Cerebral amyloid angiopathy

Cerebral amyloid angiopathy (CAA) remains a largely untreatable disease often not diagnosed until autopsy. It ranges in severity from asymptomatic amyloid deposition in otherwise normal cerebral vessels to complete replacement and breakdown of the cerebrovascular wall.^1,2 Severe CAA can cause lobar cerebral hemorrhage, transient neurologic symptoms, and dementia with leukoencephalopathy.³

Estimates of the incidence of CAA can be derived from data on lobar hemorrhages in the elderly. Data from population-based studies give annual incidence rates for symptomatic lobar hemorrhage in patients over age 70 years of 30 to 40 per 100,000.^4,5 This estimate(which may include non-CAA lobar hemorrhages, but excludes clinically silent lobar hemorrhages) accounts for one-third to one-half of all primary intracerebral hemorrhages in this age range.^4,5

Another approach to assessing the frequency of CAA is to analyze the prevalence of pathologically advanced CAA at autopsy. In 784 postmortem cases, the estimated prevalence of moderate to severe CAA was 2.3% for those aged 65 to 74 years, 8.0% for ages 75 to 84, and 12.1% for ages≥85.⁶ CAA is particularly common in association with AD. Consecutive autopsies of AD patients showed 30 of 117 (25.6%) with moderate to severe CAA and a surprisingly large proportion (6 of 117, 5.1%) with CAA-related intraparenchymal hemorrhage.⁷

These data suggest that CAA is not a rare clinical entity in the elderly and that its incidence will increase as the population ages. The clinical impact of CAA will likely be heightened by increasing use of anticoagulant and thrombolytic therapies.⁸ All of these factors point to the urgency for devising effective therapies for CAA, arguably the only major type of stroke without either preventative or acute treatment.

Diagnosis of CAA-related hemorrhage. Analysis of biopsy tissue. Pathologic tissue becomes available during hematoma evacuation or cortical biopsy. The sensitivity and specificity of biopsy diagnosis in CAA were estimated by examination of random tissue samples from postmortem brains and found to vary according to the severity of CAA used to define a positive result.⁶ The presence of at least some congophilic staining was highly sensitive for the diagnosis of CAA-related hemorrhage; more advanced CAA increased the specificity of diagnosis. Other factors influencing sensitivity and specificity were the patient's age (the presence of CAA carrying greater specificity when identified in a relatively younger brain), the number of vessels examined, and the presence of meningeal tissue in the specimen (increasing the sensitivity for detecting CAA). The lack of perfect specificity supports use of the proposed term "probable CAA with supporting pathology".

Molecular risk factors for CAA. One goal of CAA research is to identify specific circulating markers (genetic and nongenetic) that can provide accessible information about the disease's presence, prognosis, or response to therapy. Several potential nongenetic markers were investigated in small numbers of patients. CSF showed decreased concentrations of the soluble amyloid precursor protein (APP) in three individuals with a familial form of CAA-related hemorrhage⁹ and decreased amyloidβ-peptide (Aβ) and apolipoprotein E in seven cases of sporadic CAA without hemorrhage.¹⁰ The contribution of these measurements to the diagnosis of hemorrhage in a clinical setting has not been determined. Ongoing studies are focused on the Aβ species in plasma, reported to be elevated in familial but not sporadic AD.^11,12

Several genetic mutations associate with dominantly inherited CAA-related hemorrhage. These include two mutations within the genetic segment coding for Aβ^13,14 and one in the gene for the protease inhibitor cystatin C.¹⁵ These mutations are absent (with one reported exception¹⁶) from patients with the more common sporadic form of CAA-related hemorrhage^17,18 and thus not helpful in most diagnostic situations.

Apolipoprotein E genotype (APOE), although neither sensitive nor specific for the diagnosis of CAA, has emerged as a genetic risk factor of this disease and a marker of its aggregate presence in groups of patients(see below). Studies in AD identified higher than predicted frequencies of the APOE ε4 allele and lower frequencies of the ε2 allele,^19,20 suggesting specific roles forε4 in promoting and ϵ2 in protecting from AD pathology. Work from several groups showed an association as well between CAA and APOE, but as first noted by Nicoll et al.,²¹ one that differs from AD by the overrepresentation rather than underrepresentation of theε2 allele. In 68 reported cases of pathologically supported (biopsy) or verified (autopsy) CAA-related hemorrhage with APOE genotype,^21-24 the allele frequencies of both APOE ε4 (0.26) and APOEϵ2 (0.17) were significantly greater than corresponding frequencies in the general population (0.14 and 0.08²⁵) or a set of elderly control patients (0.07 and 0.09,²⁴ all p < 0.05).

Patients with probable CAA who carried the APOE ε2 orε4 alleles were also 6 to 7 years younger at their first hemorrhage than those with the ε3/ε3 genotype.²⁴ Interestingly, the earliest ages at onset were in the small group of patients with the rare ε2/ϵ4 genotype, suggesting that the two risk factors might be particularly potent when combined. The possibility that the APOE alleles might similarly affect timing of hemorrhage recurrence in CAA is under investigation.

Radiographic diagnosis of CAA. There is no radiographic technique for detecting amyloid deposition in cerebral vessels. Contrast angiography is normal in CAA, with the exception of rare cases of CAA-related vasculitis.²⁶ An alternative approach is to label vascular amyloid for in situ imaging, a method used to mark presence and progression of the systemic amyloidoses.²⁷ Arterially injected Aβ²⁸ and cisternally injected antibodies to Aβ²⁹ have been found to aggregate with preexisting amyloid deposits in aged primate brains, but not in humans.³⁰

An alternate approach focuses on radiographic demonstration of the pattern of hemorrhages characteristic of CAA. CAA-related hemorrhage, like vascular amyloid itself, favors the cerebral cortex and corticosubcortical or lobar regions of the brain.^31,32 In contrast, CAA is typically sparse or absent in brain locations characteristic of hypertensive hemorrhages, such as putamen, thalamus, pons, and cerebellum. Based on the characteristic distribution of CAA and the observed tendency for the hemorrhages to recur, the term "probable CAA-related hemorrhage" was proposed for multiple hemorrhages entirely restricted to lobar regions in an elderly(age ≥60 years) individual without other known cause of hemorrhage.²² The goal of the "probable" diagnosis of CAA, it should be noted, is not 100% accuracy, but rather to provide enough information to support clinical decision-making and research.

The most useful radiographic technique for establishing the distribution of hemorrhages is gradient-echo MRI. Gradient-echo is an MRI technique that enhances the signal dropout caused by the deposited iron products that characterize chronic hemorrhage. The signal changes detected by gradient-echo MRI, like the iron-containing deposits themselves, are essentially permanent. The technique is thus well-suited to assess the full set of hemorrhages that may have occurred over a period of years.

Our experience since initial report of gradient-echo MRI in elderly patients with lobar hemorrhage³³ indicates that the study demonstrates one or more previous hemorrhages in approximately three-quarters of such individuals and a pattern suggestive of CAA in more than half. In a prospective series of 50 patients age ≥60 years presenting with primary lobar hemorrhage, 28 (56%) demonstrated the "probable CAA" pattern of multiple hemorrhages restricted to the lobar regions. An additional 12 patients (24%) demonstrated only the single presenting lobar hemorrhage on their study; the remaining 10 (20%) showed at least one prior hemorrhage involving a deep region not typical of CAA.

Analysis of APOE genotypes in these patients gives support to the proposed criteria for diagnosis of probable CAA. Thus the frequencies of APOE ε2 (0.18) and ε4 (0.20) in the group of patients with multiple hemorrhages confined to lobar regions are similar to those noted above for pathologically diagnosed CAA (0.17 and 0.26 respectively) and significantly greater than those in populations of elderly controls or hypertensive hemorrhage patients.²⁴

In addition to its role in diagnosis of CAA, gradient-echo MRI may be able to mark the disease's progression by the appearance of new hemorrhages. In a pilot study,³⁴ 7 of 14 patients followed with a second gradient-echo examination after a 1.5 year mean asymptomatic interval were found to have between 1 and 10 new petechial hemorrhages. Interestingly, the risk of hemorrhage recurrence appeared to depend in part on the number of hemorrhages at baseline, with those having more hemorrhages initially likely to develop more new hemorrhages at follow-up. These data hold the prospect that small hemorrhages could be to CAA research what demyelinating plaques have been to multiple sclerosis: a marker of disease progression that could serve as the basis for pilot drug trials. From a clinical standpoint, these findings suggest that the number of hemorrhages at baseline may have prognostic value in assessing an individual's risk of hemorrhage recurrence.

Treatment of CAA-related hemorrhage. Acute treatment. Whereas the biology of vascular damage in CAA appears to be distinct from other types of hemorrhage, there is no evidence for differences in behavior of the acute hematoma. Early hematoma growth, for example, appears to occur with similar frequency in lobar and deep hemorrhages.³⁵ With regard to the potential risks of surgical trauma, recurrent hemorrhage following resection appears to be no more frequent in CAA than in other forms of intracerebral hemorrhage.³²

Preventative treatment. Several (but not all) outcome studies have suggested lower mortality and better functional recovery in lobar compared with deep hemorrhages.³⁶ Recurrence of lobar hemorrhage, however, is relatively common, occurring in one series in 16 of 42 patients (38%) and with high mortality (7 of 16, 44%).³⁷

Current treatment for preventing recurrent CAA-related hemorrhage is limited to withdrawal of anticoagulant or antiplatelet agents. This step is not necessarily straightforward, as when the presentation suggests other diagnoses such as transient ischemic attack³ or hemorrhagic conversion of ischemic infarction. Gradient-echo MRI can be helpful in these situations to establish the presence of previous hemorrhages suggestive of CAA. Another challenging situation arises in patients with both CAA and a valid indication for anticoagulation such as atrial fibrillation. In these cases, the high mortality of warfarin-associated hemorrhage³⁸ represents a reasonably strong contraindication to this agent.

Future approaches to preventing CAA progression are likely to focus on blocking the specific molecular and cellular steps in its pathogenesis. The pathogenesis of CAA can be divided into 1) those steps involved in the production and deposition of vascular amyloid and 2) the subsequent steps that cause amyloid-laden vessel to rupture.

Amyloid accumulation is typically patchy and segmental, such that even advanced cases will have both affected and unaffected vessels. A quantitative comparison of vessels in early and advanced cases of CAA recently observed an increase in the amount of amyloid per vessel rather than in the number of vessels affected.³⁹ These results suggest that progression of CAA may involve expansion of previously existing amyloid deposits in a susceptible subgroup of cortical vessels. The same analysis indicated that the presence of APOE ε4 enhances this aspect of CAA's pathogenesis, with progressively more amyloid per vessel noted in brains carrying 1 or 2 copies of this allele.

A potential pharmacologic target for both CAA and AD is the production of Aβ, the 39-43 amino acid constituent of vascular amyloid and senile plaques. Several features distinguish Aβ in CAA from its counterpart in plaques, including the predominance of the Aβ species ending at amino acid positions 39-40^39,40 and the tendency of particular mutant forms of Aβ to accumulate preferentially in vessels.^13,14

Data derived from transgenic mice highlight the importance of proteins other than Aβ in amyloid accumulation. A role for apolipoprotein E in this process has been established by the finding that its genetic removal prevents amyloid deposition.⁴¹ Another line of mice was engineered to overexpress both APP and the cytokine transforming growth factor-β1.⁴² These mice unexpectedly developed vascular deposits of Aβ at 2 to 3 months of age, significantly earlier than mice overexpressing APP alone. This finding suggests that vasoactive cytokines may significantly regulate the extent of CAA, possibly through modulation of the vascular extracellular matrix.⁴³

Among the first vascular changes associated with Aβ deposition is death of the vascular smooth muscle cells, perhaps reflecting direct toxicity of Aβ. In smooth muscle culture, toxicity appears related to unaggregated Aβ⁴⁴ rather than the aggregated form considered to be the toxic species in AD. Another experimental system has demonstrated production of superoxide radicals by vascular endothelial cells treated with Aβ, resulting in both endothelial damage and alterations in vascular contractility.⁴⁵ This finding is consistent with studies suggesting that antioxidant treatments limit the toxicity of Aβ⁴⁶ and may slow the progression of AD.⁴⁷

Advanced cases of CAA demonstrate structural changes to the walls of the amyloid-laden vessel such as cracking between layers (creating a"vessel-within-vessel" appearance), microaneurysm formation, and fibrinoid necrosis.^1,2 An association has recently emerged between the APOE ε2 allele and the presence of CAA-related vasculopathic changes,²⁴ suggesting that the mechanism for ε2's effect on risk of hemorrhage may be enhanced breakdown of amyloid-laden vessels. Another possible contributor to the CAA-related vasculopathy is the inflammatory response. Whereas frank inflammation^18,48 occurs rarely in CAA, studies of vessel-associated monocyte or microglial cells^48-50 suggest that an Aβ-induced immune response may be a general feature of CAA-related hemorrhage.

Based on the above sequence of pathogenic steps, it is possible to outline potential approaches to prevention of CAA progression. Prevention of amyloid accumulation might be achieved by inhibitors of Aβ cleavage from APP, potentiators of Aβ clearance, antagonists of vasoactive cytokines, or inhibitors of Aβ binding to apolipoprotein E or the vascular extracellular matrix. Breakdown of amyloid-laden vessel walls might be prevented by inhibitors of Aβ toxicity, antioxidants, or anti-inflammatory agents. It should be noted that these strategies are based on independent aspects of the biology of CAA and might be combined with additive benefit.

The treatment approaches, diagnostic methods, and disease markers described above form a foundation for designing pilot clinical drug trials in CAA. Because of overlap in the pathogeneses of CAA and AD, it is likely that new drugs developed for AD will be candidate treatments for CAA. Indeed, CAA might offer certain advantages in testing of drugs, as the measurement of hemorrhages by MRI may have less day-to-day variability than the cognitive tests used to evaluate AD. The noted differences between the pathogeneses of the two disorders, however, including potentially different roles for the 40 amino acid and unaggregated forms of Aβ, apolipoprotein E2, and the process of vascular breakdown itself, suggest that application of drugs from one disease to the other be viewed with care.

Acknowledgments

I am grateful to Drs. Carlos S. Kase, Bradley T. Hyman, G. William Rebeck, and Merit E. Cudkowicz for critical review of earlier drafts of this article and to my many other collaborating investigators.