HER2 missense mutations have distinct effects ononcogenic signaling and migrationDaniel J. Zabranskya, Christopher L. Yankaskasb, Rory L. Cochrana, Hong Yuen Wonga, Sarah Croessmanna, David Chua,ShyamM. Kavuric,1, Monica Red Brewerd, D. Marc Rosena, W. Brian Daltona, Ashley Cimino-Mathewsa,e, Karen Craveroa,Berry Buttona, Kelly Kyker-Snowmana, Justin Cidadoa,2, Bracha Erlangera, Heather A. Parsonsa, Kristen M. Mantob,Ron Bosec,f, Josh Lauringa, Carlos L. Arteagad, Konstantinos Konstantopoulosb,g, and Ben Ho Parka,b,3

aSidney Kimmel Comprehensive Cancer Center, The Johns Hopkins University School of Medicine, Baltimore, MD 21287; bDepartment of Chemical andBiomolecular Engineering, Johns Hopkins University, Baltimore, MD 21218; cDivision of Oncology, Department of Medicine, Washington University Schoolof Medicine, St. Louis, MO 63110; dDepartment of Medicine: Breast Cancer Research Program, Vanderbilt-Ingram Cancer Center, Vanderbilt University,Nashville, TN 37232; eDepartment of Pathology, The Johns Hopkins Hospital and School of Medicine, Baltimore, MD 21287; fSiteman Cancer Center,Washington University School of Medicine, St. Louis, MO 63141; and gJohns Hopkins Institute for NanoBioTechnology, Johns Hopkins University, Baltimore,MD 21218

Edited by Peter K. Vogt, The Scripps Research Institute, La Jolla, CA, and approved September 29, 2015 (received for review August 24, 2015)

Recurrent human epidermal growth factor receptor 2 (HER2)missense mutations have been reported in human cancers. Thesemutations occur primarily in the absence of HER2 gene amplificationsuch that most HER2-mutant tumors are classified as “negative” byFISH or immunohistochemistry assays. It remains unclear whethernonamplified HER2 missense mutations are oncogenic and whetherthey are targets for HER2-directed therapies that are currently ap-proved for the treatment ofHER2 gene-amplified breast cancers. Herewe functionally characterize HER2 kinase and extracellular domainmutations through gene editing of the endogenous loci in HER2 non-amplified human breast epithelial cells. In in vitro and in vivo assays,the majority of HER2 missense mutations do not impart detectableoncogenic changes. However, the HER2 V777L mutation increasedbiochemical pathway activation and, in the context of a PIK3CAmutation, enhanced migratory features in vitro. However, the V777Lmutation did not alter in vivo tumorigenicity or sensitivity toHER2-directed therapies in proliferation assays. Our results suggestthe oncogenicity and potential targeting of HER2 missense muta-tions should be considered in the context of cooperating geneticalterations and provide previously unidentified insights into func-tional analysis of HER2 mutations and strategies to target them.

HER2 missense mutations | breast cancer | mutant oncogenes |targeted therapies

Agreat success in the treatment of breast cancer has comefrom the identification of human epidermal growth factor

receptor 2 (HER2)/Neu (ERBB2) amplification/overexpression asa targetable driver in ∼20% of breast cancers (1). HER2 is amember of the ErbB family of transmembrane receptor tyrosinekinases, which includes the epidermal growth factor receptor(EGFR/ErbB1), HER3 (ErbB3), and HER4 (ErbB4) (2). Acti-vation of ErbB signaling causes receptor tyrosine autophosphor-ylation and induces interactions with cytoplasmic signal transductionpartners that promote a wide variety of cellular processes includingproliferation, motility, and escape from apoptosis. In addition totheir key role in normal cellular growth and maintenance, thedysregulation of ErbB receptors has been extensively implicated inthe development of numerous cancers (1).Whereas overexpression or amplification of HER2 has been

well described to deregulate ErbB signaling, cancer genome se-quencing studies have demonstrated that somatic point muta-tions in the HER2 gene occur in a number of cancers, including2–4% of breast cancers (3–6). Importantly, these mutations aremost often found in patients as single copies without amplifica-tion/overexpression of HER2 (HER2-“negative” breast cancers),though HER2 protein expression is often still present. Over-expression studies have implicated a number of these mutationsas activating and oncogenic (4, 7–9). Additionally, one mutation,

L755S, has been described to be associated with lapatinib re-sistance when overexpressed (4, 10).Past studies comparing overexpression of a mutant cDNA

to single nucleotide knockin of mutant oncogenes have showndramatic differences in signaling and transformed phenotypes(11–13). Additionally, we have shown KRAS G12V mutations, assingle copies, have relatively little effect in breast epithelial cells,but have a profound effect on cancerous phenotypes and tu-morigenicity when coupled with a PIK3CA oncogenic hotspotmutation (14). Given these caveats, we used two HER2 non-amplified human breast epithelial cell lines to create an isogenicpanel of HER2 missense knockin mutants, to study the effects ofsingle-copy HER2 mutations under gene expression levels com-parable to what has been observed in HER2-mutant cancers.

Significance

The discovery of human epidermal growth factor receptor 2(HER2) missense mutations in breast and other cancers poten-tially make such tumors susceptible to current and futureHER2-targeted therapies. However, the majority of HER2mutations occur in HER2 nonamplified cancers, and whetherthese mutations will predict for sensitivity to HER2-directedtherapies remains unknown. Using genome editing, the datapresented here suggest that HER2 missense mutations arefunctionally distinct and require additional oncogenic input toimpart cancerous phenotypes. These results suggest that HER2missense mutations by themselves may not be reliable pre-dictors of response to HER2-targeted therapies, a hypothesiscurrently being tested in genomically driven clinical trials.

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

Conflict of interest statement: B.H.P. is a paid consultant for Novartis and is a member ofthe scientific advisory boards of Horizon Discovery, Ltd and Loxo Oncology, and has re-search contracts with Genomic Health, Inc. and Foundation Medicine. Under separatelicensing agreements between Horizon Discovery, Ltd. and The Johns Hopkins University,B.H.P. is entitled to a share of royalties received by the university on sales of products. Theterms of this arrangement are being managed by Johns Hopkins University, in accordancewith its conflict of interest policies. All other authors declare no potential conflicts.

This article is a PNAS Direct Submission.1Present address: Lester and Sue Smith Breast Cancer Center, Baylor College of Medicine,Houston, TX 77019.

2Present address: Oncology iMED, AstraZeneca, Waltham, MA 02451.3To whom correspondence should be addressed. Email: bpark2@jhmi.edu.

This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.1073/pnas.1516853112/-/DCSupplemental.

www.pnas.org/cgi/doi/10.1073/pnas.1516853112 PNAS | Published online October 27, 2015 | E6205–E6214

MED

ICALSC

IENCE

SPN

ASPL

US

Dow

nloa

ded

by g

uest

on

May

19,

202

1

ResultsTargeted Knockin of Single Copy, Heterozygous HER2 MissenseMutations in HER2 Nonamplified Human Breast Epithelial Cell Lines.To model HER2 missense mutations as found in human cancers,we used adeno-associated virus (AAV)-mediated gene targetingto create an isogenic panel of HER2 mutant knockin MCF-10Aand MCF7 human breast epithelial cells. These two cell lines donot overexpress or have amplification of HER2, thus heterozy-gous knockin cell lines contain one wild-type and one mutantcopy of HER2 and express HER2 protein at levels consistentwith HER2 nonamplified tumors (SI Appendix, Fig. S1) (15).Three AAV gene-targeting vectors were used as backbone vec-tors to introduce each of seven previously reported mutationsinto the extracellular or kinase domains of HER2 (Fig. 1). Twoheterozygous HER2-mutant clones were generated for eachHER2 mutation. We also generated wild-type control clonesusing wild-type AAV vector backbones for both MCF-10A andMCF7. Cell lines were verified to have a single integrated copy ofthe desired mutation and equivalent expression of the mutantand wild-type alleles using PCR and RT-PCR followed by Sangersequencing (SI Appendix, Fig. S1). See SI Appendix, Table S1 fora list of cell lines generated and used in this study.

HER2 V777L Mutation Increases HER2 Signaling Pathway Activation inNontransformed MCF-10A Cells. Overexpression studies have iden-tified that HER2 mutations, including G309A, S310F, L755S,V777L, and R896C activate HER2 signaling and promote trans-formation. To assess whether these effects could be recapitulatedin genome edited clones, we initially performed Western blottinganalysis on our MCF-10A isogenic panel. MCF-10A is a non-transformed human breast epithelial cell line with a mostly diploidkaryotype that requires EGF supplementation for proliferation inculture (16). Using genome editing, we have demonstrated that

MCF-10A cells can be transformed by the introduction of onco-genic mutations both in vitro and in vivo (14).We found that MCF-10A HER2 V777L cells showed in-

creases in phosphorylation of HER2, EGFR, and ERK com-pared with control cell lines in 0.2 ng/mL EGF (physiologic)(Fig. 2A) and EGF-free (SI Appendix, Fig. S2) conditions. Nodifference in AKT phosphorylation was noted in gene targetedclones (Fig. 2A and SI Appendix, Fig. S2), indicating that activationof the HER2 receptor by this mutation preferentially activatesERK signaling. Interestingly, other HER2 mutations did notconsistently activate HER2 signaling in MCF-10A cells. Theseresults demonstrate that HER2 mutations, even those occurring inthe same domain of the protein, have distinct effects on signalingpathway activation.

HER2 Mutant MCF-10A Cells Do Not Exhibit Oncogenic Properties inVitro. Prior overexpression studies identified that HER2 muta-tions can lead to transformative changes, including increases inanchorage independent growth and aberrant morphology in 3Dculture, even without detectable increases in HER2 phosphory-lation (4, 7). Therefore, we performed a series of assays to identifytransformed properties of our HER2 mutant cell line panel.Whereas parental MCF-10A cells require EGF supplementa-

tion, EGF independence has been demonstrated to be a featureof transformation in MCF-10A cells (17). However, HER2 mu-tations did not confer EGF independence (SI Appendix, Fig. S2),and there was little difference in proliferation rates betweencontrol and HER2 mutant cells in assay media containing either0.2 ng/mL (physiologic) or 20 ng/mL (maintenance dose) EGF(Fig. 2B and SI Appendix, Fig. S2).We next tested the ability of HER2 mutations to promote

anchorage independent growth in MCF-10A cells. Knockin ofHER2 V777L was not sufficient for colony formation in soft agar,although MCF-10A cells that overexpress HER2 V777L mutant

KinaseECD TM JM Tail

Vector 3 Vector 1 Vector 2A

Alteration Domain Base Change ReportedCases

G309A ECD c.926G>C 1

S310F ECD c.929C>T 30

L755S Kinase c.2264T>C 30

I767M Kinase c.2301C>G 5

V777L Kinase c.2329G>T 11

Y835F Kinase c.2504A>T 1

R896C Kinase c.2686C>T 3

SEPT5’HA

3’HA

EndogenousAllele

TargetingConstruct

TargetedPre-Cre

TargetedPost-Cre

Exon

Exon*

SEPTExon*

Exon*

CB

Fig. 1. Generation of an isogenic HER2 cell line panel. (A) Previously identified HER2 mutations included in the isogenic panel. Three distinct gene targetingvectors were used to introduce the mutations. ECD, extracellular domain; TM, transmembrane region; JM, juxtamembrane region. (B) Genomic locus andprevalence as reported in the COSMIC database as of January 2015 for each introduced HER2 mutation. (C) Representative rAAV-mediated gene targetingstrategy for a targeting vector carrying an exonic mutation (*) within the 5′ homology arm (HA). rAAV transduction leads to integration of the targetingvector via homologous recombination of the 5′ and 3′ HAs. After neomycin selection and clone isolation, loxP (triangles) flanked SEPT cassette is excised usingCre recombinase.

E6206 | www.pnas.org/cgi/doi/10.1073/pnas.1516853112 Zabransky et al.

Dow

nloa

ded

by g

uest

on

May

19,

202

1

cDNA formed robust colonies in soft agar as previously reported(4) (Fig. 2C). Anchorage independent growth in semisolid me-dium was not observed in other MCF-10A HER2 mutant celllines (SI Appendix, Fig. S2). Next, we examined acinar mor-phology of MCF-10A HER2 mutant knockin cells in 3D culture.All HER2 mutant knockin cell lines formed spherical structureswith normal polarization similar to controls (Fig. 2D). This wasdistinct from phenotypes observed with overexpression of HER2mutant cDNAs, which led to larger, irregular, spiculated struc-tures with abnormal protrusions (Fig. 2D).

HER2 V777L and L755S Mutations Increase HER2 Pathway SignalingActivation in MCF7 Cells. Whereas biochemical signaling differ-ences were modest in MCF-10A HER2 knockin cell lines, wehypothesized that HER2 missense mutations may be highlycontext dependent and require cooperating genetic alterationsfound in cancer cells to promote additional transformative fea-tures. Therefore, we created an isogenic panel of HER2 mutantMCF7 cell lines containing single copies of HER2 mutations.The parental MCF7 cell line is derived from a metastatic pleuraleffusion in a patient with ER-positive breast cancer and does notoverexpresses HER2 protein. Two HER2 kinase domain muta-tions, L755S and V777L, and one extracellular domain mutation,G309A, were chosen for genome editing in MCF7 cells becausethese mutations have been characterized via overexpressionstudies to result in increased pathway activation, oncogenicphenotypes, and, in the case of L755S, resistance to lapatinib (4,18). It should be noted that MCF7 contains two copies of anactivating E545K PIK3CA mutation and one wild-type copy ofPIK3CA (19), a known oncogene involved in PI3 kinase andMAP kinase pathway signaling (17).We initially tested the effects of HER2 mutations in MCF7

cells by examining signaling pathway activation. Consistent withthe results in the MCF-10A background, MCF7 HER2 V777Lcells showed increases in phosphorylation of HER2, EGFR, andERK compared with parental MCF7 and control cells in serum-starved conditions (Fig. 3A) and HER2 in serum-supplemented

media (SI Appendix, Fig. S3). Furthermore, the L755S mutationincreased activation of HER2 signaling pathway proteins inMCF7 (Fig. 3A). In addition, MCF7 cells have greater basalexpression of HER3 than MCF-10A cells, and MCF7 HER2V777L and L755S cell lines exhibited an increase in HER3phosphorylation (Fig. 3A and SI Appendix, Fig. S3). Notably,AKT activation did not differ significantly among MCF7 pa-rental, control, and HER2-mutant cell lines. Despite this path-way activation, HER2 mutations did not affect proliferation inMCF7 cells (Fig. 3B and SI Appendix, Fig. S3).

HER2 Missense Mutations Cooperate with Mutant PIK3CA for Aug-mented Pathway Activation inMCF7 Cells.BecauseMCF7HER2 L755Sand V777L cells showed greater signaling activation than theirMCF-10A counterparts, we hypothesized that mutant PIK3CAE545K in MCF7 may cooperate with HER2 mutations to increasepathway activation and other transformed phenotypes. Numerousstudies have implicated PIK3CA mutations in transformation andactivation of the PI3 kinase/AKT pathway (17, 20). To test thishypothesis, we used genome editing to introduce the HER2 V777Lmutation into MCF7 cells that had previously undergone genetargeting to restore the PIK3CA alleles to wild type (referred toas MCF7 corrected) (19). The resulting cells have three wild-type copies of PIK3CA and a heterozygous HER2 V777L mutation(referred to as MCF7 corrected + V777L).MCF7-corrected + V777L cells displayed slightly lower levels

of phosphorylated HER2 and HER3 compared with MCF7HER2 V777L cells, but relatively more than both MCF7 andMCF7-corrected cells (Fig. 3C). These results suggest that theHER2 V777L mutation activates the MAP kinase pathway inMCF7 cells and that mutant PIK3CA may augment this signal-ing, similar to our past studies (17). Interestingly, both MCF7-corrected and MCF7-corrected + V777L cells had dramaticallyreduced AKT phosphorylation. Although it has been shown thatthe absence of PIK3CA mutations in MCF7 cells lowers activa-tion of the PI3 kinase/AKT pathway (19), it was unexpected thatisolated activation of HER2 signaling via the V777L mutation

MCF-10A

Targeted WT 1

Targeted WT 2

Targeted WT 3

G309A

S310FL755S

I767MV777LY835FR896C

P-HER2

Total HER2

P-ERK

Total ERK

P-EGFR

Total EGFR

GAPDH

P-AKT

Total AKT

A

C

B

D

2 5 7 90.5

1248

163264

MCF-10ATargeted WTV777L

Day

Fold

Cha

nge

in G

row

th

MCF-10ATargeted

WT G309A L755SS310F I767M

V777L Y835F R896CG309A

O/EL755S

O/EV777L

O/EMCF-10ATargeted

WT V777LV777L

O/E

Fig. 2. HER2 V777L mutation increases signaling pathway activation but does not promote transformed phenotypes in MCF-10A cells. (A) MCF-10A HER2mutant and control cell lines were grown in physiologic (0.2 ng/mL) EGF supplemented assay media and subjected to Western blotting analysis with theindicated antibodies. White lines represent lanes that have been removed from blot. (B) Relative mean (±SEM) proliferation data for V777L clones and MCF-10A controls in 0.2 ng/mL EGF supplemented assay media (n ≥ 6 per cell line, two independent experiments were performed). (C) Representative imagesdepicting soft agar colony formation for MCF-10A V777L, controls, and an MCF-10A cell line overexpressing the HER2 V777L cDNA. (Scale bar, 200 μm.) (D) 3Dacinar morphogenesis assay for MCF-10A HER2 mutant isogenic cell lines and mutant HER2 overexpression cell lines. (Scale bar, 50 μm.)

Zabransky et al. PNAS | Published online October 27, 2015 | E6207

MED

ICALSC

IENCE

SPN

ASPL

US

Dow

nloa

ded

by g

uest

on

May

19,

202

1

did not lead to any discernible increase in AKT phosphorylation.This result, along with the concurrent increases in ERK phos-phorylation, again suggests that V777L HER2 preferentially ac-tivates the MAP kinase signaling pathway rather than the PI3kinase/AKT signaling pathway. MCF7 corrected and its deriva-tives also grew similarly to each other, but more slowly thanMCF7 parental cells in serum-supplemented conditions (Fig. 3Dand SI Appendix, Fig. S3).

Acinar Morphology and Anchorage Independent Growth in MCF7HER2 Mutant Cell Lines. To determine any morphologic effects ofHER2 mutations in MCF7 cells, we conducted Matrigel assays.All cell lines appeared similar to parental and control cells.When cells were seeded in Matrigel along with the EGFR/HER2inhibitor lapatinib, there was little to no effect on the size ormorphology of the acini (Fig. 4A and SI Appendix, Fig. S4).We then performed soft agar colony formation assays to test

the ability of HER2 mutations to promote anchorage inde-pendent growth in MCF7 cells. Whereas MCF7 V777L clonesshowed a trend toward increased colony number, this wasmodest relative to controls. Similarly, MCF7 HER2 G309A andL755S did not differ significantly from controls (Fig. 4 B and Cand SI Appendix, Fig. S4). Treatment with lapatinib had a stronginhibitory effect on colony size and formation in all tested cell

lines, though there were no significant differences betweenHER2 knockin clones and controls (SI Appendix, Fig. S4).MCF7-corrected + V777L mutant cells did not exhibit trans-formed phenotypes in in vitro assays compared with controls(Fig. 4 and SI Appendix, Fig. S4). These data suggest that,whereas PIK3CA mutations may cooperate with HER2 muta-tions to accentuate signaling activation in MCF7 cells, thiscooperativity is not sufficient to increase anchorage independentgrowth or promote abnormal acinar morphology.

Cooperativity Between HER2 and PIK3CA Mutations Leads to In-creased Signaling Activation in MCF-10A Cells. Because the effectsof HER2 missense mutations as single copies were greater inMCF7 cells compared with MCF7 corrected cells, we wanted toconfirm that activating E545K PIK3CA mutations in MCF7HER2 mutant cells contributed to the observed effects. There-fore, we created PIK3CA and HER2 double mutant cell lines byknockin of the PIK3CA E545K mutation into our MCF-10AHER2 L755S and V777L cell lines. These resultant doubleknockin (DKI) cell lines (referred to as L755S DKI and V777LDKI) are heterozygous for their respective HER2 mutation andfor the PIK3CA E545K mutation.We first analyzed activated signaling pathways via Western

blot in DKI and control cell lines in media with 0.2 ng/mL orwithout EGF. It has been previously demonstrated that MCF-10A cells with knockin of PIK3CA E545K (referred to as MCF-10A + E545K) activated both the PI3 kinase and MAP kinasepathways compared with MCF-10A parental cells (14, 17). MCF-10A HER2 V777L cells had increased levels of HER2 phos-phorylation compared with MCF-10A + E545K cells but hadsimilar levels of ERK phosphorylation (Fig. 5). Additionally,both L755S DKI and V777L DKI cell lines showed increases inHER2, EGFR, and ERK phosphorylation compared with MCF-10A + E545K cells and in ERK activation compared with theirrespective single HER2 mutation knockin cell lines in mediasupplemented with 0.2 ng/mL EGF (Fig. 5A). Differences inEGFR phosphorylation were less apparent in EGF-free mediaconditions (Fig. 5B). The DKI cell lines also showed slight in-creases in AKT phosphorylation compared with PIK3CA singleknockin cells in EGF-free conditions (Fig. 5B), but not in thepresence of 0.2 ng/mL EGF (Fig. 5A). Interestingly, there was anincrease in the ratio of phosphorylated HER2 to total HER2 inL755S DKI and V777L DKI cells compared with their respectivesingle knockin cell lines in EGF-free conditions, which was alsopresent in both L755S DKI and one of two V777L DKI cell linesin 0.2 ng/mL EGF conditions (Fig. 5C). Although HER2 isclassically considered upstream of PI3K, these results indicatethat under certain conditions, interactions between mutantPIK3CA and mutant HER2 may lead to “rewiring” and dysre-gulation of signaling pathways. Similar pathway dysregulationhas been reported in other studies (17, 21, 22).Because PIK3CA mutations have been shown to confer EGF-

independent growth properties to MCF-10A cells, we tested ifHER2 mutations could augment this phenotype. L755S andV777L DKI cells proliferate in EGF-free media, but this growthwas not increased compared with MCF-10A + E545K cells (SIAppendix, Fig. S5A). Furthermore, the combination of HER2and PIK3CA mutations did not confer anchorage independentgrowth to MCF-10A cells, and proliferating colonies were notobserved in soft agar assays (SI Appendix, Fig. S5B). We nextseeded the L755S and V777L DKI cell lines in Matrigel to ex-amine their morphology. Both L755S DKI and V777L DKI cellsgrew with a morphology similar to controls that was not affectedby treatment with lapatinib (SI Appendix, Fig. S5C). Taken to-gether, our results suggest that, whereas signaling pathway acti-vation was increased in DKI cell lines, this is not sufficient topromote transformation in these assays.

C

1 3 5 70.5

1248

163264

128256

MCF7MCF7 CorrectedMCF7 Corrected + V777L

**

Day

Fold

Cha

nge

in G

rowt

h

B

1 3 5 70.1

1

10

100

1000MCF7Targeted WTG309AL755SV777L

Day

Fold

Cha

nge

in G

row

th

MC

F7Targeted W

T 1Targeted W

T 2G

309A 1G

309A 2L755S

1L755S

2V

777L 1V

777L 2

P-HER2

Total HER2

P-ERK

Total ERK

P-HER3Total HER3

P-AKT

Total AKT

GAPDH

P-EGFR

Total EGFR

A

D

MC

F7Targeted W

TV

777LM

CF7 C

orrectedM

CF7 C

orrected + V777L 1

MC

F7 Corrected + V

777L 2

P-HER2Total HER2

P-AKTTotal AKT

P-ERKTotal ERK

GAPDHTotal HER3

P-HER3

Fig. 3. HER2 and PIK3CA mutations cooperate to increase signaling path-way activation in MCF7 cells. (A) MCF7 HER2 mutant and control cell lineswere grown in serum starved conditions and subjected to Western blottinganalysis with the indicated antibodies. (B) Relative mean (±SEM) proliferationfor MCF7HER2mutants in serum supplementedmedia (n ≥ 6, two independentexperiments were performed). (C) Western blotting analysis of MCF7 andMCF7-corrected cell lines and HER2 mutant derivatives as in A. White linesrepresent lanes that have been removed from blot. (D) Mean (±SEM) relativeproliferation for MCF7, MCF7-corrected, and MCF7-corrected + V777L cells as inB (n ≥ 5, two independent experiments were performed, **P ≤ 0.01, two-wayANOVA followed by Bonferroni multiple comparison test).

E6208 | www.pnas.org/cgi/doi/10.1073/pnas.1516853112 Zabransky et al.

Dow

nloa

ded

by g

uest

on

May

19,

202

1

HER2 V777L and PIK3CA E545K Double Mutant Cell Lines Have In-creased Migratory Capacity in Vitro. Because HER2 overexpres-sion and activation of HER2 signaling have been described tolead to increases in cell migration, we next tested the migratorycapacity of HER2 mutant cell lines in scratch wound healing andchemotactic microchannel migration assays. In scratch woundexperiments, MCF-10A V777L DKI cells exhibited significantlyincreased wound closure compared with both MCF-10A + E545K

and L755S DKI cell lines. Treatment with lapatinib significantlyreduced the migration of V777L DKI cells (Fig. 6 A and B). L755SDKI cells exhibited slower wound closure than the MCF-10A +E545K cells, indicating that the ability to promote migration maybe unique to the V777L mutation. Single V777L knockin cells didnot migrate more quickly than MCF-10A parental or control cells(SI Appendix, Fig. S6). In accord with these results, the MCF7HER2 V777L cell line also showed an increase in wound closure

MCF7

Targete

d WT 1

Targete

d WT 2

G309A

1

G309A

2

L755

S 1

L755

S 2

V777L

1

V777L

2

MCF7 Corr

ected

MCF7 Corr

ected

+ V77

7L 1

MCF7 Corr

ected

+ V77

7L 2

0.0

0.5

1.0

1.5

2.0

Rel

ativ

e co

lony

cou

nt

A

CB

Lapatinib

Vehicle

MCF7Targeted

WT G309A L755S V777L

MCF7

V777L Clone 1

V777L Clone 2

Targeted WT G309A

L755S

Fig. 4. Three-dimensional morphology and anchorage independent growth for MCF7 HER2 mutant cell lines. (A) MCF7 control and HER2 mutant cells wereseeded into 3D Matrigel culture in the presence of DMSO vehicle control or lapatinib (5 μM) and photographed on day 12. (Scale bar, 100 μm.) (B) Repre-sentative images depicting soft agar colony formation in MCF7 HER2 knockin cell lines and controls. (Scale bar, 200 μm.) (C) Relative mean (±SEM) number ofcolonies formed in soft agar assays as counted visually via phase contrast microscopy (n ≥ 5 at least two independent experiments were performed).

A B

MCF-10A

Targete

d WTL7

55SV77

7L

10A +

E545K

L755

S DKI 1

L755

S DKI 2

V777L

DKI 1

V777L

DKI 2

0

5

10

15

20

25

Rel

ativ

e pH

ER

2:H

ER

2 ra

tio

MCF-10A

Targete

d WTL7

55SV77

7L

10A +

E545K

L755

S DKI 1

L755

S DKI 2

V777L

DKI 1

V777L

DKI 2

0

2

4

6

8

Rel

ativ

e pH

ER

2:H

ER

2 ra

tio

P-HER2

Total HER2

P-ERK

Total ERK

P-AKT

Total AKT

P-EGFR

Total EGFR

GAPDH

10A ParentTargeted W

TL755SV777L

MCF-10A + E545K

L755S DKI 1L755S DKI 2V777L DKI 1V777L DKI 2

10A ParentTargeted W

TL755SV777L

MCF-10A + E545K

L755S DKI 1L755S DKI 2V777L DKI 1V777L DKI 2

P-HER2

Total HER2

P-ERK

Total ERK

P-EGFR

Total EGFR

P-AKT

Total AKT

GAPDH

C

Fig. 5. Knockin of PIK3CA E545K mutation in HER2 missense cells increases signaling activation in MCF-10A. MCF-10A HER2 mutant, L755S DKI, V777L DKI,and control cell lines were grown in physiologic 0.2 ng/mL EGF supplemented (A) or EGF free (B) assay media and subjected to Western blotting analysiswith the indicated antibodies. White lines represent lanes that have been removed from blot. (C) Bar graphs represent the ratio of pHER2 to total HER2Western blot band quantification relative to the MCF-10A parental cell line in 0.2 ng/mL (Top) and EGF-free (Bottom) conditions as measured with ImageJsoftware.

Zabransky et al. PNAS | Published online October 27, 2015 | E6209

MED

ICALSC

IENCE

SPN

ASPL

US

Dow

nloa

ded

by g

uest

on

May

19,

202

1

compared with parental control, which was decreased with lapati-nib treatment and not present in MCF7-corrected + V777L cells(Fig. 6 C and D and SI Appendix, Fig. S6).We next examined the migratory capacity of cells with and

without HER2 V777L mutation in collagen-I–coated micro-channels using a microfluidic device constructed of poly-dimethylsiloxane (PDMS) (23–26). This device allows for thecharacterization of single-cell migration, rather than the col-lective migration observed during wound closure. Individualcells migrate up a chemotactic gradient inside 10-μm-tall channelsof prescribed widths. Because all tested cells do not efficientlymigrate through narrow channels (10 μm wide or less) we fo-cused our quantitative analysis on cells migrating in 50-μm and20-μm wide channels (24).

MCF-10A V777L DKI cells displayed higher migrationvelocity and persistence (defined as the ratio of net cell dis-placement to total distance traveled) inside the 50-μmchannels compared with MCF-10A + E545K cells (Fig. 6E andMovie S1). Similarly, MCF7 V777L cells exhibited an increasedmigratory propensity compared with parental MCF7 (Fig. 6Fand Movie S2). Although both MCF-10A V777L DKI andMCF7 V777L migrated faster than their corresponding controlsin 20-μm wide channels, no significant difference was detectedin persistence (Fig. 6 E and F and Movies S1 and S2). Takentogether, these studies demonstrate that HER2 V777L coop-erates with the PIK3CA E545K to confer increased migratorypotential in breast epithelial cells.

MCF7

Targete

d WT

G309A

L755

SV77

7L0.0

0.2

0.4

0.6

0.8

Vehicle

*Lapatinib

*

Rel

ativ

e Fr

actio

n of

Wou

nd C

lose

d

MCF-10A + E

545K

L755

S DKI

V777L

DKI

0.0

0.5

1.0

1.5 VehicleLapatinib

*********

Rel

ativ

e Fr

actio

n of

Wou

nd C

lose

d0 Hours 16 Hours + Lapa nib

MCF-10A + E545K

L755S DKI

V777L DKI

0 Hours 20 Hours20 Hours

+ Lapa nib

MCF7

V777L

A

C

B

D

FE

50 200.0

0.2

0.4

0.6

0.8MCF7V777L

**P

ersi

sten

ce

50 200.0

0.1

0.2

0.3

0.4

0.510A + E545KV777L DKI**

Per

sist

ence

50 200

5

10

15

20

2510A + E545KV777L DKI

*** *

50 200

10

20

30MCF7V777L

*** *

Fig. 6. Cells harboring both HER2 V777L and PIK3CA E54K mutations have increased migratory capacity in vitro and increased interaction between HER3 andp85. (A) Relative fraction of scratch wound closure (±SEM) as measured after 16 h. Cells were grown to near confluent monolayers, a wound was introduced,and assay media without EGF ± lapatinib (1 μM) was added back to wells (n ≥ 14, at least three independent experiments were performed, ***P ≤ 0.001, one-way ANOVA followed by Bonferroni multiple comparison test). (B) Representative phase contrast images of wound closure. (C) Relative fraction of scratchwound closure (±SEM) for MCF7 and HER2 mutant derivatives in serum-supplemented media ± lapatinib (5 μM) after 20 h (n ≥ 12, *P ≤ 0.05, at least twoindependent experiments were performed, one-way ANOVA followed by Bonferroni multiple comparison test). (D) Representative images from MCF7 cellline scratch wound closure experiments. (Scale bar, 500 μm in B and D.) Mean (±SEM) cell velocity (Left) and persistence (net cell displacement to total distancetraveled ratio) (Right) in microchannel migration experiments of MCF-10A + E545K and V777L double knockin cell lines (E) and MCF7 and MCF7 V777L celllines (F) (n ≥ 30 tracked cells, at least three independent experiments were performed, ***P ≤ 0.001, **P ≤ 0.01, *P ≤ 0.05, unpaired t test).

E6210 | www.pnas.org/cgi/doi/10.1073/pnas.1516853112 Zabransky et al.

Dow

nloa

ded

by g

uest

on

May

19,

202

1

HER2 Missense Mutations Do Not Lead to Increased Tumor Growth orInvasion in Vivo. The ability of HER2 mutations to increase tumorformation and xenograft growth was tested. MCF7 HER2mutantcell lines did not differ significantly in their growth comparedwith controls in the presence or absence of estrogen supple-mentation (Fig. 7 A and B). Additionally, the L755S and V777LDKI MCF-10A–derived cell lines did not form tumors in vivo(Fig. 7C), in contrast to our prior results with mutant KRAS andPIK3CA double knockin cell lines (14).We performed a tail vein injection assay to test the in vivo

invasiveness of MCF7 HER2 mutant cell lines. MCF7 has lowmetastatic capability and infrequently forms disease sites in thelungs after tail vein injection (27, 28). HER2 mutant cells did notform sites of disease in the lungs of nude mice as determinedby gross and microscopic inspection (Fig. 7D and SI Appendix,Fig. S7). These results suggest additional oncogenic alterationsare required to impart invasive phenotypes to HER2/PIK3CAmutant cells.

HER2 and PIK3CA Mutations Cooperate to Increase Interaction BetweenHER3 and p85. We examined potential interactions in the HER2signaling pathway that may be responsible for the cooperative ef-fects of HER2 and PIK3CA mutations. Previous work has shownthat overexpression of HER2 in cells with PIK3CA mutations en-hance HER2-mediated signaling through HER3 and p85 (29).Using the MCF7 cell line panel, we performed immunoprecipita-tion assays to examine the effect of HER2 and PIK3CA mutationson HER3/p85 dimerization. MCF7 HER2 L755S and V777L cellsshowed an increase in HER3–p85 interaction relative toHER2 andPIK3CA wild-type cells (Fig. 7E). Because HER3 levels are sig-nificantly lower in MCF-10A cells, this precluded our ability toperform immunoprecipitation assays in our MCF-10A cell linepanel. Therefore, we overexpressed mutant or wild-type HER2 andPIK3CA, along with wild-type HER3 in HEK293T cells to examinethe consequences of their interactions in another cellular back-ground. Similar to our results with MCF7 cells, the presence ofHER3–p85 dimers increased in cells transfected with both V777LHER2 and E545K PIK3CA compared with wild-type controls

7 14 21 28 35 420.00

0.05

0.10

0.15

0.20MCF7Targeted WTL755SV777L

Day

Tum

or V

olum

e (c

m3 )

A

E

Cell Line Tumors formed

Mice per group

MCF-10A + E545K 0 5

10A + L755S DKI 0 10

10A + V777L DKI 0 10

C

D

MCF7

MCF7 V777L

7 14 21 28 35 420.0

0.5

1.0

1.5MCF7Targeted WTG309AL755SV777L

Day

Tum

or V

olum

e (c

m3 )

MC

F7Targeted W

TL755S

V777L 1

V777L 2

MC

F7 Corrected

MC

F7 Corrected + V

777L 1M

CF7 C

orrected + V777L 2

p85

HER3

p85

GAPDH

IP:HER3

WB:

1%Input

MCF7

Targete

d WTL7

55SV77

7L

MCF7 Corr

ected

MCF7 Corr

ected

+ V77

7L0.0

0.1

0.2

0.3

0.4

IP p

85:H

ER

3 ra

tio

B

Fig. 7. HER2 mutations do not enhance tumor growth or invasion in vivo. (A) Mean (±SEM) in vivo tumor growth of MCF7 HER2 mutants in micesupplemented with estrogen pellets (n ≥ 5 animals per group, two independent experiments were performed). (B) Mean (±SEM) in vivo tumor growthfor MCF7 HER2 mutants in mice without estrogen pellet supplementation. n ≥ 5 animals per group. (C ) MCF-10A + E545K, L755S, and V777L DKI cells donot form tumors as xenografts in nude mice. (D) Representative images of H&E stained lung sections from tail vein injection assays showing benign lungparenchyma and lack of proliferating, multicellular sites of carcinoma. n ≥ 5 animals per group. (Scale bar, 100 μm.) (E ) MCF7 parental, MCF7-corrected,and HER2 mutant cell lysates were immunoprecipitated with a HER3 antibody. Antibody pulldowns were then subjected to Western blotting analysiswith the indicated primary antibodies. One percent of total protein was used as a control (Left). Bar graphs represent the ratio of p85 after immu-noprecipitation to HER3 input control Western blot band as measured with ImageJ software (Right). Bars for MCF7 V777L and MCF7 corrected + V777Lrepresent the average of two clones.

Zabransky et al. PNAS | Published online October 27, 2015 | E6211

MED

ICALSC

IENCE

SPN

ASPL

US

Dow

nloa

ded

by g

uest

on

May

19,

202

1

(SI Appendix, Fig. S7 B and C). This increase in HER3–p85 in-teraction in V777L mutant cells may represent a potential signalingpathway rewiring leading to increased MAP kinase activation viaHER3/PI3 kinase interaction as we have previously described(30). These results suggest that the V777L HER2 missense mu-tation in combination with PIK3CA mutations increase the for-mation of HER3–p85 dimers as a potential mechanism ofincreased MAP kinase pathway signaling.

HER2Mutations and Proliferative Response to HER2 Targeted Therapies.Given that overexpression models of HER2 mutations revealeddifferential sensitivities to HER2 targeted drugs, we measured theIC50 values of two HER2/EGFR tyrosine kinase inhibitors, lapa-tinib and neratinib, on our isogenic HER2 knockin cell line panel(SI Appendix, Tables S3 and S4 and Fig. S8) (4, 18).For MCF-10A cells, ∼650–1,000 nM lapatinib was necessary to

inhibit cell proliferation for cells regardless of HER2 and/orPIK3CA mutation status (SI Appendix, Table S3). Whereas MCF7cells were relatively insensitive to lapatinib,HER2mutations againdid not alter the response to lapatinib. Interestingly, cells har-boring a HER2 L755S mutation did not exhibit resistance tolapatinib in terms of proliferation or biochemical pathway acti-vation (SI Appendix, Fig. S8). Similar to lapatinib, HER2 muta-tions did not predict for increased sensitivity or resistance toneratinib in our models. However, neratinib was slightly morepotent in L755S DKI cell lines compared with controls (SI Ap-pendix, Fig. S8 and Table S4), though not so for MCF7 HER2L755S cells. Moreover, all cell lines were relatively insensitive totrastuzumab. Additionally, cell lines with both a PIK3CA and aHER2 mutation were not more sensitive to the PI3K p110α-spe-cific inhibitor BYL-719 than cells with a single PIK3CA mutation,though cells without a PIK3CA mutation were relatively resistantto BYL-719 (SI Appendix, Fig. S8 and Table S4).

DiscussionHER2 mutations in nonamplified/nonoverexpressed breast can-cers represent a phenomenon that can potentially be exploitedtherapeutically. Whereas previous overexpression studies ofmutant HER2 cDNAs have suggested a number of HER2 mu-tations are activating and promote transformation, our studyfound only the V777L mutation to be of functional significance.Additionally, the effects of the HER2 V777L mutation appear tobe accentuated by a PIK3CA mutation. Consistently across twodifferent human breast cell line models, HER2 V777L combinedwith PIK3CA E545K imparted features of increased signalingpathway activation and migration as noted by changes in West-ern blotting, scratch wound healing, and microchannel migrationassays. Notably, the HER2 V777L kinase domain mutation led toan increased interaction between p85 and HER3 in the presenceof a PIK3CA E545K mutation. These results suggest that HER2missense mutations require additional genetic alterations topromote features of transformation, leading to an increased in-teraction of known signaling partners. In addition, the HER2mutation L755S, which previous overexpression studies haveshown imparts resistance to lapatinib (4, 10, 18, 31) did not showobvious resistance phenotypes in our genome edited cell lines.Endogenous expression of L755S mutant HER2 in our modelsmay not be sufficient to produce resistance to lapatinib due tolower levels of the mutant protein. This finding may be of clinicalimportance, because patients that do not have amplification/overexpression or high expression of L755S mutant HER2 maystill be sensitive to lapatinib, an FDA approved therapy for HER2-amplified breast cancers.Trials are currently ongoing to evaluate the efficacy of tar-

geting HER2 mutations in HER2 negative cancers with FDAapproved therapies. However, given the rarity and variety ofHER2 mutations, such trials may be underpowered to de-finitively address this question. Because the majority of HER2

mutations in clinical samples are not amplified or overexpressed,our approach using somatic cell gene targeting has led to pre-viously unidentified insights into this issue. However, we recog-nize that limitations of our models, including the inability toanalyze immunologic effects of HER2-directed therapies, varyingprotein expression levels, and tumor microenvironment. None-theless, our isogenic cell line models provide useful tools forunderstanding the functional consequences of HER2 mutationsin isolation and in combination with other oncogenes, as well astesting responses to targeted therapies.Our study illustrates that different HER2 mutations impart

distinct phenotypes depending on the individual mutation andthe presence of other genetic alterations. This is consistent withrecent work demonstrating similar findings for AKT1 and HER3mutations (32, 33) and has implications for predicting responseto targeted therapies. For example, in our study, HER2 muta-tions did not overtly promote cell proliferation. Therefore, it maybe expected that HER2 targeted therapies would not show dif-ferential sensitivity to cells harboring these mutations in terms ofreducing cell number in standard growth assays. On the otherhand, the V777L HER2 mutation in the appropriate context ledto an increase in cellular migration, potentially resulting in in-creased metastatic potential that was affected by the HER2 in-hibitor lapatinib. HER2-directed therapies in this regard maypossibly afford clinical benefit. These results underscore theneed for examining multiple aspects of cancer phenotypes whenassessing response to targeted therapies.The variable effects of individual HER2 mutations coupled

with a requisite need for oncogene cooperativity may provechallenging for the use of HER2mutation status to guide therapyfor patients with breast cancer. Our studies have implicatedoncogenic mutant PIK3CA as a potential partner to HER2 mu-tations, and, indeed, HER2 kinase domain mutations and PIK3CAmutations have been reported concurrently in breast cancer se-quencing efforts (34). We tested the effects of PIK3CA mutationsin conjunction with HER2 mutations due to their close relation-ship as members of the same signaling pathway; however, it islikely that there are other genes, which when mutated, cooperatewith HER2 missense mutations to impart transformative effects.Indeed, the relatively low frequency of HER2 missense mutationsmay be explained by the need for concurrent oncogenic mutationsto achieve a selective advantage in HER2-mutant tumors. Addi-tionally, it may explain why many HER2 mutations in our isogenicmodels did not have a detectable phenotype, even though they arerecurrently found in human cancers. Further work in this regard isongoing and may help elucidate whether mechanisms of cooper-ativity are shared or unique among altered oncogenes and tumorsuppressors and HER2 mutations.In conclusion, using an isogenic panel of nonamplified/non-

overexpressed HER2-mutant cell lines, we have determined thatthe majority of HER2 mutations alone are not sufficient topromote transformative phenotypes in breast cell lines. How-ever, HER2 kinase domain mutations, notably V777L, can ex-hibit cooperativity with the activating PIK3CA E545K mutation.Cells with both HER2 V777L and PIK3CA E545K mutationsexhibited key features of transformation in vitro, including on-cogenic signaling pathway activation and increased migratorypotential, but not increased tumorigenicity in vivo. Although themajority of our genome-edited cell lines did not yield an overtphenotype, this has clinically impactful consequences given thatmost cancers with HER2 missense mutations do not have con-current overexpression/amplification of HER2. Indeed, given thecurrent landscape evaluating HER2 targeted therapies in can-cers harboring these mutations, the positive aspect of our find-ings and their importance to clinical oncology warrants emphasis.Specifically, these findings provide a previously unidentifiedcontext for the study and clinical significance of HER2 muta-tions, in that detailed analysis and study of coexisting mutations

E6212 | www.pnas.org/cgi/doi/10.1073/pnas.1516853112 Zabransky et al.

Dow

nloa

ded

by g

uest

on

May

19,

202

1

in patients may be required to make meaningful patient man-agement decisions regarding HER2-directed therapies. Takentogether, our results support that HER2 missense mutations maybe targetable in the appropriate genetic context.

Materials and MethodsCell Culture. MCF-10A, MCF7, and 293T cell lines were maintained as pre-viously described (14). Please see SI Materials and Methods for completedetails. Parental cell lines were authenticated via short tandem repeatprofiling analysis at the Johns Hopkins Genetic Resources Core Facility.

Gene Targeting and Generation of HER2 Missense Mutation Cell Lines. Genetargeting was carried out using recombinant AAV vectors as previously described(17). Vectors were generated with primers listed in SI Appendix, Table S2. See SIMaterials and Methods for details specific to HER2 gene targeting.

Overexpression of HER2, HER3, and PIK3CA cDNAs in Human Cells. HEK-293Tcells were transfected using the Fugene 6 system (Promega). pCFG5 plasmidscontaining wild-type, G309A, L755S, or V777L HER2 cDNAs, and LXSN plas-mids containing wild-type or E545K mutant PIK3CA were derived in previousstudies (4, 14). A wild-type HER3 expression vector was created by subclon-ing HER3 cDNA from the pCMV6-XL4-ERBB3 plasmid (SC118918; Origene)into a pIRES-neo3 backbone. HEK-293T cells were used in assays 48 h aftertransfection and overexpression was confirmed by Western blot.

Cell Proliferation Assays. Assays were performed as previously described (14).See SI Materials and Methods for more detail.

Drug Inhibitor Assays. Lapatinib, neratinib, and BYL-719 were obtained fromSelleck and were dissolved in dimethyl sulfoxide (DMSO). Trastuzumab wasobtained from the Johns Hopkins Research Pharmacy and was dissolved inbacteriostatic water. The 1–2 × 103 cells per well were plated into 96-well plateson day 0. MCF-10A derivatives were grown in assay media containing 20 ng/μLEGF, MCF-10A derivatives with PIK3CA E545K mutation were grown in assaymedia without EGF, whereas MCF7 and MCF7-corrected derivatives were grownin assay media containing 5% (vol/vol) FBS. See SI Materials and Methodsfor details.

Colony Formation Assay in Semisolid Medium. The 5 × 103 exponentiallygrowing cells were plated and experiments were performed as previouslydescribed (14). See SI Materials and Methods for details.

Acinar Morphogenesis Assay. Morphogenesis assays were conducted ingrowth factor reduced Matrigel (BD Biosciences) as previously described (35).For assays with inhibitors or DMSO vehicle controls (0.5%), compounds wereadded at seeding and media were refreshed every 4 d. Photographs weretaken under phase contrast microscopy (Nikon) after 12 d.

Immunoblotting. Cells were washed and seeded in the indicated media. After48 h, protein lysates were harvested, and immunoblotting was conducted aspreviously described (11). See SI Materials and Methods for details and a listof primary antibodies.

Immunoprecipitation. Cells were grown in serum-starved media conditionsand were then washed with ice-cold PBS, scraped, and lysed on ice in lysisbuffer. A total of 1 mg of protein extract was incubated with 1 μg of anti-HER3 antibody (Millipore 05–390) overnight at 4 °C, then with DynabeadsProtein G for immunoprecipitation (Life Technologies) for 4 h at 4 °C. See SIMaterials and Methods for details.

Scratch Wound Healing Assays. Cells were plated in six-well plates and grownto near confluent monolayers. Scratch wounds were introduced in a crosspattern with a 200-μL pipette tip. For assays with inhibitors, compounds wereadded after the scratch wound was introduced. Phase contrast images weretaken after media was replaced and wounds were followed over time. See SIMaterials and Methods for details.

Microchannel Migration Assays. Standard photolithography and replicamolding were used to create the polydimethylsiloxane (PDMS) microfluidicdevice as previously described (23–26). See SIMaterials andMethods for details.

Flow Cytometry and Extracellular Staining. The 1 × 106 cells were treated withHuman TruStain FcX Fc receptor blocking solution (BioLegend) before staining.Data were collected using a FACSCalibur II and analyzed using FACSCompsoftware (BD Biosciences). Gates and quadrants were set using isotype controlstaining. See SI Materials and Methods for a list of antibodies.

In Vivo Assays. The 8- to 10-wk-old female athymic nude mice (Harlan Lab-oratories) were used for in vivo assays. For xenografts, mice with or withoutestrogen pellet supplementation were injected s.c. in either flank with 200 μLmixture containing 2 × 106 cells in 20% PBS and 80% growth factor reducedMatrigel (BD Biosciences). Tumor volumes were calculated by multiplyinglength, width, and height for each individual tumor.

For tail vein injection assays, 1 × 105 cells in 200 μL of PBS were injected intothe tail vein of 8- to 10-wk-old female athymic nude mice. After 5 wk, animalswere euthanized and their lungs were examined grossly and were then fixed in10% (vol/vol) formalin, paraffin embedded, sectioned, and stained with hema-toxylin and eosin. Five stained sections were examined for evidence of pro-liferating disease per experimental group under a phase contrast microscope.

All animal experiments were performed in accordance with institutionaland The National Institutes of Health Guide for the Care and Use of Labo-ratory Animals guidelines (36).

Statistics. All statistical analyses were performed using GraphPad Prism 5 soft-ware (GraphPad Software). Significance levels are indicated using one or moreasterisks: *P ≤ 0.05, **P ≤ 0.01, and ***P ≤ 0.001. Error bars represent ± SEM.

ACKNOWLEDGMENTS. This work was supported by the Avon Foundation(B.H.P. and J.L.), NIH CA009071 (to H.A.P., K.C., and B.H.P.), GM007309 (toD.J.Z.), CA168180 (to R.L.C.), CA167939 (to S.C.), CA183804 (to K.K.-S.),Conquer Cancer Foundation Young Investigator Award 117451 (to H.A.P.),NIH Grant R01 CA080195 (to C.L.A.), Breast Cancer Specialized Program ofResearch Excellence (SPORE) P50 CA098131, and Vanderbilt-Ingram CancerCenter Support Grant P30 CA68485, and by the support of NIH P30 CA006973,the Sandy Garcia Charitable Foundation, the Commonwealth Foundation, theSanta Fe Foundation, the Breast Cancer Research Foundation, the HealthNetwork Foundation, the Marcie Ellen Foundation, the Helen Golde Trust, theAugustine Fellowship (to W.B.D.), and the Robin Page/Lebor Foundation.

1. Arteaga CL, Engelman JA (2014) ERBB receptors: From oncogene discovery to basicscience to mechanism-based cancer therapeutics. Cancer Cell 25(3):282–303.

2. Yarden Y, Sliwkowski MX (2001) Untangling the ErbB signalling network. Nat Rev MolCell Biol 2(2):127–137.

3. Ellis MJ, et al. (2012) Whole-genome analysis informs breast cancer response to ar-omatase inhibition. Nature 486(7403):353–360.

4. Bose R, et al. (2013) Activating HER2 mutations in HER2 gene amplification negativebreast cancer. Cancer Discov 3(2):224–237.

5. Shah SP, et al. (2012) The clonal and mutational evolution spectrum of primary triple-negative breast cancers. Nature 486(7403):395–399.

6. Lee JW, et al. (2006) Somatic mutations of ERBB2 kinase domain in gastric, colorectal,and breast carcinomas. Clin Cancer Res 12(1):57–61.

7. Greulich H, et al. (2012) Functional analysis of receptor tyrosine kinase mutations inlung cancer identifies oncogenic extracellular domain mutations of ERBB2. Proc NatlAcad Sci USA 109(36):14476–14481.

8. Rexer BN, et al. (2013) Human breast cancer cells harboring a gatekeeper T798Mmutation in HER2 overexpress EGFR ligands and are sensitive to dual inhibition ofEGFR and HER2. Clin Cancer Res 19(19):5390–5401.

9. 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(1):25–38.

10. Trowe T, et al. (2008) EXEL-7647 inhibits mutant forms of ErbB2 associated with la-patinib resistance and neoplastic transformation. Clin Cancer Res 14(8):2465–2475.

11. Konishi H, et al. (2007) Knock-in of mutant K-ras in nontumorigenic human epithelial cellsas a new model for studying K-ras mediated transformation. Cancer Res 67(18):8460–8467.

12. Lauring J, et al. (2010) Knock in of the AKT1 E17K mutation in human breast epithelialcells does not recapitulate oncogenic PIK3CA mutations. Oncogene 29(16):2337–2345.

13. Di Nicolantonio F, et al. (2008) Replacement of normal with mutant alleles in thegenome of normal human cells unveils mutation-specific drug responses. Proc NatlAcad Sci USA 105(52):20864–20869.

14. Wang GM, et al. (2013) Single copies of mutant KRAS and mutant PIK3CA cooperatein immortalized human epithelial cells to induce tumor formation. Cancer Res 73(11):3248–3261.

15. Subik K, et al. (2010) The expression patterns of ER, PR, HER2, CK5/6, EGFR, Ki-67 andAR by immunohistochemical analysis in breast cancer cell lines. Breast Cancer (Auckl)4:35–41.

16. Soule HD, et al. (1990) Isolation and characterization of a spontaneously immortal-ized human breast epithelial cell line, MCF-10. Cancer Res 50(18):6075–6086.

17. Gustin JP, et al. (2009) Knockin of mutant PIK3CA activates multiple oncogenicpathways. Proc Natl Acad Sci USA 106(8):2835–2840.

18. Kancha RK, et al. (2011) Differential sensitivity of ERBB2 kinase domain mutationstowards lapatinib. PLoS One 6(10):e26760.

Zabransky et al. PNAS | Published online October 27, 2015 | E6213

MED

ICALSC

IENCE

SPN

ASPL

US

Dow

nloa

ded

by g

uest

on

May

19,

202

1

19. Beaver JA, et al. (2013) PIK3CA and AKT1 mutations have distinct effects on sensitivity

to targeted pathway inhibitors in an isogenic luminal breast cancer model system.

Clin Cancer Res 19(19):5413–5422.20. Gabelli SB, et al. (2014) Activation of PI3Kα by physiological effectors and by onco-

genic mutations: Structural and dynamic effects. Biophys Rev 6(1):89–95.21. Masuda H, et al. (2012) Role of epidermal growth factor receptor in breast cancer.

Breast Cancer Res Treat 136(2):331–345.22. Hart JR, et al. (2015) The butterfly effect in cancer: A single base mutation can re-

model the cell. Proc Natl Acad Sci USA 112(4):1131–1136.23. Balzer EM, et al. (2012) Physical confinement alters tumor cell adhesion and migration

phenotypes. FASEB J 26(10):4045–4056.24. Tong Z, et al. (2012) Chemotaxis of cell populations through confined spaces at single-

cell resolution. PLoS One 7(1):e29211.25. Hung W-C, et al. (2013) Distinct signaling mechanisms regulate migration in un-

confined versus confined spaces. J Cell Biol 202(5):807–824.26. Stroka KM, Konstantopoulos K (2014) Physical biology in cancer. 4. Physical cues

guide tumor cell adhesion and migration. Am J Physiol Cell Physiol 306(2):

C98–C109.27. Li N, et al. (2011) Siva1 suppresses epithelial-mesenchymal transition and metastasis

of tumor cells by inhibiting stathmin and stabilizing microtubules. Proc Natl Acad Sci

USA 108(31):12851–12856.

28. Barkan D, et al. (2008) Inhibition of metastatic outgrowth from single dormant tumorcells by targeting the cytoskeleton. Cancer Res 68(15):6241–6250.

29. Chakrabarty A, et al. (2010) H1047R phosphatidylinositol 3-kinase mutant enhancesHER2-mediated transformation by heregulin production and activation of HER3.Oncogene 29(37):5193–5203.

30. Blair BG, et al. (2015) A phosphoproteomic screen demonstrates differential de-pendence on HER3 for MAP kinase pathway activation by distinct PIK3CA mutations.Proteomics 15(2-3):318–326.

31. Yang B, Zhang H, Wang H (2015) Atomistic insights into the lung cancer-associated L755Pmutation in HER2 resistance to lapatinib: A molecular dynamics study. J Mol Model 21(2):24.

32. Jaiswal BS, et al. (2013) Oncogenic ERBB3 mutations in human cancers. Cancer Cell23(5):603–617.

33. Yi KH, Axtmayer J, Gustin JP, Rajpurohit A, Lauring J (2013) Functional analysis ofnon-hotspot AKT1 mutants found in human breast cancers identifies novel drivermutations: Implications for personalized medicine. Oncotarget 4(1):29–34.

34. Network CGA; Cancer Genome Atlas Network (2012) Comprehensive molecular por-traits of human breast tumours. Nature 490(7418):61–70.

35. Debnath J, Muthuswamy SK, Brugge JS (2003) Morphogenesis and oncogenesis ofMCF-10A mammary epithelial acini grown in three-dimensional basement membranecultures. Methods 30(3):256–268.

36. Committee on Care and Use of Laboratory Animals (1996) Guide for the Care and Useof Laboratory Animals (Natl Inst Health, Bethesda), DHHS Publ No (NIH) 85-23.

E6214 | www.pnas.org/cgi/doi/10.1073/pnas.1516853112 Zabransky et al.

Dow

nloa

ded

by g

uest

on

May

19,

202

1