Cancer Letters 352 (2014) 187–195

Contents lists available at ScienceDirect

Cancer Letters

journal homepage: www.elsevier .com/locate /canlet

LKB1 inactivation sensitizes non-small cell lung cancerto pharmacological aggravation of ER stress

http://dx.doi.org/10.1016/j.canlet.2014.06.0110304-3835/� 2014 Elsevier Ireland Ltd. All rights reserved.

⇑ Corresponding author. Address: Norton Thoracic Institute, St. Joseph’s Hospitaland Medical Center, 445 N. 5th Street, Suite 110, Phoenix, AZ 85004, United States.Tel.: +1 (602) 406 8322; fax: +1 (602) 294 5261.

E-mail address: landon.inge@dignityhealth.org (L.J. Inge).

Landon J. Inge a,⇑, Jacqueline M. Friel a, Amanda L. Richer a, Aaron J. Fowler a, Timothy Whitsett b,Michael A. Smith a, Nhan L. Tran b, Ross M. Bremner a

a Norton Thoracic Institute, St. Joseph’s Hospital and Medical Center, Phoenix, AZ, United Statesb Cancer and Cell Biology Division, The Translational Genomics Research Institute, Phoenix, AZ, United States

a r t i c l e i n f o

Article history:Received 7 April 2014Received in revised form 5 June 2014Accepted 22 June 2014

Keywords:LKB1Lung cancerTreatmentUPRER stress

a b s t r a c t

Five-year survival rates for non-small cell lung cancer (NSCLC) have seen minimal improvement despiteaggressive therapy with standard chemotherapeutic agents, indicating a need for new treatmentapproaches. Studies show inactivating mutations in the LKB1 tumor suppressor are common in NSCLC.Genetic and mechanistic analysis has defined LKB1-deficient NSCLC tumors as a phenotypically distinctsubpopulation of NSCLC with potential avenues for therapeutic gain. In expanding on previous work indi-cating hypersensitivity of LKB1-deficient NSCLC cells to 2-deoxy-D-glucose (2DG), we find that 2DG hasin vivo efficacy in LKB1-deficient NSCLC using transgenic murine models of NSCLC. Deciphering of themolecular mechanisms behind this phenotype reveals that loss of LKB1 in NSCLC cells imparts increasedsensitivity to pharmacological compounds that aggravate ER stress. In comparison to NSCLC cells withfunctional LKB1, treatment of NSCLC cells lacking LKB1 with the ER stress activators (ERSA), tunicamycin,brefeldin A or 2DG, resulted in aggravation of ER stress, increased cytotoxicity, and evidence of ER stress-mediated cell death. Based upon these findings, we suggest that ERSAs represent a potential treatmentavenue for NSCLC patients whose tumors are deficient in LKB1.

� 2014 Elsevier Ireland Ltd. All rights reserved.

Introduction

Lung cancer, and more specifically non-small cell lung cancer(NSCLC), contributes significantly to cancer related mortalitieswithin the United States [1]. Mortality rates for NSCLC have seenlittle change over the past two decades, with relapse occurringwithin 5 years in 35–50% of patients [1,2]. The majority of NSCLCpatients undergo treatment with radiation and traditional chemo-therapeutic agents, such as alkylating agents or anti-metabolites,neither of which appreciably improve survival rates [3]. Applica-tion of genomic profiling technologies on NSCLC has led to targetedtherapies for patients whose NSCLC tumors carry either distinctmutations to EGFR or the EML4-ALK fusion oncogene [4]. However,the low incidence of these mutations within the NSCLC patientpopulation [4], especially within the US, prevents the larger major-ity of NSCLC patients from reaping the clinical benefits of thesecompounds, indicating a demonstrable need for alternative thera-peutic approaches.

STK11 (Serine/threonine-protein kinase 11) or Liver kinase B1(LKB1) was initially identified as the tumor suppressor generesponsible for the familial cancer disorder, Peutz-Jeghers syn-drome [5]. Subsequent analyses for sporadic mutations of LKB1within tumors found inactivating mutations occur at a high fre-quency in NSCLC subtypes, second in incidence to mutations tothe TP53 gene [6,7]. Bi-allelic deletion of LKB1 within geneticallyengineered mouse tumor models results in highly aggressive, met-astatic cancer cells [8–11]. Functionally, LKB1 phosphorylates andactivates a family of 14 kinases [5]. The most highly characterizedof LKB1 substrates is the 50-adenosine mono-phosphate activatedprotein kinase (AMPK) [5]. Activated by LKB1 during low ATP con-ditions, AMPK interacts with a variety of proteins to alter cellmetabolism and inhibit cell growth [5]. Accordingly, substantialdata within in vitro and in vivo systems has revealed LKB1 inactiva-tion results in alterations to cellular metabolic functions [11–14].These observations have clinical potential for treating LKB1-defi-cient NSCLC, as pharmacological targeting of altered metabolicstates in LKB1-deficient cells have been shown to have pre-clinicalefficacy and identification of targetable characteristics of LKB1-deficient NSCLC are ongoing [11,13,15].

One of the most basic requirements of all cells is the synthesisof new proteins [16]. Within eukaryotes, a portion of protein

188 L.J. Inge et al. / Cancer Letters 352 (2014) 187–195

synthesis occurs within the Endoplasmic Reticulum (ER). Stimuli,such as nutrient deprivation or hypoxia, perturb ER functionresulting in accumulation of unfolded/misfolded proteins withinthe ER lumen, commonly referred to as ER stress [16]. ER stressactivates a cascade of actions, called the Unfolded Protein Response(UPR), aimed at either restoring function to the ER or induction ofprogrammed cell death [16,17]. Although the specific mechanismsthat determine and regulate these two opposing fates are stillbeing elucidated, it is understood that the extent and severity ofER stress are integral to determining whether the UPR activatespro-survival or apoptotic functions [17–19]. Using a variety ofmodel systems, studies indicate it is extreme, irremediable levelsof ER stress that activate UPR-mediated cell death pathways, whilemoderate or low-level ER stress directs UPR functions towardsadaptation and survival [20]. Concordantly, tumor cells utilizethe cyto-protective functions of the UPR to adapt and survivelow-level ER stress induced by the hypoxic, hypoglycemic andacidic tumor microenvironment [16,21]. Evidence of ER stress/UPR activity is frequently observable within tumors and phenotyp-ically differentiates tumor cells from the majority of normal cells,whose microenvironment rarely necessitates engagement of theUPR [21]. This characteristic provides a point for therapeutic inter-vention, as a variety of studies have shown that utilizing pharma-cological compounds that perturb ER function to aggravate theexisting ER stress conditions within tumor cells, subsequentlyengages the UPR’s apoptotic functions [22]. Clinically, this has beenfound to be the mechanism of action for the proteasome inhibitor,bortezomib in multiple myeloma [23] and significant efforts areunderway to further develop this treatment avenue [22].

In the present study, we find that the glucose analog, 2-D-Deoxyglucose (2DG) has efficacy in an in vivo model of KRas/LKB1-deficient NSCLC. In investigating the molecular mechanismsof cellular toxicity of 2DG, we find LKB1-null NSCLC cells are moresensitive to aggravation of ER stress by pharmacological com-pounds that perturb ER function, compared to NSCLC cells express-ing LKB1. Absence of LKB1 in NSCLC cells correlates with decreasedcell survival, evidence of increased UPR-mediated apoptosis andincreased levels of reactive oxygen species. Collectively, these find-ings suggest that targeted aggravation of the UPR in LKB1-deficientNSCLC may be a promising therapeutic approach.

Materials and methods

Cell culture and reagents

A427, H23, H2126, H838, H1395, H2009, H358, H1975 and H441 NSCLC celllines were obtained from ATCC and maintained in RPMI 1640 (Invitrogen) understandard tissue culture conditions. Antibodies to phosphorylated AMPK, AMPK,LKB1, GRP78/BiP, CHOP, cleaved PARP, cleaved Caspase 9, phosphorylated eif2a,eif2a were all purchased from Cell Signaling Technologies. 2-D-Deoxyglucose(2DG, Sigma) was diluted to a 1M stock in sterile phosphate buffered saline. Tunica-mycin (Tm), brefeldin-A (BFA), celecoxib and bortezomib were purchased fromTocris Bioscience (MS, USA) and diluted in DMSO. The pBabe retroviral constructscontaining FLAG-tagged full length LKB1 and the kinase dead LKB1 (KDLKB1) orig-inated in Dr. Lewis Cantley’s laboratory (Harvard Medical School, MA) and wereobtained from Addgene.org (Cambridge, MA). The isogeneic H23-KDLKB1, H23-LKB1, A427-LKB1, A427-KDLKB1, H838-LKB1 and H838-KDLKB1 cell lines werederived using methods described previously [15]. CellROX™ Green reagent waspurchased from Life Technologies. Cyclohexamide (CHX) was purchased from Sel-leck and diluted in DMSO.

Cell lysis and Immunoblotting

Cells were washed once with ice-cold PBS and incubated on ice for 30 min witha lysis buffer containing 10 mM Tris–HCl, 150 mM NaCl, 0.1% SDS and 1% IGEPAL(Sigma) to which a protease inhibitor cocktail (Sigma) and two phosphatasecocktails I and II (Sigma) were added immediately before use. Lysates were trans-ferred to tubes and insoluble material was pelleted by centrifugation. For immuno-blotting, 50–100 lg of total protein lysates were separated by SDS–PAGE andtransferred to nitrocellulose. Blots were blocked in 5% milk/Tris buffered saline/

0.1% Tween 20 (TBST) and primary antibodies were diluted in 5% Bovine SerumAlbumin (Sigma)/TBST and incubated overnight at 4 �C. Blots were developed withECL plus (GE) and visualized on a Kodak Image station.

ATP measurement/cell viability

500 H838 or A427 cells stably expressing either LKB1 or KDLKB1 were plated in96 well plates. The following day, cells were dosed with the indicated treatmentsfor 48 h. Cellular ATP levels were assessed using the CellTiter-Glo Luminescent CellViability Assay substrate (Promega) according to manufacturer’s instructions. Lumi-nescence was read using the DTX 880 Multimode Detector. Cellular ATP/cell viabil-ity was calculated relative to vehicle treated control.

Flow cytometry of ROS levels

To analyze levels of reactive oxygen species (ROS), 1 � 106 H838-KDLKB1 andH838-LKB1 cells were plated and treated with the indicated drugs. Cells were thenstained with 0.5 lM CellRox™ for 30 min at 37 �C, collected by trypsinization andwashed with PBS. Following fixation with 2% Paraformaldehyde/PBS for 20 min,cells were brought to a final volume of 500 ll in PBS. CellRox™ staining of ROSwas assessed using the Accuri C6 Flow Cytometer (BD Biosciences, San Jose, CA,USA) with the threshold set at 80,000. A total of 30,000 events were collected persample and ROS staining intensity determined using FCS Express 4 Flow Cytometersoftware. Background events were gated out using unstained control. Histogramplots were overlaid and histogram subtraction was performed by subtracting theintegral of the stained control from the treated sample.

Murine transgenic models of KRas and KRas/LKB1 null NSCLC

KRasG12D/LKB1lox/lox (KRas/LKB1null) were generated by selective breeding ofLox-Stop-Lox KRasG12D mice [24], purchased from Jackson Laboratories (Bar-Harbor,ME), with LKB1lox/lox mice obtained from Mouse Repository at the NCI and describedpreviously [25] and were inbred on a FVB background. KRasG12D/LKB1wt/wt (KRas/LKB1wt) mice were generated by backcrossing Lox-Stop-Lox KRasG12D onto a FVBbackground. Induction of lung tumors was performed as described previously[9,11]. Briefly, six-week-old KRas/LKB1null and KRas/LKB1wt mice were transientlyinfected with 5 � 106 p.f.u. of Cre adenovirus (University of Iowa adenoviral core)via intranasal infection. At �6 weeks (KRas/LKB1null) or �10 weeks (KRas/LKB1wt)post- infection, mice underwent a baseline MRI scan to determine the baselinetumor burden. Mice with identified disease were randomized for daily treatmentwith either vehicle (PBS) or 500 mg/kg 2DG for three weeks, at which time micereceived a followup MRI scan, sacrificed and the presence of disease confirmed bygross assessment and histology. For data generated in Fig. 2, KRas/LKB1null andKRas/LKB1wt mice with MRI positive tumors were treated either with 2DG(500 mg/kg) or vehicle (PBS) twice a day for two days before sacrifice and necropsyof tumor tissue. Mice received daily checks to monitor for clinical signs of disease(labored breathing, weight loss) and all procedures were under a protocol approvedby the St. Joseph’s Hospital and Medical Center IACUC committee.

MRI imaging

All the scans were performed using a 7T small animal, 30-cm horizontal-boremagnet and BioSpec Advance III spectrometer (Bruker, Billerica, MA) with a72 mm quadrature volume coil. Each animal was induced and maintained underisoflurane anesthesia, 3.0% and 1.5%, respectively, and delivered with 1.5 L/min of100% oxygen. During MRI scans, the animal’s respiration was continually monitoredby a small animal monitoring and gating system (SA Instruments, Stoney Brook, NY)via a pillow sensor positioned on top of the abdomen. Mice were placed on a heatedanimal bed system (Bruker, Billerica, MA) and the normal body temperature (36–37 �C) was maintained. For MR imaging, a Fast Low Angle Shot (FLASH) sequence(TR/TE = 40 ms/2 ms; FOV = 60 mm � 60 mm; Average = 2; matrix = 256 � 256; FlipAngle = 30�) was used to acquire initial scout images, which were used for locatingthe lung region for subsequent images. Then T1 weighted images were obtainedwith the Fast Imaging with Steady State Precession (FISP) sequence (TR/TE5.262 ms/2.0 ms; 8 segments; FOV = 30 mm � 30 mm; matrix = 256 � 256; slicethickness = 1.0 mm; Flip angle, 20�; 22 averages; total acquisition time = 17 min).During the scan, motion artifacts were suppressed with multiple averages and res-piration triggering. The images were saved in analyze format, and the further imageprocessing and tumor volume measurements were performed with ImageJ(National Institutes of Health, Bethesda, Maryland, USA), using 17 sequential imagescans. Change in tumor volumes was determined by comparison of baseline tumorvolume to tumor volume following three weeks of treatment.

Clonogenic assay

Clonogenic survival assay were performed in 6 well dishes. Briefly, 200 of indi-cated cells were plated per well and allowed to attach overnight. Were indicatedcells were pre-treated with 125 ng/ml of cyclohexamide (CHX) for one hour. Indi-cated drugs were diluted fresh into media and cells were treated for 24 h, at which

L.J. Inge et al. / Cancer Letters 352 (2014) 187–195 189

time the drug containing media was removed and replaced with fresh media. After14 days, cells were fixed in 10% Methanol/10% Acetic Acid and stained with 0.5%crystal violet. After washing in ddH20, plates were allowed to dry and colonies of50 cells or more were counted. The surviving fraction was calculated according tomethods described by Franken et al. [26].

Immunohistochemical staining

Lungs from either KRas/LKB1null or KRas/LKB1wt mice with identified tumor andtreated with 2DG (500 mg/kg, two doses over a 24 h period) were fixed overnight in10% neutral buffered formalin and paraffin embedded using routine procedures.FFPE samples of KRas/LKB1null and KRas/LKB1wt NSCLC tumors were sectioned usingstandard procedures and adhered to charged microscope slides. Five lM sectionsunderwent heat induced epitope retrieval and immunohistochemical (IHC) stainingof tissue was performed using previously published procedures [27]. Antibodies topAMPK, LKB1 and BiP were purchased from Cell Signaling. Slides were scannedusing the Aperio system (Leica Biosystems, Buffalo Grove, IL) and images collectedat a magnification of 20�.

Results

2DG displays improved pre-clinical efficacy in vivo using a transgenicmodel of KRas/LKB1-deficient NSCLC

2-deoxy-D-glucose (2DG), a non-hydrolyzable analog of glu-cose, has therapeutic activity in several tumor types and hasattracted attention as a possible chemotherapeutic [28,29]. Previ-ously, we found that lack of LKB1 activity sensitized NSCLC cellsto 2DG-mediated cytotoxicity in vitro [15]. Based upon these find-ings, we assessed the role of LKB1 in modulating the effects of 2DGin NSCLC in vivo using two well-characterized transgenic models ofNSCLC (KRas/LKB1wt and KRas/LKB1null) [9,11,24,30]. Based uponprevious studies [11,30], KRas/LKB1wt and KRas/LKB1null miceunderwent an MRI scan to determine baseline tumor burden ateither 10 weeks (KRas/LKB1wt) or 6 weeks (KRas/LKB1null) post-infection with adeno-Cre (Fig. 1A). Animals with identified tumorwere randomized for treatment with vehicle (PBS) or 2DG at500 mg/kg, a dose shown to have limited efficacy in vivo [29,31]and re-imaged after three weeks of treatment to determine thechange in tumor volume. Treatment with 2DG reduced the growthof NSCLC tumors in both KRas/LKB1wt and KRas/LKB1null mice,compared to vehicle (Fig. 1A and B) after three weeks of treatment.However, KRas/LKB1null tumors displayed a statistically greaterresponse to 2DG (p = 0.0032), compared to KRas/LKB1wt tumors(Fig. 1B). Despite reducing the growth of NSCLC tumors, we didnot observe any effects upon metastasis (a reported feature ofthe KRas/LKB1null model [9]) under gross assessment at necropsyin KRas/LKB1null mice with 2DG treatment (data not shown).

Fig. 1. 2DG preferentially reduces the growth of LKB1-deficient NSCLC tumors. (A)Representative MRI images of NSCLC tumors in KRas/LKB1wt and KRas/LKB1null

transgenic mice. KRas/LKB1wt and KRas/LKB1null transgenic mice underwent MRIimaging as described in Material and Methods to determine baseline and treated(vehicle, 2DG) tumor burden. White arrows indicated tumors encircled by red line.(B) Volumetric analysis of tumor volume in KRas/LKB1wt and KRas/LKB1null NSCLCtumors treated with vehicle (PBS) or 2DG (500 mg/kg, SID). Collected MRI imageswere analyzed as described in Materials and Methods to determine the change involume (baseline vs 3 weeks of treatment). Graphs depict the mean change involume ± SE of 5 mice in each group. P = 0.0032.

2DG induction of BiP/GRP78 is LKB1 dependent

We next investigated the effects of 2DG treatment in KRas/LKB1wt and KRas/LKB1null mice by performing IHC staining forphosphorylated AMPK (pAMPK) and glucose response protein 78(HSPA5/GRP78) or BiP, a protein shown to be up-regulated inKRas/LKB1wt and KRas/LKB1null NSCLC tumors in response to met-abolic stress [11]. Tumors from KRas/LKB1wt and KRas/LKB1null

mice were also stained for LKB1 to confirm Cre-mediated deletionof the LKB1 gene. LKB1 (Fig. 2B) and pAMPK (Fig. 2D and F) wereabsent from KRas/LKB1null tumors, while staining of both LKB1(Fig. 2A) and pAMPK (Fig. 2C and E) were observable in NSCLCtumors from KRas/LKB1wt mice. IHC staining of BiP revealed thattreatment with 2DG resulted in a significant increase in BiP proteinlevels in NSCLC tumor cells of KRas/LKB1wt mice (Fig. 2G and I), butnot in tumors from KRas/LKB1null mice (Fig. 2H and J).

Fig. 2. Immunohistochemical staining of KRas/LKB1wt and KRas/LKB1null NSCLCtumors. (A and B) LKB1 staining of KRas/LKB1wt and KRas/LKB1null NSCLC tumors.Arrows (B) indicate LKB1 positive stromal cells. (C–F) Phosphorylated AMPKthr172

(pAMPK) staining in KRas/LKB1wt and KRas/LKB1null NSCLC, treated with vehicle (Cand D) or 2DG (E and F) for two days (500 mg/kg, BID). (G–J) BiP/GRP78 staining inKRas/LKB1wt and KRas/LKB1null NSCLC, treated with vehicle (G, H) or 2DG (I, J) fortwo days (500 mg/kg, BID). Magnification = 20�.

190 L.J. Inge et al. / Cancer Letters 352 (2014) 187–195

The anti-tumorigenic properties of 2DG have been attributed toaggravation of existing ER stress in tumor cells via inhibition ofglucose dependent protein glycosylation [32]. Escalation of BiP isreflective of induction of ER stress [33], leading us to investigatethe ER stress response in LKB1-deficient and LKB1-expressingNSCLC cells in response to 2DG. LKB1 expressing H2009 andLKB1-deficient H23 NSCLC cells [15] were treated with 2DG andimmunblotted for pAMPK and BiP/GRP78. Lysates were also immu-noblotted for CHOP/GADD153 (C/EBP homologous protein/DDIT3),

a protein whose expression is increased during both ER and meta-bolic stress [34–36]. Treatment with 2DG (20 mM) resulted inphosphorylation of AMPK in LKB1 expressing H2009 cells, butnot LKB1-deficient H23 cells (Fig. 3A) and apoptosis in H23 cells(supplemental Fig. 1A). In concordance with our in vivo staining(Fig. 2G–J), 2DG increased BiP in LKB1 expressing H2009 cells,while LKB1-deficient H23 cells displayed a more moderate increase(Fig. 3A). Protein levels of CHOP were also found to increase inLKB1 expressing H2009 cells upon treatment with 2DG, but notLKB1-deficient H23 (Fig. 3A). We expanded our analysis to addi-tional LKB1 expressing (H358, H1975, H441) and LKB1-deficient(H2126, H838, H1395) NSCLC cell lines in order to determine ifdecreased CHOP and BiP expression in response to 2DG persistedin the absence of LKB1. LKB1-deficient H2126, H838 and H1395NSCLC cells displayed minimal increases in CHOP and BiP with2DG treatment after 12 h of treatment and increased apoptosis(cleaved PARP) after 18 h of treatment (Fig. 3B). The lack of LKB1resulted in a failure to phosphorylate AMPK in response to 2DG(Fig. 3B). LKB1 expressing H358, H1975 and H441 cells displayedan opposite phenotype, with 2DG inducing AMPK phosphorylationand increasing CHOP and BiP protein levels (Fig. 3B). Stable expres-sion of LKB1, but not a kinase-dead LKB1 (KDLKB1) in H23 NSCLCcells resulted in increases in BiP and CHOP protein expression,AMPK phosphorylation (Fig. 3C) and improved survival in responseto 2DG treatment (supplemental Fig. 1B). To confirm the role ofLKB1 in these observations, we depleted LKB1 in H2009 andH358 NSCLC cells using siRNA and assessed the effects of 2DGtreatment. As shown in Fig. 3D, depletion of LKB1 in H2009 andH358 NSCLC cells resulted in reduced CHOP and BiP proteinexpressions (Fig. 3D), as well as reduced pAMPK (Fig. 3D) andreduced survival (supplemental Fig. 1C).

ER stress response is altered in LKB1-deficient NSCLC cells

Although 2DG can induce ER stress, it also inhibits other glucosedependent processes [37], which could manifest itself in ourobserved phenotype in LKB1-deficient NSCLC cells. Thus we com-pared the effects of two pharmacological activators of ER stress(tunicamycin [Tm], brefeldin A [BFA]) [22] to 2DG in derivativesof the LKB1-deficient NSCLC cell line, H838, stably expressingeither wild-type LKB1 or kinase-dead LKB1 (KDLKB1). Protein lev-els of BiP increased in both H838-LKB1 and H838-KDLKB1 cellsafter 12 h of treatment with BFA (30 ng/ml) or Tm (1.25 ug/ml)(Fig. 4A). Interestingly, the relative difference in BiP levels betweenH838-LKB1 cells and H838-KDLKB1 were moderate compared tothe effects upon BiP with 2DG (10 mM) treatment (Fig. 4A), per-haps related to 2DG inhibition of other glucose dependent func-tions. As such, we assessed the phosphorylation of the a subunitof the eukaryotic initiation factor 2 (eIF2/EIF2S1) that occurs inresponse to several stressors, including ER stress [34,35]. Interest-ingly, H838-KDLKB1 cells displayed increased levels of phosphory-lated eIF2a (phospho-eIF2a) following treatment with 2DG, Tm orBFA, compared to H838-LKB1 cells, contrary to the presence ofreduced BiP (Fig. 4A).

Activation of ER stress in LKB1-deficient NSCLC results in UPR-mediated cell death

Under moderate levels of ER stress, functions of the UPR areprotective, enabling adaptation and restoration of homeostasiswithin the cell. However in conditions of acute, irremediable ERstress, the UPR instead activates apoptosis [17]. Increased phos-pho-eIF2a is found during acute, irremediable ER stress and isassociated with ER stress induced cell death [18], leading us toassess whether 2DG, BFA and Tm were activating of UPR-mediatedapoptosis in LKB1-deficient NSCLC cells. Expression of CHOP and

Fig. 3. 2DG induced ER stress is altered in LKB1-deficient NSCLC cells. (A) LKB1 signaling correlates with expression of ER stress markers. LKB1 expressing H2009 or LKB1 nullH23 LA cells were treated with 20 mM 2DG for the indicated times. Following SDS–PAGE, samples were immunoblotted for LKB1, phosphorylated AMPK and CHOP and BiP,markers of ER stress and the UPR. Actin and total AMPK were used as loading controls. Blot is representative of 4 independent experiments. (B) LKB1 expression correlateswith expression of ER stress markers and decreased apoptosis in LA cells. A panel of LA cells (H358, H1975, H441-LKB1+; H2126, H838, H1395-LKB1�) was treated for 12 and18 h with 20 mM 2DG. Total protein lysates were immunoblotted with antibodies specific to LKB1, CHOP, BiP, phosphorylated AMPK and cleaved PARP. Actin was used as aloading control. Blot is representative of three independent experiments. (C) Re-expression of LKB1 in LKB1 null H23 LA cells restores expression of BiP and CHOP. H23-LKB1and H23-KDLKB1 LA cells were treated with 20 mM 2DG for the indicated times (hours) and probed as in A. Actin and total AMPK were used as loading controls. Blot isrepresentative of 4 independent experiments. (D) Reduction of LKB1 by siRNA knockdown attenuates UPR activation. LKB1 expressing H2009 or H358 were transfected with100 nM siRNA targeting human LKB1 or a scrambled control and treated with 2DG (20 mM) 48 h after transfection for four hours. Lysates were immunblotted with theindicated antibodies. Actin was used as a loading control. Blot is representative of 3 independent experiments.

L.J. Inge et al. / Cancer Letters 352 (2014) 187–195 191

activating transcription factor 4 (ATF4) function downstream ofphosphor-eIF2a in initiation of UPR-mediated cell death [38].Analysis of protein lysates from H838-LKB1 and H838-KDLKB1cells treated with with 2DG, Tm or BFA for 18 h revealed increasesin ATF4 (Fig. 4B) and equivalent levels of BiP for both cell lines(Fig. 4B). However, chronic treatment induced increased levels ofCHOP in H838-KDLKB1, relative to H838-LKB1 cells (Fig. 4B). Inaddition, H838-KDLKB1 cells showed increased cleavage of caspase9 (CASP9), a marker of ER stress-induced apoptosis [39], relative toH838-LKB1 after 18 h (Fig. 4B). Expression of ATF4 and CHOP havebeen linked to depletion of ATP, leading to cell death [38]. Consis-tent with the hypothesis, treatment with 2DG, Tm or BFA allresulted in reduced ATP (Fig. 5B) in both H838-KDLKB1 (Fig. 5A)and A427-KDLKB1 (Fig. 5B), compared to H838 and A427 cellsre-expressing LKB1 (p < 0.005) and reduced cell survival of H838-KDLKB1 cells (Fig. 5C). As these effects of ATF4 and CHOP inUPR-mediated cell death are a result of reactivation of protein syn-thesis [38], we tested whether reducing protein load would ame-liorate cytoxtoxicity of 2DG, BFA and Tm in the H838-KDLKB1cell line. Cycloheximide (CHX) has been previously shown toreduce ER stress cytoxtoxicity by lowering the overall client pro-teins within the ER [40]. The addition of CHX was able to amelio-rate 2DG, BFA and Tm cytotoxicity in H838-KDLKB1 cells,compared to 2DG, BFA or Tm alone (Fig. 5C, p < 0.05). ATF4 andCHOP activity leads to accumulation of reactive oxygen species

(ROS), which contributes to cell death [38]. Staining of ROS withCellROX™ in H838-LKB1 and H838-KDLKB1 cells showed thatH838-KDLKB1 had markedly higher basal levels of ROS (1.4-foldincrease in intensity over H838-LKB1), compared to H838-LKB1cells (Fig. 6A–C, dotted lines mark baseline [blue peaks] ROS levelof KDLKB1), consistent with recent reports of increased ROS inLKB1-deficient cells [11,14]. ROS levels in H838-KDLKB1increasedfurther, relative to H838-LKB1, following treatment with 2DG (2-fold increase in intensity over H838-LKB1), BFA (1.2-fold increasein intensity over H838-LKB1) or Tm (1.67-fold increase in intensityover H838-LKB1) (Fig. 6A–C, red peak).

Although BFA and Tm function to induce ER stress within cells,neither is currently being explored for clinical use. Thus, we nexttested bortezomib and celecoxib, which aggravates ER stress viaa mechanism unrelated to COX-2 inhibition [22], for preferentialcytotoxicity in LKB1-deficient cells. LKB1-deficient H23 NSCLC cellsexpressing KDLKB1 showed reduced expression of both BiP andCHOP following 18 h of exposure to either celecoxib (CHOP = 1.3-fold decrease; BiP = 1.6-fold decrease) or bortezomib(CHOP = 1.6-fold decrease, BiP = 9-fold decrease), compared toH23 cells with reconstituted LKB1; the presence of LKB1 correlatedwith phosphorylation of AMPK (Fig. 6D). In addition, both cele-coxib and bortezomib reduced the survival of H23-KDLKB1 cellsin a dose dependent manner, compared to H23-LKB1 NSCLC cells(Fig. 6E).

Fig. 4. Response to ERSA treatment is altered in LKB1-deficient NSCLC cells. (A andB) H838 human NSCLC cells expressing full-length LKB1 (H838-LKB1) or kinasedead LKB1 (H838-KDLKB1) were treated with 2DG (10 mM), tunicamycin (Tm,1.25 lg/ml) or brefeldin A (BFA, 30 ng/ml) for 12 (A) or 18 (B) hours. Lysates wereimmunoblotted with the indicated antibodies. Actin was used as loading controls.Blot is representative of 3 independent experiments.

Fig. 5. LKB1-deficient NSCLC cells are sensitive to ERSA treatment. (A and B) H838(A) or A427 (B) LKB1-deficient human NSCLC cell lines expressing full-length LKB1or kinase dead LKB1 (KDLKB1) were treated with 2DG (10 mM), Tm (1.25 lg/ml) orBFA (30 ng/ml) and ATP levels assessed by CellTiter Glo™. Levels of ATP werenormalized to vehicle treated control. Graphs depict mean ± SE from two indepen-dent experiments. (C) Clonogenic survival of H838-LKB1 and H838-KDLKB1 cells.H838 LKB1-deficient human NSCLC cell lines expressing full-length LKB1 or kinasedead LKB1 (KDLKB1) were treated with vehicle (DMSO), 2DG (10 mM), Tm (1.25 lg/ml) or BFA (30 ng/ml) with or without 125 ng/ml of cyclohexamide (CHX) fortwenty-four hours. Surviving fraction was calculated as described in Material andMethods. Graph depicts mean ± SE of three independent experiments.

192 L.J. Inge et al. / Cancer Letters 352 (2014) 187–195

Discussion

In the present study, we show that the non-hydrolyzable analogof glucose, 2DG has increased therapeutic efficacy in an in vivomodel of LKB1-deficient NSCLC, compared to NSCLC tumorsexpressing LKB1. Studies directed towards elucidating the mecha-nism/s behind the effect of 2DG in LKB1-deficient NSCLC tumors,reveal that the absence of LKB1 sensitizes NSCLC cells to aggrava-tion of ER stress. Lack of LKB1 activity in NSCLC cells was associ-ated with increased phosphorylation of eIF2a, CHOP and reducedsurvival, as well as evidence of UPR-mediated apoptosis (increasedROS, cleavage of caspase 9, reduced ATP). Significantly, theseeffects of 2DG in the absence of LKB1 also occurred upon aggrava-tion of ER stress with Tm and BFA, two well-characterized inducersof ER stress. Further, treatment with celecoxib and bortezomib,two FDA approved compounds with ER stress activity, also dis-played increased cytotoxicity when LKB1 activity was absent. Col-lectively, the sensitivity of LKB1-deficient NSCLC cells to ER stressaggravators (ERSAs) are suggestive of a potential treatment avenuefor NSCLC and necessitate further studies.

Although our data suggests that LKB1-deficient NSCLC cells arehypersensitive to ERSAs, the mechanism behind this phenotype areunclear. One of LKB1’s primary functions is alterations of cellularfunctions to restore ATP levels upon increased in AMP to promotecell survival [5]. Thus the inability to activate LKB1 function andrestore ATP levels upon ATF4/CHOP mediated depletion of ATPduring ER stress in LKB1-deficient NSCLC cells is a plausible mech-anism for our observations. However, LKB1 has been found to havebroad regulatory functions, leading to a variety of alterations incellular function when LKB1 is absent [5,11,12,14,41], in particularin initiation of autophagy, a catabolic mechanism responsible fordegrading damaged organelles and proteins, via phosphorylationof the ULK1 kinase [11,41]. Fittingly, autophagy is necessary for cellsurvival to ER stress, thought to alleviate proteo-toxicity via degra-dation of misfolded/unfolded proteins and the damaged ER [42].Thus, cytotoxicity of ERSAs in LKB1-deficient NSCLC cells couldbe a consequence of an inability to activate autophagy via failure

to induce AMPK phosphorylation of ULK1, contributing to irreme-diable ER stress and activation of apoptosis. Consistent with thispossibility, we found that 2DG, Tm and BFA results in ULK1 phos-phorylation in only in H838-LKB1 cells (data not shown), suggest-ing that this mechanism could be involved in our observations andis currently under study in our laboratory. Regardless, future stud-ies investigating the potential overlap in LKB1 and UPR function in

Fig. 6. ERSA treatment results in increased ROS levels in LKB1-deficient NSCLC cells. (A–C) Fluorescence-activated cell sorting (FACS) on H838-LKB1 (LKB1) and H838-KDLKB1(KDLKB1) human NSCLC cells stained with CellROX™ following treatment with (A) 2DG (10 mM), (B) Tm (1.25 lg/ml) or (C) BFA (30 ng/ml) for twelve hours or at baseline(0 h). Dotted line highlights baseline (0 h) ROS levels in H838-KDLKB1 compared to ROS levels in H838-LKB1. H838-KDLKB1 displayed a 1.4-fold increase over H838-LKB1 atbaseline and 2-fold (2DG), 1.2-fold (BFA) and 1.67-fold increase over H838-LKB1 with treatment. (D) LKB1-deficient H23 human NSCLC cells expressing full length LKB1(H23-LKB1) or kinase dead LKB1 (H23-KDLKB1) were treated with celecoxib (40 lM) or bortezomib (40 nM) for 18 h. Protein lysates were immunoblotted with the indicatedantibodies. (E) Clonogenic survival of H23-LKB1 and H23-KDLKB1 cells treated with vehicle (DMSO) or celecoxib or bortezomib at the indicated concentration for twenty-fourhours. Surviving fraction was calculated as described in Material and Methods. Graph depicts mean ± SE of three independent experiments.

L.J. Inge et al. / Cancer Letters 352 (2014) 187–195 193

cellular adaptation to stress will be critically important, as thesestudies have potential to reveal alternative targets.

Our in vitro data suggest that ERSA treatment may have efficacyin LKB1-deficient NSCLC tumors. Although daily treatment with2DG was found to have increased efficacy in LKB1-deficient NSCLCtumors, compared to NSCLC tumors with functional LKB1 (Fig. 1),we found 2DG had no observable effects upon metastasis inKRas/LKB1null mice. As metastasis is the major cause of cancerrelated mortalities, this finding which would also explain why2DG had no statistical impact upon survival in KRas/LKB1null micein longer term studies (data not shown). Further, we found thatchronic dosing with 2DG or Tm was poorly tolerated, consistentwith published reports of toxic side effects. Collectively, theseobservations indicate that 2DG may display greater efficacy if usedin defined combinations instead of as a single agent or in combina-tion with standard treatments as is currently being explored [43],an area we are currently pursing. Alternatively, it would be moreuseful to assess the effects of ERSAs with established clinical pro-files. Currently several compounds with known ERSA effects areapproved by the FDA, most notably are bortezomib, celecoxiband nelfinavir [22]. Indeed, we found that both bortezomib andcelecoxib displayed increased in vitro cytotoxicity within anisogenic LKB1-deficent NSCLC cell line (Fig. 6E). Mechanistically,celecoxib has been shown to induce ER stress by causing calciumefflux from the ER due to inhibition of the sarcoplasmic/endoplas-

mic reticulum calcium ATPase, in addition to its known ability toinhibit COX-2 [44]. Celecoxib’s ERSA activity has been theorizedas a possible mechanism for its antitumor effects, however theconsiderable cardiovascular side effects associated with celecoxib,and its lack of antitumor in advanced tumors, has led to studiesaimed at developing analogs of celecoxib with more potent ERSAactivity and eliminating its COX-2 inhibitory function [44]. ERSAactivity of both bortezomib and nelfinavir are based upon theirfunctions in inhibiting proteasomal degradation [22]. Nelfinavir,currently approved for treatment of HIV infections, has been foundto have considerable antitumor activity, inducing cytotoxicity as asingle agent, as well as enhancing the toxicity of standard treat-ments [45,46]. As such, clinical trials for nelfinavir are ongoing[47]. Although bortezomib has considerable activity towardshematologic malignancies, it has minimal activity in solid tumors,due its inability to penetrate tissues and achieve efficacious con-centrations [48]. Several second generation proteasome inhibitorshave been developed, such as MLN9708, which displays improvedtissue distribution and antitumor activity in solid tumors [48], andrepresent an attractive alternative to bortezomib in treating LKB1-deficient NSCLC. Collectively, future studies geared towards assess-ing the activity of these compounds, either alone or in combinationwith other defined treatments, are warranted based upon theeffects of Tm and BFA in our in vitro model systems. Given thedemonstrable need for new therapeutic approaches for NSCLC

194 L.J. Inge et al. / Cancer Letters 352 (2014) 187–195

and the incidence of sporadic LKB1 mutations in lung cancer, aswell as cervical and endometrial cancer, completion of these stud-ies would have considerable value.

Conflict of Interest

The authors have no conflicts to disclose.

Acknowledgments

This work was funded by Grants from the St. Joseph’s Founda-tion (Phoenix, AZ) for the Heart and Lung Institute Research Initia-tive and American Lung Association (RG-224607-N-LJ.I). Theauthors wish to thank Dr. Gregory Turner and Mr. Qingwei Liufor performing the MRI imaging and assisting in the imageanalysis.

Appendix A. Supplementary material

Supplementary data associated with this article can be found, inthe online version, at http://dx.doi.org/10.1016/j.canlet.2014.06.011.

References

[1] R. Siegel, D. Naishadham, A. Jemal, Cancer statistics, CA Can. J. Clin. 63 (1)(2013) 11–30. Epub 2013/01/22.

[2] J.M. Varlotto, L.N. Medford-Davis, A. Recht, J.C. Flickinger, E. Schaefer, M.M.DeCamp, Failure rates and patterns of recurrence in patients with respected N1non-small-cell lung cancer, Int. J. Rad. Onc., Biol., Phys. 81 (2) (2011) 353–359.Epub 2010/08/25.

[3] R. Arriagada, B. Bergman, A. Dunant, T. Le Chevalier, J.P. Pignon, J.Vansteenkiste, Cisplatin-based adjuvant chemotherapy in patients withcompletely respected non-small-cell lung cancer, The New Eng. J. Med. 350(4) (2004) 351–360. Epub 2004/01/23.

[4] T. Reungwetwattana, S.J. Weroha, J.R. Molina, Oncogenic pathways,molecularly targeted therapies, and highlighted clinical trials in non-small-cell lung cancer (NSCLC), Clin. Lung Can. 13 (4) (2012) 252–266. Epub 2011/12/14.

[5] D.B. Shackelford, R.J. Shaw, The LKB1-AMPK pathway: metabolism and growthcontrol in tumour suppression, Nat. Rev. Can. 9 (8) (2009) 563–575.

[6] R.K. Gill, S. Yang, D. Meerzaman, L.E. Mechanic, E.D. Bowman, H. Jeon, et al.,Frequent homozygous deletion of the LKB1/STK11 gene in non-small cell lungcancer, Oncogene 30 (2011) 3784–3791.

[7] L. Ding, G. Getz, D.A. Wheeler, E.R. Mardis, M.D. McLellan, K. Cibulskis, et al.,Somatic mutations affect key pathways in lung adenocarcinoma, Nature 455(2008) 1069–1075.

[8] C. Contreras, S. Gurumurthy, J. Haynie, L. Shirley, E. Akbay, S. Wingo, et al., Lossof Lkb1 provokes highly invasive endometrial adenocarcinomas, Can. Res. 68(3) (2008) 759–766.

[9] H. Ji, M. Ramsey, D. Hayes, C. Fan, K. Mcnamara, P. Kozlowski, et al., LKB1modulates lung cancer differentiation and metastasis, Nature 448 (7155)(2007) 807–810.

[10] W. Liu, K.B. Monahan, A.D. Pfefferle, T. Shimamura, J. Sorrentino, K.T. Chan,et al., LKB1/STK11 inactivation leads to expansion of a prometastatic tumorsubpopulation in melanoma, Can. Cell. 21 (2012) 751–764.

[11] D. Shackelford, E. Abt, L. Gerken, D. Vasquez, A. Seki, M. Leblanc, et al., LKB1inactivation dictates therapeutic response of non-small cell lung cancer to themetabolism drug phenformin, Can. Cell. 23 (2) (2013) 143–158.

[12] D.B. Shackelford, D.S. Vasquez, J. Corbeil, S. Wu, M. Leblanc, C.L. Wu, et al.,MTOR and HIF-1alpha-mediated tumor metabolism in an LKB1 mouse modelof Peutz-Jeghers syndrome, Proc. Nat. Acad. Sci. USA 106 (27) (2009) 11137–11142.

[13] Y. Liu, K. Marks, G.S. Cowley, J. Carretero, Q. Liu, T.J. Nieland, et al., Metabolicand functional genomic studies identify deoxythymidylate kinase as a targetin LKB1-mutant lung cancer, Can. Disc. 3 (8) (2013) 870–879. Epub 2013/05/30.

[14] B. Faubert, E.E. Vincent, T. Griss, B. Samborska, S. Izreig, R.U. Svensson, et al.,Loss of the tumor suppressor LKB1 promotes metabolic reprogramming ofcancer cells via HIF-1alpha, Proc. Nat. Acad. Sci. USA 111 (7) (2014) 2554–2559. Epub 2014/02/20.

[15] L.J. Inge, K.D. Coon, M.A. Smith, R.M. Bremner, Expression of LKB1 tumorsuppressor in non-small cell lung cancer determines sensitivity to 2-deoxyglucose, J. Thor. Cardiovas. Surg. (2008).

[16] D. Ron, P. Walter, Signal integration in the endoplasmic reticulum unfoldedprotein response, Nat. Rev. 8 (7) (2007) 519–529.

[17] G.C. Shore, F.R. Papa, S.A. Oakes, Signaling cell death from the endoplasmicreticulum stress response, Curr. Opin. Cell Biol. 23 (2) (2011) 143–149. Epub2010/12/15.

[18] D. Han, A.G. Lerner, L. Vande Walle, J.-P. Upton, W. Xu, A. Hagen, et al.,IRE1alpha kinase activation modes control alternate endoribonuclease outputsto determine divergent cell fates, Cell 138 (3) (2009) 562–575.

[19] S. Oyadomari, C. Yun, E.A. Fisher, N. Kreglinger, G. Kreibich, M. Oyadomari,et al., Cotranslocational degradation protects the stressed endoplasmicreticulum from protein overload, Cell 126 (4) (2006) 727–739.

[20] C.M. Haynes, E.A. Titus, A.A. Cooper, Degradation of misfolded proteinsprevents ER-derived oxidative stress and cell death, Mol. Cell. 15 (5) (2004)767–776. Epub 2004/09/08.

[21] Y. Ma, L.M. Hendershot, The role of the unfolded protein response in tumourdevelopment: friend or foe?, Nat Rev. Can. 4 (12) (2004) 966–977.

[22] A.H. Schonthal, Pharmacological targeting of endoplasmic reticulum stresssignaling in cancer, Biochem. Pharm. 85 (5) (2013) 653–666. Epub 2012/09/25.

[23] S. White-Gilbertson, Y. Hua, B. Liu, The role of endoplasmic reticulum stress inmaintaining and targeting multiple myeloma: a double-edged sword ofadaptation and apoptosis, Front. Gen. 4 (2013) 109. Epub 2013/06/20.

[24] E.L. Jackson, N. Willis, K. Mercer, R.T. Bronson, D. Crowley, R. Montoya, et al.,Analysis of lung tumor initiation and progression using conditional expressionof oncogenic K-ras, Genes Dev. 15 (24) (2001) 3243–3248. Epub 2001/12/26.

[25] N. Bardeesy, M. Sinha, A. Hezel, S. Signoretti, N. Hathaway, N. Sharpless, et al.,Loss of the Lkb1 tumour suppressor provokes intestinal polyposis butresistance to transformation, Nature 419 (6903) (2002) 162–167.

[26] N.A. Franken, H.M. Rodermond, J. Stap, J. Haveman, C. van Bree, Clonogenicassay of cells in vitro, Nat. Prot. 1 (5) (2006) 2315–2319.

[27] T.G. Whitsett, E. Cheng, L. Inge, K. Asrani, N.M. Jameson, G. Hostetter, et al.,Elevated expression of Fn14 in non-small cell lung cancer correlates withactivated EGFR and promotes tumor cell migration and invasion, Am. J. Pathol.181 (1) (2012) 111–120. Epub 2012/05/29.

[28] M. Kurtoglu, J. Maher, T.J. Lampidis, Differential toxic mechanisms of 2-deoxy-D-glucose versus 2-fluorodeoxy-D-glucose in hypoxic and normoxic tumorcells, Antiox. Redox Signal. 9 (9) (2007) 1383–1390.

[29] G. Maschek, N. Savaraj, W. Priebe, P. Braunschweiger, K. Hamilton, G.F.Tidmarsh, et al., 2-deoxy-D-glucose increases the efficacy of adriamycin andpaclitaxel in human osteosarcoma and non-small cell lung cancers in vivo,Can. Res. 64 (1) (2004) 31–34.

[30] Z. Chen, K. Cheng, Z. Walton, Y. Wang, H. Ebi, T. Shimamura, et al., A murinelung cancer co-clinical trial identifies genetic modifiers of therapeuticresponse, Nature 483 (2012) 613–617.

[31] J.H. Cheong, E.S. Park, J. Liang, J.B. Dennison, D. Tsavachidou, C. Nguyen-Charles, et al., Dual inhibition of tumor energy pathway by 2-deoxyglucoseand metformin is effective against a broad spectrum of preclinical cancermodels, Mol. Can. Ther. 10 (12) (2011) 2350–2362. Epub 2011/10/14.

[32] M. Kurtoglu, N. Gao, J. Shang, J. Maher, M.A. Lehrman, M. Wangpaichitr, et al.,Under normoxia, 2-deoxy-D-glucose elicits cell death in select tumor types notby inhibition of glycolysis but by interfering with N-linked glycosylation, Mol.Can. Therap. 6 (11) (2007) 3049–3058.

[33] A.S. Lee, The ER chaperone and signaling regulator GRP78/BiP as a monitor ofendoplasmic reticulum stress, Methods 35 (4) (2005) 373–381.

[34] Y.C. Chae, M.C. Caino, S. Lisanti, J.C. Ghosh, T. Dohi, N.N. Danial, et al., Control oftumor bioenergetics and survival stress signaling by mitochondrial HSP90s,Can. Cell. 22 (3) (2012) 331–344. Epub 2012/09/15.

[35] H.P. Harding, Y. Zhang, H. Zeng, I. Novoa, P.D. Lu, M. Calfon, et al., An integratedstress response regulates amino acid metabolism and resistance to oxidativestress, Mol. Cell. 11 (3) (2003) 619–633.

[36] X.Z. Wang, B. Lawson, J.W. Brewer, H. Zinszner, A. Sanjay, L.J. Mi, et al., Signalsfrom the stressed endoplasmic reticulum induce C/EBP-homologous protein(CHOP/GADD153), Mol. Cell. Biol. 16 (8) (1996) 4273–4280.

[37] G.S. Karczmar, J.M. Arbeit, B.J. Toy, A. Speder, M.W. Weiner, Selective depletionof tumor ATP by 2-deoxyglucose and insulin, detected by 31P magneticresonance spectroscopy, Can. Res. 52 (1) (1992) 71–76.

[38] J. Han, S.H. Back, J. Hur, Y.H. Lin, R. Gildersleeve, J. Shan, et al., ER-stress-induced transcriptional regulation increases protein synthesis leading to celldeath, Nat. Cell. Biol. 15 (5) (2013) 481–490. Epub 2013/04/30.

[39] A. Jimbo, E. Fujita, Y. Kouroku, J. Ohnishi, N. Inohara, K. Kuida, et al., ER stressinduces caspase-8 activation, stimulating cytochrome c release and caspase-9activation, Exp. Cell. Res. 283 (2) (2003) 156–166. Epub 2003/02/13.

[40] D.R. Fels, J. Ye, A.T. Segan, S.J. Kridel, M. Spiotto, M. Olson, et al., Preferentialcytotoxicity of bortezomib toward hypoxic tumor cells via overactivationof endoplasmic reticulum stress pathways, Can. Res. 68 (22) (2008)9323–9330.

[41] D.F. Egan, D.B. Shackelford, M.M. Mihaylova, S. Gelino, R.A. Kohnz, W. Mair,et al., Phosphorylation of ULK1 (hATG1) by AMP-activated protein kinaseconnects energy sensing to mitophagy, Science 331 (6016) (2011) 456–461.Epub 2011/01/06.

[42] M. Ogata, S-i Hino, A. Saito, K. Morikawa, S. Kondo, S. Kanemoto, et al.,Autophagy is activated for cell survival after endoplasmic reticulum stress,Mol. Cell. Biol. 26 (24) (2006) 9220–9231.

[43] L.E. Raez, K. Papadopoulos, A.D. Ricart, E.G. Chiorean, R.S. Dipaola, M.N. Stein,et al., A phase I dose-escalation trial of 2-deoxy-D-glucose alone or combinedwith docetaxel in patients with advanced solid tumors, Can. Chemo. Pharm. 71(2) (2013) 523–530. Epub 2012/12/12.

[44] P. Pyrko, A. Kardosh, Y.T. Liu, N. Soriano, W. Xiong, R.H. Chow, et al., Calcium-activated endoplasmic reticulum stress as a major component of tumor cell

L.J. Inge et al. / Cancer Letters 352 (2014) 187–195 195

death induced by 2,5-dimethyl-celecoxib, a non-coxib analogue of celecoxib,Mol. Can. Therap. 6 (4) (2007) 1262–1275.

[45] A. Bruning, K. Friese, A. Burges, I. Mylonas, Tamoxifen enhances the cytotoxiceffects of nelfinavir in breast cancer cells, Breast Can. Res.: BCR 12 (4) (2010)R45. Epub 2010/07/03.

[46] P. Pyrko, A. Kardosh, W. Wang, W. Xiong, A.H. Schonthal, T.C. Chen, HIV-1protease inhibitors nelfinavir and atazanavir induce malignant glioma deathby triggering endoplasmic reticulum stress, Can. Res. 67 (22) (2007) 10920–10928. Epub 2007/11/17.

[47] R. Rengan, R. Mick, D. Pryma, M.A. Rosen, L.L. Lin, A.M. Maity, et al., A phase Itrial of the HIV protease inhibitor nelfinavir with concurrentchemoradiotherapy for unresectable stage IIIA/IIIB non-small cell lungcancer: a report of toxicities and clinical response, J. Thor. Oncol.: Off. Publ.Int. Assoc. Study Lung Canc. 7 (4) (2012) 709–715. Epub 2012/03/20.

[48] L.R. Dick, P.E. Fleming, Building on bortezomib: second-generation proteasomeinhibitors as anti-cancer therapy, Drug Disc. Today 15 (5–6) (2010) 243–249.Epub 2010/02/02.