Companion Diagnostics & Biomarkers · Companion Diagnostics & Biomarkers Quantitative Chemical Proteomics Proﬁling Differentiates Erlotinib from Geﬁtinib in EGFR Wild-Type Non–Small

Companion Diagnostics & Biomarkers

Quantitative Chemical Proteomics Profiling DifferentiatesErlotinib from Gefitinib in EGFR Wild-Type Non–Small CellLung Carcinoma Cell Lines

Ang�elique Augustin1, Jens Lamerz2, H�el�ene Meistermann1, Sabrina Golling1, Stefan Scheiblich4,Johannes C. Hermann5, Guillemette Duchateau-Nguyen2, Manuel Tzouros1, David W. Avila1,Hanno Langen1, Laurent Essioux2, and Barbara Klughammer3

AbstractAlthough both erlotinib and gefitinib target the EGF receptor (EGFR), erlotinib is effective in patients with

EGFRwild-type ormutated tumors, whereas gefitinib is only beneficial for patients with activatingmutations.

To determine whether these differences in clinical outcomes can be attributed to their respective protein

interaction profiles, a label-free, quantitative chemical proteomics studywas conducted. Using thismethod, 24

proteins were highlighted in the binding profiles of erlotinib and gefitinib. Unlike gefinitib, erlotinib displaced

the ternary complex formed by integrin-linked kinase (ILK), a-parvin, and PINCH (IPP). The docking of

erlotinib in the three-dimensional structure of ILK showed that erlotinib has the ability to bind to the ATP-

binding site, whereas gefitinib is unlikely to bind with high affinity. As the IPP complex has been shown to be

involved in epithelial-to-mesenchymal transition (EMT) and erlotinib sensitivity has been correlatedwith EMT

status, we used a cellular model of inducible transition and observed that erlotinib prevented EMT in a more

efficientway than gefitinib by acting on E-cadherin expression aswell as on IPP levels. A retrospective analysis

of theMERIT trial indicated that, besides a high level of E-cadherin, a low level of ILK could be linked to clinical

benefit with erlotinib. In conclusion, we propose that, in an EGFR wild-type context, erlotinib may have a

complementary mode of action by inhibiting IPP complex activities, resulting in the slowing down of the

metastatic process of epithelial tumors. Mol Cancer Ther; 12(4); 520–9. �2013 AACR.

IntroductionOver the past few years, several drugs targeting protein

kinases have shown clinical efficacy as cancer therapy (1).Among these, erlotinib and gefitinib are first-generationtyrosine-kinase inhibitors (TKI) that directly inhibit theEGF receptor (EGFR; refs. 2–5). Recent clinical data haveshown that patients with non–small cell lung carcinoma(NSCLC) whose tumors harbor an activating EGFRmuta-tion respond especially well to treatment with these TKIs(3, 4, 6). Although both TKIs target EGFR and are espe-

cially potent inhibitors of activating mutation-positiveEGFR, erlotinib has also shown efficacy versus placeboin EGFR wild-type NSCLC (2, 3, 7, 8).

To determine whether differences in clinical outcomesseen in patients with EGFR wild-type tumors receivingerlotinib or gefitinib can be attributed to differences in theprotein interaction profiles of these 2 compounds, a quan-titative chemical proteomics studywas conducted. Chem-ical proteomics is a mass spectrometry–based affinitychromatography approach, which uses immobilizedsmall molecules as bait to capture and identify interactingprotein complexes from an entire proteome. This tech-nique has been successfully applied to explore the modeof action of kinase inhibitors and to assess naturallyoccurring small molecules to identify their targets (9,10). Togain specificity byfilteringout backgroundbindersgenerated by low-affinity abundant proteins, the work-flow is best accompaniedwith a competition-binding stepthat involves free compound pretreatment of the probedmixture. Here, competition-binding experiments andlabel-free liquid chromatography/mass spectrometry(LC/MS) signal detection were used to identify proteincomplexes that can be specifically displaced from anaffinity matrix by unmodified erlotinib or gefitinib. Thisreveals a potential complementary mode of action forerlotinib in the context of epithelial-to-mesenchymal

Authors' Affiliations: 1Protein and Metabolite Biomarkers, 2Bioinformat-ics and Exploratory Data Analysis, Translational Research Sciences, and3Tarceva Clinical Biomarker Subteam, Pharmaceuticals Division, F. Hoff-mann-La Roche Ltd., Basel, Switzerland; 4Discovery Research Oncology,Pharmaceuticals Division, F. Hoffmann-La Roche Ltd., Penzberg, Ger-many; and 5Discovery Chemistry, Pharmaceuticals Division, F. Hoff-mann-La Roche Ltd., Nutley, New Jersey

Note: Supplementary data for this article are available at Molecular CancerTherapeutics Online (http://mct.aacrjournals.org/).

Corresponding Authors: Ang�elique Augustin, Translational ResearchSciences, Pharmaceuticals Division, F. Hoffmann-La Roche Ltd., Grenza-cherstrasse 124, 4070 Basel, Switzerland. Phone: 41-61-688-4050; Fax:41-61-688-1929; E-mail: [email protected]; and BarbaraKlughammer, [email protected]

doi: 10.1158/1535-7163.MCT-12-0880

�2013 American Association for Cancer Research.

MolecularCancer

Therapeutics

Mol Cancer Ther; 12(4) April 2013520

on February 22, 2021. © 2013 American Association for Cancer Research. mct.aacrjournals.org Downloaded from

Published OnlineFirst January 31, 2013; DOI: 10.1158/1535-7163.MCT-12-0880

http://mct.aacrjournals.org/

transition (EMT). EMT is a known mechanism throughwhich epithelial cells lose adhesion molecules such as E-cadherin and acquire a mesenchymal phenotype, whichallows them to migrate into the stroma and becomeinvasive (11). The correlation between erlotinib sensitivityand EMT has been shown previously in vitro and in vivo(12, 13) and suggests that patients with epithelial-typeNSCLC tumors, that is, which had not undergone thetransition, could have a better outcome with an erlotinib-containing regimen (12). The complex formed by integrin-linked kinase (ILK), a-parvin, and PINCH is bound toerlotinib but not gefitinib in the chemical proteomicsprofiling of several EGFR wild-type NSCLC cell lines.This complex, also known as ILK–PINCH–Parvin com-plex (IPP) complex, forms a structural and signaling entityin the integrin pathway (14) and has been shown toinfluence EMT. ILK and PINCH overexpression are ableto trigger EMT, by inducing loss of E-cadherin (15, 16).TGF-b-1–induced EMT in mammary epithelial cellsrequired functional ILK (17). The next steps were toinvestigate a potential structural explanation for thisbinding discrepancy between the 2 compounds, as wellas using an EMT-inducible cellular model to assess theirrespective functional effect on EMT. These findings werefurther correlated with gene expression data from theMERIT open-label, multicenter, phase II clinical trial,which aimed to identify genes with potential as biomar-kers for clinical benefit from erlotinib therapy (18).

Materials and MethodsReagents and cell linesAnti-ILK, E-cadherin, anda-parvinwere obtained from

Cell Signaling Technology, anti-PINCH1 and glyceralde-hyde-3-phosphate dehydrogenase–horseradish peroxi-dase (GAPDH-HRP) from Santa Cruz Biotechnology, andanti-vimentin from BD Biosciences. All cell lines wereobtained from the American Type Culture Collection(ATCC) without further authentication. Cell culturesH292, H226, H322, H358, H441, H460, and H1299 weremaintained in RPMI-1640, and Calu6 in EMEM; bothsupplemented with 10% FBS.

Chemical proteomics profiling in H292 cellsErlotinib and gefitinib were modified with a 6-(3-ami-

nopropyloxy) side chain for covalent linking toepoxy-activated Sepharose 6B beads (GE Healthcare) togenerate affinity matrices, as described by Brehmer andcolleagues (19). Starting material 4-[(3-ethynylphenyl)amino]-7-(2-methoxyethoxy)-6-quinazolinol was pre-pared as described in WO9630347A1 (20). Coupling ofthe derivatives to the epoxy-activated beads was done in50% dimethylformamide (DMF)/0.1 mol/L Na2CO3 pH11 in thepresence of 10mmol/LNaOH, overnight at roomtemperature in the dark. After washing with 50% DMF/0.1 mol/L Na2CO3, any remaining reactive groups wereblockedwith 1mol/L ethanolamine. Subsequentwashingsteps were conducted according to the manufacturer’s

instructions. Binding efficiency was estimated by com-paring the optical densities of the compound solutionsbefore and after coupling at the maximal absorbancewavelengths 247 and 330 nm. Cells were lysed in 50mmol/L HEPES pH 7.3 buffer containing 150 mmol/LNaCl, 1 mmol/L CaCl2, 1 mmol/L MgCl2, 1% NP-40,protease inhibitors (Complete EDTA free; Roche AppliedSciences), and 1 mmol/L orthovanadate; samples werecleared by centrifugation at 12,000 � g for 15 minutes at4�C. Cell lysates were preincubated with increasing con-centrations (0, 0.0856, 0.2862, 0.957, 3.2, 10.7, 35.78, 119.6,and 400 mmol/L final concentration) of unmodified erlo-tinib or gefitinib for 1 hour at 4�C under rotation. Thepretreated samples were loaded onto Sepharose beadsimmobilized with the respective compound (erlotinib orgefitinib) for 3 hours at 4�C under rotation. Beads werewashed 3 times with lysis buffer, treated with SDS-PAGEbuffer for 5minutes at 85�C, andelutedproteins separatedon 4% to 20% Tris–glycine SDS-PAGE. After proteinfixation in 40% ethanol and 10% acetic acid, gels werestained overnight with Coomassie Brilliant blue. For pro-tein in-gel digestion, an adapted protocol of Shevchenkoand colleagues (21) was used. In brief, each gel lane wascut into 5 slices between the 20 and 150 kDa region.Proteins were reduced with 50 mmol/L dithioerythritolfor 45 minutes at 56�C, alkylated with 55 mmol/L iodoa-cetamide for 1 hour at room temperature in the dark, anddigested with trypsin (Promega) overnight at room tem-perature. Peptides were extracted twice with 1:2 (v/v)acetonitrile/25 mmol/L ammonium bicarbonate, and 1:2(v/v) acetonitrile/5% formic acid, respectively, for 15minutes at 37�C, dried downusing a speedvac, and storedat �20�C before analysis. Samples were reconstituted in2% acetonitrile/5% formic acid and run in triplicate byLC/MS. Nanoflow liquid chromatography/tandemmassspectrometry (LC/MS-MS) was carried out by couplingan Easy-nLC system (Proxeon) to an LTQ Orbitrap Velos(Thermo Fisher Scientific) equipped with a Proxeonnanoelectrospray and an active background ion reductiondevice (ABIRD; ESI Source Solutions). Peptides wereloaded on a 10-mmAqua C18 trapping (packed in-house,100 mm inner diameter) and separated on a 200-mmReprosil-Pur C18-AQ analytic (Dr. Maisch GmbH) col-umn (75 mm inner diameter, 3 mm particle size, 120 A).Individual raw data files were processed with SEQUEST(version 27.0, revision 12, Thermo Fisher Scientific) andsearches were conducted against a concatenated for-ward/reverse human UniProt/SwissProt database (June2009 release, 34,275 entries, containing splice-variants)allowing for a spectrum false-discovery rate of 2.5%. Amass tolerance of 5 ppm for precursor and 1.0 Da forfragment ions was set. Oxidized methionines (þ15.9949Da) and carbamidomethylated cysteines (þ57.0215 Da)were considered as differential and fixed modifications,respectively. Only fully tryptic peptides with not morethan one mis-cleavage were considered for data analysis.Detailed LC/MS and data analysis procedure can befound in the Supplementary Data.

Chemical Proteomics Differentiates Erlotinib from Gefitinib

www.aacrjournals.org Mol Cancer Ther; 12(4) April 2013 521




Chemical proteomics profiling of erlotinib andgefitinib in other NSCLC cell lines

H358, H322, H226, and Calu6 cells were lysed in 50mmol/L HEPES pH 7.3 buffer containing 150 mmol/LNaCl, 1 mmol/L CaCl2, 1 mmol/L MgCl2, 1% NP-40,protease inhibitors (Complete EDTA free, Roche AppliedSciences), 1 mmol/L orthovanadate, and samples werecleared by centrifugation at 12,000 � g for 15 minutes at4�C. Cell lysates were loaded on erlotinib-Sepharose andgefitinib-Sepharose, using the same experimental condi-tions as in the profiling experiment described earlierbut without competing with the unmodified erlotinib orgefitinib. The elution fractions were analyzed by LC/MS.Spectral counts corresponding to EGFR, GAK, LOK(Stk10), Abl2, SLK, MEKK1 (M3K1), ILK, a-parvin(PARVA), and PINCH1 (LIMS1) were extracted and nor-malized toward the sumof all spectral counts correspond-ing to all proteins identified in the respective fractions.

Alignment of EGFR and ILK crystal proteinstructures

Publicly available crystal structure protein data bank(pdb) files of EGFR-bound erlotinib or gefitinib and thecrystal structure of ILK complexed with ATPwere down-loaded from the crystallographic database [ref. 22; pdbcodes 1M17, 2ITY, 3KMW; (23–25)]. The 3 kinase struc-tureswere alignedusing the coordinates fromCaatomsofa subset of the residues forming the ATP-binding site[residue numbers based on EGFR numbering 691–693,701–703, 719–72 (21) 1, 751, 766–772, 820–823, 831–832].Alignment and distance measurements were conductedusing the PyMOL visualization package (25).

Testing EMT on H358 NSCLC cell lineH358 cells were treated with recombinant 10 ng/mL

TGF-b3 (Peprotech) and 5mmol/L erlotinib or gefitinib for3, 6, 10, and 13 days. Cell culture medium was renewedevery 3 days. Cell lysis was conducted in PBS containing1% NP-40 and protease inhibitors. After clearing by cen-trifugation at 12,000 � g for 15 minutes at 4�C, proteinswere separated by SDS-PAGE and processed for Westernblot analysis. Images were acquired using a compactdigital camera (Luminescent Image Analyzer; Fujifilm)and quantified with ImageQuant TL (GE Healthcare).Statistical analysis was conducted using JMP 8.0.2 (SASInstitute Inc.). To determine the influence of erlotinib andgefitinib on ILK and PINCH after TGF-b3 stimulation, thefollowingmodelwas applied: ExpressionD13� Treatmentþ ExpressionD3. For analysis of IPP expression in NSCLCcell lines, simple t tests were conducted.

Analysis of MERIT dataGene expression profilesweremeasured in tumor biop-

sy samples from 102 patients with stage IIIB/IV NSCLC,being treated with erlotinib (150 mg/day). Tumor biop-sies were analyzed using gene expression profiling withAffymetrix GeneChip Human Genome U133A Array(Affymetrix). All gene expression results described here

(fold changes andPvalues) are derived from the statisticalanalysis already reported (18). We specifically looked atdifferences between gene expression profiles of patientswith and without clinical benefit from erlotinib. Clinicalbenefit was defined as an objective response (completeresponses or partial responses were confirmed by repeat-ed assessments �4 weeks apart at any time during thetreatment period) ormaintenance of stable disease for�12weeks after study entry.

The following 21 proteins, potentially involved in tran-scriptional regulation of EMT or in IPP complex, weremapped to the corresponding genes and microarrayprobe sets: Twist, Snail (Sma1), Zeb (TCF8), Lef-1, E12/E47 basic loop helix (bHLH), Slug (Sma2), E-cadherin(CDH1), N-cadherin (CDH2), Myc, vimentin, catenin-b,catenin-d, fibronectin, epimorphin, ILK, a-parvin, b-par-vin, PINCH 1, PINCH 2, metastasis-associated protein 3(MTA3), and Y box–binding protein 1 (YB-1).

ResultsInteraction profile of erlotinib and gefitinib in H292NSCLC cell line

A quantitative chemical proteomics method wasapplied to erlotinib and gefitinib (Fig. 1A) using the samewild-type EGFR-expressing NSCLC cell line H292 asbiological target to identify the interaction profiles of thecompounds. Cell lysates were preincubated with increas-ing concentrations of unmodified erlotinib or gefitinibbefore fractionation on the respective affinity matrix,captured proteins were eluted, separated by SDS-PAGEand identified by LC/MS and protein database mining(Supplementary Fig. S1). A total of 1,285 and 1,303 proteingroups were identified in the bound fractions to erlotiniband gefitinib, respectively (Supplementary Table S1). Alinear model (27, 28) was applied on mass spectrometrysignals toderive concentration-dependent signaldecreaseof specific binders. Comparison of the binding profiles ofthe 2 compounds was carried out using a robust, simple,statistical procedure adapted fromBretz andHothorn (29)on peptide signals and extended to multiple testing ofcorresponding protein groups. Nine proteins were foundwith an adjusted P value below 5% in the samples dis-placed with erlotinib (Fig. 1B, Regions A, B and x-axis;Supplementary Table S2), and 19 in those displaced withgefitinib (Fig. 1B; regions A, C, and y-axis; SupplementaryTable S2), including EGFR, which was specifically dis-placed by both compounds, confirming the validity of theapproach. The presence of 2 other proteins, GAK andStk10,was also significant in both compoundprofiles (Fig.1B; region A; Supplementary Table S2), whereas NUD12,K0564, SN1L1, KC1E, and EPHA1 were only identified inthe gefitinib-bound fraction (Fig. 1B, y-axis; Supplemen-tary Table S2); M3K1 and LIMS1 only in the erlotinib-bound fraction (Fig. 1B, x-axis; Supplementary Table S2).

While the majority of proteins displaced by eithererlotinib or gefitinib were protein kinases (7 for erlotinib,i.e., EGFR, GAK, STK10, ILK, SLK, ABL2, andM3K1; and

Augustin et al.

Mol Cancer Ther; 12(4) April 2013 Molecular Cancer Therapeutics522




11 for gefitinib, i.e., EGFR, GAK, STK10, MKNK2, RIPK2,ADCK3; MK09; EPHB2, EPHB4, KC1E, and EPHA1), anotable difference between the 2 compounds was seenin the non-protein kinase binding profile. In this sub-group, erlotinib only displaced 2 proteins (PARVA andLIMS1), whereas gefitinib displaced 8 (NUD12, K0564,SN1L1, AP1G2, BLMH, HEAT3, SAAL1, and DFNA5).Interestingly, LIMS1 (also known as PINCH1) forms anactive complex togetherwith PARVA (a-parvin) and ILK,another protein displaced by erlotinib.

Interaction profile of erlotinib and gefitinib in otherNSCLC cell linesWe next addressed the question of whether the chem-

ical proteomics profile was cell line–specific by testing 4additionalEGFRwild-typeNSCLC cell lines (H358,H322,H226, and Calu6) for the 9 erlotinib-specific binders(EGFR, STK10, GAK, ILK, LIMS1, PARVA, SLK, M3K1,andABL2). Figure 2 shows the normalized spectral countsof these 9 proteins, including EGFR, in the bound frac-tions. All 9 proteins found in the erlotinib profile in H292were also identified in the protein purifications from all 4

of the other cell lines. In addition, we observed that thespectral counts of Abl2, Slk, MEKK1 (M3K1), ILK, a-par-vin (PARVA), and PINCH1 (LIMS1) in the erlotinib-bound fraction were higher than those in the gefitinib-bound fraction. This was particularly noticeable for theIPP, which was enriched on the erlotinib but not thegefitinib column from these EGFR wild-type NSCLC celllysates.

Interaction profile of erlotinib and gefitinib in theILK DVK loop region

Erlotinib and gefitinib have a high degree of similarityin their chemical structure (Fig. 1A). Knowing theresolved crystal structures of EGFR and ILK (24, 25), weexamined whether the binding of erlotinib to ILK wasdifferent from gefitinib, which might explain the discre-pancies in their binding profiles toward the IPP complex.As expected,when erlotinib or gefitinibwere in a complexwith EGFR, they displayed the same mode of binding tothe ATP-binding sites. The overall protein conformationof EGFR-erlotinib, EGFR-gefitinib, and the ATP-bindingpocket was very similar, resulting in an exact overlay ofthe 2 structures after aligning (data not shown).

The pseudokinase ILK shows some crucial differencesin fold and conformation to other kinases, includingEGFR(25). However, when ignoring sequence differences, theburied part of the ILK ATP-binding site is very similar tothe ATP-binding site of EGFR and allows a reliable struc-tural alignment of ILKwith EGFRwhen in a complexwitheither erlotinib or gefitinib (Fig. 3). A crucial difference inthe ATP-binding site is the conformation of the DFG loopand Ala338, the residue upstream of the DVK loop in ILK(residues 339–341, which form the typical DFG). Thisresidue is part of a turn that is not observed in EGFR andis close to the space occupied by the aniline portions oferlotinib and gefitinib.

The different substitutions in the erlotinib and gefitinibaniline rings (the 3 position is substituted with chlorine ingefitinib but with acetylene in erlotinib and the 4 positionis unsubstituted in erlotinib but substituted with fluorinein gefitinib) and the resulting differences in atom proxi-mities suggest that erlotinib ismore likely than gefitinib tobind favorably to the ATP-binding site of ILK.

Potential influence of erlotinib on the integrinsignaling pathway, connecting to EMT

The IPP complex functions as a hub in the integrinsignaling pathway (14). It plays a structural role by con-necting integrins to the cytoskeleton and also transducessignals from the extracellular matrix to various intracel-lular pathways involved in cell proliferation, migration,and survival. In addition, overexpression of both ILK andPINCH seems not only to follow the occurrence of EMT(15, 16), but ILK activity is also required for TGF-b1–mediated EMT in mammary epithelial cells (17). On thebasis of the fact that erlotinib binds to the IPP complex, asshown in our chemical proteomics approach, we hypoth-esized that in NSCLC cell lines that have not undergone

Figure 1. A, chemical structures of erlotinib and gefitinib. B, bindingprofiles of erlotinib and gefitinib in H292. P values of erlotinib versusgefitinib after log10 transformation. Lines indicate 5% probability of error.Region A shows proteins significantly binding to erlotinib and gefitinib;regions B and C show proteins significantly binding to erlotinib orgefitinib, respectively. Proteins in region D did not show significantbinding to either erlotinib or gefitinib. If a protein was identified in onecompound and not in the other, it could not be quantified, and hence noP value is available (P value, NA). These proteins are displayed outsidethe 2-dimensional plot for completeness (x- and y-axis). Note: all celllines were obtained from ATCC without further authentication.






EMT, erlotinib couldpotentially slowdown the transition,by inhibiting the IPP-mediated E-cadherin pathway.

To test this hypothesis, we used a model NSCLC cellline, H358, which had been previously described asswitching from epithelial-to-mesenchymal state uponTGF-b3 treatment (30). The results show a clear decreaseof E-cadherin after 13days of TGF-b3 treatment, indicatingthat the cells are transitioning (Fig. 4A). Treatment witherlotinib prevents this, whereas gefitinib fails to inhibit theTGF-b3–mediated decrease of E-cadherin levels.

We also examined the effects of the transition on ILKand PINCH expression (a-parvin could not be testedbecause of lowdetection byWestern blot analysis). Figure4B and C shows the average quantified Western blotanalysis signals from both proteins, for 3 and 13 days oftreatment. The expression of both ILK and PINCHwas increased by the transition. Both proteins gave alower signal when cells were treated with TGF-b3 anderlotinib.

Thus, this EMT model indicates that erlotinib reducesthe extent of EMT by inhibiting the loss of E-cadherin and

Figure 2. Purification of erlotinib binders in otherNSCLCcell lines. Additional chemical proteomics experimentswere carried out usingH226,Calu6, H358, andH322 cell lysates, loaded on erlotinib-Sepharose and gefitinib-Sepharose. Elution fractions from each purification were analyzed by LC/MS. The averagednormalized spectral counts out of 2 biological replicates, corresponding to the 9 proteins previously identified as binders to erlotinib, are represented from theerlotinib-Sepharose and gefitinib-Sepharose eluted fractions. Note: all cell lines were obtained from ATCC without further authentication.

Figure 3. ATP-binding site of aligned and overlaid ILK and EGFR crystalprotein structures showing potential interactions of erlotinib and gefitinibwith the area of the ILK DFG loop. Oxygen atoms are colored in red,nitrogen inblue, chlorine ingreen, and fluorine in light cyan.Carbonatomsof the erlotinib/EGFR complex (1M17) are colored in salmon, carbon ofthe gefitinib/EGFR (2ITY) complex in yellow, and carbon atoms of ILK arecolored in white. Picture was generated with PyMOL (25).

Augustin et al.





also seems to prevent a TGF-b3–mediated increase of bothmembers of the IPP complex, ILK and PINCH.

Expression of IPP complex in NSCLC cell linesILK and PINCH overexpression has been correlated

with tumor progression and/or aggressiveness in severalhumanmalignancies (14). At present, the 3 components ofthe complex are studiedmainly onan individual basis, notaddressing the possibility of the expression of the entirecomplex being correlated with EMT status. Eight EGFRwild-type NSCLC cell lines were selected on the basis ofthis status (13, 30), which was assessed by the expressionof E-cadherin as an epithelial marker and vimentin as a

mesenchymal marker. All the IPP proteins were detected,with varying intensities, in all 8 cell lines except fora-parvin in H358 (Fig. 5A). After quantification and nor-malization, the overall sumof IPP complex intensitieswasalmost 2 times higher in mesenchymal versus epithelialcells (Fig. 5B).

ILK expression in erlotinib-treated patients—datafrom the MERIT trial

Taking into account the data fromNSCLCcell lines, andconsidering that the EMT status of biopsies from patientswith NSCLC has been previously correlated with clinicalbenefit from erlotinib treatment (12), we evaluated the

Figure 4. Erlotinib slows down the transition of H358 by stabilizing E-cadherin. H358 cells were treated for 3, 6, 13, and 20 days with TGF-b3 alone or togetherwith erlotinib and gefitinib. The levels of expression of E-cadherin (A), ILK, and PINCH were analyzed by Western blot analysis. For ILK (B) and PINCH (C),Western blot analysis signals were quantified and averaged from 2 biological replicates. The effect of 13 days of treatment with erlotinib and gefitinibwas estimated on TGF-b3–stimulated cells after adjustment to the level at 3 days of respective treatment. Compared with TGF-b3–stimulated untreated cells,a downregulation (in the sense of a "non-upregulation") by gefitinib on ILK (one-sided P ¼ 0.085) and PINCH-1 (one-sided P ¼ 0.12) did not reachsignificance. In contrast, there was a significant downregulation by erlotinib on ILK (one-sided P¼ 0.03) and PINCH-1 (one-sided P¼ 0.02). Note: all cell lineswere obtained from ATCC without further authentication.






potential relationship between IPP, a subset of genesinvolved in EMT, and response to erlotinib. Specifically,we looked at the differences between their expressionprofiles in patients with and without clinical benefit fromerlotinib.

In total, 16 of 21 preselected genes were identifiedwith detected probe sets and statistically analyzed [Twist,Snail (Sma1), Zeb (TCF8), Lef-1, E12/E47 basic loop helix(bHLH), Slug (Sma2), E-cadherin, Myc, vimentin, catenin-b, catenin-d, fibronectin, epimorphin, ILK, a-parvin, andb-parvin]. Data for the following 5 genes: PINCH 1,PINCH 2, metastasis-associated protein 3 (MTA3), Y box–binding protein 1 (YB-1), andN-cadherinwere not available.Among the 31 probe sets (corresponding to 16 uniquegenes), only 2 were differentially expressed in tumorsamples from patients with and without clinical benefit:ILK and CDH1 (E-cadherin), which were downregulated

(0.93-fold; P ¼ 0.011) and upregulated (1.38-fold; P ¼0.033), respectively, in patients with clinical benefit. Itshould be noted that these genes exhibited only smalltrends of regulation that were not significant when aprocedure of P value adjustment for multiple testing wasapplied, as in the initial work from Tan and colleagues(ref. 18; Fig. 6).

DiscussionIn tumors dependent on anEGFR-mutated receptor, the

mutation triggers a decreased affinity of the kinase forATP but its affinity to erlotinib is kept intact. This, takentogether, leads to an increased sensitivity to the TKI. Thedifference in efficacy between erlotinib and gefitinib hasbeen at least partially attributed to a difference in phar-macokinetic properties, that is, erlotinib at 150 mg has anarea under the curve 7 times higher than gefitinib at 250mg. In this study, we hypothesize that an additionalactivity besides the inhibition of EGFR could explain thedifference in clinical benefit seen between erlotinib andgefitinib in EGFR wild-type tumors, for example, inhibi-tion of ILK. Indeed, the chemical proteomics profiling oferlotinib and gefitinib has highlighted differences in thebinding profiles of the 2 compounds in at least 5 differentEGFRwild-typeNSCLCcell lines.Manykinase inhibitors,which have been approved for clinical use, have beenprofiled with this method leading to the discovery of newactivities, for example, on DDR1 for imatinib (31) orCAMK2G for bosutinib (32). Given the conservation ofthe ATP-binding site among human protein kinases, thefocus had been placed on potential binding activity withother protein kinases. Indeed,Abl2, Slk, and STK10 (LOK)have been confirmed to bind to erlotinib in vitro with atleast 25-fold higher affinity than gefitinib (33). However,these assays revealed also at least 30-fold lower affinity oferlotinib for Abl2, Slk, and STK10 by comparison withEGFR. A recent study using native kinase binding profil-ing in HL60 and PC3 cell lines also reported inhibitoryactivities of erlotinib on STK10 (LOK), MAP3K1 (M3K1),Slk, and ILK, which correlates well with our findings (34).Further experiments would be required to confirm theeffect of erlotinib on these kinases in the appropriate invivo background. Our data show that substantial infor-mation can also be deduced from the nonkinase proteinsthat form complexes with the direct binders. Unlike gefi-nitib, erlotinib interacted with the ternary complexformed by ILK, a-parvin, and PINCH in our assay.

We propose a structural explanation for the differencein binding to IPP between erlotinib and gefitinib. Thedocking of erlotinib in the 3-dimensional structure of ILKshows that erlotinib is able to bind to theATP-binding siteof the pseudokinase, as opposed to gefitinib, which is lesslikely to bind with high affinity. Indeed, when the EGFRcrystal structures, complexed either with erlotinib orgefitinib, are aligned with the ILK structure, the differentsubstitutions of the aniline moieties of the compoundsseem to have crucial implications for their potential

Figure 5. IPP expression in NSCLC cell lines. A, Western blot analyses ofILK, a-parvin, PINCH, E-cadherin, vimentin, and GAPDH on EGFR wild-type NSCLC epithelial cell lines H292, H441, H358, H322, andmesenchymal cell lines H460, H1299, H226, and Calu6. B, quantifiedWestern blot analysis signals normalized onGAPDH levels and averagedto determine the level of expression of the IPP complex. A significantupregulation in the GAPDH normalized expression was observed fora-parvin (one-sided P¼ 0.045), ILK (one-sided P¼ 0.007), and PINCH-1(one-sidedP¼ 0.011) from epithelial-to-mesenchymal cell lines. Note: allcell lines were obtained from ATCC without further authentication.

Augustin et al.





binding to ILK. Acknowledging that the distance mea-sured between overlaid structures depends on the align-ment method used and may vary slightly, a few observa-tions canbemade fromour structural alignment.Gefitinibhas 2 substitutions on the aniline ring, whereas erlotinibhas only 1,which points inside theATP-binding site, closeto the DVK loop. The 4-position is not substituted inerlotinib, whereas in gefitinib it features fluorine. Com-pared with hydrogen, the van der Waal’s radius of thefluorine atom is larger and the carbon–fluorine bond issignificantly longer than the carbon–hydrogen bond (35).This hydrogen-to-fluorine substitution seems tobe impor-tant, as the fluorine sits in an environmentwhere it clashesand causes steric repulsions with the Cb-methyl and thecarbonyl oxygen of Ala338 (�1.5 and 2 A, respectively;Fig. 5). Furthermore, the fluorine also creates an unfavor-able electrostatic interaction with the carbonyl oxygen ofAla338. In contrast, erlotinib projects a slightly polarizedhydrogen atom from the aniline 4-position that can estab-lish an attractive interaction with the carbonyl oxygen ofAla338. The repulsion from the fluorine atom in gefitinibcould either prevent it from binding to ILK with reason-able affinity or it could trigger a rearrangement of the loopcontaining theDVKmotif uponbinding. Because this loopis connected to the activation loop of ILK that is crucial forbinding a-parvin (25), such a rearrangement is likely tohave an influence on the ability of ILK to form a complexwith a-parvin. More sophisticated methods such asmolecular dynamics simulations could be applied tofurther elucidate the favorable and unfavorable interac-

tions between ligands and proteins. Measurements of invitro binding affinity of the compounds to the purifiedcomplex will be required to experimentally validate theresults of this in silico approach.

Our study shows that erlotinib prevents EMT inducedby TGF-b3 in a more efficient way than gefitinib in theH358 cellular model. As expected, this does not influencecellular proliferation in our in vitro assays, where thedifferent cell lines used in this study were growth inhib-ited to a varying extent but equally by both compounds aspreviously reported (12, 36). EMT will be more relevantfor invasive tumor growth in vivo, and clinical studieshave already shown the link between E-cadherin (CDH1)levels and response to erlotinib (12). The retrospectiveanalysis presented here on data from the MERIT trialconfirms that patients with higher expression of E-cad-herin (CDH1) have an increased clinical benefit fromtreatment with erlotinib, as previously published (18).A similar analysis has been conducted on gefitinib byKook and colleagues and did not show any correlationof E-cadherin expression with response rate or withprogression-free survival and overall survival in patientswith inoperable stage IIIB/IV NSCLC treated withgefitinib (37). Interestingly, a recent study has shown acausative relation between loss of E-cadherin and over-expression of ILK in squamous cell carcinoma and ade-nocarcinoma of the lung (38).

In addition, in our cellular model, erlotinib showed atendency to prevent the overexpression of ILK andPINCHinducedbyTGF-b3. It shouldbenoted that neither

Figure 6. E-cadherin (CDH1) and ILK expression in patients with versus without clinical benefit. Each dot represents the gene signal (log2 transformed)measured in one sample. For each group of patients (without or with clinical benefit), a boxplot is displayed where the thick black horizontal line (within a box)represent themedian gene signal for that group. The top and bottom boundaries of the box correspond respectively to the 25th percentile and 75th percentileof the gene signals. The whiskers extend to the most extreme data point, which is no more than 1.5 times the interquartile range from the box.






erlotinib nor gefitinib have been shown to bind the TGF-breceptors in vitro (33), excluding the possibility of a directeffect in our model on TGF-b receptor signaling. Over-expression of both ILK and PINCH has been shown toinduce EMT by lowering E-cadherin levels (16, 39, 40).Interestingly, while ILK is captured in our chemical pro-teomics experiments to the same extent in all the cell lines,PINCH ismost enriched inH226 and Calu6 fractions (Fig.2), which are mesenchymal cell lines (30). Overall, theseobservations indicate that erlotinib may exert a long-termeffect on the expression of ILK and PINCH, which couldexplain the observed consequences on the EMT. Becausethe expression of both proteins is regulated, at least inpart, by proteasomal degradation (41), we could hypoth-esize that erlotinib would accelerate this degradation andtherefore counteract the EMT process.

The respective overexpression of ILK and PINCH hasbeen associated with poor prognosis in NSCLC (42–44).Although PINCH could not be monitored in the MERITtrial, low expression of ILK seems to indicate a betterbenefit with erlotinib. This strengthens our theory that inpatients with a low expression of IPP, the remaining freeerlotinib can still act on the complex, whereas in cases ofoverexpression, this effect would not be significant.

In conclusion, we propose that, in a wild-type EGFRcontext, unlike gefitinib, erlotinib targets both EGFR andthe IPP complex activities, resulting in the slowing downof the metastatic process of epithelial tumors. Furtherinvestigations on additional models will be necessary toconfirm this observation and explore the early effects oferlotinib on the activities of the IPP complex, for example

in deciphering the details of the signaling pathway lead-ing to hindrance of EMT.

Disclosure of Potential Conflicts of InterestAll authors were employees of F. Hoffmann La Roche Ltd. at the date

of the study.

Authors' ContributionsConception and design: A. Augustin, J. Lamerz, B. KlughammerDevelopment of methodology: A. Augustin, J. Lamerz, H. Meistermann,S. Golling, G. Duchateau-Nguyen, L. Essioux, B. KlughammerAcquisition of data (provided animals, acquired and managed patients,provided facilities, etc.): A. Augustin, H. Meistermann, S. Golling, S.Scheiblich, M. Tzouros, D.W. Avila, B. KlughammerAnalysis and interpretation of data (e.g., statistical analysis, biostatis-tics, computational analysis): A. Augustin, J. Lamerz, S. Golling, J.C.Hermann, G. Duchateau-Nguyen, L. EssiouxWriting, review, and/or revision of the manuscript: A. Augustin, J.Lamerz, H. Meistermann, S. Golling, J.C. Hermann, G. Duchateau-Nguyen, M. Tzouros, D.W. Avila, H. Langen, L. Essioux, B. KlughammerAdministrative, technical, or material support (i.e., reporting or orga-nizing data, constructing databases): A. Augustin, S. GollingStudy supervision: A. Augustin, H. Langen, B. Klughammer

AcknowledgmentsThe authors thank Nikos Berntenis for providing automated sequence

interpretation platform; Michel Petrovic for automating the quantitativeinterpretation platform; and Peter Berndt and Wolfgang von der Saal forseminal discussions.

Grant SupportSupport for third-party writing assistance for this article was provided

by F. Hoffmann-La Roche Ltd.The costs of publication of this article were defrayed in part by the pay-

ment of page charges. This article must therefore be hereby marked adver-tisement in accordancewith 18 U.S.C. Section 1734 solely to indicate this fact.

Received September 4, 2012; revised January 17, 2013; accepted January25, 2013; published OnlineFirst January 31, 2013.

References1. McDermott U, Settleman J. Personalized cancer therapywith selective

kinase inhibitors: an emerging paradigm in medical oncology. J ClinOncol 2009;27:5650–9.

2. Shepherd FA, Rodrigues Pereira J, Ciuleanu T, Tan EH, Hirsh V,Thongprasert S, et al. Erlotinib in previously treated non–small-celllung cancer. N Engl J Med 2005;353:123–32.

3. Cappuzzo F, Ciuleanu T, Stelmakh L, Cicenas S, Szcz�esna A, Juh�aszE, et al. Erlotinib asmaintenance treatment in advanced non–small-celllung cancer: a multicentre, randomised, placebo-controlled phase 3study. Lancet Oncol 2010;11:521–9.

4. Mok TS, Wu YL, Thongprasert S, Yang CH, Chu DT, Saijo N, et al.Gefitinib or carboplatin-paclitaxel in pulmonary adenocarcinoma.N Engl J Med 2009;361:947–57.

5. Kim ES, Hirsh V, Mok T, Socinski MA, Gervais R, Wu YL, et al.Gefitinib versus docetaxel in previously treated non–small-cell lungcancer (INTEREST): a randomised phase III trial. Lancet 2008;372:1809–18.

6. Paz-Ares L, Souli�eres D, Melezínek I, Moecks J, Keil L, Mok T, et al.Clinical outcomes in non–small-cell lung cancer patientswith EGFR mutations: pooled analysis. J Cell Mol Med 2010;14:51–69.

7. Giaccone G, Herbst RS, Manegold C, Scagliotti G, Rosell R, Miller V,et al. Gefitinib in combination with gemcitabine and cisplatin inadvanced non–small-cell lung cancer: a phase III trial–INTACT 1.J Clin Oncol 2004;22:777–84.

8. Herbst RS, Giaccone G, Schiller JH, Natale RB, Miller V, Manegold C,et al. Gefitinib in combination with paclitaxel and carboplatin in

advanced non–small-cell lung cancer: a phase III trial–INTACT 2.J Clin Oncol 2004;22:785–94.

9. RixU, Superti-FurgaG. Target profiling of smallmolecules by chemicalproteomics. Nat Chem Biol 2009;5:616–24.

10. Bantscheff M, Scholten A, Heck AJ. Revealing promiscuous drug-target interactions by chemical proteomics. Drug Discov Today2009;14:1021–9.

11. Singh A, Settleman J. EMT, cancer stem cells and drug resistance:an emerging axis of evil in the war on cancer. Oncogene 2010;29:4741–51.

12. Yauch RL, Januario T, Eberhard DA, Cavet G, Zhu W, Fu L, et al.Epithelial versus mesenchymal phenotype determines in vitro sensi-tivity and predicts clinical activity of erlotinib in lung cancer patients.Clin Cancer Res 2005;11:8686–98.

13. Thomson S, Buck E, Petti F, Griffin G, Brown E, Ramnarine N, et al.Epithelial to mesenchymal transition is a determinant of sensitivity ofnon–small-cell lung carcinoma cell lines and xenografts to epidermalgrowth factor receptor inhibition. Cancer Res 2005;65:9455–62.

14. Cabodi S, Del Pilar Camacho-Leal M, Di Stefano P, Defilippi P. Integrinsignalling adaptors: not only figurants in the cancer story. Nat RevCancer 2010;10:858–70.

15. Oloumi A, McPhee T, Dedhar S. Regulation of E-cadherin expressionand beta-catenin/Tcf transcriptional activity by the integrin-linkedkinase. Biochim Biophys Acta 2004;1691:1–15.

16. Li Y, Dai C, Wu C, Liu Y. PINCH-1 promotes tubular epithelial-to-mesenchymal transition by interacting with integrin-linked kinase.J Am Soc Nephrol 2007;18:2534–43.

Augustin et al.





17. Serrano I,McDonaldPC, Lock FE, Dedhar S. Role of the integrin-linkedkinase (ILK)/Rictor complex in TGFbeta-1-induced epithelial-mesen-chymal transition (EMT). Oncogene 2012;32:50–60.

18. TanEH, RamlauR, PluzanskaA, KuoHP, ReckM,Milanowski J, et al. Amulticentre phase II gene expression profiling study of putative rela-tionships between tumour biomarkers and clinical response witherlotinib in non–small-cell lung cancer. Ann Oncol 2010;21:217–22.

19. Brehmer D, Greff Z, Godl K, Blencke S, Kurtenbach A, Weber M, et al.Cellular targets of gefitinib. Cancer Res 2005;65:379–82.

20. Schnur RC, Arnold LD. Qinazoline derivatives. International patentapplication publication number WO1996/30347.

21. Shevchenko A, Wilm M, Vorm O, Mann M. Mass spectrometricsequencing of proteins silver-stained polyacrylamide gels. Anal Chem1996;68:850–8.

22. Berman HM, Westbrook J, Feng Z, Gilliland G, Bhat TN, Weissig H,et al. The protein data bank. Nucleic Acids Res 2000;28:235–42.

23. Stamos J, Sliwkowski MX, Eigenbrot C. Structure of the epidermalgrowth factor receptor kinase domain alone and in complex with a 4-anilinoquinazoline inhibitor. J Biol Chem 2002;277:46265–72.

24. Yun CH, Boggon TJ, Li Y, Woo MS, Greulich H, Meyerson M, et al.Structures of lung cancer-derived EGFR mutants and inhibitor com-plexes: mechanism of activation and insights into differential inhibitorsensitivity. Cancer Cell 2007;11:217–27.

25. Fukuda K, Gupta S, Chen K, Wu C, Qin J. The pseudoactive site of ILKis essential for its binding to alpha-Parvin and localization to focaladhesions. Mol Cell 2009;36:819–30.

26. The PyMOL Molecular Graphics System V, Schr€odinger, LLC.27. IrizarryRA,HobbsB,Collin F,Beazer-BarclayYD,AntonellisKJ, Scherf

U, et al. Exploration, normalization, and summaries of high densityoligonucleotide array probe level data. Biostatistics 2003;4:249–64.

28. Holder D, Raubertas RF, Pikounis VB, Svetnik V, Soper K. Statisticalanalysis of high density oligonucleotide arrays: a SAFER approach. In:Proceedings of the ASA Annual Meeting 2001; Atlanta, GA; 2001.

29. Bretz F, Hothorn LA. Testing dose–response relationships with a prioriunknown, possibly nonmonotone shapes. J Biopharm Stat 2001;11:193–207.

30. ThomsonS, Petti F, Sujka-Kwok I, Epstein D, Haley JD. Kinase switch-ing inmesenchymal-like non–small cell lung cancer lines contributes toEGFR inhibitor resistance through pathway redundancy. Clin ExpMetastasis 2008;25:843–54.

31. Bantscheff M, Eberhard D, Abraham Y, Bastuck S, Boesche M,Hobson S, et al. Quantitative chemical proteomics reveals mechan-

isms of action of clinical ABL kinase inhibitors. Nat Biotechnol2007;25:1035–44.

32. Remsing Rix LL, Rix U, Colinge J, Hantschel O, Bennett KL, StranzlT, et al. Global target profile of the kinase inhibitor bosutinibin primary chronic myeloid leukemia cells. Leukemia 2009;23:477–85.

33. KaramanMW, Herrgard S, Treiber DK, Gallant P, Atteridge CE, Camp-bell BT, et al. A quantitative analysis of kinase inhibitor selectivity. NatBiotechnol 2008;26:127–32.

34. Patricelli MP,Nomanbhoy TK,WuJ,BrownH,ZhouD, Zhang J, et al. Insitu kinase profiling reveals functionally relevant properties of nativekinases. Chem Biol 2011;18:699–710.

35. Muller K, Faeh C, Diederich F. Fluorine in pharmaceuticals: lookingbeyond intuition. Science 2007;317:1881–6.

36. Ono M, Hirata A, Kometani T, MiyagawaM, Ueda S, Kinoshita H, et al.Mol Cancer Ther 2004;3:465–72.

37. Kook EH, Kim YM, Kim HT, Koh JS, Choi YJ, Rho JK, et al. Prognosticvalue of E-cadherin expression in non–small cell lung cancer treatedwith gefitinib. Oncol Res 2010;18:445–51.

38. Yu J, Shi R, Zhang D, Wang E, Qiu X. Expression of integrin-linkedkinase in lung squamous cell carcinoma and adenocarcinoma: corre-lation with E-cadherin expression, tumor microvessel density andclinical outcome. Virchows Arch 2011;458:99–107.

39. Li Y, Yang J, Dai C, Wu C, Liu Y. Role for integrin-linked kinase inmediating tubular epithelial to mesenchymal transition and renal inter-stitial fibrogenesis. J Clin Invest 2003;112:503–16.

40. Somasiri A, Howarth A, Goswami D, Dedhar S, Roskelley CD. Over-expression of the integrin-linked kinase mesenchymally transformsmammary epithelial cells. J Cell Sci 2001;114:1125–36.

41. Fukuda T, Chen K, Shi X, Wu C. PINCH-1 is an obligate partner ofintegrin-linked kinase (ILK) functioning in cell shapemodulation, motil-ity, and survival. J Biol Chem 2003;278:51324–33.

42. OkamuraM, Yamaji S, Nagashima Y, NishikawaM, YoshimotoN, KidoY, et al. Prognostic value of integrin beta1-ILK-pAkt signaling pathwayin non–small cell lung cancer. Hum Pathol 2007;38:1081–91.

43. Wang-Rodriguez J, Dreilinger AD, Alsharabi GM, Rearden A. Thesignaling adapter protein PINCH is up-regulated in the stroma ofcommon cancers, notably at invasive edges. Cancer 2002;95:1387–95.

44. Takanami I. Increased expression of integrin-linked kinase is associ-ated with shorter survival in non–small cell lung cancer. BMC Cancer2005;5:1.






2013;12:520-529. Published OnlineFirst January 31, 2013.Mol Cancer Ther Angélique Augustin, Jens Lamerz, Hélène Meistermann, et al. Cell Lines

Small Cell Lung Carcinoma− Wild-Type NonEGFRfrom Gefitinib in Quantitative Chemical Proteomics Profiling Differentiates Erlotinib

Updated version

10.1158/1535-7163.MCT-12-0880doi:

Access the most recent version of this article at:

Material

Supplementary

http://mct.aacrjournals.org/content/suppl/2013/01/31/1535-7163.MCT-12-0880.DC1

Access the most recent supplemental material at:

Cited articles

http://mct.aacrjournals.org/content/12/4/520.full#ref-list-1

This article cites 41 articles, 12 of which you can access for free at:

Citing articles

http://mct.aacrjournals.org/content/12/4/520.full#related-urls

This article has been cited by 4 HighWire-hosted articles. Access the articles at:

E-mail alerts related to this article or journal.Sign up to receive free email-alerts

Subscriptions

Reprints and

[email protected]

To order reprints of this article or to subscribe to the journal, contact the AACR Publications Department at

Permissions

Rightslink site. Click on "Request Permissions" which will take you to the Copyright Clearance Center's (CCC)

.http://mct.aacrjournals.org/content/12/4/520To request permission to re-use all or part of this article, use this link



http://mct.aacrjournals.org/lookup/doi/10.1158/1535-7163.MCT-12-0880

http://mct.aacrjournals.org/content/suppl/2013/01/31/1535-7163.MCT-12-0880.DC1

http://mct.aacrjournals.org/content/12/4/520.full#ref-list-1

http://mct.aacrjournals.org/content/12/4/520.full#related-urls

http://mct.aacrjournals.org/cgi/alerts

mailto:[email protected]

http://mct.aacrjournals.org/content/12/4/520


Documents

Companion Diagnostics & Biomarkers · Companion Diagnostics & Biomarkers Quantitative Chemical Proteomics Proﬁling Differentiates Erlotinib from Geﬁtinib in EGFR Wild-Type Non–Small