Age-related macular degeneration and cerebral amyloid angiopathy have similar pathologies from cholesterol-APOE-amyloid-β-complement mediated inflammation

Abstract

Age-related macular degeneration (AMD) and Alzheimer’s disease (AD) are neurodegenerative conditions that afflict millions of elderly people around the world. AMD is a progressive retinal disorder that leads to central vision loss whereas AD primarily causes cognitive decline and behavioral changes. While each disease has distinct clinical manifestations, the accumulation of extracellular amyloid-β is a common histopathologic finding. Similarly, cerebral amyloid angiopathy (CAA), a vascular condition that can exist independent or with AD, is characterized by the accumulation of amyloid-β in cerebral blood vessels. While significant investigation of the pathophysiologic links between AMD and AD has been conducted, the underlying similarities and differences in the pathobiology of AMD and CAA has not been considered. In this review, we discuss the common pathological features of these two conditions. We then discuss the similar pathobiology that involves cholesterol metabolism, apolipoprotein E, amyloid-β, and complement mediated inflammation. At the same time, we discuss key differences in their pathobiology. This discussion sheds new perspective and insights of their pathobiology.

1. Introduction

Age-related macular degeneration (AMD) and Alzheimer’s disease (AD) are multifactorial, chronic, progressive neurodegenerative diseases that share pathological and biochemical features (Jabbehdari, Oganov et al. 2024). The accumulation of extracellular amyloid-β, a 38–43 amino acid peptide, in the brain is a defining characteristic of AD. The discovery of amyloid-β in drusen, a hallmark sign of AMD, has prompted significant investigation into common pathogenic pathways with AD. Several studies have shown that the risk of AD is increased among those with AMD and vice versa (Keenan, Goldacre et al. 2014, Lee, Larson et al. 2019, Wen, Wan et al. 2021, Wen, Wan et al. 2021). Over the past decade, AMD and AD researchers have independently identified similar dysregulated pathways involving oxidative stress, mitochondrial function, autophagic/lysosomal function, lipid metabolism, inflammation, and immune activation, including alterations in the complement pathway. Each of these mechanisms can damage their respective tissues, the retina in AMD, and the brain in AD.

Cerebral amyloid angiopathy (CAA) is a common cause of cortical hemorrhage in older adults, characterized by amyloid-β deposition in small to medium-sized blood vessels (Revesz, Holton et al. 2002). CAA and AD are mechanistically linked by the role of amyloid-β and can exist by itself or coexist in the same patients (Greenberg, Bacskai et al. 2020, Yin, Sun et al. 2025). Despite these discoveries, a deep understanding of the key pathophysiologic events that guides the development of effective treatment for CAA and AD has remained elusive. The clinical evidence for a link between CAA and AMD is limited (Qiu, Ding et al. 2018, Anisetti, Stewart et al. 2023). However, the histopathologic findings in CAA and AMD surrounding amyloid-β deposition is similar in cerebral blood vessels and the retinal pigmented epithelium (RPE)-Bruch’s membrane (BrM), respectively. These findings raise the possibility that identifying additional commonalities and differences in the pathobiology of these diseases will enhance our understanding of their pathogenesis. In this review, we describe the similar histopathologic features of AMD and CAA. We then discuss the prominent, linked roles of cholesterol, apolipoprotein E (APOE), amyloid-β, and complement mediated inflammation in the pathobiology of these diseases. By discussing their common and distinct mechanistic differences, we present new insights into the pathobiology of AMD and CAA.

2. Age-Related Macular Degeneration

Nearly 300 million individuals worldwide and 20 million in the United States suffer from AMD, the fourth most common cause of blindness and the leading cause of vision loss worldwide in those older than 55 years old (Wong, Su et al. 2014, Jonas, Cheung et al. 2017). AMD is a chronic progressive retinal disease that affects the macula and is classified into non-neovascular (dry) and neovascular (wet) forms (Fleckenstein, Schmitz-Valckenberg et al. 2024). While central vision loss is the primary clinical manifestation, nonophthalmic manifestations such as impaired cognitive performance, including verbal fluency and memory loss, have been reported, which suggests that AMD extends beyond the retina (Woo, Park et al. 2012, Rozzini, Riva et al. 2014, Zhuang, Madden et al. 2021). The prevalence of AMD increases with age, with a 15% prevalence in those 65–74 years old, 25% if 75–84 years old, and 30% if ≥85 years old (Klein, Peto et al. 2004). A prospective cohort study found that the overall estimated incidence of AMD was 5.3–4.1% in Caucasians, 4.5% −2.2% in Chinese, 3.3%−0.8% in Hispanics, and 1.6%−0.4% in Blacks (Fisher, Klein et al. 2016).

This review will focus on non-neovascular or “dry” AMD given the overlapping pathology with CAA. Dry AMD is characterized by a slow decline in central vision that is preceded by loss of retinal sensitivity, contrast sensitivity, and delayed dark adaptation (Eisner, Stoumbos et al. 1991, Chen, Fitzke et al. 1992, Eisner, Klein et al. 1992, Midena, Degli Angeli et al. 1997, Owsley, Huisingh et al. 2015). Clinically, patients with dry AMD present with BrM accumulations called drusen, subretinal deposits called reticular pseudodrusen (RPD), and RPE abnormalities (Figure 1) (Chew, Clemons et al. 2014). The density and/or size of drusen within the macula correlates with disease severity. Histopathologically, drusen are accumulations within the inner collagenous layer of BrM that are composed of a heterogeneous mixture of lipids, minerals, lipoproteins, and proteins (Curcio, Johnson et al. 2009). RPD are subretinal deposits that contain vitronectin, CD59, apolipoprotein E (APOE), unesterified cholesterol, IBA-1 expressing immune cells, and are associated with altered photoreceptor outer segment and RPE morphology (Rudolf, Malek et al. 2008, Ebrahimi, Fijalkowski et al. 2013, Greferath, Guymer et al. 2016). The RPE pigmentary abnormalities reflect its morphologic and functional degeneration. Over time, dry AMD can advance to geographic atrophy (GA) with loss of the functional retinal network (Thomas, Mirza et al. 2021). Histopathologically, photoreceptor remnants are adjacent to BrM due to RPE loss, and the choriocapillaris is atrophic. When GA involves the fovea, visual acuity loss is significant.

Figure 1.

A patient with intermediate AMD who progressed to geographic atrophy. Color fundus photo and OCT image of the left eye. (A) Intermediate AMD. showing drusen deposits. (B) Areas of outer retinal and RPE atrophy with hypertransmission defects into the choroid indicating progression. (C) Geographic atrophy with more severe and confluent areas of outer retinal and RPE atrophy with hypertransmission defects, as indicated by red arrowheads. The arrowheads in A and B indicate the region of GA that developed in C. (D) Histopathology of a different patient with geographic atrophy with loss of RPE and outer retina in the central macula (arrowheads). A druse (*) is seen. Bar = 25μm.

Risk factors for AMD include a combination of over 63 genetic loci that include multiple single nucleotide polymorphisms (SNPs) as well as environmental and social factors such as smoking, sunlight exposure, hypertension, and a Western or high-glycemic diet (Clemons, Milton et al. 2005, Gorman, Voloudakis et al. 2024). Along with data from epidemiologic studies, pharmacologic and clinical trials, these risk factors provide the rationale for preclinical studies designed to understand AMD pathobiology. At present, pathologic oxidative stress, lipid dysregulation, inflammation, and innate immunity including complement and inflammasome activations, extracellular matrix abnormalities, and mitochondrial and lysosomal dysfunction, are key pathogenic factors that cause degeneration of the photoreceptors and outer retina, RPE, and choriocapillaris (Campochiaro 2000, Donoso, Kim et al. 2006). While known, the relative pathogenic contribution of these factors at each disease stage is unclear.

3. Cerebral amyloid angiopathy (CAA): an amyloid-β angiopathy of cerebral blood vessels

CAA is an age-related angiopathy of small to medium blood vessels of the CNS and leptomeninges due to deposition of amyloid-β (Vonsattel, Myers et al. 1991, Vinters 1992, Revesz, Holton et al. 2002). The amyloid-β deposits induce vessel fragility that can lead to intracerebral hemorrhages (Vinters 1987). In addition, patients with CAA can have cognitive impairment (Vinters 1987). Given the combination of amyloid-β deposition and cognitive impairment, the association with AD, which is characterized by the buildup of amyloid-β in the extracellular space, creating plaques, as well as the accumulation of hyperphosphorylated tau (p-tau) tangles in the brain (Zhuang, Madden et al. 2021), has been made, especially since CAA can occur simultaneously with AD (Charidimou, Gang et al. 2012). However, CAA can occur independent of AD (Thal, Griffin et al. 2008, Arvanitakis, Leurgans et al. 2011). While the diagnosis of CAA is sometimes made using clinical and radiographic findings, the definitive diagnosis often requires a biopsy or postmortem examination of the brain (Arvanitakis, Leurgans et al. 2011).

4. CAA and AMD share similar histopathology

The accumulation of amyloid-β is a common histopathologic finding in CAA and AMD. In CAA, amyloid-β accumulates in the neurovascular unit, comprised of neurons, astrocytes, endothelial cells, vascular smooth muscle cells (Camacho, Moline et al. 2019), and pericytes, while in AMD, the RPE-BrM-choriocapillaris complex is principally involved. In CAA, amyloid-β co-localizes with APOE in blood vessels while in AMD (Okazaki, Reagan et al. 1979, Barelli, Lebeau et al. 1997, Poyuran, Mahadevan et al. 2019), amyloid-β and APOE are concentrated in drusen and BrM (Figure 2) (Johnson, Leitner et al. 2002, Isas, Luibl et al. 2010, Thompson, Reffatto et al. 2015, Ranjan, Kayastha et al. 2021).

Figure 2.

Amyloid-β deposition in human AMD and CAA. A 74-year-old man with non-neovascular AMD. (A) Basal deposits within the macular region (arrows) with partial loss of overlying RPE cells (H&E stain, original magnification 400X). (B,C) Patchy Amyloid-β immunoreactivity (brown) in the basal deposits (arrows, original magnification 400X). A surgical specimen from a 69-year-old woman with CAA. (D) Amyloid-β immunoreactivity in leptomeningeal vessels of a surgical CAA biopsy (Arrows). Diffuse amyloid plaques were also present in the cortex (arrowheads, original magnification 100X). (E) Thickened walls of a medium-size vessel in the CAA case (arrows, original magnification 400X). (F) Amyloid-β immunoreactivity in the leptomeningeal vessel (arrows, original magnification 400X).

5. CAA and AMD have similar and key differences in their pathobiology

In both CAA and AMD, alterations in APOE, cholesterol, amyloid-β, and complement mediated inflammation appear to drive their pathobiology. Genetic variations in APOE and complement genes increase the risk of both diseases. As described above, the key difference is that APOE, amyloid-β, and complement mediated inflammation pathobiology affect different cell types in CAA and AMD. The common etiologic mechanisms along with key distinctions are described below and is intended to elucidate new insights into the mechanisms of these two diseases with complex pathobiology.

5.1. Cholesterol and APOE are cornerstone pathogiolobic factors in CAA, AD, and AMD

Lipids are essential for the maintenance, repair, and growth of CNS neurons, a process that depends upon APOE, which binds lipids including cholesterol and cholesterol esters (Adibhatla and Hatcher 2007). In the brain, APOE is expressed mainly by astrocytes and microglia, which transports cholesterol to the extracellular space (Zhang, Mu et al. 2013). Once in the extracellular space, APOE can direct cholesterol in two directions. Cholesterol can be internalized by neurons through APOE receptors, such as the low-density lipoprotein (LDL) receptor (LDLR) and LDL receptor-related proteins (LRP1, LRP2 and LRP8), which are highly expressed by neurons, to maintain cellular homeostasis (Zhang, Mu et al. 2013). With neuronal stress, APOE is induced and binds lipids to facilitate their redistribution during cellular repair (Liao, Yoon et al. 2017, Long and Holtzman 2019, Husain, Laurent et al. 2021, Troutwine, Hamid et al. 2022). APOE is also involved with reverse cholesterol transport (RCT) by interacting with the transmembrane ATP-binding cassette transporters ABCA1 and ABCG1, which promote cholesterol efflux from cells to APOE, and then into HDLs for systemic removal (Hirsch-Reinshagen, Zhou et al. 2004, Wahrle, Jiang et al. 2004, Karten, Campenot et al. 2006).

In addition to its role in RCT, APOE is a component of LDL and very low-density lipoprotein (VLDL), which both have apolipoprotein B100 (APOB100) as their protein core (Campos, Perlov et al. 2001), and are essential for clearing lipids from the systemic circulation by the liver. Importantly, VLDLs carry more cholesterol than the RCT pathway. For example in response to lipid overload, cardiac myocytes secrete VLDLs to decrease cellular cholesterol and prevent lipoapoptosis (Unger and Orci 2002). In fact, when APOB100 lipoprotein secretion is prevented in cardiac myocytes, lipid accumulation is sufficient to induce lipoapoptosis that leads to death by heart failure in mice (Nielsen, Veniant et al. 1998, Nielsen, Bartels et al. 2002).

In early AMD, it has been hypothesized that the RPE becomes overloaded with cholesterol from the combination of the daily phagocytosis of the nearly 30,000 lipid rich photoreceptor outer segments and the consequences of a high fat western diet (Ershov and Bazan 2000). In response to increased cholesterol, the RPE produces and secretes APOE and cholesterol. To illustrate the importance of APOE in lipid processing, mice deficient of Apoe develop thickened BrM from lipid accumulation that correlates with decreased retinal function, as shown by impaired electroretinography (Dithmar, Curcio et al. 2000, Ong, Zorapapel et al. 2001).

With aging and early AMD, lipoprotein particles the size of VLDLs accumulate in BrM due to interaction with heparan sulfate to form a “lipid wall” (Curcio, Millican et al. 2001, Malek, Li et al. 2003, Curcio, Presley et al. 2005, Curcio, Presley et al. 2005, Li, Clark et al. 2006, Toomey, Pflugmacher et al. 2025). Age-related alterations in BrM such as the accumulation of advanced glycation endproducts (AGEs), alterations in matrix components such as heparan sulfate proteoglycans (HSPGs), CFH, pentraxin-3, and lipoprotein lipase, can impede the removal of lipoproteins that contribute to lipid retention (Handa, Verzijl et al. 1999, Ida, Ishibashi et al. 2004, Tian, Ishibashi et al. 2005, Cano, Fijalkowski et al. 2011, Toomey, Kelly et al. 2015, Wang, Cano et al. 2016, Landowski, Kelly et al. 2019, Toomey, Pflugmacher et al. 2025). Due to the high oxidative stress microenvironment, these retained lipoproteins get oxidized. Hydroxyapatite (HAP) shells form around these lipids, which then become coated with numerous proteins including amyloid-β (Ranjan, Kayastha et al. 2021). Both amyloid-β and oxidized lipids can elicit an inflammatory reaction, and the inflammatory proteins attach to these HAP shells during the growth of drusen (Thompson, Reffatto et al. 2015). Indeed, immunoglobulin light chains, vitronectin, and complement proteins such as CFH, C5, and the C5b-9 complex have been identified in drusen (Mullins, Russell et al. 2000, Malek, Li et al. 2003, Li, Presley et al. 2005, Wu, Fujihara et al. 2010, Fujihara, Cano et al. 2014).

While APOE is a key etiologic factor in both CAA and AMD, the lipid processing role of APOE is different in these diseases, which may explain in part, the pathologic changes specific for each disease. In the CNS, APOE is principally involved in RCT, which transfers lipids and toxins such as amyloid-β from the intra- to the extra-cellular space to prevent harmful cellular cholesterol and amyloid-β accumulation. Besides APOE, the RPE synthesizes and secretes APOB100 and VLDL-like lipoproteins at a level similar to cardiac myocytes (Malek, Li et al. 2003, Li, Clark et al. 2006, Fujihara, Bartels et al. 2009, Fujihara, Cano et al. 2014). In AMD, it is unclear to what extent RCT via HDLs or VLDL-like lipoproteins play a role in BrM lipid accumulation. HDLs in BrM are more pro-inflammatory and pro-oxidative than circulating HDLs from the same donor, which suggests a pathogenic role for AMD (Kelly, Grigsby et al. 2020). On the other hand, given that HDLs are small relative to VLDL-like particles, they are more likely, based on size alone without regard to their composition, to pass through BrM into the choriocapillaris. Thus, it is possible that RCT remains a protective response. In contrast, VLDL-like lipoproteins are retained in BrM due to age-related BrM alterations, and eventually trigger a pro-inflammatory response if they get oxidized. While this response is designed to remove these lipoproteins and amyloid-β, their retention promotes sustained inflammation that contributes to further protein and lipid aggregation and drusen growth. While RCT is a key factor in CAA, APOE may have an additional important role in the accumulation of VLDL-like lipoproteins in BrM during drusen biogenesis in AMD.

5.2. Amyloid-β biology in CAA and AMD

In the CNS, neurons, astrocytes, and vascular smooth muscle cells accumulate amyloid-β (Wisniewski and Wegiel 1994, Calhoun, Burgermeister et al. 1999, Frackowiak, Miller et al. 2003, Herzig, Winkler et al. 2004). Because it also binds amyloid-β, APOE regulates in part, amyloid-β abundance. In neurons, the predominant cell type that produces amyloid-β, amyloid-β accumulates through a combination of increased cholesterol from its synthesis and APOE mediated cholesterol import from astrocytes (Wang, Kulas et al. 2021). With increased cellular cholesterol, amyloid precursor protein (APP) associates with β- and γ-secretases in lipid clusters to generate amyloid-β, which either remains associated with lipid rafts in the plasma membrane or is released to the extracellular space (Wang, Kulas et al. 2021). When APOE-mediated cholesterol trafficking is limited in astrocytes, neuronal cholesterol is low, and amyloid-β accumulation is minimal. Thus, excess APOE-mediated cholesterol trafficking could result in the formation of insoluble extracellular plaques.

Amyloid-β accumulation is also influenced by its degree of degradation. Neurons degrade intracellular amyloid-β mainly through lysosomes (Li, Kanekiyo et al. 2012, Kanekiyo, Cirrito et al. 2013). Astrocytes and microglia remove the majority of extracellular amyloid-β by cellular uptake and degradation, a process that is APOE dependent (Koistinaho, Lin et al. 2004). In microglia, soluble amyloid-β is internalized by macropinocytosis while amyloid-β aggregates are internalized through phagocytosis, with subsequent degradation by lysosomes (Bamberger, Harris et al. 2003, Mandrekar, Jiang et al. 2009). In addition, after APOE binds to amyloid-β, the proteases neprilysin (NEP) and insulin-degrading enzyme are activated to degrade amyloid-β, either intracellularly or in the extracellular space after NEP is secreted (Wisniewski and Frangione 1992, Russo, Angelini et al. 1998, Jiang, Lee et al. 2008, Saido and Leissring 2012). Impairment of these degradative processes contribute to amyloid-β accumulation.

Like astrocytes in CAA, the RPE increases the expression of APP, and β- and γ-secretases to produce amyloid-β in response to elevated cholesterol and/or oxidative stress (Ohno-Matsui 2011, Wang, Ohno-Matsui et al. 2012, Guo, Alekseev et al. 2014, Francelin, Mitter et al. 2021). In the RPE, amyloid-β accululates in late endosomes and lysosomes, and is degraded by lysosomes, a process that can be increased experimentally by mTOR inhibitors (Lynn, Johnston et al. 2021, Jo, Lee et al. 2023). With amyloid-β accumulation, the lysosomes become impaired and the RPE cell undergoes functional impairment that includes an inability to process photoreceptor outer segments that have been internalized (Lynn, Johnston et al. 2021). As mentioned above, amyloid-β has been identified in drusen (Johnson, Leitner et al. 2002, Thompson, Reffatto et al. 2015, Ranjan, Kayastha et al. 2021). Dentchev et al. reported that amyloid-β was found in drusen from AMD donors, but not in small drusen from normal retinas (Dentchev, Milam et al. 2003). How the RPE releases amyloid-β is unclear. However, Kurzawa-Akanbi et al found that extracellular vesicles (EVs) secreted basally by iPSC-RPE cells derived from donors harboring the Y402H high-risk polymorphism induced amyloid-β fibril formation in the extracellular matrix that had similar morphology to drusen of human specimens (Kurzawa-Akanbi, Whitfield et al. 2022).

5.3. Cholesterol, APOE, and amyloid-β mediated inflammation is a pathobiologic factor in CAA and AMD

Polymorphisms in two regulators of complement activation, complement component (3b/4b) receptor 1 (CR1) and complement factor H (CFH) raise the risk of CAA and AMD, respectively (Edwards, Ritter et al. 2005, Hageman, Anderson et al. 2005, Haines, Hauser et al. 2005, Jakobsdottir, Conley et al. 2005, Klein, Zeiss et al. 2005, Rivera, Fisher et al. 2005). CR1, a negative regulator of both the classic and alternative complement cascade, is cell membrane bound and mediates cellular binding to debris and immune complexes that have activated complement, particularly in astrocytes of the neurovascular system, so that the opsonized immune complexes can be removed by phagocytosis (Crehan, Holton et al. 2012). Polymorphisms in CFH such as the Y402H variant, and other complement components increase AMD risk (Fritsche, Igl et al. 2016). CFH is a secreted protein that binds to cells and regulates the alternative complement pathway by regulating factor I induced C3b cleavage and decay of the C3-convertase (Rodriguez de Cordoba, Lublin et al. 1985).

In CAA, cholesterol, APOE, and amyloid-β induce inflammation to alter the functionally interlinked neurovascular unit that is comprised of neurons, astrocytes, endothelial cells, vascular smooth muscle cells, and pericytes. While the arterial endothelial cell is initially uninvolved, amyloid-β is deposited in the basal lamina and extends to the arterial internal elastic lamina in both human CAA and experimentally in animal models (Figure 3) (Rannikmae, Kalaria et al. 2014). Of note, a mouse genetic context prone to CAA (WSB/EiJ, Figure 3) was also uniquely susceptible to photoreceptor degeneration—suggesting potentially shared biological etiologies of CAA and AMD (Onos, Uyar et al. 2019, Marola, MacLean et al. 2025, Onos, Marola et al. 2025). Histopathologically, astrocytes are seen surrounding amyloid-β plaques that colocalize with APOE (Eikelenboom and Stam 1982, Kalaria and Grahovac 1990, Harr, Uint et al. 1996, Camacho, Moline et al. 2019, Taylor, Cisternas et al. 2020). In CAA mouse models, vascular amyloid-β deposition activates astrocytes. Transport across the blood brain barrier (BBB), phagocytosis, and enzymatic degradation are known clearance mechanisms. Notably, perivascular removal accounts for up to 25% of amyloid-β clearance, and is impaired in CAA, which substantially contributes to amyloid-β accumulation (Tarasoff-Conway, Carare et al. 2015, Xiang, Bu et al. 2015, Yuede, Lee et al. 2016).

Figure 3.

Cerebral amyloid angiopathy in mouse. The WSB/EiJ genetic background is susceptible to age-related cerebral amyloid angiopathy. (A) Brain hemispheres from WSB.APP/PS1 mice with severe cerebral amyloid angiopathy (arrowheads) at 8 and 14 months. Sections are stained with X34 (left: white; right: cyan) and immunoassayed for CD31 (right: red).

In the extracellular space, amyloid-β initially elicits a protective inflammatory response to avoid toxic accumulation (Frost and Li 2017). In response to amyloid-β deposition, astrocytes and microglia express and activate the complement system. For example, the anaphylatoxin C3a is formed, which by interacting with the C3a receptor (C3aR), can induce monocyte chemoattractant protein-1 to recruit microglia that degrade and remove unwanted material, including amyloid-β (Ohno-Matsui 2011, Guo, Alekseev et al. 2014, Lian, Litvinchuk et al. 2016, Lashkari, Teague et al. 2018, Pauly, Agarwal et al. 2019, Francelin, Mitter et al. 2021, Desai and Dugel 2022). With unresolved amyloid-β accumulation, complement activation is prolonged, shifting a protective immune response into a dysregulated one that causes tissue damage. In CAA, amyloid-β colocalizes with APOE, apolipoprotein A-I, complement proteins, and other inflammatory mediators such as serum amyloid P, α1-antichymotrypsin, and clusterin accumulate to cause smooth muscle cell degeneration (Eikelenboom and Stam 1982, Kalaria and Grahovac 1990, Harr, Uint et al. 1996, Camacho, Moline et al. 2019). The severity of vascular amyloid-β lesions is associated with abundant complement levels including formation of terminal C5b-9 complexes (Tanskanen, Lindsberg et al. 2005, Matsuo, Shindo et al. 2017). In addition, an astrogliotic fibrotic response ensues with increased heat shock protein family B (small) member 1 and modulator of VRAC current 1 (Schrader, Xu et al. 2024). Cumulatively, the vessel wall weakens, leading to rupture from C5b-9 complex deposition and vessel fibrosis (Kalaria and Sepulveda-Falla 2021, Kara, Gordon et al. 2023).

In response to amyloid-β, the RPE activates some but not all complement components, including CFH, CFD, and terminal complement components C5b-9. Together with systemically derived complement, the RPE activates complement to degrade and remove amyloid-β (Pauly, Agarwal et al. 2019, Desai and Dugel 2022). In addition, microglia, astrocytes, Müller glia, and vascular cells can express complement in the retina to contribute to complement mediated inflammation (Pauly, Agarwal et al. 2019). Importantly, the RPE expresses complement regulators CFH, CD46, and CD59 to maintain a controlled, protective complement response(Geerlings, de Jong et al. 2017). With aging and AMD, the expression by the RPE of these regulators is decreased (Ebrahimi, Fijalkowski et al. 2013). As a result, and/or due to impaired regulation by the Y402H CFH AMD risk variant, amyloid-β may elicit augmented complement activation that induces RPE injury, secondary photoreceptor degeneration, and deposition of cellular debris into BrM (Mullins, Aptsiauri et al. 2001, Moreira-Neto, Moult et al. 2018). Notably, amyloid-β inhibits complement factor I (CFI), a serine protease that restricts C3 and C5 convertase formation and inhibits the inactivation of C3b, both of which can augment or prolong complement activity (Lashkari, Teague et al. 2018, Papadopoulos 2020).

While complement activation is a common conceptual pathobiologic theme, the different complement components involved in CAA and AMD may influence their pathogenesis. CR1 is predominantly expressed by astrocytes and microglia while CFH is locally produced by the RPE.

For the most part, CFH is secreted while CR1 is membrane bound. These differences concentrate the complement response to the vasculature and RPE-BrM-choriocapillaris complex in CAA and AMD, respectively. Besides regulating complement activation, CR1 mediates endocytosis in phagocytic cells (Funkhouser and Vik 1999). In CAA, complement activation results in C3b binding to both amyloid-β and CR1, which initiates endocytosis of the amyloid-β-C3b-CR1 complex by CR1 for degradation (Rogers, Li et al. 2006, Qi and Ma 2017). Since the CR1 variant may impair the C3b binding sites, amyloid-β removal by CR1 binding cells such as erythrocytes for elimination by the liver or kidney, may be impaired (Rogers, Li et al. 2006, Brouwers, Van Cauwenberghe et al. 2012, Qi and Ma 2017).

Like CR1, CFH has complement independent functions. In BrM, CFH and lipoproteins compete for binding with HSPGs which could contribute to drusen formation. With aging, lipoproteins are retained in BrM by HSPGs, and can result in aggregation and subsequent drusen formation. The CFH Y402H high risk variant has impaired binding to the HSPGs and cell surfaces, and inhibits the binding of thrombospondin-1, a lipoprotein component, to CD47 on mononuclear phagocytes, which prevents their removal (Clark, Perveen et al. 2010, Calippe, Augustin et al. 2017). Furthermore, CFH can bind the lipid peroxidation product malondialdehyde, which then induces inflammation (Weismann, Hartvigsen et al. 2011, Toomey, Kelly et al. 2015). The CFH Y402H variant has impaired binding of malondialdehyde-modified lipids that triggers inflammation in BrM (Weismann, Hartvigsen et al. 2011, Landowski, Kelly et al. 2019). Thus, amyloid-β works in concert with cholesterol to both activate complement mediated inflammation and complement independent functions that accumulates around cholesterol filled HAP spherules during drusen biogenesis.

Besides RPE injury and deposition of cellular debris into BrM during drusen biogenesis, complement activation may contribute to choriocapillaris loss. In fact, choriocapillaris vascular density is inversely correlated with drusen volume (Edwards and Lutty 2021). Experimentally, amyloid-β is deposited more in choroidal vessels of Cfh deficient mice as compared to Cfh competent mice to the extent that blood flow is impaired (Aboelnour, Kam et al. 2016). A model of choriocapillaris loss due to increased complement activation including C3 and C5 deposition has been described. C5b-9 complexes are deposited in the choriocapillaris endothelial basement membrane prior to choriocapillaris dropout and in the pillars between choroidal vessels where drusen typically form (Mullins, Dewald et al. 2011, Whitmore, Sohn et al. 2015, Chirco, Tucker et al. 2016). Mice that overexpress Aβ42 and Aβ40 develop an AMD-like phenotype that includes drusen-like deposits, RPE hyperpigmentation, and retinal degeneration (Prasad, Zhu et al. 2017). In addition, aged APOE4 targeted replacement mice on a high fat high cholesterol diet develop an AMD-phenotype with amyloid-β and complement accumulation in BrM can be rescued by systemic anti-amyloid immunotherapy (Ding, Johnson et al. 2011). Thus, amyloid-β mediated complement activation may contribute to RPE degeneration, drusen biogenesis, and choriocapillaris loss.

5.4. APOE-related pathways regulate both cholesterol and amyloid-β metabolism

The APOE genetic risk in AMD and CAA/AD provides insights into important differences in their pathobiology. APOE has different isoforms due to two SNPs that cause amino acid substitutions in the APOE protein at positions 112 and 158 (Davignon, Gregg et al. 1988, Thakkinstian, Bowe et al. 2006). The wildtype and most common form, APOE3 (Cys112 Arg158) differs from APOE4 (Arg112 Arg158) and APOE2 (Cys112 Cys158). These SNPs modify risk for both AMD and AD including CAA (Jansen, Savage et al. 2019, Viturino, Neto et al. 2021). APOE4 is the strongest genetic risk factor associated with AD/CAA and related dementias (Corder, Saunders et al. 1993, Saunders, Strittmatter et al. 1993, Seshadri, Drachman et al. 1995, Slooter, Tang et al. 1997, Liu, Li et al. 2012, Hu, Wan et al. 2025). AD mouse models with the addition of APOE4 had exacerbated CAA and relevant comorbidities, including microhemorrhages, cognitive decline, and neuritic dystrophy (Fitz, Wolfe et al. 2020, Sawmiller, Koyama et al. 2023, Grenon, Papavergi et al. 2024, Hu, Wan et al. 2025, Onos, Marola et al. 2025).

In contrast, APOE4 is mildly protective for AMD—although this protection may be due in part to a drop effect of APOE4 patients with AD—while it raises risk for CAA and AD, and APOE2 increases risk for AMD and decreases risk for CAA and AD (McKay, Silvestri et al. 2011, Liu, Liu et al. 2013, Rebeck 2017, Jansen, Savage et al. 2019, Zhang, Xia et al. 2023). While APOE4 may confer a decreased risk for AMD, smoking may mitigate this effect (Levy, Lavalette et al. 2015, Xiying, Wenbo et al. 2017, Hu, Quinn et al. 2021). Nevertheless, the opposite risk profiles for AMD and CAA of these isoforms provide mechanistic insights into their pathobiology.

The APOE4 isoform increases the risk of CAA/AD relative to APOE2 and APOE3 in part, because the effects of APOE on cholesterol and amyloid-β metabolism are isoform dependent (Rannikmae, Kalaria et al. 2014). In the CNS, neurons, microglia, and astrocytes with the APOE4 isoform accumulate cholesterol ester more than with the APOE3 and APOE2 isoforms due to increased de novo cholesterol synthesis and upregulated cell surface lipid receptors that mediate cholesterol uptake (Tcw, Qian et al. 2022). As a result, cholesterol ester levels in APOE4 microglia are similar to cells with cholesterol ester storage disorders (Chan, Oliveira et al. 2012, van der Kant, Langness et al. 2019, Andreone, Przybyla et al. 2020, Nugent, Lin et al. 2020, Litvinchuk, Suh et al. 2024). At the same time, APOE4 is less efficient than APOE3 at facilitating cholesterol efflux from both astrocytes and neurons, an effect due to both differences in intermolecular dimerization of APOEs and its intramolecular domain interaction that promotes HDL particle generation (Michikawa, Fan et al. 2000, Gong, Kobayashi et al. 2002, Gong, Morita et al. 2007, Minagawa, Gong et al. 2009). As in astrocytes, APOE4 RPE have increased de novo cholesterol synthesis and cell surface lipid receptors to facilitate cholesterol uptake and then secretion to the choriocapillaris for removal, which is higher than patients with APOE2 or APOE3 (Hu, Quinn et al. 2021). Cumulatively, the increased cellular cholesterol from increased synthesis and decreased efflux stimulates amyloid-β production either in astrocytes or the RPE.

Amyloid-β clearance is influenced by APOE in an isoform-dependent manner through several mechanisms. In general, APOE4 is less effective at clearing amyloid-β than APOE3 while APOE2 is more effective. Mechanistically, APOE interacts with and stabilizes amyloid-β fibrils to enhance amyloid aggregation and deposition, a process facilitated by HSPGs (Garai, Verghese et al. 2014, Rauch, Chen et al. 2018, Andreone, Przybyla et al. 2020). The higher affinity of APOE4 for HSPGs promotes amyloid-β aggregation (Yamauchi, Deguchi et al. 2008). On the other hand, APOE4 binds to amyloid-β less efficiently than the other isoforms (Verghese, Castellano et al. 2011, Yamazaki, Zhao et al. 2019, Husain, Laurent et al. 2021, Loch, Wang et al. 2023), which allows APOE4 to compete with amyloid-β for APOE receptors that ultimately decreases amyloid-β clearance through the BBB (Deane, Sagare et al. 2008, Kanekiyo, Cirrito et al. 2013, Kanekiyo, Xu et al. 2014, Yamazaki, Zhao et al. 2019, Ingala, Mazzai et al. 2020). Specifically, APOE4 shifts amyloid-β clearance from the LRP1 to the VLDL receptor pathway, whereas amyloid-β clearance by APOE2 and APOE3 is through both the LRP1 and VLDL receptor pathways (Deane, Sagare et al. 2008). Thus, not only does APOE4 promote amyloid-β production through increased cellular cholesterol, but it impedes amyloid-β degradation and clearance, which increases amyloid-β deposition and risk in CAA.

In AMD, chronic inflammation is a major pathogenic factor. Critically, APOE modulates the immune response that is stratified by the isoform (Liao, Yoon et al. 2017, Long and Holtzman 2019, Husain, Laurent et al. 2021, Troutwine, Hamid et al. 2022). APOE4 is associated with less inflammation and drusen formation than APOE2 or APOE3, which suggests why this isoform is protective in the retina (Abyadeh, Gupta et al. 2023). In AMD, microglia or mononuclear phagocytes (MNPs) accumulate in the subretinal space (Gupta, Brown et al. 2003). In Cx3cr1-deficient mice, which are devoid of the receptor that minimizes microglia activation, Levy et al. found increased APOE expression in microglia drives their migration and accumulation in the subretinal space, which contributes to photoreceptor or RPE degeneration (Levy, Calippe et al. 2015). In mice expressing the different APOE isoforms, APOE2 mice relative to APOE3 and APOE4 mice, accumulate subretinal MNPs with higher APOE and APOE-dependent immune activation that results in photoreceptor or RPE degeneration. Compared with APOE3 mice, APOE4 in Cx3cr1-deficient mice protected against subretinal MNP accumulation through reduced APOE expression.

The APOE effect on microglia/MNPs can be explained in part by their impact on lipid rafts. APOE2 has reduced APOE uptake relative to the other isoforms, which activates microglia after cholesterol is extracted from lipid rafts. In contrast, APOE4 is less efficient at removing lipid raft cholesterol, which keeps microglia quiescent (Levy, Lavalette et al. 2015). Thus, APOE expression in microglia of the neural retina influences the degree of inflammation in the subretinal space. APOE4 by altering lipid raft cholesterol, minimizes subretinal microglia/MNP accumulation and inflammation to mitigate neural retinal or RPE injury. Since inflammation contributes to the growth of drusen, it is possible that APOE4 relative to the other isoforms, mitigates the inflammatory response to decrease the deposition of inflammatory proteins on HAP shells. Despite the lipid wall, the inflammatory phase of drusen biogenesis is minimized.

APOE4 also selectively binds to VLDL rather HDLs, the latter of which are preferred by APOE3 and APOE2 (Phillips 2014). This profile would promote its selective binding of VLDLs, which due to their larger size than HDLs, results in lipid accumulation in BrM. In support of this possibility, Malek et al. showed that APOE4 mice, and less so in APOE3 mice, when aged and given a high-fat cholesterol diet, developed an AMD-like phenotype including BrM deposits and RPE atrophy (Malek, Johnson et al. 2005). Paradoxically, the APOE4 phenotype is opposite of what the genetic risk would predict. Since the APOE4 isoform preferentially bound VLDLs, APOE4 mice had more lipid accumulation that increased amyloid-β dependent or independent inflammation. Critically, while APOE4 promotes cholesterol accumulation through VLDLs, it also minimizes inflammation from either amyloid-β or oxidized lipids that accumulate in BrM during drusen formation, which would be expected to decrease development of AMD-like features. Notably, these mice had targeted replacement of human APOE2, APOE3, or APOE4 (Sullivan, Mezdour et al. 1997). Thus, it is possible that the human APOE protein, independent of the isoform, elicited an inflammatory response sufficient to contribute to the retinal pathology. Alternatively, human APOE proteins may interact differently with mouse proteins such as LDLR, LRP, ABCA1, or HS than their respective human proteins.

Finally, it has been shown with live cell imaging of RPE cells that elevated-risk APOE2 has less effective cholesterol transport than the other isoforms, which caused ceramide accumulation (La Cunza, Tan et al. 2021). The elevated ceramide in turn, impaired autophagy and induced complement-mediated mitochondrial injury, the latter of which drove a redox state-sensitive cysteine-mediated phase separation of APOE2, forming biomolecular condensates that could nucleate as during drusen biogenesis. In contrast, the APOE4 isoform lacks these cysteines and resisted phase separation and condensate formation. Further work is needed to fully understand the mechanistic role of the APOE isoforms with regard to lipid trafficking and deposition in BrM with AMD.

6. Clinical Evidence for CAA and AMD

The evidence for a clinical link between CAA and AMD is limited in study number and the size of the cohorts studied. The Age, Gene/Environment Susceptibility (AGES)-Reykjavik Study tested the hypothesis that AMD and retinal microvascular signs are differentially associated with lobar and deep cerebral microbleeds (CMBs) because their prior work showed an association of CMBs with retinal microvascular signs and a suggestive association with geographic atrophy (Qiu, Cotch et al. 2012). 2502 participants were prospectively followed for an average of 5.2 years (Qiu, Ding et al. 2018). Geographic atrophy was associated with lobar cerebral microbleeds (CMBs) typical of CAA (multivariable-adjusted odds ratio 2.59, 95% CI 1.01–6.65) but not with deep CMBs, which are linked with hypertensive small vessel disease. Retinal microvascular changes were not associated with lobar CMBs. Anisetti et al. evaluated 256 age-matched pairs (126 with AMD and 130 without AMD) and found that late AMD was associated with increased odds (OR2.83, 95% CI 1.10–7.27, p=0.031) of CAA (Anisetti, Stewart et al. 2023). With the histopathologic similarities and differences, longitudinal clinical studies tracking AMD and CAA progression is warranted.

7. Future perspectives

While on the face of it, AMD and CAA are distinct diseases, both are induced or strongly influenced by overlapping factors, including cholesterol dysfunction, amyloid-β accumulation, APOE risk genotypes, and complement activation (Figure 4). These similarities lead us to posit that AMD and CAA share common mechanisms that considered together, can accelerate our understanding of both diseases, leading to novel biomarkers and therapeutic approaches.

Figure 4.

Schematic of similarities and differences of CAA and AMD due to cholesterol, APOE, amyloid, and complement. (A) In the RPE, cholesterol (Chol) accumulates due to photoreceptor recycling or a consequence of a high-fat cholesterol diet (HFD) to generate amyloid-β and the secretion of lipoproteins that contain apoE and apoB100 (yellow circles). With amyloid-β (black stars), these lipoproteins become a nidus for inflammation (gray circles) during drusen biogenesis in Bruch’s membrane (BrM). The reverse cholesterol transport (RCT) is a mechanism to release cholesterol from the RPE. (B) In the CNS, neurons or astrocytes release cholesterol (chol; yellow circles) and amyloid-β (black stars) by RCT into cerebral blood vessel wall, which accumulates and elicits an immune response that eventually leads to vascular injury. (C) Similarities and differences of AMD and CAA.

To enable future studies, we propose a model that centers on cholesterol metabolism, amyloid-β accumulation, and complement activation. In both CAA and AMD models, cholesterol accumulates due to increased cholesterol synthesis, increased cholesterol uptake, and decreased cholesterol efflux. The enhanced cellular cholesterol stimulates amyloid-β production. With impaired amyloid-β clearance and degradation, amyloid-β accumulates, and if sufficient, activates complement. If sufficiently prolonged or severe, the complement response will induce tissue injury. We focus particularly on CAA, albeit simalities exist with parenchymal amyloid-β (plaque) accumulation, because both CAA and AMD have a vascular component; small vessel disease in the case of CAA, and choriocapillary loss in AMD. That said, given CAA occurs in many cases of AD, understanding the relationship between parenchymal and vascular amyloid-β in AD may inform our understanding of amyloid accumulation in AMD.

APOE appears central to both AMD and CAA. Among many functions, APOE regulates cholesterol synthesis, cholesterol import, and cholesterol efflux. In both AMD and CAA models, cholesterol accumulates due to increased cholesterol synthesis, increased cholesterol uptake, and decreased cholesterol efflux. However, the APOE isoforms have opposite risk for AMD and CAA, which provides valuable insight into how the APOE and cholesterol synthesis and accumulation contribute to the pathobiology of each disease. The APOE4 isoform, relative to APOE2 and APOE3, i) promotes cholesterol accumulation due to cholesterol synthesis and increased cholesterol uptake; ii) decreases cholesterol efflux via reverse cholesterol transport because APOE4 has lower binding with HDLs than APOE2 and APOE3; and iii) induces less inflammation. The key difference in APOE isoforms between AMD and CAA pathobiology appears to be the contribution of amyloid-β and complement mediated inflammation. In CAA, APOE4, relative to APOE2 and APOE3, promotes cholesterol accumulation due to cholesterol synthesis and increased cholesterol uptake. In addition, cholesterol efflux through reverse cholesterol transport, is decreased because APOE4 has lower binding with HDLs than APOE2 and APOE3. The increased cholesterol both augments amyloid-β synthesis and decreases its degradation. With amyloid-β accumulation, complement is activated. In CAA, both amyloid-β and the complement response are concentrated in the neurovascular unit. Despite reduced inflammation with APOE4 relative to APOE2 and APOE3, amyloid-β accumulation is sufficient to activate complement that damages and ultimately ruptures blood vessels. Thus, amyloid-β accumulation is the key pathophysiologic factor in CAA. Despite its overlapping and genetic risk of AMD and CAA, the precise mechanisms of APOE are still be elucidated. Future studies exploring shared and distinct mechansims, including contrasting risk/protective affects of APOE isoforms, across common eye and brain diseases are warranted.

Complement activation is a key component of both AMD and CAA, and likely plays beneficial or damaging roles that is dependent on disease stage. Genetic studies have identified variation in the complement regulators CR1 and CFH with increased risk for CAA and AMD, respectively. These regulators influence the location of complement activity and the cell types involved. CR1 normally protects tissue by binding complement fragments (C3b, C4b) to clear immune complexes and moderate complement activation. In addition, it has a noncanonical role by mediating endocytosis of amyloid-β-C3b-CR1 complexes that are then degraded. Importantly, CR1 is located on astrocyte and microglia cell membranes, which focuses the complement activation to cells in the neurovascular unit, culminating in the injury and rupture of the vessel wall. CFH also modulates the complement response, but instead, regulates the alternative pathway. In contrast to CR1, CFH is secreted by the RPE and can regulate complement activity in BrM. In addition, CFH has noncanonical functions that are relevant to AMD pathobiology, such as its binding with HSPGs, which compete for lipoproteins to regulate lipoprotein accumulation in BrM, and its ability to bind and remove the lipid peroxidation product malondialdehyde (MDA) on oxidized lipoproteins. In AMD, these impaired noncanical functions due to the CFH risk variant promote oxidized lipoprotein accumulation in BrM. Along with amyloid-β, oxidized lipoproteins trigger an augmented complement immune response to collectively promote drusen formation. Whether the distinct genetic associations (CFH in AMD, CR1 in CAA) are pointing towards distinct mechansims, or shared mechanisms given their overlapping functions, are still be determined. Also, CFH and CR1 regulate central components of the complement cascade, and yet all of the cell types that produce the necessary components to drive complement activation (C1 through C9) and other regulators are not entirely known. Mapping these complement factors to specific cell types will improve provide the basis for understanding coordinated cellular response in AMD and CAA. We recognize that this model has been constructed on studies that relied heavily on in vitro and mouse models that may have limited translation to human disease. Since this pathobiologic comparison is new, direct comparative studies of common and different pathobiologies have not been conducted. By more precisely defining the roles of APOE, amyloid-β, and inflammatory pathways, along with their cell-specific effects, new biomarkers and therapies targeting these shared pathophysiologic mechanisms may be developed to reduce disease.

Highlights.

• Amyloid-β accumulation has etiologically linked AMD and AD.

• Amyloid-β accumulates in cerebral vessels in cerebral amyloid angiopathy (CAA).

• CAA exhibits pathological features that remarkably parallel those of AMD.

• Their pathobiology involves apoE cholesterol metabolism and complement inflammation

• Pathologic differences are due to the distinct involved cell types

Data availability

No data were used for the research described in the article.