Genome-wide analysis reveals mechanisms
modulating autophagy in normal brain
aging and in Alzheimer’s disease
Marta M. Lipinskia, Bin Zhengb, Tao Luc, Zhenyu Yand, Bénédicte F. Pya, Aylwin Nge, Ramnik J. Xaviere, Cheng Lid,
Bruce A. Yanknerc, Clemens R. Scherzerb, and Junying Yuana,1
aDepartment of Cell Biology, Harvard Medical School, Boston, MA 02115;bCenter for Neurologic Diseases, Brigham and Women’s Hospital, Harvard Medical
School, Cambridge, MA 02139;cDepartment of Pathology, Harvard Medical School, Boston, MA 02115;dDepartment of Biostatistics and Computational
Biology, Dana-Farber Cancer Institute, Harvard School of Public Health, Boston, MA 02115; andeCenter for Computational and Integrative Biology,
Massachusetts General Hospital, Harvard Medical School, Boston, MA 02114
Communicated by Michael Eldon Greenberg, Children’s Hospital Boston, Boston, MA, June 30, 2010 (received for review May 26, 2010)
Dysregulation of autophagy, a cellular catabolic mechanism essen-
tial for degradation of misfolded proteins, has been implicated in
multiple neurodegenerative diseases. However, the mechanisms
that lead to the autophagy dysfunction are still not clear. Based on
the results of a genome-wide screen, we show that reactive oxygen
of the type III PI3 kinase, which is critical for the initiation of auto-
phagy. Furthermore, ROS play an essential function in the induction
of the type III PI3 kinase and autophagy in response to amyloid β
peptide, the main pathogenic mediator of Alzheimer’s disease (AD).
However, lysosomal blockage also caused by Aβ is independent of
ROS. In addition, we demonstrate that autophagy is transcription-
ally down-regulated during normal aging in the human brain. Strik-
ingly, in contrast to normal aging, we observe transcriptional up-
regulation of autophagy in the brains of AD patients, suggesting
that there might be a compensatory regulation of autophagy. In-
terestingly, we show that an AD drug and an AD drug candidate
have inhibitory effects on autophagy, raising the possibility that
decreasing input into the lysosomal system may help to reduce cel-
lular stress in AD. Finally, weprovide a list of candidate drug targets
that can be used to safely modulate levels of autophagy without
causing cell death.
reactive oxygen species|type III PI3 kinase|neurodegeneration|
ing turnover of cellular components, plays an important role
in regulating cellular homeostasis in the nervous system (1).
Even in the absence of any other risk factors, autophagy de-
ficiency in the CNS has been shown to lead to the accumulation
of protein aggregates and progressive neurodegeneration (2).
Thus, autophagy has been established as an important mecha-
nism mediating degradation of misfolded proteins in the CNS.
Because accumulation of misfolded proteins is a common fea-
ture in multiple human neurodegenerative diseases, activation of
autophagy has been proposed as a strategy for combating neu-
rodegeneration (3). However, little is currently known about how
defects in autophagy might be involved in specific neurodegen-
erative diseases. Furthermore, as induction of autophagy is fre-
quently associated with cell death, it remains a challenge to
identify molecular targets whose inhibition can specifically acti-
vate autophagy without compromising cell viability.
Pathological evidence supports the involvement of autophagy
dysfunction in neurodegenerative diseases in humans. In Alz-
heimer’s disease (AD), one of the earliest pathological changes
include accumulation of autophagic vesicles (AVs) specifically
within damaged neuritic processes and synaptic terminals (4).
This phenotype is also observed in AD animal models and in cell-
utophagy, a lysosome-dependent catabolic process mediat-
the mechanisms leading to the accumulation of AVs and the
causal relationship to neurodegeneration are not yet established.
We have recently conducted a genome-wide screen using
siRNA library to identify genes regulating autophagy in human
cells under normal nutritional conditions (5). In this image-based
screen we took advantage of the autophagy specific GFP-LC3
reporter whose translocation from the cytosol to autophago-
somes can serve as a quantitative measure of autophagy. In this
study, we specifically explore the mechanisms that regulate
autophagy in neural cells using the hits identified in our screen.
We demonstrate that reactive oxygen species (ROS) play a gen-
eral function in mediation of autophagy upstream of the type III
PI3 kinase and that this pathway is essential for the up-regulation
of autophagy by Aβ. Interestingly, our data show that genes
regulating autophagy are differentially expressed in normal aging
and in AD patient brains. Finally, we identify candidate molec-
ular targets that may be safely manipulated to modulate auto-
phagy to treat neurodegenerative diseases.
ROS Functions Upstream of the Type III PI3 Kinase to Induce Auto-
phagy. Genes whose knock-down induced autophagy in our
screen included components of the ROS detoxification pathway
GPx2 and SOD1, whose mutations are known to cause familial
ALS. The screen hits also included several mitochondrial pro-
teins, some of which are involved in oxidative respiration and
electron transport (Table S1) as well as genes reported to be
involved in ROS homeostasis based on a literature co-citation
analysis (Table S2). This suggests a role for ROS in the induction
of autophagy and a possible function for this pathway in neur-
To determine whether ROS are sufficient to induce autophagy,
we confirmed that transfection of SOD1 siRNA led to both an
induction of autophagy in H4 human neuroblastoma cells, as well
role of ROS, treatment with an antioxidant N-acetyl-L-cysteine
(NAC) significantly attenuated the induction of autophagy (Fig.
S1D). Therefore, interference with normal cellular ROS homeo-
stasis by inactivation of a gene involved in neurodegeneration is
sufficient to induce autophagy in the absence of any other
To determine whether ROS may have a general signaling
function in autophagy, we compared the levels of autophagy in-
duced by knock-down of our screen hit genes in the presence and
Author contributions: M.M.L. and J.Y. designed research; M.M.L., T.L., and B.F.P. per-
formed research; M.M.L., B.Z., T.L., Z.Y., A.N., R.J.X., C.L., B.A.Y., and C.R.S. analyzed data;
and M.M.L. and J.Y. wrote the paper.
The authors declare no conflict of interest.
1To whom correspondence should be addressed. E-mail: firstname.lastname@example.org.
This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.
| August 10, 2010
| vol. 107
| no. 32www.pnas.org/cgi/doi/10.1073/pnas.1009485107
absence of NAC. We uncovered a large group of genes (117, or
54% of all genes tested, Fig. S2A and Table S3) that when
knocked-down led to translocation of GFP-LC3 to autophago-
ROS were required for the induction of autophagy. Interestingly,
the presence of NAC also reduced the vesicular accumulation of
FYVE-dsRed, a reporter protein for PtdIns3P, induced by the
type III PI3 kinase, this result suggests that ROS may serve an
important general signaling function in the early steps of auto-
phagic pathway mediated by the type III PI3 kinase.
Knock-down any of the remaining 98 (46%, Table S4) genes
was able to induce accumulation of GFP-LC3 in the presence of
NAC, suggesting that, in these cases, autophagy can be induced
independently of ROS. Knock-down of these genes was also able
to induce comparable average levels of vesicular FYVE-dsRed in
the presence and absence of NAC (Fig. 1A). Thus, knock-down
of this group of genes led to induction of the type III PI3 kinase
through a mechanism independent of ROS.
Because our data indicated that the type III PI3 kinase plays
an especially crucial role in the mediation and regulation of
autophagy downstream of ROS, we performed an additional
screen to determine how our hit genes are influenced by the type
III PI3 kinase activity. In addition to its prominent function in
regulation of apoptotic cell death, Bcl-2 may negatively regulate
autophagy through its interaction with beclin 1 and consequent
inhibition of the type III PI3 kinase activity (6). Consistently, we
observed a significant increase in cell viability and a decrease in
levels of PtdIns3P following knock-down of the hit genes in H4
cells expressing Bcl-2 as compared with WT controls (Fig. 1 B
and C and Fig. S2B). Knock-down of 124 (58%) of the 215 tested
hit genes was unable to induce accumulation of vesicular GFP-
LC3 in cells overexpressing Bcl-2, confirming that, in the ma-
jority of cases, up-regulation of the type III PI3 kinase activity
is necessary for the induction of autophagy (Fig. S2C and Table
S5). On the other hand, knock-down of the remaining 91 (42%)
genes was able to induce translocation of GFP-LC3 to auto-
phagosomes in the presence of Bcl-2 (Table S6). For 17 (19%) of
these genes, induction of autophagy was accompanied by an in-
crease in type III PI3 kinase activity, suggesting additional mech-
anisms that regulate production of PtdIns3P downstream of
Bcl-2 (Fig. 1D). However, knock-down of the remaining 74 genes
was able to induce autophagy without additional activation of the
type III PI3 kinase. Knock-down of 31 (34%) of these genes led
to the expansion of the lysosomal compartment as assessed by
the accumulation of Lamp-1-RFP marker, indicating that, in
these cases, a block in lysosomal degradation may contribute to
the apparent increase in autophagy (Fig. 1D). No changes in the
lysosomal function were observed for the remaining 43 (47%)
genes. These genes may function either downstream or in-
dependently of the type III PI3 kinase, suggesting that the ap-
parent inhibitory effect of Bcl-2 on the type III PI3 kinase is not
always incompatible with the induction of autophagy, which, in
these cases, can occur without an increase in PtdIns3P levels.
Interestingly, we observed a substantial overlap between genes
whose knock-down was unable to induce autophagy in the
presence of NAC or Bcl-2 (86, 40% of all genes tested; Fig. 1E).
The overlapping genes include the majority (11 of 18) of the
mitochondrial genes, confirming that mitochondria-generated
ROS induce autophagy by positively regulating the type III PI3
kinase. Because overexpression of Bcl-2 has been shown to re-
duce stress induced generation of ROS (7), our data imply that,
in addition to directly binding and suppressing the activity of
Beclin1 (6), Bcl-2 may be able to contribute to the regulation of
the type III PI3 kinase and autophagy through its ability to at-
tenuate generation of ROS.
Differential Expression of Autophagy Regulators in AD Brain Samples.
Accumulation of both ROS and AVs are early features in AD
(4). To determine whether we could detect changes in the ex-
pression of genes involved in regulation of autophagy in this
disease, we analyzed expression of the autophagy screen hit
genes in six brain regions of 34 patients with AD and 14 age-
matched normal controls (8). We observed an overall significant
underexpression of the hit genes in AD patient samples com-
pared with controls specifically in the hippocampus and ento-
rhinal cortex, the brain regions most affected by the disease (Fig.
2A). Consistent trends were observed in other brain regions af-
fected by AD (superior frontal gyrus, posterior cingulate, and
medial temporal gyrus), but did not reach statistical significance.
Notably, in the visual cortex, a brain region relatively resistant to
AD pathology, these changes were absent. Further subdivision of
the hit genes revealed that in the entorhinal cortex negative
regulators of autophagy flux were specifically negatively enriched
(Fig. 2B). A similar trend was also observed in other brain areas
affected by AD. Conversely, positive regulators of autophagy
were positively enriched in the entorhinal cortex (Fig. 2C). Such
differential expression patterns of autophagy regulators suggest
transcriptional up-regulation of autophagy in AD brains.
KD of Mitochondrial Complex IV Gene Cox5a Leads to ROS-Induced
Autophagy. The screen hits included Cox5a, a nuclear-encoded
component of the mitochondrial electron transport complex IV
(Table S1). Inactivation of this complex has been suggested to
occur in AD and to contribute to accumulation of ROS (9). We
determined that inhibition of complex IV by knock-down of
Cox5a led to the accumulation of ROS (Fig. 3A and Fig. S3 A
and B) and induction of autophagy (Fig. 3B and Fig. S3C).
Consistent with the requirement for ROS, induction of autoph-
levels of autophagy and the type III PI3 kinase activity. (A) Quantification of
average type III PI3 kinase activity following knock-down of genes able (yes)
and unable (no) to induce autophagy in the presence of NAC. H4 FYVE-dsRed
cells were transfected with hit siRNA for 72 h, followed by fixation and im-
aging on a high-throughput fluorescent microscope at 10× magnification.
Average z-scores as compared with nontargeting siRNA are shown. (B) Com-
parison of the relative average viability of WT or pBabe-Bcl-2–expressing H4
cells transfected with hit gene siRNAs for 72 h. (C) Comparison of average
type III PI3 kinase activity in WT or pBabe-Bcl-2–expressing H4 FYVE-dsRed
cells following hit siRNA transfection for 72 h. (D) Subdivision of hits whose
knock-down was able to induce autophagy under conditions of low PtdIns3P
into functional categories based on their ability to up-regulate type III PI3
kinase activity or to alter lysosomal processing. (E) Subdivision of genes
whose knock-down led to the induction of autophagy into functional cate-
gories based on their dependence on ROS and elevated levels of PtdIns3P
(PI3P). *P < 0.05, **P < 0.01 based on two-tailed t test with equal variance. All
error bars indicate SEM.
Suppression of ROS and expression of Bcl-2 lead to decrease in the
Lipinski et al.PNAS
| August 10, 2010
| vol. 107
| no. 32
agy following knock-down of Cox5a was attenuated in the pres-
ence of NAC (Fig. 3C).
Our screen data indicated that ROS is regulating autophagy by
activating the type III PI3 kinase. In agreement, knock-down of
Cox5a led to the increase in the levels of PtdIns3P (Fig. 3D).
Levels of PtdIns3P were suppressed in the presence of the type
III PI3 kinase inhibitor, 3MA, confirming the involvement of this
kinase. In addition, 3MA, as well as the knock-down of the
catalytic subunit of the kinase, Vps34, attenuated the induction
of autophagy in response to the loss of Cox5a (Fig. 3 E and F and
Fig. S3D). Together, these data indicate that knock-down of
Cox5a leads to the induction of autophagy by increasing ROS
dependent type III PI3 kinase activity.
ROS Mediate Autophagy in Response to Amyloid β. Aβ, the main
pathogenic factor in AD, has been proposed to cause mito-
chondrial damage leading to the generation of ROS (9). We
hypothesized that induction of autophagy by Aβ may also be
mediated by ROS. We observed increased levels of autophagy
following treatment of H4 cells with Aβ (Fig. 4A and Fig. S4A).
To determine whether this was due to an increase in the initia-
tion of autophagy or to a block in lysosomal degradation, we
compared the accumulation of LC3-II following Aβ treatment in
the absence and presence of lysosomal protease inhibitor, E64d
(Fig. 4A). Up to 8 h after treatment the accumulation of LC3-II
could be observed only in the presence of E64d. At 48 h after the
addition of Aβ, the increased levels of LC3-II were observed
even without E64d but were further increased in the presence of
E64d. In addition, we observed increased conjugation of Atg12-
Atg5 starting 4 h after Aβ treatment. Together these data suggest
increased initiation of autophagy in response to Aβ. However, as
previously reported (4), lysosomal proteolysis also appeared to
be inhibited, as demonstrated by the expansion of the lysosomal
compartment (Fig. 4B) and altered mobility of lysosomal pro-
teins, cathepsin D (CtsD) and Lamp-2 on a Western blot, starting
at 24 h after Aβ treatment (Fig. S4B, Left). Therefore, Aβ regu-
lates the apparent levels of cellular autophagy by both enhancing
initiation ofautophagyanddecreasing the rateofautophagosome
clearance due to lysosomal inhibition.
To determine whether ROS are involved in the induction of
autophagy by Aβ, we first checked that Aβ induced ROS (Figs.
4C and S4C). Confirming the essential function of ROS, NAC
attenuated induction of autophagy in response to Aβ (Fig. 4D).
in lysosomal proteins (Fig. S4B, Right), suggesting that lysosomal
damage may be mediated by a different mechanism.
We then investigated the involvement of the type III PI3 ki-
nase in the induction of autophagy by Aβ. We observed increased
accumulation of PtdIns3P (Fig. 4E), which was suppressed in the
presence of 3MA (Fig. 4E), confirming the involvement of the
type III PI3 kinase. In agreement with a causal role of ROS,
accumulation of PtdIns3P was suppressed in the presence of
NAC (Fig. 4F). Finally, treatment with 3MA (Fig. 4G) or knock-
down of Vps34 (Fig. S4D) was able to attenuate induction of
autophagy in response to Aβ. On the other hand, an Aβ induced
change in Lamp-2 mobility was not affected (Fig. S4D). There-
fore, Aβ influences autophagy by two distinct mechanisms: by
increasing initiation, and by blocking lysosomal degradation. The
initiation of autophagy is an earlier event and is dependent on
the accumulation of ROS and up-regulation of the type III PI3
kinase activity. The blockade of lysosomal degradation is medi-
ated by a ROS and type III PI3 kinase independent mechanism.
autophagy in Alzheimer’s disease. Forrest plots of NES estimates with SD for
the screen hit gene sets are shown. (A) GSEA analysis of overall screen hit
gene expression in different regions of AD brain as compared with unaf-
fected age-matched controls. (B and C) GSEA analysis of hit genes determined
to function as negative (B) or positive (C) regulators of autophagy flux. Blue
squares indicate enrichment signals with P ≤ 0.05 in an individual comparison.
The size of a square is inversely proportional to the respective SD.
Differential gene expression leads to transcriptional up-regulation of
of ROS and induction of autophagy. (A) Induction of ROS in H4 cells trans-
fected with two independent siRNAs against Cox5a or nontargeting siRNA
for 72 h. Cells were stained in 25 mM carboxy-H2DCFDA and Hoechst and
imaged at 40×. (B) Induction of autophagy in cells transfected with siRNAs
against Cox5a or controls for 72 h; autophagy levels were assessed with
antibodies against p62, Atg5 (Atg12-Atg5 complex is shown), and LC3.
(C) Induction of autophagy by Cox5a knock-down is suppressed in the
presence of the antioxidant NAC. Cells were prepared as in B, except that,
where indicated, 10 mM NAC was added to the culture media 24 h after
transfection. (D) Knock-down of Cox5a leads to the accumulation of
PtdIns3P in a manner dependent on the function of the type III PI3 kinase.
Quantification of FYVE-dsRed reporter is shown. Cells were transfected as in
B; where indicated, the type III PI3 kinase inhibitor 3MA (10 mM) was added
for 8 h before fixation and imaging. (E) Induction of autophagy following
Cox5a knock-down depends on the function of the type III PI3 kinase. GFP-
LC3 H4 cells were prepared and imaged as in D. (F) Levels of autophagy
induced following knock-down of Cox5a were assessed in H4 cells trans-
fected with control siRNA (nt, nontargeting; Tor, mTOR) or siRNA against
Vps34 for 72 h by Western blot. *P < 0.05, **P < 0.01 based on two-tailed
t test with equal variance. All error bars indicate SEM.
Knock-down of the mitochondrial gene Cox5a leads to generation
| www.pnas.org/cgi/doi/10.1073/pnas.1009485107Lipinski et al.
Modulation of Autophagy as Treatment Against Neurodegeneration.
To explore the feasibility of modulating autophagy as treatment
for neurodegenerative diseases, we compared our screen hits
with the known neurodegenerative disease drug targets. This
revealed that several of the genes identified in the screen are
targets of drugs currently in use or in clinical trials against
neurological and neurodegenerative diseases, including AD
(Table S7). Consistent with the predicted negative effect of the
growth hormone secretagogue receptor activity on the levels of
autophagy, treatment of H4 cells with its ligand, Ghrelin, a can-
didate drug for AD, led to a dose-dependent suppression of
autophagy (Fig. 5A). Similarly, galanthamine hydrochloride, an
agonist of the nicotinic acetylcholine receptor (CHRND) and an
AD drug, led to the suppression of autophagy levels (Fig. 5B).
Neither drug significantly affected cell viability (Fig. S5A and B).
Therefore, these drugs may be able to partially suppress up-
regulation of autophagy.
Candidate Autophagy Drug Targets Against Neurodegeneration. To
modulate autophagy as a treatment against neurodegenerative
diseases, we need to identify novel molecular drug targets that
can up-regulate autophagy flux without affecting cell viability. As
ROS can be the cause of cellular damage, these genes should up-
regulate autophagy in a ROS-independent manner. To this end,
we analyzed our screen data and identified 26 such candidate
genes (Table 1 and Table S8). We propose these genes as can-
didate inhibitory drug targets against neurodegenerative diseases
where up-regulation of autophagy is beneficial.
In addition, our screen included nine genes whose inhibition
down-regulated autophagy without affecting cell viability and
μM Aβ. (B) Expansion of the lysosomal compartment following Aβ treatment
based on the Lamp-1-RFP lysosomal reporter. Cells were treated as in A, then
Induction of autophagy by Aβ is suppressed in the presence of the antioxidant
NAC. GFP-LC3 cells were prepared as in B except that, where indicated, 10 mM
prepared and imaged as in B; where indicated, the type III PI3 kinase inhibitor
3MA (10mM) was added for 8 h before fixation. (F) Induction of the type III PI3
kinase activity by Aβ is suppressed in the presence of antioxidant. Cells were
prepared as in D. (G) Induction of autophagy by Aβ is dependent on the type III
PI3 kinase activity. H4 GFP-LC3 cells were treated and imaged asinE. **P < 0.01
based on two-tailed t test with equal variance. All error bars represent SEM.
Amyloid β up-regulates autophagy by inducing accumulation or ROS
ted with indicated concentrations of GSHR ligand Ghrelin (A) or CHRND
agonist Galanthamine (B) for 24 h and imaged at 10×. *P < 0.05, ** P < 0.01
based on two-tailed t test with equal variance. All error bars represent SEM.
Drugs against AD suppress autophagy. H4 GFP-LC3 cells were trea-
autophagy in neurodegenerative disease
Candidate inhibitory drug targets for modulation of
Gene symbolGene ID Gene name
Drug targets for up-regulation of autophagy
PPFIA4 8497Protein tyrosine phosphatase, receptor type,
f polypeptide (PTPRF), interacting protein
(liprin), alpha 4
Prostaglandin E receptor 2 (subtype EP2), 53kDa
E1A binding protein p300
Chemokine (C-X-C motif) ligand 12 (stromal
cell-derived factor 1)
Natural cytotoxicity triggering receptor 3
Gamma-aminobutyric acid (GABA) B receptor, 2
Gap junction protein, alpha 4, 37kDa
Succinate dehydrogenase complex, subunit B,
iron sulfur (Ip)
GTF2I repeat domain containing 2
Gonadotropin-releasing hormone 2
HIV type I enhancer binding protein 2
Ubiquitin specific peptidase 19
EPH receptor A6
Tachykinin receptor 2
Myosin, light chain 3, alkali; ventricular,
Calpain 1, (mu/I) large subunit
Sortilin-related VPS10 domain containing
BAI1-associated protein 2
Polycomb group ring finger 1
Matrix metallopeptidase 10 (stromelysin 2)
Cholinergic receptor, nicotinic, delta
Cyclin-dependent kinase inhibitor 2D (p19,
F-box and leucine-rich repeat protein 20
Drug targets for down-regulation of autophagy
ATG7 autophagy related 7 homolog
Multiple EGF-like-domains 10
Kinesin family member 5C
SET and MYND domain containing 3
Suppressor of defective silencing 3 pseudogene
Stromal interaction molecule 1
GRB2-associated binding protein 1
G-protein–coupled receptor 18
Cation channel, sperm associated 4
Lipinski et al.PNAS
| August 10, 2010
| vol. 107
| no. 32
which are not known to affect ROS homeostasis (Table 1 and
Table S8). These genes represent potential drug candidates
against diseases such as late-stage AD, in which reducing auto-
phagic input might be advantageous.
Transcriptional Regulation of Autophagy in Normal Brain Aging. To
determine whether the regulation of autophagy may have wider
implications in normal aging of the human brain, we analyzed
expression of the autophagy screen hit genes in a set of younger
versus older human brain samples (10). We observed differential
significantly (P < 0.05) up-regulated and 46 down-regulated with
age (Fig. 6A and Fig. S6 A and B and Table S9). Gene ontology
biological process analysis revealed that the age up-regulated
group was highly enriched in genes involved in mediation and
regulation of the MAP kinase pathway (P = 1.6 × 10−4). An in-
crease in the activity of MAP kinase pathway was predicted by our
previous analysis to lead to the suppression of autophagy (5).
Conversely, expression of the key autophagy genes, such as Atg5
andAtg7, was down-regulated in aging. This is consistent with our
previous data demonstrating transcriptional down-regulation of
beclin 1, in normal human brain aging (11). Together, this sug-
gests, that unlike AD, the normal aging process may lead to
transcriptional down-regulation of autophagy.
To further define the biological processes affected by down-
regulation of autophagy in aging, we used gene ontology ca-
nonical pathway analysis. It revealed a significant enrichment in
the “Axon guidance” (P = 0.0009) and “Regulation of actin
cytoskeleton” (P = 0.038) pathways, suggesting a connection
between regulation of autophagy, axon guidance and actin dy-
namics. Construction of protein–protein interaction networks
anchored by the hit genes belonging to these pathways (12, 13)
revealed two related networks encompassing, respectively, 27
(11%) and 61 (26%) of the hit genes (Fig. S6 C and D). Im-
portantly, both networks directly connect to the known autoph-
agy machinery through the interaction of the RIP kinase (RIPK1)
and PKCζ (PRKCZ) with p62/sequestrosome (SQSTM1). In
addition, syndecan 2 (SDC2), a part of the “Regulation of actin
cytoskeleton” network, interacts with syntenin, a binding partner
of ULK1, the human ortholog of yeast Atg1 (14). ULK1 is known
to play a role in the regulation of endocytic processes involved
in axon guidance (15) and to promote synapse formation in
Drosophila (16). These data suggest that some of the molecular
networks involved in the regulation of autophagy are closely
connected to those regulating endocytosis, actin dynamics, and
neuronal axon guidance, and that autophagy may play a wider
role in the development and maintenance of neuronal function.
In this study, we demonstrate that the type III PI3 kinase plays
a fundamental role in the regulation of autophagy and that ROS
function as general mediators of autophagy induction upstream
of this kinase. This pathway has an essential function in the
initiation of autophagy in response to mitochondrial damage
following exposure to Aβ, the main pathogen of AD. At the same
time, Aβ is able to slow down autophagic processing through
ROS independent inhibition of lysosomal degradation. In addi-
tion, our analysis of expression of the autophagy screen hits
suggests that autophagy is differentially regulated at the tran-
scriptional level in normal human aging and in AD, with overall
levels decreased in normal aging but elevated in AD.
ROS Are General Mediators of Autophagy Upstream of Type III PI3
Kinase. Although ROS have been implicated in the regulation of
starvation-induced autophagy (17), their general significance in
mediation of autophagy has not been appreciated. Furthermore,
the biochemical step at which ROS intersect with the autophagic
pathway has remained unknown. Our data demonstrate that up-
regulation of autophagy and the type III PI3 kinase activity by the
knock-down of a large fraction of our autophagy screen hit genes
can be inhibited in the presence of the antioxidant NAC. Thus,
ROS might serve a previously unappreciated role as a common
autophagy upstream of the type III PI3 kinase. Although ROS
have been implicated in regulation of different cellular events,
such as cell death, such extensive involvement in a specific sig-
naling pathway has not been appreciated before. Because an in-
crease in the levels of ROS is a frequent consequence of the
intracellular signal for the homeostatic activation of autophagy
under basal physiological conditions as well as in neurodegener-
ative diseases including AD.
Both excessive generation of ROS and accumulation of AVs
are some of the earliest hallmarks of AD (4, 18, 19). Until now,
however, these phenotypes were not considered to be related.
Although further confirmation in primary neurons will be nec-
essary, our data suggest a mechanistic link between the genera-
tion of ROS and the early accumulation of AVs observed in AD.
In addition, they point to the previously unappreciated in-
volvement of the type III PI3 kinase and the accumulation of
PtdIns3P in the etiology of AD.
Differential Expression of Autophagy Regulators in Normal Aging and
in AD. Our gene expression data suggest that autophagy is also
differentially regulated at the transcriptional level in normal
human brain aging versus in AD. Because autophagy is known to
play a protective role against onset of neurodegeneration in
animal models (2, 3, 20, 21), its down-regulation in normal aging
could contribute to the observed age-dependent predisposition
to development of chronic neurodegenerative diseases. In addi-
tion, the extensive overlap of the autophagy screen hits with
in younger (≤40 y old) versus older (≥70 y old) human brain samples, based on (i) minimum 1.2-fold change between the average expression, and (ii) P value
<0.05 using unpaired t test.
Expression of autophagy screen hit genes in normal human aging. Clustering analysis (dChip) of mRNA expression levels of select autophagy hit genes
| www.pnas.org/cgi/doi/10.1073/pnas.1009485107Lipinski et al.
regulatory networks mediating actin dynamics and endocytic
processes involved in axonal guidance suggests that autophagy
may directly contribute to the regulation of growth and mainte-
nance of neuronal processes. Therefore, progressive transcrip-
tional down-regulation of autophagy in normal human brain aging
may also contribute to age related decline in synaptic plasticity,
memory, and cognitive function.
Conversely, our data indicate that autophagy is specifically up-
regulated in AD, due both to ROS-dependent activation of the
type III PI3 kinase as well as at the transcriptional level. We
hypothesize that this may represent, respectively, an acute and
a long-term attempt by the affected neuronal cells to rid them-
selves of the harmful effects of Aβ exposure, such as accumu-
lation of defective mitochondria and protein aggregates.
Surprisingly, our study also demonstrated that drugs in clinical
trials or used against AD lead to the inhibition of autophagy.
Because increasing autophagy is predicted to be beneficial for the
treatment of neurodegenerative diseases, one may wonder if this
inhibition may be an unintended side effect, as autophagy was not
a parameter tested when developing these drugs. Alternatively,
inhibition of autophagy could contribute to the efficacy of those
drugs. Our data indicate that treatment with Aβ leads to both an
increase in the initiation of autophagy as well as a decrease in the
lysosomal degradation. We hypothesize that as the downstream
degradative pathway is partially blocked, increasing the autopha-
gic input may further exacerbate the stress on the system. In this
case, reducing the initiation of autophagy might be beneficial by
decreasing input into the lysosomal system, thus allowing for
up-regulation of autophagy might be beneficial as a preventive
measure during normal aging, as well as to combat some neuro-
degenerative diseases including early stages of AD (21, 22), in-
creasing autophagic input may backfire if applied to later stages of
AD, in which the lysosomal blockage is preeminent.
Dissociation of Autophagy from Cell Death for Clinical Benefit. An-
other important consideration in developing autophagy-based
therapies is the ability to change levels of autophagy without
inducing neuronal cell death. As a first important step toward
this goal, we provide a list of potential molecular drug targets for
the modulation of autophagy in neurodegenerative diseases.
Interestingly, this list includes calpain 1 which, as we have pre-
viously shown, can be inhibited by fluspirilene, an inhibitor of
intracellular Ca2+flux, to induce autophagy by reducing cleavage
of full-length Atg5 and increasing conjugated Atg12-Atg5 with-
out inducing cell death (23). We propose that by developing
inhibitors against these gene targets, autophagy can be safely
manipulated without harming neuronal cells, thus providing
potential novel therapies against neurodegenerative diseases.
Materials and Methods
Cell Lines and Culture Conditions. H4 human neuroblastoma cells were cul-
tured and treated as described elsewhere (5). GFP-LC3, FYVE-dsRed, GFP-LC3
pSRP-Beclin1 knock-down (11) and Lamp-1-RFP (5) H4 cells have been
For antioxidant assay, cells were treated with 2.5 mM NAC (Sigma). Aβ was
prepared using a modified method from ref. 24.
Imaging and Image Quantification. Cells were imaged as previously described
(5) on a CellWoRx microscope (Applied Precision) at ×10 magnification.
Images were quantified using VHSscan and VHSview software (Cellomics).
Quantification of Cellular ROS Levels. ROSwerequantifiedusingImage-iTLIVE
facturers instructions. Images were acquired on a Nikon Eclipse E800 micro-
scope at ×40 magnification and quantified using CellProfiler software (25).
Bioinformatics Analysis. Gene set enrichment analysis. Associations of autoph-
agy gene sets with AD were evaluated using GSEA (26) and gene expression
data from laser-capture microdissected non-tangle-bearing neurons of 34
late-onset AD-afflicted individuals with a mean age at death of 79.9 ± 6.9 y
and 14 neurologically normal healthy elderly controls (27) (GEO accession
Analysis of hit gene expression in aging. Analysis was based on Affymetrix HG-
U133_Plus_2 microarray data of young (≤40 y old) and old (≥70 y old) human
brain samples (10). Array normalization, expression value calculation and clus-
teringanalysis were performedusingthedChipsoftware(28)(www.dchip.org).
ACKNOWLEDGMENTS. We thank the members of the Yuan laboratory and
Institute of Chemistry and Cell Biology (ICCB)-Longwood for help during this
work. This work was supported in part by National Institutes of Health Grants
R37 AG012859 (to J.Y.), PO1 AG027916 (to B.A.Y. and J.Y.), AI062773 and
DK043351 (to R.J.X.), as well as R01 NS064155, R21 NS060227, P01 NS058793,
and an RJG Foundation grant (to C.R.S.). A.N. is supported by the Crohn’s
and Colitis Foundation of America.
1. Levine B, Klionsky DJ (2004) Development by self-digestion: Molecular mechanisms
and biological functions of autophagy. Dev Cell 6:463–477.
2. Hara T, et al. (2006) Suppression of basal autophagy in neural cells causes
neurodegenerative disease in mice. Nature 441:885–889.
3. Ravikumar B, et al. (2004) Inhibition of mTOR induces autophagy and reduces toxicity
of polyglutamine expansions in fly and mouse models of Huntington disease. Nat
4. Nixon RA, et al. (2005) Extensive involvement of autophagy in Alzheimer disease: An
immuno-electron microscopy study. J Neuropathol Exp Neurol 64:113–122.
5. Lipinski MM, et al. (2010) Multiple mTORC1 independent signaling pathways regulate
autophagy through type III PI3 kinase under normal nutritional conditions. Dev Cell
6. Pattingre S, et al. (2005) Bcl-2 antiapoptotic proteins inhibit Beclin 1-dependent
autophagy. Cell 122:927–939.
7. Kane DJ, et al. (1993) Bcl-2 inhibition of neural death: Decreased generation of
reactive oxygen species. Science 262:1274–1277.
8. Liang WS, et al. (2008) Alzheimer’s disease is associated with reduced expression of
energy metabolism genes in posterior cingulate neurons. Proc Natl Acad Sci USA 105:
9. Takuma K, et al. (2005) ABAD enhances Abeta-induced cell stress via mitochondrial
dysfunction. FASEB J 19:597–598.
10. Loerch PM, et al. (2008) Evolution of the aging brain transcriptome and synaptic
regulation. PLoS ONE 3:e3329.
11. Shibata M, et al. (2006) Regulation of intracellular accumulation of mutant
Huntingtin by Beclin 1. J Biol Chem 281:14474–14485.
12. Ho Y, et al. (2002) Systematic identification of protein complexes in Saccharomyces
cerevisiae by mass spectrometry. Nature 415:180–183.
13. Ito T, et al. (2001) A comprehensive two-hybrid analysis to explore the yeast protein
interactome. Proc Natl Acad Sci USA 98:4569–4574.
14. Hara T, et al. (2008) FIP200, a ULK-interacting protein, is required for autophagosome
formation in mammalian cells. J Cell Biol 181:497–510.
15. Zhou X, et al. (2007) Unc-51-like kinase 1/2-mediated endocytic processes regulate
filopodia extension and branching of sensory axons. Proc Natl Acad Sci USA 104:
16. Shen W, Ganetzky B (2009) Autophagy promotes synapse development in Drosophila.
J Cell Biol 187:71–79.
17. Scherz-Shouval R, et al. (2007) Reactive oxygen species are essential for autophagy
and specifically regulate the activity of Atg4. EMBO J 26:1749–1760.
18. Lustbader JW, et al. (2004) ABAD directly links Abeta to mitochondrial toxicity in
Alzheimer’s disease. Science 304:448–452.
19. Reddy PH, Beal MF (2008) Amyloid beta, mitochondrial dysfunction and synaptic
damage: Implications for cognitive decline in aging and Alzheimer’s disease. Trends
Mol Med 14:45–53.
20. Komatsu M, et al. (2006) Loss of autophagy in the central nervous system causes
neurodegeneration in mice. Nature 441:880–884.
21. Pickford F, et al. (2008) The autophagy-related protein beclin 1 shows reduced
expression in early Alzheimer disease and regulates amyloid beta accumulation in
mice. J Clin Invest 118:2190–2199.
22. Hung SY, Huang WP, Liou HC, Fu WM (2009) Autophagy protects neuron from Abeta-
induced cytotoxicity. Autophagy 5:502–510.
23. Xia HG, et al. (2010) Control of basal autophagy by calpain1 mediated cleavage of
ATG5. Autophagy 6:61–66.
24. Walsh DM, et al. (1999) Amyloid beta-protein fibrillogenesis. Structure and biological
activity of protofibrillar intermediates. J Biol Chem 274:25945–25952.
25. Carpenter AE, et al. (2006) CellProfiler: Image analysis software for identifying and
quantifying cell phenotypes. Genome Biol 7:R100.
26. Subramanian A, et al. (2005) Gene set enrichment analysis: A knowledge-based
approach for interpreting genome-wide expression profiles. Proc Natl Acad Sci USA
27. Liang WS, et al. (2008) Altered neuronal gene expression in brain regions
differentially affected by Alzheimer’s disease: A reference data set. Physiol Genomics
28. Li C, Wong WH (2001) Model-based analysis of oligonucleotide arrays: Expression
index computation and outlier detection. Proc Natl Acad Sci USA 98:31–36.
Lipinski et al.PNAS
| August 10, 2010
| vol. 107
| no. 32