Our findings suggest a highly conserved mechanism () in which Aβ monomers stimulate angiogenesis by a process that may be critical to our understanding of AD pathogenesis. Alzheimer’s risk is increased by conditions that hinder vascular flow, including atherosclerosis, diabetes, and a sedentary lifestyle 
. Those factors reduce Aβ efflux from the brain and increase blood vessel branching, which over many years could result in the formation of dense, highly branched blood vessel networks, as occur in AD brain and in Aβ over-expressing mouse models 
. Early angiogenic sprouting could temporarily restore flow rates and reduce brain Aβ, but persistently high levels of longer Aβ species would eventually result in hyper-vascularization and a reduction of perfusion efficiency. Those changes could set the stage the accumulation of much higher levels of Aβ in the parenchyma that could cause coalesce into plaques and trigger neurodegeneration; further, it could diffuse into adjacent neural tissue where is would change the vascular structure and cause the cycle to repeat and spread AD pathology. It is worth noting that AD pathology spreads continguously from entorhinal cortex, hippocampus, basal forebrain and neocortex. This mechanism may explain the temporal nature of Alzheimer’s disease with initial symptoms that often stabilize for a few years before rapid deterioration.
Proposed mechanism for hypervascularization in response to high concentrations of Aβ.
A possible mechanism for feedback inhibition of Aβ peptides on the γ-secretase complex may rest with hydrophobic residues near the carboxyl terminus. The amino-terminus of C99 is bound by Nicastrin, which mediates substrate entry into the γ-secretase complex 
. Once inside the complex, internal amino acids from both C-terminal and N-terminal fragments (CTF and NTF, respectively) of Presenilin mediate substrate cleavage. Initially, the substrate is cleaved at the cytoplasmic interface 
, releasing the intracellular fragment, and then amino acids are removed from the transmembrane region before the extracellular fragment is released by the complex. Variability of Aβ species (39–43 amino acids) depends on the strength of those interactions and may explain why AD-linked mutations in Presenilin-1 cause more Aβ1–42 to be produced 
. Hydrophobic residues from long Aβ species (Aβ1–42 and Aβ1–43) may transiently insert into the plasma membrane of cells, and possibly enter the endosome pathway (), resulting in slower tissue clearance rates than Aβ species with short hydrophobic tails (Aβ1–39
). Importantly, the amino-termini of these peptides project outward and contain the same Nicastrin-binding motif as C99. Brains that produce higher levels of Aβ could accumulate enough peptide on the outer surface, or within endosomes, to compete for Nicastrin with NEXT. This effect may also happen with the extracellular products of other γ-secretase substrates including, cadherin intermediates.
Amyloid-Notch cross-talk 
may explain the failure of γ-secretase inhibitors in human trials for Alzheimer’s disease, and could clarify the source of vascular complications that have been reported with anti-Aβ antibody therapies 
. Our findings suggest that high levels of Aβ monomer may set the stage for ensuing Alzheimer’s pathology by disrupting γ-secretase processing of NEXT and, as such, GSI compounds should exacerbate this effect. Indeed, clinical trials with the GSI Semagacestat worsened AD symptoms in some subjects, who did not improve when administration of the drug was stopped 
. Antibodies directed against Aβ have been more promising in clinical trials, though there are concerns about the frequency of MRI signal intensity shifts, indicative of micro-hemorrhages that were first reported in mouse studies with similar antibodies 
. If anti-Aβ antibodies are transported across brain endothelial cells they would quickly encounter Aβ in the perivascular-interstitial space. Antibody binding would thereby prevent Aβ from interacting with Nicastrin, resulting in a loss of γ-secretase inhibition on the blood vessels in AD affected brain areas. A consequent burst of γ-secretase activity acting on the backlog of unprocessed substrates, including NEXT and cadherin intermediates, would strongly induce NICD and TCF/LEF target genes. Such a combination of factors may disrupt VE−/E-cadherin-mediated interactions between endothelial cells, leading to transient leakiness of the blood-brain barrier, as was reported for Bapineuzumab in the early stages of treatment 
γ-secretase is a major integrator of cellular activity and the mechanism described here suggests that its extracellular cleavage products, such as Aβ, may feedback to regulate γ-secretase dependent signaling pathways. Future studies will determine how feedback inhibition from the spectrum of γ-secretase products may impact biological processes that rely on this important complex.