Alzheimer disease (AD) is a progressive neurodegenerative disorder characterised by impaired memory and cognitive function. A proportion of patients present with visual symptoms which could result from either anterior or posterior visual pathway dysfunction. Retinal and optic nerve abnormalities have in recent years been studied intensively in Alzheimer disease (AD) and are reviewed in this chapter.
2. Evidence of retinal nerve fibre layer thinning in AD
2.1. Histological evidence
Retinal ganglion cell (RGC) degeneration or optic neuropathy is one of the features of AD that has been identified in several histological, imaging, and electroretinogram (ERG) studies. Analyses of neuronal numbers in the RGC layer of severe AD patients and age-matched control subjects have revealed extensive neuronal loss throughout the entire retina in AD when compared with control eyes: the mean (RGC) number is 696,871 in AD which is significantly less than that of the controls (1,095,904) representing a loss of 36% (p<0.004) (Blanks et al., 1996b). The findings are in good agreement with other histological studies (Hinton et al., 1986; Sadun & Bassi, 1990; Trick et al., 1989) and compatible with diminished contrast sensitivity which may be secondary to afferent visual pathway dysfunction in AD patients (Crow et al., 2003). A post-mortem study revealed widespread optic nerve fibre degeneration with thinning of the retinal nerve fibre layer and reduced ganglion cell numbers in AD patients aged 76-89 years (Hinton et al., 1986), any potential correlation with the degree of cognitive impairment was not studied. In one study the axonal loss was prominent in the posterior part of the optic nerve suggesting that the process involved may be one of retrograde degeneration of the retinal ganglion cell axons (Sadun & Bassi, 1990). Neurons in the ganglion cell layer (GCL) in AD patients are reduced by 25-40% throughout the entire retina (Blanks et al., 1996b). The greatest density of RGCs is located in the macular region and the total numbers of ganglion cell fibres in the fovea/parafovea are reduced by 25% (p < 0.001) in AD as compared to normal eyes (Blanks et al., 1996a). The loss in the central retina is greatest in the temporal area, which is surprisingly different from that found in the periphery (Blanks et al., 1996a). The neuronal loss reaches its peak in the superior and inferior quadrants of the peripheral retina (Blanks et al., 1996b). Unlike what is found in normal ageing, the RGC loss in AD is not related to age (Blanks et al., 1996a, 1996b). It was also found that both small- and large-diameter RGC fibres are affected equally throughout the retina in AD eyes in some studies (Blanks et al., 1996b; Curcio & Drucker, 1993) but only large M-cell degeneration has been identified by the others (Sadun & Bassi, 1990; Trick et al., 1989; Blanks et al., 1989, Miller, 1990). The total number of astrocytes in the GCL was found to be 16% greater in the AD patients than in controls but the increase did not reach statistical significance (Blanks et al., 1996b). The ratio of astrocytes to neurons in the GCL is significantly raised in the AD retinas, resulting from both an increase in astrocytes and the decrease in neurons (Blanks et al., 1996b). A study comparing patients with at least a 4 year history of severe AD aged 67-86 years and age-matched controls revealed no significant difference in the number of RGCs (Curcio & Drucker, 1993). This is not the only study to find no significant difference in RGC number between AD patients and controls (Davies et al., 1995). It is notable that there is no data whether AD patients in this review are familial or sporadic type.
2.2. Imaging evidence
By employing fundus photography, RNFL degeneration was observed in AD patients as compared to age-matched control subjects (Hedges et al., 1996; Tsai et al., 1991). A study utilizing scanning laser ophthalmoscopy (SLO) has demonstrated a reduction of optic nerve fibres in AD patients when compared to age-matched controls (Danesh-Meyer et al., 2006). However, another SLO study showed no difference (Kergoat et al., 2001). It was shown that there was no significant difference in mean overall RNFL thickness and RNFL thickness in each quadrant between the AD and age-matched controls. The patients selected for this study were classed as mild to moderate dementia with a range of mini mental-state examination of 11-29 (mean 21.57) and a mean duration of 3 years. Optical coherence tomography (OCT) has been recently developed and has found wide application in neurology and ophthalmology both in the clinic and in research (Jindahra et al., 2009). It has been employed to measure RNFL thickness in several AD studies. In AD eyes, the mean of overall peripapillary (Parisi et al., 2001, 2003; Iseri, et a., 2006; Paquet et al., 2007; Lu et al., 2010; Valenti, 2007) and macular RNFL thickness (Paquet et al., 2007) as well as the mean of total macular volume (Paquet et al.,2007) measured by OCT are lower when compared with age-matched control subjects. It has been proposed that the retinal involvement might have occurred early in the course of the disease as the peripapillary RNFL loss is identified in mild cognitive impairment (MCI) (Paquet et al., 2007). The reduction of RNFL thickness is statistically significant in MCI (Paquet et al., 2007), mild AD (Parisi, 2003; Iseri et al., 2006; Paquet et al., 2007), and moderate to severe AD (Parisi, 2003; Iseri et al., 2006; Paquet et al., 2007), compared to controls. No difference was found between the results observed in MCI and mild AD patients (Paquet et al., 2007) but the measurements of RNFL thickness seen in moderate to severe cases are significantly thinner than those in MCI cases (Paquet et al., 2007). The RNFL thickness was found to be thinner than the controls in all four quadrants in MCI (Paquet et al., 2007) and AD cases (Parisi et al, 2001, 2003; Paquet et al., 2007). However in one study the temporal quadrant was unaffected (Iseri et al., 2006) and significant thinning was found only in the superior quadrant in mild to moderate AD cases compared with controls in a study and a case report (Berisha et al., 2007; Valenti, 2007). Yet another study, the RNFL was found to be significantly thinner than in normal subjects in the superior and inferior quadrants in AD patients and hence the patients no longer showed the double peak RNFL pattern (Lu et al., 2010). The thinning in the nasal and temporal quadrants did not reach statistical significance (Lu et al., 2010). It seems that the retinal damage due to AD may be localized preferentially to the vertical quadrants (Lu et al., 2010) and as such may be considered to mimic the pattern seen in glaucoma (Figure 1, Table 1 & 2). By using OCT it has been shown that the reduction in total macular volume is highly related to the severity of cognitive impairment (MMSE) (Iseri et al., 2006).
2.3. Electrodiagnostic evidence
The results of electrodiagnostic testing are conflicting. ERG studies have failed to demonstrate changes in AD patients (Justino et al., 2001; Kergoat et al., 2001, 2002; Davies et al., 1995). Scotopic and photopic electroretinograms and oscillatory potentials in patients with mild Alzheimer disease were compared with normal individuals in one study (Justino et al., 2001). The amplitude and latency of a and b waves in mild AD patients were normal, reflecting intact function of the outer retina. The oscillatory potentials were also unremarkable in this study. Pattern electroretinogram (PERG) recordings have shown a significant delay in P50 and N95 implicit times and reduction in both P50 and N95 amplitudes in mild to severe AD when compared with the results obtained in control eyes (Parisi et al, 2001, 2003). This might indicate that the dysfunction lies in both ganglionic and preganglionic elements (Parisi et al, 2001, 2003). The delayed P50 and N95 implicit times and the reduced P50 and N95 amplitudes are significantly correlated with the reduced overall mean of the RNFL thickness measured by OCT (Parisi et al., 2001, 2003). No significant difference was found in the latency of the pattern visually evoked potential (PVEP) P100 of AD patients and control subjects (Iseri et al., 2006). The normal PVEP responses revealed no evidence for any abnormality of primary visual cortex or of optic nerve function despite considerable RNFL loss (Iseri et al., 2006). However, some earlier studies did find abnormalities of the flash VEP (Wright et al., 1986; Norman et al., 1995) but probably not useful clinically (Coburn et al., 2003). However using PERG and PVEP, another study revealed a reduction in amplitude of N95 and increased latency of P100 wave in most AD eyes (Krasodomska et al., 2010).
2.4. Optic disc morphology in AD eyes
A large cup-to-disc ratio, thin rim area and volume are identified in AD eyes compared to age-matched normal eyes (Tsai et al., 1991; Danesh-Meyer et al., 2006). In one of these studies, the patients had a MMSE result of 21±4 taking an upper limit of the vertical cup-to-disc ratio as 0.42 gives a sensitivity of 0.45 and specificity of 0.84 (Danesh-Meyer et al., 2006). Pallor area to disc area ratio did not significantly differ between AD patients and normal subjects in one study (Tsai et al., 1991). However patients with a higher ratio had a higher Alzheimer disease assessment scale and longer duration of illness (Tsai et al., 1991). The changes are not in a uniform pattern for all AD patients (Berisha et al., 2007).
2.5. Information from Down syndrome
It has been shown that all adults with Down syndrome (DS) over 35-40 years old who had autopsies performed have AD pathology in their brains i.e., beta amyloid plaques and neurofibrillary tangles (Malamud, 1972). Amyloid precursor protein gene on the locus of the proximal part of the long arm of chromosome 21 is over-expressed in DS patients (Goldgaber et al., 1987), leading to AD development (Prasher et al., 1998). The DS brain pathology is comparable to AD brain and may be useful in further AD studies (Hof et al., 1995). Regarding visual functions, DS patients have impaired colour discrimination, stereoacuity, and contrast sensitivity, similarly to AD patients (Rocco et al., 1997). Moreover abnormal spatial vision in DS children has been detected without other ophthalmologic abnormalities (Suttle & Turner, 2004). A literature review of children with Down syndrome age 0-16 years revealed that refractive error, strabismus, poor acuity, nystagmus, and blepharitis were common ophthalmologic findings whereas cataract and glaucoma were less common (Creavin & Brown, 2009). A pattern reversal VEP study demonstrated significantly longer P100 latency and smaller amplitude in DS patients (16/36 cases) as compared to age-matched controls (Kakigi et al., 1993). By employing achromatic transient VEP, children with DS had small or undetectable N75 but normal latency as compared to normal developing children (Suttle & Turner, 2004). Patients with DS also responded abnormally to chromatic transient VEP (Suttle & Lloyd, 2005). As far as we are aware, there has been no current report about RNFL measurement in DS eyes.
3. Hypotheses to explain RNFL thinning in AD
Three hypotheses to explain retinal ganglion cell fibre damage in AD have been proposed.
3.1. Is the retina affected by AD pathology directly?
AD pathology might develop not only in the cortex but also in the retina, perhaps in the ganglion cell layer (Lu et al., 2010). Beta amyloid, amyloid associated proteins, tau and amyloid precursor protein are expressed in the human retina at the level of ganglion cells and fibres in older eyes and in the retinal pigment epithelium in retinitis pigmentosa and age-related macular degeneration (Löffler et al., 1995). However neurofibrillary tangles, senile plaques, and amyloid angiopathy have not been identified in the retina even in association with extensive neuronal loss (Leuba & Kraftsik, 1994; Blanks et al., 1989, 1996b). A further study identified neither neurofibrillary degeneration nor amyloid angiopathy in AD patients’ retinas (Hinton et al., 1986). Glial fibrillary acidic protein (GFAP) localized to Muller cells and astrocytes in the GCL is increased in AD eyes; as is found in retinal injuries and in the AD brain (Blanks et al., 1996b). It indicates that the retinal degeneration is accompanied by a glial response as GFAP is a major cytoskeletal component of astrocytes (Blanks et al., 1996b). Amyloid beta or Abeta deposition was found mainly in the outer and inner plexiform layer in the retina of the APPswe/PS1DeltaE9 transgenic (tg) mouse model of Alzheimer disease (Perez et al., 2009). Likewise, Abeta plaques with increased retinal microvascular deposition of Abeta and neuroinflammation in Tg2576 mouse retinas were detected chiefly from the GCL to the inner plexiform layer and some plaques were also identified in the outer nuclear layer, the photoreceptor layer, and the optic nerve (Liu et al., 2009). Abeta deposits reduced with abeta vaccinations (Liu et al., 2009). Hyperphosphorylated tau was demonstrated in areas adjacent to the plaques (Liu et al., 2009). Furthermore, abeta deposition was observed in the cytosol of lens fibre cells along with equatorial supranuclear cataracts in AD patients as compared to age-matched controls (Goldstein et al., 2003). No supranuclear cataracts were identified in any normal individual in this study (Goldstein et al., 2003). Like AD, an evaluation of lens in patients with Down syndrome revealed supranuclear opacification with accelerated supranuclear abeta accumulation (Moncaster et al., 2010).
3.2. AD and glaucoma
Wostyn has proposed a link between glaucoma and AD suggesting that an abnormal high trans-lamina cribrosa pressure difference in AD eyes has led to glaucomatous optic neuropathy (Wostyn et al., 2009). Glaucoma is characterized by a progressive loss of RNFL and a resulting visual field defect. Elevated IOP is a strong risk factor but not all patients with glaucoma have high IOP (Johanson et al., 2008; Berdahl et al., 2008b). The subgroup is classified as normal tension glaucoma (NTG) (Johanson et al., 2008). The retinal ganglion cell fibres in NTG eyes might be vulnerable to normal IOP as it is relatively high in the NTG eyes. The cause of NTG is still unknown. Recent studies have revealed that AD patients may have a higher risk of developing glaucoma than normal subjects (Bayer et al., 2002a; Tamura, 2006) and that glaucoma in AD patients tends to be more progressive than glaucoma in non-AD cases (Bayer & Ferrari, 2002b). A case control study (Chandra et al., 1986), investigating all death certificates (1,930,627) for the United States in 1978, compared 7195 cases who had senile and presenile dementia as the cause of death with other patients who died from other conditions. Gluacoma was associated with these demented patients with odd ratio of 2.6. Early RNFL loss in glaucoma occurs in the temporal inferior and temporal superior regions (Hoyt et al., 1973; Pederson & Anderson, 1980; Tuulonen & Airaksinen, 1991; Jonas et al., 1993) as found in some of the AD studies described above. Trans-lamina cribrosa pressure differences may have resulted in the glaucomatous like RNFL changes. A study (Jonas et al., 2003) revealed that lamina cribrosa forms a barrier between the intraocular space and retrobulbar space. The lamina cribrosa has been found to be thinner in glaucomatous eyes than in control eyes (Fig. 2). The outer part of the cribrosa that is directly in contact with pia mater or indirectly with cerebrospinal fluid (CSF) was significantly thinner in the glaucomatous eyes as compared to the controls, and the shortest distance between the intraocular space and the CSF space was significantly less in the glaucoma patients. The optic disc is situated close to this area. Trans-lamina cribrosa pressure (the pressure gradient across the lamina cribrosa) is derived from the intraocular pressure minus the retrobulbar CSF pressure (Jonas et al., 2003). Normally the intraocular pressure is higher than that in the CSF. Abnormal pressure from either side of the lamina cribrosa may be involved in the pathogenesis of several ocular and neurological conditions. In vivo, high IOP glaucoma can damage the optic nerve head and very low IOP can cause swollen discs.
Reduced ICP in patients with normal tension glaucoma (NTG) could cause abnormal trans-lamina cribrosa pressure (Berdahl et al., 2008a). Trans-lamina cribrosa pressure was significantly greater in patients with primary open angle glaucoma (POAG) and NTG than in normal individuals (Ren et al., 2010). CSF pressure in severe AD patients tends to be disproportionately low (Silverberg et al., 2006) and therefore may create a situation where there is relatively high IOP in their eyes (still within a normal reference range). Glaucomatous-like RNFL changes may then be expected to occur. There is evidence of choroid plexus (CP) degeneration in AD brains and their CSF production is affected (Serot et al., 2003). The choroid plexus consists of villi covered by a single layer of ciliated cuboidal epithelium and extends through the lateral, 3rd, and 4th ventricles, acting as a blood-CSF barrier (Serot et al., 2003). It is contiguous with ependyma; produces CSF; synthesizes several molecules; and carries nutrients from blood to the CSF (Serot et al., 2003; Silverberg, et al., 2001). Two-thirds of the CSF secretion is derived from the CPs, the remainder coming from brain interstitial fluid drainage, which is produced by the capillary-astrocyte complex found in the blood brain barrier (Johanson et al., 2008). The production rate of the blood-CSF barrier is substantially greater than that of the blood brain barrier (Johanson et al., 2008). Another source of CSF production is likely to be ependyma lining the ventricles (Pollay & Curl, 1967). CSF reabsorption (Johanson et al., 2008) occurs along sleeves of subarachnoid spaces surrounding cranial nerves that enter the nose and eyes; through the cribriform plate, nasal mucosa, and cervical lymphatic system eventually. CSF is also drained along spinal nerve arachnoid pathways. Arachnoid villi in the dural sinuses absorb the CSF when ICP is elevated. CSF pressure (Johanson et al., 2008) is normally higher than venous pressure in the dural sinuses. It is steady when CSF formation and reabsorption are balanced. CSF pressure measured by lumbar puncture in a lateral recumbent position is directly proportional to CSF production rate and outflow resistance. CSF pressure is determined by hydrodynamic and haemodynamic parameters. Regarding hydrodynamic factors, reduced CSF production or increased outflow resistance will decrease CSF pressure. In ageing the CP epithelium becomes atrophic; its basement membrane thickens; and CSF secretion decreases by 50% (Serot et al., 2003). These changes in AD choroid plexuses appear more pronounced than in
normal ageing and additionally stroma fibrosis has also been demonstrated in AD (Serot et al., 2003). Abeta proteins have been detected in the choroid plexus in AD brain (Kalaria et al., 1996). Ig and C1q depositions are frequently found along the basement membrane of the plexus in AD brains, suggestive of immunological processes in this location (Serot et al., 2003). As a consequence the choroid plexus cannot function normally (Serot et al., 2003). In young adults, the CSF production rate is 0.4 ml/min or 500 – 600 ml per day, the CSF volume is 150 ml, and the CSF turnover rate is 4 volumes per day (Johanson et al., 2008). In AD patients, on the contrary, the CSF production rate is 0.2 ml/min, the CSF volume is 250 ml due to brain atrophy, and the CSF turnover rate is 1.2 volumes per day (Johanson et al., 2008). In addition to the affected CSF production, the resistance of CSF outflow in AD is becoming greater for there is evidence of abeta depositions in the meninges (Silverberg et al., 2003; Hamano et al., 1997; Kalaria et al., 1996). Further studies are needed to confirm low CSF pressure in AD patients; to establish a relationship between the severity of cognitive impairment, brain atrophy, ventricular volume, CP morphology with CSF pressure; and to establish whether the trans-laminar cribrosa pressure difference plays an important role in the pathogenesis of RNFL thinning in AD.
3.3. Retrograde trans-synaptic degeneration secondary to cortical pathology
Lastly, we hypothesize that the RGC loss in AD could be partly due to retrograde trans-synaptic degeneration. RGC loss following an occipital injury, which is a consequence of retrograde trans-synaptic degeneration of geniculo-cortical towards retino-geniculate pathways; and anterograde degeneration of cortico-geniculate pathway, has been identified in the visual pathway (Cowey, 1974; Mehta & Plant, 2005b; Jindahra et al., 2009; Bridge et al., 2011). Neuronal loss, neurofibrillary tangles (NFT) and senile plaques have been identified in several neocortex areas including primary visual cortex (Pearson et al., 1985; Leuba & Kraftsik, 1994). Senile plaques were also identified in the LGN (Leuba & Kraftsik, 1994; Leuba & Saini, 1995) along with NFTs and degenerating axons or threads in the white matter underlying area 17 (Leuba & Saini, 1993; Leuba & Saini, 1995), reflecting a spread of the degeneration along the cortico-geniculate axons (Leuba & Kraftsik, 1994; Leuba & Saini, 1995). In a study (Leuba & Saini, 1995), senile plaques were found more in the parvocellular layer of the dLGN than the magnocellular layer, the interlaminar zones, and the optic radiation. No neuritic degeneration (NFT, neuritic plague, and thread) was demonstrated in the LGN in this study, suggesting mainly amyloid deposition in this area. The finding was in good agreement with another that showed mild tau pathology in the LGN of AD patients (Dugger et al., 2011). In addition, senile plaques and NFTs were detected in the pyramidal and non-pyramidal cells in layer 5 and 6 of the primary visual cortex (Leuba & Saini, 1995). The degeneration in the visual cortex varied greatly among individuals possibly due to different AD subtypes as presented below. Neuronal loss, glial cell proliferation, NFT, and neuritic plaque (NP) deposition have been demonstrated in visual cortex, area 17 in particular, in AD patients with mean age of 76.1 +/- 8.1 years when compared with age-matched controls (Leuba & Kraftsik, 1994). The tangles in layer 5 are twice in number as in layer 3 in the occipital lobe except area 18 and the neuritic plaques are found in all layers in one study (Pearson et al., 1985). In a study (Hof et al., 1989), AD patients with Balint syndrome which is characterized by optic apraxia (impairment of target pointing under visual guidance), ocular apraxia (inability to shift gaze to a new visual target), and simultagnosia (perception and recognition in a small part of visual field) had greater NFT density and NP area in all cortical layers of area 17 and 18 as compared to AD patients without Balint syndrome. The NFT density in the superior frontal cortex of AD with Balint syndrome appeared much less than that in AD without Balint syndrome. Patients with Balint syndrome from stroke did not show NFT or high NP numbers in the visual cortex. The mean number of Meynert cells in layer 5 and 6 of area 17 in AD with Balint syndrome was significantly lower than that in AD without Balint syndrome. This study suggested a disruption of occipito-parietal connections or dorsal stream in these AD with Balint syndrome cases. Another study used SMI32-immunoreactive staining technique which represented pyramidal neurons in a small subset of total neuronal population (Hof & Morrison, 1990). It had been shown that the Meynert cell (the largest SMI32-ir neurons) counts in area 17 and 18 were significantly lower in AD patients than in age-matched controls only in a small magnitude. The loss appeared more pronounced in temporal and prefrontal cortices. The neuronal loss was confined to area 4b in area 17 and layer 3, 6 in area 18. The findings might have indicated the degeneration of projections of Meynert cells in these regions to area V5 that is responsible for visuospatial skills. NFT in area 17 and 18 were less numerous than area 9 and 20 in AD patients (Hof & Morrison, 1990). NFT were dominant in layer 2-3 in the visual cortices whereas in layer 5 in area 9 and 20. NP were numerous in layer 2-4 with the greatest density in layer 4 of area 17 in layer 2-3 in area 18.
A study (Arnold et al., 1991) revealed the distribution of NFT and NP among 39 cortical regions in 11 AD patients, having mean age 80.2 years (range 63-88 years) and mean duration of disease 7.5 years (range 3-15 years). It had been shown that NFT in the limbic and temporal lobes were substantially higher than the frontal, parietal, and occipital lobes. NPs were evenly distributed throughout the cortex with the highest density in the temporal and occipital lobes. When comparing NFT among visual cortices namely area 17, 18, and 20 (inferior temporal gyrus) in 8 AD patients aged 48-82 years, the number of NFT was low in area 17 but progressively increased in area 18 and 20 respectively, which paralleled to the hierarchical visual organization (Lewis et al., 1987). NFTs were found predominantly in layer 3 and 5, which contained cortico-cortical and cortico-fugal projecting fibres (Lewis et al., 1987). A substantial number of NPs was identified equally in all three regions. They were present across all cortical layers (Lewis et al., 1987). A study (Kiyosawa et al., 1989) of AD patients with and without impaired visual functions i.e., figure copying, colour vision tested by isochromatic plates, and steropsis showed no change in their primary visual cortices in 18F-fluoro-2-deoxyglucose positron emission tomography (PET) as compared to the results in age-matched controls. Additionally AD cases with impaired visual functions showed significantly decreased glucose metabolism in visual association and inferior parietal areas compared with the controls. AD patients with good vision showed no significant change in these areas. No neuronal loss in area 17 of AD brains was demonstrated in another study (Mountjoy et al., 1983).
There is increasing evidence of RNFL thinning or RGC loss in patients with AD but the relationship between the degree of cognitive impairment and the degree of RNFL loss has not been established yet. There are a few possibilities that could explain the findings. These include AD change in the retina, abnormal trans-lamina cribrosa pressure, and retrograde trans-synaptic degeneration. The degenerative changes in the brain and retina vary among AD patients because of different AD subtype, severity, and duration. It seems that the RNFL measurement has a good potential to be a monitoring tool in AD patients in the near future. Further investigations are required to understand more about AD pathology in these areas.