sensitizing triple-negative breast cancer to pi3k inhibition …...small molecule therapeutics...

Small Molecule Therapeutics

Sensitizing Triple-Negative Breast Cancer toPI3K Inhibition by Cotargeting IGF1RKlaas de Lint1,2, Jos B. Poell1,2, Hayssam Soueidan1, Katarzyna Jastrzebski1,Jordi Vidal Rodriguez, 1,2, Cor Lieftink1,3, Lodewyk F.A.Wessels1, andRoderick L. Beijersbergen1,2,3

Abstract

Targeted therapies have proven invaluable in the treatmentof breast cancer, as exemplified by tamoxifen treatment forhormone receptor–positive tumors and trastuzumab treatmentfor HER2-positive tumors. In contrast, a subset of breast cancernegative for these markers, triple-negative breast cancer(TNBC), has met limited success with pathway-targeted ther-apies. A large fraction of TNBCs depend on the PI3K pathwayfor proliferation and survival, but inhibition of PI3K alonegenerally has limited clinical benefit. We performed an RNAi-based genetic screen in a human TNBC cell line to identifykinases whose knockdown synergizes with the PI3K inhibitorGDC-0941 (pictilisib). We discovered that knockdown of insu-lin-like growth factor-1 receptor (IGF1R) expression potently

increased sensitivity of these cells to GDC-0941. Pharmacologicinhibition of IGF1R using OSI-906 (linsitinib) showed a strongsynergy with PI3K inhibition. Furthermore, we found that thecombination of GDC-0941 and OSI-906 is synergistic in 8 linesfrom a panel of 18 TNBC cell lines. In these cell lines, inhibitionof IGF1R further decreases the activity of downstream PI3Kpathway components when PI3K is inhibited. Expression anal-ysis of the panel of TNBC cell lines indicates that the expressionlevels of IGF2BP3 can be used as a potential predictor forsensitivity to the PI3K/IGF1R inhibitor combination. Our datashow that combination therapy consisting of PI3K and IGF1Rinhibitors could be beneficial in a subset of TNBCs. Mol CancerTher; 15(7); 1545–56. �2016 AACR.

IntroductionBreast cancer is currently the most prevalent invasive cancer

in women (1). Although most breast cancers have targetedtreatment options due to hormone receptor or HER2 expres-sion, the subgroup of breast cancer which does not expressthese markers, triple-negative breast cancer (TNBC), compris-ing 15% to 20% of all breast cancers, currently lacks effectivetargeted treatment options (2). Compared with the other sub-types, TNBC has a worse prognosis (3) and is more likely todevelop chemotherapy-resistant disease (4).

The heterogeneity observed in this subtype of breast cancer (5)poses a challenge for the development of targeted treatmentstrategies. The most commonly mutated oncogenic driver inTNBC is PIK3CA, the gene encoding the catalytic alpha subunitof PI3K, with gain-of-function mutations in only about 12% ofcases (6, 7). However, analysis of The Cancer Genome Atlas(TCGA) breast tumor data shows amplification of the PIK3CAgene in 60% of these tumors (7).

In addition to PIK3CA deregulation, mutation or deletion ofother genes in the PI3K signaling axis, such as AKT and PTEN,also occurs (7). This renders the PI3K pathway an attractivetarget for pharmacologic intervention. Therefore, several clin-ical trials have been initiated using specific inhibitors for PI3Kpathway components, mainly mTOR (8). PI3K inhibitors arecurrently in phase II clinical trials in patients with TNBC(ClinicalTrial: NCT01629615), often combined with chemo-therapy (ClinicalTrials: NCT01918306 and NCT01740336).

In preclinical models, PI3K inhibition leads to an increase inoverall survival, but most tumors still progress after prolongedtreatment (9, 10). Results from clinical trials show limitedefficacy of PI3K-inhibitor monotherapy (11) and emergenceof resistance (12). Given the fact that few tumors regress upontreatment, PI3K inhibitors merely decrease proliferation ratesor induce a cell-cycle arrest rather than causing apoptosis andcell death. As a consequence, cells are potentially able toovercome the effects of the PI3K inhibition by an adaptiveresponse. This phenomenon is already apparent in vitro,because apoptosis upon PI3K inhibition is low or absent andcells can adapt to relatively high concentrations of inhibitor(13). Therefore, we sought to identify genes that are requiredfor sustained proliferation and survival under PI3K inhibitionin TNBC cells. The identification of key genes underlyingresistance and adaptive response may lead to improved strat-egies for targeted treatment of TNBC.

Materials and MethodsCell lines, cell cultures, and reagents

TNBC cell lines except SK-BR-7 were acquired in 2011 fromeither the ATCC or DSMZ and authenticated by short tandem

1Division of Molecular Carcinogenesis, The Netherlands Cancer Insti-tute, Amsterdam, the Netherlands. 2Cancer Systems Biology Center(CSBC), The Netherlands Cancer Institute, Amsterdam, the Nether-lands. 3The NKI Robotics and Screening Center (NRSC), The Nether-lands Cancer Institute, Amsterdam, the Netherlands.

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

K. de Lint and J.B. Poell contributed equally to this article.

Corresponding Author: Roderick L. Beijersbergen, The Netherlands CancerInstitute, Plesmanlaan 121, 1066CX Amsterdam, the Netherlands. Phone:31205121960; Fax: 31205121954; E-mail: [email protected]

doi: 10.1158/1535-7163.MCT-15-0865

�2016 American Association for Cancer Research.

MolecularCancerTherapeutics

www.aacrjournals.org 1545

on August 13, 2021. © 2016 American Association for Cancer Research. mct.aacrjournals.org Downloaded from

Published OnlineFirst May 11, 2016; DOI: 10.1158/1535-7163.MCT-15-0865

http://mct.aacrjournals.org/

repeat (STR) profiling by the vendors. Cells were expanded andstored in liquid N2, and passaged for a maximum of 3 monthsafter which a new cell aliquot was obtained from the frozen stock.Cell lines were propagated in the media recommended by thesuppliers at 37�C, 5% CO2. SK-BR-7 cells were acquired from theinternal NKI cell bank in 2014, authenticated by STR profiling,and expanded and stored in liquid N2. SK-BR-7 cells were pas-saged for a maximum of 3 months after which a new cell aliquotwas obtained from the frozen stock. SK-BR-7 cells were propa-gated in DMEM/F-12 þ 10% FBS and 1% pen/strep (Life Tech-nologies). RPMI, DMEM, MEM, and DMEM-F12 were purchasedfrom Life Technologies, and FBS was purchased from ThermoScientific (Hyclone; lot# RWB35889). Supplements used wereGlutamax from Life Technologies; insulin, glutathione, andinsulin-transferrin-sodium selenite from Sigma-Aldrich; andhuman EGF from BD Biosciences.

RNAi screenHEK293T cells (internal NKI cell bank) were transfected with

a subset of the TRC library consisting of 4,774 constructstargeting the human kinome together with the helper plasmidspRSV-Rev, pHCMV-VSV-G, and pMDLg/pRRE using 4 mg poly-ethylenimine (CAS Number 9002-98-6; Sigma-Aldrich) per mgof plasmid DNA. Thirty-six hours after transfection, the mediumcontaining lentiviral particles was harvested and cleared througha 0.45-mm filter and frozen in aliquots in liquid N2.

HCC1806 cells were infected with a 1:60 dilution of the viralsupernatant (corresponding to a multiplicity of infection of 0.2)in medium containing hexadimethrine bromide (8 mg/mL; Sig-ma-Aldrich). Twenty-four hours after transduction, puromycin (2mg/mL; Sigma-Aldrich) was added to the transduced cells. Afterselection, a t ¼ 0, sample was taken and the remaining cells wereused to create three control populations as well as three popula-tions which were treated with 1 mmol/L of the PI3K inhibitorGDC-0941. All populations consisted of 5 � 106 cells at the timeof seeding and were grown for an additional eight populationdoublings.

Genomic DNA was isolated using the DNeasy Blood & TissueKit (Qiagen), and the shRNA cassettes were amplified using thePhusion High-Fidelity DNA Polymerase Kit (Thermo Scientific)from 32 mg of genomic DNA per population by performing atwo-step PCR amplification using the following conditions:(1) 98�C, 30 seconds; (2) 98�C, 10 seconds; 3) 60�C, 20seconds; (4) 72�C, 1 minute; (5) to step 2, 16 cycles; (6) 72�C,5 minutes; (7) 4�C. Indices (2 per population as an internalcontrol) and adaptors for deep sequencing (Illumina) wereincorporated into PCR primers.

From the first PCR, 5 mL was used as a template for the secondPCR reaction (same PCR conditions as first PCR, but 18 cycles).The final PCR products were purified using a Qiagen PCR puri-fication kit and run on a 1.5% agarose gel to confirm that eachsample yielded a comparable amount of product. Samples werepooled, quantified by bioanalyzer, and sequenced in one lane ofan Illumina Hiseq2000. Each read was aligned to the TRC library;reads that did not match the library were discarded. All primersused are listed in the Supplementary Table S1.

Cell viability and drug sensitivity assaysCells of each TNBC cell line were seeded in 384-well plates in

50 mL of medium at numbers previously determined to yieldapproximately 75% confluence in untreated wells after 96 hours

of proliferation (BT-20: 1250, BT-549: 500, CAL-120: 500, CAL-148: 1500, CAL-51: 800, HCC1187: 5000, HCC1395: 1500,HCC1806: 800, HCC1937: 2000, HCC38: 800, HCC70: 5000,HS578T: 800, MDA-MB-157: 1250, MDA-MB-231: 2500, MDA-MB-436: 2000, MDA-MB-468: 2500, MFM-223: 2000, SK-BR-7:1250). GDC-0941, OSI-906, and PD-0325901 (Selleck Chemi-cals), dissolved to 10 mmol/L in DMSO, were added 24 hoursafter seeding using a Tecan HP D300 Digital Dispenser in arandomized pattern not using edge wells. GDC-0941 andPD-0325901 were used at concentrations ranging from32 nmol/L to 10 mmol/L, whereas OSI-906 was added at 0.25,0.5, 1, and 2 mmol/L. IGF2 was purchased from Sigma-Aldrich(I2526), and anti-IGF2 antibodies were purchased from Abcam(#ab9574). Four replicate wells were assayed for each condition.Phenylarsine oxide (PAO; Sigma-Aldrich) was used at 10 mmol/Las a zero cell viability control. Seventy-two hours after additionof the compounds to the plate, cell viability was assayed byadding 10 mL of a 1:4 dilution of CellTiter-Blue (Promega) inmedium. After 3 hours of incubation at 37�C, plate fluorescence(560Ex/590Em) was measured on a PerkinElmer Envision platereader. Relative viability for each well was calculated from thefluorescence readings as: (value-(value(PAO))/((value(untreat-ed)-(value(PAO)). Graphpad Prism 6 was utilized to fit a four-parameter dose–response curve to the viability data. The GI50swere defined as the drug concentration at which the fitted curvescorrespond to 0.5 of the untreated viability.

Quantification of synergySensitization to GDC-0941 by increasing concentrations of

OSI-906 was quantified by calculating the change in area underthe curve (DAUC) at each concentration of OSI-906. The Loeweadditivity model was used to distinguish between additivity andsynergy in cell line HCC1187.

Overexpression of IGF1RCell lines HCC1806, CAL-51, CAL-148, and MFM-223 over-

expressing IGF1Rwere generated by lentiviral integration of cloneID#101931727 from the CCSB-Broad Lentiviral ExpressionLibrary supplied by Thermo Scientific. The plasmid wassequenced from the 50 and 30 ends to confirm its identity.

Dynamic measurements of confluence and apoptosisConfluence and apoptosis levels were quantified by IncuCyte

FLR and IncuCyte Zoom live-cell imaging systems from EssenBioscience (Ann Arbor). Confluence was determined by IncuCyteFLR software 2011A Rev2 and IncuCyte Zoom software 2013BRev1 using phase-contrast images, and doubling times were cal-culated using Graphpad Prism 6 software. Apoptosis was quanti-fied by addition of CellPlayer NucView Caspase-3/7 substrate(Essen Bioscience) to a final concentration of 5 mmol/L andmeasuring green fluorescent object confluence. Relative apoptosisfor each cell line was calculated by dividing the confluence offluorescent apoptotic cells by total confluence andnormalizing thisvalue to the amount of apoptosis induced by 1 mmol/L stauros-porin (Sigma-Aldrich). Wells were imaged at 4-hour intervals.

Western blottingWhole-cell lysateswere preparedbywashing cellswith cold PBS

and lysing with RIPA buffer (1% NP40, 0.5% DOC, 0.1% SDS,150 mmol/L NaCl, and 50 mmol/L Tris, pH 8.0) containingprotease inhibitors (Complete; Roche) and a phosphatase

de Lint et al.

Mol Cancer Ther; 15(7) July 2016 Molecular Cancer Therapeutics1546




inhibitor cocktail (PhoStop; Roche). Samples were spun down at14,000 rpmat4�C for 15minutes, supernatantwasmixedwith4�NuPage Sample buffer and 10� Reducing agent, heated to 70�Cfor 10minutes, and separated on 4% to 12%Bis-Tris NuPage gels.Proteins were transferred to PVDF membrane (Millipore), afterwhich the membrane was blocked for 1 hour with PBS/0.05%Tween-20 supplemented with 4% (w/v) BSA. Membranes wereincubated with primary antibodies (Cell Signaling) at a 1:1,000dilution overnight at 4�C, then with secondary antibodies con-jugated to horseradish peroxidase (Bio-Rad) at a 1:10,000 dilu-tion for 1 hour at room temperature. Membranes were washed inPBS/0.05% Tween-20. Signal was detected by adding ECL (Bio-Rad) followed by exposure to film (AmershamHyperfilmMP; GEHealthcare) or imaging on the Chemidoc Touch (Bio-Rad).

RNA isolation, cDNA synthesis, and qPCRRNA was isolated using the Quick-RNA MiniPrep Kit (Zymo

Research) and reverse transcribed to cDNAusing theMaxima FirstStrand cDNA Synthesis Kit (Thermo Scientific). qRT-PCRs wererun on a 7500 Fast Real-Time PCR System (Applied Biosystems)using FastStart Universal SYBRGreenMaster mix (Roche, AppliedScience).mRNA levels were normalized to the housekeeping geneGAPDH. qPCR primer sequences are listed in SupplementaryTable S1.

Gene expression analysis by RNA sequencingTotal RNA was collected from 70% confluent cycling popula-

tions. Samples were prepared for sequencing using the IlluminaTruSeq RNA Sample Preparation Kit according to the manufac-turer's protocol. Ten samples were sequenced per lane on anIllumina HiSeq 2000, and reads were aligned to hg19. Readcounts for the IGF1R- and PI3K-related genes were extracted andused for gene set analysis. The read counts are provided inSupplementary Table S2.

Gene set analysisThe list of IGF1R- and PI3K-related genes was composed

based on literature and by selecting all genes with an interac-tion with either PIK3CA, IGF1R, or IGF2 in the String 9.1database of known and predicted protein–protein interactionsbased on default settings (14). Correlation between geneexpression and DAUC of the cell line panel excluding HCC1187was calculated for each gene from RNA-Seq data. We repeatedlyselected random gene sets of size 54, in each iteration rankedthese genes based on the R2 value, and then created the distri-bution of the R2 values for each rank position from which themedian and 95% confidence intervals (CI) shown in Fig. 6Awere derived. Q values were calculated using the Benjamini–Hochberg method on P values derived from the observedcorrelations with 15 degrees of freedom.

ResultsKnockdown of IGF1R sensitizes HCC1806 to PI3K inhibitionby GDC-0941

To select the most appropriate cell line model to enhancesensitivity to PI3K inhibitors, we first tested a panel of 18 TNBCcell lines for their sensitivity to PI3K inhibition with GDC-0941.Of this panel, 14 cell lines are sensitive to moderately sensitive toGDC-0941 with a GI50 below 10 mmol/L (Fig. 1A). This indicatesthat a majority of these lines at least partly depend on PI3K

signaling for their proliferation. We observed that PIK3CAmuta-tion, but not PTEN loss, correlated with sensitivity to PI3Kinhibition, as demonstrated before for breast cancer regardlessof hormone receptor or HER2 status (15).

To screen for enhancers of response to PI3K inhibitors, weselected the PIK3CA-amplified cell line HCC1806, a cell line withmoderate sensitivity to GDC-0941. We hypothesized that cellularmechanisms attenuating the effect of PI3K inhibition eitherdirectly or indirectly involve kinases. To identify these kinases,we performed a pooled shRNA sensitizer screen using 4,774constructs from the TRC library targeting the human kinome.HCC1806 cells were transduced using the TRC Kinome libraryand cultured in the presence or absence of 1 mmol/L GDC-0941(Fig. 1B). After eight population doublings, relative abundance ofeach individual shRNA vector was determined by Illumina deepsequencing. DeSeq analysis was performed to identify hairpinssignificantly depleted in the treated population. The followingcriteria were used for hit-calling: a minimum 2-fold depletion inthe exposed population, at least 100 reads in the untreatedpopulation, and an FDR of 10%. We identified IGF1R as the tophit with nine independent hairpins targeting this gene fulfillingthe selection criteria (Fig. 1C; see Supplementary Table S3 for allthe screen data). To validate that knockdown of IGF1R sensitizesHCC1806 to PI3K inhibition, we generated two stable IGF1Rknockdown lines with different shRNA constructs (Fig. 1D). Wethen calculated the population doublings per day at differentconcentrations of GDC-0941 for both of these lines comparedwith an empty vector–transduced cell line (Fig. 1E). Both IGF1Rknockdown lines proliferated at a rate comparable with the wild-type line in the absence of inhibitor, but their doubling timesweremore strongly affected by inhibition of PI3K. This decrease inproliferation was also observed in a long-term colony outgrowthassay at 1 mmol/L GDC-0941 (Fig. 1F).

Pharmacologic inhibition of IGF1R synergizes with PI3Kinhibition in HCC1806 by impeding downstream signaling

IGF signaling has been recognized as an event specificallyassociated with malignancy (16), and both IGF ligands andIGF1R have been the subject of drug development efforts (17).We tested whether pharmacologic inhibition of IGF1R by OSI-906 could recapitulate the effect of IGF1R knockdown onsensitivity to GDC-0941 in HCC1806 cells. Indeed, althoughtreatment with OSI-906 itself had no discernable effect oncell proliferation, it sensitized HCC1806 cells to PI3K inhibi-tion by GDC-0941, causing a 3-fold reduction of the GI50 from4.2 mmol/L to 1.4 mmol/L at 2 mmol/L OSI-906 (Fig. 2A).Analogous results were obtained using the IGF1R inhibitorNVP-AEW541 (Supplementary Fig. S1). The sensitization toPI3K inhibition by OSI-906 resulted in a pronounced reductionin colony outgrowth after prolonged culture (Fig. 2B).

To further dissect the interaction between PI3K and IGF1Rinhibition, we examined the effect of GDC-0941 and OSI-906on PI3K signaling inHCC1806 cells (Fig. 2C). Treating HCC1806cells with GDC-0941 led to a rapid and strong decrease inphospho-AKT, which was partially recovered 24 hours after treat-ment. This correlated well with AKT activity as measured byphosphorylation of its substrate PRAS40. Exposure to OSI-906decreased AKT phosphorylation to a lesser extent, with onlyminor effects on phospho-PRAS40. However, in the combinationtreatment, phospho-AKT and phospho-PRAS40 were dramatical-ly decreased. These data suggest that IGF1R signals toPI3K in these

Sensitizing TNBC to PI3K Inhibition by Cotargeting IGF1R

www.aacrjournals.org Mol Cancer Ther; 15(7) July 2016 1547




kinomeLentivirallibraryshRNA

Pooled infection

HCC1806

selectionPuromycin

1 µmol/L GDC-0941No drug

No drug

8 population doublings

sequencesIdentify depleted shRNAby deep sequencing

Log 2

fold

cha

nge

Mean reads of untreated samples

shIGF1R

pLK

O

#2#1

IGF1R

HSP901 10 100 1,000 10,000 100,000

Pop

ulat

ion

doub

lings

pe

r day

GDC-0941 (log10 µmol/L)

GI 5

0 G

DC

-094

1 (μ

mol

/L)

No drug

GDC-0941(1 µmol/L)

pLKO

shIGF1R

#2#1

BA

C

D

FE

Figure 1.

Knockdown of IGF1R sensitizes HCC1806 to PI3K inhibition. A, sensitivity of 18 TNBC cell lines to GDC-0941. The GI50 of each cell line was determined in a72-hour drug exposure assay. B, schematic outline of the shRNA dropout screen for enhancers of GDC-0941 sensitivity. Human TRC kinome shRNA librarypolyclonal virus was produced and used to infect HCC1806 cells. After puromycin selection, cells were left untreated (control) or treated with 1 mmol/LGDC-0941 for eight population doublings. Next, shRNA inserts from both populations were recovered by PCR and quantified by deep sequencing. C,representation of the relative abundance of the shRNA barcode sequences from the screen depicted in B. The y-axis represents enrichment on a log2scale (relative abundance of GDC-0941 treated/untreated), whereas the x-axis represents intensity (average sequence reads in the untreated populations)of each shRNA. Using cutoffs of a minimum 2-fold depletion in the exposed population and an FDR of 10%, nine independent shRNAs targeting IGF1R(indicated in red) were identified. D, the levels of knockdown of IGF1R in the cell lines used in E and F were determined by assaying IGF1R proteinlevels by Western blotting. E, the effect of GDC-0941 on cell proliferation analyzed by cell doublings per day based on well confluence as measured byIncuCyte. Doubling times were determined for HCC1806 transduced with empty pLKO and with two independent shRNAs directed against IGF1R. F,cell lines transduced with the indicated vectors were seeded in 6-well plates at a density of 10,000 cells/well and grown in the absence or presence of1 mmol/L GDC-0941. Cells were fixed, stained, and imaged after 10 days.

de Lint et al.





cells, but its contribution to signaling only becomes rate limitingwhen PI3K is inhibited.

The IGF1R inhibitor OSI-906 potentiates PI3K inhibition in asubset of TNBC cell lines

We extended our synergy experiments to a panel of 18 TNBCcell lines. We found that IGF1R inhibition by itself has noappreciable effect on viability in all lines except HCC1187, in

which OSI-906 shows a modest effect on proliferation (Fig. 3A).Because OSI-906 has no or little effect by itself in all cell linesexcept HCC1187, and hence no four-parameter curve can begenerated for this compound, the Loewe additivity model couldnot be used to quantify synergy in amajority of our cell lines (18).Instead, we quantified the sensitization to PI3K inhibition byOSI-906 by calculating the differential area under the GDC-0941dose–response curve for each cell line examined (DAUC; Fig. 3B;Supplementary Fig. S2 for all dose–response curves). IGF1Rinhibition decreased sensitivity to PI3K inhibition in one cellline (HCC70), had no or little sensitizing effect in nine cell lines,and augmented the effects of GDC-0941 in eight cell lines(Fig. 3C). These data suggest that a substantial subset of TNBCcould benefit from combined PI3K-IGF1R inhibition. Interest-ingly, the levels of IGF1R and phospho-IGF1R do not predictsensitivity to the drug combination in our cell line panel (Sup-plementary Fig. S3).

In combination-sensitive cell lines, cotargeting IGF1Rdecreases residual PI3K signaling in the presence ofGDC-0941 and increases levels of apoptosis

To investigate whether the synergy between PI3K and IGF1Rinhibition is governed by common mechanisms in these celllines, we measured signaling response in two combination-sensitive lines, HCC1806 and CAL-51, as well as two combi-nation-insensitive lines, CAL-148 and MFM-223, after 24-hourdrug exposure (Fig. 4A). In CAL-51, like in HCC1806, phospho-AKT was more effectively reduced by the drug combination.This is also reflected in the phosphorylation of its down-stream target PRAS40. In CAL-148 and MFM-223, the additionof OSI-906 had no effect on signaling downstream of PI3K.

It has been observed that IGF1R can also activate signalingthrough the mitogen-activated protein kinase (MAPK) pathway(16). However, in our cell lines, no significant effects of GDC-0941, OSI-906, or the combination of both drugs on phos-phorylation levels of ERK were detected (Fig. 4A). This indicatesthat in the combination-sensitive cell lines, the decrease inviability is not explained by an inhibitory effect on MAPKsignaling. The absence of a role for MPK signaling in theproliferation and survival of our panel of TNBC cell lines isfurther illustrated by the lack of response to an inhibitor ofMitogen/Extracellular signal-regulated Kinase (MEK), PD-032590. None of the cell lines, with the exception of MDA-MB-231 with a GI50 of 1.6 mmol/L, displayed sensitivity toconcentrations below 10 mmol/L (Supplementary Fig. S4). Thisimplies that both combination-sensitive and -insensitive TNBCcell lines in the panel do not directly depend on the MAPKpathway for their proliferation.

By activating the prosurvival protein AKT, PI3K plays anantiapoptotic role in many cell systems (19). In addition toachieving a stronger proliferative arrest, the drug combinationmay therefore also increase apoptosis. To quantitatively mea-sure apoptosis over time, we monitored cells cultured in thepresence of a Caspase-3/7 activity reporter (Fig. 4B). Peak apop-totic activity occurred at 24 hours after drug administration(Supplementary Fig. S5). OSI-906 by itself did not leadto significantly increased apoptosis in any of the cell lines.However, OSI-906 approximately doubled the induction ofapoptosis caused by GDC-0941 in CAL-51 and HCC1806,whereas it did not lead to increased apoptosis in CAL-148and MFM-223 (Fig. 4C). The results from the real-time

A

B

C

Rel

ativ

e vi

abili

ty


0 µmol/L OSI-906

1 μmol/L GDC-09412 μmol/L OSI-906

µmol/L OSI-9062

1052.51.250.6250

No drug –1.50

0.25

0.5

0.75

1

–1 –0.5 0 0.5

0 μmol/L OSI-9060.25 μmol/L OSI-9060.5 μmol/L OSI-9061 μmol/L OSI-9062 μmol/L OSI-906

1


++

IGF1R

pIGF1R

pAKTS473

pAKTT308

pPRAS40

HSP90

––

––++

++–

––

–++

++–

––

–++

++–

––

–++

72 h48 h24 h4 h

Figure 2.

Inhibition of IGF1R sensitizes HCC1806 to PI3K inhibition by abrogating signalingthrough PI3K. A, dose–response curves of HCC1806 for GDC-0941 at theindicated concentrations of OSI-906 as determined in a 72-hour cell viabilityassay. B, colony formation assay of HCC1806. Cellswere grown in the absence orpresence of 2 mmol/L OSI-906 together with GDC-0941 at the indicatedconcentrations. Cells were grown for 9 days and then fixed and stained. C,biochemical responses of HCC1806 cells mock treated (DMSO) or exposedto 1 mmol/L GDC-0941, 2 mmol/L OSI-906, or a combination of both drugs.Cells were harvested at the indicated time points after drug treatment, and(phospho)protein levels were assayed by Western blotting. Heat shockprotein 90 (HSP90) served as a loading control.






apoptosis assay correspond well with the levels of cleaved PARPobserved 24 hours after drug exposure (Fig. 4A).

The contribution of IGF signaling to PI3K activity mayincrease over time through feedback mechanisms such asupregulation of the expression of tyrosine kinase receptors. Touncouple the direct effect of IGF1R inhibition from effectscaused by transcriptional feedback mechanisms, we deter-mined phospho-AKT and phospho-PRAS40 levels 1 hour afterdrug exposure. The added effect of IGF1R inhibition on PI3Kinhibition was clearly evident in downstream signaling ofthe combination-sensitive lines HCC1806 and CAL-51. Theaddition of OSI-906 decreased the minimal concentration ofGDC-0941 that caused a clear reduction in phospho-AKT andphospho-PRAS40 (Fig. 4D). Together, our results suggest thata subset of TNBC cell lines is sensitized to PI3K inhibitionby limiting upstream signaling input from IGF1R, the effectsof which are evident on downstream activity immediately upondrug treatment.

Levels of the ligand IGF2 modulate the sensitivity to PI3Kinhibition, while overexpression of IGF1R does not decreasesensitivity to GDC-0941

Feedback regulation of receptor tyrosine kinases (RTK) upondrug treatment has been amply described and is frequentlyimplicated in resistance mechanisms to pathway-targeted ther-apies (20–22). We noticed in the response of HCC1806 cells toGDC-0941 that total levels of IGF1R protein steadily increaseupon prolonged exposure to the drug. In addition, long-termexposure of HCC1806 to high concentrations of GDC-0941led to adaptive resistance (Supplementary Fig. S6A). Moreover,the resistant population was still sensitive to the combinationof GDC-0941 and OSI-906 (Supplementary Fig. S6B). Wetherefore hypothesized that overexpression of IGF1R mayinstill resistance to PI3K inhibition in TNBC cells. To test thispossibility, we created IGF1R overexpression lines for thecombination-sensitive lines, HCC1806 and CAL-51, as well asthe combination-insensitive lines CAL-148 and MFM-223.

0

0.2

0.4

0.6

0.8

1

1.2

Rel

ativ

e vi

abili

ty a

t 2 μ

mol

/L O

SI-9

06

TNBC panel (n = 18)

–0.25

–0.2

–0.15

–0.1

–0.05

0

0.05

0.1

0.15

CA

L-51

HC

C18

06

HC

C11

87

MD

A-M

B-4

36

BT5

49

CA

L-12

0

MD

A-M

B-4

68

SK

-BR

-7

HC

C38

MD

A-M

B-2

31

BT2

0

MD

A-M

B-1

57

MFM

-223

HC

C13

95

Hs5

78T

HC

C19

37

CA

L-14

8

HC

C70ΔA

UC

Strongsynergy

Intermediatesynergy Weak/no synergy or antagonism

HCC1187

0

0.25

0.5

0.75

1

No drug –1.5 –1 –0.5 0 0.5 1

ΔAUC

GDC-0941 (log μmol/L)

Rel

ativ

e vi

abili

ty

A B

C

0.25 μmol/L OSI-906 0.5 μmol/L OSI-906 1 μmol/L OSI-906 2 μmol/L OSI-906

Figure 3.

A subset of TNBC cell lines is sensitized toPI3K inhibition by cotargeting IGF1R. A,relative viability of the TNBC panel after 72-hour exposure to 2 mmol/L OSI-906. B,sensitization to GDC-0941 was quantified bycalculating the change in area underthe dose–response curve (DAUC). C, theDAUC of the GDC-0941 dose–response curveas determined for the 18 TNBC cell lines fordifferent concentrations of OSI-906 rangingfrom 0.25 mmol/L to 2 mmol/L.

de Lint et al.





Surprisingly, the overexpression of IGF1R did not decrease thesensitivity to GDC-0941 in any of these cell lines (Supplemen-tary Fig. S7) nor did it improve colony outgrowth in long-termassays (Fig. 5A).

We next tested whether ligand availability is rate limiting forthe induction of PI3K inhibitor resistance, by adding IGF2 tothe culture media. We observed decreased sensitivity to GDC-0941 with the addition of IGF2 to HCC1806 and CAL-51 cells,but not for CAL-148 and MFM-223 cells (Fig. 5B). Correspond-ingly, colony outgrowth under PI3K inhibition was markedlyenhanced by the addition of the ligand in HCC1806 and CAL-51, but not CAL-148 and MFM-223 cells (Fig. 5A; Supplemen-tary Fig. S8). These effects were reversible by the addition ofIGF1R inhibitor. In line with these findings, stimulation withIGF2 could only overcome a PI3K block in combination-sen-sitive cells, as evidenced by phosphorylation of AKT andPRAS40 (Fig. 5C).

To further establish the role of ligands, we generated PI3Kdose–response curves in the presence or absence of neutralizinganti-IGF2 antibodies in HCC1806, CAL-51, CAL-148, and MFM-223 cells. IGF2-neutralizing antibodies increased sensitivity toPI3K inhibition inHCC1806andCAL-51, but not inCAL-148 andMFM-223 cells (Fig. 5D; Supplementary Fig. S9). These datademonstrate that the levels of IGF2 can modulate sensitivity ofHCC1806 and CAL-51 to PI3K inhibition.

Our functional experiments suggest that ligand levels are morelimiting to PI3K signaling than IGF1R abundance. We therefore

hypothesized that IGF2 is also likely to be upregulated in com-bination-sensitive cells after exposure to GDC-0941. Indeed,levels of IGF2 mRNA increased after 24 hours of exposure toGDC-0941 in CAL-51 and HCC1806, but not in CAL-148 andMFM-223 cells (Fig. 5E). IGF1 mRNA levels were below thedetection limit of our qRT-PCR (data not shown) and was alsonot detected in RNAseq of HCC1806 and with a single read onlyin CAL-51, and is therefore not likely to play an important rolein IGF1R signaling in these cell lines.

Molecular predictors of sensitivity to PI3K/IGF1Rcombination treatment

Although promising in a subset of TNBC cell lines, com-bined PI3K/IGF1R inhibition has no beneficial effect on anequally large subset of cell lines. The different responses amongTNBC cell lines most likely reflect the clinical situation. Theidentification of predictive biomarkers for the PI3K/IGF1Rdrug combination is therefore instrumental for further clinicaldevelopment. We have measured protein (reverse-phase pro-tein array) and mRNA (RNA-Seq) expression for each cell lineof our TNBC panel, and set out to employ these data to identifypredictive markers of drug synergy. We first determined wheth-er IGF1R expression is associated with a positive response tothe drug combination. Neither protein nor mRNA levels ofIGF1R showed any significant correlation with response tocombination treatment (Supplementary Fig. S3 and Supple-mentary Table S4; Pearson's correlation for mRNA ¼ �0.08;

++

IGF1R

pIGF1R

pAKTS473

pAKTT308

pPRAS40

HSP90

––

––++

++–

––

–++

++–

––

–++

++–

––

–++

MFM-223CAL-148CAL-51HCC1806

OSI-906GDC-0941

pERK

PARP

cl PARP

5 µmol/L GDC-09410 µmol/L GDC-0941

0 µm

ol/L

OS

I-906

2 µm

ol/L

OS

I-906

pIGF1R

IGF1R

pAKTS473

pAKTT308

pPRAS40

HSP90

+–– +–+

GDC-0941 (µmol/L)

+–– +–+ +–– +–+

10.10

HCC1806

Nor

mal

ized

apo

ptos

is

CAL-5100

0.04

0.08

0.12

0.16

0.2

0.2

0.4

0.6

0.8

1

CAL-148 MFM-223

ns

ns

No drug

2 μmol/L OSI-906

5 μmol/L GDC-0941

GDC+OSI

10.1010.10

+–– +–+

10.10

OSI-906

CBA

D

Figure 4.

Combined PI3K and IGF1R inhibition abrogates PI3K signaling and increases apoptosis in lines sensitive to the combination. A, biochemical responses ofcombination-sensitive (HCC1806 and CAL-51) and combination-insensitive (CAL-148 and MFM-223) TNBC cell lines mock treated (DMSO) or exposed to1 mmol/L GDC-0941, 2 mmol/L OSI-906, or the combination of both drugs. Cells were harvested 24 hours after drug treatment, and (phospho)proteinlevels were assayed by Western blotting. Heat shock protein 90 (HSP90) served as a loading control. B, apoptosis in CAL-51 cells 24 hours after exposureto vehicle (DMSO), 5 mmol/L GDC-0941, 2 mmol/L OSI-906, or the combination of both drugs as imaged by IncuCyte using a Caspase-3 activated DNAdye (green). Purple lines indicate detected confluence. C, levels of apoptosis for CAL-51, HCC1806, CAL-148, and MFM-223 24 hours after exposure tovehicle (DMSO), 5 mmol/L GDC-0941, 2 mmol/L OSI-906, or the combination of both drugs. Apoptosis was quantified by confluence of the Caspase-3activated DNA dye relative to the total cell confluence. The level of apoptosis was normalized to the amount of apoptosis observed at 1 mmol/L staurosporin(¼1) for each cell line. � , P < 0.01; ns, not significant. D, biochemical responses of combination-sensitive (HCC1806 and CAL-51) and combination-insensitive(CAL-148 and MFM-223) TNBC cell lines exposed to indicated concentrations of GDC-0941 in the presence or absence of 2 mmol/L OSI-906. Cells wereharvested after 1 hour of drug exposure, and (phospho)protein levels were assayed by Western blot. HSP90 served as a loading control.






HCC1806

Control

No drug

Rel

ativ

e vi

abilit

y

0

0.25

0.5

0.75

1

–1.5 –1GDC-0941 (log10 μmol/L)

–0.5 0 0.5 1

2 μmol/L OSI-90650 ng/mL IGF2IGF2 + OSI

Control

No drug

Rel

ativ

e vi

abilit

y

0

0.25

0.5

0.75

1

–1.5 –1GDC-0941 (log10 μmol/L)

–0.5 0 0.5 1


Control

No drug

Rel

ativ

e vi

abilit

y

0

0.25

0.5

0.75

1

–1.5 –1GDC-0941 (log10 μmol/L)

–0.5 0 0.5 1


Control

No drug

Rel

ativ

e vi

abilit

y

0

0.25

0.5

0.75

1

–1.5 –1GDC-0941 (log10 μmol/L)

–0.5 0 0.5 1


HCC1806

CAL-51

CAL-148

MFM-223

MFM-223CAL-148CAL-51

No drug (5 days)

IGF1REmpty vector IGF1REmpty vector

5 µmol/L GDC-0941 (14 days)

GDC-0941only

2 µmol/LOSI-906

50 ng/mLIGF2

IGF2 + OSI-906

GDC-0941only

2 µmol/LOSI-906

50 ng/mLIGF2

IGF2 + OSI-906

CAL-51HCC1806

OSI-906

IGF2

++––

––

+––

+

HCC1806empty vector

+–– ++GDC-0941

++––

––

+––

+

HCC1806IGF1R

+–– ++

++––

––

+––

+

CAL-51empty vector

+–– ++

++––

––

+––

+

CAL-51IGF1R

+–– ++

++––

––

+––

+

CAL-148empty vector

+–– ++

++––

––

+––

+

CAL-148IGF1R

+–– ++

++––

––

+––

+

MFM-223empty vector

+–– ++

++––

––

+––

+

MFM-223IGF1R

+–– ++

pIGF1R

IGF1R

pAKTS473

pPRAS40HSP90

A

B

C

ED

No drug

Rel

ativ

e vi

abili

ty

0

0.25

0.5

0.75

1

–1.5 –1

GDC-0941 (log10 μmol/L)–0.5 0 0.5 1

Control2 μmol/L OSI-906Anti-IGF2

No drug

Rel

ativ

e vi

abili

ty

0

0.25

0.5

0.75

1

–1.5 –1

GDC-0941 (log10 μmol/L)–0.5 0 0.5 1

0–

HCC1806 CAL-51 CAL-148 MFM-223+ – + – + – +

2

4

150

IGF2

exp

ress

ion

(arb

itrar

y un

its)

200

250

300

*, P < 0.01Control2 μmol/L OSI-906Anti-IGF2

Figure 5.

Availability of IGF1R ligandsmodulates the sensitivity to PI3K inhibition in combination-sensitive cells. A, colony formation assayof control (empty vector) and IGF1R-expressingHCC1806, CAL-51, CAL-148, andMFM-223 cell lines treatedwith the indicated combinations of 5mmol/LGDC-0941, 2mmol/LOSI-906, and50 ng/mL IGF2.Control wells (no drug) were fixed and stained after 5 days, whereas the GDC-0941 exposed wells were fixed and stained after 14 days. B, GDC-0941 dose–response curve for HCC1806, CAL-51, CAL-148, andMFM-223 cells, in combinationwith 2 mmol/L OSI-906, 50 ng/mL IGF2, or the combination of OSI-906 (2 mmol/L)and IGF2 (50 ng/mL) as determined in a 72-hour cell viability assay. C, biochemical responses of control (empty vector) and IGF1R-expressing HCC1806,CAL-51, CAL-148, and MFM-223 cell lines treated with the indicated combinations of 1 mmol/L GDC-0941, 2 mmol/L OSI-906, and 50 ng/mL IGF2. Cells were drugtreated for 1 hour, harvested, and (phospho)protein levels assayed by Western blotting. Heat shock protein 90 (HSP90) served as a loading control. D, GDC-0941dose–response curve of HCC1806 and CAL-51 cells, in combination with 2 mmol/L OSI-906 or anti-IGF2 antibodies (0.5 mg/mL). E, IGF2 mRNA expressionlevels in HCC1806, CAL-51, CAL-148, and MFM-223 in the absence or presence of 1 mmol/L GDC-0941. Cells were exposed to drug for 24 hours, and IGF2mRNA levels were quantified by qRT-PCR and normalized to GAPDH for each cell line.

de Lint et al.





95% CI, �0.53, 0.40; and Pearson's correlation for protein ¼�0.19; 95% CI, �0.60 to 0.31). This observation is in agree-ment with the results obtained with the IGF1R-overexpressingcell lines (Supplementary Figs. S7 and S8). As a more func-tional readout of activity, we looked at phospho-IGF1R, whichalso did not show a correlation with response (SupplementaryFig. S3). We then analyzed the mRNA expression of a set ofIGF1R- and PI3K-related genes (Supplementary Table S5).Within this gene set, IGF2BP3 expression was most stronglycorrelated with the synergistic interaction (Fig. 6A and B).

Protein levels yielded similar results, with the highly synergis-tic cell lines CAL-51, HCC1806, and MDA-MB-436 showinghighest IGF2BP3 expression (Supplementary Fig. S3).IGF2BP3, also known as IMP-3, is an oncofetal protein thathas been implicated by several independent studies as a bio-marker for an unfavorable outcome in TNBC (23, 24) as well asin other tumor types (25, 26). Analysis of IGF2BP3 mRNAlevels in 959 TCGA breast cancer samples (7) confirmedthat the expression of this gene is strongly associated with theTN status with high expression levels mostly occurring in TN

Figure 6.

Molecular predictors of sensitivity to PI3K/IGF1R combination treatment. A, correlation between the mRNA expression levels of a curated set of genesinvolved in IGF1R signaling and the DAUC upon exposure to the combination of GDC-0941 and OSI-906. The green area indicates a 95% CI for randomlyselected gene sets. q values represent Benjamini–Hochberg corrected P values derived from the observed correlations with 15 degrees of freedom. B,correlation between DAUC and gene expression of IGF2BP3 in the TNBC cell line panel. Combination-sensitive cell lines (HCC1806 and CAL-51) areindicated in green, and combination-insensitive cell lines (CAL-148 and MFM-223) in red. C, expression of IGF2BP3 mRNA in normal tissue and breasttumors. Data derived from the TCGA database.






tumors (P value ¼ 2.3�� 10�35, two sample Wilcoxon test ofIGF2BP3 normalized counts for n¼ 114 TN vs. n¼ 845 non-TNbreast cancers, Fig. 6C). Furthermore, IGF2BP3 mRNA levelsdisplay a bimodal distribution in TN breast cancers. Wehypothesize that the population of TN breast cancers expres-sing high levels of IGF2BP3 are highly enriched for tumorssensitive to the combination treatment.

DiscussionAs PIK3CA is the most frequently mutated and amplified

oncogenic driver in TNBC, inhibition of this kinase is an obviouscandidate for targeted therapy in these tumors.We speculated thatdirectly inhibiting PI3K instead of downstream components ofthe pathway provides better specificity toward cancer cells. Also,targeting downstream effectors, such as AKT or mTOR, does notprevent PI3K signaling through PDPK1 and other proteinsrecruited to the plasma membrane by PIP3, the product of PI3Kactivity.

PI3K inhibitors efficiently inhibit proliferation of manyTNBC cells. However, clinical regression of tumors is usuallynot observed (27). We sought to find enhancers of responseto the PI3K inhibitor GDC-0941 that could augment its effecton tumor cells. Supporting the need for these enhancers,results from the first-in-human phase I study of GDC-0941in patients with advanced solid tumors show a modest re-sponse rate, and the authors postulate that >90% inhibition ofAKT phosphorylation is required to effectively inhibit tumorcell proliferation (28).

Our kinome-wide genetic screen in the TNBC cell lineHCC1806 identified IGF1R as a gene whose knockdown signif-icantly sensitizes these cells to PI3K inhibition.Our study suggeststhat perhaps as much as half of TNBC patients could benefit fromthe addition of IGF1R inhibitors to a PI3K inhibition regime. Toour knowledge, this is a novel combination of two anticancerdrugs that warrants preclinical follow-up.

On a molecular level, we show that in cell lines with asynergistic response to our drug combination, simultaneousinhibition of PI3K and IGF1R is more effective than eitherdrug target alone in abrogating signaling downstream of PI3K.Decreased phosphorylation of AKT and its substrate, PRAS40,correlates to a strong decrease in proliferation and increas-ed apoptosis. These findings are in line with several recentreports on combination therapies in breast cancer, whichshow that drugs further inhibiting the PI3K pathway are mostefficacious (29–31).

Although it has been observed that MAPK activation cancompensate for PI3K pathway inhibition, in particular in mela-noma and colorectal cancer, we did not observe a compensatoryrole ofMAPK signaling in our breast cancer cell lines. Sensitivity toMEK inhibition was found in only one cell line in the panel, andno kinases downstream of RTKs in the MAPK pathway wereidentified as hits in our screen. We did not observe an effect ofthe inhibitors or their combination on pERK, a marker of MAPKpathway activity. Together, these findings indicate that these celllines do not primarily depend on the MAPK pathway for theirproliferation and that abrogation of PI3K signaling alone issufficient to impede proliferation and survival in the combina-tion-sensitive cell lines.

IGF signaling has long been suggested to play an importantrole in cancer (16, 32), and early gene expression studies

pinpointed IGF2 to be specifically highly expressed in tumorcells compared with their normal counterparts (33). This madeIGF1R a promising drug target and the subject of intensiveresearch and drug development during the late 1990s (34, 35).Initial trials with monoclonal antibodies as monotherapy werenevertheless disappointing, and further development of thesetherapies is in question (17, 36). To further illustrate this,recent phase I trials of OSI-906 as a single agent demonstratedevidence of clinical benefit (37), and it was speculated thatmodest coinhibition of the insulin receptor (IR) prevents therelief of a negative feedback loop stimulating tumor growth byelevated insulin levels in vivo (38). In contrast, in a phase IIIstudy in patients with adrenocortical carcinoma, a cancer typewith frequent IGF2 overexpression, no improvement of overallsurvival was observed with OSI-906 (39).

IGF1R inhibitors as a single treatment were shown to have amodest effect on proliferation selectively in KRAS mutantcancer cell lines (40), but in these cancers, IGF1R is not theprimary driver. Therefore, it is not surprising that IGF1R inhi-bition alone was insufficient to achieve prolonged response.Indeed, in KRAS mutant colon cancer, IGF1R inhibition alonehad limited effect on cell survival (41). However, when IGF1Rinhibition was combined with MEK inhibition, cells underwentapoptosis and proliferation was severely impaired. The com-bination of MEK and IGF1R inhibitor was also found to bemore potent than either alone in lung cancer similar to colo-rectal cancer (40, 42).

IGF1R signaling has also been implicated in sensitivity to drugstargeting the PI3K/AKT pathway. IGF1R has been suggested toplay a role in ovarian cancer due to its upregulation after com-bined PI3K and mTOR inhibition (43) as well as in a range ofovarian cancer cell lines after AKT inhibition (20), but was notpursued with functional inhibitor experiments in these studies.These and other (44–47) examples demonstrate that IGF1Rinhibition may prove a valuable addition to inhibitors targetingthe MAPK and PI3K pathways.

Our results indicate that the PI3K/IGF1R inhibition com-bination is not expected to be effective in all TNBC. Therefore,a predictive biomarker that distinguishes responders fromnonresponders would be crucial for further clinical develop-ment of the PI3K/IGF1R inhibitor combination therapy. Inall subtypes of breast cancer, phospho-IGF1R can be detect-ed and is correlated with poor survival (48). In our cellline panel, levels of phospho-IGF1R or total IGF1R did notpredict sensitivity to IGF1R inhibition alone or in combi-nation with PI3K inhibition. Among genes related to IGF1Rand PI3K pathway, expression of IGF2BP3 correlated bestwith response to the combination. IGF2BP3 is an oncofetalprotein, which, in breast cancer, is expressed predominantlyin TNBC. It has been shown to bind to IGF2 mRNA andpromote its translation in leukemia cells (49); in glioblasto-ma, it was shown to activate the PI3K pathway by increasingexpression of IGF2 (50). Since we identified IGF2BP3 frombase-level mRNA expression, it is a potential predictive bio-marker for benefit of IGF1R inhibition alongside PI3K inhi-bition. Whether IGF2BP3 plays an active role in mediatingthe contribution of IGF1R to PI3K inhibitor sensitivity is asyet unclear.

The work presented here suggests that the combination ofPI3K and IGF1R inhibitors has potential application in thetreatment of TNBC and warrants testing in a clinical setting.

de Lint et al.





Disclosure of Potential Conflicts of InterestNo potential conflicts of interest were disclosed.

Authors' ContributionsConception and design: K. de Lint, J.B. Poell, J.V. Rodriguez, R.L. BeijersbergenDevelopment of methodology: K. de Lint, J.B. PoellAcquisition of data (provided animals, acquired and managed pati-ents, provided facilities, etc.): K. de Lint, J.B. Poell, K. Jastrzebski,J.V. RodriguezAnalysis and interpretation of data (e.g., statistical analysis, biostatistics,computational analysis): K. de Lint, J.B. Poell, H. Soueidan, J.V. Rodriguez,C. Lieftink, L.F.A. WesselsWriting, review, and/or revision of the manuscript: K. de Lint, J.B. Poell,H. Soueidan, K. Jastrzebski, L.F.A. Wessels, R.L. BeijersbergenAdministrative, technical, or material support (i.e., reporting or organizingdata, constructing databases): K. de LintStudy supervision: R.L. Beijersbergen

AcknowledgmentsThe authors thank Pasi Halonen and Ben Morris from the NKI Robotics

and Screening Center and the NKI Genomics Core Facility for technicalsupport and members from the Beijersbergen and Bernards labs for discus-sion and suggestions.

Grant SupportR.L. Beijersbergen and L.F.A. Wessels received a Cancer Systems Biology

Center grant (853.00.120) from the Netherlands organization for ScientificResearch (NWO).

The costs of publication of this article were defrayed in part by thepayment of page charges. This article must therefore be hereby markedadvertisement in accordance with 18 U.S.C. Section 1734 solely to indicatethis fact.

Received October 30, 2015; revisedMarch 3, 2016; accepted March 26, 2016;published OnlineFirst May 11, 2016.

References1. DeSantis CE, Lin CC, Mariotto AB, Siegel RL, Stein KD, Kramer JL, et al.

Cancer treatment and survivorship statistics, 2014. CA Cancer J Clin2014;64:252–71.

2. Lehmann BD, Pietenpol JA, Tan AR. Triple-negative breast cancer: Molec-ular subtypes and new targets for therapy. Am Soc Clin Oncol Educ Book2015;35:e31–9.

3. Venkitaraman R. Triple-negative/basal-like breast cancer: Clinical,pathologic and molecular features. Expert Rev Anticancer Ther 2010;10:199–207.

4. Rakha EA, El-Sayed ME, Green AR, Lee AH, Robertson JF, Ellis IO. Prog-nostic markers in triple-negative breast cancer. Cancer 2007;109:25–32.

5. Banerji S, Cibulskis K, Rangel-Escareno C, Brown KK, Carter SL, FrederickAM, et al. Sequence analysis of mutations and translocations across breastcancer subtypes. Nature 2012;486:405–9.

6. Shah SP, Roth A, Goya R, Oloumi A, Ha G, Zhao Y, et al. The clonal andmutational evolution spectrum of primary triple-negative breast cancers.Nature 2012;486:395–9.

7. Cancer Genome Atlas Network. Comprehensive molecular portraits ofhuman breast tumours. Nature 2012;490:61–70.

8. Crown J, O'Shaughnessy J, Gullo G. Emerging targeted therapies in triple-negative breast cancer. Ann Oncol 2012;23Suppl 6:vi56–65.

9. Klarenbeek S, vanMiltenburgMH, Jonkers J. Genetically engineeredmousemodels of PI3K signaling in breast cancer. Mol Oncol 2013;7:146–64.

10. Raynaud FI, Eccles SA, Patel S, Alix S, Box G, Chuckowree I, et al. Biologicalproperties of potent inhibitors of class I phosphatidylinositide 3-kinases:From PI-103 through PI-540, PI-620 to the oral agent GDC-0941.Mol Cancer Ther 2009;8:1725–38.

11. Rodon J, Dienstmann R, Serra V, Tabernero J. Development of PI3Kinhibitors: Lessons learned from early clinical trials. Nat Rev Clin Oncol2013;10:143–53.

12. Klempner SJ, Myers AP, Cantley LC. What a tangled web we weave:Emerging resistance mechanisms to inhibition of the phosphoinositide3-kinase pathway. Cancer Discov 2013;3:1345–54.

13. Vora SR, Juric D, Kim N, Mino-Kenudson M, Huynh T, Costa C, et al. CDK4/6 inhibitors sensitize PIK3CA mutant breast cancer to PI3K inhibitors.Cancer Cell 2014;26:136–49.

14. Franceschini A, Szklarczyk D, Frankild S, Kuhn M, Simonovic M, Roth A,et al. STRING v9. 1: Protein-protein interaction networks, with increasedcoverage and integration. Nucleic Acids Res 2013;41(D1):D808–815.

15. O'BrienC,Wallin JJ, SampathD,GuhaThakurtaD, SavageH, Punnoose EA,et al. Predictive biomarkers of sensitivity to the phosphatidylinositol30 kinase inhibitor GDC-0941 in breast cancer preclinical models.Clin Cancer Res 2010;16:3670–83.

16. Pollak M. Insulin and insulin-like growth factor signalling in neoplasia.Nat Rev Cancer 2008;8:915–28.

17. GuhaM. Anticancer IGF1R classes take more knocks. Nat Rev Drug Discov2013;12:250.

18. Foucquier J, Guedj M. Analysis of drug combinations: Current methodo-logical landscape. Pharmacol Res Perspect 2015;3:e00149.

19. Kennedy SG, Wagner AJ, Conzen SD, Jordan J, Bellacosa A, Tsichlis PN,et al. The PI 3-kinase/Akt signaling pathway delivers an anti-apoptoticsignal. Genes Dev 1997;11:701–13.

20. Chandarlapaty S, Sawai A, Scaltriti M, Rodrik-Outmezguine V, Grbovic-Huezo O, Serra V, et al. AKT inhibition relieves feedback suppression ofreceptor tyrosine kinase expression and activity. Cancer Cell 2011;19:58–71.

21. Stommel JM, Kimmelman AC, Ying H, Nabioullin R, Ponugoti AH,Wiedemeyer R, et al. Coactivation of receptor tyrosine kinases affects theresponse of tumor cells to targeted therapies. Science 2007;318:287–90.

22. Sun C, Bernards R. Feedback and redundancy in receptor tyrosine kinasesignaling: relevance to cancer therapies. Trends Biochem Sci 2014;39:465–74.

23. Walter O, Prasad M, Lu S, Quinlan RM, Edmiston KL, Khan A. IMP3 is anovel biomarker for triple negative invasive mammary carcinoma associ-ated with a more aggressive phenotype. Hum Pathol 2009;40:1528–33.

24. Sidoni A, Cartaginese F. IMP3 expression in triple-negative breast carci-noma. Hum Pathol 2010;41:1355–6.

25. Kobel M, Xu H, Bourne PA, Spaulding BO, Shih Ie M, Mao TL, et al.IGF2BP3 (IMP3) expression is a marker of unfavorable prognosis inovarian carcinoma of clear cell subtype. Mod Pathol 2009;22:469–75.

26. Lochhead P, Imamura Y,Morikawa T, Kuchiba A, YamauchiM, LiaoX, et al.Insulin-like growth factor 2messenger RNAbinding protein 3 (IGF2BP3) isa marker of unfavourable prognosis in colorectal cancer. Eur J Cancer2012;48:3405–13.

27. JuricD,Rodon J,Gonzalez-Angulo AM,BurrisHA, Bendell J, Berlin JD, et al.BYL719, a next generation PI3K alpha specific inhibitor: Preliminary safety,PK, and efficacy results from the first-in-human study. Cancer Res 2012;72(8 Suppl; abstr CT-01).

28. Sarker D, Ang JE, Baird R, Kristeleit R, Shah K, Moreno V, et al. First-in-human phase I study of pictilisib (GDC-0941), a potent pan-class Iphosphatidylinositol-3-kinase (PI3K) inhibitor, in patients with advancedsolid tumors. Clin Cancer Res 2015;21:77–86.

29. Costa C, Ebi H, Martini M, Beausoleil SA, Faber AC, Jakubik CT, et al.Measurement of PIP3 levels reveals an unexpected role for p110beta inearly adaptive responses to p110alpha-specific inhibitors in luminalbreast cancer. Cancer Cell 2015;27:97–108.

30. Elkabets M, Vora S, Juric D, Morse N, Mino-KenudsonM, Muranen T, et al.mTORC1 inhibition is required for sensitivity to PI3K p110alpha inhibi-tors in PIK3CA-mutant breast cancer. Sci Transl Med 2013;5:196ra99.

31. Bosch A, Li Z, Bergamaschi A, Ellis H, Toska E, Prat A, et al. PI3K inhibitionresults in enhanced estrogen receptor function and dependence in hor-mone receptor-positive breast cancer. Sci Transl Med 2015;7:283ra51.

32. Pollak MN, Perdue JF, Margolese RG, Baer K, Richard M. Presence ofsomatomedin receptors on primary human breast and colon carcinomas.Cancer Lett 1987;38:223–30.

33. Gray A, Tam AW, Dull TJ, Hayflick J, Pintar J, Cavenee WK, et al. Tissue-specific and developmentally regulated transcription of the insulin-likegrowth factor 2 gene. DNA 1987;6:283–95.






34. Cullen KJ, YeeD, SlyWS, Perdue J, Hampton B, LippmanME, et al. Insulin-like growth factor receptor expression and function inhumanbreast cancer.Cancer Res 1990;50:48–53.

35. Baserga R. The insulin-like growth factor I receptor: A key to tumor growth?Cancer Res 1995;55:249–52.

36. Chen HX, Sharon E. IGF-1R as an anti-cancer target–trials and tribulations.Chin J Cancer 2013;32:242–52.

37. Jones RL, Kim ES, Nava-Parada P, Alam S, Johnson FM, Stephens AW, et al.Phase I study of intermittent oral dosing of the insulin-like growth factor-1and insulin receptors inhibitor OSI-906 in patients with advanced solidtumors. Clin Cancer Res 2015;21:693–700.

38. Yee D. A tale of two receptors: Insulin and insulin-like growth factorsignaling in cancer. Clin Cancer Res 2015;21:667–9.

39. Fassnacht M, Berruti A, Baudin E, Demeure MJ, Gilbert J, Haak H, et al.Linsitinib (OSI-906) versus placebo for patients with locally advanced ormetastatic adrenocortical carcinoma: Adouble-blind, randomised, phase 3study. Lancet Oncol 2015;16:426–35.

40. Molina-Arcas M, Hancock DC, Sheridan C, Kumar MS, Downward J.Coordinate direct input of both KRAS and IGF1 receptor to activation ofPI3 kinase in KRAS-mutant lung cancer. Cancer Discov 2013;3:548–63.

41. Flanigan SA, Pitts TM, Newton TP, Kulikowski GN, Tan AC, McManus MC,et al.Overcoming IGF1R/IR resistance through inhibitionofMEK signalingin colorectal cancer models. Clin Cancer Res 2013;19:6219–29.

42. Chen R, Sweet-Cordero EA. Two is better than one: Combining IGF1R andMEK blockade as a promising novel treatment strategy against KRAS-mutant lung cancer. Cancer Discov 2013;3:491–3.

43. Muranen T, Selfors LM,Worster DT, Iwanicki MP, Song L,Morales FC, et al.Inhibition of PI3K/mTOR leads to adaptive resistance in matrix-attachedcancer cells. Cancer Cell 2012;21:227–39.

44. Chakraborty AK, Zerillo C, DiGiovanna MP. In vitro and in vivo studies ofthe combination of IGF1R inhibitor figitumumab (CP-751,871) withHER2 inhibitors trastuzumab and neratinib. Breast Cancer Res Treat2015;152:533–44.

45. Zhang Z, Wang J, Ji D, Wang C, Liu R, Wu Z, et al. Functional geneticapproach identifies MET, HER3, IGF1R, INSR pathways as determinants oflapatinib unresponsiveness in HER2-positive gastric cancer. Clin CancerRes 2014;20:4559–73.

46. Di Cosimo S, Sathyanarayanan S, Bendell JC, Cervantes A, SteinMN, BranaI, et al. Combination of the mTOR inhibitor ridaforolimus and the anti-IGF1R monoclonal antibody dalotuzumab: Preclinical characterizationand phase I clinical trial. Clin Cancer Res 2015;21:49–59.

47. Fox EM, Kuba MG, Miller TW, Davies BR, Arteaga CL. Autocrine IGF-I/insulin receptor axis compensates for inhibition of AKT in ER-positivebreast cancer cellswith resistance to estrogendeprivation. BreastCancer Res2013;15:R55.

48. Law JH, Habibi G, Hu K, Masoudi H, Wang MY, Stratford AL, et al.Phosphorylated insulin-like growth factor-i/insulin receptor is present inall breast cancer subtypes and is related to poor survival. Cancer Res2008;68:10238–46.

49. Liao B, Hu Y, Herrick DJ, Brewer G. The RNA-binding protein IMP-3 is atranslational activator of insulin-like growth factor II leader-3 mRNAduring proliferation of human K562 leukemia cells. J Biol Chem 2005;280:18517–24.

50. Suvasini R, Shruti B, Thota B, Shinde SV, Friedmann-Morvinski D, NawazZ, et al. Insulin growth factor-2 binding protein 3 (IGF2BP3) is a glioblas-toma-specific marker that activates phosphatidylinositol 3-kinase/mito-gen-activatedprotein kinase (PI3K/MAPK)pathways bymodulating IGF-2.J Biol Chem 2011;286:25882–90.


de Lint et al.




2016;15:1545-1556. Published OnlineFirst May 11, 2016.Mol Cancer Ther Klaas de Lint, Jos B. Poell, Hayssam Soueidan, et al. Cotargeting IGF1RSensitizing Triple-Negative Breast Cancer to PI3K Inhibition by

Updated version

10.1158/1535-7163.MCT-15-0865doi:

Access the most recent version of this article at:

Material

Supplementary

http://mct.aacrjournals.org/content/suppl/2016/05/11/1535-7163.MCT-15-0865.DC1

Access the most recent supplemental material at:

Cited articles

http://mct.aacrjournals.org/content/15/7/1545.full#ref-list-1

This article cites 49 articles, 20 of which you can access for free at:

Citing articles

http://mct.aacrjournals.org/content/15/7/1545.full#related-urls

This article has been cited by 3 HighWire-hosted articles. Access the articles at:

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

Subscriptions

Reprints and

[email protected]

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

Permissions

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

.http://mct.aacrjournals.org/content/15/7/1545To request permission to re-use all or part of this article, use this link



http://mct.aacrjournals.org/lookup/doi/10.1158/1535-7163.MCT-15-0865

http://mct.aacrjournals.org/content/suppl/2016/05/11/1535-7163.MCT-15-0865.DC1

http://mct.aacrjournals.org/content/15/7/1545.full#ref-list-1

http://mct.aacrjournals.org/content/15/7/1545.full#related-urls

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

mailto:[email protected]

http://mct.aacrjournals.org/content/15/7/1545


sensitizing triple-negative breast cancer to pi3k inhibition …...small molecule therapeutics...

Documents