Parkinsonism and Related Disorders

Transcription

1 Parkinsonism and Related Disorders 19 (2013) 1e14 Contents lists available at SciVerse ScienceDirect Parkinsonism and Related Disorders journal homepage: Review Foveal vision is impaired in Parkinson s disease Ivan Bodis-Wollner * State University of New York (SUNY) Downstate Medical Center, 450 Clarkson Avenue, Box 1213, Department of Neurology, Brooklyn, NY , USA article info abstract Article history: Received 6 July 2010 Received in revised form 16 July 2012 Accepted 21 July 2012 Keywords: Retina Visual cortex Fovea Contrast sensitivity Isoluminance colour contrast PERG Optical coherence tomography (OCT) ERG/VEP Dopamine Higher visual dysfunction Purpose: The article aims to review foveal involvement in Parkinson s disease. Scope: Clinical observations as well as electrophysiological and anatomical studies in animal models provide evidence that Parkinson s disease (PD) affects vision. The retina is the most distal locus of visual dysfunction in PD as shown by electroretinographic (ERG) and optical coherence tomographic (OCT) studies. Thinning of the retinal nerve fibre layer (RNFL) and the fovea has been reported in PD. This review summarises retinal physiology and foveal visual dysfunction in PD and quantification of retinal thinning as reported in different studies and using different instruments. At this point due to methodological diversity and relatively low number of subjects studied, a meta-analysis is not yet possible. Results obtained on one equipment are not yet transferable to another. The author also briefly alludes to some links of visual processing deficits beyond visual detection, such as visual discrimination, visual categorisation and visuospatial orientation in PD. Conclusions: There are some promising results suggesting the potential applicability of ST-Oct as a biomarker in PD. Furthermore, these data raise some interesting neurobiological questions. However, there are identifiable pitfalls before OCT quantification may be used as a biomarker in PD. Analysis standardisation is needed on a larger than existing healthy and patient population. Furthermore, longitudinal studies are needed. The exact relationship between retinal foveal deficits and visuocognitive impairment in PD remains a challenging research question. Ó 2012 Elsevier Ltd. All rights reserved. Contents 1. Foveal vision e the retina and primary visual cortex Retinal architecture and visual functions of the fovea Fovea Retinal ganglion cells The effect of PD on foveal vision Visual evoked potentials Contrast sensitivity The electroretinogram Colour vision Open questions about differences in ERG and CS OCT and in vivo retinal morphology Retinal thinning in PD Caveats of OCT use Discussion Dopaminergic circuitry of the retina and foveal processing in PD Differential diagnosis of retinal thinning Proteomics of the PD retina Visual cortex and beyond * Tel.: þ ; fax: þ address: ivan.bodis-wollner@downstate.edu /$ e see front matter Ó 2012 Elsevier Ltd. All rights reserved.

2 2 I. Bodis-Wollner / Parkinsonism and Related Disorders 19 (2013) 1e14 8. Conclusions and recommendations Institutional review board approval Conflict of interest Acknowledgements References This review examines the effects of Parkinson s disease (PD) on foveal vision in humans and in primate models of the disease. Several comprehensive reviews on dopamine in retinal processing have been published previously [1e5]. The current review focusses on foveal signal processing and PD, and considers its potential contribution to higher order visual processing. Although the fovea represents a tiny anatomical region of the retina, the number of cortical neurons per unit of visual field devoted to foveal input is much larger and contains more neurons than the network devoted to perifoveal vision [6e8]. Functionally, the fovea mediates the highest contrast sensitivity (CS) [9]. CS is a critical, low-level visual process involved in object recognition [10]. Electroretinography (ERG) and optical coherence tomography (OCT) [11,12] reveal retinopathy in PD. The amplitude of the pattern ERG (PERG) and the multifocal ERG (mferg) are dominated by foveal processing [13,14] and are reduced in PD. A comparison of different OCT studies is difficult because of differences in equipment and varying factory algorithms of analysis. In addition, as with many imaging techniques, OCT yields masses of data. Foveal CS contributes to both the detection and discrimination of visual stimuli [15]. It was originally suggested in 1984 [16] that visual impairment is relevant to gait and balance, which are among the major movement defects in PD. Some studies [17e20] suggest that even in early stages of PD, patients demonstrate decreased stability and that changes in the visual input impair their postural control [18e20]. However, it remains to be seen whether or not impaired foveal visual processing in PD causes visuospatial deficits [19] and upsets multisensory interactions necessary for maintaining gait and posture [18e20]. The potential effect of impaired vision on motor impairment in PD is beyond the scope of this review. have amplified signals that capture attention for relatively small stimuli in a complex, large-scale visual scene [21]. In summary, three important properties of foveal vision include: a) high CS; b) high density of cortical neurons per unit area of this part of the visual field; and c) preference for these signals in visuospatial attention in realistic visual scenes containing many potential targets. 2. Retinal architecture and visual functions of the fovea The retina is a multilayered structure with distinct neural elements in each layer (Fig. 1). Receptors in the outer layers convert light energy to a change in membrane potential, signalling the bipolar cells in the middle of the retina, which then synapse onto ganglion cells in the inner layer. The axons of the ganglion cells leave the retina as the optic nerve. Lateral and feedback connections are mediated by two other cell types: horizontal cells and amacrine cells. Amacrine cells, including those that use the neurotransmitter dopamine, are located in the layer closest to ganglion cells (Fig. 2). The inner retina (IRL) includes the NFL, the ganglion cell layer and the inner plexiform and inner nuclear layers, while the outer retina (ORL) consists of layers starting from inner nuclear layer up to and including the retinal pigment epithelium. The special role of the fovea in anthropoids is subserved by its architecture. All vertebrates have a central retinal area, often called area centralis, mediating the highest visual acuity. It is called fovea because in several species it appears as a depression or pit in crosssection. The fovea of anthropoids is characterised by a high density of cone photoreceptors underneath the depression. In humans, this area is approximately 1.5e2 mm in diameter within the larger macula lutea where the retina thins out greatly. The term macula refers to a circular area including and surrounding the fovea. The 1. Foveal vision e the retina and primary visual cortex Fixating on visual targets for better scrutiny enlists the fovea, which is the region of the retina containing the highest concentration of photoreceptors. Additional foveal processing occurs within the retina before ganglion cell axons (comprising the nerve fibre layer e NFL) leave the retina to carry this information to the brain via the optic nerve. The density of the photoreceptors per unit area in the fovea provides exquisite visual acuity, and interactions between these and other retinal neurons are responsible for high CS. The fovealeextrafoveal distinction is more complicated when moving stimuli are used, but evidence shows that motion processing may also be affected in PD, although this is not discussed in this chapter. Retinal output from the fovea is further augmented by the computational properties of the visual cortex, which participates in object recognition. Foveal vision is subserved by a small area of the retina that is greatly magnified in the representation of the central visual field in the visual cortex [6e8]. Furthermore, object selection and visual attention is mediated by the processing that takes place in the connections between visual cortical areas in the occipital lobe and additional visual cortical areas in the temporal and parietal lobes. Receptive fields of neurons located nearest to the foveola Fig. 1. The foveal region of the human retina. The colour coded illustration is an average, derived from a spectral domain optical coherence tomography (SD-OCT) study. Different retinal layers can be visually identified above the layer labelled retinal pigment epithelium. For further explanation and not averaged, but for single passage OCT-s see Fig. 10. (From an on-line Wikipedia entry).

3 I. Bodis-Wollner / Parkinsonism and Related Disorders 19 (2013) 1e14 3 feedback neural interactions facilitate maximum quantum efficiency in spatiotemporal vision. Maximum sensitivity is obtained from pooled information, including summation and centreesurround interaction (discussed in more detail later) from area roughly 1 in diameter [22]. This means that for optimum efficiency, a visual stimulus needs to cover more than the very centre of the fovea, the foveola. Maximum efficiency is quantified with CS, the minimum contrast needed for detection, as a function of the pattern element size covering the central visual field. Pattern element size is conveniently defined as spatial frequency of the repetitive pattern, such as a sinusoidal grating. Temporal frequency is the number of times in a second a stimulus changes its colour, luminance or spatial contrast. Visual acuity, a well-proven and often-used clinical measure to assess the optical integrity of the eye, only measures spatially limited neuronal organisation compared with the spatial contrast transfer function. The spatial transfer function of the overall system, including optics and neural apparatus, can be divided into optical and neural parts [9]. Studies in neurological [24,25] and ophthalmological patients show that impaired spatial CS is not predictable from visual acuity (discussed in more detail later). A specific deficit of spatiotemporal CS was first shown in PD patients [26]. Fig. 2. TH labelled dopaminergic amacrine cell of the retina. Structural organization of dopaminergic neurons in the rat retina. (a) Vertical section through the rat retina showing a DA perikaryon and DA processes visualized by immunocytochemical staining with an antibody against tyrosine hydroxylase (high signal intensity). On top, the typical location of the cell body is shown in the inner nuclear/- inner plexiform layer (see Fig. 9) boundary. The arrow points to an ascending process which reaches the outer plexiform layer. (b) A horizontal view showing DA cell bodies, primary dendrites and fine processes disposed in rings in the distal portion of the inner plexiform layer the spatial distribution of the scarce but nearly evenly spaced array of dopaminergic amacrine cells and their extensive connections (inner plexiform layer) is shown. The 20 mm marker bar serves for both (a) and (b). gel, ganglion cell layer; inl, inner nuclear layer; ipl, inner plexiform layer; onl, outer nuclear layer; opl, outer plexiform layer. By permission of the author P. Witkowsky [3] and Documenta Ophthalmologica. very centre of the fovea is called the foveola. The layers and the presence of diverse neuronal types in the different layers surrounding the fovea depend on its distance from the very centre of regard. The ganglion cells and the innermost layers are pushed aside. At the very centre, the fovea has only four external layers: the internal limiting membrane, the outer plexiform layer, the outer nuclear layer and the photoreceptors [2]. Light first activates the photoreceptors. Their signals are transmitted to bipolar cells and some other second order neurons and finally reach the ganglion cells, the output of the retina. Between these three types of cells, there are many interconnections, both lateral and in feedback. An oblique outward shifting of all layers (except the pigment epithelium layer) begins at the foveola. The fovea has sloping walls (clivus) and contains a few rods in its periphery. The foveola contains the highest density of cones and is adapted to yield highest visual acuity, each central cone being connected to only one ganglion cell [22,23]. There are some recent suggestions that melanopsin-containing ganglion cells make feedback synaptic contact on to dopaminergic (DA) amacrines. At this point, the functional significance of this loop is not yet clear. 4. Retinal ganglion cells Kuffler (1953) [27] used spot-like visual stimuli to introduce the concept of centre and surround receptive fields in the ganglion cells of the mammalian retina. Later, sinusoidal gratings were used for a more analytical description of retinal ganglion cell-receptive fields [28]. The inverted U-shaped response curve of individual ganglion cells to different spatial frequencies (SFs) is a quantitative measure of the antagonistic centre/surround retinal ganglion cell receptive field organisation, and of the linearly operating [28] central foveal neurons (Fig. 3). The introduction of this approach to studying ganglion cell-receptive fields has provided an analytical technique for studying human vision in neurological and ophthalmological disorders [29]. The CS of human observers to sinusoidal gratings over a specified range of SFs can be understood as an envelope-transfer function of ganglion cells in the fovea and subsequent foveal visual pathways in humans and other primates. CS measures visual functions in the entire retina-to-visual cortex circuitry; thus, abnormal CS does not localise pathology to any particular part of the pathway. However, the CS function provides 3. Fovea The fovea was originally thought to subserve only high-acuity- (smallest detectable image) and two-point discrimination (smallest distance between two small dots). However, work during the early sixties introduced an approach to quantifying optical and neural transfer functions of the eye [9] showing that lateral and Fig. 3. The centre-surround model of the receptive field (RF) organization of foveal ganglion cells. Spatial contrast sensitivity relies on the antagonistic interaction of centre and surround. (After Enroth-Cugell and Robson 1966, [28]). The model explains the tuned characteristics of contrast sensitivity (Fig. 6) and the ERG (see Fig. 7).

4 4 I. Bodis-Wollner / Parkinsonism and Related Disorders 19 (2013) 1e14 a bridge between retinal ganglion cell physiology and primate psychophysical and electrophysiological findings and it is most relevant to understanding processing deficits in the dopaminedeficient retina. As explained below, DA amacrines, located in the inner retina, provide signals to the front end of the retina for the centre and surround organisation of the eventual ganglion cellreceptive field [2]. 5. The effect of PD on foveal vision 5.1. Visual evoked potentials The very first evidence of visual involvement in PD and its relationship to dopaminergic dysfunction was obtained with visual evoked potential (VEP) measurements [30]. Stimulation with patterned stimuli results in occipitally recordable signals now known as VEPs; this measure was first introduced for clinical diagnosis of multiple sclerosis (MS) and optic neuropathy. The relatively large amplitude signal (5e10 mv) is dominated by foveal stimulation of the central 6e8 of the visual field [31,32]. In PD, the optimal spatial frequency is above 2.0 cycles per degree (cpd) [33,34]. Routine clinical diagnosis uses a checkerboard stimulus of up to 15 diameter, with an SF of roughly 1 cpd (Fig. 4). Although Fig. 4. Illustrates a so-called checkerboard pattern frequently used for visual stimulation in clinical ERG and VEP studies. The size of each individual check is about 1 in diameter. The pattern is centred on the foveola in a schematic drawing of the retinal surface, as seen by ophthalmoscope. Notice that with this pattern element size, as shown here individual ganglion cells of the fovea would receive local flicker or flickering edge stimulation. For PD diagnosis the optimal check size is a factor of 2e3 smaller. Smaller checks or higher spatial frequency gratings evoke in an equal size field larger amplitude ERG signals than this pattern. (See Fig. 7 and text) (From: Bodis- Wollner I, Ghilardi M and Mylin L (l986) the importance of stimulus selection in VEP practice: the clinical relevance of visual physiology. In: Frontiers of Clinical Neuroscience, Vol. 3, Evoked Potentials, R. Cracco and I Bodis-Wollner (eds), New York: Alan R. Liss, Inc., pp. 15e27). the stimulus field is large, the VEP amplitude is dominated by the central region of the retina. With even lower SF (larger check sizes) each check is larger than the largest receptive field centre of foveal neurons. Hence, the stimulus presents temporal modulation of luminance change or an edge rather than spatial contrast to individual foveal neurons. In PD, the VEP is considerably delayed and somewhat attenuated, and can show interocular asymmetry [35]. However, as with abnormalities in CS, VEP deficits can represent pathology in any portion of the visual pathway from the retina to the cortex. The specific role of DA neurons [36,37] in foveal vision was first shown in a VEP study [30]. The specific role of dopamine is also supported by the effects of dopaminergic blockade in PD and in non-pd observers, by CS, ERG and VEP [33e35,38e48] measures (discussed below). Since the publication of these and subsequent VEP studies, evidence of a functional deficit in visual processing in PD has been complemented by neuropharmacological and retinal electrophysiological studies of retinal impairment in PD [3e5] Contrast sensitivity A contrast sensitivity (CS) curve represents the minimum contrast needed to detect a pattern (usually a sinusoidal grating) as plotted over a range of spatial frequencies. CS tuning reflects the fact that the highest sensitivity is obtained over a specific range of middle spatial frequencies, whereas CS declines to either side, resulting in an inverted U-shaped curve. The curve may be distorted by neurological or ophthalmological pathology either at its end point (equivalent to a measure of visual acuity), at all spatial frequencies, or over a particular range of spatial frequencies [24,25,49]. Consistent with VEP studies, behaviourally measured CS is most reduced in PD above 2 cpd. [26,43,44,50,51]. While visual acuity is only minimally affected in well-corrected PD patients, these patients loose CS in the fovea to patterns for which normal observers demonstrate the most sensitivity (i.e., that need the least contrast to detect) [52]. Contrast detection is ultimately a cortical process; however, CS impairments can be separated from cognitive impairments [53]. Nevertheless, ocular pathology may affect the interpretation of a CS deficit. For example, an increased incidence of pathology affecting the ocular media, lens and cornea is evident in the typical age group of PD. [54]. These pathologies are foremost in degrading visual acuity. As mentioned, however, visual acuity is only minimally affected in well-corrected PD patients [26]. In contrast, a more than a threefold CS deficit is unlikely to result from optical defects, and is more likely to be neural [26]. Conversely however, some studies left open the possibility that deficits attributed to PD were actually caused by optical reasons or by other eye pathology. A detailed review of all CS studies in PD is however beyond the scope of this review. On the whole, most studies attested to the effect of PD on CS for medium SF stimuli. In addition, CS deficits change in tandem with the state of DA efficiency (oneoff) [26], and no oneoff effect on the pupil is evident in PD (Fig. 5). Furthermore, levodopa treatment [50,51] improves CS loss in PD, consistent with VEP and ERG studies [40,41,43,44]. CS tuning is enhanced by nomifensine, a drug that increases the amount of synaptic norepinephrine and dopamine by blocking their reuptake [42]. Finally, CS loss is more profound when the stimulus grating (spatial contrast) is also temporally modulated at 4e8 Hz[26,55,56]. Optical degradation is unlikely to cause specific temporal deficits. The difference between input and output can be analysed by using incommensurate sinusoidal temporal frequency inputs [56]. For

5 I. Bodis-Wollner / Parkinsonism and Related Disorders 19 (2013) 1e14 5 Fig. 5. Spatiotemporal contrast sensitivity in a foveally centred 9 field is shown for 6 spatial and 6 temporal frequencies in a three dimensional surface plot for a PD patient. Top left: the three dimensional surface is highest (highest CS) around 3 cycles per degree and between 4 and 8 Hz modulation rate. Slicing horizontally versus vertically through this surface shows the spatial versus temporal sensitivity profile. This ST CS represents roughly the optimal (foveal) sensitivity for the human eye in photopic conditions. (Watson BA, Barlow HB, Robson JG. What does the eye see best? Nature. 1983; 302:419e21). Top right: ST of CS when the same patient is in the on condition. Notice a lack of a peak. Bottom: the difference 3-D surface shows the effect of dopamine on ST CS in photopic foveal vision. (From the author: Bodis-Wollner I, et al. Visual dysfunction in Parkinson s disease: Loss in spatiotemporal contrast sensitivity. Brain. 1987; 110:1675e98). By permission of Oxford University Press. instance, an input at 4.19 Hz and one at 4.32 Hz each generates second harmonics and both first- and second-order intermodulation frequencies. They and second-order intermodulation frequencies can be separately analysed. For instance, even though the fundamental frequency responses of the primary input are unaffected, their intermodulation frequencies are affected. Thus, the affected nonlinear product must arise at an early stage of processing at the input level in the retina where nonlinear interactions take place. This site is likely to be in the retina [56], suggesting that DA deficiency probably affects temporal processing early on, at the retina [56]. In clinical practice, CS is easily testable with wallmountable printed charts. These charts have less specificity but allow rapid screening that can be valuable in instances of PD The electroretinogram PERG represents the average response of all ganglion cells covering the area stimulated [57], with a small contribution from the preceding inner layers [58]. A tightly tuned electroretinogram (ERG) reflects the optimal balance between the antagonistic centre/surround organisation of foveal ganglion cell-receptive fields. When the PERG response amplitude of healthy observers is plotted against SF in humans or monkeys [59], the resulting curve is non-monotonic (Fig. 6); the inverted U-shaped curve shows a peak that is similar to the CS curve The exact location of the peak along the SF axis is dependent on many factors such as field size, luminance and technique. The important point is the non-monotonicity of the function; tuning is attenuated or absent in the PERG of PD patients. Impaired retinal ganglion cell processing in the fovea of patients with PD has been demonstrated with PERG and mferg [60e69]. PERG studies in idiopathic PD reveal mid-sf and thus spatial tuning loss (Fig. 6), similar to the ERG in the monkey model of PD, created by the neurotoxin methyl-phenyl-tetra-hydropyridine (MPTP) [70] and 6-hydroxydopamine [71]. D2 receptor blocking causes similar changes in the human [47] and in the monkey PERG [72]. In humans with PD and in the monkey model, the PERG shows lack of tuning, similar to the CS curve in PD [26]. ERG tuning becomes evident in levodopa-treated patients [45]. Experimental neuropharmacological treatments in healthy controls and the MPTP monkey model yielded somewhat more detailed and converging evidence, complementing the human data. Based on the effects of selective D1 and D2 receptor blockers on the PERG of the monkey, we [73,74] modelled the DA circuit that modulates the balance of the centre and surround antagonism of foveal ganglion cells. The model quantifies how DA amacrine cells, although sparsely distributed, control the tuning of foveal ganglion cells via separate D1- and D2-linked receptors. DA amacrine cell dysfunction may result in absent SF tuning (Fig. 6) [70,71]. The mferg (Fig. 7) technique provides a large array of stimulus elements, typically in a 20e30 field. Each stimulus is modulated at a slightly different temporal sequence. The final ERG output, obtained with a single recording electrode, can be separated into as many responses as are in the stimulus display. In essence, each small region is tagged by a temporal sequence. A topographical representation of mferg response then reflects topographical differences in retinal activity. The mferg [14,58,75] reflects outer and also inner retinal responses and receives a strong contribution of the foveal inner retina [58] Colour vision. Colour vision is commonly tested with clinical ophthalmological tests. When more rigorous psychophysical and electrophysiological

6 6 I. Bodis-Wollner / Parkinsonism and Related Disorders 19 (2013) 1e14 magnocellular pathways. Significant impairment in all three pathways was found, which was more marked along the protan/deutan (RG) axis than the tritan (BY). This pattern of losses sharply contrasts with that typically seen in ageing, which primarily affects the tritan axis. Ocular diseases in the elderly include glaucoma, where the tritan axis is affected most (but not exclusively). In Best macular dystrophy, the colour axis is most affected, depending on the stage of the disease. Colour vision deficits were originally described with psychophysical tools [76e79]. ERG and VEP [66,67,81] studies using isoluminant colour contrast stimuli yield evidence of the retinal versus retino-cortical primary visual pathway origin [66,67] of certain colour deficits in PD. Reduced chromatic and achromatic PERG have been shown in PD and not in multiple system atrophy (MSA) [65] Open questions about differences in ERG and CS Fig. 6. The inverted U-shape of the PERG. PERG amplitude as a function of spatial frequency in healthy controls and levodopa treated and untreated PD patients (After Tagliati et al., 1996). The normal PERG curve is tuned : it shows a peak, somewhat similar to the CS curve shown in Fig. 6. The normally strongly bandpass PERG amplitude function (top curve) shows low-pass shape in levodopa treated (middle curve) and untreated PD patients. The horizontal line, parallel with the horizontal axis, shows the noise level during the recordings. In this study a signal to noise ratio of above 1.3 was considered significant. Noise was established from the ERG recording obtained concurrently with the signal. The stimulation frequency was 7.5 Hz, the response was at even frequencies, the dominant one at 15 Hz. The bandwidth of the Fourier decomposition of the response was 0.5 Hz. Noise was quantified by the mean amplitude of the averaged adjacent non-harmonic components Notice a lack of tuning in untreated PD patients. These data show that dopamine has an essential role in spatial frequency tuning in the human retina. tests are used, the results suggest the involvement of different colour-opponent pathways in PD. Achromatic vision is conemediated through specific, segregated visual pathways. The parvocellular pathway is mediated by small retinal ganglion cells, RGCs (P cells), and terminates in the parvocellular layers of the lateral geniculate nucleus (LGN). The koniocellular pathway is mediated by bistratified RGCs and synapses in the interlaminar layers of the LGN. Chromatic information is transmitted by large RGCs (M cells) in the magnocellular pathway. Clinical, psychophysical and electrophysiological tests of colour vision [66,67,76e82] have all been applied to the PD population, but each test has drawbacks. The Farnsworth- Munsell 100 Hue test (FM) and the D-15 L Anthony test (D-15) are the most widely used clinical tests, requiring participants to arrange coloured discs into a smoothly graduated colour sequence. A thorough discussion of the use and limitations of these commonly used tests goes beyond the scope of this review. For example, there are some concerns about the limited quantification power and the variability in testeretest scores [76]. Nonetheless, PD patients demonstrate significantly higher error rates on the FM test than age-matched controls [76e79]. Less dramatic, but statistically significant deficits are also seen in colour discrimination tasks devoid of the motor requirements of the FM and D-15 tasks [66,67,77,78,81,82]. Silva and co-workers [82] probed chromatic and achromatic CS changes in PD using complex psychophysical measures designed to isolate parvocellular, koniocellular and While PERG and CS show a tuning deficit in PD, the specific spatial frequency of the peak response is nearly a factor of 2 different. Whether this is attributable to different foveal sampling areas for the two techniques is unknown. Another possible reason for this discrepancy may be that CS is established by quantifying threshold responses, whereas the ERG (and VEP) uses high-contrast stimuli. It is known that low- and highcontrast stimuli shift the stimulation between so-called parasol (large) and midget (small) ganglion cells. They differ somewhat in their sensitivity to contrast and receptive field size. They also differ in their respective retinal organisation. Furthermore, their densities differ as a function of distance from the foveola. The model of the action of retinal dopamine by Bodis-Wollner and Tzelepi [73,74] was restricted to a single type of neuron with strong centreesurround organisation. These are more likely to be morphologically midget cells. However, it has been reported [82], based on examining a number of visual functions, that PD affects all three types of cells (midget, magno- and konio-). It may be necessary to exploit stimulus conditions (for instance, retinal area stimulated) and temporal effects to understand why the ERG responds better to levodopa therapy than does the VEP [46]. An understanding of colour and luminance processing and the involvement of three types of ganglion cells in PD is not yet clear. Does it represent a non-selective loss of axons in the optic nerve? The relationship of motor symptoms and ERG abnormalities is contradictory [45,48,61,63]. The dependence of CS and VEP responses on orientation [83,84] of the stimulus has not been satisfactorily explained. 6. OCT and in vivo retinal morphology In the past decade, it has become possible to obtain highresolution in vivo images of the human retina using commercially available equipment. OCT is an optical interferometrictechnique used to image biological tissues. It can achieve very fine, submicrometreresolution employing near-infraredlight of very widespectrum sources emitting over a w100-nm wavelength range. OCT equipment creates colour-coded three-dimensional images. Gross retinal histology in humans can be quantified and imaged in vivo using time-domain (TD-OCT) [11] and, more recently, spectral-domain OCT (SD-OCT) [12]. Both are used to image the human retina. TD-OCT takes consecutive slice measurements, whereas SD-OCT emits multiple rays and creates interferometric data. SD-OCT has an imaging speed of w axial scans s 1, which is approximately 50 times faster than TD-OCT [85], which is still widely used. The overall image quality of SD-OCT is higher than that of TD-OCT, because of the increased speed of the SD-OCT, which eliminates many motion artifacts. The axial resolution of

7 I. Bodis-Wollner / Parkinsonism and Related Disorders 19 (2013) 1e14 7 Fig. 7. An illustration of the concept and method of multifocal ERG (MF-ERG) [14,75]. A. The stimulus array, composed of tightly packed hexagonal elements is shown centred on the foveola. Three perifoveal distances are shown. The centre most circle corresponds to the fovea B and C. each hexagon is stimulated on/off at a different temporal sequence, so-called M sequence. D. The ERG response is the sum of all single temporal functions. Since each individual hexagon is stimulated with a different sequence, spatial location is tagged. To create a spatial map, the composite function is cross correlated with each individual temporal sequence. Highest correlation pinpoints the hexagon which was stimulated with the particular sequence. E. The cross correlation data are represented as an amplitude map of the whole stimulated region. It is evident that the MF-ERG is dominated by foveal stimulation (after Sutter E. [75]). Notice that retinal thickness and MF-ERG amplitude negatively correlate in PD. ([90], see text). a TD-OCT is 8e10 um, while for a Fourier-domain it is w5 um, resulting in a more accurate representation of retinal topography. Different manufacturers provide clinical protocols intended for specific diseases of the optic nerve and retinal ganglion cells, such as glaucoma, and the macula such as in maculopathy, diabetic retinopathy and retinitis pigmentosa, to name a few. There are no commercially available programs for testing vision in patients with PD. When using the same equipment, OCT measurements have good reproducibility, but there are differences in actual values provided by equipments of different manufacturers [85]. The most popular OCT equipments are the time-domain Stratus (Zeiss, Oberkochen, Germany), the frequency-domain Cirrus (Zeiss), RTVue (Optovue, Fremont, CA, USA) and the Spectralis (Heidelberg Instruments, Heidelberg, Germany). As with every imaging technique, OCT yields masses of data. In order to use OCT as a clinical tool, it would be desirable to provide a reduced number of end point measures. This is not easily achieved. Part of the problem is generated by the uncertainty regarding the critical retinal area of measurement. Currently, most PD studies quantify the peripapillary NFL most applicable for glaucoma and other diseases of the optic nerve. Some PD studies use macular programmes designed for maculopathies and for diabetes. One can measure full retinal thickness or only a portion of it, divided into the inner and outer retina (for an image of the retina see Figs. 1, 2, 8, 9). Actual values of retinal thickness, either for the full retina or only for the outer or inner retina, have been provided by different instruments or are not identical. Part of the difference is due to different algorithms used by different manufacturers for establishing reference lines and segmenting retinal layers. A measured full retinal thickness value depends on the reference lines, usually for the full retinal thickness from the inner limiting membrane (ILM) to the top of the pigment epithelium. One type of equipment (RTVue) provides automated segmentation of the inner retina, that is, the thickness from ILM to the inner nuclear layer, which only includes the NFL, ganglion cell and inner plexiform layers. Other types of equipment allow manual layer-by-layer segmentation with a calliper. Many still use the lower resolution TD-OCT [85e87]; however, the reference line is rarely straight and

8 8 I. Bodis-Wollner / Parkinsonism and Related Disorders 19 (2013) 1e14 and a differential increase in the thickness of different inner layers, consistent with actual histology [88] Retinal thinning in PD Fig. 8. Histology of the human fovea. Section through the adult fovea embedded in glycol methacrylate and stained with azure II and methylene blue. Both the GCL and INL are absent from the foveola. Retinal blood vessels are present on the upper foveal slope (arrows) but not on the lower slope or in the foveola. At the very centre of the primate retina, the fovea, not all layers are present. The very centre of the fovea, the foveola, contains only a thinnest sheet of photoreceptors. From the foveola, in a roughly 1.8 mm diameter central foveal pit, the other layers of the retina are displaced concentrically Notice the gradual emergence of different inner retinal layers on the slope (clivus) of the foveal pit. The inner nuclear layer (INL) and then the ganglion cell layer (GCL) emerge. Interrupted lines at radial distances of 0.75e1.50 mm indicate the emergence of the vascular zone. (Courtesy of ARCH OPHTH, figure from Provis and Hendrickson). This region was identified using ROC where inner retinal thickness measurements show the highest sensitivity and specificity for discriminating PD patients from controls. The fovea illustrated in a control and in a PD subject as provided by the OCT equipment. Left: Foveal retina in healthy subject (65 years old). Right: Foveal retina in PD patient (67 years old). Same magnification. In the OCT reconstruction of the retina, one can distinguish from the nerve fiber (ganglion cell axon) layer (towards the receptors on the bottom), the IRL sub-layers, ONL and OPL, receptors and pigment epithelium. Notice the thinner retina and wider foveal pit on the PD patient [91]. The foveola is the centre of the pit. Around the central foveal pit the other layers of the retina are displaced concentrically both in the control subject and in the patient. There are few studies directly quantifying foveal slope in PD. Visual inspection of the foveal pit in several studies shows this image of the widened pit in PD and in about 7 percent of healthy controls [114]. However, for potential clinical use of the foveal OCT, large scale studies will need to be done. the measuring callipers are not necessarily orthogonal at each point of the baseline. Even after comparing four different pieces of SD- OCT equipment, further development of segmentation algorithms and quantitative features are needed to assist clinicians in objective use of these newer instruments [86]. While different studies yield discrepant absolute thickness values in vivo architecture of the fovea, they appear to be consistent. All images show a gradual (rather than abrupt) thickness change from the foveola Thickness values of the retina are based on one of three methods: a) an automated equipment programme; b) manual measurements from the images provided (manual segmentation); and c) actual data from the volume of individual voxels. Currently, this method is available with only one type of equipment, as actual data must be obtained by a company representative. Most studies evaluate NFL thickness with automated programmes. Some studies in healthy controls include for measurements more than e retinal region for the analysis. In PD subjects there are far too few studies which attempted to quantify diverse ROI-s. Thinning of the peripapillary retinal NFL was first shown by Inzelberg et al. (2004) in 10 patients [89]. NFL thinning was subsequently shown by some [67,89,93,95,96] but not in all [92,94] studies. Absolutely no difference was found between the mean values of controls and 51 patients in one study [94]. This negative study used a low-resolution TD-OCT. However, the equipment difference hardly explains the negative data, as TD-OCT was also used in three positive studies. The infero-temporal segment of the NFL was thinned in four studies. However, the sensitivity and specificity of the OCT for individual eyes need to be established. Macular volumes are commonly quantified using the software based on the Early Treatment Diabetic Retinopathy Study (ETDRS) protocol, designed for large-scale diabetic studies. One TD-OCT and two SD-OCT studies reported volume loss. Altintas et al. [90] were the first to quantify macular volumes in PD eyes. They also reported a deficit in the inner (foveal) macula (Zone 1). Thinning in the fovea was reported in two studies but the methods used were different. We examined foveal architecture by exporting voxel-by-voxel values into a MATLABÔ computational environment. We studied 45 eyes of 24 early PD patients and 31 eyes of 17 control subjects in the same age range [96]. A difference of the thickness of the inner but not the outer foveal retina was significant. The diagnostic yield of 1.5 SD was about 78%. However, in another SD-OCT study of 10 patients [92], foveal thickness manually measured from the OCT images was the same in PD and in control subjects [92]. In this study, macular thickness in the outer superior subfield was 2.8% Fig. 9. Dopamine s action in the retina. A sketch of the model of pre-ganglionic dopaminergic control of the centre-surround receptive field of foveal ganglion cells (after Bodis- Wollner and Tzelepi 1998, 2002 [73,74]). Notice presynaptic cross-inhibition from the D2 pathway to the D1 feedback route to the photoreceptors.

9 I. Bodis-Wollner / Parkinsonism and Related Disorders 19 (2013) 1e14 9 thinner (P ¼ 0.026), while the outer nasal and inner inferior subfields were respectively 2.8% (P ¼ 0.016) and 2.7% (P ¼ 0.001) thicker compared to published normal values. Thickening of the retina has not yet been reported in any study of PD, but NFL thickening has been noted in acute Leber s hereditary optic neuropathy (LHON), a mitochondrial disorder affecting the first step (complex I) in the respiratory chain, as in models of PD and postulated for PD Caveats of OCT use (a) Some OCT studies of the slope, depth and rim of the fovea do come to the conclusion, that dark-skinned individuals have a thinner retina than Caucasians and there may be also gender differences [97e102]. All of these studies except one [102] used TD-OCT. The reported effects varied though: some studies found one and not the other variable to have an effect on macular volumes and foveal thickness. Other studies [103e105] found no effects of either variable. The methods of analysis quantified macular volumes, using the ETDRS protocol and computed full retinal thickness (FRT) and different individual methods of specifying foveal morphology. An analysis of comparing these methods however goes beyond this review. Mean macular volumes average out potential differences in the contribution of different retinal layers and cell types. However, one study [106] quantified segmented foveal thickness and examined gender and age effects. Inner retinal thickness negatively correlated with age. The age group of controls and PD patients is generally equated in all OCT studies. However, the parametric effect of age on inner retinal thickness will need a large-scale study as some investigators believe that PD represents accelerated ageing. (b) The comparability of measurements varies with different manufacturers. Few studies used the same OCT equipment. Different manufacturers provide different canned measures and manual measurements are not standardised either across equipments or investigators [107e110]. Whether using company standards or manual measurements, the definition of the reference line is variable and thus absolute thickness values are not easily comparable from one type of equipment to another. All programmes quantify the NFL thickness around the optic disc. Given the effect of PD in the preganglionic retina, we approached the problem by looking at the origin of the NFL in the macular area and quantified the thickness of the ganglion cell complex (GC), which includes the cell bodies of the macular nerve fibres. Besides ganglion cells, the GC also includes the inner plexiform layer, consisting of processes and interconnections of amacrine cells. (c) The selection of control and PD subjects can be biased. A diagnostic application of OCT necessarily involves patients with concurrent ophthalmic or medical conditions. They should not be excluded; a statistical model approach using multivariate analysis should be developed instead. In the search for a quantitative clinical tool to use as a biomarker for PD (i.e., SD-OCT), a larger normative database is needed for the parametric analysis of the effects of age, race, gender and axial length on the inner foveal retina. Neurodegenerative diseases increase with age and a number of them have been shown to affect the retina. The selection of our patients and controls was based on identical and rigorous ophthalmological and neurological inclusion and exclusion criteria. Our study points to the importance of strict neurological selection criteria in aged subjects presumed to be controls. We are not aware of any relevant published OCT study in which controls would have been screened for neurological conditions. We recommend rigorous standards when building a large, gold standard normative database of SD-OCT in aged controls. Especially glaucoma needs to be considered in the differential diagnosis of retinal NFL thinning. Glaucoma is a disease of the elderly and is often discovered on careful eye examination only, as early on many patients do not complain of any visual loss. According to some studies, the incidence of glaucoma is about 23% among PD patients [111] compared to its incidence of 5e12% in the healthy, same age group [112]. Glaucoma needs to be considered in the exclusionary criteria for both patient and control subjects. Some subjects without visual changes and others with typical glaucomatous optic-disc changes reveal narrow angles and vulnerability for glaucoma. This needs to be carefully considered in selecting both patients and controls for studies measuring RNFL thickness. The existence of interocular differences mandates that not only one eye be randomly selected for study in PD. A study [113] in 100 British healthy individuals quantified minimum foveal thickness (MFT), central 1 mm average foveal thickness (AFT) and total macular volume. AFT was symmetrical between the eyes of each individual and there was no effect of age, gender or race on interocular differences. In 110 eyes of 57 subjects, including Europeans and Afro- Caribbeans, Tick et al. [114] reported a high degree of symmetry in all examined morphometric parameters (CFT, pit depth, pit diameter and mean retinal thickness in healthy controls), while we [91] reported interocular asymmetry in PD. These studies indicate that one should not select randomly one or another eye in patients or controls in OCT diagnostic studies of PD. In addition, one needs to consider that therapy with different DA agents may have differential effects on retinal thickness [115]. 7. Discussion 7.1. Dopaminergic circuitry of the retina and foveal processing in PD Direct morphological evidence shows that DA cells [116,117] are abnormal in the postmortem PD retina. They have low dopamine content and fewer tyrosine hydroxylase (TH) labelled neurons [118]. There is also evidence that dopaminergic amacrine cells are affected in the MPTP monkey retina as dopamine metabolites are reduced [119]. DA neurons are the main elements of an intraretinal centrifugal pathway [120]. Through dopaminergic signalling [121e123], via dopamine receptors in the outer retina [124e128], dopamine optimises spatial transfer properties of the retinal ganglion cells. The effect of DA cell dysfunction on ganglion cell receptive-field organisation has been quantified in a formal model [73,74]. According to this model, DA amacrine cells control ganglion cell receptive-field organisation with a balancing action of D1- and D2-type receptors on centre/surround interaction of primate and human ganglion cells. The model includes presynaptic crossinhibition between the D1 and the D2 pathway. The experimental results in man and monkey show that for a low-dose sulpiride, PERG amplitude is higher at low SFs than in the untreated eye and the response to the peak spatial frequency is attenuated [47,59,72,73]. Consistent with lower vertebrate data, it has been postulated that a feedback pathway arising from the proximal retina makes contacts on distal retinal elements [124e128]. This pathway is in the position to regulate nonlinear local luminance responses of the ganglion cell (Fig. 10) [56,129]. Such a regulatory mechanism in the distal retina could possibly influence both luminance and chromatic properties of diverse ganglion cells [66,67,76e81]. A circuit abnormality prior to the ganglion cells is not contradictory to a deficit in the processing of all three classes of retinal ganglion cells in PD [82].

10 10 I. Bodis-Wollner / Parkinsonism and Related Disorders 19 (2013) 1e14 Our model [73,74] of how dopamine deficiency affects the retinal circuitry is based on experimental pharmacological results in humans and monkeys [2,5]. However, the model is impoverished by not taking into account the possible localisation of D1 receptors in the inner plexiform layer, including amacrine cells [127,128] in the primate. Further understanding of dopaminergic synaptic connections in the inner retina may lead to a more detailed and refined model and better understanding of the retinogram. One study [68] correlated mf-erg results with OCT thinning in PD. Such morphological/functional correlation is reassuring. Amplitude of the mferg is determined by inner and outer retina and it is dominated by the fovea (Fig. 7). There is an apparent paradox however: the OCT results show not only the IRL but also ganglion cell losses in the PD retina. Ganglion cells are not dopaminergic, yet the PERG abnormalities reflecting ganglion cell responses both in man and in monkey respond to levodopa or dopaminergic ligands [59,62,65,70,72]. This apparent paradox can be explained by the details of the retinal circuitry. DA cells control receptive-field organisation of foveal ganglion cells via a feedback loop from the inner to the outer retinal system, which contacts the receptors (Fig. 9). The focus of this review is foveal vision. Both the PERG [13,65] and the mferg [14,75] (Figs. 7 and 10) are dominated by the foveal retina. The peak of the PERG response function corresponds to the CS of retinal ganglion cell properties in the central 2 of visual field [13,14]. Thus, it is conceivable that degraded control of dopaminergic feedback to receptors is expressed in attenuated mferg and their effect on ganglion cells on the PERG. However, a direct comparison of the mferg and the PERG on the same eyes in PD has not been conducted. Altintas et al. [90] examined 17 PD eyes and reported a correlation between disease severity and inner foveal thickness, and not with full macular volume or peripapillary NFL thickness. This result is consistent with the study of Hajee et al. [91], which shows about 15e20% inner retinal thinning in the fovea. Perhaps this modest loss is the reason for the absence of disc pallor in PD despite the reported ganglion cell damage [90,91]. However, the 15e20% loss in total IRL thickness does not necessarily cause a minor visual loss. The most compelling evidence of foveal retinal thinning was demonstrated by showing a correlation with the mferg [68] Differential diagnosis of retinal thinning It needs to be established whether OCT findings contribute a quantitative measure of the early diagnosis of PD by perhaps joining a constellation of early signs [48,93,130e132]. The OCT results in PD are potentially relevant for the ophthalmologist. Fig. 10. Three dimensional plot of multi-focal electroretinogram of left and right eye of a PD patient.the multifocal electroretinogram in each eye of a patient with Parkinson Disease. Compare the healthy control data (Fig. 8) with this image. Notice on the colour coded depth, in difference to Fig. 8, of a healthy control. Top: right eye; bottom: left eye. Notice also the asymmetry of the ERG profile of the foveal pit between the eyes: (From: Moschos M et al. Morphologic changes and functional retinal impairment in patients with Parkinson disease without visual loss European Journal of Ophthalmology 2011, 21; 24e29 with permission).