To explore the formation of amyloid plaques and to determine the effects of newly formed dense-cored plaques on the microarchitecture of the brain, we have developed a novel in vivo
multiphoton imaging technique. This recognizes newly formed plaques and allows monitoring of their immediate vicinity thereafter to determine the rate of their formation and the temporal sequence of pathophysiological events. We imaged young (5- to 6-month-old) B6C3-YFP mice, an age when plaques begin to appear4
(). We used three-colour imaging to establish fiduciary markers for repeated imaging: YFP positive neurons, dendrites and axons in the cortex, methoxy-XO4-labelled fibrillar amyloid-β deposits in the parenchyma and on vessel walls, and a fluorescent angiogram with Texas red dextran to image blood vessels. A volume of cortex (lamina I–III) that initially did not contain plaques was re-imaged until repeat imaging detected a new plaque, establishing its ‘birthday’. To ensure that the appearance of a new plaque did not simply reflect a greater depth of imaging or a slightly different imaging volume, we went through each image stack and compared them with previous sessions. New plaques were accepted only if a uniquely identifiable fiduciary point, such as a blood vessel or a dendritic process, could be unambiguously noted in a deeper imaging plane.
Appearance of a novel plaque is a rapid process
We postulated that we would occasionally observe the appearance and growth of new plaques within an imaging volume if the time interval between imaging sessions was long enough. From one weekly imaging session to the next, most of the sites remained unchanged (Supplementary Fig. 1a–c
). However, we identified 14 new plaques: instances in which a plaque appeared in a second imaging session in a volume that had clearly been unoccupied in the first images one week earlier ().
We examined the spatial relation between newly identified plaques and blood vessels. Measurements of the distance between vessel wall and the edge of a plaque confirmed that dense-core plaques develop close to but not within blood vessels (9.1 ± 3.9 µm from blood vessels). As a control, 70 randomly placed, plaque-sized objects had an average distance of 8.4±11.2 µm from a vessel. New plaques therefore do not form any closer to vessels than would be expected by chance, in accord with an earlier study of human Alzheimer’s disease 5
. Furthermore, multiphoton microscopic images showed that newly formed plaques were not penetrated by blood vessels6
, suggesting only a limited direct role of blood vessels in the formation of dense-core plaques.
To examine whether the phenotype of plaque formation in as short a period as one week was unique to the aggressive APP/PS1 transgenic mouse model, we used a mouse line that has a slower progression of disease (Tg2576)7
. Seven Tg2576 transgenic mice (11 months) were imaged weekly, and fourteen additional new plaques were observed, suggesting that the rapid plaque formation is not restricted to one mouse model (Supplementary Fig. 2
We next imaged the B6C3-YFP mice on a daily basis for up to six days in a row and/or on a weekly basis for up to three weeks (). To our surprise, senile plaque formation is a very rapid event, with five new plaques appearing precipitously within 24 h after the last imaging session. However, plaque formation is a rare event. Combining our experiments of daily and weekly imaging, a total of 26 new plaques were found in 14 animals over 238 sites, imaged a total of 1,285 times.
These newly formed plaques were then re-imaged over days to weeks to determine their growth pattern. Measurements of plaque area over several imaging sessions revealed that they do not change in size after about the first 24 h, regardless of whether they had small or large diameters when they were first imaged (). To examine whether imaging procedures had affected the observed plaque characteristics, we compared in vivo
measures with those obtained from post-mortem analysis of mice either after our imaging protocols or with naive mice that had never been imaged. These analyses confirmed that the newly formed plaques are not significantly different in area or diameter from those observed post mortem and therefore represent typical plaques (Supplementary Fig. 3
). Furthermore, the size distribution of new plaques compares well with that of all plaques seen post mortem in non-imaged mice (Supplementary Fig. 3
). The shape of this histogram is also quite similar to that observed in analyses of human patients with Alzheimer’s disease8
Several studies have shown that microglia surround senile plaques both in patients with Alzheimer’s disease and transgenic mouse models 9,10
although their role remains elusive11,12
. It has been proposed that activated microglia clear amyloid deposits13
, are the nidus for initiation of amyloid fibrils14
or restrict plaques15
. To study the interaction between newly formed amyloid plaques and microglia, we imaged microglia before and after plaque formation () in PDAPP mice crossed with a line that expresses enhanced green fluorescent protein (EGFP) in microglia16,17
. In this third APP transgenic line, we again found rapid formation of new plaques, with seven new plaques occurring within 24 h. Microglia were attracted to the site of plaque formation within a day (). None of the new plaques occurred immediately adjacent to resident microglia, suggesting that microglia do not form the nidus of new plaques. We did not observe any plaques being cleared by this microglial response. This suggests that, unless further activated18
, microglia do not successfully clear plaques but instead may well restrict their growth, leading to the observed steady state of plaque size after initial formation.
Microglia recruitment follows plaque formation
Stokin et al.
recently proposed that amyloid deposits followed axonal trafficking defects, marked morphologically as neuritic dystrophies. In contrast, amyloid deposition has been postulated to cause neuronal alterations1,19
. To study the temporal relation between newly formed plaques and dystrophic neurites, we compared the shape and trajectories of yellow fluorescent protein (YFP) fluorescent neurites before and after plaque formation in B6C3-YFP animals (). These neurites were morphologically normal in the volume of cortex one day before plaque formation. By contrast, in the days after plaque formation, progressive neuritic alterations were evident: from a smooth bend around the new plaque to very tortuous changes identical in appearance to dystrophic neurites seen in Alzheimer’s disease20
. The degree of neuritic deformation is best illustrated in individual image slices () as well as in three-dimensional reconstructions ().
Plaque formation has no immediate effect on neuritic curvature
Neurite curvature was quantitatively analysed20,21
from the in vivo
images around each new plaque (n
= 12) as well as more than 50 µm away from it, and compared with equivalent measures in randomly selected fields from YFP control mice. As expected, neurites were essentially straight, and unchanged in morphology over several weeks of imaging in YFP mice. In the B6C3-YFP mice, in the immediate vicinity of plaques the tortuosity increased gradually over the first week (, P
< 0.0001). Post-mortem analysis of neuritic curvature (n
= 218) taken of randomly selected plaques observed at the time of death confirmed that neuritic curvature measured around new plaques one week after their formation is indistinguishable from the same measurements taken around all plaques (1.048±0.02 versus 1.052±0.014). This suggests that neuritic changes occur rapidly and to essentially a maximal extent over the first week of a plaque’s presence.
We next analysed changes in neurites near plaques on a daily basis. No change from baseline in neuritic curvature was detected on the day of plaque appearance (). Neuritic alterations were first evident one day after the occurrence of a new plaque, indicating that neuritic deformation is a secondary effect of plaque development (n
= 15; analysis of variance (ANOVA P
= 0.02)). After two days the damage became more prominent; after five days it resembled the more robust neuritic damage seen after one week (). Thus the neurite changes progressed over days after a short lag phase (although the imaging would not detect more subtle changes in cytoskeleton that might occur even earlier). In several instances, frank neuritic dystrophies appeared on a previously normal dendrite three to four days after the first appearance of a new senile plaque (). In almost all instances of new plaque formation (11 of 13 new plaques in which concurrent observation of YFP filled neurites was obtained), frank neuritic dystrophies followed plaque formation within one week (Supplementary Fig. 4a, b
< 0.001). Post-mortem immunostaining suggested that both dendritic and axonal elements contribute to these dystrophic neurites (Supplementary Fig. 5
Axonal dystrophies can be observed both near plaques and in the neuropil without a plaque. Such dystrophies have been hypothesized to anticipate the location of plaques2
. We tested the idea that areas with a high density of dystrophies would be a prime location for plaque development. We tracked sites containing dystrophies but no plaques over days to weeks, but never observed a new plaque appearing at a site of high dystrophic neurite density. Instead, it seemed that these dystrophic processes are quite dynamic. Out of ten examples, 40% of the dystrophic neurites changed morphologies (with some even resolving completely) and only 60% were unaltered over one to two weeks of imaging (Supplementary Fig. 4c–f
Because our in vivo
amyloid imaging agent, methoxy-XO4, would not report amyloid-β deposits that do not contain β-pleated sheets, we compared immunostaining for amyloid-β and methoxy-XO4 in histologic specimens to look for diffuse plaques in 8-month-old B6C3-YFP mice. All of the plaques were dense-core in morphology, with co-staining of amyloid-β immunofluorescence and methoxy-XO4 (Supplementary Fig. 6
). Because these mice develop innumerable plaques, the absence of any plaques with a diffuse morphology argues against the idea that dense-core plaques represent a remodelled form of diffuse, ‘primitive’ plaques.
Longitudinal in vivo
imaging of plaque formation provides a new appreciation for the kinetics of the amyloid deposition process in vivo
, and demonstrates the temporal relations between amyloid deposits, microglial recruitment and activation, and neuritic changes. Although it takes months for many plaques to accumulate, even in accelerated mouse models of Alzheimer’s disease7,16
, our data show that an individual dense-core plaque’s formation unexpectedly represents an acute event. These results are surprising because it has been generally accepted, based on in vitro
studies of protein aggregation, that amyloid-β aggregation is time dependent and follows a relatively slow nucleation-dependent polymerization process22
. It is possible that submicroscopic nuclei may be present in the cortex23
, perhaps related to the recently reported amyloid-β oligomeric forms24,25
, as precursors to this sudden growth.
Observing the ‘birthday’ of new plaques provides the opportunity to examine directly whether dystrophic neurites near amyloid deposits precede or follow amyloid deposition. Our data strongly argue in favour of the latter, and suggest a period of several days after plaque formation during which progressive cytoskeletal derangements occur in neurites near a plaque. The observation of a local microenvironment in which microglia are recruited and activated after plaque formation further supports a model in which plaques act as a reservoir of bioactive molecules (Supplementary Fig. 7
), which subsequently lead to neuronal alterations19
including local loss of dendritic spines and axonal dystrophies26
. These data thus lead to a model consistent with a prediction of the amyloid hypothesis in which amyloid deposition, activation and recruitment of microglia and local neuritic changes play out as a sequential cascade leading to neurodegeneration1
The current observations provide narrow parameters on the kinetics of plaque growth and stabilization through microglia, and raise the possibility that the years-long degenerative process of Alzheimer’s disease is marked by innumerable sudden changes in cortical structure. Altering the kinetics of this process may well change the rate of progression of Alzheimer’s disease.