Autophagy limits the efficacy of temozolomide and curcumin combination in

glioblastomas: role of DNA damage response and MAPK pathways

Alfeu Zanotto-Filho1, Elizandra Braganhol2, Karina Klafke1, Fabrício Figueiró1, Sílvia

Resende Terra1, Francis Jackson Paludo1, Maurílio Morrone1, Ivi Juliana Bristot1, Ana

Maria Battastini1, Cassiano Mateus Forcelini3, Daniel Pens Gelain1, José Cláudio

Fonseca Moreira1.

1 Departamento de Bioquímica, Universidade Federal do Rio Grande do Sul (UFRGS),

Porto Alegre, RS, Brazil. 2 Departamento de Ciências Básicas da Saúde – Universidade Federal de Ciências da

Saúde de Porto Alegre (UFCSPA), Porto Alegre, RS, Brasil. 3 Hospital São Vicente de Paulo, Universidade de Passo Fundo (UPF), Passo Fundo,

RS, Brazil.

Corresponding author:

Alfeu Zanotto-Filho, PhD.

Depto. Bioquímica (ICBS-UFRGS)

Rua Ramiro Barcelos, 2600/Anexo, CEP 90035-003, Porto Alegre, Rio Grande do Sul,

Brazil.

Phone +55(51)3308-5578.

Email: alfeuzanotto@hotmail.com

Running title: Autophagy limits temozolomide and curcumin efficacy

Keywords: curcumin; temozolomide; glioblastoma; autophagy; ERK1/2.

Disclosure statement: The authors disclose no potential conflicts of interest.

Abstract

Decades of research have proven that glioblastoma therapy is still challenging due to

factors such as tumor particular localization, cell growth kinetics and chemoresistance.

In this study, we tested the usefulness of combining the clinical anti-glioblastoma agent

temozolomide (TMZ) with curcumin - a phytochemical described to affect glioblastoma

growth in vitro and in vivo - and the mechanisms involved. The data showed that

curcumin and TMZ synergism was not reached due to redundant molecular

mechanisms leading to G2/M checkpoint activation and autophagy as early events

preceding apoptosis. Autophagy prevented apoptotic machinery activation in

glioblastoma cells in vitro, a phenomenon also observed in C6 brain implants in rats.

Appropriate blocking of this response with early and late autophagy inhibitors (3-

methyl-adenine, ATG7 siRNA and chloroquine) made cells susceptible to TMZ,

curcumin and TMZ/curcumin combinations by increasing apoptotic rates. We depicted

that while curcumin effects were mediated by inhibition of STAT3, NFκB and PI3K/Akt

pathways, TMZ-induced autophagy was dependent on DNA damage response

machinery (as ATM and MSH6) and MAPKs activation. Amidst the MAPKs, we

identified ERK1/2 as a common mechanism of TMZ and curcumin-induced autophagy,

which could be blocked to achieve apoptosis. Last, we validated that resveratrol -

which is a brain barrier permeable drug - switched cells to apoptosis through inhibition

of ERK1/2-mediated autophagy in vitro and in vivo. Taken together, the data show that

autophagy impairs efficacy of TMZ, curcumin and its combinations and support the use

of autophagy modulators in optimizing glioblastoma therapy.

Introduction

Over the last decades, cumulating evidences have shown that the polyphenolic

curcumin is active against a variety of cancers [1]. The ability of interfering with cancer

overexpressed pathways as NFκB, Jak/STAT3, Sonic Hedgehog and PI3K/Akt is

pivotal in understanding the efficacy as well as the selectivity of curcumin in inducing

cell death in tumor while sparing healthy tissues [2]. Our group has sought to

understand the mechanisms whereby curcumin affects glioblastomas aiming to use it

for therapeutic advantage. We have reported that despite the concerns regarding its

oral biovailability [1], intraperitoneal curcumin is able to decrease brain-implanted

glioblastomas [3]. Others have ran to same directions, with similar dose-effect patterns

in vitro and other animal models [4;5;6;7]. With collaborators, we designed curcumin-

loaded nanocapsules with improved preclinical efficacy [8].

Despite the advances in the therapy of diverse cancers, glioblastomas basically

relies on limited efficacy of the alkylating temozolomide (TMZ) in combination with

neurosurgery and/or radiotherapy, and although efforts have been done - for instance

the development of targeted therapies [9] - prognosis remains dismal, with median

survival ~14 months. Most of difficulties in treating glioblastomas are consequent of its

particular localization, which limit the bioavailability of chemotherapies due to

restrictions imposed by the blood-brain barrier (BBB), but also come from its highly

proliferative and infiltrative phenotypes caused by frequent p53, EGFR and PTEN

mutations [10].

Given the preclinical efficacy of curcumin and the clinical usefulness of TMZ in

treating glioblastomas, we tested whether curcumin increases the efficacy of TMZ in

vitro and in vivo. The experiments were undertaken in order to investigate not only the

synergism but also the molecular mechanisms orchestrating the effects of

curcumin/TMZ combination. Our data pointed a role of autophagy in limiting

glioblastomas to undergo apoptosis, and appropriate blocking of this phenomenon

rescued cells toward more sensitive phenotypes in vitro and in vivo.

Material and Methods

Reagents

Curcumin, temozolomide, propidium iodide/PI, acridine orange, MTT (3-(4,5-dimethyl)-

2,5-diphenyl tetrazolium), BAY117082, 3-methyl-adenine (3-MA), UO126, SP600129,

SB203508, Stattic and LY294002 were from Sigma (São Paulo, Brazil). SDS/PAGE

and immunoblot reagents were from Bio-Rad (Hercules, CA, USA). The phospho-

p65Ser536 (#3033), phospho-STAT3Tyr705 (#9131), Bax (#2772), cyclin D1 (#2926),

phospho-Chk1Ser345 (#2348), phospho-cdc25cSer216 (#9528), caspase-3 (#9665), PARP

(#9532), BiP/GRP78 (#9956), CHOP (#2895), LC3B (#2775) and ATG7 (#8558), p-

ERK1/2Thr202/Tyr204 (#9101) and anti-rabbit/mouse IgG HRP-linked antibodies were from

Cell Signaling Technology (USA).

Cell lines

C6, U251MG and U87MG cell lines were from American Type Culture Collection

(ATCC) (Rockville, Maryland, USA). Primary astrocytes were isolated from cortex of 2-

days old Wistar rats by mechanical dissociation with Ca+2/Mg+2-free Hank’s balanced

salt solution and plated in poly-L-lysine-coated plates as described [11].The cells were

maintained in low-glucose DMEM plus antibiotics (Gibco BRL, Carlsbad, USA) in a

humidified incubator. TMZ, curcumin and inhibitors stocks were dissolved in DMSO.

Inhibitors were pre-treated for 3 h.

Viability assays

MTT reduction by cellular dehydrogenase was used to estimate cellular viability

[11].The leakage of lactate dehydrogenase (LDH) into culture medium was measured

following the CytoTox 96-NonRadioactive Cytotoxicity Assay kit instructions (Promega).

For PI uptake, treated cells were incubated with 6 uM PI in DMEM for 1 h, and images

were obtained using a Nikon Eclipse TE 300 inverted microscope set up with

rhodamine filter. Manual counting was performed by using a hemocytometer.

Clonogenic survival assays

Exponentially growing cells (1 x 106) were 6-well plated and then treated for 72 h. After,

the cells were trypsinized and 104 Trypan blue-negative viable cells were re-plated in 6-

well plates, and maintained for additional 9 days. Viable colonies were stained with

MTT (2 mg/mL) and manually counted under magnification.

Cell cycle analysis

Treated cells were trypsinized, centrifuged and resuspended in 500 μl lysis buffer (10

mM PBS, 0.1% v/v Nonidet P-40, 1.2 mg/mL spermine, 5 µg/ml RNAse, 2.5 µg/mL

propidium iodide/PI, pH 7.4). Cells were vortexed and incubated for 10 min on ice for

permeabilization and PI staining. DNA content was determined by FACS, and analyzed

using CellQuest® (BD Biosciences, USA).

Western blots

Proteins (~40 μg) were separated by SDS-PAGE and electrotransfered onto nitrocellulose

membranes. The membranes were Ponceau-stained, rinsed with TBS-T, blocked with 5%

non-fat dry milk in TBS-T (1 h), and then incubated with primary antibodies (1:1000; 12

h/4°C). Following incubation with HRP-linked anti-IgG (1:3000, 2h/room temperature),

chemiluminescence was detected using X-ray films.

Annexin-V staining

Annexin V-FITC Apoptosis Detection Kit (Sigma-Aldrich) was used for quantification of

apoptosis agreeing.. After trypsinization, externalized phosphatidylserine was labeled

with annexin-V-FITC-conjugated for 15 min on ice. PI (1 µg/mL) was added 10 min

prior to FACS analysis. Viable (annexin-/PI-), early (annexin+/PI-) and late apoptotic

(annexin+/PI+), and necrotic cells (annexin-/PI+) were assigned. Data were analyzed

by CellQuest® (BD Biosciences, USA).

Caspase-3/7 activity

Caspase-3/7 activity was assessed following CASP3F Fluorimetric kit instructions

(Sigma, Saint Louis/MI). After whole-cell extracts preparation, 150 μg proteins

(Bradford method) were mixed with 200 μL assay buffer containing Ac-DEVD-AMC, a

caspase-3-specific substrate. Ac-DEVD-AMC cleavage was monitored for 1 h at 37°C

in a fluorescence reader (Ex/Em: 360/460 nm).

Comet assay

Briefly, 2 x 104 cells in 20 µl DMEM was mixed with 80 µl of low-melting agarose

(0.75%), and added to a 1.5% agarose-coated microscope slide. The slides were

placed into a lysis solution (2.5 M NaCl, 100 mM EDTA and 10 mM Tris, pH 10.5, plus

1% Triton X-100 and 10% DMSO) overnight at 4°C, electrophoresed in alkaline

solution (300 mM NaOH and 1 mM EDTA) during 20 min at 25 V (0.90 V/cm; 300 mA),

neutralized with 400 mM Tris (pH 7.5), washed with dH2O, and stained with SYBR

Safe. Hundred cells per sample were counted and classified based on tail length using

a fluorescence microscope. Comet Index (CI) was calculated as follow: CI =

[(n°type1)x1+(n°type2)x2+(n°type3)x3+(n°type4)x4], where “n° type X” is the number of

cells carrying a type X of DNA damage.

Acridine orange staining

Acridine orange (AO) is a probe that fluoresces green in the whole cell except in acidic

compartments, where it fluoresces red. Vacuolar acidification of autophagosomes is a

characteristic of efficient autophagy, thus the red fluorescence is proportional to

autophagy. Treated cells were incubated with AO (1 µg/mL) for 15 min, trypsinized,

centrifuged, and resuspended in PBS. The green and red (FL1/FL3-H) fluorescences

were detected using a FACSCalibur, and analyzed using CellQuest® (BD Biosciences,

San Jose, USA).

Small interference RNA (siRNA)

Cells were transfected using the Lipofectamine RNAiMax (Invitrogen, USA) by reverse

transfection (20 to 30 nM siRNA). Protein knockdown was confirmed by Western blot.

Human ATG7 (Autophagy-related 7 homolog) siRNA was from Cell Signaling (#6604);

ATM siRNA (sc-29761) was from Santa Cruz; and MSH6 siRNA (#144497) was from

Invitrogen. Silencer® Select Control#1 siRNA was used as a scrambled control.

STAT3 and NFκB DNA-binding activity

TransAM™ STAT3 and TransAM® Flexi NFκB p65 ELISA kits were used to address

DNA-binding activity of these transcription factors according to instructions (Active

Motif Inc., USA). Nuclear extracts were prepared as previously described [11] and 10

ug of soluble nuclear fractions were assayed.

Immunofluorescence

Cover slips seeded cells were treated and fixed in 4% paraformaldehyde/PBS, rinsed

and permeabilized with ice-cold 100% methanol (10 min, –20°C). Cells were blocked

with 5% horse serum/PBS (1h), and then incubated with rabbit anti-LC3B antibody

(1:500; overnight/4ºC) followed by anti-rabbit IgG Alexa 594-conjugated (1:500/2h at

room temperature). For nuclear staining, Prolong Gold Antifade Reagent with DAPI

(Molecular Probes, P36931) was used. Images were taken with an Olympus

FluoView™ 1000 confocal microscope and analyzed by ImageJ®.

Immunohistochemistry

Cryostat sections from rat brain or human glioblastoma specimens were deparaffinized,

unmasked using citrate method, and blocked in 1% albumin. Anti p-ERK1/2 (E4, sc-

7383, 1:100; Santa Cruz) or anti-LC3B (1:200) primary antibodies were incubated

overnight at 4°C. The sections were then incubated with biotinylated secondary

antibody followed by streptavidin-avidin-biotin (kit Lsab, Dako, CA, USA). The

peroxidase reaction was performed using DAB counterstained with Harris hematoxylin.

Human glioblastoma specimens were obtained from Hospital Sao Vicente de Paulo,

Universidade de Passo Fundo (UPF), and approved by UPF Ethical Committee

(Approval N° 200/2011).

Animal studies: C6 glioblastoma model

C6 cells brain implants were performed as previously [3;8]. C6 cells (0.8×106/3 μL)

were injected using a Hamilton microsyringe coupled with an infusion pump into the

right striatum of 8-weeks-old male Wistar rats anesthetized by ketamine/xylazine.

Animal studies followed the “Policy on Humane Care and Use of Laboratory Animals”

of the NIH, and were approved by local Ethics Committee for Animal Experimentation

(Protocol: 23810). After 7 days for tumor establishment, animals were grouped as

follows (n=9-11/group): Vehicle (DMSO); Curc (Curcumin 50mg/Kg); RSV (Resveratrol

10mg/Kg); TMZ (temozolomide 10mg/Kg); TMZ+Curc (TMZ 10mg/Kg + Curcumin

50mg/Kg); TMZ+RSV (TMZ 10mg/Kg + RSV 10mg/Kg); TMZ+Curc+RSV (TMZ

10mg/Kg + Curcumin 50mg/Kg + RSV 10mg/Kg); TMZ+CQ (TMZ 10mg/Kg +

Chloroquine 20mg/Kg); TMZ+Curc+CQ (TMZ 10mg/Kg + Curcumin 50mg/Kg +

Chloroquine 20mg/Kg); CQ (Chloroquine 20mg/Kg). TMZ was administrated thrice a

week; curcumin, RSV and CQ were daily (i.p.). In combinations, the solutions were

mixed, and a single injection was applied. The rats were decapitated after 14 days of

treatment. The brain was removed, fixed with 10% paraformaldehyde and paraffin

embedded. Three H&E-stained coronal sections (3 μm thick) from each animal were

analyzed. Images were captured, and the tumor area (mm3) was determined as

described [3;8].

Statistical analysis

The in vitro experiments were performed three times independently in triplicates. For

the in vivo model, number of animals was based on our previous calculations [3;8].

ANOVA followed by Tukey and Newman-Keuls post-hoc were used (Prism

Graphpad®). P<0.05 were assumed as significant.

Results

The efficacy and selectivity of TMZ/curcumin combinations

We first determined the effect of TMZ on viability of glioblastoma cells lines harboring

different mutations. C6 (p53wt/PTENmut/p16del) showed higher sensitivity to TMZ than

U251MG (p53mut/PTENnull/p16del) and U87MG (p53wt/PTENmut/p16del) (fig. 1A). As we

previously validated curcumin efficacy in vitro and in vivo [3], we were expecting that

curcumin would sensitize glioblastomas to TMZ. However, viability assays showed that

curcumin effects were additive instead of synergistic with TMZ (fig. 1A). Noteworthy,

co-treatments were harmless to astrocytes, indicating a selective toxicity to

transformed glia (fig. 1A).

Up to 48 h, TMZ and curcumin-treated cultures showed a decreased

density/numbers, but none significant PI incorporation was observed (Fig. 1A top-right

and 1B). Time-course experiments showed that decreases in cell proliferation (counting

assays) preceded toxicity (PI uptake), suggesting growth inhibition as an early step (fig.

1A, top right). On the other hand, pronounced changes in cell morphology and PI

incorporation were observed at high curcumin (≥30 uM for glioblastomas; ≥60 uM for

astrocytes) (fig. 1B), thus making co-treatments with TMZ unfeasible due to toxicity. In

addition, clonogenic survival assays indicated that a 72 h incubation with

TMZ/curcumin could lead to prolonged growth inhibition of surviving clones compared

to single agents (fig. 1C).

The therapeutic benefit of TMZ depends on its ability to methylate DNA at N-7

or O-6 positions of guanine, which leads to single and double-strand breaks and

activation of DDR (DNA damage response). Curcumin, and TMZ as expected, caused

significant amount of DNA damage, and TMZ/curcumin co-treatments produced higher

levels of comet index (fig. 1D). Phosphorylation of the DNA double-strand breaks

sensors H2AX and ATMSer1981 were readily detectable in curcumin and TMZ mono-

treatments, and no further increase was obtained in co-treatments. Based on the non-

synergistic patterns of viability changes and kinetic of cell death with this first

experimental set, we hypothesized that either curcumin/TMZ use some redundant

mechanism of cell growth inhibition or cells redundantly respond to these chemicals.

Thus, hereafter we sought to determine the mechanisms whereby cells resist to

TMZ/curcumin combinations.

G2/M checkpoint activation is an early step of glioblastomas response

TMZ is described to cause G2/M arrest in glioblastomas [12]. Therefore, we asked

whether TMZ-induced changes in the cycle could be modulated by curcumin co-

treatments. The results showed that both curcumin and TMZ redundantly induced

accumulation of cells in G2/M, and co-treatments were able to increase even more the

percentage of cells in G2/M, mainly in U251MG and C6 (fig. 2A). Western blots showed

that diverse controllers of the G2/M checkpoint were affected by curcumin and TMZ.

Phosphorylation of WeeSer642, Cdc2Tyr15, CHK1Ser345 and Cdc25cSer216 were induced by

curcumin and TMZ alone, and co-treatments had minor further induction; except

phospho-CHK1 levels whose activation was higher in co-treatments (Fig. 2C).

Corroborating, phosphorylation of cyclin B1 and cyclin D1 decreased in drug

combination-independent manner (Fig. 2C). Extending drug incubations for longer

periods (96 h in C6 and U251MG, and 120 h in U87MG), we were able to detect sub-

G1 apoptotic phenotypes across the cell lines (Fig. 2B). These findings confirm fig. 1 to

show that cell cycle arrest and inhibition of proliferation preceded cell death, and

redundant mechanisms ending up in G2/M checkpoint took place in co-treatments.

Autophagy precedes apoptosis in TMZ and curcumin-treated cells

Different glioblastomas displayed differing kinetics to succumb to apoptosis. In C6,

apoptosis was detected from 48 h incubation whereas U251MG and U87MG apoptosis

was only detectable after 96 h, and in a lower magnitude (fig. 3A and B). The inhibitory

effect of the pan-caspase inhibitor Z-VAD-fmk suggests a caspase-mediated

mechanism (fig. 3B). Caspase-3/7 activity was readily detectable in TMZ and curcumin-

treated C6 and U251MG with higher levels in co-treatments; an effect less pronounced

in U87MG (fig. 3B, bottom right). As a control, the proteasome inhibitor MG132 induced

pronounced caspase-3 activation [13], suggesting that the limited effect of curcumin

and TMZ upon apoptosis was not related to an intrinsic dysfunction of cells to trigger

apoptosis, but likely to activation of protective mechanisms.

Then we decided to test whether cell resistance could be attributed, at least in

part, to autophagy. In this intent, we performed FACS to detect acridine orange stained

autophagosomes, which indicate efficient autophagosome formation. We observed that

both curcumin and TMZ promoted autophagy, which was even more pronounced in co-

treatments (Fig 3C and D). Autophagy occurred earlier than apoptosis, being evident

as soon as 24 h treatment across the cell lines (Fig. 3C). Curcumin and TMZ increased

the conversion of LC3B-I to LC3B-II and induced ATG7 – two proteins typically

involved in vacuolar formation (fig. 3F). LC3B vacuolar compartmentalization was

confirmed by immunofluorescence in glioblastoma cells (Fig. 3E), and IHC of

glioblastoma multiforme specimens from patients treated with TMZ (Fig. 3G), indicating

that it is also a phenomenon in TMZ clinics. Markers of endoplasmic reticulum (ER)

stress - which is an upstream inducer of autophagy by TMZ in glioblastomas [14;15] -

as the ER chaperone GRP78/BIP and CHOP up-regulated in curcumin/TMZ treatments

independent on doses or drug combination (fig. 3F).

Autophagy inhibition abrogates G2/M arrest and favors apoptosis

Figure 4A shows that TMZ/curcumin-induced autophagy was blocked by ATG7

knockdown and class III PI3K inhibitor 3-MA, which inhibit the early steps of vacuole

formation (see knockdown validation in Fig. S1). Interestingly, inhibition of autophagy

potentiated both curcumin- and TMZ- as well as combinations-induced decreases in

viability, and enhanced LDH leakage (fig. 4B). This cytotoxicity seems to be attributed

to a switch from autophagy toward apoptosis as the levels of caspase-3/7 activity and

Annexin-V positive cells increased with ATG7 knockdown and 3-MA (fig. 4B). Similar

patterns were observed when cells were treated with the late autophagy inhibitor

chloroquine (15 uM). Chloroquine ≥ 20 uM was particularly difficult to manage due to

high toxicity; probably attributed to lysosomal dysfunction as already described in

glioblastomas [16]. Autophagy inhibition increased Bax, cleaved-caspase-3 and

cleaved-PARP immunocontents (Fig. 4D). Interestingly, blocking of autophagy

attenuated G2/M arrest as determined by cell cycle analysis, and decreased p-

WeeSer642, p-Ccd2tyr15 and p-Cdc25C. While C6 cells arrested in G1/S, U251MG

arrested in the S-phase in the presence of 3-MA (fig 4C); the levels of p-Rb increased

under these experimental conditions (fig. 4D). When cells were treated with

curcumin/TMZ for 24 h and then post-treated with 3-MA, it was not sufficient to block

G2/M arrest (data not shown), showing that early autophagy is someway coordinated

with G2/M checkpoint activation. We also observed that modulating autophagy in vivo

with chloroquine (CQ) - which was already shown to block oncolytic adenovirus dl922-

947 [17] and PI3K/mTOR dual inhibitor NVP-BEZ235-induced autophagy in

glioblastomas in vivo [18] – improved the efficacy of TMZ and TMZ/curcumin in C6

brain implanted rats (fig. 4E). Total inhibition of tumors was not achieved in CQ co-

treatments, probably due to tumor acidic pH-mediated CQ inactivation [19], and/or

restrictions imposed by BBB.

ERK1/2 activation is in the overlap of autophagy induction by curcumin and TMZ.

DDR via ATM and the mismatch repair protein MSH6 were recently described to

mediate TMZ-induced autophagy in LN229 and U87MG cells [20], Also, MAPKs

modulate either TMZ chemoresistance or/and G2/M arrest [21;22], but few is known

regarding TMZ-induced autophagy. Likewise, we and others have reported that

curcumin blocks glioblastoma relevant pathways as PI3K/Akt, NFκB and STAT3.

Based on that, we screened the contribution of these pathways on

autophagy/apoptosis balance in TMZ/curcumin co-treatments. Besides activation of

DDR in figure 1D, TMZ stimulated JNK1/2 and p38 (fig. 5A). Curcumin increased

neither basal nor TMZ-induced p38 nor JNK1/2 phosphorylation. Silencing of MSH6

and ATM (siRNAs) and p38 and JNK1/2 inhibition (with SB203508 and SP600129,

respectively) caused switch from autophagy to apoptosis in TMZ and TMZ/curcumin

treatments, but not curcumin alone, indicating a dominant role of the DDR members

and JNK/p38 on TMZ-induced autophagy (fig. 5C)

Curcumin blocked both basal and TMZ-induced p-p65 and NFκB DNA-binding

activity. STAT3 phosphorylation did not alter following TMZ but curcumin itself

decreasedp-STAT3 and STAT3 DNA-binding activity (fig 5A). By using specific

inhibitors, we observed that MAPKs and PI3K inhibition did not cause apoptosis or

autophagy itself, but decreased cell viability via inhibition of proliferation (Fig. 5B and

fig. S1) whereas NFκB and STAT3 inhibitors (BAY117082 and Stattic, respectively)

promoted autophagy, apoptosis and massive toxicity (Fig. 6C), which recapitulated the

mechanisms whereby curcumin affected glioblastomas.

Amidst the tested, ERK1/2 seems to be one of the overlapping mechanisms as

both TMZ and curcumin and its co-treatments stimulated ERK1/2 (fig. 5A). We also

took attention to PI3K/Akt - which is constitutively active and putative therapeutic target

in glioblastomas – which was downregulated by curcumin (fig. 5A). We found that

MEK/ERK and PI3K/Akt inhibition with UO126 and LY294002, respectively, attenuated

curcumin-, TMZ- and TMZ/curcumin-induced autophagy (fig. 5C and E) and enhanced

apoptosis (fig. 5G) as well as caspase-3 cleavage/activation and bax expression (fig.

5H). ERK1/2 and PI3K/Akt seem to participate in the early steps of autophagy as UO

and LY inhibited LC3B conversion to LC3B-II, an early step in vacuole formation (fig.

5H). Agreeing with 3-MA data (fig. 4), blocking of early autophagy with UO126 and

LY294002 was associated with inhibition of G2/M checkpoint as determined by FACS

and p-Wee, p-Cdc2 and p-Cdc25C western blots (fig. 5H).

When cells treated for 3 days with TMZ/curcumin plus inhibitors were kept for

additional 3 days to recover, the chemicals that inhibited autophagy also made cells

more susceptible to TMZ/curcumin (fig. 5D). Reminiscent cells were detected in every

treatment at 6 days, indicating that resistant clones or long-term processes like

senescence were also taking place (fig. S2)

Inhibition of ERK1/2-dependent autophagy by resveratrol improves TMZ and

curcumin efficacy in vivo

In this last part, we aimed to apply the knowledge on how autophagy limits the efficacy

of TMZ/curcumin association in glioblastomas, and use it for therapeutic advantage.

Then we looked for other potential small-molecule inhibitors with ability to cross BBB

and modulate autophagy in glioblastomas. Substantial evidences show that resveratrol

(RSV), a non-toxic stilbenoid phenolic, is able to block TMZ-induced autophagy to

enhance apoptosis through inhibition of ERK1/2 phosphorylation in U87MG and

GBM8401 cells in vitro [23]. In contrast, a second study found RSV to cause autophagy

in U87MG and U138MG [24] and potentiate TMZ effects by blocking G2/M checkpoint

[25]. We observed that RSV synergized with TMZ and/or curcumin in short (fig. 6A)

and long-term viability assays (fig. 6B). Although [24] and [23] pointed to opposite

results of RSV on apoptosis and autophagy, our data revealed that this dual effect was

dose-related (fig. 6C). We found that low levels RSV (up to 15 uM) blocked TMZ,

curcumin and TMZ/curcumin-induced autophagy (Fig. 6D and E) and LC3B conversion

(fig. 6F), thus switching cell program to apoptosis (fig. 6E). In contrast, higher levels of

RSV (30-60 uM) induced autophagic and apoptotic subpopulations (fig. 6C). Low level

RSV also inhibited G2/M checkpoint and ERK1/2 phosphorylation (fig. 6D and F),

which emulated the 3-MA and UO126 effects previously observed. In this context, our

collaborators have performed complete pharmacokinetic settings to show that RSV (i.p.

and oral) reaches the brain of Wistar rats at pharmacologically active levels (1 to 5 ug/g

tissue) enough to attenuate Alzheimer-related dysfunctions [26;27], therefore

supporting this chemical to be associated with TMZ/curcumin in C6-brain implants.

In vivo, curcumin and TMZ mono-treatments decreased C6 brain implanted

tumors, and combination of both drugs did not elicit significant improvement in reducing

tumor sizes; neither the in vitro additive effects were observed in vivo. Co-regimen of

TMZ/curcumin plus RSV yielded significantly smaller tumors compared to TMZ/Curc

and TMZ/RSV (fig. 6G). IHC and western blot analysis showed that TMZ/curcumin

combination increased p-ERK1/2, LC3B and ATG7 in tumor tissues from C6 implants

(fig 6H), suggestive of autophagy in vivo. These alterations were reversed by RSV co-

treatment (fig. 6H), proving that RSV in vitro findings upon ERK and autophagy may be

feasible in vivo to improve efficacy.

Discussion

Even though TMZ has been the chemotherapeutic that achieved the best clinical

performance in glioblastoma patients, combined therapies aiming to improve TMZ

efficacy are still a pressing goal because patients frequently chemoresist and relapse in

the course of therapy. Given the efficacy of curcumin in a variety of in vitro and

preclinical models [3;4;5;6;7], we developed a series of experiments in order to

evaluate the usefulness of curcumin/TMZ combination in glioblastomas. In vitro,

TMZ/curcumin co-treatments were more cytotoxic than its respective mono-treatments,

although the effects were additive rather than synergistic. In vivo, combining curcumin

with TMZ did not translate in reduction of tumors compared to TMZ alone, and in vitro

investigations set out autophagy as a mechanism involved in such resistance. Both

curcumin and TMZ induced autophagy as an early-response, and TMZ/curcumin

combination instead of switching cells to apoptosis redundantly caused even more

autophagy.

Autophagy has been pointed as a mechanism of glioblastoma resistance to

xenobiotics, including TMZ [18;20;23]. In our model, autophagy occurred concomitantly

with G2/M arrest, and these events preceded apoptosis. Blocking of autophagy using

3-MA, chloroquine, ATG7 siRNA impaired cell viability, and increased in LDH leakage

as well as apoptotic markers, suggesting a protective role. Corroborating with our

findings in U251MG and C6, 3-MA potentiated TMZ toxicity by rendering LN229 and

U87MG cells to apoptosis [20]. Diverse pathways collaborated to promote autophagy.

TMZ, as an alkylating agent, activated DDR-dependent autophagy, which involved

DNA strand breaks sensors and mismatch repair proteins as ATM and MSH6, agreeing

with [20]. Curcumin acted through inactivation of NFκB, STAT3 and PI3K/Akt as a

primary mechanism. These pathways are constitutively up-regulated in glioblastomas

[2;3;6;28], and experiments with specific inhibitors pinpointed that interfering with these

signals is sufficient to trigger autophagy and apoptosis. Noteworthy, to STAT3 inhibition

was attributed part of curcumin antiglioblastoma activity in vivo [7].

Not exclusive to DDR, TMZ pro-autophagy also involved the three MAPKs.

JNK1/2 and p38 promote G2/M checkpoint and survival to TMZ [21;22;29]; here we

showed that JNK/p38 also controls autophagy. Although crosstalks between DDR and

MAPKs consist a fruitful question, we focused to depict some mechanism which made

curcumin to fail in switching TMZ-treated cells to apoptosis. ERK1/2 pinpointed as an

overlap component of TMZ/curcumin-induced autophagy as both drugs activated

ERK1/2. MEK/ERK1/2 axis regulated the early steps of autophagy as UO126

attenuated LC3B conversion, one of the first steps of autophagosome formation [30].

Although we were interested in optimizing TMZ/curcumin combination, a previous

publication reported that curcumin (40 uM) stimulated ERK1/2-dependent autophagy in

vitro, and decreased tumors in flank-implanted U87MG [6]. TMZ alone was recently

described to cause ERK1/2-dependent autophagy in U87MG in vitro and in vivo [23]. In

our model, TMZ/curcumin co-treatments increased ERK1/2 phosphorylation as well as

LC3B conversion and ATG7 in brain-implanted C6 tumors, which carry the advantage

of parse out immune anti-tumor responses [31]. This support that TMZ/curcumin-

induced ERK1/2-dependent autophagy is not exclusive of in vitro conditions. CQ and

RSV experiments showed that autophagy could be targeted as a mean to achieve

better therapeutic efficacy in vivo. Also, the potentiation of both TMZ and

TMZ/curcumin efficacy through inhibition of ERK1/2 and autophagy set out novel

mechanisms of RSV, which is well-tolerated and supported by extensive literature,

including clinical trials [32]. Blocking of autophagy inhibited G2/M arrest in our model.

Although it is subject of more in-depth investigation, there is some evidence that

autophagy inhibition blocked G2/M arrest caused by artesunate in MCF7 and MDA-

MB231 breast cancers [33]. ATG7 silencing, 3-MA or bafilomycin A1 disrupted AO-

1012-induced S-phase arrest, leading to caspase-9 and PARP cleavage [34].

Despite the yet described preclinical efficacy of curcumin and the clinical

usefulness of TMZ to treat human glioblastomas, our data predict that TMZ and

curcumin synergism is unlike to be achieved due to redundant mechanisms leading to

protective autophagy. Autophagy may be modulated to re-sensitize cells to TMZ and

curcumin in vitro and, mainly, in vivo. Extending to broader landscapes, our findings

support that small-molecule autophagy inhibitors able to cross BBB could optimize

TMZ therapy.

Acknowledgements: This study was supported by Brazilian funding agencies:

CAPES; FAPERGS (PqG 12/1060-6); Conselho Nacional de Desenvolvimento

Científico e Tecnológico (CNPq; Projeto Universal 485758/2013-0 and 470973/2012-

9); A. Zanotto-Filho was recipient of a DOCFIX grant (Edital CAPES/FAPERGS n°

09/2012).

Figure legends:

Figure 1: (A) MTT experiments to determine the toxicity of TMZ and curcumin in

glioblastomas and astrocytes. Top right graph: Cell counting versus PI uptake in C6

and U251MG treated with 200 uM TMZ plus/or 15 uM curcumin (B) Representative

microphotographs (and PI uptake inserts) of TMZ/curcumin-treated glioblastomas at 48

h; 10x magnification. (C) Long-term clonogenic survival, and representative C6

colonies staining. (D) Comet assays of TMZ and curcumin-treated U251MG cells and

quantification of damage indexes. Bottom right: representative westerns showing H2AX

and ATM phosphorylation status in U251MG cells. If not specified, 72 h treatments

were performed. * Different from control (vehicle-treated); #different from control and

TMZ at equivalent TMZ doses.

Figure 2: Cell cycle analysis of (A) short-term 72 h and (B) long-term curcumin and

TMZ treatments. (C) Representative immunoblots showing the phosphorylation of cell

cycle regulatory proteins by curcumin and TMZ at 72 h.

Figure 3: (A) Annexin-V-FITC/PI assays showing the time-course of apoptosis in

glioblastoma cells. (B) Representative Annexin-V/PI FACS and quantification in C6 and

U251MG cells. Caspases-3/7 activity is also shown. (C) Time-course quantification of

autophagy and; (D) Representative FACS of acridine orange (AO) staining.(E)

Immunofluorescence showing LC3B-positive vacuoli in U251MG. (F) Western blots

showing the immunocontent of autophagic (ATG7 and LC3B) and ER stress (GRP78

and p-PERK) proteins in U251MG. (G) IHC for LC3B and Ki67 (proliferation marker) in

tumor specimens from glioblastoma patients. 200 uM TMZ, 15 uM curcumin and 5 uM

MG132 were used. If not specified, 72 h treatment was performed. * Different from

control (vehicle treated); #different from control and TMZ at equivalent TMZ doses.

Figure 4: (A) Validation of 3-MA and ATG7 siRNA effect on autophagy of U251MG

cells (AO staining). (B) MTT, LDH leakage, caspase-3/7 activity and Annexin-V/PI

FACS showing the impact of autophagy inhibitors on viability and apoptosis of

U251MG. (C) Cell cycle distribution in C6 and U251MG cells, and (D) immunoblots for

cell cycle and apoptotic markers in U251MG in the presence/absence of 3-MA. (E)

Effect of CQ on C6 tumors growth in TMZ and TMZ/curcumin-treated rats. Legends: 3-

MA (4 mM), CQ (chloroquine, 15 uM in vitro); ATG7si (ATG7 siRNA). Data are

expressed as average ± SD. * Asterisks denote signification for indicated comparisons.

Figure 5: (A) Phosphorylation status of p65-NFκB, STAT3 and Akt, and ELISA for

NFκB and STAT3 DNA-binding activity in nuclear extracts of TMZ and/or curcumin-

treated U251MG (24 h). (B) Effect of MAPK, PI3K/Akt, JAK/STAT3 and IKK/NFκB

inhibitors on viability (MTT), autophagy and apoptosis of U251MG at 72 h. (C-D) Effect

of various inhibitors on (C) autophagy and apoptosis and (D) long-term survival of

TMZ/curcumin-treated U251MG cells. FACS assays showing (E) AO staining, (F) cell

cycle distribution, (G) Annexin-V-FITC/PI staining, and (H) western blots for cell cycle

and apoptotic effectors; caspase-3/7 activity in TMZ/curcumin-treated U251MG cells in

the presence of UO126 or LY294002 (72 h). UO126 (UO), SP600129, SB208503 and

LY294002 (LY) at 20 uM, and 5 uM BAY117082 and 1 uM Stattic were used.*Asterisks

denote signification for indicated comparisons or different from control cells.

Figure 6: (A) MTT and (B) long-term clonogenic assays showing the sensitizing effect

of RSV in U251MG. (C) Dose-effect of RSV-induced autophagy and apoptosis in

U251MG. (D-E) Low levels RSV (15 uM) blocks TMZ/curcumin-induced autophagy and

G2-M arrest in U251MG cells. (F) Immunoblots showing RSV effect on the

phosphorylation cdc2, Wee, ERK1/2 and Akt, caspase-3 cleavage, and LC3B in

U251MG. U251MG were treated with 200 uM TMZ and 15 uM Curcumin; if not

specified, 72 h treatments were carried out. (G) C6 brain-implanted tumors size and (H)

IHC of C6 tumors showing the levels of p-ERK1/2 and LC3B in the presence/absence

of RSV. ERK1/2 forms, ATG7 and LC3B immunocontent in C6 brain tumor lysates is

also shown. *Different from control; #different from control, TMZ and TMZ/curcumin at

equivalent doses; asterisks also denote signification for indicated comparisons; a

Denotes the effect of RSV in inhibiting TMZ/curcumin effects on autophagy.

References:

1. Prasad S, Tyagi AK, Aggarwal BB. Recent developments in delivery,

bioavailability, absorption and metabolism of curcumin: the golden pigment from

golden spice. Cancer Res Treat 2014 Jan; 46(1):2-18.

2. Hasima N, Aggarwal BB. Cancer-linked targets modulated by curcumin. Int J

Biochem Mol Biol 2012; 3(4):328-51.

3. Zanotto-Filho A, Braganhol E, Edelweiss MI, Behr GA, Zanin R, Schröder R et

al. The curry spice curcumin selectively inhibits cancer cells growth in vitro and

in preclinical model of glioblastoma. J Nutr Biochem 2012; 23(6):591-601.

4. Du WZ, Feng Y, Wang XF, Piao XY, Cui YQ, Chen LC et al. Curcumin

suppresses malignant glioblastoma cells growth and induces apoptosis by

inhibition of SHH/GLI1 signaling pathway in vitro and vivo. CNS Neurosci Ther

2013; 19(12):926-36.

5. Perry MC, Demeule M, Régina A, Moumdjian R, Béliveau R. Curcumin inhibits

tumor growth and angiogenesis in glioblastoma xenografts. Mol Nutr Food Res

2010; 54(8):1192-201.

6. Aoki H, Takada Y, Kondo S, Sawaya R, Aggarwal BB, Kondo Y. Evidence that

curcumin suppresses the growth of malignant glioblastomas in vitro and in vivo

through induction of autophagy: role of Akt and extracellular signal-regulated

kinase signaling pathways. Mol Pharmacol 2007; 72(1):29-39

7. Weissenberger J, Priester M, Bernreuther C, Rakel S, Glatzel M, Seifert V et al.

Dietary curcumin attenuates glioblastoma growth in a syngeneic mouse model

by inhibition of the JAK1,2/STAT3 signaling pathway. Clin Cancer Res 2010;

16(23):5781-95.

8. Zanotto-Filho A, Coradini K, Braganhol E, Schröder R, de Oliveira CM, Simões-

Pires A et al. Curcumin-loaded lipid-core nanocapsules as a strategy to improve

pharmacological efficacy of curcumin in glioblastoma treatment. Eur J Pharm

Biopharm 2013; 83(2):156-67.

9. Zheng Q, Han L, Dong Y, Tian J, Huang W, Liu Z et al. JAK2/STAT3 targeted

therapy suppresses tumor invasion via disruption of the EGFRvIII/JAK2/STAT3

axis and associated focal adhesion in EGFRvIII-expressing glioblastoma. Neuro

Oncol 2014; 16(9):1229-43.

10. Cancer Genome Atlas Research Network. Comprehensive genomic

characterization defines human glioblastoma genes and core pathways. Nature

2008; 455(7216):1061-8.

11. Zanotto-Filho A, Braganhol E, Schröder R, de Souza LH, Dalmolin RJ, Pasquali

MA et al. NFκB inhibitors induce cell death in glioblastomas. Biochem

Pharmacol 2011; 81(3):412-24.

12. Hirose Y, Berger MS, Pieper RO. Abrogation of the Chk1-mediated G(2)

checkpoint pathway potentiates temozolomide-induced toxicity in a p53-

independent manner in human glioblastoma cells. Cancer Res 2001;

61(15):5843-9.

13. Zanotto-Filho A, Braganhol E, Battastini AM, Moreira JC. Proteasome inhibitor

MG132 induces selective apoptosis in glioblastoma cells through inhibition of

PI3K/Akt and NFkappaB pathways, mitochondrial dysfunction, and activation of

p38-JNK1/2 signaling. Invest New Drugs 2012;30(6):2252-62.

14. Shen S, Zhang Y, Wang Z, Liu R, Gong X. Bufalin induces the interplay

between apoptosis and autophagy in glioblastoma cells through endoplasmic

reticulum stress. Int J Biol Sci 2014;10(2):212-24

15. Lin CJ, Lee CC, Shih YL, Lin CH, Wang SH, Chen TH et al. Inhibition of

mitochondria- and endoplasmic reticulum stress-mediated autophagy augments

temozolomide-induced apoptosis in glioblastoma cells. PLoS One 2012;

7(6):e38706.

16. Geng Y, Kohli L, Klocke BJ, Roth KA. Chloroquine-induced autophagic vacuole

accumulation and cell death in glioblastoma cells is p53 independent. Neuro

Oncol. 2010; 12(5):473-81.

17. Botta G, Passaro C, Libertini S, Abagnale A, Barbato S, Maione AS et al.

Inhibition of autophagy enhances the effects of E1A-defective oncolytic

adenovirus dl922-947 against glioma cells in vitro and in vivo. Hum Gene Ther

2012; 23(6):623-34.

18. Cerniglia GJ, Karar J, Tyagi S, Christofidou-Solomidou M, Rengan R, Koumenis

C et al. Inhibition of autophagy as a strategy to augment radiosensitization by

the dual phosphatidylinositol 3-kinase/mammalian target of rapamycin inhibitor

NVP-BEZ235. Mol Pharmacol 2012; 82(6):1230-40.

19. Pellegrini P, Strambi A, Zipoli C, Hägg-Olofsson M, Buoncervello M, Linder S et

al. Acidic extracellular pH neutralizes the autophagy-inhibiting activity of

chloroquine: implications for cancer therapies. Autophagy 2014; 10(4): 562-71.

20. Knizhnik AV, Roos WP, Nikolova T, Quiros S, Tomaszowski KH, Christmann M

et al. Survival and death strategies in glioblastoma cells: autophagy,

senescence and apoptosis triggered by a single type of temozolomide-induced

DNA damage. PLoS One 2013; 8(1):e55665.

21. Hirose Y, Katayama M, Berger MS, Pieper RO. Cooperative function of Chk1

and p38 pathways in activating G2 arrest following exposure to temozolomide. J

Neurosurg 2004; 100(6):1060-5.

22. Ohba S, Hirose Y, Kawase T, Sano H. Inhibition of c-Jun N-terminal kinase

enhances temozolomide-induced cytotoxicity in human glioblastoma cells. J

Neurooncol 2009; 95(3):307-16

23. Lin CJ, Lee CC, Shih YL, Lin TY, Wang SH, Lin YF et al. Resveratrol enhances

the therapeutic effect of temozolomide against malignant glioblastoma in vitro

and in vivo by inhibiting autophagy. Free Radic Biol Med 2012; 52(2):377-91.

24. Filippi-Chiela EC, Villodre ES, Zamin LL, Lenz G. Autophagy interplay with

apoptosis and cell cycle regulation in the growth inhibiting effect of resveratrol in

glioblastoma cells. PLoS One 2011; 6(6):e20849.

25. Filippi-Chiela EC, Thomé MP, Bueno e Silva MM, Pelegrini AL, Ledur PF,

Garicochea B et al. Resveratrol abrogates the temozolomide-induced G2 arrest

leading to mitotic catastrophe and reinforces the temozolomide-induced

senescence in glioblastoma cells. BMC Cancer 2013; 13:147.

26. Frozza RL, Bernardi A, Paese K, Hoppe JB, da Silva T, Battastini AM et al.

Characterization of trans-resveratrol-loaded lipid-core nanocapsules and tissue

distribution studies in rats. J Biomed Nanotechnol 2010; 6(6):694-703.

27. Frozza RL, Bernardi A, Hoppe JB, Meneghetti AB, Matté A, Battastini AM et al.

Neuroprotective effects of resveratrol against Aβ administration in rats are

improved by lipid-core nanocapsules. Mol Neurobiol 2013; 47(3):1066-80

28. Dhandapani KM, Mahesh VB, Brann DW. Curcumin suppresses growth and

chemoresistance of human glioblastoma cells via AP-1 and NFkappaB

transcription factors. J Neurochem 2007; 102(2):522-38.

29. Carmo A, Carvalheiro H, Crespo I, Nunes I, Lopes MC. Effect of temozolomide

on the U-118 glioblastoma cell line. Oncol Lett 2011; 2(6):1165-70.

30. Gallagher LE, Chan EY. Early signalling events of autophagy. Essays Biochem

2013; 55:1-15.

31. Parsa AT, Chakrabarti I, Hurley PT, Chi JH, Hall JS, Kaiser MG et al.

Limitations of the C6/Wistar rat intracerebral glioblastoma model: implications

for evaluating immunotherapy. Neurosurgery 2000; 47(4):993-9.

32. Tomé-Carneiro J, Larrosa M, González-Sarrías A, Tomás-Barberán FA, García-

Conesa MT, Espín JC. Resveratrol and clinical trials: the crossroad from in vitro

studies to human evidence. Curr Pharm Des 2013; 19(34):6064-93.

33. Chen K, Shou LM, Lin F, Duan WM, Wu MY, Xie X et al. Artesunate induces

G2/M cell cycle arrest through autophagy induction in breast cancer cells.

Anticancer Drugs. 2014; 25(6):652-62

34. Liou JS, Wu YC, Yen WY, Tang YS, Kakadiya RB, Su TL et al. Inhibition of

autophagy enhances DNA damage-induced apoptosis by disrupting CHK1-

dependent S phase arrest. Toxicol Appl Pharmacol 2014; 278(3):249-58.