ARTICLE

Although the mechanism whereby A deposition may lead to dementia in Alzheimer disease (AD) is unknown, compelling genetic evidence suggests that aggregation of A to form senile plaques (SP) is an essential component of AD pathophysiology (1-3). Biochemical studies suggest that these A deposits are insoluble, and their formation process is viewed as irreversible. From inspection of AD tissue samples, it is evident that a wide variety of morphologies and textures of SP are present in the AD brain. Their morphologies cannot be explained by known aggregation models (4-8).

A is a 39- to 42-amino acid amphipathic peptide derived from a portion of the transmembrane domain and extracellular region of the A precursor protein (9). A is a normal cellular product and is present in nanomolar concentrations in biological fluids (10, 11). In vitro, at higher concentrations, it is extremely insoluble and precipitates to form aggregates (12-15). In the AD brain, A deposits form -pleated, sheets which are the major constituent of SP. Racemized amino acids have been found in A, suggesting that at least some of the deposits are long-lived (16). Given the insoluble nature of A, it is reasonable to predict that plaques would continue to grow in size and number as the disease progresses. However, experimental data show that this is not the case; instead, plaque size and A burden (total percentage) appear to remain relatively constant over a wide range of disease durations (17-19).

Optical microscopy of the AD brain reveals innumerable A deposits of various sizes and shapes. In an effort to understand how A deposition occurs and evolves over time, we have examined the fine structure of the SP using confocal scanning laser microscopy (20, 21) and immunofluorescence techniques for A immunostaining. The confocal microscope is able to obtain optical sections that are 0.3- to 0.5-µm thick, allowing the reconstruction of the three-dimensional fine structure of a plaque with a resolution close to the theoretical limit of the order of the wavelength of visible light.

Standard immunostaining in thick sections might suggest that plaques are relatively solid (Fig. 1a), but our examination of individual cross-sections using confocal microscopy reveals cavities and inner structure, suggesting that the three-dimensional structure of A aggregates in SPs is porous [Fig. 1b]. By stepping through the A deposit in a chosen direction, sequential optical sections similar to those shown in Fig. 1c can be obtained and reconstructed to analyze the three-dimensional structure of the plaque (Fig. 1d).

The first approach to quantify the geometry of these individual plaques is to calculate the density-density correlation function (22) g(r), which is defined to be the probability that two points of space at a distance r are both part of the SP. Results from cross-sections of typical cortical samples (Fig. 3a) indicate that, in contrast to that of a solid disk (shown by a solid line), the correlation function displays three regimes: (i) a central region that approximates the solid disk curve and (ii) an inner and (iii) an outer regime that deviate. The deviations from the solid line in the small r regime indicate the existence of "pockets" of higher and lower density than the average density, corroborating the conclusion derived by simple inspection of confocal pictures (Fig. 1 b and c). For larger distances, the deviation indicates that the density of the plaque decays slowly (diffuse ring) as the distance from the center of the plaque increases.

To further study the characteristics of the inner structure of the plaques and thus gain some insight into the formation of SP, we performed the correlation function analysis again, but now exclusively over the cross-sectional area of the plaque confined to its interior. In this way, there are no surface effects on the calculation, so this analysis yields information only about the plaque's inner structure. Over 500 images were collected; correlation function analysis of middle sections of plaques of diameters ranging from 20 to 90 µm (Fig. 3b) indicated that the average linear size of these pores or "pockets" of different density is, at most, only weakly diameter-dependent, being roughly 5 ± 2 µm (see the Inset of Fig. 3b). Double immunostaining using 4, 6-diamidino-2-phenylindole for nuclei, LN-3 for microglia, or glial fibrillary acidic protein for astrocytes showed that these pores are infrequently occupied by cellular elements.

Thus, these quantitative analyses of plaque structure revealed two new features not evident by qualitative inspection. A typical plaque consists of (i) a porous core with pores of a characteristic size and (ii) a diffuse ring whose density slowly decays from the center of mass of the SP. Recognition of these two features immediately leads to the question of how these structures are formed in the AD brain and what kind of mechanisms could produce such morphologies. Consideration of general principles of aggregation leads to several possibilities that depend on the diffusion constant of A (4-8, 22-26). If the diffusion of the solute A is slower than the speed of aggregation, the growth will occur at the tips of the aggregate, leading to a ramified tree-like structure belonging to the diffusion limited aggregation universality class (5-8, 22-26) rather than the structure observed. This case is very unlikely to occur in AD brains because, it is believed, aggregation is a slow process that may continue for years while the diffusion of A in the brain is much faster. On the other hand, if the diffusion of A is faster than the aggregation, then A is equally likely to aggregate at any point on the surface of the SP. The outcome is a very compact spherical structure with only a few very small pores, belonging to the Eden universality class (5-8, 22-26), which is also quite different than the experimentally observed morphology. Physicochemical models based on nucleation-dependent polymerization, as suggested by Jarrett and Lansbury (27), would yield a compact object if the process is continued beyond the nucleation and growth steps.

We suggest an alternative possibility in which aggregation occurs simultaneously with disaggregation. Objects grown by our proposed process can lead to the formation of porous objects whose size distributions, number, and A burden are constant if the aggregation is in dynamic equilibrium with disaggregation (17). Our hypothesis is not inconsistent with the nucleation process proposed by Jarrett and Lansbury (27), but extends it by adding a disaggregation process that, once the thermodynamic equilibrium is reached, would yield a porous structure similar to that found experimentally. This also does not preclude the possibility that, in addition to a dynamic equilibrium between soluble and deposited A, some A undergo irreversible biochemical changes to long-lived species (16).

To corroborate the expected outcomes from these hypothesized competing aggregation and disaggregation mechanisms, we developed a dynamical model. The model incorporates the experimental observation that the amount of A burden varies within a narrow range and is independent of duration or severity of illness (28). In the model, an ensemble of plaques, as well as individual plaques of various sizes, are grown on a three-dimensional lattice. A collection of model plaques, grown in a computer simulation starting from a configuration of isolated occupied lattice points, is shown in Fig. 2a (to be compared with actual plaques shown in Fig. 1a). The size distribution of configurations of model plaques, like the one represented in Fig. 2a, was found to exhibit a peaked distributionin agreement with earlier experimental work (28). The two computer-generated model plaques presented in Fig. 2b show the importance of the inclusion of surface diffusion in the model, which allows the model plaques to acquire a smoother surface (note the difference between the two model plaques in Fig. 2b). In addition, as a consequence of the surface diffusion, the model exhibits well defined pores in its interior (lower model plaque in Fig. 2b). Fig. 2 c and d shows cross-sections and three-dimensional reconstruction of a typical model plaque, respectively (compare with Fig. 1 c and d). To make quantitative comparisons between the experiment and the model, in Fig. 3 c and d we present results for the correlation function of the model plaques. Similarly to Fig. 3a, Fig. 3c exhibits a porous core and diffuse ring around the core for the model plaques. Comparing the Insets of Fig. 3 b and d, we concluded that the model was able to reproduce the distribution of average pore and pocket sizes in the cross-sections of SPs.

In summary, using laser scanning confocal microscopy, we were able to obtain three-dimensional images of SP. Using correlation function analysis, we examined the fine structure of SPs, and we have discovered within the plaque's internal morphological structure the presence of characteristic size pores and pockets of higher density (Figs. 1 and 3b). This structure serves to decrease the likelihood of several potential formation mechanisms based solely on aggregation and favors instead a model in which both aggregation and disaggregation processes are in dynamic steady-state equilibrium. The requirement of a disaggregation mechanism in the model is consistent with the possibility that A deposition may not be irreversible. The plausibility of this model is further supported by (i) analyses of a small number of patients studied at both biopsy and autopsy, in whom a decrease (or no increase) in A deposits was seen (29, 30), and (ii) in vitro studies suggesting that microglia can respond to and ingest A aggregates (31, 32).