Recent advances in our understanding of intracellular protein trafficking have shown that subtle alterations in the intracellular distribution of APP can have strong effects on whether it is processed by the non-amyloidogenic or amyloidogenic pathways 
. In the current study we present data showing that autophagy is a degradative pathway that has the capacity to reduce cellular levels of APP and its metabolites when activated either physiologically (starvation), through pharmacological treatment (rapamycin or thapsigargin), or by lentiviral overproduction of Becn1. Conversely, reduced expression of Becn1 or pharmacological inhibition of autophagosomal degradation (bafilomycin A1, chloroquine, ammonium-chloride/leupeptin) led to an increase in APP and its metabolites. We conclude that Becn1 is a key regulator of cellular APP turnover.
Autophagy is a physiological mechanism that can have both beneficial and detrimental effects on neurons, depending on the circumstances 
. Whether or not autophagy is increased in AD and whether such an increase reflects a protective attempt by cells to possibly degrade APP and Aβ, or a neurotoxic process promoting autophagic cell death has been debated. However, recent publications indicate that pharmacological stimulation of autophagy can be beneficial and reduce Aβ mediated toxicity 
. In human brains and AD mouse models autophagosomes can be readily detected by electron microscopy and they appear to accumulate in swollen dystrophic neurites 
. This is most commonly interpreted as a sign of impaired autophagosomal degradation 
. Furthermore APP-cleaving secretases and Aβ have been localized to autophagosomes and the accumulation of autophagosomes in AD brains and APP/PS1 mice has been interpreted as evidence that autophagy could promote AD pathology 
. In agreement with these neuropathological findings, we observed that APP transgenic mice accumulate lysosomal and autophagosomal vesicles and that Becn1 deficiency in APP mice further promotes this pathology 
. In addition, we confirm here that autophagy is activated in AD by detecting increased levels of LC3-II in AD brains ().
However, at the same time, we and others found BECN1 
and in the current study PIK3C3, reduced in AD tissue ( and Table S1
), suggesting an impairment in the initiation of autophagy. To reconcile these apparently contradictory findings we postulate a dual role for BECN1: one in autophagy initiation, in a complex with PIK3C3, and another in autophagosomal flux and degradation, potentially in a complex with other proteins (). BECN1 has been shown to execute various functions depending on its binding partners and siRNA mediated knockdown of Becn1
has been demonstrated to impair autophagosomal degradation and cause LC3-II accumulation at the same time 
, similar to our findings (). Different experimental models therefore appear to reflect different aspects of this dual role. On one hand Becn1
heterozygous knockout mice have reduced autophagosomes and reduced LC3-II 
, reflecting impaired autophagosomal initiation. On the other hand, Becn1
siRNA treated CHO cells have increased LC3-II levels (). This reflects impaired autophagosomal degradation similar to the pathology observed in AD brains. In either role, reducing BECN1 leads to pathological accumulations of APP and its metabolites through impaired autophagy ().
Effects of BECN1 deficiency in AD.
Our observations regarding the effect of Becn1
siRNA on autophagy confirm that the BECN1-PI3K complex has a crucial role during the initiation stages of autophagy, but they also show that reduction of Becn1 protein levels can have effects on the availability of PIK3C3 and vice versa (). This is supported by recent findings of similar Becn1 reduction after Pik3c3
, although a reduction of Pik3c3 after Becn1
knockdown had not been reported. It will be important do determine if other proteins that are part of the BECN1 complex (Atg14L, PIK3R4, UVRAG, Ambra1, Vps15, Bif-1, or Rubicon) are also reduced in AD or in response to BECN1 reduction, respectively, as this could help explain the (possibly indirect) effects of BECN1 reduction on autophagosomal degradation (). Atg14L and UVRAG are especially interesting candidates for this since both proteins have been shown to determine the stability of Becn1 
and Atg14L knockdown causes LC3-II accumulation similar to Becn1 siRNA 
. Further studies will be needed to precisely determine the role of Becn1 and its binding partners in the modulation of autophagic flux and autophagosomal maturation. Nevertheless, with respect to APP metabolism, Becn1 seems to play central role, since Pik3c3 siRNA does not cause a comparable effect on APP accumulation in our in vitro
Aiming to validate our cell culture findings in AD brain tissue, we measured the levels of PIK3C3, LC3, and ATG5. We found a reduction not only of BECN1, but also of its binding partner PIK3C3, similar to our cell culture model using Becn1 siRNA (). Importantly, we observed a linear relationship between the levels of these two proteins () similar to the cell culture studies, supporting the idea that reduction in one of the proteins can cause instability of the PI3K complex and increased degradation or reduced production of the respective binding partner. The levels of ATG5 on the other hand were not significantly changed, arguing for a specific disruption of the PI3K complex in AD rather than a general deficiency in the autophagy pathway and signaling cascade. The reduction in PI3K complex components appear to have an inhibiting effect on the degradation rate of autophagosomes, which may lead to the build-up of LC3 protein in brain tissue and a subsequent accumulation of APP and its metabolites.
Which comes first, BECN1/PIK3C3 deficiency or APP accumulation? While the data from the transgenic mice suggested an important role of Becn1 levels on AD pathology 
, it was unclear if this effect is upstream of APP pathology or partially a consequence of disrupted intracellular trafficking due to APP overexpression. Our cell culture data from wildtype human APP overexpressing cell lines demonstrate now that APP overexpression alone does not lead to reduced Becn1 and Pik3c3 levels, leaving the possibility that autophagy disturbance could precede APP/Aβ pathology in vivo
, and that the observed reduction of BECN1 in human AD brain tissue is unlikely due to elevated levels of APP or its metabolites alone. Instead, it suggest that an escalating disturbance in autophagosomal flux and degradation could have a negative impact on BECN1 and/or PIK3C3 levels, presumably via a negative feedback loop, downregulating autophagy induction in response to abundant autophagosome numbers (). Such a loop could be in place to prevent an uncontrolled run-off activation of autophagy with potentially disastrous consequences for the cell. In support of such a model, pharmacological inhibition of autophagosomal-lysosomal fusion using BafA causes a strong accumulation of autophagosomes, accompanied with APP and APP-CTF accumulation in those autophagosomes and other intracellular vesicles. This in turn leads to decreased levels of Becn1 and, at least under some treatment conditions, of Pik3c3 (). These findings suggest that disturbances in autophagosome turnover can further inhibit proper induction and execution of autophagy, potentially worsening the cellular capacity to degrade APP and its metabolites.
The initial factor that impairs autophagy in AD and reduces BECN1/PIK3C3 still has to be determined. This study however identifies autophagy as an important degradative pathway for APP and suggests that once autophagosomal flux and turnover is impaired an escalating cycle of APP/APP-CTF/Aβ accumulation and further reduced initiation of autophagy occurs (). Future studies of conditional knockout mice for proteins that are part of the BECN1-PI3K complex will help to deepen our understanding of the sequence of events that lead to the disruption of autophagy and how this contributes to the development of AD pathology.