Cerebral amyloid angiopathy : new insights from transgenic mice

Cerebral amyloid angiopathy (CAA) is characterized by the deposition of congophilic material
within the walls of small to medium-sized blood vessels of the brain and leptomeninges. The
incidence of CAA increases with aging, and in its most severe stages, the vascular amyloid
causes a breakdown of the blood vessel wall which results in spontaneous, often recurrent, lobar
intracerebral hemorrhage. CAA is estimated to account for four to twenty percent of all
nontraumatic intracerebral hemorrhages. Besides this major complication, extensive CAA has
been associated with ischemic white matter damage with progressive dementia, perivascular
inflammation, and secondary vasculitis. CAA occurs as a sporadic disorder in the elderly and in
association with Alzheimer's disease (AD) with virtually all AD patients showing some degree of
vascular amyloid in addition to parenchymal plaques. There are also familial forms of CAA such
as hereditary cerebral hemorrhage with amyloidosis-Dutch type (HCHWA-D). The vascular
amyloid in these disorders mainly consists of β -amyloid peptide (Aβ ) that is produced by
proteolytic cleavage from its precursor, which is the β -amyloid precursor protein (APP). The
major Aβ species that is deposited in the vasculature is Aβ 40, while parenchymal amyloid is
mainly composed of Aβ 42. One major difficulty in studying CAA is that it can be definitely
diagnosed only postmortem. Moreover, spontaneous CAA occurs only in old primates and dogs,
both of which are not practical models to study the pathogenesis and therapy of CAA. Rodents
do not spontaneously develop CAA.
The purpose of this thesis was to provide useful model systems to study the pathomechanism of
vascular amyloid formation and associated pathology. To this end we generated and used mice
that are transgenic for human genes bearing mutations that are well known to cause either
hereditary Aβ -CAA or classical familial AD. In a first study we analyzed CAA and CAA-associated
pathological changes in APP23 transgenic mice. These mice overexpress human APP bearing the
Swedish K670N/M671L double mutation, a typical early-onset AD-causing mutation, under the
control of the neuron-specific Thy-1 promoter. In addition to parenchymal amyloid plaques,
APP23 mice show consistent amyloid within leptomeningeal, neocortical, hippocampal, and
thalamic vessel walls. Both CAA frequency and severity significantly increase with aging,
demonstrating that not only more vessels are affected, but also that the amyloid burden of
individual vessels increases with the progression of amyloid deposition. Cerebrovascular amyloid
causes degeneration of vascular smooth muscle cells (SMCs). In severely affected vessels, SMCs
are completely replaced by the amyloid. Similar to humans, amyloid depositing APP23 mice
develop spontaneous hemorrhages, some of them being recurrent. The bleedings are associated
with amyloid-laden vessels and therefore, their anatomical distribution appears very similar to
that of CAA. In aged mice, a quantitative analysis revealed a positive correlation between
hemorrhages and CAA. Interestingly, no significant relationship between hemorrhages and total
amyloid load was observed. Occasionally, CAA-associated vasculitis is seen in animals with
extensive vascular amyloid.
In a second study, we generated transgenic mice that express human APP E693Q under the
control of the same neuron-specific Thy-1 promoter (APPDutch mice) that has been used in
APP23 mice. In HCHWA-D patients, the APP E693Q Dutch mutation causes severe CAA with
recurrent cerebral hemorrhagic strokes often leading to death early in their fifties, or to
dementia in patients that survive the strokes. In contrast to AD patients that show parenchymal
amyloid plaques, HCHWA-D patients exhibit few parenchymal amyloid deposits. Similar to
HCHWA-D, aged APPDutch mice show extensive Aβ deposits mainly within the walls of
leptomeningeal vessels followed by cortical vessels. Parenchymal Aβ deposits are mostly absent.
In severely affected vessels, the SMCs are completely displaced by the amyloid. In regions with
CAA, fresh and old hemorrhages are observed, and activated perivascular microglia and reactive
astrocytes are found. To examine the mechanism that leads to the almost exclusive vascular
amyloid formation in APPDutch mice, we compared the mice with transgenic mice
overexpressing wild-type (wt) human APP using the same neuronal promoter (APPwt mice). As
they age, APPwt mice develop parenchymal plaques with limited vascular amyloid deposits. A
biochemical analysis of Aβ 40 and Aβ 42 levels revealed significant higher Aβ 40:42 ratios in
amyloid depositing and pre-depositing APPDutch mice compared to APPwt mice. To
demonstrate that the high Aβ 40:42 ratio in APPDutch mice is linked to the almost exclusive
vascular amyloid deposition, we crossed APPDutch mice with mice that overexpress human
presenilin-1 bearing the G384A mutation (PS45 mice) that is known to dramatically increase the
production of Aβ 42. Strikingly, young APPDutch/PS45 double-transgenic mice develop massive
diffuse and compact parenchymal amyloid with only very little CAA. Thus, shifting the Aβ 40:42
ratio towards Aβ 42 is sufficient to redistribute the amyloid pathology from the vasculature to the
parenchyma.
A third series of experiments using neurografting techniques was performed to investigate the
mechanisms involved in the initiation of cerebral amyloidosis in vivo . Cell suspensions of
transgenic APP23 and wild-type B6 embryonic brain tissue were injected into the neocortex and
hippocampus of both APP23 and B6 mice, respectively. In wild-type hosts, APP23 grafts did not
show amyloid deposits up to 20 months after grafting. Interestingly, transgenic and wild-type
grafts in young APP23 hosts develop amyloid plaques as early as three months after grafting.
Although the majority of the amyloid is of the diffuse type, some compact and congophilic
amyloid plaques are observed in the wild-type grafts. These congophilic amyloid lesions are
surrounded by neuritic changes and gliosis, comparable to the amyloid-associated pathology
that has previously been described in APP23 mice. These results support the importance of
neuronally secreted Aβ for the development of cerebral amyloidosis which can be initiated
distant from the site of Aβ production, a finding that supports the observation of the above
mentioned APPDutch mouse model.
In summary, we demonstrate that APP23 and APPDutch mice recapitulate CAA and CAAassociated
pathology observed in humans and thus are valuable models for studying the human
disease. Our results stress the importance of neuronally secreted Aβ for the development of CAA
and emphasize the Aβ 40:42 ratio as an important factor in determining parenchymal versus
vascular amyloid deposition. The understanding that different Aβ species can drive amyloid
pathology in different cerebral compartments not only provides insights into the
pathomechanism of sporadic and familial CAA but also has implications for current anti-amyloid
therapeutic strategies.