Functional analysis of receptor tyrosine kinasemutations in lung cancer identifies oncogenicextracellular domain mutations of ERBB2Heidi Greulicha,b,c,d,1, Bethany Kaplana,d, Philipp Mertinsd, Tzu-Hsiu Chend, Kumiko E. Tanakaa,d, Cai-Hong Yune,Xiaohong Zhanga, Se-Hoon Leea, Jeonghee Choa, Lauren Ambrogiod, Rachel Liaoa,d, Marcin Imielinskia,d,Shantanu Banerjia,d, Alice H. Bergera,d, Michael S. Lawrenced, Jinghui Zhangf, Nam H. Phoa,d, Sarah R. Walkera,Wendy Wincklerd, Gad Getzd, David Franka, William C. Hahna,b,d,g, Michael J. Eckh, D. R. Manid, Jacob D. Jaffed,Steven A. Carrd, Kwok-Kin Wonga,b,c, and Matthew Meyersona,d,g,i,j

aDepartment of Medical Oncology, gCenter for Cancer Genome Discovery, and hCancer Biology, Dana–Farber Cancer Institute, Boston, MA 02115;Departments of bMedicine and iPathology, Brigham and Women’s Hospital, Boston, MA 02115; Department of cMedicine and jPathology, Harvard MedicalSchool, Boston, MA 02115; dBroad Institute of Harvard and MIT, Cambridge, MA 02142; eDepartment of Biophysics, Peking University Health Science Center,Beijing 100191, China; and fDepartments of Biotechnology and Computational Biology, St. Jude Children’s Research Hospital, Memphis, TN 38105

Edited by William Pao, Vanderbilt–Ingram Cancer Center, Nashville, TN, and accepted by the Editorial Board July 24, 2012 (received for review February23, 2012)

We assessed somatic alleles of six receptor tyrosine kinase genesmutated in lung adenocarcinoma for oncogenic activity. Five ofthese genes failed to score in transformation assays; however, novelrecurring extracellular domain mutations of the receptor tyrosinekinase gene ERBB2 were potently oncogenic. These ERBB2 extra-cellular domain mutants were activated by two distinct mecha-nisms, characterized by elevated C-terminal tail phosphorylationor by covalent dimerization mediated by intermolecular disulfidebond formation. These distinct mechanisms of receptor activationconverged upon tyrosine phosphorylation of cellular proteins,impacting cell motility. Survival of Ba/F3 cells transformed to IL-3independence by the ERBB2 extracellular domain mutants wasabrogated by treatment with small-molecule inhibitors of ERBB2,raising the possibility that patients harboring such mutationscould benefit from ERBB2-directed therapy.

HER2 | breast cancer | bladder cancer

Lung cancer is the leading cause of cancer death, accountingfor over 150,000 deaths annually in the United States alone

(1). Current treatment options are thus inadequate for the majorityof patients and additional therapies are needed. Mutationally ac-tivated oncogenes that promote tumorigenesis represent poten-tial drug targets due to frequent dependency of tumor cells onsuch oncogenes (2, 3), and somatically altered receptor tyrosinekinases in particular have been successfully exploited as thera-peutic targets in several cancers.The prototypical therapy targeted to a somatically activated

tyrosine kinase oncogene is imatinib mesylate, which targets theBCR-ABL fusion protein in chronic myelogenous leukemia (4).Targeted therapies developed for lung cancer include gefitiniband erlotinib, small-molecule inhibitors of mutationally activatedEGFR in lung adenocarcinoma (5–8), and crizotinib, a small-molecule inhibitor of the EML4-ALK translocation product inlung adenocarcinoma (9). Trastuzumab, a monoclonal antibodyinhibitor targeting ERBB2, and the small-molecule EGFR/ERBB2 inhibitor lapatinib are effective in ERBB2-amplifiedpatients with breast cancer (10, 11).The advent of next-generation sequencing technologies has

enabled compilation of large somatic mutation datasets fromcancer sequencing studies. Statistical methods that examinedifferences in gene mutation frequency can reveal evidence ofpositive selection; however, demonstration of the contribution ofa mutated gene to tumorigenesis additionally requires functionalvalidation. To identify new lung cancer oncogenes, we system-atically assessed somatic alleles of significantly mutated receptortyrosine kinase genes reported in patients with lung adenocar-cinoma (12) for activity in cellular transformation assays.

Although most receptor tyrosine kinase mutations tested failedto score, novel extracellular domain mutations of ERBB2 wereoncogenic. Our results indicate a unique therapeutic opportunityfor patients with lung and breast cancer who harbor extracellulardomain mutations of ERBB2.

ResultsExtracellular Domain Mutations of ERBB2 Found in Cancer areOncogenic. In the most comprehensive lung adenocarcinoma tar-geted sequencing experiment thus far, 623 genes were sequenced in188 lung adenocarcinomas, identifying 1,013 nonsynonymous so-maticmutations and 26 significantly mutated genes (12). In additionto mutated genes already well characterized in lung adenocarci-noma (13), the significant genes included known tumor suppressorsand several receptor tyrosine kinases, putative but unprovenoncogenes. In an effort to determinewhether these uncharacterizedreceptor tyrosine kinase mutations are oncogenic, we analyzed thefour most significantly mutated receptor tyrosine kinase genesidentified by multiple statistical methods, EPHA3, ERBB4, FGFR4,and NTRK3, and two genes that failed to achieve statistical signif-icance, NTRK2 and ERBB2, due to a cluster of mutations in thekinase domain of NTRK2 and an extracellular domain mutationof unknown significance in ERBB2 (Fig. S1). We expressed themutant alleles in NIH 3T3 cells and examined oncogenic activity insoft agar assays.None of the somatic alleles of EPHA3, ERBB4, FGFR4,NTRK2,

or NTRK3 were found to support anchorage-independent pro-liferation in soft agar assays (Figs. S1 and S2A). In contrast, ectopicexpression of FGFR4 V550E, recurrent in rhabdomyosarcoma andoncogenic inNIH 3T3 cells (14), andFGFR4K645E,modeled afterthe activating FGFR3 K650E mutation found in multiple can-cers (15), resulted in soft agar colony formation (Fig. S2A).Moreover, we could not detect EPHA3 protein expression inthree lung cancer cell lines harboring EPHA3 mutations (Fig.S2B). Somatic mutations of EPHA3, ERBB4, FGFR4, NTRK2,andNTRK3 reported in lung adenocarcinoma thus do not conferphenotypes expected of receptor tyrosine kinase oncogenes.

Author contributions: H.G., P.M., S.R.W., W.W., D.F., J.D.J., S.A.C., and K.-K.W. designedresearch; H.G., B.K., P.M., T.-H.C., K.E.T., S.-H.L., L.A., R.L., and W.W. performed research;M.S.L. and G.G. contributed new reagents/analytic tools; H.G., P.M., C.-H.Y., X.Z., J.C., M.I.,S.B., A.H.B., M.S.L., J.Z., N.H.P., W.W., G.G., D.F., W.C.H., M.J.E., D.M., K.-K.W., and M.M.analyzed data; and H.G. and M.M. wrote the paper.

The authors declare no conflict of interest.

This article is a PNAS Direct Submission. W.P. is a guest editor invited by the EditorialBoard.1To whom correspondence should be addressed. E-mail: heidig@broadinstitute.org.

This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.1073/pnas.1203201109/-/DCSupplemental.

14476–14481 | PNAS | September 4, 2012 | vol. 109 | no. 36 www.pnas.org/cgi/doi/10.1073/pnas.1203201109

Dow

nloa

ded

by g

uest

on

Feb

ruar

y 23

, 202

0

Of the four mutations reported in ERBB2, S310F andA775_G776insYVMA (“insYVMA”) are predicted to encodethe full-length protein (Fig. S1). Whereas the insYVMA mutationof the kinase domain of ERBB2 is already well characterized(16, 17), mutations of the extracellular domain have not beenfunctionally analyzed. We therefore focused on the S310F mu-tation in exon 8 of ERBB2, found in 1/188 lung adenocarcinomasamples (12). Additional reports of extracellular domain muta-tions of ERBB2 included G309E in 1/183 breast cancer samplesand S310Y in 1/63 squamous lung cancer samples (18), S310F in2/112 breast cancers (19), 1 S310F and 1 S310Y in 258 lungadenocarcinomas sequenced by the Cancer Genome Atlas Net-work (Fig. S3 A and B), S310F in 1/65 breast cancers (20), andS310F in 1/316 ovarian cancers (21). An S310F mutation wasalso found in a bladder cancer cell line, 5637 (22).We examined genomic data for samples with extracellular do-

main mutations of ERBB2. One breast cancer sample harboredan additional kinase domain mutation of ERBB2, L755S, and onelung cancer sample harbored a mutation of KRAS, G12F (Fig.S3C); none had mutations of EGFR. One breast cancer sampleexhibited high-level amplification of ERBB2 in the genome, whereasthe other samples did not (Fig. S3C). Two of four patients with lungcancer were former smokers. However, given the small number ofsamples analyzed, we lack power to determine whether thereare any systematic associations of ERBB2 extracellular domainmutations with the presence or absence of other known drivermutations, ERBB2 amplification, or smoking status.NIH 3T3 cells overexpressing wild-type ERBB2 exhibited a

weak anchorage-independent phenotype (Fig. 1 A and B), con-sistent with previous reports (23). In contrast, the G309E, S310F,and S310Y mutants supported robust colony formation in soft agar(Fig. 1 A and B), similar to an ERBB2 kinase domain insertionmutant (16, 24–26). A kinase-inactive mutant, D845A, failed toform any colonies. AALE human lung epithelial cells were simi-larly transformed to anchorage independence by the extracellularmutants of ERBB2 (Fig. 1 C and D). We have thus identifiedoncogenic somatic mutations of the extracellular domain ofERBB2 in lung and breast cancer, occurring at a rate of about 1%,approximately half that of the ERBB2 kinase domain mutationspreviously described in lung adenocarcinoma (12, 24, 26).ERBB2 was reported to be a significantly mutated gene in

glioblastoma (27). Paradoxically, only three of the reportedmutations, C311R, E321G, and C334S, were transforming (Fig. 1E and F). Upon closer inspection, it became apparent that all 7glioblastoma samples harboring ERBB2 mutations were froma single TCGA sample batch. Because the reported data werederived from 91 samples in four sample batches, it was unlikelythat all 7 samples would be in the same sample batch by chance(Fisher’s exact P = 0.000002). Nor could prior patient treatmentaccount for this cluster of ERBB2-mutated samples. This analysisraised the possibility that the reported mutations were artifactsof whole-genome amplification, as the mutations in the TCGAstudy were not validated in unamplified DNA.Sequenom genotyping of the original unamplified glioblas-

toma DNA samples, corresponding to the whole-genome am-plified material in which the ERBB2 mutations were discovered,failed to detect the reported ERBB2 mutations (Fig. S4A),whereas most mutations reported in other significantly mutatedgenes in this sample batch were present in the unamplified DNA(Fig. S4B). Consistent with the possibility that these ERBB2mutations are artifacts, no mutations of ERBB2 were reported in aparallel study (28). Because of these inconsistencies, we checkedunamplified tumor DNA for the three mutations in lung andbreast cancer originally detected in whole-genome amplifiedmaterial. All three mutations were confirmed in native DNA(Figs. S3C and S4 C–F).

ERBB2 Extracellular Domain Mutations Activate the Receptor by TwoDistinct Mechanisms. The oncogenic mutations of the ERBB2 ex-tracellular domain cluster in subdomain II, a region characterizedby 11 disulfide bonds (29). Because two ERBB2 mutants with

in vitro transforming activity, C311R and C334S, affect cysteineresidues, we examined the crystal structure of ERBB2 (29) to askwhether these changes affect disulfide bonds. Both C311 and C334are involved in disulfide bond formation, with C299 and C338,respectively (Fig. 2A). These intramolecular disulfide bonds arepresumably disrupted in the C311R and C334S mutants.We hypothesized that disruption of intramolecular disulfide

bonds might result in formation of intermolecular disulfide bondsby the remaining unpaired cysteines. We tested this hypothesis byrunning lysates on nonreducing and reducing gels in parallel.Whereas wild-type ERBB2 showed no evidence of dimerizationunder these conditions, C311R and C334S formed high-molecular-weight species consistent with ERBB2 dimers on nonreducingSDS/PAGE gels (Fig. 2B). E321G did as well, possibly due todisruption of salt bridges that E321 forms with K369 and R434that stabilize the structure of the disulfide-bonded loops (Fig. 2 Aand B). In contrast, there was no evidence of dimerization by thetransforming insYVMA kinase domain mutant (Fig. 2B).We then examined the mechanism of activation of the mutants

found in breast and lung cancer. The S310F mutant proteinwas hyperphosphorylated, similar to the kinase domain mutantinsYVMA (Fig. 3A). However, the C-terminal tail of the G309Emutant was not hyperphosphorylated (Fig. 3A), like that of other

pBp wt L49H T216S C311R

C334SD326GE321GN319D

insV

V750E

D845AV777A insYVMA

E

FpBp

wt

L49H

T216S

C311R

N319D

E321G

D326G

C334S

V750E

V777A

insYVMA

insV

D845A

ERBB2

Phospho-ERBB2Y1221/1222

Actin

A

B

wt

G309E

S310F

insYVM

A

ERBB2

Actin

wt G309E S310FpBp S310Y

D845AinsYVMA

S310Y

D845A

pBp

CD

wt

G309E

S310F

insYVMA

S310Y

D845A

pBp

ERBB2

Vinculin0

20

40

60

80

100

pB

p wt

G3

09

E

S31

0F

S31

0Y

insY

VM

A

D8

45

A

Nu

mb

er o

f C

olo

nie

s

AALE

Fig. 1. Extracellular domain mutations of ERBB2 found in lung and breastcancer are oncogenic. (A) NIH 3T3 cells expressing ERBB2 extracellularmutants were assessed for colony formation in soft agar. (B) Anti-ERBB2immunoblot on NIH 3T3 lysates. (C) AALE human airway epithelial cellsexpressing ERBB2 extracellular mutants also exhibited an increase in softagar colony formation. (D) Anti-ERBB2 immunoblot on AALE lysates. (E) NIH3T3 cells expressing ERBB2 mutants reported in glioblastoma were assessedfor colony formation in soft agar. (F) Immunoblot analysis of ERBB2 proteinand phosphorylation state on lysates of NIH 3T3 expressing ERBB2mutationsreported in glioblastoma. pBp, pBabe puro vector; insYVMA, A775_G776-insYVMA; insV, ERBB2 A775_G776insV/G776C; WT, wild-type ERBB2.

Greulich et al. PNAS | September 4, 2012 | vol. 109 | no. 36 | 14477

CELL

BIOLO

GY

Dow

nloa

ded

by g

uest

on

Feb

ruar

y 23

, 202

0

mutants of ERBB2 that dimerized by intermolecular disulfidebonding (Figs. 1F and 3A). We therefore investigated the di-merization capacity of ERBB2 G309E and found that this mutantdid indeed form reduction-sensitive dimers (Fig. 3B).There are six cysteine residues involved in three intramolecular

disulfide bonds in the region below the dimerization arm ofERBB2; replacement of any of these six cysteines with serineconferred the ability to form reduction-sensitive dimers and trans-form NIH 3T3 cells (Fig. S5 A and B). A decrease in C-terminalphosphorylation was also observed on the ERBB2 cysteinemutants despite a robust soft agar phenotype (Fig. S5C).

ERBB2 Extracellular Domain Mutants Effect Transformation UsingCommon Downstream Machinery.We have thus defined two distinctmechanisms of activation of extracellular domain mutants ofthe ERBB2 receptor tyrosine kinase: elevation of C-terminalphosphorylation and formation of disulfide-linked dimers. Inorder to determine whether these two classes of ERBB2 mutantsuse similar pathways to effect oncogenic transformation, we usedstable isotope labeling by amino acids in cell culture (SILAC)combined with immunoaffinity enrichment of tyrosine-phosphor-ylated peptides to compare differences in global protein tyrosinephosphorylation using quantitative mass spectrometry.Whereas only a slight increase in phosphopeptide ratios was

seen in the ERBB2 G309E-expressing cells over wild type, thecells expressing ERBB2 S310F exhibited a more substantial in-crease in peptide phosphorylation (Fig. S6), correlating with thegreater oncogenic activity of S310F. Forty-four of 47 endogenousproteins with peptides phosphorylated twofold or higher in the

ERBB2 S310F-expressing cells (Table 1) were also hyper-phosphorylated in the G309E-expressing cells (Dataset S1). Fur-thermore, the 92 individual peptides phosphorylated twofold orhigher in the ERBB2 S310F-expressing cells compared withERBB2 wild-type–expressing cells exhibit a fold change distri-bution that is skewed toward the top of the list of hyper-phosphorylated peptides in the G309E-expressing cells in astatistically significant manner, with a rank-test P value of 2.2 ×10−16 (SI Experimental Procedures). These data indicate that de-spite activation by distinct mechanisms, the two ERBB2 mutantsuse similar downstream effector pathways to transform cells.ERBB2 itself was hyperphosphorylated in S310F-expressing

cells but not G309E-expressing cells, consistent with immunoblotdata (Table 1, Dataset S1, and Fig. 3A). Interestingly, the EGFR/ERBB2 inhibitor MIG6, encoded in human DNA by ERRFI1,was hyperphosphorylated in both the G309E- and S310F-expressing cells (Table 1 and Dataset S1), correlating with thepreviously described dependence of association with ERBB2 onERBB2 activity but not C-terminal autophosphorylation (30).A number of proteins regulating cytoskeletal dynamics and

cell motility were found to be prominently hyperphosphorylatedin the ERBB2 S310F cells (Table 1), including the murine homo-logs of CRK, DLG1, CCD88A, IQGAP, and PEAK1, as well ascomponents of the cytoskeleton (Table 1). Altered cell motility maythus be closely linked to the transformed phenotype measured bythe soft agar assay. PTPN11, a phosphatase involved in activationof Erk proteins in response to growth factor stimulation andintriguingly required for growth and metastasis of HER2-positivebreast cancer cells (31, 32), was also prominently phosphorylatedin the ERBB2 S310F cells. Of note, there was considerable overlapbetween the proteins phosphorylated in response to ERBB2 S310Fexpression and proteins reported to be phosphorylated in humanmammary epithelial cells in response to knockdown of PTPN12,a negative regulator of EGFR and ERBB2 (33).

Oncogenic Activity of ERBB2 Extracellular Domain Mutants Is Sensitiveto Treatment with ERBB2 Inhibitors. Introduction of a kinase-inacti-vating D845A mutation into cDNAs harboring ERBB2 extracel-lular domain mutations prevented soft agar colony formation bytransduced NIH 3T3 cells (Fig. S5B, Bottom Right). To facilitateinhibitor testing, we expressed the ERBB2 mutants in murine Ba/F3 cells and derived IL-3 independent lines. Expression of the

A

B

pBp

wt

C311R

E321G

C334S

V777A

insYVMA

D845A

pBp

wt

C311R

E321G

C334S

V777A

insYVMA

D845A

Nonreducing Reducing

ERBB2

Actin

Monomer

Dimer

Fig. 2. Oncogenic extracellular domain mutations of ERBB2 reported inglioblastoma cause disulfide bond remodeling. (A) Model of the ERBB2 di-mer made by superimposing the human [Protein Data Bank (PDB) ID code2A91] and rat (PDB ID code 1N8Y) ERBB2 extracellular domain crystalstructures onto an EGF-bound EGFR extracellular domain dimer crystalstructure (PDB ID code 1IVO). Intramolecular disulfide bonds are indicated ingreen. (B) Immunoblot analysis of ERBB2 extracellular mutants reported inglioblastoma reveals formation of covalent dimers on nonreducing gels.pBp, pBabe puro vector; WT, wild-type ERBB2.

pBp

wt

C311R

E321G

insYVMA

G309E

S310F

ERBB2

Phospho-ERBB2

Actin

A

B

pBp

wt

G309E

S310F

S310Y

E321G

insYVMA

D845A

ERBB2Nonreducing

ERBB2Reducing

ActinNonreducing

Fig. 3. ERBB2 mutants found in lung and breast cancer form reduction-sensitive dimers that exhibit diminished C-terminal tail phosphorylation. (A)Immunoblot analysis of ERBB2 protein and tyrosine 1221/1222 phosphory-lation state on lysates of NIH 3T3 expressing ERBB2 extracellular domainmutations. (B) Immunoblot analysis of ERBB2 dimers trapped by non-reducing SDS/PAGE. pBp, pBabe puro vector; WT, wild-type ERBB2.

14478 | www.pnas.org/cgi/doi/10.1073/pnas.1203201109 Greulich et al.

Dow

nloa

ded

by g

uest

on

Feb

ruar

y 23

, 202

0

extracellular mutations of ERBB2 conferred IL-3 independencemore efficiently than wild-type ERBB2, whereas the vectorcontrol and kinase-inactive form of ERBB2 were not able tosupport IL-3–independent growth (Fig. 4A).Ba/F3 cells transformed with the ERBB2 extracellular domain

mutants were treated with the irreversible ERBB2 inhibitorsneratinib and afatinib, resulting in effective abrogation of cellsurvival, with IC50s in the low nanomolar range (Fig. 4 B and C).Cells expressing the extracellular domain mutants exhibited in-creased sensitivity to these inhibitors relative to cells expressing

the wild-type ERBB2 or the kinase domain mutant, insYVMA.Importantly, the 95% confidence intervals of the IC50s for ex-tracellular domain mutants S310F, S310Y, and E321G in responseto treatment with small-molecule inhibitors were generally lowerthan the corresponding limits for wild-type ERBB2 or insYVMA(Fig. S7A). Inhibitor efficacy furthermore correlated with in-hibition of ERBB2 phosphorylation (Fig. S7 B–E).The reversible inhibitor lapatinib was 5- to 10-fold less effective

than neratinib and afatinib (Fig. 4D), perhaps due to the moreefficient recovery of receptor activity, evidenced by increases inphospho-ERBB2 and phospho-Akt following inhibitor washout inlapatinib-treated cells but not in neratinib-treated cells (Fig. S8).However, cells expressing the extracellular domain mutants weresignificantly more sensitive to lapatinib than cells expressinginsYVMA (Fig. 4D). Trastuzumab treatment effectively inhibi-ted survival of Ba/F3 cells expressing mutants of G309 and S310,but curiously had less of an effect on cells transformed by theother mutants (Fig. 4E). Although the cancer-derived mutationsare located in the same region of the receptor as the epitopebound by trastuzumab, these results indicate that mutations ofG309 or S310 do not inhibit trastuzumab binding.We have previously shown that the lung cancer cell line NCI-

H1781, harboring an ERBB2 kinase domain mutation, is sensitiveto treatment with the irreversible inhibitor afatinib (34). Incontrast, the endometrial cancer cell line AN3CA is character-ized by FGFR2 mutation but not by ERBB2 mutation (35). Usingthese two cancer cell lines as controls, we tested whether ERBB2inhibition affected survival of a bladder cancer cell line, 5637,harboring an ERBB2 S310F mutation (22). Whereas ERBB2inhibition alone was effective against the NCI-H1781 cells,a combination of ERBB2 inhibition and Mek inhibition wasnecessary for abrogation of 5637 cell survival with an IC50comparable to that for the NCI-H1781 cells (Fig. 4F). Neitherinhibitor had a significant effect on the AN3CA cells alone or incombination. These results suggest a possible treatment option forpatients with lung and breast cancer harboring these mutations.

DiscussionWe functionally analyzed mutated receptor tyrosine kinase genesfound in lung adenocarcinomas. None of the somatic alleles ofEPHA3, ERBB4, FGFR4, NTRK3, or NTRK2 were found to beoncogenic in NIH 3T3 cells. There are three possible explan-ations for the lack of oncogenic transformation by these mutantreceptor tyrosine kinase genes. The reported significantly mu-tated genes may in fact be tumor suppressor genes, contributingto tumorigenesis but not scoring in a transformation assaydesigned to detect dominant gain-of-function oncogenes. Theabsence of nonsense and frameshifting mutations of EPHA3,ERBB4, NTRK2, and NTRK3 argues against a role in tumorsuppression, as all known significantly mutated tumor suppressorgenes found in the lung adenocarcinoma study harbored muta-tions resulting in premature termination.A second explanation for the absence of oncogenic trans-

formation is that the tested somatic alleles are in fact gain-of-function and oncogenic, and we simply used the wrong assay touncover an oncogenic phenotype. This argument is difficult torefute; however, the ability of FGFR4 V550E and K645E and anETV6-NTRK3 gene fusion (34) to support anchorage-independentgrowth argues against such an explanation. Furthermore, we find itunlikely that three of three lung adenocarcinoma cell lines har-boring EPHA3 mutations would fail to express detectableEPHA3 if such mutations were in fact oncogenic.A third explanation for the absence of oncogenic trans-

formation is that the reported mutations are passenger muta-tions, and more refined statistical methods are needed to detectevidence of positive selection. For example, as nonexpressedgenes may exhibit higher mutation rates than expressed genes(37), incorporation of sample-specific gene expression data fromparallel RNA sequencing may assist in a more accurate estimationof gene-specific background mutation rates. Moreover, the effectof replication timing on mutation of individual genes (38, 39) was

Table 1. Proteins containing peptides phosphorylated twofoldor higher in cells expressing ERBB2 S310F than in cells expressingwild-type ERBB2 implicate events that impact cell motility

Gene S310F/WT* No. phosphopeptides†

CSNK1A1 83.5 1CRK 19.3 1DLG3 17.1 1AHNAK 14.8 2DLG1 11.1 1ERBB2 10.2 11CCDC88A 9.3 2IQGAP1 9.3 1ERRFI1 9.1 3SEMA4C 8.0 1PEAK1 7.0 3STXBP3 7.0 1CAV1 6.4 3PTPN11 6.2 1ERBB2IP 5.6 4ANXA1 5.4 2TLN1 5.3 2ANXA2 5.1 9DOK1 5.1 5G6PD 4.9 1TNK2 4.8 1SH2B3 4.6 1LAYN 4.4 2CFL1 4.3 2SHC1 4.2 2ACTB 3.5 1VIM 3.5 3AXL 3.5 3EPS8 3.0 1PLCG1 2.9 1PABPC3 2.8 1CALM1 2.7 2LDLR 2.7 1CTNND1 2.6 1GAB1 2.6 1EEF1A2 2.5 1CDK2 2.4 1VASP 2.2 1RIN1 2.2 1STAT3 2.2 2EEF1A1 2.2 1PIK3R2 2.2 1CDK1 2.2 1MAPK1 2.1 1ABI1 2.1 1RPS27 2.1 2RPS10 2.1 1MYO9B 2.0 1

*Mean fold increase in phosphorylation of the most phosphorylated peptidefor each protein.†Total number of distinct phosphopeptides detected for each protein.

Greulich et al. PNAS | September 4, 2012 | vol. 109 | no. 36 | 14479

CELL

BIOLO

GY

Dow

nloa

ded

by g

uest

on

Feb

ruar

y 23

, 202

0

not modeled into the background mutation rate calculation andmay similarly confound the results of significance testing.In contrast to the other receptor tyrosine kinase mutants

tested, an extracellular domain mutation of ERBB2 transformedNIH 3T3 cells to anchorage independence. This result demon-strates that infrequently mutated but genuine oncogenes can befound in cancer sequencing data even if the genes fail to achievestatistical significance. ERBB2 was reported as significantly mu-tated in glioblastoma (27) but the observed mutations could not beconfirmed in patient-matched unamplified tumor DNA, raisingthe possibility that the reported mutations were artifacts of whole-genome amplification. In response to these data, the TCGAconsortium is no longer to our knowledge performing any se-quence analysis of whole-genome–amplified DNA.We have identified a unique mechanism of activation of ERBB2

in tumor cells, namely, formation of covalent dimers linked byintermolecular disulfide bonds in subdomain II. It is straightforwardto envision how the cysteine substitution mutants of the ERBB2extracellular domain may lead to intermolecular disulfide bonding.G309 is located in close proximity to the C299-C311 disulfide bond;replacement of this compact residue with a bulkier residue suchas glutamate may prevent formation of this intracellular di-sulfide bond, leaving unpaired cysteine residues available for in-termolecular disulfide bond formation. We furthermore speculatethat the S310F and S310Y mutations may result in hydrophobicinteractions between the aromatic rings of the newly introduced310F or 310Y with Y274 and F279 of the neighboring molecule,promoting noncovalent dimerization and kinase activation.There is precedent in the literature for activation of receptor

tyrosine kinases by disulfide-mediated dimerization. Somatic andgerm-line mutations of the extracellular domain of the receptortyrosine kinase FGFR3 that introduce unpaired cysteine residueshave been described in bladder cancer and thanatophoric dysplasia,respectively (15). These mutations caused reduction-sensitive di-mer formation, activated FGFR3 kinase activity, and supportedanchorage-independent proliferation of NIH 3T3 cells (40, 41).

The rat neu oncogene, an ERBB2 ortholog identified as thetransforming agent in nitrosoethylurea-induced rat neuroblasto-mas (42), harbors a mutation corresponding to V659E in thetransmembrane domain of human ERBB2 (43). Mutant neu butnot wild-type neu formed covalent high-molecular-weight speciesunder nonreducing conditions, consistent with disulfide bond-mediated receptor dimerization (44). Mice expressing a mousemammary tumor virus encoding wild-type neu developed mam-mary tumors from which spontaneously mutated forms of neu,characterized by in-frame deletions in domain IV of the extra-cellular domain, could be isolated (45). These spontaneous de-letion mutants, which typically removed a single cysteine residue,were oncogenic and migrated as dimers on nonreducing gels (46).The availability of inhibitors effective against the extracellular

ERBB2 mutants present in lung, breast, ovarian, and bladdercancer raises therapeutic possibilities. The efficacy of a combi-nation of ERBB2 and Mek inhibition on a bladder cancer cellline harboring an ERBB2 S310F mutation further indicates theclinical utility of this approach; however, it is not yet mechanis-tically clear why both inhibitors are necessary for this effect,requiring further study. More broadly, our results suggest thata clinical trial of ERBB2 inhibition, alone or in combinationwith other agents, in patients with cancer harboring extracellulardomain mutations of ERBB2 across tumor types is warranted.

Experimental ProceduresCell Culture. NIH 3T3 cells (ATCC) were maintained in DMEM (Cellgro) sup-plemented with 10% (vol/vol) calf serum (Invitrogen). Ba/F3 cells weremaintained in RPMI 1640 (Cellgro) with 10% FCS (Gemini Bioproducts) and10 ng/mL interleukin-3 (BD Biosciences). AALE cells were grown in SAGMmedia (Lonza). NCI-H1781 and 5637 cells were grown in RPMI 1640 with 10%FCS, and AN3CA cells were grown in MEM (Cellgro) with 10% FCS.

Retroviral Transduction. cDNAs were ectopically expressed in NIH 3T3, AALE,and Ba/F3 using a Gateway-modified pBabe puro vector. Details can be foundin SI Experimental Procedures.

BA

Vec+IL-3 wild-type G309E S310F S310Y E321G insYVMA

IC50 (µM) 1.3090 0.0057 0.0032 0.0027 0.0029 0.0038 0.0054

Vec+IL-3 wild-type G309E S310F S310Y E321G insYVMA

IC50 (µM) 0.7031 0.0023 0.0011 0.0013 0.0012 0.0020 0.0023

Vec+IL-3 wild-type G309E S310F S310Y E321G insYVMA

IC50 (µM) 2.5260 0.0201 0.0107 0.0094 0.0106 0.0207 0.1755

Vec+IL-3 wild-type G309E S310F S310Y E321G insYVMA

IC50 (µg/ml) 0.2612 0.1563 0.0882 0.0508 0.0531 0.2955 0.2512

DC

H1781 AN3CA 5637 H1781+PD AN3CA+PD 5637+PD

IC50 (µM) 0.220 14.480 2.032 0.112 3.128 0.215

E F

Fig. 4. Ba/F3 cells transformed to IL-3 independencewith the ERBB2 extracellular domain mutants aresensitive to ERBB2 inhibition. (A) Proliferation of Ba/F3 cells expressing mutant forms of ERBB2 upon IL-3withdrawal. (B) Survival of ERBB2-transformed Ba/F3cells in response to neratinib. (C) Survival of ERBB2-transformed Ba/F3 cells in response to afatinib. (D)Survival of ERBB2-transformed Ba/F3 cells in responseto lapatinib. (E) Survival of ERBB2-transformed Ba/F3cells in response to trastuzumab. (F) Response ofcancer cell lines NCI-H1781, AN3CA, and 5637 to acombination of Mek and ERBB2 inhibition. Theconcentration of Mek inhibitor PD184352, 1 μM,was chosen for lack of an effect alone on survivalof these cell lines.

14480 | www.pnas.org/cgi/doi/10.1073/pnas.1203201109 Greulich et al.

Dow

nloa

ded

by g

uest

on

Feb

ruar

y 23

, 202

0

Soft Agar Assay. Soft agar assays were performed as described in ref. 47.Briefly, 5 × 103 to 5 × 104 cells were suspended in media containing 0.33%Select agar (Invitrogen) and plated on a bottom layer of media containing0.5% Select agar in a six-well plate. Plates were incubated at 37 °C 2 wkbefore imaging. AALE colonies were photographed at 5× magnification andone field was counted 2–3 wk after plating.

Inhibitor Assays. Totals of 2,000–4,000 Ba/F3 or 3,000 NCI-H1781, 5637, andAN3CA cells were seeded in 96-well plates, incubated with the indicatedconcentrations of inhibitor for 3 d, and assessed for cell survival with theWST-1 reagent (Roche). PD184352 (CI-1040) was obtained from Sigma. Afati-nib, neratinib, and trastuzumab were purchased commercially, and lapatinibwas purified from patient-discarded tablets by James Bradner. Survival datawere analyzed using the Prism GraphPad software.

Dimerization Assay. The dimerization assay was performed as described in ref.46. Briefly, cells were washed twice with cold PBS containing 10 mM iodoace-tamide (Sigma) and lysed with 500 μL TGP buffer (50 mM Tris, pH 7.4, 1% Triton,10% glycerol, 10 mM iodoacetamide) containing protease inhibitors (Roche) and

phosphatase inhibitors (Calbiochem). One aliquot was boiled in NuPage LDS 4×sample buffer (Invitrogen) containing 100 mM DTT (final concentration of 20mM DTT) and one aliquot was boiled in sample buffer without reducing agent.Samples were run on 4–12% gradient polyacrylamide gels (Invitrogen).

SILAC Experiments. Experiments were performed as described in ref. 47.Full experimental details and statistical methods can be found in SIExperimental Procedures.

ACKNOWLEDGMENTS. We thank Dr. James Bradner for providing lapatinib,Dr. Jesse Boehm for supplying receptor tyrosine kinase cDNAs, Dr. Somase-kar Seshagiri for genotyping of ERBB2 mutations found in ref. 18 in nativeDNA, Dr. Emanuele Pelscandolo and Ms. Christina Go of the Dana-FarberCancer Institute Center for Cancer Genome Discovery for genotyping theERBB2 mutation found in ref. 12 in native DNA, Drs. Angela Brooks andAndrew Cherniack for assistance with genomic data, and Drs. Hideo Wata-nabe and Rameen Beroukhim for helpful discussions. H.G. is supported bya grant from Uniting Against Lung Cancer. This work was also supported inpart by National Cancer Institute Grants R01CA109038, R01CA116020, andP20CA90578 (to M.M.); the American Lung Association; the Seaman Foun-dation; and the Monopoli Foundation (M.M.).

1. Siegel R, Ward E, Brawley O, Jemal A (2011) Cancer statistics, 2011: The impact ofeliminating socioeconomic and racial disparities on premature cancer deaths. CACancer J Clin 61:212–236.

2. Sharma SV, Settleman J (2010) Exploiting the balance between life and death: Tar-geted cancer therapy and “oncogenic shock” Biochem Pharmacol 80:666–673.

3. Weinstein IB (2002) Cancer. Addiction to oncogenes—the Achilles heal of cancer.Science 297:63–64.

4. Druker BJ, et al. (2001) Activity of a specific inhibitor of the BCR-ABL tyrosine kinase inthe blast crisis of chronic myeloid leukemia and acute lymphoblastic leukemia withthe Philadelphia chromosome. N Engl J Med 344:1038–1042.

5. Lynch TJ, et al. (2004) Activating mutations in the epidermal growth factor receptorunderlying responsiveness of non-small-cell lung cancer to gefitinib. N Engl J Med350:2129–2139.

6. Mok TS, et al. (2009) Gefitinib or carboplatin-paclitaxel in pulmonary adenocarci-noma. N Engl J Med 361:947–957.

7. Paez JG, et al. (2004) EGFR mutations in lung cancer: Correlation with clinical responseto gefitinib therapy. Science 304:1497–1500.

8. Pao W, et al. (2004) EGF receptor gene mutations are common in lung cancers from“never smokers” and are associated with sensitivity of tumors to gefitinib and erlo-tinib Proc Natl Acad Sci USA 101:13306–13311.

9. Kwak EL, et al. (2010) Anaplastic lymphoma kinase inhibition in non-small-cell lungcancer. N Engl J Med 363:1693–1703.

10. Geyer CE, et al. (2006) Lapatinib plus capecitabine for HER2-positive advanced breastcancer. N Engl J Med 355:2733–2743.

11. Slamon DJ, et al. (1987) Human breast cancer: Correlation of relapse and survival withamplification of the HER-2/neu oncogene. Science 235:177–182.

12. Ding L, et al. (2008) Somatic mutations affect key pathways in lung adenocarcinoma.Nature 455:1069–1075.

13. Greulich H (2010) The genomics of lung adenocarcinoma: Opportunities for targetedtherapies. Genes Cancer 1:1200–1210.

14. Taylor JG, 6th, et al. (2009) Identification of FGFR4-activating mutations in humanrhabdomyosarcomas that promote metastasis in xenotransplanted models. J Clin In-vest 119:3395–3407.

15. Greulich H, Pollock PM (2011) Targeting mutant fibroblast growth factor receptors incancer. Trends Mol Med 17:283–292.

16. Minami Y, et al. (2007) The major lung cancer-derived mutants of ERBB2 are onco-genic and are associated with sensitivity to the irreversible EGFR/ERBB2 inhibitor HKI-272. Oncogene 26:5023–5027.

17. Wang SE, et al. (2006) HER2 kinase domain mutation results in constitutive phos-phorylation and activation of HER2 and EGFR and resistance to EGFR tyrosine kinaseinhibitors. Cancer Cell 10:25–38.

18. Kan Z, et al. (2010) Diverse somatic mutation patterns and pathway alterations inhuman cancers. Nature 466:869–873.

19. Banerji S, et al. (2012) Sequence analysis of mutations and translocations across breastcancer subtypes. Nature 486:405–409.

20. Shah SP, et al. (2012) The clonal and mutational evolution spectrum of primary triple-negative breast cancers. Nature 486:395–399.

21. Cancer Genome Atlas Research Network (2011) Integrated genomic analyses ofovarian carcinoma. Nature 474:609–615.

22. Barretina J, et al. (2012) The Cancer Cell Line Encyclopedia enables predictive mod-elling of anticancer drug sensitivity. Nature 483:603–607.

23. Di Fiore PP, et al. (1987) erbB-2 is a potent oncogene when overexpressed in NIH/3T3cells. Science 237:178–182.

24. Shigematsu H, et al. (2005) Somatic mutations of the HER2 kinase domain in lungadenocarcinomas. Cancer Res 65:1642–1646.

25. Shimamura T, et al. (2006) Non-small-cell lung cancer and Ba/F3 transformed cells har-boring the ERBB2 G776insV_G/C mutation are sensitive to the dual-specific epidermalgrowth factor receptor and ERBB2 inhibitor HKI-272. Cancer Res 66:6487–6491.

26. Stephens P, et al. (2004) Lung cancer: Intragenic ERBB2 kinase mutations in tumours.Nature 431:525–526.

27. Cancer Genome Atlas Research Network (2008) Comprehensive genomic characteriza-tion defines human glioblastoma genes and core pathways. Nature 455:1061–1068.

28. Parsons DW, et al. (2008) An integrated genomic analysis of human glioblastomamultiforme. Science 321:1807–1812.

29. Cho HS, et al. (2003) Structure of the extracellular region of HER2 alone and incomplex with the Herceptin Fab. Nature 421:756–760.

30. Fiorentino L, et al. (2000) Inhibition of ErbB-2 mitogenic and transforming activity byRALT, a mitogen-induced signal transducer which binds to the ErbB-2 kinase domain.Mol Cell Biol 20:7735–7750.

31. Aceto N, et al. (2012) Tyrosine phosphatase SHP2 promotes breast cancer progressionand maintains tumor-initiating cells via activation of key transcription factors anda positive feedback signaling loop. Nat Med 18:529–537.

32. Shi ZQ, Yu DH, Park M, Marshall M, Feng GS (2000) Molecular mechanism for theShp-2 tyrosine phosphatase function in promoting growth factor stimulation of Erkactivity. Mol Cell Biol 20:1526–1536.

33. Sun T, et al. (2011) Activation of multiple proto-oncogenic tyrosine kinases in breastcancer via loss of the PTPN12 phosphatase. Cell 144:703–718.

34. Li D, et al. (2008) BIBW2992, an irreversible EGFR/HER2 inhibitor highly effective inpreclinical lung cancer models. Oncogene 27:4702–4711.

35. Dutt A, et al. (2008) Drug-sensitive FGFR2 mutations in endometrial carcinoma. ProcNatl Acad Sci USA 105:8713–8717.

36. Wai DH, et al. (2000) The ETV6-NTRK3 gene fusion encodes a chimeric protein tyro-sine kinase that transforms NIH3T3 cells. Oncogene 19:906–915.

37. Lee W, et al. (2010) The mutation spectrum revealed by paired genome sequencesfrom a lung cancer patient. Nature 465:473–477.

38. Chen CL, et al. (2010) Impact of replication timing on non-CpG and CpG substitutionrates in mammalian genomes. Genome Res 20:447–457.

39. Stamatoyannopoulos JA, et al. (2009) Human mutation rate associated with DNAreplication timing. Nat Genet 41:393–395.

40. Chesi M, et al. (2001) Activated fibroblast growth factor receptor 3 is an oncogenethat contributes to tumor progression in multiple myeloma. Blood 97:729–736.

41. d’Avis PY, et al. (1998) Constitutive activation of fibroblast growth factor receptor 3by mutations responsible for the lethal skeletal dysplasia thanatophoric dysplasiatype I. Cell Growth Differ 9:71–78.

42. Padhy LC, Shih C, Cowing D, Finkelstein R, Weinberg RA (1982) Identification ofa phosphoprotein specifically induced by the transforming DNA of rat neuroblasto-mas. Cell 28:865–871.

43. Bargmann CI, Hung MC, Weinberg RA (1986) Multiple independent activations of theneu oncogene by a point mutation altering the transmembrane domain of p185. Cell45:649–657.

44. Weiner DB, Liu J, Cohen JA, Williams WV, Greene MI (1989) A point mutation in theneu oncogene mimics ligand induction of receptor aggregation. Nature 339:230–231.

45. Siegel PM, Dankort DL, Hardy WR, Muller WJ (1994) Novel activating mutations inthe neu proto-oncogene involved in induction of mammary tumors. Mol Cell Biol14:7068–7077.

46. Siegel PM, Muller WJ (1996) Mutations affecting conserved cysteine residues withinthe extracellular domain of Neu promote receptor dimerization and activation. ProcNatl Acad Sci USA 93:8878–8883.

47. Ong SE, Mann M (2006) A practical recipe for stable isotope labeling by amino acids incell culture (SILAC). Nat Protoc 1:2650–2660.

Greulich et al. PNAS | September 4, 2012 | vol. 109 | no. 36 | 14481

CELL

BIOLO

GY

Dow

nloa

ded

by g

uest

on

Feb

ruar

y 23

, 202

0