Association of apolipoprotein E ϵ2 and vasculopathy in cerebral amyloid angiopathy

Cerebral amyloid angiopathy progresses to cerebral hemorrhage through a specific series of pathogenic events. CAA itself, defined as deposition of congophilic material in the media of cerebral vessels,^1,2 is common in the elderly brain. CAA-related cerebral hemorrhage, however, occurs only in a relatively small subset of cases. Those brains that develop hemorrhage demonstrate both extensive vascular amyloid and a set of vasculopathic changes that can include cracking and fibrinoid necrosis of the vessel wall and paravascular leaking of blood.^3-6

Amyloid deposited in vessels in CAA, similar to amyloid in senile plaques in Alzheimer's disease (AD), is composed largely of the amyloid β-peptide(Aβ). We have previously proposed⁷ that CAA and AD might share some mechanisms and risk factors, particularly those related to Aβ deposition, and might differ in other molecular features, such as those related to the vasculopathic changes.

Apolipoprotein E (APOE for gene, apoE for protein) has been identified as an important risk factor for Aβ deposition and Aβ-related diseases. The ϵ4 allele of APOE is associated with increased Aβ deposition, increased risk of disease, and earlier onset in both AD^8-14 and CAA.^10,15-18 APOE ϵ2, conversely, appears to protect from senile plaques and AD.^11,12,19-22 These results have led to the suggestion that apoE might exert both its positive and negative effects on AD by regulation of amyloid deposition.

Surprisingly, however, the ϵ2 allele of APOE appears to be overrepresented rather than underrepresented in cases of CAA-related hemorrhage, as demonstrated first in pathologic specimens²³ and subsequently in a clinically defined series.¹⁸ This striking difference from AD has raised the possibility that apoE2 might affect molecular steps specific to CAA such as the degenerative changes in the amyloid-laden blood vessels. The latter possibility leads to a specific and testable prediction: that the ϵ2 allele be overrepresented in cases of CAA with vasculopathic changes such as cracking of the vessel wall and paravascular leaking of blood, relative to brains with extensive vascular amyloid but without these vasculopathic changes.

Methods. Pathologic and clinical cases of CAA. Postmortem cases were chosen from the Harvard Brain Tissue Resource Center(BTRC) at McLean Hospital (Belmont, MA). Brains had been systematically graded for CAA severity by review of 18 blocks of brain tissue, including seven from cerebral cortical regions and one from cerebellum as described in detail elsewhere.²⁴ Sections were prepared with Luxol fast blue, hematoxylin and eosin, and Congo red and examined with and without polarized light; some sections were also immunostained for Aβ(6F3D, DAKO, Carpinteria, CA 1:150) to confirm the identity of congophilic material. Cases were graded according to the most advanced degree of CAA present among all vessels examined, an approach previously shown to correlate with risk of hemorrhage.⁴ All pathologic grading was performed and recorded prior to determination of genotype.

To examine specific determinants of the vasculopathic changes in CAA, we selected only brains with complete amyloid replacement of vessel walls, represented by the CAA grades of moderate or severe.^4,15 Moderate CAA denotes complete replacement of the blood vessel media with amyloid but without evident vasculopathy(figure 1A), whereas severe CAA is defined by the additional changes of cracking of the vessel wall and at least one focus of paravascular blood leakage (figure 1B). Seventy-five brains were selected, including 28 from the set of pathologic cases reported previously¹⁵ (in which moderate-severe cases were analyzed as a single group) and an additional 47 representing all other moderate-severe CAA brains received by the BTRC between July 1993 and June 1996 with available frozen tissue for genotyping. Mean density of neuritic plaques was estimated according to Consortium to Establish a Registry for Alzheimer's Disease (CERAD) criteria,²⁵ with grade 1 = sparse density, 2 = moderate, and 3 = severe.

Clinical analysis of APOE genotype was performed on 48 consecutive patients age ≥50 years with a diagnosis of probable or definite CAA-related hemorrhage,¹⁵ 27 described previously¹⁸ and an additional 21 consecutive patients presenting between January, 1996 and February, 1997. These patients were compared with 49 patients age ≥50 with probable hypertensive (HTN) hemorrhage (18 described previously,¹⁸ 31 consecutive additional patients from the above interval), and 87 elderly volunteers without cerebral hemorrhage ("no hemorrhage") presenting for routine general examination during this interval. Mean age (± SD) of the CAA patients was 77.7 ± 8.1 years, for HTN 71.8 ± 10.9, and for the no-hemorrhage patients 69.0 ± 11.0. All but eight of the studied patients (four black, four Hispanic) were white. Clinical and radiographic diagnosis was performed as described previously.^15,18 History of prior dementia was ascertained by interview with family members. Determination of diagnosis, previous dementia, and age at first hemorrhage were recorded before analysis of APOE genotype, performed by polymerase chain reaction (PCR) as described previously.¹⁵ This study was performed with approval of the institutional review board and informed consent of participants or family members.

Statistical analysis. Allele frequencies (proportion of chromosomes in which the allele was present) were compared using 2 × 2 tables with Fisher's exact test for significance. Ages of first CAA-related hemorrhage were normally distributed (Shapiro-Francia test for normality) and compared with Student's t-test or multivariate linear regression. All significance tests were two-sided.

Results. Compared with patients with either HTN hemorrhage or no hemorrhage, those with CAA-related hemorrhage demonstrated significantly increased frequencies of both the ϵ4 and ϵ2 alleles(table 1). In comparison with the ϵ3/ϵ3 genotype, the presence of one copy of either APOE ϵ2 or ϵ4 increased the odds ratio for CAA-related hemorrhage by approximately four-fold. Allele frequencies similar to those in the full CAA cohort were seen in the subgroup of CAA patients without history of preceding dementia(0.21 for ϵ2, 0.24 for ϵ4, n = 36) and the subgroup of CAA-related hemorrhages diagnosed with definite or supporting pathologic evidence (15)(0.15 for ϵ2, 0.35 for ϵ4, n = 10).

Within the cohort of patients with definite/probable CAA, those who carried ϵ2 or ϵ4 had their first hemorrhage 6 to 7 years earlier than those with the ϵ3/ϵ3 genotype (table 2). The four individuals with the ϵ2/ϵ4 genotype demonstrated the earliest age of onset of any genotype (67.3 ± 4.0 years, p< 0.02 [corrected for multiple comparisons] versus ϵ3/ϵ3 patients). Multivariate analysis demonstrated independent and significant effects of APOE ϵ2 and APOE ϵ4 on earlier age of first hemorrhage. Neither allele significantly affected age at onset in the HTN hemorrhage group (data not shown).

Given the specific effect of APOE ϵ2 on the presence and onset of CAA-related hemorrhage and the absence of a potentiating effect of APOEϵ2 on vascular amyloid deposition,^15,22 we reasoned that this allele might exert its effect by accelerating the formation of the CAA-related vasculopathic changes that lead to hemorrhage. This hypothesis predicted that among brains with extensive vascular amyloid, APOE ϵ2 should be present at greater frequency in the subset of brains with vasculopathic changes than in brains without vasculopathy.

We therefore compared two groups of postmortem brains, each with complete replacement of vessel walls with amyloid: one group without evidence of vasculopathy ("moderate CAA," see figure 1A) and one with the CAA-related vasculopathic changes, such as cracking of the vessel wall that are associated with hemorrhage ("severe CAA,"figure 1B). The groups did not significantly differ in age at death (79.0 ± 7.0 without vasculopathy, 78.4 ± 9.9 with vasculopathy) or density of plaques (mean CERAD grades 2.6 ± 0.7 versus 2.6±0.9). Frequency of APOE ϵ2 was significantly greater in the brains with CAA-related vasculopathic changes than those without(table 3). APOE ϵ4 frequency was high in each group and did not significantly differ between brains with and without vasculopathic changes. These results are consistent with a role for APOE ϵ2 in promoting vasculopathy in CAA.

Discussion. The results presented offer several lines of evidence to suggest that APOE ϵ2, like APOE ϵ4, is positively associated with CAA-related hemorrhage: (1) it is overrepresented among CAA cases, confirming an independent series²³; (2) it correlates with age of first CAA-related hemorrhage (not seen in a previous smaller, less stringently defined cohort¹⁸); and(3) it is associated with CAA-related vasculopathy. Further series of patients will be required before this unexpected association is fully established, as highlighted by the absence of elevated ϵ2 frequency in a pathologic series of 13 CAA + AD cases.¹⁶ Althoughϵ2 may increase the risk of CAA-related hemorrhage, it is absent in the majority of patients (see table 1) suggesting the existence of other factors that can independently promote hemorrhage.

We found the frequency of APOE ϵ2 in the overall group of pathologic cases with extensive vascular amyloid to be relatively low (seetable 3), in keeping with this allele's protective influence on amyloid deposition.^12,15,22 The frequency of 0.09 observed in brains with CAA-related vasculopathy was, however, significantly greater than the frequency of ϵ2 seen in the cases without vasculopathy, and also above the range of allele frequencies (0.005 to 0.04) reported in brains with extensive amyloid deposition from Alzheimer's disease.^11,19-21 These results indicate that when ϵ2 is present in cases with extensive CAA, the vessels frequently proceed to develop vasculopathic changes, such as cracking.

The data suggest a model for apoE and CAA in which apoE4 promotes creation of amyloid-laden cerebral vessels and apoE2 causes the vessels to progress toward rupture. This model, by postulating an action of apoE2 specific to the CAA pathway, offers a potential explanation for this allele's overrepresentation in CAA-related hemorrhage but not AD.

The pathologic criteria used in the current study-cracking of an amyloid-laden vessel wall and paravascular leaking of blood-were chosen from among the described CAA-related vasculopathic changes (also including vascular ectasia, microaneurysms, infiltration of inflammatory cells, and fibrinoid necrosis³) because of their demonstrated association with CAA-related hemorrhage. A previous study using the same methods for analysis and grading found severe CAA as defined here in all 13 cases of hemorrhage with full postmortem examination and no other cause of hemorrhage, but only 10 of 70 brains from patients with other neurologic or psychiatric illnesses and only one of 66 elderly patients from a general hospital autopsy series.⁴ These vasculopathic changes thus appear to be a necessary precondition for CAA-related hemorrhage.

Differences between the Aβ-dependent pathologies of CAA and AD might pertain to the different effects of apoE2 on the two diseases. Vascular Aβ, unlike plaque amyloid, appears to contain large amounts of the 40 amino acid form of the peptide.^26-29 Further, data from cell culture suggest that cerebrovascular smooth muscle cells might be specifically sensitive to the unaggregated form of Aβ,³⁰ the opposite of findings from cultured neurons.³¹ ApoE is present in vessels early in CAA,^27,32 where it could affect the aggregation or other biological properties of Aβ.^33,34 It will therefore be interesting to explore whether apoE2 affects specific steps relevant to CAA, such as maintaining Aβ in a state toxic to smooth muscle cells.

Although no clear role remains for determination of APOE genotype in the diagnosis of CAA, its association with CAA-related hemorrhage has several potential clinical implications. Acting at different pathogenic steps in CAA, the apoE isoforms might identify patient subgroups with particular risks of hemorrhage recurrence. Elucidation of the molecular basis of vasculopathy in CAA might also identify new and useful targets for preventative treatments.

Acknowledgments

We are grateful to Larry Cherkas for preparation of figures, James A.R. Nicoll, MD, for sharing data prior to publication, and the Harvard Brain Tissue Resource Center (NIH grant MH/NS 31862).