From Dennis Selkoe: Principally from the extracellular space.
Amyloid b-protein (Ab) is generated intracellularly by constitutive proteolytic cleavages of the b-amyloid precursor protein, first by the protease(s) referred to as b-secretase(s) and then by the proteases called g-secretase(s). At present, it appears that the most abundant Ab species, Ab1-40, is generated in considerable part in the endosomal system and released into the extracellular space as endosomes recycle to and fuse with the plasma membrane. Ab40 can also be generated in the secretory (exocytic) pathway, including in the Golgi. The less abundant but pathogenically important Ab42 appears to be primarily generated in the secretory pathway, including the ER and Golgi, although it may be that Ab42 is also produced to some extent in other vesicular compartments. Ab42, like Ab40, is constitutively secreted from the cell, although it could be that some Ab42 accumulates inside cellular vesicles, presumably to be degraded, at least in normal cells. Although Ab exists both in the intra- and extracellular spaces, it is currently unresolved whether one or both of these loci are the principal sites for Ab aggregation and cytotoxicity.
I believe that evidence available at this time favors the hypothesis that Ab exerts its cytotoxic effects principally from the extracellular side of the plasma membrane. In my opinion, the following points support this hypothesis:
1. Ab is abundantly detected in AD brain in a variety of deposits outside cells. Most cases of AD have large numbers of diffuse and "mature" (neuritic/glial) plaques (and a spectrum of intermediate forms) in the extracellular space of hippocampus, amygdala, entorhinal cortex and cerebral neocortex. The same antibodies that readily detect even small deposits of extracellular Ab (small diffuse plaques, punctate or stellate deposits, wispy subpial deposits, etc.) in AD and aged brains fail to detect unambiguous intraneuronal Ab aggregates or deposits. Moreover, while extracellular ("ghost") tangles can show Ab immunoreactivity, intraneuronal tangles have not shown reproducible Ab reactivity.
2. Electron microscopy of AD brain (both biopsied and autopsied) has not revealed clear intraneuronal amyloid fibrils, even though these are readily detectable in large amounts in the adjacent extraneuronal space. PHF (and related straight filaments) containing tau proteins are easily visualized by EM inside many neurons in AD, but intraneuronal amyloid fibrils have not been detected in these (or any) neurons. While occasional reports of amyloid fibrils apparently inside microglia have been published, similar reports of fibrils inside neurons in AD are few or non-existent.
3. Down's syndrome (DS) shows great neuropathological similarity or identity to AD. Several immunocytochemical studies report some or many diffuse plaques of extracellular Ab in young DS brains (as early as age 12) -- prior to the appearance of compacted, neuritic plaques, neurofibrillary tangles and other AD cytopathology. At a time long before full-blown AD neuropathology and any attendant behavioral changes, these early DS brains do not show clear-cut intraneuronal Ab reactivity. Thus, it appears that extracellular Ab aggregation and accumulation can occur before or in the absence of any detectable intracellular accumulation.
4. A very similar argument to point #3 can be made about non-AD ("normal") aged brains - both in humans and lower primates, dogs and cats.
5. In the reported transgenic mouse models, abundant diffuse and compacted Ab deposits accumulate with age, and yet few or no intraneuronal Ab deposits accompany this process. Masliah and colleagues have observed intracellular amyloid fibrils by EM in occasional cells of the PDAPP transgenic mouse brain, but these appear to be far less abundant than the robust extracellular deposits these mice have. These APP transgenic mice would be the ideal setting in which to expect abundant intraneuronal Ab accumulation and aggregation, since the mice have enormous intraneuronal expression of V717F mutant APP (driven by a largely neuron-specific promoter) from birth on. Nevertheless, the neuropathological phenotype is overwhelmingly extracellular Ab deposition, as judged both by light and electron microscopy.
6. Various proteins that have been implicated as facilitators/enhancers of Ab deposition in AD are found overwhelmingly in the extraneuronal space, e.g., ApoE4, HSPG, ACT, etc. If these have a role in promoting the b-amyloidotic state, as many laboratories have suggested, they appear more likely to be exerting their activities in the extracellular rather than intraneuronal space.
7. Many studies have demonstrated that the extracellular addition of aggregated Ab to cultured neurons, astrocytes or microglia reproducibly alters these cells in ways resembling the cytological changes observed in AD brain tissue.
8. Virtually all other human amyloidotic diseases are believed to develop as a result of extracellular accumulation of the respective amyloidotic protein, with abundant accumulation of 6-10 nm amyloid fibrils very similar to those of AD in the extracellular spaces of affected tissues.
9. The unequivocal evidence that Ab peptides, particularly Ab42, are made intracellularly in early secretory compartments makes it important to search for intracellular oligomers and polymers of Ab. These are likely to exist, particularly in high Ab42-producing cells such as those harboring presenilin mutations. The question of whether these induce early internal neurotoxicity and ultimately cell death must be answered. But if this were a significant basis for neuronal injury in presenilin-caused forms of AD, one might expect a somewhat different neuropathological phenotype in PS cases than in "sporadic" AD, namely the presence of intraneuronal Ab deposits. Yet this is not the case; instead, mutant PS AD brains have been shown to have excessive extracellular Ab42 deposits compared to sporadic AD brains.
10. If intraneuronal toxicity of Ab were the principal mechanism for the effect of Ab in AD, then the intensive extracellular deposition of Ab that appears to precede detectable neuronal alteration in AD would presumably be an unnecessary and superfluous process.
Cell death in AD must surely be the combined result of many intracellular and extracellular processes. In this sense, the dichotomy posed by our initial question is artificial, although interesting to debate. Ab originates intracellularly and may even oligomerize there. But there is strong evidence that the development of abundant extracellular Ab aggregates is a very early and invariant feature of the disease cascade. I believe it is likely that extracellular Ab is necessary for the development of AD-type neuronal injury and plays the principal role in initiating microglial activation, astrocytosis and, ultimately, neurodegeneration.