Overlapping Protein Accumulation Profiles of CADASIL and CAA

Cerebral SVD is a prevalent vascular disorder of the brain and a major contributor to neurologic deterioration in our elderly population. Recent estimates report that sporadic SVD affects over half of the population aged >65 years and significantly elevates the risk of stroke, vascular dementia, and progression of neurologic diseases.^, In addition to sporadic SVD, monogenic causes of SVD have also been described, which may shed light on pathomechanisms. The most common type of SVD is CADASIL, caused primarily by stereotypical cysteine mutations in the NOTCH3 gene that result in altered cysteine number.

Clinically, patients with CADASIL present with an accelerated and often more severe clinical course compared with sporadic SVD. Although there can be variability in disease presentations, patients with CADASIL typically present with migraine with aura, subcortical ischemic events, mood disturbances, apathy, and cognitive impairment. In most patients, migraine with aura is the first observed symptom, often occurring roughly 15 years earlier than ischemia.

CADASIL was first described in 1993 when several unrelated families presented with a mendelian syndrome that caused recurrent strokes. Initially, these syndromes were reported under different names until the affected gene was mapped to chromosome 19q12. Subsequent linkage analysis of additional families led to the mapping of the CADASIL critical region to the NOTCH3 gene. Although the prevalence of genetically proven CADASIL is estimated to be roughly 2 per 100,000 adults, the actual prevalence is thought to be higher when taking into account sporadic mutations.

NOTCH3 gene encodes a vascular smooth muscle transmembrane protein involved in smooth muscle cell differentiation and vascular development. NOTCH3 contains 34 extracellular epidermal growth factor–like repeats, and each epidermal growth factor–like repeat contains six cysteines that are predicted to form three disulfide bonds important for protein structure. Virtually all CADASIL-causing mutations in NOTCH3 affect cysteines, leading to a change in cysteine number. This provides support for the idea that aberrant NOTCH3 undergoes abnormal disulfide bonding that is important for disease pathogenesis. In addition, three-dimensional modeling shows that at least a subset of known NOTCH3 mutations leads to domain misfolding, supporting the notion of misfolded proteins driving CADASIL. It is thus postulated that cysteine mutants of NOTCH3 are associated with abnormalities in protein folding and could possess neomorphic properties responsible for CADASIL pathology. Recently, there have also been reports of CADASIL in patients without stereotypical cysteine mutations in NOTCH3. Many of the noncysteine involving NOTCH3 mutations are also thought to disrupt NOTCH3 protein structure, and in vitro examination of cysteine-sparing NOTCH3 mutants shows that they also form aggregates similar to those of typical cysteine mutants.

As of now, no clear genotype-phenotype correlation has been identified, although studies have emerged that suggest a relationship between the position of mutations and disease severity. For example, Rutten et al provides evidence to suggest that C-terminal mutants/polymorphisms could contribute to delayed onset small-vessel disease that is less severe compared with the classic CADASIL phenotype. In addition, a recent review of 224 CADASIL case reports suggests that the pathogenicity of CADASIL mutations is related to the location of the mutation. Currently, it is thought that both genetic and environmental factors play a role in disease presentation.

Presently, the gold standard for diagnosis of CADASIL is genetic testing. Screening 23 exons of NOTCH3 for mutations has a specificity and sensitivity close to 100%. Another useful diagnostic tool is skin biopsy. An ultrastructural hallmark of CADASIL is the deposition of granular osmiophilic material (GOM) within vessels. GOMs are electron-dense extracellular deposits typically found between neighboring vascular smooth muscle cells and most readily visualized by electron microscopy. In addition to being found in the cerebrovasculature, GOMs can also be detected in extracerebral vasculature, such as that of the skin. Thus, skin biopsy can be used to diagnose CADASIL with a specificity of 100% and a sensitivity that ranges from 45% to 100%. Currently, the exact GOM composition remains unclear. However, NOTCH3 ectodomain, N-terminal fragment of NOTCH3 (NTF), metalloproteinase inhibitor 3 (TIMP3), vitronectin (VTN), latent transforming growth factor-β 1 (LTBP1), amyloid P (SAP), annexin 2, and periostin have been identified to be components, suggesting that GOMs consist of abnormal protein aggregates.^{,  ,  ,  ,}  Finally, presence of subcortical infarcts and leukoencephalopathy is a feature of CADASIL best detected by magnetic resonance imaging (MRI). MRI lesions become apparent at a mean age of 30 years and are found in virtually all patients with CADASIL aged >35 years. Typically, MRI changes precede development of other CADASIL symptoms and worsen with age and disease progression.^, In particular, MRI involvement of the anterior temporal pole has approximately 90% sensitivity and approximately 90% specificity for CADASIL.

Pathologic characteristics of CADASIL include GOMs, as mentioned earlier, that surround the vascular smooth muscle cells of arterioles and the degeneration of vascular smooth muscle cells in arterial walls. In addition, CADASIL brains show significantly thickened arteriolar walls and reduced diameters of small penetrating arteries. It is thought that these thickened arteries and GOMs result from disease-related abnormal protein accumulation that includes molecules such as NOTCH3 ectodomain.

Notch signaling plays a critical role in development and involves processing of NOTCH proteins. NOTCH3 is a transmembrane protein, composed of both an extracellular and a membrane tethered intracellular domain. On activation by Notch signaling ligands, a series of proteolytic cleavages occur by ADAM-TACE and γ-secretase, releasing the NOTCH3 intracellular domain to the nucleus to affect gene transcription.

Although CADASIL is characterized by mutations in NOTCH3, it seems unlikely that a loss of NOTCH3 signaling is the sole driver of disease. Several studies report connections between CADASIL-causing NOTCH3 mutations and NOTCH3 receptor function, which alter canonical Notch signaling.^{,  ,  ,}  However, other mouse model studies do not support this idea. For example, despite having increased susceptibility to stroke, Notch3 knockout mice do not demonstrate classic features of CADASIL. On the other hand, mice that overexpress mutant Notch3 better model the human disorder, although it is also unlikely that CADASIL leads to a gain of Notch signaling function because CADASIL mutations do not enhance downstream signaling. The early accumulation of GOMs and NOTCH3 aggregates, found in these mouse models and patients with CADASIL, therefore suggests a neomorphic role of mutant NOTCH3, although it remains unknown whether these aggregates cause disease or are consequences of disease.

In addition to physiological NOTCH3 processing involved in Notch signaling, post-translational modification of NOTCH3 protein has also been identified as a disease-specific feature of CADASIL. Using conformation-specific antibodies, Zhang et al identified a reduced form of NOTCH3 that accumulates specifically in disease-affected CADASIL vessels compared with normal-appearing vessels from age-matched control subjects. It is conceivable that a cysteine-involving mutation results in a change in NOTCH3 tertiary structure or aggregation state. Support for disease-related post-translational alterations of NOTCH3 also comes from Arboleda-Velasquez et al, who demonstrated impairment of glycosylation and cleavage in mutant CADASIL protein. Both reduced glycosylation and aberrant multimerization of mutant NOTCH3 may play a role in abnormal protein accumulation in CADASIL. In addition, studies identify accumulation of NTF in pathologic vessels compared with normal-appearing vessels, suggesting enhanced cleavage of NOTCH3 protein in disease that is unrelated to Notch signaling. In vitro studies demonstrate that reduction of NOTCH3, which breaks NOTCH3 disulfide bonding and destabilizes protein structure, enhances nonenzymatic NOTCH3 fragmentation in CADASIL vessels. Given the disease-specific nature of these NOTCH3 forms, there are likely multiple abnormalities of post-translational processing of NOTCH3 in CADASIL.

Both wild-type NOTCH3 and mutant NOTCH3 are capable of forming dimers, oligomers, and higher-order multimers in vitro. However, mutant NOTCH3 demonstrates increased propensity to form higher-order oligomers compared with wild-type protein.^{,  ,}  NOTCH3 oligomerization does not require presence of cofactors and is mediated in part by disulfide bonding.^, Furthermore, NOTCH3 binding partners, such as thrombospondin 2 and Fringe, have also been identified to co-aggregate with mutant NOTCH3, suggesting a role of protein aggregation in disease.^,

In addition, NTF, which localizes to pathologically affected vessels in CADASIL, has also been shown to undergo spontaneous thiol-mediated oligomerization in vitro. NTF can form covalent conjugates with vascular catecholamines that enhance multimerization of the NOTCH3 fragmentation product. More recent studies have highlighted the presence of NTF multimers in CADASIL tissue, implicating a role for NTF multimers in disease (unpublished data).

CADASIL brains demonstrate abnormal protein accumulation that is often thought to be linked to disease. For example, CADASIL vessels feature accumulation of the NOTCH3 ectodomain without simultaneous accumulation of the intracellular domain, NOTCH3 aggregates, abnormal NOTCH3 conformations enriched in disease, and NOTCH3 fragments, including an NTF.^,, In addition, other proteins, such as proteoglycans, collagens (COLs), and other extracellular matrix (ECM) proteins, have been shown to accumulate within the vessels as well, suggesting formation of extracellular protein complexes in disease.^{,,  ,}  The extracellular matrix is a network of highly cross-linked proteins that plays important roles in cell structure and support. ECM proteins are also thought to influence fundamental cellular processes, such as differentiation, survival, and proliferation. Traditionally, the ECM is thought to consist of collagens, proteoglycans, and glycoproteins. However, there are also ECM-associated proteins, such as ECM regulators, enzymes, or proteins, that modify the ECM and ECM-affiliated proteins.

The observed enrichment of ECM proteins in CADASIL can be explained by increased secretion of proteins or decreased protein turnover in disease. In mouse models of CADASIL, NOTCH3 ectodomain accumulation is one of the earliest observed events in disease pathogenesis, suggesting a role in disease initiation.

One model of enhanced protein recruitment suggests that NOTCH3 mutations lead to abnormal accumulation of mutant NOTCH3 protein, which then plays a role in recruiting and sequestering other proteins. Recent studies have identified the enrichment of ECM proteins in CADASIL, such as TIMP3, VTN, collagens, LTBP1, clusterin, decorin, biglycan, and laminins, among others.^{,,  ,  ,} In vitro studies suggest potential physical interaction between NOTCH3 and some of the proteins, including TIMP3 and LTBP1.^, Transforming growth factor (TGF)-β signaling activity is linked to regulation of fibrotic events in the vasculature, and is shown to be activated by fibronectin, fibrillin-1, and other members of the LTBP family. This supports a model where dysregulation of TGF-β signaling results from aggregation and/or accumulation of NOTCH3 in the cerebrovasculature of patients with CADASIL and promotes abnormal recruitment of ECM proteins. The resulting aggregation of proteins can result in a change in biological function and contribute to disease. Interestingly, TGF-β activity has been implicated in other vasculopathies, including cerebral autosomal recessive arteriopathy with subcortical infarcts and leukoencephalopathy (CARASIL) and CAA.

Another model of abnormal protein accumulation involves decreased protein turnover and proposes a loss of serine protease HTRA1 activity in CADASIL. HTRA1 is a negative regulator of TGF-β signaling and an extracellular serine protease that is known to be mutated in CARASIL, an autosomal recessive disorder that shares clinical overlap with CADASIL. In a recent liquid chromatography–tandem mass spectrometry study of CADASIL brain tissue, significantly increased levels of HTRA1 protein and substrates were identified (clusterin, vitronectin, elastin, and LTBP1), suggesting an impairment of HTRA1 activity in CADASIL.

CAA is a common age-related disease that can occur both with and without Alzheimer disease (AD). Although CAA can be caused by several amyloidogenic proteins, such as cystatin C, transthyretin, and others, this review will focus on amyloid-β (Aβ) CAA. Population-based studies demonstrate that CAA affects roughly 20% to 40% of elderly populations without dementia and 50% to 60% of elderly populations with dementia. In AD, CAA is predicted to be present in roughly 80% to 90% of patients. Classic clinical presentation of CAA involves spontaneous lobar intracerebral hemorrhage, cognitive impairment and/or dementia, and transient focal neurologic episodes.

Most cases of CAA are sporadic. However, some CAA has been linked to specific genetic loci that include apolipoprotein E (APOE) e4 and e2 alleles, mutations in presenilin 1 (PS1) and PS2, and mutations in amyloid precursor protein (APP). Mutations in APP that cause familial forms of CAA (eg, hereditary cerebral hemorrhage with amyloidosis, Dutch type) tend to cluster around residues 21 to 23 and have been proposed to decrease proteolytic degradation of Aβ or attenuate Aβ protein clearance from the brain into the circulation. On the other hand, two isoforms of APOE are thought to promote CAA via different mechanisms. For example, APOE e4 has been linked to increased amyloid deposition, whereas APOE e2 is thought to accelerate the formation of vasculopathies that promote vessel rupture. Both APOE e2 and e4 alleles have been associated with early recurrence of lobar intracerebral hemorrhage following a prior lobar intracerebral hemorrhage. In addition, alterations in TGF-β1 have been linked to CAA, along with polymorphisms in a1-antichymotrypsin, neprilysin, low-density lipoprotein receptor protein 1, CR1, and angiotensin-converting enzyme genes.

Definitive diagnosis of CAA relies on identification of CAA-related vascular damage, multiple lobar hemorrhages, and absence of alternative pathologies on postmortem examination of brain, as assessed by the Boston criteria. Diagnosis of probable CAA includes use of clinical data and MRI imaging to identify the presence of multiple hemorrhages or microbleeds in regions typical of CAA, and more recently in the modified Boston criteria, the presence of superficial siderosis. Although many imaging features are shared between CADASIL and CAA (cerebral microbleeds and white matter hyperintensities), cortical superficial siderosis is absent in a large cohort of patients with CADASIL from a clinical study examining imaging features of CADASIL, CAA, and control subjects. Thus, this study posits that in individuals with imaging features suggestive of small-vessel disease, presence of cortical superficial siderosis is highly suggestive of CAA.

Histopathologically, CAA is characterized by extracellular Aβ protein deposits in the cerebrovasculature. The most common forms of vascular amyloid deposits include Aβ_1-40 and Aβ_1-42.^, Many sources of Aβ have been proposed to explain its deposition within the vessel wall, including transport across the blood-brain barrier (BBB) from the circulation, the vascular smooth muscle cells, and neurons as a result of impaired perivascular drainage.^{,  ,}  In most cases, there is no direct evidence of overproduction of Aβ within the vessel wall. Instead, the perivascular drainage model suggests that the Aβ accumulation within the vasculature is likely the result of impaired Aβ drainage through perivascular pathways that typically serve as lymphatic-like drainage pathways from the brain. The impaired elimination of proteins results in the accumulation of both soluble and insoluble protein aggregates in the extracellular spaces within arterial and capillary walls. Alterations of these extracellular spaces in disease or aging might further contribute to decreases in drainage capacity and increased Aβ deposition.

CAA vessels also feature marked degeneration of smooth muscle cells within the medial layer of arterioles and hyalinization. Severely affected vessels demonstrate disruption of the vascular architecture, leading to microaneurysm formation, fibrinoid necrosis, and Aβ deposition in the surrounding neuropil (alias dysphoric changes). The two major types of CAA include type 1, which involves cortical capillaries and/or larger vessels, and type 2, which is limited to larger leptomeningeal and cortical arteries.^, CAA type 1 leads to capillary occlusion and alterations in vascular flow, contributing to the development of dementia.

Aβ is generated from cleavage of APP. APP is a membrane glycoprotein intimately involved in neuronal development, maintenance of neuronal homeostasis, cellular signaling, and intracellular transport. APP can generate many different cleavage products, some of which are thought to be major contributors to Aβ deposits throughout the brain. Under physiological conditions, APP is cleaved by α-secretase to shed the ectodomain, which does not produce amyloidogenic fragments. Alternatively, ectodomain shedding can occur by β secretase activity. Aβ is generated by β-secretase and γ-secretase cleavage to generate two predominant forms: Aβ_1-40 and Aβ_1-42. Aβ_1-40 is more often found within the vessel wall, whereas Aβ_1-42 is more often found in senile plaques and capillaries involved in CAA. The major β-secretase is BACE1, which has a main cleavage site at position 1D of Aβ, but also has an alternative site at 11E, generating Aβ_11-40/42. A second cleavage process occurs via the γ-secretase complex, consisting of presenilin, nicastrin, anterior pharynx-defective-1, and presenilin enhancer-2. The γ-secretase complex results in the intramembranous proteolysis of the β-secretase cleaved product to generate Aβ_1-40/42. The γ-secretase complex is also essential for physiological processing of NOTCH3.

Aβ monomers can form various types of assemblies, such as oligomers, protofibrils, and amyloid fibrils. In AD, insoluble amyloid fibrils can further assemble into amyloid plaques most commonly found in the neocortex of AD brains, whereas soluble amyloid oligomers can deposit throughout the brain. The dominant form in CAA is fibrillar Aβ, although a recent study found that most mutations in the Aβ sequence that promote CAA and AD do not have obvious stabilizing or destabilizing effects on Aβ fibrils derived from AD brains. Both soluble and insoluble Aβ assemblies are thought to contribute to cerebrovascular dysfunction.

Accumulation of Aβ has been proposed to exert neuronal toxicity through various means, and it is thought that assemblies of oligomeric Aβ result in activation of microglia and astrocytes, oligomerization and aggregation of tau protein, and progressive neuronal loss. Alternatively, accumulation of fibrillar Aβ in the vasculature has been proposed to impact blood vessel integrity and function.^, Deposition of amyloid in capillaries has been linked to degeneration of the lumen, endothelium, and basal lamina, resulting in ischemia and neurodegeneration. Amyloid fibrils found in the perivascular space, indicating dysphoric changes, have also been linked to neurite dystrophy. In larger vessels, Aβ deposition is associated with degeneration of vascular smooth muscle cells. Furthermore, it is thought that accumulation of Aβ in the vessel wall can impair perivascular draining of Aβ, further promoting accumulation of protein.

Numerous protein-based studies have highlighted CAA-specific differential regulation of proteins in addition to amyloidogenic proteins, with many of the differentially regulated proteins related to extracellular structure and matrix organization. Several proposed mechanisms exist to explain the increase in ECM proteins in CAA. For example, this could be explained by either increased secretion of extracellular matrix components or decreased turnover.

One model suggests that tissue injury or cellular processes result in enhanced synthesis of ECM proteins. Disruption of the BBB has been implicated in CAA both with and without AD pathology, with identification of BBB leakage markers, such as fibrinogen, and a decrease in tight junction proteins (occludin, claudin-5, and tight junction protein 1) important for maintenance of the BBB.^{,  ,}  Thus, it is conceivable that enhanced cross-linking of ECM proteins might result as a response to strengthen weakened vessel walls and prevent further BBB leakage. Covalent protein cross-linking results in increased formation and stability of extracellular protein complexes. Tissue transglutaminase is a known modifier of proteins implicated in AD,^{,  ,}  and enhanced tissue transglutaminase activity has been identified in hereditary cerebral hemorrhage with amyloidosis, Dutch type, and CAA with AD pathology. Tissue transglutaminase cross-links proteins, such as Aβ and APP, and additional ECM proteins, such as fibronectin and laminin, potentially leading to their enhancement in CAA.^,

In the ECM, tissue transglutaminase and tissue injury are also capable of activating TGF-β, which stimulates synthesis of ECM proteins and protease inhibitors that prevent enzymatic breakdown of the ECM.^, In support of this notion, expression of TGF-β1 and TGFβR2 is increased in hereditary cerebral hemorrhage with amyloidosis, Dutch type, a genetic form of CAA that results in both accelerated and more severe CAA pathology and symptoms. In addition, immunohistochemical staining of downstream TGF-β pathway signaling molecules, such as phosphorylated SMAD2/3, and TGF-β regulated proteins, such as fibronectin, collagen, and TIMP3, has been identified.^,,, Similarly, TGF-β activity has been implicated in other ECM affecting diseases, such as AD, CADASIL, and CARASIL.^,

Another potential explanation for the increase in ECM proteins is the decreased turnover and removal of ECM components. This could be due to aggregation and sequestration of chaperones, contributing to further aggregation and deposition of proteins. In CAA, enrichment of extracellular chaperones, such as clusterin, APOE, and HTRA1, was identified.^, APOE features chaperone function that targets Aβ assemblies, and clusterin and HTRA1 are also thought to be involved in removal of protein aggregates.^, Sequestration of protein chaperones can thus contribute to aggregation of Aβ and other extracellular proteins. In addition, tissue endogenous inhibitors (TIMPs) and matrix metalloproteinases (MMPs) are involved in regulation of the ECM, and dysregulation of these proteases can result in decreased turnover of ECM components and/or damage the integrity of the BBB, contributing to lobar hemorrhage.^,, MMP2 and MMP9 are enhanced in CAA with AD pathology, and matrix metalloproteinase inhibitor TIMP3 is enhanced in CAA both with and without underlying AD pathology.^,,, MMP2 and MMP9 have also been shown to proteolyze various Aβ peptides in AD.^, Thus, dysregulation of proteases can contribute to accumulation of ECM proteins identified in CAA.

CADASIL and CAA can be clearly distinguished by histologic evaluation. Key histologic findings of CADASIL include accumulation of NOTCH3 protein ectodomain within the vasculature and the pathognomonic presence of ultrastructural extracellular GOMs.^, In addition, CADASIL vessels demonstrate dramatic intimal hyperplasia and accumulation of intimal proteins (Figure 1A).^, In comparison, CAA is characterized by cerebrovascular accumulation of amyloidogenic protein, such as Aβ (Figure 1A).^, Routine histopathologic distinction is straightforward, because CADASIL vessels have not been shown to contain amyloid.

There is significant clinical overlap in CADASIL and CAA disease presentation, with both disorders resulting in increased risk of dementia, stroke, and intracerebral hemorrhage.^, At the histopathologic level, CADASIL and CAA both harbor abnormal accumulation of proteins within the vessel wall, although in CADASIL the major protein is NOTCH3 ectodomain and in CAA the major protein is Aβ (Figure 1A).^, Furthermore, both diseases feature degeneration of the vascular smooth muscle cells with accompanying hyalinization of vessels (Figure 1A).^, Finally, many characteristics of CADASIL at the molecular level are highly reminiscent of known amyloid pathology. For example, as discussed above, abnormal protein cleavage, processing, and oligomerization have been identified in both CADASIL and CAA. Both disorders are also thought to recruit abnormal accumulation of additional proteins other than NOTCH3 and Aβ. Given the numerous common qualities, shared molecular mechanisms may be involved in CADASIL and CAA. Thus, published work was reviewed to compare and contrast the protein profiles of the two cerebrovascular diseases.

A recent review of CADASIL and CAA proteomics-based studies by Haffner identified six proteins that demonstrated increased abundance in both diseases compared with controls: SAP, TIMP3, VTN, APOE, clusterin, and HTRA1. To expand on this, both proteomics-based studies and immuno-based studies of human CADASIL and CAA tissue were examined. In addition, CAA studies that did not include patients with underlying AD pathology were specifically targeted to compare CADASIL and CAA as primarily vascular disorders. The criteria for inclusion of studies in this review can be found in Table 1.^{,,  ,  ,,,,,,,,,,  ,  ,,,,,,  ,  ,,,  ,  ,  ,  ,  ,  ,  ,  ,  ,  ,  ,  ,  ,  ,  ,  ,}  Limitations of including immuno-based studies include the lack of specificity for some protein subtypes. For example, numerous immunohistochemical studies have observed enrichment of various collagens in both CADASIL and CAA. However, unlike with proteomics-based approaches, immunohistochemical studies often cannot discern the specific collagen subtype (eg, COL1A1 versus COL1A2). In these cases, only the more specific collagen subtypes from liquid chromatography–tandem mass spectrometry studies were included, when available. When unavailable, the general protein (eg, COL4) was included. In later gene expression analyses, all subtypes of the protein were included. Occasionally, studies indicated opposite directions of change for some proteins (eg, glial fibrillary acidic protein and basement membrane–specific heparan sulfate proteoglycan core protein). These cases were noted, and the proteins were graded in the direction supported by most studies (≥50%).