EGF receptor uses SOS1 to drive constitutive activationof NFκB in cancer cellsSarmishtha De, Josephine Kam Tai Dermawan, and George R. Stark1

Department of Molecular Genetics, Lerner Research Institute, Cleveland Clinic, Cleveland, OH 44195

Contributed by George R. Stark, July 2, 2014 (sent for review April 9, 2014)

Activation of nuclear factor κB (NFκB) is a central event in theresponses of normal cells to inflammatory signals, and the abnor-mal constitutive activation of NFκB is important for the survival ofmost cancer cells. In nonmalignant human cells, EGF stimulates ro-bust activation of NFκB. The kinase activity of the EGF receptor(EGFR) is required, because the potent and specific inhibitor erlotinibblocks the response. Down-regulating EGFR expression or inhibit-ing EGFR with erlotinib impairs constitutive NFκB activation inseveral different types of cancer cells and, conversely, increasedactivation of NFκB leads to erlotinib resistance in these cells. Weconclude that EGF is an important mediator of NFκB activation incancer cells. To explore the mechanism, we selected an erlotinib-resistant cell line in which the guanine nucleotide exchange factorSon of Sevenless 1 (SOS1), well known to be important for EGF-dependent signaling to MAP kinases, is overexpressed. Increasedexpression of SOS1 increases NFκB activation in several differenttypes of cancer cells, and ablation of SOS1 inhibits EGF-inducedNFκB activation in these cells, indicating that SOS1 is a functionalcomponent of the pathway connecting EGFR to NFκB activation.Importantly, the guanine nucleotide exchange activity of SOS1 isnot required for NFκB activation.

Nuclear factor κB (NFκB), an important mediator of thenormal response to inflammatory signals, is abnormally

constitutively activated in most cancer cells, promoting re-sistance to apoptosis and contributing to tumorigenesis by driv-ing cell proliferation and metastasis (1, 2). The five members ofthe mammalian NFκB family, RelA (p65), RelB, cRel, p105/p50(NFκB1), and p100/p52 (NFκB2), form a variety of homo- andheterodimers. In normal unstimulated cells, NFκB dimers arekept inactive as cytoplasmic complexes, bound to a member ofthe inhibitor of κB (IκB) family. Many pathways that releaseNFκB from IκB use IκB kinase (IKK), which phosphorylatesIκB, leading to its ubiquitination and proteasome-mediateddegradation, liberating NFκB dimers, which then translocate tothe nucleus where they activate the transcription of target genes(3). The activation of NFκB is regulated by many differentstimuli in virtually all cell types, with many different functionalconsequences (4, 5). Specific and highly regulated control ofNFκB is critical for its normal transient activation in response tostress and proinflammatory signals. Aberrant constitutive acti-vation of NFκB in cancer (6, 7) contributes to malignant pro-gression and therapeutic resistance, both in cell lines and intumors (1, 8, 9).The EGF receptor (EGFR/HER-1/ErbB1) is a member of the

ErbB family of receptor tyrosine kinases. Upon stimulation,EGFR undergoes homodimerization or heterodimerization withother family members (HER-2/neu/ErbB2, HER-3/ErbB3 andHER-4/ErbB4) (10), leading to autophosphorylation and asso-ciation with a set of intracellular signaling proteins that havebeen intensively studied for many years (11). Activation ofdownstream pathways facilitates cell growth, survival, and pro-liferation. Activation and mutation of EGFR have been observedin up to 30% of many different solid tumors, including head andneck, colorectal, breast, nonsmall cell lung, pancreatic, and gas-tric cancers, and usually correlate with a poor prognosis (12, 13).Thus, there is great interest in EGFR as a target for anticancer

therapies that use small molecule inhibitors of its tyrosine kinaseactivity (14).In contrast to the intensively studied pathways of NFκB acti-

vation as a part of the inflammatory response, the mechanismsunderlying its activation in cancer are diverse and have not beenwell defined. For example, Lu et al. (6) showed that differentcancer cell lines secrete several different cytokines and growthfactors, each of which is capable of activating NFκB, includingsome for which this activity was not anticipated, such as trans-forming growth factor β and fibroblast growth factor 5. Fur-thermore, it is well known that some genes encoding cytokinesthat activate NFκB are themselves NFκB targets, including IL-1βand TNF-α, revealing positive feedback loops. NFκB can also beactivated by EGF (15, 16), and different laboratories have in-dicated roles for several different proteins in this pathway, in-cluding RIP (17), NIK (18), GRB2-associated binder 1 (19),mTORC2 (20), CARMA3 (21), and FER (22). Although EGFR-dependent activation of NFκB has been reported before, itsimportance as a means of constitutive NFκB activation in canceris unclear, and it is fair to say that the pathway has not yet beenwell defined.To address these issues, we examined the ability of EGF/

EGFR to activate NFκB in both nonmalignant and malignanthuman cell lines. Inhibition of the kinase activity of EGFR by thespecific inhibitor erlotinib or knockdown of EGFR expressionimpaired the activation of NFκB. Conversely, activating NFκBwith IL-1β increased resistance to erlotinib in cancer cells. Toexplore the EGFR-NFκB connection in an unbiased way, weused insertional mutagenesis to up-regulate the expression ofproteins that mediate resistance to erlotinib, selecting a resistantcell line in which the guanine nucleotide exchange factor, Son ofSevenless 1 (SOS1), is overexpressed. SOS1 promotes RAS and

Significance

In normal cells, quiescent nuclear factor κB (NFκB) is activated byinflammatory stimuli. In most cancers, the abnormal constitutiveactivation of NFκB contributes to malignant progression and re-sistance to therapy. Overexpression or constitutive activation ofthe EGF receptor (EGFR) in many cancers contributes to theirproliferation and survival. We find that the constitutive activationof NFκB in several cancer cell lines is decreased by EGFR knock-down or by the EGFR inhibitor erlotinib. We used insertionalmutagenesis to find that overexpression of Son of Sevenless 1(SOS1), a component of EGF-dependent pathways that facilitatecell growth and survival, causes erlotinib resistance and increasesNFκB activation. SOS1 is required for EGF-dependent activationof NFκB but its GDP–GTP exchange activity is not, revealinga novel function for this protein.

Author contributions: S.D., J.K.T.D., and G.R.S. designed research; S.D. and J.K.T.D. per-formed research; S.D. contributed new reagents/analytic tools; S.D., J.K.T.D., and G.R.S.analyzed data; and S.D. and G.R.S. wrote the paper.

The authors declare no conflict of interest.1To whom correspondence should be addressed. Email: starkg@ccf.org.

This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.1073/pnas.1412390111/-/DCSupplemental.

www.pnas.org/cgi/doi/10.1073/pnas.1412390111 PNAS | August 12, 2014 | vol. 111 | no. 32 | 11721–11726

CELL

BIOLO

GY

RAC activation (23) downstream of a wide variety of receptortyrosine kinases (24). In response to EGF, SOS1 interacts withactivated EGFR through the adaptor protein GRB2, leading tothe activation of RAS through the juxtaposition of SOS and RAS

at the membrane (25, 26). We now show that, in addition tothese well-known activities, SOS1 is required for the activation ofNFκB in response to EGF.

ResultsEGFR-Driven NFκB Activation in Human Mammary Epithelial Cells. Toinvestigate the pathway for EGF-dependent activation of NFκBwithout complications from a variety of genetic changes in dif-ferent cancer cell lines, we began by studying nonmalignanthuman mammary epithelial (HME) cells. When EGF-starvedHME cells were stimulated with EGF for different times (Fig.1A), robust phosphorylation of EGFR, IKK, and IκB and sub-stantial degradation and resynthesis of IκB were observed. Be-cause the IκB gene NFKBIA is a highly specific NFκB target,the resynthesis of the IκB protein makes it clear that NFκB-dependent gene expression is driven very well by EGF in thesecells. Because ERK and AKT are major downstream targets ofEGF, we also analyzed their phosphorylation, which was in-creased by EGF, as expected (Fig. 1A). Quantitative real-timePCR showed a significant increase in the mRNA levels of theNFκB-targeted genes IL8 and IL1B in response to EGF (Fig.1B). To determine whether the kinase activity of EGFR is re-quired for signaling to NFκB, we treated EGF-starved cells withthe potent and specific inhibitor erlotinib for 1 h and thenstimulated them with EGF for 5 min (Fig. 1C). As expected,erlotinib blocks the phosphorylation of AKT and ERK, and italso blocks the phosphorylation of IKK and IκB. These resultsreveal robust EGF- and EGFR-dependent activation of NFκBin nonmalignant human epithelial cells.

EGFR Mediates NFκB Activation in Several Cancer Cell Lines. NFκBis constitutively activated in many cancers (6, 7) and thus, asexpected, shRNA-mediated depletion of p65/RELA leads togreatly increased cell death in the nonsmall cell lung cancer

Fig. 1. EGFR-dependent activation of NFκB in HME cells. (A) The cells wereEGF starved for 24 h and then stimulated with 10 ng/mL EGF for the in-dicated times. Stimulation of the known EGF-activated pathways that acti-vate PI3K/ATK and MAPK, as well as the NFκB signaling components IKK andIκB, was detected with phospho-specific antibodies. Anti-IκB antibodydetected the degradation and resynthesis of this protein. (B) Analysis by real-time PCR of the activation of the IL8 and IL1B genes in response to EGF.(C) Effects of erlotinib on EGF-treated HME cells. The cells were deprived ofEGF for 24 h and then treated with EGF (100 ng/mL) for 10 min, with orwithout pretreatment with erlotinib (50 μM for 45 min). The levels ofphosphorylated and total proteins were analyzed by the Western method.The experiments above were repeated three times, with very similar results.

Fig. 2. EGFR mediates NFκB activation in cancer.(A and B) NFκB activation and cancer cell survival.(Upper) PC9 cells (A) and A549 cells (B) were infec-ted with empty vector or with a vector encodingNFκB shRNA and analyzed by the Western methodfor expression of p65 and β-actin. (Lower) Cell sur-vival assays were performed in control and p65knockdown cells. The cells were lysed with 0.5 MNaOH and the A260 was measured as an indicationof the total amount of nucleic acid. Means of threeexperiments are shown; each measurement wasperformed in triplicate; bars, SD. (C ) Erlotinibinhibits constitutive NFκB activation in cancer cells.Cells without EGF pretreatment were treated witherlotinib for different times, and the levels ofphosphorylated EGFR and IκB were determined bythe Western method. β-Actin was the loading con-trol. (D) Constitutive NFκB activity is not inhibited byerlotinib in some cancer cells. Cells without EGFpretreatment were treated with erlotinib and thelevels of phosphorylated and total proteins wereanalyzed by the Western method. (E) EGFR knock-down inhibits constitutive NFκB activation in cancercells. A549 and SKOV3 cells were infected witha vector encoding either a nontargeted (NT shRNA)or an EGFR shRNA and analyzed for total EGFRby the Western method. Phosphorylations of IKKand IκB were analyzed by the Western method.The experiments were repeated twice, with verysimilar results.

11722 | www.pnas.org/cgi/doi/10.1073/pnas.1412390111 De et al.

(NSCLC) cell lines PC9 and A549 (Fig. 2 A and B). As notedabove, several different mechanisms of NFκB activation in re-sponse to EGF have been reported in cell lines derived fromseveral different tumor types. To investigate the role of EGFR inconstitutive NFκB activation in cancer cells further, we treatedthe NSCLC cell lines A549 and HCC827, the ovarian cancer cellline OVCAR3, and the small cell lung cancer (SCLC) cell linesH1048 and SW1271 with erlotinib. Note that these cells were not

pretreated with EGF or any other NFκB activator. A549,OVCAR3, H1048, and SW1271 cells have wild-type EGFR (27–30), whereas HCC827 cells carry an activating EGFR mutation(31). As expected, erlotinib inhibits EGFR phosphorylation in allof these cells and, importantly, it also decreases greatly thephosphorylation of IκB, within 2–3 h (Fig. 2C). However, theactivation of NFκB was not inhibited by erlotinib in the breastcancer cell lines MDAMB468 and MDAMB436, or the coloncancer cell lines DLD1 and RKO (Fig. 2D). To confirm a role ofEGFR in activating NFκB in some cancer cells, we down-regu-lated its expression in A549 and SKOV3 cells with an EGFR-specific shRNA. Consistent with the effects of erlotinib, thephosphorylations of IKK and IκB were decreased in the EGFRknockdown cells (Fig. 2E). In summary, EGFR is an importantactivator of NFκB in some cancer cells, but in others alternativemechanisms must be responsible.

High Expression of SOS1 Increases Resistance to Erlotinib. To beginto explore the mechanisms of EGFR-mediated NFκB activationin cancer cells, we isolated and characterized an erlotinib-resistant cell line. Using validation-based insertional mutagenesis(VBIM), which employs lentiviral vectors (32), we inserted thestrong CMV promoter approximately randomly into the genomeof the erlotinib-sensitive NSCLC cell line PC9 (33). The cells weretreated with 4 μM erlotinib every 72 h. After 10 d, we isolatedthe erlotinib-resistant clone SD3.1, which was subsequentlyinfected with a lentiviral vector encoding CRE recombinase, toexcise the floxed CMV promoter (34). The cells were thentreated with erlotinib to reveal reversion of the phenotype,providing genetic proof that the insertion caused resistance tothis drug (Fig. 3A). To identify the mRNA responsible forresistance, we performed RNA-based cloning, using vectorsequences present in the bicistronic mRNA expressed from theinserted promoter (32). The mRNA we identified encodesa major portion of the SOS1 protein. The full coding sequence ofSOS1 translates to 1,333 amino acids, of which 355 residues ofthe N terminus were lost in the truncated protein driven by theinserted promoter. However, the portion of SOS1 remainingretains the catalytic domain (23). Using a 5′ VBIM-specificprimer and a 3′ gene-specific primer, analysis by RT-PCR showedoverexpression of the truncated SOS1 mRNA in erlotinib-resistant SD3.1 cells, and expression of CRE recombinaseeliminated this expression (Fig. 3B). To determine whether bothtruncated and full-length SOS1 can mediate erlotinib resistancein unmutagenized cells, cDNAs encoding both forms of theprotein were introduced into naive PC9 cells, resulting in stablepools of cells in which the two forms were expressed. Both poolswere significantly more resistant to erlotinib than were thecontrols (Fig. 3 C and D). Similar results were obtained whenfull-length SOS1 was overexpressed in another EGFR mutantNSCLC cell line, HCC827 (Fig. 3E). Therefore, increased ex-pression of SOS1 mediates erlotinib resistance. To determinewhether the catalytic activity of SOS1 is required, the catalyti-cally dead F929A mutant (35, 36) was stably expressed in PC9cells. The resistance of these cells was similar to cells in whichthe wild-type SOS1 was overexpressed (Fig. 3F), indicating that,surprisingly, the catalytic function of SOS1 is not required toconfer erlotinib resistance.

SOS1 Overexpression Increases NFκB Activation in Cancer Cells. Be-cause EGFR is an important activator of NFκB in cancer cells,we investigated whether increased activation of NFκB itself leadsto erlotinib resistance. We used the NSCLC cell line HCC827,which carries an activating EGFR mutation, making it verysensitive to erlotinib (31). Stimulation of these cells with IL-1βfor 4 h led to substantial increases both in NFκB activation (Fig.4A) and in resistance to erlotinib (Fig. 4B). Because NFκBactivation and SOS1 overexpression both mediate erlotinib

Fig. 3. SOS1 overexpression increases erlotinib resistance. (A and B) Iden-tification of SOS1 in an erlotinib-resistant clone. (A) Resistant SD3.1 cellswere infected with an empty lentiviral vector or with a vector encoding CRErecombinase. The cells were plated into six-well plates at 200,000 cells perwell. The next day, the cells were treated with erlotinib (4 or 8 μM) andstained with methylene blue 72 h later. (B) Truncated SOS1 mRNA levelswere analyzed by RT-PCR, using 5′ VBIM-specific and 3′ gene-specific primersin parental, SD3.1 (mutant), or SD3.1CRE cells. GAPDH was used as a loadingcontrol. (C and D, Upper) PC9 cells were infected with empty vector (Vec) ora vector encoding truncated or full-length SOS1 and expression was exam-ined by the Western method. (Lower) Control and SOS1-expressing cellswere plated in triplicate at 200,000 cells per well in six-well plates andallowed to attach overnight. The cells were treated with erlotinib for 72 hand lysed with 0.5 M NaOH, and the A260 was measured. (E) The level ofSOS1 in HCC827 cells over expressing the full-length protein was determinedby the Western method (Upper). The cells were treated with erlotinib andcell survival was determined after 72 h (Lower). (F) The level of SOS1 inHCC827 cells expressing the empty vector (Vec), wild-type SOS1 (WT SOS1),or mutant F929A SOS1 (FA SOS1) protein was determined by the Westernmethod (Upper). The cells were treated with erlotinib and cell survival wasdetermined after 72 h (Lower). Means of three experiments are shown; eachmeasurement was performed in triplicate; bars, SD. *P < 0.05; **P < 0.005.

De et al. PNAS | August 12, 2014 | vol. 111 | no. 32 | 11723

CELL

BIOLO

GY

resistance, we investigated whether SOS1 regulates NFκB. WhenPC9 and HCC827 cells with integrated NFκB-dependent lucif-erase reporters were transduced with a vector expressing SOS1to generate stable pools of cells overexpressing this protein, weobserved ∼3.5- to 4-fold increases in reporter activity (Fig. 4 Cand D). To show that NFκB activation by SOS1 is not limited toNSCLC cell lines, the protein was overexpressed in several ad-ditional cancer cell lines (SKOV3, H1048, and HCT116) andNFκB activation was analyzed (Fig. 4E). SOS1 overexpressionincreased NFκB activation by ∼3-fold in SKOV3 and H1048cells, and by ∼1.6-fold in HCT116 cells. Analysis by real-timePCR revealed that the levels of mRNAs expressed from theNFκB target genes IL1B and IL8 increased in PC9 cells in whichSOS1 was overexpressed (Fig. 4F). The role of SOS1 in NFκBactivation was further analyzed by downregulating its expressionin A549 cells, leading to substantial decreases in the constitutivephosphorylation of IKK and IκB (Fig. 4G). Decreased NFκBactivation upon SOS1 depletion was also seen in a reporter assay(Fig. 4H). In investigating whether the catalytic function of SOS1is required for activating NFκB, we found that overexpression inPC9 and H1048 cells of the catalytically dead F929A mutant ofSOS1 increased NFκB activation comparably to overexpressionof wild-type SOS1 (Fig. 5A). An increased level of IL8 mRNAwas also observed in H1048 cells overexpressing F929A SOS1(Fig. 5B). These results indicate that the catalytic activity of SOS1is not necessary for NFκB activation. As expected, the phos-phorylation of ERK or AKT in HCC827 cells was abolished by theMEK inhibitor PD0325901 or the PI3K inhibitor GDC0941 (Fig.S1 A and B). However, there was only slight inhibition of SOS1-dependent NFκB activation by PD0325901 or GDC0941 (Fig. S1A and B). Similarly, a RAF inhibitor GW5074 also did notabolish SOS1-dependent NFκB activation (Fig. S1C). We con-clude that NFκB activation by SOS1 is not primarily dependenton the canonical pathway through which SOS1 promotesERK activation.

SOS1 Mediates EGF-Dependent NFκB Activation. SOS1 is well knownto be involved in EGF-dependent signaling pathways that facil-itate cell growth and survival. Because EGFR plays an importantrole in NFκB activation in cancer cells and because SOS1

overexpression increases NFκB activation, it was logical to in-vestigate whether SOS1 is on the pathway of EGFR-dependentNFκB activation. Ablation of SOS1 in A549 cells did impairEGF-induced activation of an NFκB reporter gene (Fig. 6A).Note that these cancer cells already have high constitutive levelsof NFκB activation, limiting the ability of EGF to drive a furtherincrease. To extend this result, we stimulated with EGF A549and SKOV3 cells in which SOS1 expression was down-regulatedand analyzed NFκB-dependent signaling. In cells with little SOS1expression the basal levels of IKK and IκB phosphorylationwere decreased, and EGF-dependent increases in the phosphory-lation of IKK and IκB were not observed. The degradation andresynthesis of total IκB was also not observed upon EGF stim-ulation when SOS1 expression was ablated (Fig. 6B and Fig. S2).

Fig. 4. Overexpression of SOS1 increases NFκBactivation in cancer cells. (A and B) Increased acti-vation of NFκB causes erlotinib resistance. (A) HCC827cells were stimulated with IL-1β for 4 h, and NFκBactivation was determined in a reporter assay. Theresults of triplicate luciferase assays are shown asmeans ± SD **P < 0.005. (B) The cells were plated intriplicate at 200,000 cells per well in six-well platesand allowed to attach overnight. They were thenstimulated with IL-1β for 4 h or left unstimulated,then treated with erlotinib for 72 h and lysed with0.5 M NaOH. The A260 was then determined, asa measure of the total amount of nucleic acid. Thefraction of surviving cells was determined by nor-malizing the data from erlotinib-treated cells to un-treated controls. Means of three experiments areshown; each measurement was performed in tripli-cate; bars, SD. **P < 0.005. (C–E) Stable pools of PC9(C); HCC827 (D); and SKOV3, H1048, and HCT116(E) cells expressing an NFκB luciferase reporter wereinfected with a lentivirus encoding SOS1 or with theempty vector (Vec). SOS1 expression was analyzed bythe Western method. NFκB activity was deter-mined by using a luciferase assay. The results oftriplicate luciferase assays are shown as means ± SD**P < 0.005. (F) Analysis by real-time PCR of the activation of the IL1B and IL8 genes in PC9 cells overexpressing SOS1. Means of two experiments are shown;each measurement was performed in triplicate; bars, SD. (G and H) Decreased NFκB activation in SOS1 knockdown cells. The levels of phosphorylated IKK andIκB were detected by the Western method (G) and NFκB activity was measured in a luciferase assay (H). The results of triplicate luciferase assays are shown asmeans ± SD.

Fig. 5. The catalytic activity of SOS1 is not required for NFκB activation incancer cells. (A and B) Stable pools of PC9 and H1048 cells expressing an NFκBluciferase reporter were infected with a lentivirus encoding wild-type SOS1(WT SOS1), F929A SOS1 (FA SOS1), or empty vector (Vec). SOS1 expressionwas analyzed by the Western method. NFκB activity was measured in a lucif-erase assay. The results of triplicate luciferase assays are shown as means ± SD.(B) Analysis by real-time PCR of the activation of the IL8 gene in H1048 cellsexpressing WT SOS1 and F929A SOS1. Means of two experiments are shown;each measurement was performed in triplicate; bars, SD.

11724 | www.pnas.org/cgi/doi/10.1073/pnas.1412390111 De et al.

As expected, SOS1 down-regulation decreased the basal levelsof phosphorylated ERKs and also inhibited EGF-stimulatedERK phosphorylation (Fig. 6B and Fig. S2). These results in-dicate that SOS1 is involved in the pathway of EGFR-mediatedNFκB activation.

DiscussionThe EGF/EGFR Pathway Is Responsible for NFκB Activation in SomeCancers. NFκB can be activated by EGF in nonmalignant cells,including the human embryonic kidney cell line 293 (7) andhuman kidney proximal tubule cells (37). We have now foundthat EGFR also mediates NFκB activation in nonmalignant hu-man mammary epithelial cells and that this activation is inhibitedby the EGFR tyrosine kinase inhibitor erlotinib. Whereas NFκBactivation is tightly regulated as a component of the normal in-flammatory response, deregulated constitutive activation of NFκBis a hallmark of most cancers, where it contributes to resistance toapoptosis, proliferation, and the propensity to metastasize (1).Therefore, understanding the causes of constitutive NFκB acti-vation in cancer is an important issue.EGFR is commonly overexpressed or constitutively activated

in cancer cells and contributes to their uncontrolled proliferationand survival (38). Although treatment with EGF is known toactivate NFκB (39), the basis of constitutive NFκB activation incancer is complex (6) and the contribution of EGFR to NFκBactivation in the absence of treatment with exogenous EGF isunknown. Many different specific mutations provide a variety ofopportunities to connect EGFR to NFκB in different cancercells. We have found that treatment with erlotinib or down-regulation of EGFR expression inhibits NFκB activation inseveral different types of cancer cells in tissue culture, indicatingthat this pathway is likely to be responsible for constitutive NFκBactivation in some cancers. However, additional mechanisms aresure to be responsible for constitutive NFκB activation in cancer,because treatment with erlotinib had no effect on NFκB acti-vation in other cancer cells, and indeed some of these cellsshowed little or no constitutive phosphorylation of EGFR.Consistent with our results, it has been reported recently that thedual tyrosine kinase inhibitor lapatinib, which interrupts both the

HER2/neu and EGFR pathways, inhibits NFκB activation inbreast cancer cells overexpressing HER2 (40).Constitutive NFκB activation contributes importantly to re-

sistance to therapies in many cancers (1, 41, 42), and we showhere that increasing NFκB activation by treating cells with IL-1βgreatly increased resistance to erlotinib. We conclude that thereis an important connection between EGFR and NFκB activationin cancer cells. Patients treated with erlotinib may receive ther-apeutic benefit not only from the ability of this drug to suppressthe growth stimulatory activation of the RAS/ERK pathway, butalso from its ability to inhibit NFκB activation. Recently it hasbeen reported that inhibiting NFκB sensitizes NSCLC cellsto erlotinib-induced cell death (42) and, similarly, quinacrine,which inhibits NFκB (43, 44), has been combined with erlotinibin a phase I/II clinical trial to test the synergistic effect of this-combination in patients with advanced or metastatic NSCLC(NCT01839955).

Mechanistic Insight. Previously, we used the VBIM method ofinsertional mutagenesis to show that overexpression of kinesinsmediates resistance to docetaxel (34) and that overexpression ofFER confers resistance to quinacrine (22). We have now usedthis method to isolate an erlotinib-resistant cell line in whichSOS1 is overexpressed. SOS1 is well known to participate inEGF-dependent signaling pathways that facilitate cell growthand survival (26, 45). SOS1 is a guanine nucleotide exchangefactor that promotes RAS and RAC activation downstream ofEGFR and other growth factor receptors. Stimulation of cellswith growth factors leads to the association of SOS–GRB2complexes with the activated receptors, leading to the activationof RAS through the juxtaposition of SOS and RAS at themembrane (25). Growth factor-induced phosphorylation of ser-ine/threonine residues of SOS1, mediated by MAP kinases, altersits association with GRB2 and inhibits its function, providinga negative feedback mechanism (45–47). SOS1 has been repor-ted to be involved in several different cancers including breastcancer, hematological malignancies, and skin cancer (23). Am-plification of GRB2 and SOS1 has been reported in bladdercancer (48), and overexpression of SOS1 has been seen inprostate (49) and kidney cancer cells (50). The requirement ofEGFR for SOS1-dependent skin tumor development has beenshown in a transgenic mouse model (51). The ability of increasedexpression of SOS1 to confer resistance to erlotinib in cancercells prompted us to explore its role in the EGFR–NFκB path-way. Suppression of SOS1 inhibits EGF–induced NFκB activa-tion, indicating that SOS1 is indeed involved in this pathway. Wehave also shown that overexpression of SOS1 increases NFκBactivation dramatically in cancer cells, and that constitutiveNFκB activation is impaired by SOS1 depletion. These resultsunderscore the importance of SOS1 in cancer cell survival, whichis well known to be mediated by up-regulation of NFκB activa-tion. To our knowledge ours is the first report showing that SOS1is an activator of NFκB. The F929A mutation of SOS1 abrogatesits ability to catalyze guanine nucleotide exchange (35, 36) butnot its ability to activate NFκB. We demonstrate that F929Amutant SOS1 increases erlotinib resistance and NFκB activationwith an efficiency similar to that of wild-type SOS1. Consistently,SOS1-dependent activation of NFκB is not inhibited by a MEKinhibitor, a PI3K inhibitor, or a RAF inhibitor (Fig. S1 A–C).These results indicate that SOS1-mediated EGFR–NFκB acti-vation is independent of the catalytic activity of SOS1 and thata currently unknown pathway connects SOS1 to NFκB. It is in-teresting and relevant that a recent study has shown that SOS1plays an adaptor role in RAC and p38 activation in which itscatalytic activities are also dispensable (36).

Fig. 6. SOS1 mediates EGF-dependent NFκB activation: (A, Upper) A549cells were infected with vector only or SOS1 shRNA and immunoblotted forSOS1 and β-actin. (Lower) Cells were stimulated with EGF (100 ng/mL) for 5 hor left unstimulated, and NFκB-dependent luciferase activity was measured.The cell lysates were normalized for protein concentration. (B, Upper) Thecells were serum starved for 24 h and then stimulated with 100 ng/mL EGFfor the indicated times. Phosphorylated EGFR, ERK, IKK, and IκB weredetected with phospho-specific antibodies. Anti-IκB antibody detected thedegradation and resynthesis of this protein. β-Actin was used as a loadingcontrol. (Lower) Densitometric analyses are determined by the ratio of ex-pression of phosphorylated IKK or IκB and β-actin, measured by using NIHsoftware (ImageJ 1.48v). The experiments above were repeated twice, withvery similar results.

De et al. PNAS | August 12, 2014 | vol. 111 | no. 32 | 11725

CELL

BIOLO

GY

Materials and MethodsThe hTERT-HME1 cell line was purchased from Clontech. The cancer cell linesHCC827, A549, HCT116, H1048, and SW1271 were obtained from AmericanTissue Culture Collection. VBIM vector constructs were described previously (32,34). Cell survival assays were done as described previously (34). Luciferase

assays were done by using the NFκB reporter construct. Detailed materials andmethods are provided in SI Materials and Methods.

ACKNOWLEDGMENTS. We thank Maojing Yang for excellent technicalassistance. This work was supported by National Institutes of Health GrantP01CA062220.

1. Baldwin AS (2001) Control of oncogenesis and cancer therapy resistance by thetranscription factor NF-kappaB. J Clin Invest 107(3):241–246.

2. Perkins ND (2012) The diverse and complex roles of NF-κB subunits in cancer. Nat RevCancer 12(2):121–132.

3. Hayden MS, Ghosh S (2008) Shared principles in NF-kappaB signaling. Cell 132(3):344–362.

4. Perkins ND, Gilmore TD (2006) Good cop, bad cop: The different faces of NF-kappaB.Cell Death Differ 13(5):759–772.

5. Baltimore D (2011) NF-κB is 25. Nat Immunol 12(8):683–685.6. Lu T, Sathe SS, Swiatkowski SM, Hampole CV, Stark GR (2004) Secretion of cytokines

and growth factors as a general cause of constitutive NFkappaB activation in cancer.Oncogene 23(12):2138–2145.

7. Sethi G, Ahn KS, Chaturvedi MM, Aggarwal BB (2007) Epidermal growth factor (EGF)activates nuclear factor-kappaB through IkappaBalpha kinase-independent but EGFreceptor-kinase dependent tyrosine 42 phosphorylation of IkappaBalpha. Oncogene26(52):7324–7332.

8. Yu HG, et al. (2003) Increased expression of RelA/nuclear factor-kappa B proteincorrelates with colorectal tumorigenesis. Oncology 65(1):37–45.

9. Tang X, et al. (2006) Nuclear factor-kappaB (NF-kappaB) is frequently expressed inlung cancer and preneoplastic lesions. Cancer 107(11):2637–2646.

10. Ladanyi M, Pao W (2008) Lung adenocarcinoma: Guiding EGFR-targeted therapy andbeyond. Mod Pathol 21(Suppl 2):S16–S22.

11. Holbro T, Civenni G, Hynes NE (2003) The ErbB receptors and their role in cancerprogression. Exp Cell Res 284(1):99–110.

12. Wykosky J, Fenton T, Furnari F, Cavenee WK (2011) Therapeutic targeting of epi-dermal growth factor receptor in human cancer: Successes and limitations. Chin JCancer 30(1):5–12.

13. Yang W, et al. (2011) Nuclear PKM2 regulates β-catenin transactivation upon EGFRactivation. Nature 480(7375):118–122.

14. Hammerman PS, Jänne PA, Johnson BE (2009) Resistance to epidermal growth factorreceptor tyrosine kinase inhibitors in non-small cell lung cancer. Clin Cancer Res15(24):7502–7509.

15. Brown M, Cohen J, Arun P, Chen Z, Van Waes C (2008) NF-kappaB in carcinomatherapy and prevention. Expert Opin Ther Targets 12(9):1109–1122.

16. Yang W, et al. (2012) EGFR-induced and PKCe monoubiquitylation-dependent NF-κBactivation upregulates PKM2 expression and promotes tumorigenesis. Mol Cell 48(5):771–784.

17. Habib AA, et al. (2001) The epidermal growth factor receptor engages receptor in-teracting protein and nuclear factor-kappa B (NF-kappa B)-inducing kinase to activateNF-kappa B. Identification of a novel receptor-tyrosine kinase signalosome. J BiolChem 276(12):8865–8874.

18. Chen D, et al. (2003) NIK is a component of the EGF/heregulin receptor signalingcomplexes. Oncogene 22(28):4348–4355.

19. Kapoor GS, Zhan Y, Johnson GR, O’Rourke DM (2004) Distinct domains in the SHP-2phosphatase differentially regulate epidermal growth factor receptor/NF-kappaBactivation through Gab1 in glioblastoma cells. Mol Cell Biol 24(2):823–836.

20. Tanaka K, et al. (2011) Oncogenic EGFR signaling activates an mTORC2-NF-κB path-way that promotes chemotherapy resistance. Cancer Discov 1(6):524–538.

21. Jiang T, et al. (2011) CARMA3 is crucial for EGFR-Induced activation of NF-κB andtumor progression. Cancer Res 71(6):2183–2192.

22. Guo C, Stark GR (2011) FER tyrosine kinase (FER) overexpression mediates resistanceto quinacrine through EGF-dependent activation of NF-kappaB. Proc Natl Acad SciUSA 108(19):7968–7973.

23. Pierre S, Bats AS, Coumoul X (2011) Understanding SOS (Son of Sevenless). BiochemPharmacol 82(9):1049–1056.

24. Buday L, Downward J (2008) Many faces of Ras activation. Biochim Biophys Acta1786(2):178–187.

25. Jorge R, et al. (2002) HSos1 contains a new amino-terminal regulatory motif withspecific binding affinity for its pleckstrin homology domain. J Biol Chem 277(46):44171–44179.

26. Zhao C, Du G, Skowronek K, Frohman MA, Bar-Sagi D (2007) Phospholipase D2-generated phosphatidic acid couples EGFR stimulation to Ras activation by Sos. NatCell Biol 9(6):706–712.

27. Tracy S, et al. (2004) Gefitinib induces apoptosis in the EGFRL858R non-small-cell lungcancer cell line H3255. Cancer Res 64(20):7241–7244.

28. Bianco C, et al. (2000) Antitumor activity of combined treatment of human cancercells with ionizing radiation and anti-epidermal growth factor receptor monoclonalantibody C225 plus type I protein kinase A antisense oligonucleotide. Clin Cancer Res6(11):4343–4350.

29. Ogino H, et al. (2011) E7080 suppresses hematogenous multiple organ metastases oflung cancer cells with nonmutated epidermal growth factor receptor. Mol CancerTher 10(7):1218–1228.

30. Ohashi K, et al. (2013) Characteristics of lung cancers harboring NRAS mutations. ClinCancer Res 19(9):2584–2591.

31. Pao W, et al. (2005) Acquired resistance of lung adenocarcinomas to gefitinib or er-lotinib is associated with a second mutation in the EGFR kinase domain. PLoS Med2(3):e73.

32. Lu T, et al. (2009) Validation-based insertional mutagenesis identifies lysine deme-thylase FBXL11 as a negative regulator of NFkappaB. Proc Natl Acad Sci USA 106(38):16339–16344.

33. Okabe T, et al. (2007) Differential constitutive activation of the epidermal growthfactor receptor in non-small cell lung cancer cells bearing EGFR gene mutation andamplification. Cancer Res 67(5):2046–2053.

34. De S, Cipriano R, Jackson MW, Stark GR (2009) Overexpression of kinesins mediatesdocetaxel resistance in breast cancer cells. Cancer Res 69(20):8035–8042.

35. Jeng HH, Taylor LJ, Bar-Sagi D (2012) Sos-mediated cross-activation of wild-type Rasby oncogenic Ras is essential for tumorigenesis. Nat Commun 3:1168.

36. Jun JE, Yang M, Chen H, Chakraborty AK, Roose JP (2013) Activation of extracellularsignal-regulated kinase but not of p38 mitogen-activated protein kinase pathways inlymphocytes requires allosteric activation of SOS. Mol Cell Biol 33(12):2470–2484.

37. Häussler U, von Wichert G, Schmid RM, Keller F, Schneider G (2005) Epidermal growthfactor activates nuclear factor-kappaB in human proximal tubule cells. Am J PhysiolRenal Physiol 289(4):F808–F815.

38. Zhang Z, Stiegler AL, Boggon TJ, Kobayashi S, Halmos B (2010) EGFR-mutated lungcancer: A paradigm of molecular oncology. Oncotarget 1(7):497–514.

39. Alberti C, et al. (2012) Ligand-dependent EGFR activation induces the co-expression ofIL-6 and PAI-1 via the NFkB pathway in advanced-stage epithelial ovarian cancer.Oncogene 31(37):4139–4149.

40. Ma C, et al. (2013) Lapatinib inhibits the activation of NF-κB through reducingphosphorylation of IκB-α in breast cancer cells. Oncol Rep 29(2):812–818.

41. Li F, Sethi G (2010) Targeting transcription factor NF-kappaB to overcome chemo-resistance and radioresistance in cancer therapy. Biochim Biophys Acta 1805(2):167–180.

42. Bivona TG, et al. (2011) FAS and NF-κB signalling modulate dependence of lungcancers on mutant EGFR. Nature 471(7339):523–526.

43. Gurova KV, et al. (2005) Small molecules that reactivate p53 in renal cell carcinomareveal a NF-kappaB-dependent mechanism of p53 suppression in tumors. Proc NatlAcad Sci USA 102(48):17448–17453.

44. Jani TS, DeVecchio J, Mazumdar T, Agyeman A, Houghton JA (2010) Inhibition ofNF-kappaB signaling by quinacrine is cytotoxic to human colon carcinoma cell lines andis synergistic in combination with tumor necrosis factor-related apoptosis-inducingligand (TRAIL) or oxaliplatin. J Biol Chem 285(25):19162–19172.

45. Douville E, Downward J (1997) EGF induced SOS phosphorylation in PC12 cells in-volves P90 RSK-2. Oncogene 15(4):373–383.

46. Buday L, Warne PH, Downward J (1995) Downregulation of the Ras activationpathway by MAP kinase phosphorylation of Sos. Oncogene 11(7):1327–1331.

47. Orton RJ, Sturm OE, Gormand A, Wolch W, Gilbert DR (2008) Computational mod-elling reveals feedback redundancy within the epidermal growth factor receptor/extracellular-signal regulated kinase signalling pathway. IET Syst Biol 2(4):173–183.

48. Watanabe T, et al. (2000) Significance of the Grb2 and son of sevenless (Sos) proteinsin human bladder cancer cell lines. IUBMB Life 49(4):317–320.

49. Timofeeva OA, et al. (2009) Enhanced expression of SOS1 is detected in prostatecancer epithelial cells from African-American men. Int J Oncol 35(4):751–760.

50. Guerrero C, et al. (1996) Expression of alternative forms of Ras exchange factors GRFand SOS1 in different human tissues and cell lines. Oncogene 12(5):1097–1107.

51. Sibilia M, et al. (2000) The EGF receptor provides an essential survival signal for SOS-dependent skin tumor development. Cell 102(2):211–220.

11726 | www.pnas.org/cgi/doi/10.1073/pnas.1412390111 De et al.