PtdIns(4,5)P2 signaling regulates ATG14 and autophagyXiaojun Tana,1, Narendra Thapaa,1, Yihan Liaoa, Suyong Choia, and Richard A. Andersona,2

aProgram in Molecular and Cellular Pharmacology, School of Medicine and Public Health, University of Wisconsin, Madison, WI 53706

Edited by Jennifer Lippincott-Schwartz, National Institute of General Medical Sciences, Bethesda, MD, and approved August 5, 2016 (received for reviewNovember 23, 2015)

Autophagy is a regulated self-digestion pathway with fundamentalroles in cell homeostasis and diseases. Autophagy is regulated bycoordinated actions of a series of autophagy-related (ATG) proteins.The Barkor/ATG14(L)–VPS34 (a class III phosphatidylinositol 3-kinase)complex and its product phosphatidylinositol 3-phosphate [PtdIns(3)P]play key roles in autophagy initiation. ATG14 contains a C-terminalBarkor/ATG14(L) autophagosome-targeting sequence (BATS) domainthat senses the curvature of PtdIns(3)P-containing membrane. TheBATS domain also strongly binds PtdIns(4,5)P2, but the functionalsignificance has been unclear. Here we show that ATG14 specificallyinteracts with type Iγ PIP kinase isoform 5 (PIPKIγi5), an enzyme thatgenerates PtdIns(4,5)P2 in mammalian cells. Autophagosomes haveassociated PIPKIγi5 and PtdIns(4,5)P2 that are colocalized with lateendosomes and the endoplasmic reticulum. PtdIns(4,5)P2 generationat these sites requires PIPKIγi5. Loss of PIPKIγi5 results in a loss ofATG14, UV irradiation resistance-associated gene, and Beclin 1 and ablock of autophagy. PtdIns(4,5)P2 binding to the ATG14–BATS do-main regulates ATG14 interaction with VPS34 and Beclin 1, and thusplays a key role in ATG14 complex assembly and autophagy initia-tion. This study identifies an unexpected role for PtdIns(4,5)P2 sig-naling in the regulation of ATG14 complex and autophagy.

autophagy | PIPKIγi5 | PtdIns(4,5)P2 | ATG14 | ATG14–BATS domain

Macroautophagy (referred to as autophagy from here on) isan intracellular self-digestion process conserved in all eu-

karyotes that maintains cell homeostasis and protects cells fromvarious stresses. Autophagy involves the formation of double-membrane autophagosomes that contain target cytoplasmic con-tents for lysosomal degradation (1). Autophagosome formation isdriven by coordinated functions of a number of autophagy-related(ATG) proteins (2). The ATG14 (also called Barkor or ATG14L)-Beclin 1 (atg6 in yeast)-VPS34 (an enzyme involved in the cellularprocess of autophagy; class III PI3K) complex plays essential rolesin the initiation steps of autophagy by generating phosphatidyli-nositol 3-phosphate [PtdIns(3)P] and recruiting PtdIns(3)P effec-tors (3–6). ATG14 functions by recruiting VPS34 to punctatestructures associated with the endoplasmic reticulum (ER) (7),and then PtdIns(3)P, the product of VPS34, in turn binds theBarkor/Atg14 autophagosome targeting sequence (BATS) domainof ATG14 and regulates its membrane curvature sensing (8). TheATG14–BATS domain also binds PtdIns(4,5)P2 without unknownfunctional relevance (8).Phosphoinositides are important lipid messengers in various

membrane trafficking events. PtdIns(4,5)P2 plays multiple rolesat the plasma membrane, such as in endocytosis, exocytosis, focaladhesion assembly, and so on (9). However, recently studies haveargued for roles for PtdIns(4,5)P2 at intracellular compart-ments (10). The majority of PtdIns(4,5)P2 in mammalian cells aregenerated by the type I phosphatidylinositol 4-phosphate (PIP)5-kinase (PIPKI), of which there are three genes (α, β, and γ) eachwith multiple splicing isoforms (11). PIPKIα controls a nuclearPtdIns(4,5)P2 pathway that regulates the mRNA processing ofselect genes activated by different cellular stresses (12). PIPKIγi2partially targets recycling endosomes and promotes the recyclingof integrins and epithelial cadherin by generating PtdIns(4,5)P2and regulating exocyst and adaptor protein complexes (13–15),whereas PIPKIγi5 is found at endosomal compartments where it

mediates PtdIns(4,5)P2 signaling to regulate the endolysosomaltrafficking of epidermal growth factor receptor (EGFR) (16, 17).PtdIns(4,5)P2 mediates endocytosis at the plasma membrane andthe trafficking of autophagic membrane precursors, both of whichmay provide membrane sources for autophagy initiation at the ERsurface (18, 19). PtdIns(4,5)P2 binding proteins such as SNX18,Myo1c, and ATG14 have roles in autophagy (8, 18, 20). However,a role for PtdIns(4,5)P2 and the kinases that synthesize this mes-senger in autophagy initiation has not been reported.In this study, we tested the hypothesis that ATG14 is a PtdIns

(4,5)P2 effector that associates with and is regulated by a specificPIP kinase. We identified PIPKIγi5 as an ATG14 binding partnerthat controls ATG14 protein stability, complex assembly, andautophagy initiation. This study reveals an unexpected role forPtdIns(4,5)P2 in the regulation of a PtdIns(3)P-generating com-plex and autophagy.

ResultsATG14 Interacts and Colocalizes with PIPKIγi5 at the ER–EndosomeInterface. ATG14 membrane targeting depends on its carboxyl-terminal BATS domain that binds both PtdIns(3)P and PtdIns(4,5)P2(8). PtdIns(4,5)P2 binding proteins are often regulated by associationwith PIP kinases, which generates localized PtdIns(4,5)P2 (9). Toidentify a PIP kinase that associates with ATG14, a coimmuno-precipitation (co-IP) assay was performed to examine ATG14 in-teraction with different PIPKIs. ATG14 associated specificallywith PIPKIγi5 (Fig. 1A), and this was confirmed by co-IP of theendogenous proteins (Fig. 1B) and with other subunits of theATG14 complex, but serum starvation that stimulates autophagydid not influence these interactions (Fig. 1B and Fig. S1).

Significance

Autophagy is a conserved lysosomal degradation pathway, thederegulation of which is found in many human diseases, includingcancers, neurodegeneration diseases, and aging. Understandingthe molecular mechanisms of autophagy regulation is the basis forimproving clinical therapeutics. Phosphatidylinositol 3-phosphate[PtdIns(3)P] is a key lipid messenger directly involved in autophagyinitiation. Here we discovered that PtdIns(4,5)P2 plays an equallyimportant role in autophagy. The autophagy-related protein 14(ATG14) specifically interacts with PIPKIγi5, a PtdIns(4,5)P2 gener-ating enzyme that controls ATG14 stability and protein–proteininteractions. PtdIns(4,5)P2 directly binds ATG14 and regulates theATG14–VPS34 (a class III phosphatidylinositol 3-kinase) complexassembly. Our study identifies a PtdIns(4,5)P2 signaling pathwaythat is directly involved in autophagosome membrane initiationand argues for a previously unappreciated role for endosomes inautophagy initiation.

Author contributions: X.T. and R.A.A. designed research; X.T., N.T., Y.L., and S.C. per-formed research; X.T., N.T., Y.L., and S.C. analyzed data; and X.T. and R.A.A. wrotethe paper.

The authors declare no conflict of interest.

This article is a PNAS Direct Submission.1X.T. and N.T. contributed equally to this work.2To whom correspondence should be addressed. Email: raanders@wisc.edu.

This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.1073/pnas.1523145113/-/DCSupplemental.

10896–10901 | PNAS | September 27, 2016 | vol. 113 | no. 39 www.pnas.org/cgi/doi/10.1073/pnas.1523145113

Dow

nloa

ded

by g

uest

on

Feb

ruar

y 7,

202

0

To assess the subcellular localization of PIPKIγi5 and ATG14,we coexpressed EGFP–ATG14 and monomeric DsRed–PIPKIγi5in MDA–MB-231 cells. PIPKIγi5 and ATG14 partially colocalizedin punctate intracellular compartments (Fig. 1C, Upper). PIPKIγi5has been previously shown to associate with endosomes (16, 17),but ATG14 puncta form on the ER surface (7, 21), which wasconfirmed here (Fig. 1C, Lower). Of note, the subcellular locali-zation of PIPKIγi5 is sensitive to its expression levels, as highexpression of PIPKIγi5 disrupted its specific targeting (Fig. S2A).When expressed at physiologic levels, PIPKIγi5 localized topunctate structures with endosomal but not with mitochondrial orthe Golgi complex markers (Fig. S2B). Late endosomes, are oftenin contact with the ER (22), and it is possible that PIPKIγi5 andATG14 concentrate at these sites. Consistently, PIPKIγi5 andATG14 associate with both ER and late endosomes, as shown bycostaining with the ER marker Calnexin and the late endosomemarker LAMP1 (Fig. 1 D and E). The ER targeting of PIPKIγi5and ATG14 is also observed using ER Tracker (Fig. S2C). TheATG14 endosomal association is specific for late but not earlyendosomes (Fig. S3A). This is confirmed with ATG14 targetingto LAPTM4B (Fig. S3B, Upper), another endosome marker pri-marily localized to late endosomes (17).ATG14 and ATG5 puncta formation is associated with auto-

phagy initiation (21, 23). ATG5 puncta, which is usually used as aphagophore marker, also localized to LAPTM4B positive endo-somes (Fig. S3B, Lower). Similarly, LC3 was found to have ex-tensive contacts with LAMP1 compartments (Fig. 1 F and G) andwith EGFR+ endosomes (Fig. S3C) that are involved in autophagyinitiation (24). PIPKIγi5 and ATG14 also extensively colocalizedwith LC3 and LAMP1 (Fig. 1 F and G), and so did ATG5 (Fig.S3D). The data suggest that PIPKIγi5 and ATG14 might dy-namically interact at an ER–endosome interface.

PIPKIγi5 Controls the Stability of the ATG14–Beclin 1 Complex. Smallinterference RNA (siRNA)-mediated knockdown of PIPKIγi5,but not PIPKIγi2, caused a loss of ATG14, Beclin 1, and UV ir-radiation resistance-associated gene (UVRAG) (Fig. 2 A and B),which was also observed in another cell line (Fig. S4A). To con-firm that the loss of ATG14 and Beclin 1 was due to PIPKIγi5

depletion, two PIPKIγi5-knockout cell lines were generated, andthis confirmed that ATG14 and Beclin 1 were lost in these cells(Fig. S4B). Whereas PIPKIγi5 is required for ATG14 levels,knockdown or knockout of ATG14 did not affect PIPKIγi5 ex-pression (Fig. S4 C and D).Although PIPKIγi5 associated with ATG14, Beclin 1, and

VPS34 in overexpressed co-IP (Fig. S4E), PIPKIγi5 did not affectthe levels of VPS34 and Rubicon (Fig. 2A). Similarly, whereasPIPKIγi5 knockdown resulted in a loss of the ATG14–Beclin1–VPS34 complex (Fig. 2B), the Beclin 1–Rubicon complex wasnot down-regulated (Fig. 2B). The protein turnover of ATG14 isregulated by ubiquitination and proteasome degradation (25, 26).In cells with stable ectopic ATG14 expression, knockdown ofPIPKIγi5 nonetheless caused a loss of ATG14 (Fig. 2C) that wasrescued by MG132, a proteasome inhibitor, indicating thatPIPKIγi5 did not control ATG14 levels by transcriptional regu-lation. Consistently, loss of PIPKIγi5 increased ubiquitination ofATG14 (Fig. 2D). ATG14 ubiquitination and turnover is regu-lated by Cullin3 complexes (25, 26); knockdown of Cullin3 res-cued the ATG14 levels in PIPKIγi5-depleted cells (Fig. 2E).Thus, PIPKIγi5 controls the stability of ATG14 by regulating itsubiquitination and degradation.

PIPKIγi5 Is Required for Autophagy.Endocytosis and the trafficking ofautophagic membrane precursors are modulated by PtdIns(4,5)P2(18, 19), but a direct role for PtdIns(4,5)P2 in the regulation ofautophagy initiation has not been demonstrated. The PIPKIγi5interaction with and regulation of ATG14 stability suggest thatPIPKIγi5-mediated PtdIns(4,5)P2 signaling might be involvedin autophagy initiation. The double FYVE-containing protein 1(DFCP1) is a PtdIns(3)P effector and early autophagic markerdownstream of the ATG14 complex and upstream of LC3 lip-idation (1). Consistent with a role for PIPKIγi5 in ATG14 com-plex stability, the puncta formation of DFCP1 was stronglysuppressed in PIPKIγi5-knockdown cells (Fig. 3 A and B), con-sistent with a role for PIPKIγi5 in autophagy initiation. DFCP1puncta were observed in control siRNA transfected MDA–MB-231 cells cultured in either normal or serum free-media (Fig. 3A),suggesting relatively high basal autophagic activity in this cell line.

Fig. 1. PIPKIγi5 interacts and colocalizes with ATG14. (A) ATG14 specifically co-IPs with PIPKIγi5. HEK293T cells were cotransfected with GFP–ATG14 and Myc-tagged PIPKI isoforms. Twenty-four hours after transfection, cells were harvested and WCLs used for IP with anti-Myc. The immunocomplex was analyzed byimmunoblotting. (B) Endogenous co-IP of PIPKIγi5 and ATG14 in MDA–MB-231 cells. Cells were cultured in normal medium with serum (N) or serum-freemedium (S) for 24 h, andWCLs were used for IP with anti-PIPKIγi5. The immunocomplex was analyzed by immunoblotting. (C) GFP–ATG14 partially colocalizeswith PIPKIγi5 and Calnexin. MDA–MB-231 cells were cotransfected with DsRed(monomer) –PIPKIγi5 and EGFP–ATG14 or singly transfected with EGFP–ATG14.Forty-eight hours after transfection, cells were fixed and stained as indicated. (D–G) PIPKIγi5 and ATG14 puncta are localized to endosome–autophagosomeand endosome–ER-associated structures. MDA–MB-231 cells were transfected with DsRed(monomer) –PIPKIγi5 or mCherry–ATG14, and 48 h later, cells werefixed and stained for indicated subcellular markers. Boxes are selected regions for magnified view. (Scale bars, 10 μm.)

Tan et al. PNAS | September 27, 2016 | vol. 113 | no. 39 | 10897

CELL

BIOLO

GY

Dow

nloa

ded

by g

uest

on

Feb

ruar

y 7,

202

0

A431 cells showed no clear LC3 or DFCP1 puncta staining innormal media, reflecting lower basal autophagy, and serum-star-vation–induced DFCP1 puncta formation in this cell line was alsosuppressed by PIPKIγi5 knockdown (Fig. S5). The data suggestthat PIPKIγi5 controls initiation of both basal autophagy andserum-starvation–induced autophagy.To further investigate the role for PIPKIγi5 in autophagy, LC3

lipidation was analyzed. Knockdown of PIPKIγi5 strongly blockedautophagic LC3-II turnover, as indicated by the loss of LC3-IIaccumulation upon chloroquine (CQ) treatment (Fig. 3C). Similarresults were observed in both A431 and MDA–MB-231 cells(Fig. 3D). Loss of PIPKIγi2 did not affect LC3-II turnover (Fig.S6A), consistent with a specific interaction between ATG14 andPIPKIγi5 (Fig. 1A) as well as a requirement for PIPKIγi5 but notPIPKIγi2 in ATG14 stability (Fig. 2A). Overexpression of PIPKIγi5also strongly suppressed autophagy initiation (Fig. S6B), indicatingthat PIPKIγi5 levels must be tightly controlled for proper func-tioning in autophagy. This is generally the case for PIP kinases asthey lose normal targeting when overexpressed (9). Of note, al-though LC3-II generation was blocked in PIPKIγi5-knockdowncells, the basal LC3-II levels without lysosomal inhibition wereconsistently higher than in control cells (Fig. 3 C and D and Fig.S6C), implying inhibition of both autophagy initiation and matu-ration upon loss of PIPKIγi5. This hypothesis was confirmed usinga cell line stably expressing mCherry–EGFP–LC3, in which redpuncta correspond to matured and acidified autophagosomes or

autolysosomes as the EGFP signal is quenched in acidic com-partments, whereas yellow puncta represent autophagosome pre-cursors, phagophores or newly formed autophagosomes (27). InPIPKIγi5-knockdown cells, more yellow LC3 puncta were observed(Fig. 3E), indicating a block of autophagosome maturation. Cos-taining of LC3 and the late endosome/lysosome marker LAMP1confirmed the accumulation of immature autophagosomes, most ofwhich have not fused with endolysosomes (Fig. S6D). Together,these results indicate that loss of PIPKIγi5 causes not only a sup-pression of autophagy initiation as reflected by less LC3-II accu-mulation upon lysosomal inhibition, but it also causes a moresevere block of autophagosome–endolysosome fusion.

PIPKIγi5 Generates PtdIns(4,5)P2 for Autophagic Signaling. The datasuggest that PtdIns(4,5)P2 signals are required for autophagy ini-tiation and potentially also autophagosome maturation. To detectPtdIns(4,5)P2 generation, GFP–PLCδ–PH, a specific PtdIns(4,5)P2probe, was expressed, and the cells were then costained forautophagic membrane marker LC3 and endosome markerLAMP1. Although high expression of PLCδ–PH primarily causedplasma membrane targeting, clear localization of PLCδ–PH tointracellular compartments was observed in cells expressing lowlevels of PLCδ–PH (Fig. S7). Remarkably, although all cells

Fig. 2. PIPKIγi5 is required for the protein stability of ATG14 and Beclin 1.(A) Knockdown of PIPKIγi5 but not PIPKIγi2 causes loss of ATG14 and Beclin 1.MDA–MB-231 cells were transfected with indicated siRNA and after 72 h, celllysates were used for immunoblotting. (B) Knockdown of PIPKIγi5 destabilizesthe Beclin 1–ATG14–VPS34 complex, but not the Beclin 1–Rubicon complex.MDA–MB-231 cells were transfected with indicated siRNA, and 72 h later, cellswere harvested for IP assay using ATG14 or Beclin 1 antibodies, followed byimmunoblotting analysis of the immunocomplex. (C) Knockdown of PIPKIγi5destabilizes ectopically expressed ATG14–Flag. MDA–MB-231 cells stablyexpressing ATG14–Flag were transfected with control or PIPKIγi5 siRNA. Se-venty-two hours later, cells were treated as indicated with MG132 (20 μM) for6 h and harvested for immunoblotting. (D) Knockdown of PIPKIγi5 enhancesubiquitination of ATG14. MDA–MB-231 cells were transfected with indicatedsiRNA; 72 h later, cells were treated with 20 μM MG132 for 5 h and thenharvested for IP using ATG14 antibody, followed by immunoblotting with ananti-ubiquitin antibody. (E) Cullin3 is required for ATG14 loss in PIPKIγi5-knockdown cells. MDA–MB-231 cells were transfected with indicated siRNAsand cell lysates were harvested 72 h later for immunoblotting.

Fig. 3. PIPKIγi5 is required for autophagic turnover. (A and B) Knockdown ofPIPKIγi5 suppresses DFCP1 puncta formation in MDA–MB-231 cells. Cells weretransfected with indicated siRNA; 48 h later, cells were transfected with Myc-tagged DFCP1 DNA and further cultured in normal or serum-free media for24 h before being fixed for immunostaining of Myc–DFCP1. DAPI stains thenuclei. The average numbers of DFCP1 puncta were quantified in B. Mean + SD;n = 3. **P < 0.01. (C andD) Knockdown of PIPKIγi5 blocks LC3-II turnover in A431(A) and MDA–MB-231 (B) cells. Cells were transfected with indicated siRNA, andWCLs were prepared 72 h later for immunoblotting. The levels of LC3-II werequantified and normalized to actin levels. Mean + SD; n = 3. *P < 0.05; **P <0.01. (E) Knockdown of PIPKIγi5 blocks autophagosome maturation. MDA–MB-231 cells stably expressing mCherry–EGFP–LC3 were transfected with in-dicated siRNA, and 72 h later, cells were fixed for fluorescence microscopy.Boxes are selected regions for magnified view. (Scale bars, 10 μm.)

10898 | www.pnas.org/cgi/doi/10.1073/pnas.1523145113 Tan et al.

Dow

nloa

ded

by g

uest

on

Feb

ruar

y 7,

202

0

showed similar levels of LAMP1 staining, the PLCδ–PH punctawere observed specifically in cells with more LC3 puncta, wherethey extensively associated with LC3 and LAMP1 compartments(Fig. 4A). This finding suggests that cells with more autophagicactivity may require more spatial generation of PtdIns(4,5)P2.Similar to the localization of ATG14 and PIPKIγi5, PLCδ–PHextensively associated with ER and endosomes (Fig. 4B), but it didnot show evident association with Golgi or mitochondria (Fig. S7).When coexpressed in MDA–MB-231 cells, PIPKIγi5 and PLCδ–PHcolocalized on microdomains of LAMP1+ or LC3+ compartments(Fig. 4C), indicating that the ectopically expressed PIPKIγi5 still haskinase activity and is able to produce PtdIns(4,5)P2. Knockdown ofPIPKIγi5 completely blocked the labeling of intracellular puncta byPLCδ–PH, although more LC3 puncta accumulated in the knock-down cells (Fig. 4D), indicating that PIPKIγi5 is the major enzymethat produces intracellular PtdIns(4,5)P2 signals detectable byPLCδ–PH at these sites and under these conditions.

PtdIns(4,5)P2 Binding Controls ATG14 Function in Autophagy. AsPtdIns(4,5)P2, the product of PIPKIγi5, binds the ATG14–BATSdomain, the minimally required region for membrane targetingof ATG14 (8), it suggests that PtdIns(4,5)P2 binding might reg-ulate ATG14 function in autophagy. PtdIns(4,5)P2 binding of theBATS domain was confirmed in liposome binding assays (Fig.5A). The BATS domain consists of 80 aa, in which there are only

four basic residues (R423, R442, K486, and R492) that couldpotentially bind negatively charged PtdIns(4,5)P2. Four BATSmutants were generated, each with one basic residue mutatedinto alanine, but the K486A mutant was not expressed inEscherichia coli. The other three BATS mutants were allexpressed well and purified for liposome binding assay to testPtdIns(4,5)P2 and PtdIns(3)P binding. As shown in Fig. 5B,BATS–R492A had normal PtdIns(4,5)P2 binding and slightlyreduced PtdIns(3)P binding, whereas BATS–R423A and BATS–R442A lost PtdIns(4,5)P2 binding and kept PtdIns(3)P binding.Thus, R423A and R442A are two point mutants that specificallylose PtdIns(4,5)P2 binding.When either of the two point mutations was introduced into

full-length ATG14, it reduced ATG14 interaction with VPS34and Beclin 1, but it was not the case when introducing the R492Apoint mutation that retains PtdIns(4,5)P2 binding (Fig. 5C), al-though all three mutants showed normal puncta formation (Fig.S8A). Consistent with literature (3–6), knockdown of ATG14strongly blocked autophagy initiation (Figs. S8B and S9). InATG14 knockdown cells, the ATG14 R423A and R442A mu-tants, but not the R492A mutant, failed to rescue autophagicLC3-II turnover (Fig. 5D). Surprisingly, the ATG14 R423A andR442A mutants did not cause LC3-II accumulation as observedin PIPKIγi5 knockdown cells. Costaining of LC3 and LAMP1showed normal autophagosome–endolysosome fusion in ATG14knockdown cells rexpressing either wild-type, R423A, or R442Amutant of ATG14 (Fig. S9). This might be explained by theirnormal interaction with syntaxin 17 (STX17) (Fig. S10A), as theATG14–STX17 interaction has been recently shown to promoteautophagosome fusion with endolysosomes (28).The ATG14 R423A and R442A mutants also have normal in-

teraction with PIPKIγi5 (Fig. S10B), which might confer a func-tion of ATG14 in autophagosome–endolysosome fusion withoutPtdIns(4,5)P2 binding to its BATS domain. Moreover, the ATG14R423A and R442A mutants are as stable as wild-type protein (Fig.5 C and D), indicating that PIPKIγi5 controls ATG14 stability bymechanisms other than PtdIns(4,5)P2 binding to its BATS do-main. Compared with these ATG14 mutants, loss of PIPKIγi5causes more severe defects, such as loss of ATG14, Beclin 1, andUVRAG. This finding could explain why the ATG14 mutants donot fully phenocopy PIPKIγi5 knockdown.Our results identified PIPKIγi5 as a specific interacting protein

of ATG14 and it controls ATG14 stability and autophagic activity.PtdIns(4,5)P2, the product of PIPKIγi5, binds to the ATG14–BATS domain and controls ATG14 interactions with VPS34 andBeclin 1, which in turn regulates ATG14 function in autophagyinitiation (Fig. 5E). In addition, PIPKIγi5 also regulates autophago-some degradation, which is, however, independent of PtdIns(4,5)P2binding to the ATG14–BATS domain. This study revealed animportant role for PtdIns(4,5)P2 signaling in ATG14 complexassembly and autophagy initiation.

DiscussionPhosphoinositides are present in all cellular membranes andregulate membrane trafficking events including autophagicmembrane trafficking. PtdIns(3)P has been considered as thesole phosphoinositide lipid messenger that controls autophagyinitiation (29). The VPS34 complex generates PtdIns(3)P at theER surface, which functions as a lipid platform for the re-cruitment of downstream effectors to initiate the phagophoremembrane, such as DFCP1 and WD repeat domain, phos-phoinositide interacting 2 (WIPI2) (1). However, recent evi-dence has argued for a role for PtdIns(5)P, instead of PtdIns(3)P, in autophagy initiation upon glucose starvation (30). Thecurrent study further expands this lipid pool to include PtdIns(4,5)P2, which regulates the VPS34 complex. Our results es-tablish that the PIPKIγi5-mediated PtdIns(4,5)P2 signalingfunctions upstream of PtdIns(3)P by stabilizing ATG14 and

Fig. 4. PIPKIγi5 generates autophagic PtdIns(4,5)P2 signaling. (A and B) MDA–MB-231 cells were transfected with GFP–PLCδ–PH, and 48 h later, cells werefixed and stained for endogenous LAMP1 and LC3 (A) or Calnexin (B).(C) PIPKIγi5 colocalizes with PLCδ–PH at LAMP1- and LC3-associated mem-branes. MDA–MB-231 cells were cotransfected with DsRed(monomer) –i5 andGFP–PLCδ–PH, and 48 h later, cells were fixed and stained for endogenousLAMP1 or LC3. (D) Knockdown of PIPKIγi5 blocks PLCδ–PH puncta formation atLC3-associated membranes. MDA–MB-231 cells were treated with control orPIPKIγi5 siRNA, and 48 h later, cells were transfected with GFP–PLCδ–PH. Cellswere fixed 48 h after DNA transfection and stained for endogenous LC3. Boxesare selected regions for magnified view. (Scale bars, 10 μm.)

Tan et al. PNAS | September 27, 2016 | vol. 113 | no. 39 | 10899

CELL

BIOLO

GY

Dow

nloa

ded

by g

uest

on

Feb

ruar

y 7,

202

0

Beclin 1 and promoting the ATG14–Beclin 1–VPS34 complexassembly (Fig. 5E). These data suggest that additional PtdIns(4,5)P2 effectors and lipid metabolizing enzymes such as PI4K,PtdIns(4,5)P2 phosphatases, SNX18, and others (18, 19, 31)might also be involved in fine tuning autophagy initiation.Autophagy initiates at preautophagosomal structure (PAS) near

the vacuole in yeast (32), which provides a basis to couple auto-phagosome formation and the subsequent fusion with the vacuole.In mammalian cells, most autophagosomes initially fuse with lateendosomes or multivesicular endosomes before finally fusing withthe lysosome, suggesting that autophagosomes might initiate closeto or in contact with late endosomes. The endosomal localizedEGFR controls both basal and serum-starvation–induced auto-phagy initiation (24), consistent with autophagy initiation in close

proximity to endosomes. In support of this finding, most LC3puncta were found closely associated with EGFR+ endosomes (Fig.S3C), which are usually also positive for the lysosomal-associatedprotein transmembrane 4B (LAPTM4B) (24), another endosomemarker (17). The mammalian PAS has been represented by mul-tiple ATG14 and ATG5 puncta that associate with the ER (7, 8,21). However, the relationship between mammalian PAS and lateendosomes/lysosomes has not been addressed to date, although lateendosomes have been demonstrated to constantly associate with theER (22). The current study suggests that PtdIns(4,5)P2 signalingmight function as a connection between the ATG14 puncta andendosomes, as both PIPKIγi5 and ATG14 extensively associate withthe ER and endosomes. Interestingly, both PIPKIγi5 and ATG14puncta are specifically localized to punctate but not tubular ERstructures that colocalize with endosomes (Fig. 1 and Fig. S2C).These punctate ER subdomains might correlate to ER patches (33)that also contribute to the previously described autophagy-initiatingER–mitochondria contacts (21, 34). However, future studies areneeded to dissect the detailed molecular basis for the contacts be-tween ER patches and endosomes in this context.It is noteworthy that PIPKIγi5 has putative roles not only in

autophagy initiation as shown by less LC3-II accumulation uponlysosome inhibition but also in the autophagosome–endolysosomefusion, as loss of PIPKIγi5 blocked the acidification of autophagiccompartments. ATG14 has been recently reported to promoteautophagosome fusion with endolysosomes by interacting withSTX17 and priming the ternary STX17–SNAP29–VAPM8 solubleN-ethylmaleimide-sensitive factor attachment protein receptors(SNARE) complex formation that directly mediates the fusionstep (28). Whether PIPKIγi5 regulates the fusion step by con-trolling the role for ATG14 in STX17 complex assembly is in-trinsically difficult to test because ATG14 is lost upon PIPKIγi5knockdown. In addition, the ATG14 mutants without PtdIns(4,5)P2binding to the BATS domain show defects only in autophagy ini-tiation with no detected LC3-II accumulation, suggesting that eitherPtdIns(4,5)P2 regulates the fusion step by binding to other regionson ATG14 outside of the BATS domain or PIPKIγi5 has anothereffector that contributes to the fusion step, such as Myo1c, a myosinthat specifically binds PtdIns(4,5)P2 and regulates autophagosomefusion with lysosomes (20).A debate over intracellular PtdIns(4,5)P2 signaling is whether

significant PtdIns(4,5)P2 exists on intracellular compartments or ifthis PtdIns(4,5)P2 only reflects a biological noise. Increasing evi-dence supports the specific and essential functions of intracellularPtdIns(4,5)P2 (10). However, PtdIns(4,5)P2 is poorly detectedon intracellular compartments likely because of the lack of freePtdIns(4,5)P2, as it is bound to effectors that associate with PIPkinases (9). More PtdIns(4,5)P2 exists on the plasma membraneand high expression of PLCδ–PH could mask its intracellulartargeting with only apparent localization to the plasma membrane.Interestingly, more PLCδ–PH puncta was observed in cells withmore LC3 punctate staining, indicating a correlation betweenautophagy and intracellular PtdIns(4,5)P2 signaling. The intracel-lular targeting of PLCδ–PH was lost upon knockdown of PIPKIγi5,consistent with its role in PtdIns(4,5)P2 generation at these sites. Itis likely that PIPKIγi5 produces more free PtdIns(4,5)P2 signalsthan PIP kinases in other intracellular pathways, such as PIPKIγi2that controls the endosomal recycling of E-cadherin (13) andintegrins (14).In summary, this study has revealed an endosome- and ER-

associated PtdIns(4,5)P2 signaling pathway that controls ATG14complex stability and assembly required for autophagy. It revealsan unexpected role for PtdIns(4,5)P2 in autophagic membranetrafficking and argues for a role of ER–late endosome contacts inautophagy regulation as a potential mammalian PAS.

Materials and MethodsSee SI Materials and Methods for detailed description.

Fig. 5. PtdIns(4,5)P2 binding controls ATG14 function in autophagy.(A) ATG14–BATS domain binds PtdIns(4,5)P2 in liposome binding assay. A totalof 1 μg of purified T7-tagged BATS domain was incubated with PtdIns(4,5)P2liposomes for 10 min at room temperature. Liposomes with BATS domainbound were collected by centrifugation. The amount of T7–BATS in super-natants and pellets was examined by immunoblotting. (B) Identification ofBATS domain mutants losing PtdIns(4,5)P2 binding. wild-type or mutatedT–BATS domain were purified and assayed in liposome binding as described inA. (C) Full-length ATG14 with PtdIns(4,5)P2-binding defective BATS domainshows diminished VPS34 and Beclin 1 association. HEK293T cells were trans-fected with indicated ATG14 constructs, and 24 h later, cells were harvested forIP assay. (D) ATG14 mutants losing PtdIns(4,5)P2 binding to the BATS domainare defective in autophagy. MDA–MB-231 cells stably expressing siRNA-resistant wild-type or mutated ATG14 were transfected with ATG14 siRNA.Seventy-two hours later, cells were treated as indicated with 80 μM CQ for 2 hand harvested for immunoblotting analysis of LC3-II turnover. LC3-II levelswere quantified and normalized to Actin; mean + SD; n = 4. *P < 0.05; ***P <0.001. (E) Proposed model for PIPKIγi5 and PtdIns(4,5)P2 regulation of theATG14 complex in autophagy. PIPKIγi5 interacts with ATG14 and generatesPtdIns(4,5)P2 to regulate the stability and activity of the ATG14–VPS34 complexat endosome- and ER-associated membrane. PIPKIγi5 also controls autopha-gosome fusion with endolysosomes. PIP2, PtdIns(4,5)P2.

10900 | www.pnas.org/cgi/doi/10.1073/pnas.1523145113 Tan et al.

Dow

nloa

ded

by g

uest

on

Feb

ruar

y 7,

202

0

Reagents, Cell Culture, and Treatments.MDA–MB-231, A431, and HEK293 cellswere cultured in Dulbecco’s Modified Eagle’s Medium (DMEM, Corning)supplemented with 10% (vol/vol) FBS. DNA and siRNA transfection wascarried out using Lipofectamine 3000 (Thermo Fisher Scientific) and Oligo-fectamine RNAiMax (Thermo Fisher Scientific), respectively, following man-ufacturer’s protocols, except that medium was changed 2 h aftertransfection with Lipofectamine 3000. For serum starvation, cells werewashed twice with serum-free DMEM and then cultured with serum-freeDMEM for indicated periods. MG132 was dissolved in DMSO as a 20-mMstock and directly added to cell culture plates for treatments.

Co-IP. IP of indicated proteins was performed using nondenaturing whole celllysates (WCLs). The immunocomplexes were isolated using indicated anti-bodies and protein-G–conjugated beads and separated by SDS/PAGE, fol-lowed by analysis by immunoblotting as indicated.

Autophagy Assays. Autophagic activity was analyzed by Western blottingdetection of LC3-II turnover in the presence and absence of autophagy in-hibitor CQ, which is well established in literature. For autophagosomematuration, an MDA–MB-231 cell line stably expressing mCherry–EGFP–LC3was used to track acidification of the LC3 compartments, as the fluorescencesignals from EGFP, but not mCherry, are quenched by acidic pH uponautophagosome fusion with lysosomes.

Immunofluorescence Microscopy. Immunostaining of endogenous and taggedproteins was carried out following standard protocols. Cells on glass coverslipswere washed, fixed in paraformaldehyde (PFA), permeabilized, and blocked inBSA. Incubation with primary antibodies was performed at 37 °C for 2 h or 4 °Covernight and secondary antibodies at room temperature for 1 h. Fluorescenceimages were obtained using MetaMorph with a Nikon Eclipse TE2000-U mi-croscope and further processed and assembled in MetaMorph and AdobePhotoshop.

Liposome Binding Assay. PtdIns(4,5)P2 PolyPIPosomes (Echelon) were used forthe liposome binding assay. A total of 1 μg of purified BATS domain and 10 μLPolyPIPosomes were incubated at room temperature for 10 min. The lipo-somes were precipitated by centrifugation at 16,000 × g for 10 min. The lipid–protein pellets were washed twice with binding buffer before being dissolvedin SDS-loading buffer and analyzed by Western blot with HRP-conjugatedanti-T7 tag antibody.

ACKNOWLEDGMENTS. We thank all members of the R.A.A. group for dis-cussions and reagents. This study was supported by NIH Grants CA104708and GM057549 (to R.A.A.); Howard Hughes Medical Institute InternationalStudent Research Fellowship 59107631 (to X.T.); and a Hilldale Undergrad-uate/Faculty Research fellowship (to Y.L.).

1. Lamb CA, Yoshimori T, Tooze SA (2013) The autophagosome: Origins unknown,biogenesis complex. Nat Rev Mol Cell Biol 14(12):759–774.

2. Feng Y, He D, Yao Z, Klionsky DJ (2014) The machinery of macroautophagy. Cell Res24(1):24–41.

3. Itakura E, Kishi C, Inoue K, Mizushima N (2008) Beclin 1 forms two distinct phos-phatidylinositol 3-kinase complexes with mammalian Atg14 and UVRAG.Mol Biol Cell19(12):5360–5372.

4. Sun Q, et al. (2008) Identification of Barkor as a mammalian autophagy-specific factorfor Beclin 1 and class III phosphatidylinositol 3-kinase. Proc Natl Acad Sci USA 105(49):19211–19216.

5. Matsunaga K, et al. (2009) Two Beclin 1-binding proteins, Atg14L and Rubicon, re-ciprocally regulate autophagy at different stages. Nat Cell Biol 11(4):385–396.

6. Zhong Y, et al. (2009) Distinct regulation of autophagic activity by Atg14L and Ru-bicon associated with Beclin 1-phosphatidylinositol-3-kinase complex. Nat Cell Biol11(4):468–476.

7. Matsunaga K, et al. (2010) Autophagy requires endoplasmic reticulum targeting ofthe PI3-kinase complex via Atg14L. J Cell Biol 190(4):511–521.

8. Fan W, Nassiri A, Zhong Q (2011) Autophagosome targeting and membrane curva-ture sensing by Barkor/Atg14(L). Proc Natl Acad Sci USA 108(19):7769–7774.

9. Choi S, Thapa N, Tan X, Hedman AC, Anderson RA (2015) PIP kinases define PI4,5P2signaling specificity by association with effectors. Biochim Biophys Acta 1851(6):711–723.

10. Tan X, Thapa N, Choi S, Anderson RA (2015) Emerging roles of PtdIns(4,5)P2: Beyondthe plasma membrane. J Cell Sci 128(22):4047–4056.

11. Schramp M, Hedman A, Li W, Tan X, Anderson R (2012) PIP kinases from the cellmembrane to the nucleus. Subcell Biochem 58:25–59.

12. Mellman DL, et al. (2008) A PtdIns4,5P2-regulated nuclear poly(A) polymerase con-trols expression of select mRNAs. Nature 451(7181):1013–1017.

13. Xiong X, et al. (2012) An association between type Iγ PI4P 5-kinase and Exo70 directsE-cadherin clustering and epithelial polarization. Mol Biol Cell 23(1):87–98.

14. Thapa N, et al. (2012) Phosphoinositide signaling regulates the exocyst complex andpolarized integrin trafficking in directionally migrating cells. Dev Cell 22(1):116–130.

15. Ling K, et al. (2007) Type I gamma phosphatidylinositol phosphate kinase modulatesadherens junction and E-cadherin trafficking via a direct interaction with mu 1Badaptin. J Cell Biol 176(3):343–353.

16. Sun Y, Hedman AC, Tan X, Schill NJ, Anderson RA (2013) Endosomal type Iγ PIP5-kinase controls EGF receptor lysosomal sorting. Dev Cell 25(2):144–155.

17. Tan X, et al. (2015) LAPTM4B is a PtdIns(4,5)P2 effector that regulates EGFR signaling,lysosomal sorting, and degradation. EMBO J 34(4):475–490.

18. Knævelsrud H, et al. (2013) Membrane remodeling by the PX-BAR protein SNX18promotes autophagosome formation. J Cell Biol 202(2):331–349.

19. Moreau K, Ravikumar B, Puri C, Rubinsztein DC (2012) Arf6 promotes autophagosomeformation via effects on phosphatidylinositol 4,5-bisphosphate and phospholipase D.J Cell Biol 196(4):483–496.

20. Brandstaetter H, Kishi-Itakura C, Tumbarello DA, Manstein DJ, Buss F (2014) Loss offunctional MYO1C/myosin 1c, a motor protein involved in lipid raft trafficking, dis-rupts autophagosome-lysosome fusion. Autophagy 10(12):2310–2323.

21. Hamasaki M, et al. (2013) Autophagosomes form at ER-mitochondria contact sites.

Nature 495(7441):389–393.22. Friedman JR, Dibenedetto JR, West M, Rowland AA, Voeltz GK (2013) Endoplasmic

reticulum-endosome contact increases as endosomes traffic and mature. Mol Biol Cell

24(7):1030–1040.23. Ge L, Melville D, Zhang M, Schekman R (2013) The ER-Golgi intermediate compart-

ment is a key membrane source for the LC3 lipidation step of autophagosome bio-

genesis. eLife 2:e00947.24. Tan X, Thapa N, Sun Y, Anderson RA (2015) A kinase-independent role for EGF re-

ceptor in autophagy initiation. Cell 160(1–2):145–160.25. Zhang T, et al. (2015) G-protein-coupled receptors regulate autophagy by ZBTB16-

mediated ubiquitination and proteasomal degradation of Atg14L. eLife 4:e06734.26. Liu C-C, et al. (2016) Cul3-KLHL20 ubiquitin ligase governs the turnover of ULK1 and

VPS34 complexes to control autophagy termination. Mol Cell 61(1):84–97.27. Kimura S, Fujita N, Noda T, Yoshimori T (2009) Monitoring autophagy in mammalian

cultured cells through the dynamics of LC3. Methods in Enzymology (Academic,

New York), Vol 452, Chapter 1, pp 1–12.28. Diao J, et al. (2015) ATG14 promotes membrane tethering and fusion of autopha-

gosomes to endolysosomes. Nature 520(7548):563–566.29. Dall’Armi C, Devereaux KA, Di Paolo G (2013) The role of lipids in the control of

autophagy. Curr Biol 23(1):R33–R45.30. Vicinanza M, et al. (2015) PI(5)P regulates autophagosome biogenesis. Mol Cell 57(2):

219–234.31. Billcliff PG, et al. (2016) OCRL1 engages with the F-BAR protein pacsin 2 to promote

biogenesis of membrane-trafficking intermediates. Mol Biol Cell 27(1):90–107.32. Suzuki K, Ohsumi Y (2010) Current knowledge of the pre-autophagosomal structure

(PAS). FEBS Lett 584(7):1280–1286.33. Lu L, et al. (2013) The small molecule dispergo tubulates the endoplasmic reticulum

and inhibits export. Mol Biol Cell 24(7):1020–1029.34. Area-Gomez E, et al. (2009) Presenilins are enriched in endoplasmic reticulum mem-

branes associated with mitochondria. Am J Pathol 175(5):1810–1816.35. Mizushima N, Sugita H, Yoshimori T, Ohsumi Y (1998) A new protein conjugation

system in human. The counterpart of the yeast Apg12p conjugation system essential

for autophagy. J Biol Chem 273(51):33889–33892.36. Itakura E, Kishi-Itakura C, Mizushima N (2012) The hairpin-type tail-anchored SNARE

syntaxin 17 targets to autophagosomes for fusion with endosomes/lysosomes. Cell

151(6):1256–1269.37. Stauffer TP, Ahn S, Meyer T (1998) Receptor-induced transient reduction in plasma

membrane PtdIns(4,5)P2 concentration monitored in living cells. Curr Biol 8(6):

343–346.38. Schill NJ, Anderson RA (2009) Two novel phosphatidylinositol-4-phosphate 5-kinase

type Igamma splice variants expressed in human cells display distinctive cellular tar-

geting. Biochem J 422(3):473–482.39. Cong L, et al. (2013) Multiplex genome engineering using CRISPR/Cas systems. Science

339(6121):819–823.

Tan et al. PNAS | September 27, 2016 | vol. 113 | no. 39 | 10901

CELL

BIOLO

GY

Dow

nloa

ded

by g

uest

on

Feb

ruar

y 7,

202

0