Remodeling of the fovea in Parkinson disease
- First Online:
- Cite this article as:
- Spund, B., Ding, Y., Liu, T. et al. J Neural Transm (2013) 120: 745. doi:10.1007/s00702-012-0909-5
- 414 Views
To quantify the thickness of the inner retinal layers in the foveal pit where the nerve fiber layer (NFL) is absent, and quantify changes in the ganglion cells and inner plexiform layer. Pixel-by-pixel volumetric measurements were obtained via Spectral-Domain optical coherence tomography (SD-OCT) from 50 eyes of Parkinson disease (PD) (n = 30) and 50 eyes of healthy control subjects (n = 27). Receiver operating characteristics (ROC) were used to classify individual subjects with respect to sensitivity and specificity calculations at each perifoveolar distance. Three-dimensional topographic maps of the healthy and PD foveal pit were created. The foveal pit is thinner and broader in PD. The difference becomes evident in an annular zone between 0.5 and 2 mm from the foveola and the optimal (ROC-defined) zone is from 0.75 to 1.5 mm. This zone is nearly devoid of NFL and partially overlaps the foveal avascular zone. About 78 % of PD eyes can be discriminated from HC eyes based on this zone. ROC applied to OCT pixel-by-pixel analysis helps to discriminate PD from HC retinae. Remodeling of the foveal architecture is significant because it may provide a visible and quantifiable signature of PD. The specific location of remodeling in the fovea raises a novel concept for exploring the mechanism of oxidative stress on retinal neurons in PD. OCT is a promising quantitative tool in PD research. However, larger scale studies are needed before the method can be applied to clinical follow-ups.
KeywordsParkinson disease (PD)Retinal foveal pitOptical coherence tomography (OCT)Receiver operating characteristics (ROC)Dopaminergic neuronsFoveal avascular zone (FAZ)
Parkinson disease (PD) is associated with dopaminergic deficiency. Both idiopathic and neurotoxin-induced PD cause a motor ‘Parkinsonian’ syndrome and impair retinal processing and vision (Bodis-Wollner et al. 1983; Bodis-Wollner 1990; Archibald et al. 2009; Bodis-Wollner 2009). Previous studies using optical coherence tomography (OCT), an optical signal acquisition and processing method (Huang et al. 1991), revealed that the nerve fiber layer (NFL), which represents axons of the output neurons of the retina, is thinned in PD. However, there is an apparent paradox. The cell bodies of the NFL are ganglion cells, while dopaminergic (DA) neurons and their interconnecting plexus are below the ganglion cells (Dowling and Ehinger 1978; Frederick et al. 1982; Hokoc and Mariani 1987; Mariani and Hokoc 1988; Witkovsky 2004), and several synapses removed from ganglion cell processing. Since the axons of DA amacrine cells do not directly enter the NFL, thinning of the NFL on its own does not explain decreased retinal dopamine (Harnois and DiPaolo 1990) or attenuated but levodopa-responsive ERG (Gottlob et al. 1989; Stanzione et al. 1990; Tagliati et al. 1995; Peppe et al. 1998; Sartucci et al. 2006).
Subjects and methods
Consecutive patients with PD and healthy control subjects were recruited from the Parkinson disease and Related Disorders Clinic and from the corresponding author’s practice. Written informed consent was obtained from each subject after the nature and possible consequences of the study were explained. All subjects, both PD subjects and controls, had identical comprehensive neurological and ophthalmic examinations [Electronic Supplementary Material (ESM) I]. The diagnosis of PD was based on established UK Brain Bank criteria (Hughes et al. 1992). The Unified Parkinson’s Disease Rating Scale (UPDRS) was administered to each PD patient and control subject. The staging of patients’ PD was based on the Hoehn and Yahr scale (Hoehn and Yahr 1967). Inclusionary criteria for the study were: being willing and able to give informed consent and comply with study protocol; UK Brain Bank Criteria for PD; Mini Mental State Score >27. Diabetics, ocular hypertensives (>20 mm Hg), or those with narrow anterior chambers were excluded as were those with unsuspected retinal pathology, such as macular drusen or epiretinal membrane (ESM I). Healthy control subjects were, whenever possible, recruited amongst spouses and caretakers of the patients. Ultimately, 30 PD patients (50 eyes) and 27 control subjects (50 eyes) fulfilled the criteria. The number of untreated patients was five. 14 were on monotherapy (Levodopa 10, Amantadine 3, Selegiline 1) and 11 on combination therapy (three on Levodopa and Amantidine, three on Dopamine Agonist and Amantidine, two on levodopa and Selegiline, 1 on Levodopa and Dopamine Agonist therapy. Two patients were on Levodopa, Selegiline and Amantidine. 1 had Levodopa, a dopamine agonist and Selegiline and 1 was on levodopa, Amantidine and Selegiline. In other words of those on treatment, 17 had Levodopa alone or in combination. All subjects provided written informed consent approved by SUNY Downstate Medical Center Institutional Review Board.
Equipment and measurements
Subjects had Fourier-domain OCT (RTVue Model RT 100; Optovue, Inc.; Fremont, CA, USA) MM5 or EMM5 scans, using grids of 5 × 5 mm (MM5) or 6 × 6 mm (EMM5) sections centered on the foveola (ESM Fig. 1 and ESM II). Both have the same sampling rate and use identical horizontal rules. Raw data files were exported and analyzed using MATLAB™ engineering software. This method allowed uniform thickness measurements at desired points and reconstruction of three-dimensional models of each fovea.
Individual and group analysis of foveal thickness
Receiver operating characteristics
Generalized Wald tests (Engle 1983) were used to compare ROCs for each eye of each subject. Additional cutoff-thresholds (for positive and for negative) were created by generating a series of standard deviation increments above and below the template. The above analysis was repeated for each of these standard deviation increments, for each distance. SAS Release 9.2 (SAS Institute, Cary, NC, USA) statistical software was used.
Thickness of the perifoveal retina: a statistical comparison of the healthy and PD groups
Perifoveolar inner and full retinal thickness values for PD patients and control subjects
Retinal thickness (μ)
As we shall explain below, using ROC to define the sensitivity and specificity of the difference at each radial distance, allows us to assert the statistical significance of the difference, not simply on the lumped average, but for each perifoveolar distance (see also “Subjects and methods”).
Sensitivity and specificity of retinal thinning in the perifoveolar inner retina
We used ROC calculations at each measured perifoveolar distance for defining sensitivity and specificity for best discrimination between healthy and PD retinae (Fig. 3). These calculations relate to the standard Wilcoxon statistics (see “Subjects and methods”) and were subjected to. From the ROC we calculated the area under the curve (AUC) and plotted AUC for each perifoveolar distance (Fig. 3). While the AUC curve is not sharply tuned, IRL thickness comparison at radial distances between 1 and 1.5 mm from the foveola discriminate optimally patients from controls. Figure 4 shows that the perifoveolar IRL zone where the greatest raw thickness difference (red) occurs between control and PD retinae coincides with the results of sensitivity (AUC) calculations (Fig. 3; Table 1). About 78 % of eyes can be discriminated from healthy controls based on inner retinal thickness (Fig. 1) at the perifoveolar zone of the foveal pit (Table 1).
Foveal remodeling and dopaminergic neurons
Based on previous studies in humans and monkeys (Harnois and DiPaolo 1990; Nguyen-Legros 1998; Djamgoz et al. 1997; Ghilardi et al. 1988, 1989; Tagliati et al. 1994) it is reasonable to assume that a link exists between changes found with OCT and dopaminergic deficiency in the retina in PD (Bodis-Wollner 1990). Dopaminergic (DA) amacrine cells possess a high affinity uptake system (Hendley and Snyder 1972) explaining both their vulnerability to the selective neurotoxin methyl-phenyl-tetrahydropyridine (MPTP) and their response to exogenous levodopa therapy. Retinal dopamine concentration is decreased in PD (Harnois and DiPaolo 1990) but nearly normal in those who received levodopa therapy shortly before death. In the experimentally induced monkey model of PD using MPTP, there is a loss of one subset of dopaminergic amacrine cells in the monkey retina. In the experimentally induced rat model of PD using rotenone, retinal histology shows the loss of DA cells (Biehlmaier et al. 2007) and its intraocular injection causes predominant, if not exclusive thinning of the inner nuclear and plexiform layer. Postmortem studies in the human retina (Nguyen-Legros 1998; Djamgoz et al. 1997) and histological and neuropharmacological studies in MPTP-treated monkeys and monkey eyes using 6-hydroxydopamine (Ghilardi et al. 1988, 1989), are consistent with experimental electrophysiological (Gottlob et al. 1989; Stanzione et al. 1990; Ikeda et al. 1994; Tagliati et al. 1995; Peppe et al. 1998) and psychophysical data (see Bodis-Wollner in PD. However, loss of DA amacrine cells may not be solely responsible for inner retinal thinning: levodopa therapy does not completely restore the ERG in PD. The average thinning in the perifoveolar zone in PD is about 8–11 μm. This represents roughly a loss of 15 % of the inner plexiform/ganglion cell thickness based on quantification of thickness by Bagci et al. (2008) who used the same OCT equipment as we used. Thus it is conceivable but not convincing that this tissue loss represents only that of dopaminergic amacrine cells (see a review by Bodis-Wollner 2012).
OCT was first applied to the retina of PD patients by Inzelberg et al. (2004). The results inspired many subsequent studies. Most of these (Altintas et al. 2008; Cubo et al. 2010; Moschos et al. 2011; La Morgia et al. 2011) corroborated that the NFL comprising the axons of retinal ganglion cells is thinned in PD. Aaker et al. (2010) found significant change only in one segment of the perifoveolar area, while Archibald et al. (2011) found no evidence to support this conclusion (for a summary of OCT in PD see Bodis-Wollner 2012). Dopaminergic (DA) neurons and their interconnecting plexus are, in both human and monkey, below the ganglion cells (Frederick et al. 1982; Hokoc and Mariani 1987; Mariani and Hokoc 1988), several synapses removed from ganglion cell processing. Thus NFL thinning on its own does not explain decreased retinal dopamine (Harnois and DiPaolo 1990) and levodopa responsiveness of the ERG and there is no good explanation of NFL thinning, based on dopaminergic deficiency of the retina, as the processes of DA amacrine cells do not directly synapse onto ganglion cells.
Our SD-OCT study provides direct evidence of the thinning of the pre-nerve fiber layer of the retina in PD. Below we discuss questions raised by remodeling of the fovea in the PD based on pre NFL thinning.
Proteomics and the retina
It has been proposed that non-motor PD symptoms may be related to synuclein pathology that is initiated prior to the onset of the cardinal motor symptoms. Postmortem studies (Braak et al. 2002) show pathological and anatomical evidence indicating that the hallmark of PD, pathological folding of synuclein (Tan et al. 2009), initially affects neurons other than those which are dopaminergic, and neurites. Synuclein pathology progresses from peripheral to central CNS structures from the gut, olfactory bulb, and N. dorsalis vagus. Although α-synuclein is found in the healthy retina (Martínez-Navarrete et al. 2007), it has hitherto not been considered in postmortem retinal studies of PD.
The ubiquitin–proteasome system (UPS) is a degradation pathway for folded and damaged proteins and plays a neuroprotective role in response to a number of cellular stresses. Elements of UPS, the ubiquitin carboxyl-terminal hydrolase L1 (UCH-L1) proteasomal system and parkin genes are expressed at the mRNA and protein levels in several cell types in the human retina, including tyrosine hydroxylase-positive (dopaminergic) amacrine cells (Martínez-Navarrete et al. 2007; Esteve-Rudd et al. 2010). In the healthy retina, UCHL-1 and Parkin, encoded by the PARK2 gene, are thought to exert a protective function against neuronal stress in amacrines and ganglion cells. While retinal Parkin is widespread, UCH-L1 is more specific to horizontal cells and subtypes of bipolars, amacrine cells, ganglion cells and the nerve fiber layer.
Oxidative stress and foveal remodeling
We quantified the shape of the foveal pit in both healthy and PD subjects based on finely sampled pixel-by-pixel thickness values. The difference becomes evident between 0.5 and 2 mm from the foveola. Based on ROC calculations inner retinal thickness in the radial zone of 0.75–1.5 mm yields the optimal area for discriminating PD from healthy control subjects. Histology reveals the fovea as an easily recognizable pit (Fig. 1). Its wall is curved and different inner retinal layers emerge at different distances from the foveola. The pit is free of the NFL. Different layers are identified by OCT by differences in refractive properties at the boundaries of layers with different cellular composition. Using spectral-domain OCT (SD-OCT) imaging (Wojtkowski et al. 2004) we (Hajee et al. 2009) made use of OptovueTM retinal segmentation and observed on reconstructed images that the perifoveolar preganglionic inner retina is thinned in PD. We now applied finely sampled thickness measurements of retinal layers (Bagci et al. 2008; Loduca et al. 2010) in the foveal pit. The results specify an annular perifoveolar zone where healthy and PD subjects’ retinae differ most reliably. The specific distance of the annular zone suggests that the inner plexiform layer below the retinal ganglion cells is dominantly affected in the fovea in PD. This layer contains a rich plexus of amacrine cells, including interconnecting fibers of DA amacrine cells (Witkovsky 2004; Witkovsky et al. 2008).
There is no evidence of a distribution of the dopaminergic amacrines and their plexus in a perifoveal ring-like distribution in the healthy retina. Hence both the location of maximum thinning around 1.25–2 mm from the foveola and the magnitude of the maximum individual thickness difference (18 μm) between the healthy and PD fovea may not occur due to death of dopaminergic amacrine cells alone. The thinned annular zone overlaps and extends the so-called foveal avascular zone (FAZ) of the adult retina (Provis and Hendrickson 2008). This geography of the remodeled foveal pit in PD raises a challenging thought concerning the genesis of oxidative stress in the PD retina. It further poses an experimentally amenable question whether foveal remodeling in PD is linked to the capillary network. It was reported in 110 healthy normal eyes that there is a continuum ranging from a shallow to a normal depth pit (Tick et al. 2011). The study noted a continuity from the inner nuclear layer (INL) lying over the center (seven eyes; 6.7 %), to a complete separation of inner layers overlying a flat and thinner central outer nuclear layer (ONL; eight eyes; 7.3 %). We did not see in 50 PD eyes an overlay of the inner layers over the foveola. It was reported that in healthy subjects, central foveal thickness correlated inversely to the size of the FAZ (Dubis et al. 2012). There is considerable variability of the size of the FAZ (Chui et al. 2012). Interestingly, Dubis et al. (2012) derived the conclusion that the findings “support a developmental model in which the size of the FAZ determines the extent of centrifugal migration of inner retinal layers, which counteracts in some way the centripetal packing of cone photoreceptors.” In our study, we did not find any difference in the photoreceptor thickness in the central fovea; nevertheless, the potential role of foveal oxygen supply and the role of capillaries in foveal remodeling may warrant further studies in PD.
The potential clinical value of the OCT of the fovea: prospects and caveats
Functional correlations of OCT in PD
Moschos et al. (2011) correlated multifocal ERG (mfERG) in the central retina in PD with OCT thinning of the NFL. The mfERG technique provides a large array of stimulus elements, typically in a 20–30° field and reflects outer and also inner retinal responses. However, it is the foveal inner retina, which dominates the peak of the mf ERG response in the center.
Altintas et al. (2008) reported that in PD macular and nerve fiber layer thickness is reduced. However, while foveal thickness did not distinguish PD and healthy controls, foveal thickness nevertheless did correlate with the severity of PD. In fact the measure of foveal thickness in their study was the thickness of the outer retina underneath the foveola. We agree: foveolar thickness in the very center, the foveola, does not distinguish PD from HC, suggesting that there is minimal if any photoreceptor thinning in the foveola.
Further correlative studies of functional vision such as contrast sensitivity, color vision and the pattern ERG could be revealing. These are all primarily reflecting foveal visual processing.
Nevertheless, it needs to be established whether a sampling of the retina beyond the area we quantified and including the NFL, will improve group statistical differences of PD and healthy controls. At this point, the OCT is best adapted for central retinal measurements, but with special equipment, zones in the periphery could also be examined.
OCT is non-invasive, widely available and easily performed. However, the disease specificity of foveal remodeling in PD will need to be determined in a larger scale study. Furthermore, longitudinal studies of foveal architecture are needed to ascertain that OCT can serve as one biomarker in PD.
The National Parkinson Foundation and Michael J. Fox Foundation provided Partial indirect Funding/Support. William Brunken, PhD, John Danias, MD, PhD, Douglas Lazzaro, MD, Marilee P. Ogren-Balkema, PhD and John Dowling, PhD provided critical revisions and suggestions on an early and Edward Chay, MD on a later version of the manuscript. Dr. M. Asim Javaid provided much assistance during the study and made comments on earlier versions of this work. Galina Glazman, Hunter College student, and Dr. N. Sohail provided much laboratory assistance. The SUNY Downstate Academic Center for Scientific Computing (Jeremy Weedon, PhD and Matt Avitable, PhD) had advised us on several aspects of the study. We benefited from very substantial discussions with Samantha Slotnick, OD and Jerome Sherman, OD of the SUNY College of Optometry and SUNY Eye Institute.
Conflict of interest
Authors do not have a financial relationship with the National Parkinson Foundation and Michael J. Fox Foundation that sponsored the research. The authors declare that they have no conflict of interest.
All human studies have been approved by SUNY Downstate Medical Center Institutional Review Board and have therefore been performed in accordance with the ethical standards laid down in the 1964 Declaration of Helsinki. All persons gave their informed consent prior to their inclusion in the study.