Failure of Perivascular Drainage of β‐amyloid in Cerebral Amyloid Angiopathy

Abstract

In Alzheimer's disease, amyloid‐β (Aβ) accumulates as insoluble plaques in the brain and deposits in blood vessel walls as cerebral amyloid angiopathy (CAA). The severity of CAA correlates with the degree of cognitive decline in dementia. The distribution of Aβ in the walls of capillaries and arteries in CAA suggests that Aβ is deposited in the perivascular pathways by which interstitial fluid drains from the brain. Soluble Aβ from the extracellular spaces of gray matter enters the basement membranes of capillaries and drains along the arterial basement membranes that surround smooth muscle cells toward the leptomeningeal arteries. The motive force for perivascular drainage is derived from arterial pulsations combined with the valve effect of proteins present in the arterial basement membranes. Physical and biochemical changes associated with arteriosclerosis, aging and possession of apolipoprotein E4 genotype lead to a failure of perivascular drainage of soluble proteins, including Aβ. Perivascular cells associated with arteries and the lymphocytes recruited in the perivenous spaces contribute to the clearance of Aβ. The failure of perivascular clearance of Aβ may be a major factor in the accumulation of Aβ in CAA and may have significant implications for the design of therapeutics for the treatment of Alzheimer's disease.

Anatomy of the Cerebral Vasculature

Arteries penetrate the cerebral cortex perpendicular to the pial surface 25. As they enter the cortex from the subarachnoid space, arteries gain a complete layer of pia mater, separating the subarachnoid space from the perivascular Virchow‐Robin spaces 95. Macrophages recruited in part from the pool of peripheral monocytes reside within these perivascular spaces 6. Cortical arteries are, in fact, arterioles as they do not have elastic laminae, but they possess a tunica media, with multiple layers of smooth muscle, separated by a spiral basement membrane (Figure 1A). Basement membranes are thin (∼150 nm) sheets of glycoproteins and proteoglycans that provide functional and structural support to the vessel 94. The larger proteins, type IV collagen and laminins, are linked by integrins and nidogen and bound by chondroitin and heparan sulfate proteoglycans 66, 93. As the penetrating arteries give rise to smaller arterioles, the layer of pial cells becomes incomplete, leading to an absence of both pia and perivascular spaces at the capillary level 95. Cerebral capillaries have an endothelium with tight junctions, lined by an endothelial basement membrane that is fused with an astroglial basement membrane (Figure 1B). Pericytes are contractile cells completely surrounded by basement membrane 28, 35, 45. Cerebral veins do not possess a smooth muscle layer and have very little basement membrane. Perivascular spaces around the vein are confluent with the subpial space and only small numbers of pial cells are associated with the vessel wall 81. Lymphocytes pass into the brain across the venous endothelium and dilate perivenous Virchow‐Robin spaces 6, 26, 84. The recruitment of neutrophils requires G protein‐coupled receptor‐dependent activation of β2 integrins and binding to ICAM‐1 present on the endothelium 34. The crawling of neutrophils upon the venous endothelium is mediated by endothelial ICAM‐1 and ICAM‐2 and neutrophil LFA‐1 and Mac‐1 34.

Figure 1.

A. Diagrammatic representation of a cross section through a cerebral cortical arteriole. There are no elastic laminae, but there is a tunica media with smooth muscle cells (green) and basement membrane (orange) interposed between the smooth muscle cells. The astrocyte end‐feet (light blue) abut upon the leptomeningeal sheet that reflects upon the artery as it penetrates the brain from the subarachnoid space. B. Diagrammatic representation of a capillary in the cerebral cortex. The endothelial cells are connected by tight junctions. On the abluminal border of the endothelial cells there is a basement membrane that results from the fusion of the basement membranes of the endothelium and the astrocyte end‐feet. Pericytes are completely surrounded by basement membrane.

Perivenous Infiltrates of Lymphocytes: Significance for Immunological Aspects of Neurodegenerative Diseases

Taking into account the anatomy of brain vessels, immune cells need to penetrate two distinct barriers in their way from the blood stream to the brain parenchyma: the vascular wall with its basement membranes (first step of neuroinflammation) and the glia limitans which also includes a separate basement membrane (second step of neuroinflammation). The vascular and the glial basement membranes differ in their content of laminin isoforms 71. As the glial basement membrane is anchored to dystroglycan, which is cleaved by matrix metalloproteases (MMP) 2 and 9 1, blockage of MMP using BB‐94 blocks the penetration of the glia limitans and restricts immune cells to perivascular spaces 76. The first step is mainly driven by CCL‐2 76, which is crucial for the recruitment of Ly‐6ChiCCR2 + monocytes, for example, after irradiation 53. Thus, while recruitment of immune cells into perivascular spaces requires release of CCL2, additional signals are required for the induction of MMPs. While the latter causes intraparenchymal infiltration and neurological symptoms, the former may be regarded as a form immune surveillance.

Cerebral Capillaries and Arteries Host the Pathways for Perivascular Drainage of Interstitial Fluid (ISF ISF)

The pathways for perivascular drainage of interstitial fluid and solutes have been identified both in experimental animals 13 and in humans 61, 85. The pathway for the bulk flow drainage of ISF and solutes from the brain has been characterized in detail by injecting 0.5 μL of fluorescent dextran of 3–10 kDa or ovalbumin 49 kDa into the gray matter of the mouse 13 (Figure 2A). Drainage of solutes from the extracellular spaces appears to be very fast, occurring within 5–15 minutes. By 5 minutes after injection, tracers had spread diffusely through the extracellular spaces and entered the walls of blood vessels. Confocal microscopy showed that tracers were co‐localized with laminin in the basement membranes of capillaries and in the basement membranes of the tunica media of arteries 13. Microinfusion studies using ¹⁴C‐labeled glucose and d‐lactate demonstrate their perivascular drainage from the inferior colliculus to the arteries at the base of the brain 4.

Figure 2.

A. Diagrammatic representation of perivascular clearance of solutes, normally entering the walls (basement membranes) of capillaries and then arteries, but not of veins. B. Cross section through a human artery at the surface of the brain. The tissue has been immunostained for collagen IV (a marker of basement membranes, shown here in blue), smooth muscle cells (immunostaining for smooth muscle actin, shown in green) and amyloid‐β (Aβ), using anti‐4G8 antibody, shown here in red. Note the presence of Aβ surrounding the smooth muscle cells, leaving the endothelial basement membrane (blue) free of Aβ. There is some Aβ accumulating at the surface of the brain, with the collagen IV present in glia limitans. Confocal microscopy. Scale bar: 20 μm.

Within 5 minutes of injections, in the arterial walls, the tracers were selectively located in the basement membranes between the smooth muscle cells in the tunica media and not in the endothelial basement membranes nor in the outer basement membrane at the interface between the artery and the extracellular matrix. By 3 h after injection, the tracers had disappeared from the extracellular spaces of the gray matter and from the basement membranes in the walls of capillaries and arteries. At this time, tracers were observed in the perivascular macrophages within the ipsilateral parenchyma and in the leptomeninges 13. At 24 h after injection, the tracers were present in the perivascular macrophages adjacent to the ipsilateral intracerebral arteries and around leptomeningeal arteries on the surface of the brain 13.

Particles of 1000 nm size do not enter capillary or arterial basement membranes and remain confined to the injection site, whereas particles of 20 nm enter the basement membranes of capillaries and the Virchow‐Robin perivascular spaces, where particulate matter can be observed for up to 2 years 13, 43. The exact capacity of the cerebrovascular basement membranes for clearance is still unclear and requires an in vivo experimental approach using a range of nanoparticulate sizes 30.

Because of the limited sensitivity of fluorescent markers injected in low volumes into the ISF, it was not possible to determine whether they reach cervical lymph nodes. However, older studies using radioactive tracers show that they drain along the tunica media and the tunica adventitia of the major cerebral arteries, through the base of the skull to deep cervical lymph nodes 72. The perivascular drainage of ISF appears to be separated from the cerebrospinal fluid (CSF), as only 10%–15% of tracers injected into the deep gray matter of the brain can be identified in the CSF 72.

Motive forces for the perivascular drainage of interstitial fluid

It is difficult to investigate experimentally the motive force for perivascular drainage of ISF and solutes from the brain, as the timescale is fast and the pathways are inaccessible and small (100–150 nm) 36. However, the observation that perivascular transport only occurs in living animals and ceases immediately after cardiac arrest suggests that pulsations in artery walls may generate the motive force for the transport of ISF and solutes out of the brain 4. Mathematical models indicate that perivascular transport of ISF and solutes may be driven by the contrary (reflection) waves that follow each pulse wave 67, 80. A valve‐like action is required in order to prevent reflux during the passage of the major pulse wave along the vessel wall 67. As the route of drainage is within basement membranes 13, it is possible that the valve action results from changes in the conformation of the basement membrane or from different biochemical interactions. During cardiac systole and the pulse wave, the arterial internal elastic lamina and basement membranes are stretched as the vessels expand 83. During diastole, the vessel wall expands allowing the contrary wave to pass back along the artery wall and the basement membranes to also expand. These rhythmical changes in the dimensions and conformation of the basement membranes may provide the necessary valve action to ensure that ISF is driven along the artery walls toward leptomeningeal arteries 67. Mathematical fluid flow modeling demonstrates that peristaltic contractions in the walls of arteries, comparable to the pulse contractions, are enough to drive ISF along perivascular drainage pathways 80.

A recent detailed transcardial Doppler analysis of the blood flow through the circle of Willis demonstrated that mean flow velocity and the pulsatility index are reduced with age alone and decrease further with disease 63. This suggests that aging vessels become stiff, less elastic and arteriosclerotic, particularly in humans, interfering with perivascular drainage of ISF and solutes in elderly individuals 82. Administering the calcium channel antagonist nimodipine did not affect the pattern or the rate of perivascular drainage of soluble dextrans 5. In this study, nimodipine was co‐infused with the tracer, resulting in a local vasodilation, rather than a general systemic effect on the pulsatile blood flow.

Cerebral Amyloid Angiopathy (CAA): Failure of Perivascular Drainage of Amyloid‐β (Aβ)

Aβ accumulates in the human brain with age and in Alzheimer's disease (AD) as insoluble plaques in brain parenchyma and in the walls of blood vessel as CAA 55. In both humans and AD transgenic mouse models, it is mainly the more soluble Aβ40 that is deposited in the vessel walls 64, whereas the less soluble Aβ42 is found mainly in the senile plaques 23. Raised levels of soluble Aβ in the cerebral cortex and the severity of CAA appear to correlate better with the severity of cognitive decline and dementia than does the number of insoluble plaques of Aβ in brain parenchyma 48, 57, 79.

In familial forms of AD and in transgenic mouse models of AD, Aβ is overproduced or there is production of aberrant forms of Aβ resulting in the accumulation of Aβ in the brain and blood vessel walls 91. However, there is little evidence for overproduction of Aβ in the more common sporadic form of AD. It is, therefore, likely that accumulation of Aβ in the brain and in the vessel walls in AD is caused by failure of elimination of Aβ from the aging brain.

A number of different mechanisms and pathways for the elimination of Aβ from the brain have been identified in experimental animals and in humans. Enzymes such as the catabolic peptidase, neprilysin, and the metalloproteinase, insulin‐degrading enzyme, degrade Aβ in brain parenchyma 54, 70. Neprilysin activity declines significantly with age in the mouse hippocampus 27. Upregulation of neprilysin in aged SwAPP transgenic mice does not lower the level of toxic Aβ oligomers and does not improve cognition, despite the removal of Aβ plaques from the brain 51. Peripherally administered neprilysin reduce levels of Aβ in the blood, but not in the brain 39.

Antibodies against Aβ in the serum remove Aβ from the brain 22 and microglia degrade Aβ in vivo and in vitro, in mice and humans immunized against the Aβ peptides 2, 88. Patrolling monocytes and perivascular macrophages regulate the severity of CAA via mechanisms involving scavenger receptor A, class B type 1 and CD36 47, 52, 58, 75. Astrocytes and smooth muscle cells also contribute to the uptake and degradation of Aβ in a process that may be mediated by the expression of aquaporin‐4 40, 62. Aβ is absorbed into the blood by low‐density lipoprotein receptor‐related protein‐1 (LRP) and P‐glycoprotein, with α_2‐macroglobulin and apolipoprotein E also acting upon these pathways of clearance 7, 21, 68. However, LRP‐1 expression in endothelial cells and the effectiveness of this pathway are reduced in older mice and possibly in AD 7. Concomitant with enzymatic, cellular and transendothelial transport, soluble Aβ is also cleared by the perivascular bulk flow drainage of ISF 86.

The predominant distribution of Aβ in the walls of large cortical and leptomeningeal arteries in CAA support the hypothesis that Aβ is deposited in the perivascular pathways by which interstitial fluid normally drains from the brain 14 (Figure 2B). Soluble tracers, including Aβ40, injected into the ISF of the adult mouse brain distribute along basement membranes in the same pattern as Aβ deposition in human CAA 37. Increased severity of vascular load of Aβ42 is observed following anti‐Aβ immunotherapy experiments in mice and humans, suggesting that Aβ solubilized from parenchymal plaques drains along and becomes entrapped within perivascular clearance pathways 10. Other amyloid species, including cystatin, transthyretin and ABri 50, also accumulate in the vessel walls of the brain, indicating impairment of ISF drainage pathways as a common factor in the pathogensesis of CAA regardless of the amyloidogenic peptide involved. Such accumulations are often associated with alternations in basement membranes and precede endogenous CAA development in the TGF‐β1 transgenic mice 89, 90.

Factors that Disrupt Perivascular Drainage of Aβ from the Brain

Age

Age is a major risk factor for both AD and cerebrovascular disease (CVD). CVD encompasses a number of changes in the cerebral blood vessels of aging humans. There is progressive fibrosis of the intima and media in major cerebral arteries and flattening of internal elastic laminae, reflecting the generalized stiffening and loss of elasticity of artery walls with age (arteriosclerosis) 12. Focal plaques of fibrosis and lipid deposition result in atherosclerosis, and arteries may be occluded by thrombi and emboli, leading to infarction and poor vascular perfusion 8. In addition to directly affecting cerebral perfusion, CVD may contribute to AD dementia by altering the efficiency of perivascular clearance of Aβ from the elderly brain. Reduction in the amplitude of arterial pulsations is predicted to reduce the motive force driving perivascular clearance of Aβ, leading to its accumulation 67. This hypothesis is supported by experimental findings that solute clearance along basement membranes is decreased in aged mice 36, 37 and when perfusion was impaired by occlusion of the middle cerebral artery 33. Histological observations also confirm that thrombotic occlusion of cortical arteries is associated with deposition of Aβ in the walls of the capillaries served by those arteries, suggesting that Aβ is entrapped in capillary walls that form the proximal portions of the perivascular interstitial fluid drainage pathways 92.

Aging is also associated with changes in the biochemical composition of basement membranes in both humans and experimental models. Age‐dependent downregulation in collagen IV and laminin and upregulation of perlecan and fibronectin are observed in the blood vessels of mouse brains 36, 37. Small‐diameter vessels from AD patients show a decrease in the amount of collagen IV, compared to aged‐matched controls 17, while levels of heparan sulfate proteoglycans are increased 8, 69. Laminin, nidogens and collagen IV inhibit aggregation of Aβ and destabilize pre‐formed fibrils of Aβ 11. Heparan sulfate proteoglycans such as fibronectin and perlecan accelerate Aβ aggregation through high‐affinity interactions 15, 19. This age‐related shift toward increased expression of pro‐amyloidogenic basement membrane proteins may promote CAA by increasing the affinity of the vessel wall for Aβ. CAA deposition may further impair perivascular drainage pathways 2, 37, leading to a feedforward impairment in the elimination of not only Aβ but also other toxic solutes from the brain that may contribute to cognitive decline in AD.

Apolipoprotein E (ApoE) genotype

ApoE is the major carrier of cholesterol in the brain, regulating lipid homeostasis and is expressed by astrocytes, smooth muscle cells and neurons 46. The ApoE gene codes for three different alleles (ε2, 3, 4). Possession of one or two copies of ApoE4 increases the risk of developing AD and reduces the median age of onset 3, 16, 74, 77. Further, deposition of Aβ in capillary walls and increased CAA severity are observed in the brains of humans and transgenic mice expressing human apoE4 31, 60, 74.

ApoE genotype influences the efficiency of most of the Aβ clearance mechanisms in the brain, including enzymatic 18, 24, microglial/monocytic 78, 96 and transendothelial 20. Perivascular drainage pathways are also affected by the expression of ApoE4. Immunocytochemical and Western blot analysis of basement membranes demonstrated that levels of laminin and collagen IV were higher, while fibronectin expression was lower, in the brains of 3‐month old targeted replacement (TR) mice expressing the human APOE4 (TRE4) allele compared to those expressing the human APOE3 (TRE3) allele and wild‐type controls 41. In 16‐month‐old mice, laminin and collagen IV levels were unchanged between wild‐type and TRE3 mice but lower in the blood vessels of TRE4 mice. The pattern of distribution of Aβ40 following injection into the hippocampus was significantly disrupted in the blood vessels of TRE4 mice, as early as 3 months of age 38. In 16‐month‐old TRE4 mice, soluble Aβ40 aggregated within the periarterial drainage pathways, suggesting that APOE4 is associated with a significant risk of failure of clearance of Aβ40. Interestingly, although ApoE co‐localized with laminin in the blood vessels of both TRE3 and TRE4 mice, no co‐localization was noted between ApoE and deposits of Aβ in the vessels of either TRE3 or TRE4 mice 38. Thus, the role of ApoE in the clearance of Aβ along perivascular pathways remains to be fully elucidated.

Immune complexes

Anti‐Aβ immunotherapy trials for the treatment of AD in humans were introduced following successful studies in transgenic mice demonstrating that insoluble Aβ plaques were removed from the brain following immunization of mice with Aβ42 65. However, despite the clearance of plaques in patients with AD 29, 49, 56 active immunization significantly increased arterial CAA load in both transgenic animals and in humans 56, 87. As these vascular Aβ deposits were composed principally of Aβ42, it was hypothesized that Aβ solubilized from Aβ42‐positive plaques became trapped in the perivascular drainage pathway 59, 82. Both active 9, 32 and passive 41, 42 immunotherapy trials have encountered side effects comprising focal abnormalities in the cerebral white matter on imaging. The pathophysiology underlying these side effects is as yet unclear but it may be caused by the effects of immune complexes in relation to the cerebral vasculature.

In a mouse model of immune complex formation in the brain, the antigen (ovalbumin), IgG and complement C3 were located in basement membranes of artery walls 24 h after challenge with antigen 73. It is likely that the presence of immune complexes in the walls of arteries will be associated with significant disturbance of the periarterial drainage pathways.

Implications for Therapy of AD

Increasing evidence suggests that cerebrovasculature health is important for both the prevention and the treatment of AD. Risk factors for the development of cardiovascular disease, such as diabetes, hypertension and hypercholesterolemia, also increase the risk of AD 44. Maintaining the health of cerebral vessels by preventing or minimizing the impact of such modifiable risk factors may therefore help to promote efficient clearance of Aβ across the blood–brain barrier and along perivascular pathways of the aging brain. Moreover, as indicated by findings from the anti‐Aβ immunization trials, the success of Aβ‐based therapies is strongly dependent on the robustness of cerebral blood vessels and such factors must be taken into consideration in the design of new therapeutics for the treatment of AD.

Conclusions

Perivascular drainage along cerebrovascular basement membranes is a major route by which Aβ is eliminated from the brain. Theoretical and experimental models support the hypothesis that efficient drainage of Aβ along perivascular pathways is dependent on the biochemical composition of the basement membrane and the force of arterial pulsations. Factors that affect cerebrovascular health, such as age and APOE genotype, alter both the structure of the blood vessels and the expression of the basement membrane proteins such that the efficiency of perivascular drainage of Aβ is reduced. As increasing amounts of Aβ become entrapped within the drainage pathways, it causes damage to the underlying vasculature, further reducing the functionality of the vessel and creating a feedforward mechanism by which increasing amounts of Aβ accumulate as CAA. Finally, diffusion of soluble Aβ and ISF through brain tissue is blocked by insoluble Aβ in the extracellular spaces, levels of soluble Aβ and other metabolites in brain parenchyma rise and dementia ensues.