (OSCAR-IB), data from 549 eyes of 293 subjects (230 MS, 63 healthy controls) were included. Compared with control data, there was a significant amount of atrophy of the inner retinal layers in MS following optic neuritis (ON) and also in absence of ON. For both scenarios, atrophy stopped at the level of the inner nuclear layer. In contrast, there was significant localised atrophy of the primary visual cortex and secondary visual cortex in MS following ON, but not in MS in absence of ON.
Interpretation These data suggest that retrograde (trans-synaptic) axonal degeneration stops at the inner nuclear layer, a neuronal network capable of plasticity. In contrast, there seems to be no neuroplasticity of the primary visual cortex, rendering the structure vulnerable to anterograde (trans-synaptic) degeneration.
- Multiple Sclerosis
Statistics from Altmetric.com
In the last two decades, great progression in monitoring and treating the inflammatory pathology of multiple sclerosis (MS) has occurred.1–3 Despite these advances, a large proportion of patients suffering from MS continue to progress, due to axonal and neuronal degeneration.4 To date, it is however not known what mechanism drives neurodegeneration, but early neuroplasticity was suggested to be of prognostic relevance already at disease onset.5 If correct, neuroplasticity should protect against axonal loss, the main reason for sustained disability, over the long term.
The hypothesis is that axonal loss and neurodegeneration are related and can be quantified by atrophy measurements. Anterograde axonal degeneration has been known for some time,6 but only with the recent description of in vivo trans-synaptic axonal degeneration a bidirectional pathophysiological mechanism was available to explain progression of neurodegeneration in the human central nervous system (CNS).7 ,8 This discovery was made possible by use of retinal optical coherence tomography (OCT). Retinal OCT has matured into a highly sensitive and accurate tool for quantifying minute (1–2 µm) amounts of atrophy, which permits to separate individual retinal layers.9 This opens avenues for investigating the structural cascade of anterograde and retrograde (trans-synaptic) axonal degeneration in MS. Clearly, postmortem data demonstrated atrophy of retinal layers after long-standing disease.10 In absence of direct damage to the retina it is likely that retinal layer atrophy was caused by retrograde trans-synaptic axonal degeneration following more central damage to the optic pathways. There is yet no comparable in vivo data from patients with long-standing MS, but a meta-analysis of the literature on patients with a shorter disease duration suggests that atrophy of the retinal nerve fibre layer (RNFL) may be caused by retrograde axonal degeneration following optic neuritis (MSON), and trans-synaptic retrograde axonal degeneration in absence of ON (MSNON).11 Besides retinal atrophy, caused by retrograde (trans-synaptic) degeneration there is also evidence for anterograde trans-synaptic degeneration in the visual system.12–14 Anterograde degeneration along the visual pathway takes place in the lateral geniculate nucleus and visual cortex following ON or diffuse white matter axonal pathology.14 These data are consistent with optic tract diffusion abnormalities and reduced magnetisation transfer ratio in the occipital cortices following ON.15–17
Taken together, there seems to be evidence for retrograde and anterograde (trans-synaptic) axonal degeneration in MS. If left unrestricted, this process may be allowed to spread through the visual pathway and, given the many connections between the visual system and the rest of the brain, potentially the entire human CNS like a ‘domino effect’ (denoting a chain of trans-synaptic events). While there is evidence for the former,14–17 the latter was never observed. Therefore, the aim of this study was to confirm trans-synaptic axonal degeneration and to investigate if there was a physiological barrier to this process within the visual system of patients with long-standing MS.
Study design and patient population
Patients and controls were enrolled from the VU University Medical Centre, Amsterdam, between March 2011 and August 2012 for this cross-sectional study. This study was approved by the medical ethical committee (protocol number 2010/336) and the scientific research committee (protocol number CWO/10–25D) of the VU University Medical Centre in Amsterdam, the Netherlands. Written informed consent was obtained from all included subjects.
Patients were eligible for inclusion if they completed a written informed consent, were aged 18–80 years and had a diagnosis of either relapsing remitting (RR), secondary progressive (SP) or primary progressive (PP) MS at the time of their assessment (defined by Lublin-Reingold criteria).18 Patients were excluded if they fulfilled any of the following criteria: pregnancy; received a course of steroids or had a relapse within the preceding month; HIV or other immunodeficiency syndrome; history of substance abuse (drug or alcohol) in the past 5 years and specific MRI findings that could interfere with evaluation.
Healthy control (HC) subjects were included if they were willing to sign an informed consent form, aged 40–60 years at the time informed consent was signed and if they had no history of any neurological or psychiatric disease. Exclusion criteria for HC subjects were pregnancy and if they were related (1st or 2nd degree) to a patient with MS.
OCT images were acquired with a SD-OCT (Spectralis, Heidelberg Engineering, Heidelberg, Germany, Software V.126.96.36.199), using dual beam simultaneous imaging, with the eye tracking function enabled for optimal measurement accuracy.19
In order to obtain data on the peripapillary region, a 12° ring scan (corresponding to an approximately 3.4-mm-diameter circle in a normal eye) was used. The ring scan was centred manually on the papilla after activating the eye tracking function in each subject. Likewise, the macular volume scan (20×20° field, 25 B-scans) was manually centred over the macula after activation of the eye tracking function. Both scan areas are depicted in figure 1A (peripapillary) and B (macula).
All OCT scans were performed by four trained and certified technicians. Scans were excluded from the analyses if they violated international consensus quality control criteria (OSCAR-IB).20
Image postprocessing was performed as follows. The peripapillary ring scan and the macular volume scan were analysed with the aid of new segmentation software (Heidelberg Engineering, Heidelberg, Germany). The software allows for reliable segmentation of three retinal layers at the macular region; the RNFL, the ganglion cell complex (GCC=ganglion cell layer (GCL)+inner plexiform layer (IPL)) and the inner nuclear layer (INL). Furthermore, the outer retinal layers (ORLs) were calculated for the macular region by subtracting the RNFL, GCC and INL from the complete retinal thickness (shown in figure 1C, bottom image). For every layer, the OCT software generated a thickness map on a 1 mm, 3 mm and 6 mm grid, as defined by the Early Treatment Diabetic Retinopathy Study (EDTRS). Of the nine areas provided by this grid, only the mean of the four inner regions composing the perimacular rim were used in this study (grey shaded area in figure 1B). This was done because segmentation of the individual layers in the outer regions was shown to be of poorer reproducibility, compared with the four inner Early Treatment Diabetic Retinopathy Study sectors.21 The foveolar thickness is dominated by the ORLs, which precluded any reliable quantification of inner retinal layers from this area.
In the peripapillary region, the software enabled segmentation of four distinct layers; the RNFL, INL, outer nuclear layer and the outer plexiform layer (shown in figure 1C, upper image). For all layers, a thickness map was provided of which the layer specific global mean values were used for statistical analysis.
After the segmentation process, all scans and all layers were verified by two independent OCT technicians. In case of obvious algorithm failure on a small area (due to a vessel or a small artefact), the deviating line was corrected manually. Scans were excluded if the algorithm failure covered the complete OCT B-scan area. The retinal layer segmentation software used in this study has shown to be reliable, with intraclass correlation coefficients (ICCs) above 0.94 for all layers, except the IPL (ICC 0.70), which was an important source of disagreement. This issue was solved by combining the GCL and IPL to form the GCC (ICC of 0.99).21
The Expanded Disability Status Scale was determined by a certified examiner to assess level of disability. Visual acuity (VA) testing was performed monocularly in all patients, using the Snellen acuity chart (5 m distance, using decimal notation). Medical history with respect to visual symptoms was obtained in all patients and controls.
MRI data acquisition and analysis
MRI was performed on a 3 T whole body scanner (GE Signa HDxt, Milwaukee, Wisconsin, USA) using an eight-channel phased array head coil. The protocol included a three-dimensional T1-weighted fast spoiled gradient echo sequence (FSPGR; repetition time 7.8 ms, echo time 3 ms, flip angle 12°, 240×240 mm2 field of view, 176 sagittal slices of 1 mm thickness, 0.94×0.94 mm2 inplane resolution) for cortical thickness measurements, and a three-dimensional fluid attenuated inversion recovery image (FLAIR; repetition time 8000 ms, echo time 125 ms, inversion time 2350 ms, 250×250 mm2 field of view, 132 sagittal slices of 1.2 mm thickness, 0.98×0.98 mm2 inplane resolution) for lesion detection.
White matter lesion segmentation was automatically performed using the FLAIR and FSPGR images.22 Cortical thickness measurements were performed using the FreeSurfer pipeline (V.5.1) as previously described.23 ,24 Before applying FreeSurfer, the binary lesion mask was used to minimise the impact of white matter lesions on the cortical thickness measurements by applying a lesion filling algorithm.25 After running the FreeSurfer pipeline, the primary visual cortex (V1) and secondary visual cortex (V2) were segmented in each subject by surface based registration of an atlas.26 The resulting cortical parcellations were then used to compute the average thickness of V1 and V2. In order to assess localised atrophy of V1/V2, normalised V1/V2 thicknesses were calculated by dividing the V1/V2 cortical thickness by complete cortical thickness.
Demographic data was compared between patients and controls using independent sample T tests, χ2 tests and linear regression analyses. For all segmented retinal layers the relative thickness (individually normalised to complete retinal thickness) was calculated in order to investigate the layer-specific contribution to the complete retina. This was done for the perimacular rim and the peripapillary area.
Differences in retinal layer thickness between eyes with a history of ON (MSON), without a history of ON (MSNON) and HC eyes, were analysed with generalised estimation equations. This method was used because the analyses were performed on eye level instead of subject level. In the generalised estimation equations analysis (adjusting for intrasubject intereye correlation, using an exchangeable correlation structure), age was used as a covariate to adjust for confounding.
In order to assess the relationship between cortical thickness of V1 and V2, (absolute and normalised to total mean cortical thickness) and retinal layer thickness, linear multiple regression analyses (with adjustments for age and sex) were used. In order to avoid contamination of the study groups by pooling of MSON eyes (showing a severe degree of RNFL loss) with MSNON eyes (showing a small degree RNFL loss),11 these analyses included only patients in whom both eyes had the same history of ON (both eyes MSON or both eyes MSNON). In these analyses, the mean of both eyes was used.
Statistical analyses were performed using SPSS V.20.0, with a statistical significance level of 0.05. A Bonferroni correction was applied to adjust for multiple testing.
A total of 230 patients with MS (140 RRMS, 61 SPMS, 29 PPMS) and 63 HCs were included (table 1). Patients had a mean disease duration of 20.4 (±7.0) years (median: 20.2 years, IQR: 9.4, total range 8.8–45.9). Patients were older (median: 54.1 years, IQR: 14.4, total range: 30.7–81.1) compared with control subjects (median: 50.8 years, IQR: 8.6, total range: 27.5–62.9, p=0.001). An episode of unilateral MSON was experienced by 27%, while an episode of bilateral (simultaneous or sequential) MSON by 19%. Only one patient showed signs of unilateral microcystic macular oedema. Exclusion of data from this patient did not change the level of significance for any of the following calculations. The VA was worse for MSON (0.74±0.25) compared with MSNON eyes (0.86±0.19, p<0.001). Although visual cortical thicknesses of V1 and V2 were less in patients compared with controls (p=0.005 and p<0.001, respectively), no localised atrophy of V1 or V2 was observed in patients, as the normalised visual cortical thicknesses did not differ between patients and controls (p=0.634, p=0.102, respectively).
OCT scans were performed in 549 eyes (423 in MS, 126 in HC). The remaining eyes were not scanned, essentially due to problems with visual fixation using either an internal or external target. The exclusion of the acquired scans after quality control (OSCAR-IB) and layer segmentation is depicted in a flow chart (figure 2).
Optic neuritis causes severe retrograde degeneration of inner retinal layers
Table 2 (upper part) shows the relative thickness of all retinal layers for MSON, MSNON and HC eyes. In the macular area, there was a significant amount of atrophy of the inner, but not the outer layers in MSON and MSNON eyes, compared with HC (table 2). Specifically, the macular RNFL and GCC showed the largest relative thickness in HC eyes (8.2% and 28.3%, respectively), compared with MSON and MSNON eyes (table 2). In contrast, the INL and ORLs of the macular area showed the largest relative thickness in MSON eyes (13.9% and 56.0%, respectively), compared with MSNON or HC eyes (p<0.001 for all comparisons).
Importantly, the analysis of the relative thickness values from the peripapillary region clearly demonstrated that the INL did not differ significantly between groups while the thickness of the innermost layer (RNFL) was significantly reduced in MSON and MSNON eyes compared with HC eyes. In contrast, the outermost layers (counting from the INL) were significantly thicker in MSON eyes compared with MSNON or HC eyes (p<0.001 in all comparisons). Figure 3 shows the direct comparison between MSNON and HC eyes for all retinal layers. For transparency, the corresponding absolute thickness data of all retinal layers are also shown in table 2 (bottom part).
Anterograde trans-synaptic axonal degeneration causes atrophy of the primary visual cortex
Axons from the temporal visual field cross at the chiasm, demanding stratification of patients into two subgroups which are either bilateral normal (MSNON, n=107) or bilateral MSON (n=39). The average data of the right and left sides were used.
The mean thicknesses of the complete cortex, V1 and V2 (absolute and normalised) were all significantly thinner in patients with MS compared with HC subjects, irrespective of the presence of MSON. The thickness of the complete cortex did not differ between patients with MSON and patients with MSNON (2.45 vs 2.47, p=0.425). In contrast, the thicknesses of V1 and V2 (absolute and normalised) were consistently thinner in patients with MSON compared with patients with MSNON. These differences were significant for all comparisons, except for absolute V2 thickness, where statistical significance was narrowly missed (table 3, p=0.053).
Retinal layer atrophy is not related to thickness of the visual cortex
There was no significant association observed between retinal layer thicknesses and V1 thickness, either in patients with MSON, patients with MSNON or HCs, after Bonferroni correction for multiple comparisons. Results of the regression analyses, with adjustments for disease duration or age, are shown in table 4.
Likewise, there were no significant associations between retinal layer thicknesses with V2 thickness (data not shown). Moreover, results did not change when normalised V1/V2 thicknesses were used (data not shown).
This study identified a possible physiological barrier to axonal degeneration within the visual system (figure 4). Retrograde (trans-synaptic) axonal degeneration is directed towards and is possibly being blocked in the retina. Anterograde (trans-synaptic) axonal degeneration is directed towards the V1 where it causes localised atrophy. The present data suggests that retrograde (trans-synaptic) axonal degeneration is halted at an anatomical structure capable of neuroplasticity, the INL of the retina. The INL represents a neuronal network of bipolar, amacrinal and horizontal cells, which appears to act as a dam to retrograde (trans-synaptic) axonal degeneration resulting from axonal loss during an episode of MSON or diffuse axonal pathology (figure 4). In contrast with the extensive damage observed in the inner retinal layers, almost no atrophy could be shown beyond the INL after an average disease duration of 20 years. Therefore this study also closes a gap of the literature, biased towards cross-sectional studies of the early disease course. Importantly, plasticity of the V1 is blocked early in life.27 This may be the reason why anterograde (trans-synaptic) degeneration leads to atrophy of V1 and V2. A finding which is consistent with other studies.15–17
The observed differences in RNFL thickness between MSON and MSNON eyes are also consistent with the literature.11 The thinning of the RNFL was more severe in MSON eyes (retrograde axonal degeneration) compared with MSNON eyes (trans-synaptic retrograde axonal degeneration). This consistent observation can be explained by the fact that the optic nerve is a vulnerable structure because of the anatomically high density of the closely packed, hard-wired, axons. Thus, an episode of ON will immediately damage a large amount of axons, eventually resulting in axonal degeneration. Ganglion cell loss and RNFL thinning will be relatively less severe following diffuse axonal pathology in the white matter of the MS brain (MSNON) because axons are spread out over a large area in Meyer's loop.28 ,29
Although large differences between MSON and MSNON eyes were observed, it can be concluded that after two decades of MS, MSON and MSNON, eyes revealed substantial atrophy (especially of the inner layers of the retina), compared with healthy eyes.
The present data suggest that neuroplasticity may be a mechanism by which retrograde (trans-synaptic) axonal degeneration can be halted.
The damage observed in the inner retinal layers is caused by retrograde (trans-synaptic) axonal degeneration and it appears that this mechanism mainly affects physiologically hard-wired neuroaxonal connections such as the dominant projections from the first order neurons (retinal ganglion cells), to the second order neurons (within the lateral geniculate nucleus) and finally third order neurons, located within layer four of V1.
Retinal bipolar cells in the INL differ from the hard-wired 1st, 2nd and 3rd order neurons because of their extensive synaptic tree.30 Amacrinal and horizontal cells feed into this synaptic tree which has an extraordinary capacity for plasticity.31 Plasticity of the INL is also exploited for retinal implants.32 ,33 Likewise, neuroplasticity is the recognised mechanism for cortical filling-in of circumscribed scotomata (V1) following retinal lesions.34 ,35 The likely anatomical substrate for neuroplasticity following scotomata are horizontal within-cortical connections because the hard-wired connections from the lateral geniculate nucleus to V1 remain unaltered.36
In the present study, clear differences in visual cortical thickness were observed between MSON, MSNON and HC. Nonetheless, no association was observed between retinal layer thickness and visual cortical thickness in MSON, MSNON or HC. These findings are consistent with a study published while this paper was under revision, reporting clear atrophy of the visual cortex in patients with severe MSON, and to a lesser extent in patients with mild MSON, compared with patients with MSNON.37 Furthermore, Gabilondo et al reported a significant association between RNFL thickness and visual cortex volume in the complete patient group, but this effect disappeared when patients with MSNON and MSON were analysed separately. The lack of association between retinal layer thickness and visual cortical thickness may be potentially explained by presence of silent lesions in the optic radiations in patients with MSON and possible subclinical episodes of ON.To the best of our knowledge, this study is the first study to examine retinal layer atrophy in patients with long-standing MS. This is relevant because axonal degeneration in the CNS can be an extremely slow process.38 ,39 Therefore, the short disease duration of previous studies does not permit the drawing of any definite conclusions in the long term.40–44 Therefore our data predicts that longitudinal studies are likely to find a plateau effect for inner retinal layer atrophy with preservation of the outermost layers.
In human retinal postmortem samples, extensive retinal atrophy has been demonstrated after about 20 years of disease duration.10 This matches the time lag of the present study, which therefore closes the gap between in vivo and postmortem data on retinal atrophy.
The main limitation of this study is that it relies on indirect data only, as a direct investigation of retinal structures and visual pathways by histology will not be possible in the acute phase, with the exception of some extraordinary rare autopsy reports. Although long-term, the data remains cross-sectional, though we predict a plateau effect for retrograde axonal degeneration. Another shortcoming is that testing of visual function was restricted to high contrast VA. As Balcer et al45 ,46 pointed out, this is a rather poor measure for visual function in MS. Next, controls were younger than patients. This was taken into account by using the individually normalised (relative) data and age-adjusted statistical tests. Finally, records of MSON were based on a mix of clinically confirmed episodes and a patient history on earlier events during the disease. Although this approach is consistent with other studies in the field, it may have introduced a bias towards more severe episodes of MSON.
In summary, this study describes neuroplasticity of the INL as a possible physiological barrier, to retrograde (trans-synaptic) axonal degeneration.
The MS Centre Amsterdam has received funding from the Dutch MS research foundation, but had no involvement in the study design. The authors would like to thank the patients and healthy controls for participating and their substantial time commitment as well as the staff of the Departments of Ophthalmology and Neurology, VUmc for supporting this study. The authors also thank Heidelberg Engineering for providing a purpose-built algorithm and image analysis software.
Contributors LJB contributed to designing the study concept, performing data collection, revision and QC assessment of OCT scans, automated and manual layer segmentation, statistical analyses and prepared the first draft of the manuscript. JWRT assisted with statistical analyses and revised the manuscript. MDS and MD performed data collection, MRI processing and revised the manuscript. PT performed data collection, revision and QC assessment of OCT scans and revised the manuscript. JK, BMJU and CHP contributed to designing the study concept and revised the manuscript. AP contributed to designing the study concept, revision and QC assessment of OCT scans, automated and manual layer segmentation, statistical analyses, preparation of figures, critical review and revising the manuscript. All authors gave final approval of the version submitted.
Competing interests LJB, MDS, MD and AP are supported by the Dutch MS Research Foundation through a program grant to the VUmc MS Center Amsterdam (grant number 09-538d). JWRT and PT report no disclosures. MD was supported by a private sponsorship to the VUmc MS Center Amsterdam. JK has accepted speaker and consulting fees from Merck-Serono, Biogen-Idec, Novartis, Teva, Genzyme and Novartis. BMJU has received consultation fees from Novartis, Merck Serono, Biogen Idec and Danone Research. CHP received an institutional grant from Novartis; consultancy fees from Actelion, Biogen Idec, Bayer Schering, TEVA, Merck-Serono, Novartis, Glaxo SK, UCB, Roche, Antisense Ther; expert testimony: Biogen. AP sits on the Novartis steering committee for a multicentre observational study: ‘A 3-year, open-label, multicentre, multicohort, parallel-group study to validate optical coherence tomography in patients with multiple sclerosis’.
Patient consent Obtained.
Ethics approval Medical ethical committee VU University Medical Centre .
Provenance and peer review Not commissioned; externally peer reviewed.
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.