Axl mediates acquired resistance of head and neck cancer...

Axl mediates acquired resistance of head and neck cancercells to the epidermal growth factor receptor inhibitor erlotinibGiles, K., Kalinowski, F., Candy, P., Epis, M., Zhang, P., Redfern, A., ... Leedman, P. (2013). Axl mediatesacquired resistance of head and neck cancer cells to the epidermal growth factor receptor inhibitor erlotinib.Molecular Cancer Therapeutics, 12(11), 2541-2558. DOI: 10.1158/1535-7163.MCT-13-0170

Published in:Molecular Cancer Therapeutics

DOI:10.1158/1535-7163.MCT-13-0170

Document VersionPeer reviewed version

Link to publication in the UWA Research Repository

General rightsCopyright owners retain the copyright for their material stored in the UWA Research Repository. The University grants no end-userrights beyond those which are provided by the Australian Copyright Act 1968. Users may make use of the material in the Repositoryproviding due attribution is given and the use is in accordance with the Copyright Act 1968.

Take down policyIf you believe this document infringes copyright, raise a complaint by contacting [email protected]. The document will beimmediately withdrawn from public access while the complaint is being investigated.

Download date: 13. May. 2018

https://doi.org/10.1158/1535-7163.MCT-13-0170

http://research-repository.uwa.edu.au/en/publications/axl-mediates-acquired-resistance-of-head-and-neck-cancer-cells-to-the-epidermal-growth-factor-receptor-inhibitor-erlotinib(728a3d09-2c0a-4a45-afe4-0b2c9b910eab).html

UWA Research Publication

Giles, K.M., Kalinowski, F.C., Candy, P.A., Epis, M.R., Zhang, P.M., Redfern, A.D., Stuart, L.M., Goodal, G.J. & Leedman, P.J. (2013). Axl mediates acquired resistance of head and neck cancer cells to the epidermal growth factor receptor inhibitor erlotinib. MOLECULAR CANCER THERAPEUTICS, 12(11), 2541-2558.

©2013 American Association for Cancer Research.

This is pre-copy-editing, author-produced version of an article accepted for publication in Molecular Cancer Therapeutics, following peer review. The definitive published version (see citation above) is located on the article abstract page of the publisher, the American Association for Cancer Research. This version was made available in the UWA Research Repository on 1st of November 2014, in compliance with the publisher’s policies on archiving in institutional repositories.

Use of the article is subject to copyright law.

http://www.uwa.edu.au/

Giles et al., Revision MCT-13-0170

1

Axl mediates acquired resistance of head and neck cancer cells to the epidermal 1

growth factor receptor inhibitor erlotinib 2

3

Keith M. Giles1, Felicity C. Kalinowski1, Patrick A. Candy1,2, Michael R. Epis1, 4

Priscilla M. Zhang1, Andrew D. Redfern1,2,3, Lisa M. Stuart1, Gregory J. Goodall4,5,6 5

and Peter J. Leedman1,2 6

1Laboratory for Cancer Medicine, Western Australian Institute for Medical Research 7

and University of Western Australia Centre for Medical Research, Perth, WA 6000, 8

Australia 9

2School of Medicine and Pharmacology, University of Western Australia, Nedlands, 10

WA 6008, Australia 11

3Department of Medical Oncology, Royal Perth Hospital, Perth, WA 6000, Australia 12

4Centre for Cancer Biology, SA Pathology, Adelaide, SA 5000, Australia 13

5School of Molecular and Biomedical Science, University of Adelaide, SA 5000, 14

Australia 15

6Department of Medicine, University of Adelaide, SA 5000, Australia 16

17

Corresponding author: Peter Leedman, PhD, Western Australian Institute for Medical 18

Research, Level 6, MRF Building, Rear 50 Murray Street, Perth, WA 6000, Australia. 19

Tel: 618-9224-0323. Fax: 618-9224-0322. E-mail: [email protected] 20

Keywords: head and neck cancer; epidermal growth factor receptor; Axl; epithelial-21

mesenchymal transition; erlotinib 22

23

Total number of figures and tables: 8 figures + 3 tables 24

Running title: Axl promotes erlotinib resistance in head and neck cancer 25


2

ABSTRACT 26

Elevated expression and activity of the epidermal growth factor receptor (EGFR) is 27

associated with development and progression of head and neck cancer (HNC) and a 28

poor prognosis. Clinical trials with EGFR tyrosine kinase inhibitors (TKIs; eg. 29

erlotinib) have been disappointing in HNC. To investigate the mechanisms mediating 30

resistance to these agents, we developed a HNC cell line (HN5-ER) with acquired 31

erlotinib resistance. In contrast to parental HN5 HNC cells, HN5-ER cells exhibited 32

an epithelial-mesenchymal (EMT) phenotype with increased migratory potential, 33

reduced E-cadherin and epithelial-associated miRNAs, and elevated vimentin 34

expression. Phosphorylated RTK profiling identified Axl activation in HN5-ER cells. 35

Growth and migration of HN5-ER cells was blocked with a specific Axl inhibitor, 36

R428, and R428 re-sensitized HN5-ER cells to erlotinib. Microarray analysis of HN5-37

ER cells confirmed the EMT phenotype associated with acquired erlotinib resistance, 38

and identified activation of gene expression associated with cell migration and 39

inflammation pathways. Moreover, increased expression and secretion of interleukin 40

(IL)-6 and IL-8 in HN5-ER cells suggested a role for inflammatory cytokine signaling 41

in EMT and erlotinib resistance. Expression of the tumor suppressor miR-34a was 42

reduced in HN5-ER cells and increasing its expression abrogated Axl expression and 43

reversed erlotinib resistance. Finally, analysis of 302 HNC patients revealed that high 44

tumor Axl mRNA expression was associated with poorer survival (HR 1.66, 45

p=0.007). In summary, our results identify Axl as a key mediator of acquired erlotinib 46

resistance in HNC and suggest that therapeutic inhibition of Axl by small molecule 47

drugs or specific miRNAs might overcome anti-EGFR therapy resistance. 48

49


3

INTRODUCTION 50

Head and neck cancer (HNC) is a leading cause of worldwide cancer death (1). Many 51

HNC patients present with locally advanced or metastatic disease, and despite 52

advances in surgery, radiotherapy and chemotherapy the prognosis for such patients 53

remains poor (2). The epidermal growth factor receptor (EGFR) receptor tyrosine 54

kinase (RTK) is overexpressed in more than 80% of HNCs, is associated with a poor 55

clinical outcome (3), and promotes tumor growth, metastasis, treatment resistance and 56

angiogenesis, making it an attractive therapeutic target. Disappointingly, clinical trials 57

to date with anti-EGFR monoclonal antibodies (eg. cetuximab) or small molecule 58

EGFR tyrosine kinase inhibitors (TKIs, eg. erlotinib or gefitinib) have shown that 59

these agents have only modest activity in recurrent or advanced HNC. While most 60

HNCs are inherently resistant to EGFR inhibition, approximately 5-15% of HNC 61

patients experience an initial anti-tumor response (4), with minor shrinkage or disease 62

stabilization that rarely persists beyond a few months, presumably because tumors 63

rapidly acquire EGFR inhibitor resistance. 64

65

Understanding the mechanisms that mediate EGFR inhibitor resistance in HNC might 66

allow better selection of patients most likely to benefit from anti-EGFR therapy, and 67

may identify novel strategies to improve the efficacy of these agents. This is of 68

particular importance given the large number of active trials investigating anti-EGFR 69

agents such as erlotinib in various HNC settings. A number of research studies have 70

investigated the inherent and acquired resistance to EGFR inhibition in HNC, but the 71

precise mechanisms remain poorly understood. EGFR tyrosine kinase domain 72

mutations determine EGFR inhibitor sensitivity in non-small cell lung cancer 73

(NSCLC), but do not occur in HNC (5), while KRAS mutations associated with 74


4

EGFR inhibitor resistance in colorectal cancer are rarely found in HNC (6). Other 75

studies suggested that EGFR gene amplification and the levels of activated ErbB2 and 76

total ErbB3 may determine gefitinib sensitivity in HNC cell lines (7), while ErbB2 77

inhibition sensitizes HNC cells to cetuximab (8), implying that combined RTK 78

blockade might be an effective approach to overcome EGFR inhibitor resistance. A 79

recent study demonstrated that inhibition of Signal Transducer and Activator of 80

Transcription 3 (STAT3) activity sensitizes HNC cells to EGFR inhibition (9), while 81

EGFR-independent activation of Akt has been associated with inherent and acquired 82

gefitinib resistance in HNC cell lines (10, 11), supporting the concept of sustained Akt 83

activity and tumor progression being mediated through compensatory RTK signaling. 84

Finally, mounting evidence has implicated the process of epithelial-mesenchymal 85

transition (EMT), a state of altered cell morphology and migration, in EGFR inhibitor 86

resistance in a range of solid tumors, including HNC and NSCLC (12, 13), and EMT 87

has itself been associated with increased Akt signaling in HNC (14). 88

89

Recognizing that a small subset (5-15%) of unselected HNC patients experience 90

transient clinical responses to EGFR inhibitors, we reasoned that the inevitable 91

treatment failure and disease progression in these cases results from tumors rapidly 92

acquiring resistance to anti-EGFR drugs. We generated and characterized an erlotinib-93

resistant HNC cell line (HN5-ER) to investigate the molecular mechanisms associated 94

with erlotinib resistance in HNC. We show that erlotinib-resistant HNC cells exhibit a 95

classical EMT phenotype with increased migratory potential, loss of epithelial 96

markers and gain of mesenchymal markers, reduced expression of microRNAs 97

(miRNAs) functionally associated with EMT, and increased expression and activation 98

of Axl, an RTK known to promote tumor growth, metastasis and treatment resistance. 99


5

Importantly, a small molecule selective inhibitor of Axl, R428, reduced growth and 100

migration of erlotinib-resistant cells, and restored their sensitivity to erlotinib. 101

Microarray analyses identified alterations in several cancer-associated gene 102

expression programs that were associated with acquired erlotinib resistance in HNC, 103

including activation of pro-inflammatory signaling, with increased expression and 104

secretion of IL-6 and IL-8 and elevated STAT3 activity. Finally, we demonstrate a 105

role for the tumor suppressor miRNA miR-34a in the regulation of Axl expression, 106

EMT, and EGFR inhibitor resistance in HNC cells. Together, our findings identify 107

novel targets to develop strategies to circumvent this resistance and improve the 108

efficacy of anti-EGFR therapy in HNC. 109

110

MATERIALS AND METHODS 111

HN5 cell line, cell culture and reagents, cell imaging 112

HN5 cells (15) were provided by A/Prof. Terrance Johns (Monash Institute of 113

Medical Research, Australia) and their identity was verified by short tandem repeat 114

(STR) profiling at CellBank Australia (Children’s Medical Research Institute, 115

Westmead, Australia). STR profiling also confirmed the parental HN5 origin for 116

HN5-ER cells. The HNC cell line FaDu (16) was derived from a pharyngeal 117

carcinoma from a 56 year old male and was obtained from the American Type Culture 118

Collection (ATCC). Cells were cultured in Dulbecco’s Modified Eagles Medium 119

(DMEM; Invitrogen) with 10% fetal bovine serum (FBS) at 37°C in 5% CO2. 120

Erlotinib (LC Laboratories; Massachusetts) was prepared as a 23 mM stock solution 121

in 96% (v/v) dimethyl sulfoxide (DMSO; Sigma-Aldrich). R428 (Symansis, 122

California) was prepared as a 10 mM stock solution in DMSO. Gefitinib (Selleck 123


6

Chemicals) was prepared as a 20 mM stock solution in DMSO. Tocilizumab (Roche) 124

was supplied as a 20 mg/ml stock solution. Ruxolitinib (LC Laboratories; 125

Massachusetts) was prepared as a 33 mM stock solution in DMSO. Synthetic miR-126

34a (hsa-miR-34a-5p; Catalog no. AM17100, Assay ID PM11030) and miR-negative 127

control (miR-NC; negative control #1, Catalog no. AM17110) miRNA precursor 128

molecules were sourced from Ambion (Austin, Texas). Axl siRNAs (Hs_AXL_3 129

Catalog no. SI00131355, Hs_AXL_9 Catalog no. SI00605304, Hs_AXL_10 Catalog 130

no. SI00605311 and Hs_AXL_12 Catalog no. SI02626743) and AllStars negative 131

control siRNA (Catalog no. 1027280) were obtained from QIAGEN. 132

Cells were photographed with an Olympus IX71 inverted microscope using an 133

Olympus DP70 Digital Camera System at 100X magnification. 134

Generation of an erlotinib-resistant HN5 subline, HN5-ER 135

HN5 cells were cultured in increasing concentrations of erlotinib to a concentration of 136

12.5 µM over 3 months, maintained in 12.5 µM erlotinib for 6 months, and tested to 137

confirm they had stably acquired erlotinib resistance. The erlotinib-resistant HN5 cell 138

subline was designated HN5-ER. 139

Drug sensitivity and cell viability assays 140

Assays were performed as described (17). Briefly, to determine the sensitivity of HN5 141

and HN5-ER cell lines to erlotinib, gefitinib and/or R428, 5.0 x 103 cells were seeded 142

per well in 96 well plates, and fresh media containing erlotinib or R428 added 24 h 143

after cell plating. Cell viability was measured 3 d after addition of 144

erlotinib/R428/gefitinib using the CellTitre 96 Aqueous One Solution Cell 145

Proliferation Assay Kit (Promega) and a FLUOstar OPTIMA microplate reader 146


7

(BMG Labtech). EC50 values for erlotinib and R428 were calculated for each cell line 147

using GraphPad Prism (GraphPad software). 148

To assess the efficacy of combining R428 and erlotinib or gefitinib in HN5-ER cells, 149

an ineffective concentration of erlotinib or gefitinib was used (7.5 or 10 μM). Cell 150

viability was measured 3 d after addition of erlotinib or gefitinib and/or R428 as 151

above. 152

To assess the efficacy of combining miR-34a and erlotinib in HN5-ER cells, cells 153

were transfected with miR-34a or miR-NC for 3 d and then treated with an ineffective 154

concentration of erlotinib (10 μM) for a further 2 d, after which cell viability was 155

determined. 156

miRNA precursor and siRNA transfections 157

miRNA transfection experiments were performed as described (18). Briefly, HN5-ER 158

cells were seeded in 6-well plates at a density of 5.0 x 105 cells per well and 159

transfected with 30 nM miR-34a or miR-NC for 24 h prior to RNA extraction or 160

protein extraction and immunoblotting as outlined below. 161

For transfections with Axl siRNA, 2.5 x 105 cells per well were seeded in 6-well 162

plates and transfected with 10 nM Axl siRNA or AllStars negative control siRNA 163

(NC siRNA) for 3 d prior to RNA extraction or protein extraction and immunoblotting 164

as outlined below. 165

Drug treatments for immunoblotting studies 166

For characterization of the effect of combined erlotinib and R428 treatment on HN5-167

ER cells, cells were cultured overnight in DMEM + 0.2% FBS, and then treated with 168


8

either 5 µM R428, 5 µM erlotinib, a combination of 5 µM R428 and 5 µM erlotinib, 169

or DMSO for 4 h prior to preparing cytoplasmic protein extracts for immunoblotting 170


To assess the effect of R428 on HN5 cells, cells were cultured overnight in DMEM + 172

0.5% FBS, and then treated with 10 µM R428 or DMSO for 4 h, prior to protein 173

extraction and immunoblotting as outlined below. 174

To assess the effect of ruxolitinib or tocilizumab on HN5-ER cells, cells were cultured 175

overnight in DMEM + 0.2% FBS, and then treated with DMSO, or 3 µM ruxolitinib 176

for 24 h, or 1 µM tocilizumab for 3 d, prior to protein extraction and immunoblotting 177


Reverse transcription and quantitative polymerase chain reaction (RT-qPCR) 179

Total RNA was extracted from HN5 and HN5-ER cell lines with TRIzol reagent 180

(Invitrogen). For RT-qPCR analysis of mRNA expression, 0.5 μg of total RNA was 181

reverse transcribed into cDNA using a QuantiTect Reverse Transcription Kit 182

(QIAGEN). Real-time PCR was performed on a Corbett Rotor-Gene 6000 183

thermocycler (QIAGEN) using QuantiTect SYBR Mix (QIAGEN) and validated 184

QuantiTect primer assays (QIAGEN) for EGFR (QT00085701), AXL (QT00067725), 185

CDH1 (QT00080143), CDH2 (QT00063196), PTGS2 (QT00040586), ZEB1 186

(QT01888446), IL-6 (QT00083720), IL-8 (QT00000322), ALAS1 (QT00073122) and 187

GAPDH (QT01192646). Expression of EGFR, AXL, CDH1, CDH2, PTGS2, ZEB1, 188

IL-6 and IL-8 mRNA was determined relative to ALAS1 and GAPDH mRNA using 189

the 2-ΔΔCt method (19). 190


9

For RT-qPCR analysis of miRNA expression, reverse transcription and PCR were 191

carried out using a TaqMan miRNA assay kit (Applied Biosystems) for hsa-miR-34a-192

5p (miR-34a; Life Technologies, Assay ID 000426), miR-200a, miR-200b, miR-200c 193

(Life Technologies, Assay IDs 000502, 002251, 002300) and RNU44 small nucleolar 194

RNA (U44 snRNA) (Life Technologies, Assay ID 001094) with a Corbett Rotor-195

Gene 6000 thermocycler (Qiagen) according to manufacturer’s instructions. 196

Expression of mature miR-34a, miR-200a, miR-200b, or miR-200c miRNAs relative 197

to U44 snRNA was determined using the 2−ΔΔCt method. 198

Protein extraction and immunoblotting 199

Cytoplasmic protein extracts were prepared from cell lines and immunoblotting 200

performed as described (18). For detection of vimentin, cell lysates were sonicated. 201

Polyvinylidene difluoride (PVDF) membranes (Roche) were probed with anti-EGFR 202

rabbit monoclonal antibody (1:5000, Abcam ab52894-100), anti-phospho-EGFR 203

(Tyr1173) goat polyclonal antibody (1:750, Santa Cruz Biotechnology sc-12351), 204

anti-Akt rabbit polyclonal antibody (1:1000, Cell Signaling Technology 9272), anti-205

phospho-Akt (Ser473) rabbit monoclonal antibody (1:500, Cell Signaling Technology 206

4060S), anti-E-cadherin rabbit monoclonal antibody (1:1000, Cell Signaling 207

Technology 3195S), anti-Vimentin mouse monoclonal antibody (1:1000, Cell 208

Signaling Technology 3390), anti-Axl rabbit monoclonal antibody (1:1000, Cell 209

Signaling Technology 8661), anti-ERK1/2 rabbit polyclonal antibody (1:750, Cell 210

Signaling Technology 9102), anti-phospho-ERK1/2 (Thr202/Tyr204) rabbit 211

polyclonal antibody (1:750, Cell Signaling Technology 9101), anti-STAT3 rabbit 212

monoclonal antibody (1:1000, Cell Signaling Technology 4904), anti-phospho-213

STAT3 (Tyr705) rabbit monoclonal antibody (1:1000, Cell Signaling Technology 214


10

9145), anti-CD44 mouse monoclonal antibody (1:1000, Cell Signaling Technology 215

3570), anti-NF-κB p65 (RELA) rabbit monoclonal antibody (1:1000, Cell Signaling 216

Technology 8242), anti-phospho-NF-κB p65 (P-RELA) rabbit monoclonal antibody 217

(1:1000, Cell Signaling Technology 3033) or anti-β-actin mouse monoclonal antibody 218

(1:15,000, Abcam ab6276-100). Secondary horseradish peroxidase linked anti-rabbit-219

IgG (1:10,000, GE Healthcare NA934V), horseradish peroxidase linked anti-mouse-220

IgG (1:10,000, GE Healthcare NA931V) and horseradish peroxidase linked anti-goat-221

IgG antibodies (1:10,000, Santa Cruz Biotechnology sc-2020) were used prior to 222

detection with an ECL Plus Western Blotting Detection System (GE Healthcare). 223

Quantitation of specific proteins of interest was performed using Quantity One 224

software (Bio-Rad). 225

Phospho-receptor tyrosine kinase (P-RTK) array 226

P-RTK arrays (R&D Systems ARY001) were performed according to manufacturer’s 227

instructions using 500 µg of HN5 or HN5-ER protein. P-RTKs of interest were 228

quantitated using Quantity One software (Bio-Rad). 229

To confirm the inhibitory action of R428 on Axl phosphorylation, HN5 and HN5-ER 230

cells were cultured overnight in DMEM + 0.5% FBS, and then treated with 10 µM 231

R428 for 2 h, prior to protein extraction and P-RTK array analysis as above. 232

P-Axl, IL-6 and IL-8 enzyme-linked immunosorbent assay (ELISA) 233

P-Axl expression in HN5 and HN5-ER cells was quantitated using a DuoSet IC 234

Intracellular Human Phospho-Axl ELISA Kit (R&D Systems DYC2228-2) and an 235

ELISA Development System Troubleshooting Pack (R&D Systems TSP01-B), 236

according to manufacturer’s instructions. Samples were run in duplicate with 17 µg of 237


11

cytoplasmic protein extract added per well of the 96 well ELISA plate. IL-6 or IL-8 238

secretion by HN5 and HN5-ER cells was quantitated using a Hu IL-6 239

Chemiluminescence ELISA Kit (Life Technologies KHC0069) or an IL-8 Human 240

ELISA Kit (Abcam ab100575), according to manufacturer’s instructions. HN5 and 241

HN5-ER cells were seeded at a density of 6.5 x 105 cells/well in 6 well plates (total 242

volume of 1.5 ml), and media harvested 48 h after plating, cleared by centrifugation 243

and analyzed in triplicate for IL-6 and IL-8 secretion. 244

Cell migration assay 245

Migration of HN5 and HN5-ER cells was measured using an xCELLigence real time 246

migration assay system (Roche). HN5 or HN5-ER cells were plated at 3.0 x 104 247

cells/well into the upper chamber of CIM-plate 16 xCELLigence plates (Roche 248

05665817001). DMEM + 20% FBS was used in the lower chambers as a 249

chemoattractant, while serum-free media (SFM) was the control for no cell migration. 250

Cell migration into the lower chamber was measured over 24 h and expressed as a 251

baseline cell index relative to SFM control migration. 252

For some studies, HN5-ER cells (3.0 x 105 cells/well) in 6 well plates were pre-treated 253

with 5 µM R428 or DMSO for 4 h, and cell migration assessed as described above. 254

cDNA microarray expression profiling of HN5 and HN5-ER cells 255

Total RNA was extracted from HN5 and HN5-ER cells using TRIzol reagent (Life 256

Technologies), and its quantity and integrity of extracted RNA confirmed using a 257

2100 Bioanalyzer (Agilent Technologies) before samples were deemed suitable for 258

microarray analysis. Gene expression profiling was performed with three biological 259

replicate samples for each cell line by the Australian Genome Research Facility 260


12

(AGRF; Australia) using HumanHT-12 v4 array chips (Illumina). Data normalization 261

was performed by AGRF and data analyzed as described previously (17). Parametric 262

two-tailed Student’s t-test was used to calculate significance of variation, fold change 263

was calculated as mean ratio. Probes with an unadjusted p-value <0.05 and an 264

absolute fold change >1.5 or more were defined as being differentially expressed 265

between HN5-ER and HN5 cells. Microarray expression data has been deposited in 266

the Gene Expression Omnibus (GEO) under Accession Number GSE49135. 267

Microarray data analysis 268

Clustering, volcano and scatter plots that showed the distributions and correlations of 269

differential gene expression across the HN5/HN5-ER cell line microarrays were 270

produced from normalized data using the R ‘graphics’ package. A heat map was 271

produced with the R ‘gplots’ package that showed a comparison of significant 272

differential gene expressions across the HN5/HN5-ER cell line microarrays. Ingenuity 273

Pathway Analysis software was used to define functional pathways affected by the 274

acquisition of erlotinib resistance. 275

HCC 827 NSCLC microarrays (GEO accession number: GSE38121, (20)) contained 276

information on gene expression upon acquisition of erlotinib resistance in HCC 827 277

cell lines, and were processed using the R ‘affy’ and ‘limma’ packages. Genes that 278

were differentially expressed in both erlotinib resistant cell lines (HCC 827 ER1, 279

HCC 827 ER2) relative to the parental HCC 827 cell line were identified and 280

compared to our HN5/HN5-ER expression data. Venn diagrams comparing 281

differential gene expression related to erlotinib resistance in both HNC and NSCLC 282

were produced using the R ‘vennDiagram’ package. 283


13

Overrepresentation of a set of predicted and validated miR-34a target genes involved 284

in EMT among all genes upregulated in HN5-ER cells was determined using 285

Ingenuity Pathway Analysis software. TargetScan (Version 6.0: November 2011) was 286

used for miR-34a target predictions. 287

Statistics and investigation of Axl in HNC patients 288

Results are presented as mean ± standard deviation (SD). Statistical significance was 289

calculated using the Student’s t-test (two-tailed, unpaired) and the level of 290

significance was set at p<0.05. Statistical analysis of RT-qPCR data was performed 291

using GenEx software (MultiD), and normality of data confirmed using the 292

Kolmogorov-Smirnov test (KS test). 293

To investigate the clinical significance of Axl mRNA overexpression, we analyzed 294

clinicopathologic and tumor RNA-seq expression data from 334 HNC patients as 295

shown in Table 1, that were downloaded from The Cancer Genome Atlas (TCGA) 296

(https://tcga-data.nci.nih.gov/tcga/tcgaDownload.jsp; accessed June 2013). Matched 297

clinical and mRNA expression data was available for 302 HNC patients. The 298

prognostic significance of Axl overexpression was assessed using the R packages 299

“Survival” and “Graphics” to perform Cox regressions with ties addressed by Efron’s 300

method and produce Kaplan-Meier survival curves, with all deaths used as an end 301

point. To avoid confounding, analyses were adjusted for age, gender and smoking 302

status. The validity of proportional hazards assumptions were assessed by scaled 303

Schoenfeld residuals (21). Dichotomous high and low Axl mRNA expression levels 304

were determined using a median cut point and the reads per kilobase per million 305

mapped reads (RPKM) method (22). 306


14

RESULTS 307

An erlotinib-resistant HN5 cell subline (HN5-ER) with an epithelial-308

mesenchymal phenotype and increased cell motility 309

To investigate the mechanisms associated with acquired EGFR inhibitor resistance in 310

HNC, we utilized HN5 cells, as they have EGFR gene amplification (23) and are 311

EGFR-dependent and highly sensitive to EGFR inhibition in vitro and in vivo (24). To 312

select for acquired erlotinib resistance, HN5 cells were exposed to gradually 313

increasing concentrations of erlotinib for a period of six months. Microscopic analysis 314

indicated that while parental HN5 cells (HN5) exhibited a distinctive epithelial 315

morphology, with tight cell-cell junctions, HN5-ER cells possess a mesenchymal, 316

fibroblast-like morphology with increased cell spreading (Figure 1A). We measured 317

erlotinib sensitivity using cell titre assays, and confirmed that HN5-ER cells had 318

>200-fold higher resistance to erlotinib than HN5 cells (Figure 1B; HN5 EC50 319

erlotinib = 0.1 μM, HN5-ER EC50 erlotinib >20 μM). We also found that HN5-ER 320

cells were cross-resistant to the EGFR TKI gefitinib (Supplementary Figure 1A). 321

Interestingly, HN5-ER cells maintained their erlotinib resistance and mesenchymal 322

cell morphology when grown for several weeks in erlotinib-free media 323

(Supplementary Figure 1B), suggesting that they had acquired stable resistance to 324

erlotinib. 325

326

Previous reports indicate that the inherent resistance of several different tumor types 327

to EGFR inhibitors, including erlotinib, is associated with EMT, where mesenchymal 328

tumor cells are typically more resistant to EGFR inhibition than epithelial tumor cells 329

(13). To gain an insight into the signaling mechanisms associated with acquired 330

erlotinib resistance, we analyzed the expression and activity of components of the 331


15

EGFR signaling pathway together with several established EMT-related markers in 332

HN5 and HN5-ER cells (Figure 1C). These studies revealed that acquired resistance 333

to erlotinib was associated with reduced EGFR expression and activity (P-EGFR), 334

increased Akt expression and activity (P-Akt), and reduced ERK1/2 activity (P-335

ERK1/2). Consistent with their epithelial phenotype, HN5 cells had strong E-cadherin 336

expression but did not express detectable levels of vimentin, an established 337

mesenchymal marker. Conversely, HN5-ER cells lacked E-cadherin protein 338

expression but expressed vimentin, confirming their mesenchymal cancer cell 339

phenotype. We confirmed these findings by analysing expression of EGFR, Akt, P-340

Akt and E-cadherin in an independent pool of erlotinib-resistant HN5 cells (HN5-341

ER2) by immunoblotting (Supplementary Figure 2), and observed changes in 342

expression and activity of these molecules consistent with those in HN5-ER cells. We 343

next used TaqMan microRNA (miRNA) RT-qPCR assays to measure the levels of 344

miR-200a, miR-200b, and miR-200c in HN5-ER cells, as this miRNA family has 345

been reported to promote epithelial differentiation and its expression is reduced with 346

EMT, tumor progression, and EGFR inhibitor resistance (25). These experiments 347

showed significantly lower expression of miR-200a/b/c in HN5-ER cells compared 348

with HN5 cells (Figure 1D), consistent with their EMT phenotype. As increased cell 349

migration, invasion and metastasis is a hallmark of mesenchymal tumor cells (26), we 350

performed xCELLigence real-time cell migration studies (27) and found that HN5-ER 351

cells had an increased rate of migration compared to parental HN5 cells (Figure 1E). 352

Taken together, our findings indicate that acquired resistance of HNC cells to erlotinib 353

is associated with development of an EMT phenotype, an increased level of oncogenic 354

Akt activity despite a reduced dependence on EGFR signaling, and increased cell 355

migration. 356


16

357

Acquired erlotinib resistance is associated with increased Axl expression and 358

activity, and decreased miR-34a expression 359

We hypothesized that compensatory activation of other RTKs allows HN5-ER cells to 360

maintain oncogenic Akt signaling in the presence of erlotinib. We used phospho-RTK 361

(P-RTK) arrays to simultaneously measure the activity (tyrosine phosphorylation) of 362

42 RTKs known to be associated with cancer. These experiments identified several 363

RTKs with altered activity between HN5 and HN5-ER cells (Figure 2A). We 364

observed decreased activity of ErbB2, ErbB3, VEGFR1, IGF1R and increased activity 365

of EphA7 and Axl in HN5-ER cells. The increased Axl activity in HN5-ER cells was 366

of particular interest in light of previous reports showing that Axl mediates EMT, 367

chemotherapeutic resistance, cell migration and invasion, and metastasis in breast 368

cancer, non-small cell lung cancer (NSCLC), and glioblastoma multiforme (28). We 369

confirmed the increase in Axl phosphorylation in HN5-ER cells by densitometry 370

(Figure 2B), and also using ELISA assays specific for P-Axl with protein extracts 371

from HN5 and HN5-ER cells (Figure 2C). Next, we compared total Axl protein 372

expression between HN5 and HN5-ER cells with immunoblotting assays, and 373

observed a substantial increase in total Axl levels in HN5-ER cells (Figure 2D). We 374

also found that Axl was abundantly expressed in FaDu HNC cells that are resistant to 375

erlotinib (Figure 2D; (17)). As an inverse association between the levels of Axl and 376

the tumor suppressor miRNA miR-34a has been reported in NSCLC, colorectal 377

cancer and breast cancer cell lines, and miR-34a was found to repress Axl expression 378

by its direct binding to the Axl mRNA 3’-untranslated region (3’-UTR) (29), we used 379

TaqMan miRNA assays to compare miR-34a levels between HN5 and HN5-ER cells. 380

We found miR-34a levels were significantly lower in HN5-ER cells than in HN5 cells 381


17

(Figure 2E). Next, we determined whether restoring miR-34a expression to HN5-ER 382

cells following transient transfection with synthetic miR-34a could alter Axl levels. 383

These experiments showed that addition of miR-34a decreased Axl expression, 384

confirming Axl as a miR-34a target in HNC cells; we also observed a concomitant 385

increase in E-cadherin expression, suggesting that miR-34a might also act as a 386

negative regulator of EMT in this system (Figure 2F). Finally, to determine the 387

functional significance of a loss of miR-34a expression to erlotinib resistance, we 388

transfected HN5-ER cells with miR-34a or miR-NC, followed by treatment with an 389

ineffective concentration of erlotinib (10 μM) or vehicle (DMSO). We found that 390

restoring miR-34a expression to HN5-ER cells sensitized them to erlotinib (Figure 391

2G). Together, these results indicate that the acquisition of erlotinib resistance in 392

HNC is associated with increased Axl expression and activity, and suggest that Axl 393

overexpression and erlotinib resistance may in part be due to a decrease in miR-34a 394

levels. 395

396

An Axl-specific inhibitor, R428, reduces HN5-ER cell viability, reverses acquired 397

erlotinib resistance, and blocks cell migration 398

To investigate the functional significance of increased Axl expression and activity, 399

we treated HN5-ER cells with R428 (Rigel Pharmaceuticals; now BerGenBio 400

BFB324), a selective small molecule Axl inhibitor that is in clinical development for 401

the treatment of cancers (30). We first confirmed that R428 inhibited Axl activity (P-402

Axl) in HN5-ER cells using P-RTK arrays (Supplementary Figure 3A). Cell titre 403

assays indicated that both HN5-ER and HN5 cell lines were sensitive to R428 (Figure 404

3A; HN5-ER EC50 R428 = 1.4 μM, HN5 EC50 R428 = 1.3 μM). Parental HN5 cells 405

are EGFR-dependent and thus highly sensitive to EGFR inhibitors, but lack 406


18

detectable expression of Axl (Figure 2D). R428 inhibits EGFR with >100-fold lower 407

selectivity than for Axl (30), and we found that treatment of HN5 cells with R428 (10 408

μM) for 4 h caused a significant reduction in P-EGFR levels (Supplementary Figure 409

3B), demonstrating that Axl and EGFR may be targets of R428 in HNC cells. To 410

determine whether Axl promotes resistance of HN5-ER cells to erlotinib, we 411

performed cell titre assays in which cells were treated with sub-effective 412

concentrations of R428 (ranging from 0.1-1 μM) in the presence or absence of an 413

ineffective concentration of erlotinib (10 μM). Erlotinib alone did not significantly 414

reduce cell viability, nor did R428 in the absence of erlotinib (Figure 3B). However, 415

the combination of erlotinib and R428 yielded a significant and dose-responsive 416

inhibition of cell viability. As a control, we confirmed the sensitivity of HN5-ER cells 417

to Axl inhibition by treatment with 5 μM R428, observing a >95% reduction in cell 418

viability (data not shown). To further investigate the combined growth inhibitory 419

effect of erlotinib and R428 on HN5-ER cell signaling pathways, we evaluated P-Akt 420

levels in cells treated with the same ineffective concentrations of erlotinib/DMSO, 421

R428/DMSO, and erlotinib/R428 used in Figure 3B. Each drug on its own caused a 422

modest reduction in P-Akt levels, but a much greater reduction in P-Akt was observed 423

when erlotinib was combined with R428 (Figure 3C), suggesting that R428 could 424

reverse the acquired resistance of HNC cells to erlotinib and thus re-sensitize them to 425

this agent. Similarly, we found that R428 sensitized HN5-ER cells to an ineffective 426

concentration of gefitinib (7.5 μM; Supplementary Figure 4), supporting the role of 427

Axl in mediating resistance to EGFR inhibition. To confirm this finding we used 428

RNAi to deplete HN5-ER cells of Axl expression (Supplementary Figure 5A and 5B), 429

and using cell titre assays we found that Axl knockdown increased the sensitivity of 430

HN5-ER cells to erlotinib (Supplementary Figure 5C). Finally, as we observed an 431


19

increased rate of migration of HN5-ER cells compared with HN5 cells (Figure 1E), 432

and as Axl promotes cancer cell migration, invasion and metastasis (28), we tested 433

the effect of R428 on HN5-ER cell migration. Our results indicate that R428 434

completely blocked HN5-ER cell migration over a 24 h period, with the measured 435

cell index for these samples being equivalent to the serum-free media (SFM; no 436

migration) controls (Figure 3D). These results suggest that Axl may be a valid 437

therapeutic target in HNCs that are resistant to EGFR inhibitors such as erlotinib, and 438

that Axl inhibition can restore erlotinib sensitivity and block tumor cell migration. 439

440

A pro-inflammatory gene signature is associated with acquired erlotinib 441

resistance in HN5-ER cells 442

To gain further insight into the molecular mechanisms driving acquired EGFR 443

inhibitor resistance in HNC, we performed comparative microarray analysis of HN5 444

and HN5-ER cells. We refined the lists of genes with differential expression between 445

HN5-ER and HN5 cells by assigning a cut off for upregulation or downregulation of 446

at least 1.5 fold and with a significance of p<0.05 (Figure 4A and 4B). Cluster 447

analysis of differentially expressed genes between HN5-ER and HN5 cells (Figure 448

4C) identified 247 mRNAs with significantly higher expression in HN5-ER than in 449

HN5 cells (Supplementary Table 1), and 626 mRNAs with significantly lower 450

expression in HN5-ER than in HN5 cells (Supplementary Table 2). We used 451

Ingenuity Pathway Analysis (IPA) software to assign biological significance to the 452

genes upregulated and downregulated with acquired erlotinib resistance, based on 453

annotated functional pathways (Supplementary Table 3). Pathways that were 454

significantly enriched for genes altered between HN5 and HN5-ER cells predicted the 455

activation of terms that include: “Cancer (tumorigenesis)”, “Cellular movement 456


20

(invasion of cells)”, “Cellular growth and proliferation (proliferation of cells)”, 457

“Inflammatory response (inflammation)” and “Cancer (head and neck cancer)”, and 458

decreased activation of pathways that included “Cell death (apoptosis of tumor cell 459

lines)” (Table 2). Therefore, in addition to alterations in cell growth and survival, 460

these findings suggest that acquired erlotinib resistance in HNC is associated with a 461

pro-inflammatory, pro-metastatic phenotype. 462

463

To confirm our microarray findings, we used RT-qPCR to measure the mRNA 464

expression of a panel of genes that were differentially expressed between HN5 and 465

HN5-ER cells: E-cadherin (epithelial marker), N-cadherin and ZEB1 (mesenchymal 466

markers), COX-2, IL-6, IL-8 (pro-inflammatory cytokines), EGFR and Axl. These 467

studies confirmed both the direction of change and significance for each of these 468

differentially expressed genes based on the microarray data (Figure 4D). As the 469

cytokines IL-6 and IL-8 were strongly upregulated at the mRNA level in HN5-ER 470

cells, and are associated with EMT and metastasis in various epithelial cancers, 471

including HNC (31, 32), we performed ELISA studies to compare IL-6 and IL-8 472

secretion from HN5 and HN5-ER cells. HN5-ER cells secreted significantly higher 473

levels of IL-6 and IL-8 into cell culture media than HN5 cells (Figure 4E), a finding 474

that is consistent with the respective ~100- and ~300-fold increase in IL-6 and IL-8 475

mRNA levels in HN5-ER cells (Figure 4D). We also observed increased STAT3 476

expression and activity (P-STAT3) in HN5-ER cells (Figure 4F), suggesting that IL-6 477

may be driving pro-survival STAT3 signaling in these cells. This was of particular 478

interest as STAT3 activation has recently been shown to promote resistance to EGFR 479

inhibitors in HNC (9). Using immunoblotting we compared NF-κB activity between 480

HN5 and HN5-ER cells, but did not detect elevated phosphorylation of RELA (NF-481


21

κB p65 subunit) in HN5-ER cells (Supplementary Figure 6A), possibly because both 482

EGFR and Axl may contribute to downstream NF-κB signaling. To investigate the 483

regulation of NF-kB activity in HN5-ER cells in more detail, we transfected HN5-ER 484

cells with miR-34a or miR-NC, or with Axl siRNAs or NC siRNA, and analyzed P-485

RELA levels by immunoblotting (Supplementary Figure 6B). We did not observe a 486

change in P-RELA activity with miR-34a transfection, suggesting that NF-κB activity 487

is not entirely dependent upon miR-34a or its downstream target Axl. We also 488

observed that relative to HN5 cells, HN5-ER cells have increased expression of CD44 489

(Figure 4F), a marker associated with elevated levels of IL-6 in erlotinib-resistant, 490

mesenchymal NSCLC cells (33). As R428 has been shown to suppress IL-6 491

expression in breast cancer cells (30), and Axl inhibition by RNAi impairs IL-6 492

expression and STAT3 activity in prostate cancer cells (34), we treated HN5-ER cells 493

with R428 (5 μM for 4 h) to assess the significance of Axl upregulation on STAT3 494

signaling by immunoblotting. We found that R428 blocked expression of both P-495

STAT3 and P-Akt in HN5-ER cells (Figure 4G), suggesting that Axl contributes to 496

elevated STAT3 activity. Furthermore, we treated HN5-ER cells with two agents that 497

inhibit the IL-6/STAT3 signaling axis: tocilizumab (1 μM for 3 d), a monoclonal 498

antibody against interleukin-6 receptor (IL-6R) (35) or ruxolitinib (3 μM for 24 h), a 499

small molecule inhibitor of Janus kinase 1/2 (JAK1/2) (36). Both of these agents 500

inhibited STAT3 activity in HN5-ER cells (Figure 4H) without altering Axl 501

expression (Supplementary Figure 6C), suggesting that inhibitors of IL-6/STAT3 502

signaling could be evaluated alone or together with therapies targeting Axl to 503

circumvent EGFR inhibitor resistance in HNC. Together, our findings identify 504

changes in cellular pathways associated with inflammation, growth, metastasis, and 505

cell death in HN5-ER cells, and suggest that elevated pro-inflammatory signaling 506


22

(Figure 5) is a feature of erlotinib-resistant HNC cells, which are CD44 positive and 507

have increased secretion of IL-6 and IL-8, elevated COX-2 expression, and activation 508

of STAT3 signaling. 509

510

A common gene signature of acquired erlotinib resistance between head and 511

neck cancer and non-small cell lung cancer 512

A recent report described activation of Axl in EGFR-mutant NSCLC cells with 513

acquired erlotinib resistance (20), a finding that is consistent with our observations in 514

HNC and suggests that common mechanisms of erlotinib resistance might exist 515

between HNC and NSCLC. To address this issue, we integrated our HNC microarray 516

data with that from the above-mentioned NSCLC study, which compared gene 517

expression between parental HCC 827 NSCLC cells and two sublines with acquired 518

erlotinib resistance: HCC 827 ER1 and HCC 827 ER2. Our analysis identified 95 519

genes whose expression was upregulated, and 75 genes whose expression was 520

downregulated, across all three erlotinib-resistant HNC and NSCLC cell lines (HN5-521

ER, HCC 827 ER1, HCC 827 ER2) (Figure 6). To assign biological significance to 522

these altered gene sets, we assigned them to functional pathways using IPA software 523

(Supplementary Table 4). We found that acquired erlotinib resistance in HNC and 524

NSCLC is associated with a common overrepresentation of genes involved with 525

biological terms that included “Cancer”, “Head and neck cancer”, “Cell movement of 526

tumor cell lines”, “Apoptosis of tumor cell lines”, “Proliferation of tumor cell lines”, 527

and “Inflammatory response” (Table 3). Expression of Axl, IL-6 and IL-8 was also 528

increased in HN5-ER, HCC 827 ER1 and HCC 827 ER2 cells relative to their 529

parental HNC and NSCLC cell lines. Therefore, some of the functional alterations 530

that mediate acquired resistance to erlotinib appear to be in common between HNC 531


23

and NSCLC. 532

533

Axl mRNA expression predicts survival in HNC patients 534

Overexpression of Axl is associated with a worse prognosis in various cancers, 535

including NSCLC, breast and pancreatic cancer, and acute myeloid leukemia (28). 536

We utilized publicly-available gene expression data from The Cancer Genome Atlas 537

(TCGA) with matching clinical information to investigate the prognostic value of Axl 538

mRNA expression in HNC. At the time of preparation of this manuscript, matched 539

data was available for 302 HNC patients (demographics are detailed in Table 1). 540

When data was adjusted for age, gender and smoking status, Axl mRNA expression 541

was significantly correlated with poor survival (Figure 7), with a 66% increased 542

incidence of death in HNC patients with high tumor Axl mRNA expression (HR 1.66, 543

p=0.007). This finding suggests that Axl may be associated with treatment resistance 544

and progression of HNC. 545

546

DISCUSSION 547

There is considerable evidence linking EMT to cancer progression, metastasis, and 548

EGFR inhibitor resistance in multiple tumor types (12, 37-41). Our results confirm 549

the acquisition of erlotinib resistance in HN5-ER cells is associated with an EMT 550

phenotype with fibroblastic cell morphology, increased cell migration, loss of E-551

cadherin and miR-200 family miRNA expression, and gain of vimentin and N-552

cadherin. Interestingly, HN5-ER cells have strong expression of CD44 and reduced 553

EGFR levels; this combination of markers was recently associated with a 554

subpopulation of HNC cells that are highly resistant to treatment with radiation, 555

cisplatin, cetuximab and gefitinib (42). We also observed increased Akt activity, but 556


24

reduced ERK1/2 activity, in HN5-ER cells. The latter finding was unexpected as 557

TGFβ-induced EMT in H358 NSCLC cells is reported to promote EGFR-independent 558

activation of ERK1/2 signaling (43), and SCC-1 HNC cells with acquired resistance 559

to cetuximab, gefitinib, or erlotinib display elevated ERK1/2 and Akt activity (10). 560

The precise mechanisms responsible for the increased Akt expression in HN5-ER 561

cells are unclear, but Akt gene amplification and overexpression has been described 562

in ovarian, pancreatic and breast cancers (44-46). Elevated Akt activity promotes 563

tumorigenesis and is inversely associated with E-cadherin expression and patient 564

survival in many cancers, including HNC (47), and the combination of a novel Akt 565

inhibitor, MK-2206, with erlotinib synergistically inhibits growth of NCI-H292 566

NSCLC cancer cells and tumor xenografts (48), implying that targeted inhibition of 567

Akt activity represents a potential strategy for the management of EGFR inhibitor-568

resistant HNCs. 569

570

We used P-RTK arrays to identify RTKs that compensated for loss of EGFR 571

dependence and promoted Akt activity in the context of acquired erlotinib resistance, 572

and reasoned that these RTKs could be targeted therapeutically in EGFR inhibitor-573

resistant HNC. The decreased activation of EGFR, ErbB2 and ErbB3 in HN5-ER 574

cells was consistent with findings in mesenchymal NSCLC cells resistant to EGFR 575

inhibition (43), whereas the reduced activity of VEGFR1 and IGF-1R in HN5-ER 576

cells is in contrast with reports suggesting that these RTKs mediate resistance to anti-577

EGFR therapies in colon, prostate, breast, lung and epidermoid cancer cells (49, 50). 578

While ligand-independent activation of c-Met has been shown to promote erlotinib 579

resistance in HNC (51), we did not observe altered c-Met activity in HN5-ER cells. 580

The significance of increased EphA7 activity in HN5-ER cells is unclear; the role of 581


25

Eph signaling in cancer appears to be complex and loss of EphA7 expression is 582

reported in certain cancers (52). Several lines of evidence support a role for elevated 583

Axl expression and activation in cancer progression. Firstly, Axl expression is 584

correlated with metastasis and poor prognosis across multiple tumor types (53-55). 585

Secondly, Axl activation promotes oncogenic processes such as cell growth, 586

migration, invasion, survival and angiogenesis (28). Thirdly, targeted inhibition of 587

Axl expression or activity reduces expression of EMT-associated molecules (eg. Slug, 588

Snail, Twist) and decreases tumor cell invasion (54, 56). Lastly, Axl confers 589

resistance to targeted and chemotherapeutic cancer drugs (57, 58), including ErbB-2 590

inhibitors in breast cancer (59) and erlotinib in NSCLC (20). A number of strategies 591

have been developed to inhibit Axl clinically (reviewed in (28)). R428 is an orally 592

bioavailable, potent and selective small molecule Axl TKI that inhibits several Axl-593

dependent processes in breast cancer cells, including Akt activation, cell invasion and 594

pro-inflammatory cytokine production, and represses EMT and angiogenesis and 595

extends survival in mice bearing metastatic breast tumors (30). In HN5-ER cells, 596

R428 reduced growth, blocked migration and restored erlotinib sensitivity. We also 597

observed that parental HN5 cells were growth inhibited by R428; this was unexpected 598

given the apparently undetectable expression of Axl protein in HN5 cells in 599

immunoblotting studies (Figure 2D). Several possibilities exist that may account for 600

this effect: (i) a low but detectable level of P-Axl is found in HN5 cells (Figure 2A 601

and 2B), (ii) R428 inhibits EGFR and other related RTKs with >100-fold less 602

selectivity than for Axl (30), and thus the apparent sensitivity of HN5 cells to R428 603

could result from this ‘off-target’ inhibition of P-EGFR in a highly EGFR-dependent 604

cell line, and (iii) the HN5 cell line contains a small mesenchymal subpopulation of 605

highly tumorigenic HN5-ER cells. In support of the latter hypothesis, Yao and 606


26

coworkers identified erlotinib-resistant mesenchymal-like cells within NSCLC cell 607

lines and tumors before erlotinib treatment (33), suggesting that acquired erlotinib 608

resistance might arise at least in part from selection for this mesenchymal 609

subpopulation. Our findings emphasize the role of Axl in promoting EMT, cell 610

migration and erlotinib resistance in HNC, and support the rationale for therapeutic 611

blockade of Axl in EGFR inhibitor-resistant tumors, including HNC. More broadly, 612

our results suggest that EGFR inhibitor resistance may arise via one or more distinct 613

mechanisms, underscoring the need to identify relevant biomarkers to properly select 614

cancer patients for appropriate targeted therapies. 615

616

Functional roles for specific miRNAs in the regulation of EMT are well established 617

(reviewed in (60)). In view of reports showing that miR-34a expression is reduced in 618

NSCLC, colorectal and breast cancer cell lines due to promoter methylation, that Axl 619

is a miR-34a target molecule (29), and that miR-34a is part of a miRNA expression 620

signature associated with epithelial morphology and erlotinib sensitivity in lung 621

cancer cells (61), we demonstrated loss of miR-34a in HN5-ER cells and found that 622

restoring miR-34a expression blocked Axl expression, increased levels of E-cadherin, 623

and sensitized cells to erlotinib. Decreased expression of miR-34a in colon, breast and 624

lung cancer cells is thought to result, in part, from loss of wild-type p53 function, and 625

is associated with increased expression of molecules known to promote EMT, 626

including Snail1, β-catenin, LEF1, and Axin2 (62). We propose that a loss of miR-627

34a facilitates coordinate expression of a set of target genes that maintain a 628

mesenchymal cancer cell phenotype associated with EGFR inhibitor resistance and 629

metastasis. Among the genes we found upregulated in HN5-ER cells relative to HN5 630

cells, nine are known to promote EMT and are predicted or experimentally validated 631


27

targets of miR-34a (Figure 7); these include: Axin2 (63), Axl (29), CA9 (64), 632

CXCL10 (65), FOSL1 (66), FUT8 (67), GAS1 (68), KLF6 (69), and PODXL (70). 633

Therefore, miR-34a may oppose EMT via the coordinate downregulation of a 634

network of miR-34a target mRNAs that promote EMT, metastasis and EGFR 635

inhibitor resistance. Therapeutic upregulation of miR-34a could reverse the metastatic 636

and EGFR inhibitor-resistant EMT phenotype in tumors, and may ultimately improve 637

treatment responses and patient survival. This may be a particularly attractive strategy 638

given the demonstrated feasibility of delivery and efficacy of miR-34a in preclinical 639

tumor models (71), and the recent commencement of a Phase I clinical trial of miR-640

34a in cancer patients (May 2013; Mirna Therapeutics, Austin, Texas). 641

642

Our study identified a putative role for pro-inflammatory signaling in erlotinib 643

resistance in HNC, whereby elevated synthesis and secretion of IL-6 and IL-8 may 644

activate STAT3 signaling and target gene expression to promote inflammation, cell 645

survival and metastasis (Figure 5). IL-6 is known to promote EMT and metastasis in 646

HNC (32), and its expression may be induced by NF-κB signaling, which in turn also 647

activates the STAT3 pathway (72). Interestingly, IL-6 secretion is regulated by Axl 648

signaling in animal tumor models (30), STAT3 activity is increased in SCC-1 HNC 649

cells with acquired resistance to EGFR inhibitors (10), and may suppress express of 650

miR-200 miRNAs (73), and increased IL-6 secretion promotes erlotinib resistance in 651

mesenchymal NSCLC cells (33). IL-6 may also induce EMT by blocking expression 652

of miR-200c in breast cancer cells (74). IL-8 induces EMT and activation of NF-κB 653

(75) and Akt activation in HNC, where it is associated with a poor prognosis (31). 654

Low expression of IL-8 is significantly associated with longer overall survival in 655

colorectal cancer patients treated with cetuximab (76), suggesting that IL-8 levels 656


28

might also have prognostic value in HNC patients receiving anti-EGFR therapy. 657

Finally, we also observed increased expression of the pro-inflammatory mediator 658

COX-2 in HN5-ER cells. COX-2 is overexpressed in a variety of human tumors, 659

including HNC (77), where it is implicated in cancer growth and progression and is 660

associated with a worse clinical outcome (reviewed in (78)). COX-2 inhibition has 661

been proposed as a strategy to overcome resistance of HNCs to EGFR inhibitors (79), 662

with combinations of EGFR TKIs and COX-2 inhibitors producing synergistic anti-663

tumor effects in preclinical studies with HNC (80). COX-2 induces a mesenchymal, 664

metastatic phenotype in NSCLC by repressing expression of E-cadherin and 665

promoting expression of genes such as Snail, ZEB1, and IL-6 (81), suggesting that it 666

might in part maintain the EMT phenotype of HN5-ER cells. Our findings implicate 667

pro-inflammatory mediators, including IL-6, IL-8 and COX-2, in driving EMT, 668

STAT3 signaling and EGFR inhibitor resistance in HNC, and suggest that targeted 669

inhibition of IL-6, IL-8, and/or COX-2 could sensitize CD44 positive, mesenchymal 670

HNC cells to anti-EGFR therapies. 671

672

Despite EGFR being widely expressed in HNC, EGFR inhibitors have produced 673

disappointing clinical results. A thorough understanding of the mechanisms behind 674

the inherent and acquired resistance of HNCs to anti-EGFR therapy is required to 675

better select patients most likely to benefit from these treatments and to design new 676

strategies to improve clinical responses. Based on our findings with an in vitro model 677

of erlotinib resistance in HNC, we propose a model (Figure 8) in which EMT, EGFR 678

inhibitor resistance and metastasis in HNC are mediated in part by increased 679

expression/activity of Axl, increased synthesis and secretion of IL-6, and reduced 680

expression of miR-34a. Inhibition of Axl activity with R428 decreased cell 681


29

proliferation, blocked migration, and restored erlotinib sensitivity. There are 682

prognostic and therapeutic implications of these molecules for the use of EGFR 683

targeted therapies in HNC. 684

685

ACKNOWLEDGEMENTS 686

The authors thank Dianne Beveridge for helpful discussion regarding the manuscript 687

and Vance Matthews for technical advice with ELISA experiments. 688

689

GRANT SUPPORT 690

This work was supported by the National Health and Medical Research Council 691

(NHMRC; Grant Number 634375) and the University of Western Australia Research 692

Development Award scheme (to K.M. Giles). K.M. Giles is supported by a Royal 693

Perth Hospital Medical Research Foundation fellowship. 694

695

ABBREVIATIONS 696

The abbreviations used are: miRNA, microRNA; mRNA, messenger RNA; HNC, 697

head and neck cancer; RT-qPCR, reverse transcriptase quantitative polymerase chain 698

reaction; 3′-UTR, 3′-untranslated region; miR-NC, negative control miRNA; SD, 699

standard deviation; GAPDH, glyceraldehyde 3-phosphate dehydrogenase; KS test, 700

Kolmogorov-Smirnov test; EGFR, epidermal growth factor receptor; IGF-1R, insulin-701

like growth factor 1 receptor; P-EGFR, phosphorylated EGFR, P-Akt, phosphorylated 702

Akt; ELISA, enzyme-linked immunosorbent assay; TKI, tyrosine kinase inhibitor; 703

RTK, receptor tyrosine kinase; P-RTK, phosphorylated RTK; EMT, epithelial-704

mesenchymal transition; NSCLC, non-small cell lung cancer; HN5-ER, erlotinib 705


30

resistant HN5 cells; snRNA, small nucleolar RNA; IPA, Ingenuity Pathway Analysis 706

software; IL-6, interleukin-6; IL-8, interleukin-8. 707


31

REFERENCES 708 1. Baba Y, Fujii M, Tokumaru Y, Kato Y. Present and Future of EGFR 709 Inhibitors for Head and Neck Squamous Cell Cancer. J Oncol. 2012;2012:986725. 710 2. Loeffler-Ragg J, Schwentner I, Sprinzl GM, Zwierzina H. EGFR inhibition as 711 a therapy for head and neck squamous cell carcinoma. Expert Opin Investig Drugs. 712 2008;17:1517-31. 713 3. Grandis JR, Zeng Q, Drenning SD, Tweardy DJ. Normalization of EGFR 714 mRNA levels following restoration of wild-type p53 in a head and neck squamous 715 cell carcinoma cell line. Int J Oncol. 1998;13:375-8. 716 4. Cohen EE. Role of epidermal growth factor receptor pathway-targeted therapy 717 in patients with recurrent and/or metastatic squamous cell carcinoma of the head and 718 neck. J Clin Oncol. 2006;24:2659-65. 719 5. Cohen EE, Kane MA, List MA, Brockstein BE, Mehrotra B, Huo D, et al. 720 Phase II trial of gefitinib 250 mg daily in patients with recurrent and/or metastatic 721 squamous cell carcinoma of the head and neck. Clin Cancer Res. 2005;11:8418-24. 722 6. Sheu JJ, Hua CH, Wan L, Lin YJ, Lai MT, Tseng HC, et al. Functional 723 genomic analysis identified epidermal growth factor receptor activation as the most 724 common genetic event in oral squamous cell carcinoma. Cancer Res. 2009;69:2568-725 76. 726 7. Erjala K, Sundvall M, Junttila TT, Zhang N, Savisalo M, Mali P, et al. 727 Signaling via ErbB2 and ErbB3 associates with resistance and epidermal growth 728 factor receptor (EGFR) amplification with sensitivity to EGFR inhibitor gefitinib in 729 head and neck squamous cell carcinoma cells. Clin Cancer Res. 2006;12:4103-11. 730 8. Yonesaka K, Zejnullahu K, Okamoto I, Satoh T, Cappuzzo F, Souglakos J, et 731 al. Activation of ERBB2 signaling causes resistance to the EGFR-directed therapeutic 732 antibody cetuximab. Sci Transl Med. 2011;3:99ra86. 733 9. Sen M, Joyce S, Panahandeh M, Li C, Thomas SM, Maxwell J, et al. 734 Targeting Stat3 abrogates EGFR inhibitor resistance in cancer. Clin Cancer Res. 735 2012;18:4986-96. 736 10. Benavente S, Huang S, Armstrong EA, Chi A, Hsu KT, Wheeler DL, et al. 737 Establishment and characterization of a model of acquired resistance to epidermal 738 growth factor receptor targeting agents in human cancer cells. Clin Cancer Res. 739 2009;15:1585-92. 740 11. Young NR, Liu J, Pierce C, Wei TF, Grushko T, Olopade OI, et al. Molecular 741 phenotype predicts sensitivity of squamous cell carcinoma of the head and neck to 742 epidermal growth factor receptor inhibition. Mol Oncol. 2012. 743 12. Haddad Y, Choi W, McConkey DJ. Delta-crystallin enhancer binding factor 1 744 controls the epithelial to mesenchymal transition phenotype and resistance to the 745 epidermal growth factor receptor inhibitor erlotinib in human head and neck 746 squamous cell carcinoma lines. Clin Cancer Res. 2009;15:532-42. 747 13. Yauch RL, Januario T, Eberhard DA, Cavet G, Zhu W, Fu L, et al. Epithelial 748 versus mesenchymal phenotype determines in vitro sensitivity and predicts clinical 749 activity of erlotinib in lung cancer patients. Clin Cancer Res. 2005;11:8686-98. 750 14. Grille SJ, Bellacosa A, Upson J, Klein-Szanto AJ, van Roy F, Lee-Kwon W, 751 et al. The protein kinase Akt induces epithelial mesenchymal transition and promotes 752 enhanced motility and invasiveness of squamous cell carcinoma lines. Cancer Res. 753 2003;63:2172-8. 754 15. Easty DM, Easty GC, Carter RL, Monaghan P, Butler LJ. Ten human 755 carcinoma cell lines derived from squamous carcinomas of the head and neck. Br J 756 Cancer. 1981;43:772-85. 757


32

16. Rangan SR. A new human cell line (FaDu) from a hypopharyngeal carcinoma. 758 Cancer. 1972;29:117-21. 759 17. Kalinowski FC, Giles KM, Candy PA, Ali A, Ganda C, Epis MR, et al. 760 Regulation of epidermal growth factor receptor signaling and erlotinib sensitivity in 761 head and neck cancer cells by miR-7. PLoS One. 2012;7:e47067. 762 18. Giles KM, Barker A, Zhang PM, Epis MR, Leedman PJ. MicroRNA 763 regulation of growth factor receptor signaling in human cancer cells. Methods Mol 764 Biol. 2011;676:147-63. 765 19. Livak KJ, Schmittgen TD. Analysis of relative gene expression data using 766 real-time quantitative PCR and the 2(-Delta Delta C(T)) Method. Methods. 767 2001;25:402-8. 768 20. Zhang Z, Lee JC, Lin L, Olivas V, Au V, LaFramboise T, et al. Activation of 769 the AXL kinase causes resistance to EGFR-targeted therapy in lung cancer. Nat 770 Genet. 2012;44:852-60. 771 21. Grambsch PM, Therneau TM. Proportional Hazards Tests and Diagnostics 772 Based on Weighted Residuals. Biometrika. 1994;81:515-26. 773 22. Mortazavi A, Williams BA, McCue K, Schaeffer L, Wold B. Mapping and 774 quantifying mammalian transcriptomes by RNA-Seq. Nature methods. 2008;5:621-8. 775 23. Kwok TT, Sutherland RM. Differences in EGF related radiosensitisation of 776 human squamous carcinoma cells with high and low numbers of EGF receptors. Br J 777 Cancer. 1991;64:251-4. 778 24. Rusnak DW, Lackey K, Affleck K, Wood ER, Alligood KJ, Rhodes N, et al. 779 The effects of the novel, reversible epidermal growth factor receptor/ErbB-2 tyrosine 780 kinase inhibitor, GW2016, on the growth of human normal and tumor-derived cell 781 lines in vitro and in vivo. Mol Cancer Ther. 2001;1:85-94. 782 25. Adam L, Zhong M, Choi W, Qi W, Nicoloso M, Arora A, et al. miR-200 783 expression regulates epithelial-to-mesenchymal transition in bladder cancer cells and 784 reverses resistance to epidermal growth factor receptor therapy. Clin Cancer Res. 785 2009;15:5060-72. 786 26. Guarino M, Rubino B, Ballabio G. The role of epithelial-mesenchymal 787 transition in cancer pathology. Pathology. 2007;39:305-18. 788 27. Limame R, Wouters A, Pauwels B, Fransen E, Peeters M, Lardon F, et al. 789 Comparative analysis of dynamic cell viability, migration and invasion assessments 790 by novel real-time technology and classic endpoint assays. PLoS One. 2012;7:e46536. 791 28. Linger RM, Keating AK, Earp HS, Graham DK. Taking aim at Mer and Axl 792 receptor tyrosine kinases as novel therapeutic targets in solid tumors. Expert Opin 793 Ther Targets. 2010;14:1073-90. 794 29. Mudduluru G, Ceppi P, Kumarswamy R, Scagliotti GV, Papotti M, Allgayer 795 H. Regulation of Axl receptor tyrosine kinase expression by miR-34a and miR-199a/b 796 in solid cancer. Oncogene. 2011;30:2888-99. 797 30. Holland SJ, Pan A, Franci C, Hu Y, Chang B, Li W, et al. R428, a selective 798 small molecule inhibitor of Axl kinase, blocks tumor spread and prolongs survival in 799 models of metastatic breast cancer. Cancer Res. 2010;70:1544-54. 800 31. Li XJ, Peng LX, Shao JY, Lu WH, Zhang JX, Chen S, et al. As an 801 independent unfavorable prognostic factor, IL-8 promotes metastasis of 802 nasopharyngeal carcinoma through induction of epithelial-mesenchymal transition 803 and activation of AKT signaling. Carcinogenesis. 2012;33:1302-9. 804 32. Yadav A, Kumar B, Datta J, Teknos TN, Kumar P. IL-6 promotes head and 805 neck tumor metastasis by inducing epithelial-mesenchymal transition via the JAK-806 STAT3-SNAIL signaling pathway. Mol Cancer Res. 2011;9:1658-67. 807


33

33. Yao Z, Fenoglio S, Gao DC, Camiolo M, Stiles B, Lindsted T, et al. TGF-beta 808 IL-6 axis mediates selective and adaptive mechanisms of resistance to molecular 809 targeted therapy in lung cancer. Proc Natl Acad Sci U S A. 2010;107:15535-40. 810 34. Paccez JD, Vasques GJ, Correa RG, Vasconcellos JF, Duncan K, Gu X, et al. 811 The receptor tyrosine kinase Axl is an essential regulator of prostate cancer 812 proliferation and tumor growth and represents a new therapeutic target. Oncogene. 813 2013;32:689-98. 814 35. Venkiteshwaran A. Tocilizumab. mAbs. 2009;1:432-8. 815 36. Seavey MM, Dobrzanski P. The many faces of Janus kinase. Biochem 816 Pharmacol. 2012;83:1136-45. 817 37. Buck E, Eyzaguirre A, Barr S, Thompson S, Sennello R, Young D, et al. Loss 818 of homotypic cell adhesion by epithelial-mesenchymal transition or mutation limits 819 sensitivity to epidermal growth factor receptor inhibition. Mol Cancer Ther. 820 2007;6:532-41. 821 38. Fuchs BC, Fujii T, Dorfman JD, Goodwin JM, Zhu AX, Lanuti M, et al. 822 Epithelial-to-mesenchymal transition and integrin-linked kinase mediate sensitivity to 823 epidermal growth factor receptor inhibition in human hepatoma cells. Cancer Res. 824 2008;68:2391-9. 825 39. McConkey DJ, Choi W, Marquis L, Martin F, Williams MB, Shah J, et al. 826 Role of epithelial-to-mesenchymal transition (EMT) in drug sensitivity and metastasis 827 in bladder cancer. Cancer Metastasis Rev. 2009;28:335-44. 828 40. Thiery JP. Epithelial-mesenchymal transitions in tumour progression. Nat Rev 829 Cancer. 2002;2:442-54. 830 41. Thomson S, Buck E, Petti F, Griffin G, Brown E, Ramnarine N, et al. 831 Epithelial to mesenchymal transition is a determinant of sensitivity of non-small-cell 832 lung carcinoma cell lines and xenografts to epidermal growth factor receptor 833 inhibition. Cancer Res. 2005;65:9455-62. 834 42. La Fleur L, Johansson AC, Roberg K. A CD44high/EGFRlow subpopulation 835 within head and neck cancer cell lines shows an epithelial-mesenchymal transition 836 phenotype and resistance to treatment. PLoS One. 2012;7:e44071. 837 43. Thomson S, Petti F, Sujka-Kwok I, Epstein D, Haley JD. Kinase switching in 838 mesenchymal-like non-small cell lung cancer lines contributes to EGFR inhibitor 839 resistance through pathway redundancy. Clin Exp Metastasis. 2008;25:843-54. 840 44. Bellacosa A, de Feo D, Godwin AK, Bell DW, Cheng JQ, Altomare DA, et al. 841 Molecular alterations of the AKT2 oncogene in ovarian and breast carcinomas. Int J 842 Cancer. 1995;64:280-5. 843 45. Nakatani K, Thompson DA, Barthel A, Sakaue H, Liu W, Weigel RJ, et al. 844 Up-regulation of Akt3 in estrogen receptor-deficient breast cancers and androgen-845 independent prostate cancer lines. J Biol Chem. 1999;274:21528-32. 846 46. Cheng JQ, Ruggeri B, Klein WM, Sonoda G, Altomare DA, Watson DK, et al. 847 Amplification of AKT2 in human pancreatic cells and inhibition of AKT2 expression 848 and tumorigenicity by antisense RNA. Proc Natl Acad Sci U S A. 1996;93:3636-41. 849 47. Lim J, Kim JH, Paeng JY, Kim MJ, Hong SD, Lee JI, et al. Prognostic value 850 of activated Akt expression in oral squamous cell carcinoma. J Clin Pathol. 851 2005;58:1199-205. 852 48. Hirai H, Sootome H, Nakatsuru Y, Miyama K, Taguchi S, Tsujioka K, et al. 853 MK-2206, an allosteric Akt inhibitor, enhances antitumor efficacy by standard 854 chemotherapeutic agents or molecular targeted drugs in vitro and in vivo. Mol Cancer 855 Ther. 2010;9:1956-67. 856


34

49. Bianco R, Rosa R, Damiano V, Daniele G, Gelardi T, Garofalo S, et al. 857 Vascular endothelial growth factor receptor-1 contributes to resistance to anti-858 epidermal growth factor receptor drugs in human cancer cells. Clin Cancer Res. 859 2008;14:5069-80. 860 50. Tandon R, Kapoor S, Vali S, Senthil V, Nithya D, Venkataramanan R, et al. 861 Dual epidermal growth factor receptor (EGFR)/insulin-like growth factor-1 receptor 862 (IGF-1R) inhibitor: a novel approach for overcoming resistance in anticancer 863 treatment. Eur J Pharmacol. 2011;667:56-65. 864 51. Stabile LP, He G, Lui VW, Henry C, Gubish CT, Joyce S, et al. c-Src 865 Activation Mediates Erlotinib Resistance in Head and Neck Cancer by Stimulating c-866 Met. Clin Cancer Res. 2013;19:380-92. 867 52. Pasquale EB. Eph receptors and ephrins in cancer: bidirectional signalling and 868 beyond. Nat Rev Cancer. 2010;10:165-80. 869 53. Gjerdrum C, Tiron C, Hoiby T, Stefansson I, Haugen H, Sandal T, et al. Axl is 870 an essential epithelial-to-mesenchymal transition-induced regulator of breast cancer 871 metastasis and patient survival. Proc Natl Acad Sci U S A. 2010;107:1124-9. 872 54. Koorstra JB, Karikari CA, Feldmann G, Bisht S, Rojas PL, Offerhaus GJ, et 873 al. The Axl receptor tyrosine kinase confers an adverse prognostic influence in 874 pancreatic cancer and represents a new therapeutic target. Cancer Biol Ther. 875 2009;8:618-26. 876 55. Shieh YS, Lai CY, Kao YR, Shiah SG, Chu YW, Lee HS, et al. Expression of 877 axl in lung adenocarcinoma and correlation with tumor progression. Neoplasia. 878 2005;7:1058-64. 879 56. Zhang YX, Knyazev PG, Cheburkin YV, Sharma K, Knyazev YP, Orfi L, et 880 al. AXL is a potential target for therapeutic intervention in breast cancer progression. 881 Cancer Res. 2008;68:1905-15. 882 57. Hong CC, Lay JD, Huang JS, Cheng AL, Tang JL, Lin MT, et al. Receptor 883 tyrosine kinase AXL is induced by chemotherapy drugs and overexpression of AXL 884 confers drug resistance in acute myeloid leukemia. Cancer Lett. 2008;268:314-24. 885 58. Mahadevan D, Cooke L, Riley C, Swart R, Simons B, Della Croce K, et al. A 886 novel tyrosine kinase switch is a mechanism of imatinib resistance in gastrointestinal 887 stromal tumors. Oncogene. 2007;26:3909-19. 888 59. Liu L, Greger J, Shi H, Liu Y, Greshock J, Annan R, et al. Novel mechanism 889 of lapatinib resistance in HER2-positive breast tumor cells: activation of AXL. Cancer 890 Res. 2009;69:6871-8. 891 60. Bullock MD, Sayan AE, Packham GK, Mirnezami AH. MicroRNAs: critical 892 regulators of epithelial to mesenchymal (EMT) and mesenchymal to epithelial 893 transition (MET) in cancer progression. Biol Cell. 2012;104:3-12. 894 61. Bryant JL, Britson J, Balko JM, Willian M, Timmons R, Frolov A, et al. A 895 microRNA gene expression signature predicts response to erlotinib in epithelial 896 cancer cell lines and targets EMT. Br J Cancer. 2012;106:148-56. 897 62. Kim NH, Kim HS, Li XY, Lee I, Choi HS, Kang SE, et al. A p53/miRNA-34 898 axis regulates Snail1-dependent cancer cell epithelial-mesenchymal transition. J Cell 899 Biol. 2011;195:417-33. 900 63. Cha YH, Kim NH, Park C, Lee I, Kim HS, Yook JI. MiRNA-34 intrinsically 901 links p53 tumor suppressor and Wnt signaling. Cell Cycle. 2012;11:1273-81. 902 64. Lock FE, McDonald PC, Lou Y, Serrano I, Chafe SC, Ostlund C, et al. 903 Targeting carbonic anhydrase IX depletes breast cancer stem cells within the hypoxic 904 niche. Oncogene. 2012. 905


35

65. Martins VL, Vyas JJ, Chen M, Purdie K, Mein CA, South AP, et al. Increased 906 invasive behaviour in cutaneous squamous cell carcinoma with loss of basement-907 membrane type VII collagen. J Cell Sci. 2009;122:1788-99. 908 66. Yang S, Li Y, Gao J, Zhang T, Li S, Luo A, et al. MicroRNA-34 suppresses 909 breast cancer invasion and metastasis by directly targeting Fra-1. Oncogene. 2012. 910 67. Lin H, Wang D, Wu T, Dong C, Shen N, Sun Y, et al. Blocking core 911 fucosylation of TGF-beta1 receptors downregulates their functions and attenuates the 912 epithelial-mesenchymal transition of renal tubular cells. Am J Physiol Renal Physiol. 913 2011;300:F1017-25. 914 68. Katoh Y, Katoh M. Hedgehog signaling, epithelial-to-mesenchymal transition 915 and miRNA (review). Int J Mol Med. 2008;22:271-5. 916 69. Holian J, Qi W, Kelly DJ, Zhang Y, Mreich E, Pollock CA, et al. Role of 917 Kruppel-like factor 6 in transforming growth factor-beta1-induced epithelial-918 mesenchymal transition of proximal tubule cells. Am J Physiol Renal Physiol. 919 2008;295:F1388-96. 920 70. Meng X, Ezzati P, Wilkins JA. Requirement of podocalyxin in TGF-beta 921 induced epithelial mesenchymal transition. PLoS One. 2011;6:e18715. 922 71. Wiggins JF, Ruffino L, Kelnar K, Omotola M, Patrawala L, Brown D, et al. 923 Development of a lung cancer therapeutic based on the tumor suppressor microRNA-924 34. Cancer Res. 2010;70:5923-30. 925 72. Squarize CH, Castilho RM, Sriuranpong V, Pinto DS, Jr., Gutkind JS. 926 Molecular cross-talk between the NFkappaB and STAT3 signaling pathways in head 927 and neck squamous cell carcinoma. Neoplasia. 2006;8:733-46. 928 73. Guo L, Chen C, Shi M, Wang F, Chen X, Diao D, et al. Stat3-coordinated Lin-929 28-let-7-HMGA2 and miR-200-ZEB1 circuits initiate and maintain oncostatin M-930 driven epithelial-mesenchymal transition. Oncogene. 2013. 931 74. Rokavec M, Wu W, Luo JL. IL6-mediated suppression of miR-200c directs 932 constitutive activation of inflammatory signaling circuit driving transformation and 933 tumorigenesis. Molecular cell. 2012;45:777-89. 934 75. Waugh DJ, Wilson C. The interleukin-8 pathway in cancer. Clin Cancer Res. 935 2008;14:6735-41. 936 76. Vallbohmer D, Zhang W, Gordon M, Yang DY, Yun J, Press OA, et al. 937 Molecular determinants of cetuximab efficacy. J Clin Oncol. 2005;23:3536-44. 938 77. Chan G, Boyle JO, Yang EK, Zhang F, Sacks PG, Shah JP, et al. 939 Cyclooxygenase-2 expression is up-regulated in squamous cell carcinoma of the head 940 and neck. Cancer Res. 1999;59:991-4. 941 78. Aggarwal BB, Shishodia S, Sandur SK, Pandey MK, Sethi G. Inflammation 942 and cancer: how hot is the link? Biochem Pharmacol. 2006;72:1605-21. 943 79. Kao J, Sikora AT, Fu S. Dual EGFR and COX-2 inhibition as a novel 944 approach to targeting head and neck squamous cell carcinoma. Curr Cancer Drug 945 Targets. 2009;9:931-7. 946 80. Chen Z, Zhang X, Li M, Wang Z, Wieand HS, Grandis JR, et al. 947 Simultaneously targeting epidermal growth factor receptor tyrosine kinase and 948 cyclooxygenase-2, an efficient approach to inhibition of squamous cell carcinoma of 949 the head and neck. Clin Cancer Res. 2004;10:5930-9. 950 81. Dohadwala M, Yang SC, Luo J, Sharma S, Batra RK, Huang M, et al. 951 Cyclooxygenase-2-dependent regulation of E-cadherin: prostaglandin E(2) induces 952 transcriptional repressors ZEB1 and snail in non-small cell lung cancer. Cancer Res. 953 2006;66:5338-45. 954 955


36

TABLES 956 957

Table 1: Clinical information for TCGA HNC patient cohort. 958

959

Patient feature High Axl (n) Low Axl (n)

Total (n)

All cases 334 (100%) Axl evaluated 153 (50.5%) 150 (49.5%) 303 (90.7%) Gender Male 109 (36%) 113 (37.2%) 240 (71.9%) Female 45 (14.9%) 36 (11.9%) 94 (28.1%)

Age at diagnosis ≤ 59 66 (21.8%) 61 (20.1%) 142 (42.5%) 60-69 50 (16.5%) 59 (19.5%) 118 (35.3%) > 70 38 (12.5%) 29 (9.6%) 73 (22%)

Smoking status Current smoker 48 (15.8%) 52 (17.2%) 113 (33.8%) Quit smoking ≤ 15 years ago 54 (17.8%) 31 (10.2%) 92 (27.6%) Quit smoking > 15 years ago 22 (7.3%) 25 (8.3%) 56 (16.8%) Lifelong non-smoker 24 (7.9%) 37 (12.2%) 63 (18.9%) Unknown 6 (2%) 4 (1.3%) 10 (3%)

Alcohol consumption 0-1 drinks/day 21 (6.9%) 22 (7.3%) 42 (13.9%) 2-3 drinks/day 22 (7.3%) 12 (4%) 36 (11.9%) ≥ 4 drinks/day 21 (6.9%) 23 (7.6%) 50 (16.5%) Unknown 90 (29.7%) 92 (30.4%) 195 (64.4%) Cancer stage I 7 (2.3%) 3 (1%) 15 (4.5%) II 34 (11.2%) 28 (9.2%) 68 (20.4%) III 29 (9.6%) 38 (12.5%) 75 (22.5%) IV 84 (27.7%) 80 (26.4%) 176 (52.7%)

Survival status Living 88 (29%) 95 (31.4%) 210 (62.9%) Deceased 66 (21.8%) 54 (17.8%) 124 (37.1%)

960

Clinicopathologic features relative to Axl mRNA expression levels in HNC 961


37

Table 2: Functional pathways altered with acquired erlotinib resistance in head 962

and neck cancer cells. 963

964

Functional Pathways

Number of molecules affected

Predicted activation

state p-value Cancer (tumorigenesis) 306/7483 (4%) Increased 5.56x10-21 Cellular movement (invasion of cells) 60/3046 (2%) Increased 1.50x10-11 Cellular growth and proliferation (proliferation of cells) 127/5733 (2%) Increased 6.20x10-9 Cell death (apoptosis of tumor cell lines) 81/4535 (2%) Decreased 1.17x10-5 Inflammatory response (inflammation) 57/1142 (5%) Increased 2.45x10-5 Cancer (head and neck cancer) 47/818 (6%) Increased 1.45x10-4

965

A series of cancer-associated pathways altered in HNC cells with acquired resistance 966

to erlotinib, including functions related to tumorigenesis, head and neck cancer, cell 967

growth and death, and inflammation were identified using Ingenuity Pathway 968

Analysis (IPA) software. The total number of molecules upregulated or 969

downregulated in HN5-ER cells relative to HN5 cells is indicated as a proportion of 970

the total number of genes associated with that function, as is the predicted activation 971

state of that pathway in HN5-ER cells. 972

973

974

975

976


38

Table 3: Functional pathways altered with acquired erlotinib resistance in both 977

head and neck cancer and non-small cell lung cancer cells. 978

979

Functional pathway

Upregulated genes in HNC-ER and NSCLC-ER

Downregulated genes in HNC-ER and

NSCLC-ER

Cancer 58/95 (p=3.03 x 10-10) 27/75 (p=4.70 x 10-2)

Head and neck cancer 10/95 (p=3.29 x 10-3) 7/75 (p=1.27 x 10-2)

Cell movement of tumor cell lines 19/95 (p=9.4 x 10-8) 7/75 (p=3.88 x 10-2)

Apoptosis of tumor cell lines 16/95 (p=3.68 x 10-3) 10/75 (p=3.86 x 10-2)

Proliferation of tumor cell lines 19/95 (p=2.1 x 10-3) N.E.

Inflammatory response 6/95 (p=1.83 x 10-2) N.E.

N.E., no enrichment of functional pathway among that gene set. 980

981

Gene expression signatures related to acquired erlotinib resistance that were common 982

to both HNC and NSCLC were assigned biological significance using IPA software, 983

with a focus on functional pathways that were identified initially in erlotinib-resistant 984

HNC (Table 2). 95 genes were commonly upregulated, and 75 genes were commonly 985

downregulated, with acquired erlotinib resistance in HNC and NSCLC. The 986

proportion of each gene set implicated in these pathways is indicated in the table. 987

988

989

990

991

992


39

FIGURE LEGENDS 993 Figure 1. Acquired erlotinib resistance in HNC cells is associated with 994

development of an EMT phenotype, altered EGFR pathway signaling, and 995

increased cell migration. 996

(A) Photomicrograph of HN5 (top) and HN5-ER (bottom) cells at 100x total 997

magnification. (B) Cell titre analysis of erlotinib sensitivity of HN5 and HN5-ER 998

cells. Cells were seeded in 96-well plates, treated with the EGFR inhibitor erlotinib 999

(final concentration 0-100 μM), and the half maximal effective concentration (EC50) 1000

of erlotinib was determined for each cell line 3 d after the addition of erlotinib. Data 1001

are normalized to the lowest concentration of erlotinib. (C) Immunoblotting analysis 1002

comparing expression and activity of EGFR signaling pathway molecules (EGFR, P-1003

EGFR, Akt, P-Akt, ERK1/2, P-ERK1/2) and expression of EMT markers (E-cadherin, 1004

vimentin) between HN5 and HN5-ER cells. β-actin is a loading control. (D) TaqMan 1005

miRNA RT-qPCR analysis of miR-200a/b/c expression in HN5 and HN5-ER cells. 1006

Data was normalized to U44 snRNA expression and expressed relative to HN5. (E) 1007

Real time xCELLigence analysis of migration (represented by cell index) of HN5 1008

(black) and HN5-ER (red) cells. Serum free media (SFM; no migration) controls are 1009

included. 1010

Error bars represent standard deviations. Data are representative of three independent 1011

experiments. *, p<0.05; **, p<0.01, HN5-ER vs HN5. 1012

1013

Figure 2. Increased expression and activity of Axl and decreased miR-34a 1014

expression in HN5-ER cells. 1015

(A) Phosphorylated RTK (P-RTK) analysis of protein lysates from HN5 (top) and 1016

HN5-ER (bottom) cells. Spots of interest are boxed and numbered as follows: 1: 1017

positive control, 2: P-EGFR, 3: P-ErbB2, 4: P-ErbB3, 5: P-EphA7, 6: P-VEGFR1, 7: 1018


40

P-IGF-1R, 8: P-Axl. (B) Densitometry analysis of P-Axl pixel density based on P-1019

RTK profiling of HN5 and HN5-ER protein lysates. Mean pixel density across 1020

duplicate spots is expressed relative to HN5 cells. (C) ELISA quantitation of P-Axl 1021

expression in HN5 and HN5-ER cell lysates. Data is expressed relative to HN5 cells. 1022

(D) Immunoblotting analysis of Axl expression in HN5, HN5-ER and FaDu cells. 1023

β-actin is included as a loading control. (E) TaqMan miRNA RT-qPCR analysis of 1024

miR-34a expression in HN5 and HN5-ER cells. Data was normalized to U44 snRNA 1025

expression and expressed relative to HN5. (F) Immunoblotting analysis of Axl and E-1026

cadherin expression in HN5-ER cells 3 d after transfection with either synthetic miR-1027

34a, a negative control synthetic miRNA (miR-NC), or mock transfection (vehicle, 1028

LF2000). β-actin is a loading control. (G) Cell titre analysis of HN5-ER cells 1029

transfected with miR-34a (grey bars) or miR-NC (black bars) (0.5-1.0 nM) or LF2000 1030

(white bars) for 3 d and then treated for a further 3 d with either erlotinib (10 μM) or 1031

vehicle (DMSO). Data is expressed relative to LF2000/DMSO-treated HN5-ER cells. 1032


experiments. *, p<0.02, HN5-ER vs HN5; **, p<0.01, HN5-ER vs HN5; ***, 1034

p<0.001, miR-34a/erlotinib vs miR-NC/erlotinib; ****, p<0.01, miR-34a/erlotinib vs 1035

miR-34a/erlotinib. 1036

1037

Figure 3. Axl inhibitor R428 inhibits HN5-ER cell growth and migration and 1038

restores erlotinib sensitivity. 1039

(A) Cell titre analysis of sensitivity of HN5 and HN5-ER cells to R428. Cells were 1040

seeded in 96-well plates, treated with the Axl inhibitor R428 (final concentration 0-1041

100 μM) for 3 d, and the half maximal effective concentration (EC50) of R428 was 1042

determined for each cell line. Data are normalized to the lowest concentration of 1043


41

R428. (B) Cell titre analysis of HN5-ER cells measured 3 d after treatment with an 1044

ineffective concentration of erlotinib (10 μM) or vehicle (DMSO) and increasing 1045

concentrations (0-1.0 μM) of the Axl inhibitor R428. Grey bars indicate the absence 1046

of erlotinib, black bars indicate the presence of erlotinib. (C) Immunoblotting analysis 1047

of Akt expression and activity of HN5 and HN5-ER cells treated with sub-effective 1048

concentrations of erlotinib only, R428 only, and the combination of each drug (as per 1049

(B)). β-actin is a loading control. (D) Real time xCELLigence analysis of migration 1050

(represented by cell index) of HN5-ER cells that were pre-treated with R428 (black) 1051

or vehicle (DMSO; red). Serum free media (SFM; no migration) controls are included 1052

for each treatment (vehicle/SFM: blue; R428/SFM: green). 1053


experiments. *, p<0.001, R428/erlotinib vs vehicle/erlotinib. 1055

1056

Figure 4. Differential gene expression between HN5 and HN5-ER cells. 1057

Following microarray data normalization (see Materials and Methods), volcano (A) 1058

and scatter (B) plots and a heat map (C) were generated showing the distributions, 1059

correlations and replicate consistency of genes with significantly altered expression 1060

between HN5 and HN5-ER samples. (D) Microarray validation by RT-qPCR analysis 1061

of E-cadherin, N-cadherin, COX-2, EGFR, Axl, ZEB1, IL-6, and IL-8 mRNA 1062

expression in HN5 and HN5-ER cells. Data were normalized to ALAS1 and GAPDH 1063

mRNA expression and expressed relative to HN5. (E) ELISA analysis of IL-6 (left) 1064

and IL-8 (right) secretion into cell culture media. Data were expressed relative to 1065

HN5. (F) Immunoblotting analysis of STAT3, P-STAT3 and CD44 expression in 1066

HN5 and HN5-ER cells. β-actin is included as a loading control. (G) Immunoblotting 1067

analysis of P-STAT3, STAT3, P-Akt, and Akt levels in HN5-ER cells treated with 1068


42

R428 (5 μM for 4 h). β-actin is included as a loading control. (H) Immunoblotting 1069

analysis of P-STAT3 and STAT3 levels in HN5-ER cells treated with tocilizumab (1 1070

μM for 3 d) or PBS, or with ruxolitinib (3 μM for 24 h) or DMSO. β-actin is included 1071

as a loading control. 1072


experiments. *, p<0.0001, HN5-ER vs HN5. 1074

1075

Figure 5. Increased pro-inflammatory signaling in HNC cells with acquired 1076

erlotinib resistance. 1077

IPA software was used to map and combine differentially expressed genes associated 1078

with acquired erlotinib resistance into a custom merged pathway showing components 1079

of canonical NF-κB, IL-6, IL-8 and IFN-γ signaling, including Axl, IL-6, IL-8 and 1080

COX-2. The density of red shading represents the fold-change upregulation of a given 1081

gene upon acquisition of erlotinib resistance. 1082

1083

Figure 6. A common signature of acquired erlotinib resistance between HNC and 1084

NSCLC. 1085

Venn diagram comparisons are shown between the HN5/HN5-ER microarray study 1086

herein, and the NSCLC study of Zhang and coworkers: HCC 827/HCC 827 ER1/HCC 1087

827 ER2 (20). Genes commonly upregulated or downregulated with acquired erlotinib 1088

resistance in both HNC (gray) and NSCLC (blue) models are indicated by the 1089

intersection of each dataset and are listed in Supplementary Table 4. 1090

1091

Figure 7. Axl mRNA overexpression predicts poor survival in HNC. 1092


43

Kaplan-Meier plot of age-, gender- and smoking status-adjusted survival for 302 HNC 1093

patients according to median tumor Axl mRNA expression over a ten-year follow up 1094

period after initial surgery. HR, hazard ratio. 1095

1096

Figure 8. Model for miR-34a, Axl, and IL-6 in the regulation of EMT and EGFR 1097

inhibitor resistance in HNC. 1098

Acquired resistance of HNC cells to erlotinib is associated with elevated Axl 1099

expression and activity, increased synthesis and secretion of IL-6, and reduced levels 1100

of miR-34a. Loss of miR-34a permits increased expression of a set of proven or 1101

predicted miR-34a target genes that confer an EMT phenotype, including Axl, 1102

AXIN2, CA9, CXCL10, FOSL1, FUT8, GAS1, KLF6 and PODXL. Axl and IL-6 1103

activate Akt, NF-κB and STAT3 signaling, which in turn promotes EMT, erlotinib 1104

resistance, metastasis and expression of pro-inflammatory cytokines. Finally, 1105

increased levels of IL-6 may induce EMT via suppression of miR-200c expression. 1106

Boxes indicate findings in HN5-ER cells presented in this manuscript. Numbers in 1107

figure correspond to references for that association. 1108

Figure 1

A B

C

µLog10 erlotinib [ M]

Rela

tive n

um

ber

of

vi a

bl e

cells

-3 -2 -1 0 1 2 30.0

0.5

1.0

1.5HN5-ER

EC50

HN5

EC50 = 0.1 µM

> 20 µMµ

µ

DEGFR

β-actin

P-EGFR

Akt

P-Akt

ERK1/2

P-ERK1/2

E-cadherin

vimentin

HN5 HN5-ER

Re

lati

ve

miR

-20

0 e

xp

res

sio

n

(no

rma

lize

d t

o U

44

sn

RN

A)

0

0.5

1

1.5

HN5 HN5-ER

miR-200a miR-200b miR-200c

HN5 HN5-ER HN5 HN5-ER

E

Ce

ll in

de

x

0 9 18 3627 45

Time (h)

HN5HN5-ER

HN5-ER (SFM)

HN5 (SFM)-0.1

0.9

1.9

2.9

HN5-ER

HN5

****

Figure 2

CA BHN5

5

8

0

1

23

4

5

6

78

D

0

0.5

1

1.5

2

2.5

3

HN5-ERHN5 HN5-ERHN5

HN5-ER

6

7

Re

lati

ve

P-A

xl d

en

isit

y

Re

lati

ve

P-A

xl

(pg

/µg

to

tal p

rote

in)

0

0.2

0.4

0.6

0.8

1.0

1.2

1.4

HN5-ERHN5

Re

lati

ve

miR

-34

a e

xp

res

sio

nE-cadherin

β-actin

Axl

miR

-34a

LF20

00

miR

-NC

HN5-ERE F

1

1

1

1

1

1

1

1

2

2

3

3

4

4

56

87

**

*

β-actin

Axl

HN5-

ER

HN5

FaDu

G

LF20

00

Re

lati

ve

nu

mb

er

of

via

ble

ce

lls

0

0.2

0.4

0.6

0.8

1.0

1.2

1.4

miR

-34a

0.5

nM

miR

-34a

1.0

nM

miR

-NC 0

.5 n

M

miR

-NC 1

.0 n

M

LF20

00

miR

-34a

0.5

nM

miR

-34a

1.0

nM

miR

-NC 0

.5 n

M

miR

-NC 1

.0 n

M

DMSO 10 µM erlotinib

*******

Figure 3

B

C D

0

0.2

0.4

0.6

0.8

1

1.2

1.4

0 0.1 0.5 1

Rel a

tive n

um

ber

of

vi a

ble

cel ls

[R428] (µM)

β-actin

Akt

erlo

tinib

vehi

cle

P-Akt

R42

8

+ R42

8

HN5-ER

HN5-ER

Log R428 [µM]

Rela

tive n

um

ber

of

via

ble

cells

HN5-ER

EC50

HN5

EC50 = 1.4 µM

= 1.3 µMµ

10

-3 -2 -1 0 1 20.0

0.5

1.0

1.5

A

**

erlo

tinib

Ce

ll in

de

x

3

Time (h)

2.5

1.5

0.5

0 6 9 12 15 18 21 24 27-0.5

HN5-ER / vehicle

HN5-ER / R428

HN5-ER / vehicle / SFM

HN5-ER / vehicle / SFM

erlotinib (10µM)- + - + - + - +

Figure 4

A B

0 2000 4000 6000 8000 10000 12000 14000

0

2000

4000

6000

8000

10000

12000

14000

HN5, probe intensity

HN

5-E

R, p

rob

e in

ten

sit

y

HN5 HN5-ER

Row z-score

0-1 1

E

E-cadherin

No

rma

liza

ed

mR

NA

ex

pre

ss

ion

0

2

1

3

4

5

6

HN5

HN5-

ER

HN5

HN5-

ER

HN5

HN5-

ER

HN5

HN5-

ER

N-cadherin

COX-2

EGFR

7

8

9

HN5

HN5-

ER

Axl

*

*

*

*

*

No

rma

liza

ed

mR

NA

ex

pre

ss

ion

0

200

100

300

400

500

600

ZEB1

IL-6

IL-8

HN5

HN5-

ER

HN5

HN5-

ER

HN5

HN5-

ER

C

**

*

*

No

rma

liza

ed

IL

-6 c

on

ce

ntr

ati

on

(p

g/m

l)

HN5 HN5-ER

0

200

400

600

800

1000

HN5 HN5-ER

*

No

rma

liza

ed

IL

-8 c

on

ce

ntr

ati

on

(n

g/m

l)

0

log2(fold change)

-4 -2 0 2 4

HN5-ER/HN5-l

og

10(p

-va

lue

)

0

2

4

6

8

D

20

40

60

80

100

120

F

STAT3

HN5-ERHN5

P-STAT3

CD44

β-actin

G H

P-STAT3

DM

SO

ruxo

litin

ib

STAT3

β-actin

P-STAT3

PBS

toci

lizum

ab

STAT3

β-actinP-Akt

β-actin

Akt

P-STAT3

STAT3

DM

SO

(0 h

)R42

8 (4

h)

DM

SO

(4 h

)

Figure 5

Figure 6

HNC NSCLC

159195152

Genes upregulated with

acquired erlotinib resistance

HNC NSCLC

108375551

Genes downregulated with

acquired erlotinib resistance

Figure 7

low Axlhigh Axl

Follow up (months)

Pe

rce

nt

su

rviv

al (%

)

0 20 40 60 80 100 120

0

20

40

60

80

100

HR = 1

HR = 1.66, p = 0.007

Figure 8

AXIN2, CA9, CXCL10

FOSL1, FUT8, GAS1,

KLF6, PODXL

miR-34a Axl

P-Akt

NF-κB

P-STAT3

IL-6

IL-6R/JAK miR-200c

EMT

metastasis

EGFR inhibitor

resistance

(30)

(25)

(74)

(73)

(9)(14)

(63-70)

Axl mediates acquired resistance of head and neck cancer...

Documents

Transcript of Axl mediates acquired resistance of head and neck cancer...