Research ArticleCancer

The ERBB network facilitates KRAS-driven lung tumorigenesis

See allHide authors and affiliations

Science Translational Medicine  20 Jun 2018:
Vol. 10, Issue 446, eaao2565
DOI: 10.1126/scitranslmed.aao2565

A new role for kinase inhibitors

The KRAS oncogene is frequently mutated in a variety of cancer types, including lung cancer. Lung cancers with KRAS mutations are usually difficult to target, and conventional thinking dictates that these tumors are resistant to receptor tyrosine kinase inhibitors because those act upstream of the constitutively active KRAS protein. However, it appears that receptor tyrosine kinase signaling may have an effect on KRAS-driven lung tumors after all, by amplifying their growth beyond the effects of KRAS alone. Kruspig et al. and Moll et al. independently reached this conclusion and identified approved multi-kinase inhibitors that are effective in the setting of KRAS-mutant lung cancer in multiple mouse models, suggesting that this may be a potential treatment strategy for human patients as well.


KRAS is the most frequently mutated driver oncogene in human adenocarcinoma of the lung. There are presently no clinically proven strategies for treatment of KRAS-driven lung cancer. Activating mutations in KRAS are thought to confer independence from upstream signaling; however, recent data suggest that this independence may not be absolute. We show that initiation and progression of KRAS-driven lung tumors require input from ERBB family receptor tyrosine kinases (RTKs): Multiple ERBB RTKs are expressed and active from the earliest stages of KRAS-driven lung tumor development, and treatment with a multi-ERBB inhibitor suppresses formation of KRASG12D-driven lung tumors. We present evidence that ERBB activity amplifies signaling through the core RAS pathway, supporting proliferation of KRAS-mutant tumor cells in culture and progression to invasive disease in vivo. Brief pharmacological inhibition of the ERBB network enhances the therapeutic benefit of MEK (mitogen-activated protein kinase kinase) inhibition in an autochthonous tumor setting. Our data suggest that lung cancer patients with KRAS-driven disease may benefit from inclusion of multi-ERBB inhibitors in rationally designed treatment strategies.


Cancers of the lung account for more than 1.5 million deaths per annum worldwide, and 5-year survival rates remain between 10 and 15% in many developed countries (1). The majority of lung cancers are classified as non–small cell (NSCLC), and adenocarcinoma is the most common histological subtype of NSCLC. Activating mutations in KRAS occur in a third of lung adenocarcinoma (LuAd) cases (2). RAS proteins have historically proven to be elusive targets for selective inhibition, although the recent development of G12 mutant KRAS-selective tool compounds suggests that therapeutic targeting of KRAS may, in time, be possible (3, 4). In the interim, there is a pressing need to develop alternative strategies for more effective treatment of KRAS-driven disease.

The ERBB family of receptor tyrosine kinases (RTKs) is composed of four members—EGFR (ERBB1), HER2 (ERBB2, NEU), ERBB3, and ERBB4—all of which can homo- or heterodimerize, and dimerization is required for signaling activity. ERBB dimers are activated upon binding a spectrum of soluble ligands including EGF, epiregulin (EREG), amphiregulin (AREG), and neuregulin (NRG), together forming a network for ERBB-driven signal transduction (5). EGFR is a well-recognized driver of LuAd, with genetic alterations present in up to 18% of cases (2). ERBB2 and ERBB3 are highly expressed in embryonic lungs of humans and rodents, and expression of both persists into adulthood (6, 7). Overexpression of ERBB2 in the absence of gene amplification is common in human LuAd (8, 9), and functionality of ERBB2/ERBB3 heterodimers in NSCLC-derived cell lines was previously shown (10). Amplification of any of the four ERBB RTKs is associated with poor prognosis in lung cancer (11), and high expression of the promiscuous ERBB ligand EREG has previously been linked to disease progression and aggressive phenotypes in models of EGFR- and KRAS-driven lung cancer (12, 13).

In a wild-type setting, ligand-activated signaling through ERBB RTKs activates KRAS (14). Mutation of KRAS is generally thought to confer independence from upstream regulation, a view that is reinforced by the mutual exclusivity of activating mutations in KRAS and EGFR in LuAd, and by the failure of EGFR-selective inhibitors to show therapeutic benefit against KRAS-driven cancers (15, 16). However, several recent results suggest that the independence of mutant KRAS from upstream signaling may not be absolute: In KRAS-mutant NSCLC cell lines, activation of phosphatidylinositol 3-kinase (PI3K) is contingent upon basal activity of wild-type insulin-like growth factor receptor (IGFR), establishing an important precedent for coordination of oncogenic and normal signal transduction (17); genetic deletion of EGFR was shown to suppress development of KRASG12D-driven pancreatic ductal adenocarcinoma (18, 19); induced expression of ERBB2 and ERBB3 was found to underlie resistance of KRAS-mutant lung and colorectal cell lines to MEK (mitogen-activated protein kinase kinase) inhibition (20). Strikingly, in the latter study, MEK inhibitor–induced ERBB2/3 expression was associated with recovery of extracellular signal–regulated kinase (ERK) phosphorylation downstream of KRAS, suggesting a surprising role for upstream signaling in sustaining pathway activity despite the presence of activated KRAS.

We therefore examined the requirement for ERBB activity in an inducible model of progressive autochthonous LuAd, driven by the combination of endogenously expressed KRASG12D and modest overexpression of c-MYC. We present evidence that redundant signal transduction through multiple ERBB RTKs supports development and progression of mutant KRAS-driven lung tumors. Our data suggest that frontline use of multi-ERBB inhibitors may show clinical benefit in KRAS-driven LuAd.


ERBB activity is required for KRasG12D-driven lung tumor formation

Induced expression of ERBB family RTKs is associated with resistance of KRAS-mutant NSCLC cell lines to MEK inhibition (20). We therefore examined expression of ERBB RTKs and their ligands in microdissected early-stage lung tumors, using a CRE-inducible mouse model of autochthonous LuAd driven by KRASG12D combined with modestly increased MYC (lsl-KRasG12D;Rosa26-lsl-MYC—henceforth KM), the latter expressed from the Rosa26 locus in amounts that alone fail to provoke a phenotype (fig. S1A). In tumor samples harvested 6 weeks after allele activation, we found strong expression of Erbb2 and Erbb3 mRNA, whereas Egfr was weakly expressed, and Erbb4 was not detected in tumors from two of four KM mice (Fig. 1A). Multiple ERBB ligands were expressed, with Areg, Tgfα, and Hbegf showing strongest expression, whereas Egf, Ereg, Nrg3, and Nrg4 were also clearly detected (Fig. 1B). The presence of both RTKs and multiple cognate ligands suggested that ERBB RTKs may actively signal in developing KRASG12D-driven lung tumors. Lysates prepared directly from multiple individual tumors, harvested 6 weeks after induction and immunoblotted for phospho-ERBB RTKs, consistently showed readily detectable phospho-EGFR, phospho-ERBB2, and phospho-ERBB3, suggesting that these RTKs are active in developing KRASG12D-driven lung tumors (Fig. 1C and fig. S1B). To determine whether ERBB signaling functionally contributes to tumor development, we treated tumor-bearing mice acutely with a multi-ERBB inhibitor, neratinib (21). Treatment of mice for 3 days with neratinib suppressed activity of ERBB RTKs, reduced tumor cell proliferation, and increased apoptosis, suggesting that ERBB activity promotes growth of KRASG12D-driven tumors (Fig. 1, D to F). Continuous daily treatment of mice from 2 weeks after allele induction almost completely suppressed the emergence of tumors, indicating that the combination of endogenously expressed KRASG12D and constitutively expressed MYC requires input from ERBB RTKs to give rise to lung tumors (Fig. 1, G and H). In contrast, daily erlotinib treatment failed to show the same effect, consistent with reports that EGFR inhibition in isolation is ineffective in KRAS-driven lung cancer (15, 18).

Fig. 1 ERBB activity is required for KRAS-driven lung tumor formation.

(A) Expression of ERBB family RTKs in KM lung tumors harvested 6 weeks post-allele induction (PI), measured by RNA-sequencing (RNA-seq). Mean ± SD read counts in tumors from four mice are shown. (B) Expression of ERBB family ligands in KM lung tumors harvested 6 weeks PI, as per (A). (C) Immunoblots of lysates generated from 10 individual KM tumors, harvested 6 weeks PI, using the indicated antibodies. (D) Representative images of immunohistochemistry (IHC) for Ki67 and terminal deoxynucleotidyl transferase–mediated deoxyuridine triphosphate nick end labeling (TUNEL) staining of KM mice treated for 3 days with neratinib (80 mg/kg) (n = 3) or vehicle control (n = 3). Scale bars, 100 μm. (E) Quantification (mean ± SEM) of staining in five tumors from each mouse as per (D). Left panel shows % of tumor cells expressing Ki67; right panel shows % of TUNEL-positive tumor cells; vc, vehicle control. P values are from two-tailed t tests. (F) Immunoblots of three individual KM tumors from mice treated for 3 days with neratinib or vehicle control. (G) Representative hematoxylin and eosin (H&E) images from KM mice treated daily with neratinib (2 × 40 mg/kg) or erlotinib (2 × 50 mg/kg), commencing 2 weeks PI, and harvested at 6 weeks PI. Scale bar, 1 mm. (H) Quantification of tumor burden from (G). Box and whisker plots show median, interquartile, and 99% range of tumor area, expressed as a percentage of total lung tissue area, measured across >25 sections from each mouse. n = 5 vehicle control (vc), n = 4 erlotinib, and n = 3 neratinib. Analysis of variance (ANOVA) followed by Tukey test; ns, not significant.

Progression of KRas-driven lung tumors is associated with increased ERBB network expression

The above analysis was performed on tumors at 5 to 6 weeks after adeno-CRE–mediated allele activation. At this time after induction, the lungs of all induced KM mice contained dozens of individual tumors that presented with uniform histology resembling human papillary adenocarcinoma in situ (Fig. 2A, left panels). A small minority of tumors (2 to 5%) also contained a second, more disorganized population with more aggressive histological features, including increased cytosolic and nuclear volume, prominent nucleoli, and increased morphological heterogeneity (Fig. 2A, right panels). IHC analysis revealed continuous expression of the type II pneumocyte marker SPC across both populations in individual tumors, suggesting that the second population represents the emergence of aggressive subclones (fig. S1C). Progression of KRASG12D-driven lung tumors to more aggressive disease is associated with a pronounced increase in ERK1/2 phosphorylation (22, 23). Accordingly, IHC analysis of phospho-ERK (p-ERK) revealed much higher expression of p-ERK in these disorganized subclones compared with the rest of the same tumors and across the entire tumor population (Fig. 2, A and B). By comparison, KM tumors harvested at 4 to 6 months after induction showed widespread expression of p-ERK, because these more aggressive subclones gradually came to predominate (Fig. 2B, bottom panel). Moreover, KM metastases to the liver also show high amounts of epithelial p-ERK staining (Fig. 2, C to F).

Fig. 2 KM lung tumor progression is associated with increased ERK phosphorylation.

(A) Images of H&E-stained (upper panels) and p-ERK–stained (lower panels) KM lung tumors harvested at 6 weeks PI illustrating histological changes associated with tumor progression: Left panels are representative of >95% of total tumor area at 6 weeks PI, whereas right panels represent 2 to 5% of total tumor area at 6 weeks PI. Scale bar, 50 μm. (B) p-ERK staining in KM tumors harvested at 6 weeks PI (top and middle panels) versus 5 months PI (bottom panel). Scale bars, 1 mm (top and bottom panels) and 200 μm (middle panel). (C) Detection of Hprt-lsl-IRFP expression in primary lung tumors (left) and a liver metastasis (right) in a KM mouse harvested 6 months PI. Scale bar, 5 mm. (D) Histological confirmation of liver metastasis stained by H&E. (E) IHC detection of SPC and p-ERK in the same metastasis as (D). Scale bars, 50 μm (D and E). Images are representative of six mice. (F) Halo quantification of p-ERK–positive cells in individual metastases, expressed as % of tumor cells.

Increased p-ERK was previously shown to be associated with amplification of the mutant KRas locus in a KP lung cancer model, wherein KRASG12D expression was combined with loss of functional p53 (24). We used laser-capture microdissection coupled with RNA-seq lysis to compare gene expression in p-ERKLow with p-ERKHigh KM tumor regions from four individual mice (see schematic, fig. S1D). Expression of KRas was modestly higher (less than twofold) in p-ERKHigh tumor regions, suggesting that locus amplification is not the underlying driver of progression in the KM model (Fig. 3A). However, we detected a sharp increase in expression of promiscuous ERBB family ligands, Ereg and Areg, along with more modest increases in Hbegf and Tgfα (Fig. 3, B to D, and fig. S1E). Increased expression of Ereg and Areg was previously reported to be associated with more aggressive tumors in an inducible model of KRAS-driven mammary cancer (25) and in human NSCLC cells enriched for metastatic behavior (26). In situ hybridization (ISH) revealed a clear overlap between expression of Ereg, Areg, and extensive phosphorylation of ERK (by IHC) in tumor epithelium, confirming the RNA-seq data (Fig. 3C and fig. S2, A and B). The same phenotype of focal progression to p-ERKhigh accompanied by markedly increased Areg expression was also observed in primary lung tumors upon induction with a more lineage-restricted adeno-SPC-CRE (fig. S2, A and B), and in liver metastases after induction with the standard adeno-CMV-CRE (Fig. 3D and fig. S2C), arguing against the possibility that this phenotype might arise from inadvertent activation of the driver oncogenes in other lung-resident populations (27). In addition, RNA-seq analysis revealed increased expression of Adam9 and Adam10 sheddases, which process membrane-bound ERBB ligands into their more potent soluble forms (28); increased expression of the signaling scaffold Iqgap2; and increased expression of LamC2, Fabp5, and Keratin19, all recently shown to enhance signaling through EGFR/ERBB family RTKs (26, 29, 30). Expression of Erbb2 and Erbb3 was not significantly changed in p-ERKHigh regions (Fig. 3, E and F).

Fig. 3 Expression of the ERBB network increases during progression from p-ERKlow to p-ERKhigh KM tumors.

(A) Normalized expression of RAS genes in laser-capture microdissected p-ERKHigh KM tumor regions relative to p-ERKLow regions from tumors in the same mice (n = 4 mice), measured by RNA-seq. False discovery rate (FDR) shown for KRAS. (B) Mean and SEM RNA-seq reads of Ereg and Areg mRNA from p-ERKLow and p-ERKHigh KM tumor regions from four mice. Adjusted P values were calculated in R. (C) Serial sections of KM tumors stained by IHC for p-ERK (left panels) or by ISH for Ereg (middle panels) or Areg (right panels). Scale bars, 200 μm (upper panels) and 25 μm (lower panels). (D) Representative images of p-ERK (IHC) and Areg (ISH) expression in KM liver metastasis. Scale bar, 200 μm. Images are representative of four metastases with sufficient material. (E) Normalized expression of ERBB network genes showing mean fold increase (Δ) in expression in p-ERKHigh relative to p-ERKLow KM tumor regions from four mice as per (A). (F) Diagrammatic representation of up-regulated components of the ERBB-RAS-ERK pathway associated with KM tumor progression.

These data suggest an alternative route to increased RAS pathway signaling that does not require KRas amplification but rather involves increased activity of the ERBB network, defined here as consisting of ERBB ligands, RTKs, and RTK accessory molecules such as LamC2 and Fabp5. Notably, examination of The Cancer Genome Atlas (TCGA) LuAd data set via cBioPortal (31) revealed overexpression and/or amplification of one or more constituents of the ERBB network in the majority of KRAS-mutant human LuAd (fig. S3).

ERBB signaling amplifies RAS pathway activity in KRAS-mutant human NSCLC cells

Combined inhibition of both EGFR and ERBB2 is required to prevent outgrowth of MEK inhibitor–resistant clones of human KRAS-mutant NSCLC cells; however, the effect of multi-ERBB inhibition on treatment-naïve cells was not previously explored (20). Treatment of multiple KRAS-mutant human NSCLC lines with the multi-ERBB inhibitor neratinib suppressed proliferation in a dose-dependent manner (Fig. 4A). In contrast, inhibition of EGFR in isolation showed little effect. Immunoblotting confirmed that the doses of neratinib used strongly suppressed activity of ERBB RTKs (Fig. 4B). Consistent with previous reports that neratinib drives increased ERBB turnover (32), ERBB2 and ERBB3 protein expression was also reduced by neratinib in multiple cell lines. Despite the presence of G12 mutant KRAS in all such cells, neratinib consistently reduced p-ERK, whereas effects on AKT and signal transducer and activator of transcription 3 (STAT3) varied across the cell lines tested (Fig. 4B and fig. S4A). We therefore examined RAF binding as a direct measure of RAS signaling activity in A549 cells, which are homozygous for mutant KRAS. Acute ERBB inhibition reduced both total RAS:RAF and specific KRAS:RAF binding by >50%, consistent with the observed partial reduction in p-ERK (Fig. 4C). In addition, murine lung tumor cells carrying a spontaneously amplified KRasG12D allele also showed sensitivity to neratinib, albeit somewhat lower than that of cells carrying a single copy of the allele (fig. S4B), and RAS-less mouse embryonic fibroblasts (MEFs) reconstituted with G12D mutant KRAS were likewise sensitive to neratinib, albeit less so than those reconstituted with wild-type KRAS (fig. S4, C and D). Together, these results suggest that ERBB amplifies signaling through the core RAS→ERK module, even in cells with activating mutation of KRAS, whereas sensitivity to ERBB suppression is imparted through both mutant and wild-type KRAS.

Fig. 4 A feed-forward ERBB signaling loop drives proliferation of KRAS-mutant human NSCLC cells.

(A) Growth curves of three KRAS-mutant human NSCLC lines upon treatment with increasing doses of the EGFR-selective inhibitor erlotinib or the multi-ERBB inhibitor neratinib, measured by IncuCyte time-lapse video microscopy. Error bars show SD for technical triplicates. Data are representative of at least two independent experiments. (B) Lysates from KRAS-mutant NSCLC cells treated with increasing doses of neratinib, immunoblotted with the indicated antibodies. Asterisks, where present, indicate the correct band. (C) RAS immunoblots of RAF-coated bead precipitates from lysates of A549 cells treated with neratinib or vehicle control for 2 hours. Lysate input aliquots were also immunoblotted with the indicated antibodies. Right panel shows mean ± SEM for quantification of KRAS band intensities from three independent experiments (arbitrary units). P values are from two-tailed t tests. (D) Top 20 significantly modulated pathways associated with the transition to p-ERKHigh disease in the KM model, identified using Metacore GeneGO analysis of RNA-seq expression data. Segment size in the pie chart (left panel) reflects ranking of the pathways by FDR. Right panels show that 18 of the same pathways are modulated in each of the three KRAS-mutant human NSCLC lines after overnight treatment with neratinib (250 nM A549 and H2009; 25 nM H358). Numbers and pie segment sizes reflect ranking by FDR. (E) Expression of ERBB ligands in the indicated cells treated overnight with vehicle (black) or neratinib (red), measured by RNA-seq as per (D). Mean and SEM of biological triplicates shown. *P < 0.05, **P < 0.01, ***P < 0.0001, two-tailed t tests. EMT, epithelial-mesenchymal transition; IL-4, interleukin 4; NGF, nerve growth factor; TGF, transforming growth factor.

To gain further mechanistic insight into the growth inhibitory effects of neratinib, we performed unbiased transcriptomic analysis after overnight treatment of three KRAS-mutant NSCLC cell lines. Metacore GeneGO pathway analysis revealed a remarkable degree of consistency across the three cell lines tested. Moreover, 18 of the 20 most highly modulated pathways associated with progression of murine KM tumors to p-ERKHigh disease were reciprocally modulated in the human cells upon neratinib treatment (Fig. 4D and table S1). In all three cell lines, ERBB inhibition reduced expression of the same ERBB ligands that increased as KM tumors progressed to p-ERKHigh disease (Fig. 4E). These findings indicate that ERBB activity establishes a feed-forward loop that sustains KRAS-mutant NSCLC proliferation in vitro and drives tumor progression in vivo at least in part by amplifying signaling through the core RAS-ERK module.

ERBB inhibition enhances the potency of MEK inhibition in vitro and extends survival of mice with LuAd

With the exception of H358 cells, ERBB inhibition alone did not result in death of KRAS-mutant NSCLC cells in vitro; however, neratinib increased apoptosis induced by inhibition of MEK downstream of KRAS in multiple cell lines (Fig. 5A). Substitution of neratinib with a second multi-ERBB inhibitor, afatinib, replicated the effects of neratinib, confirming the on-target specificity of the drug (fig. S5A). Moreover, both drugs combined with MEK inhibition to suppress colony formation (Fig. 5B). Consistent with published results (20), MEK inhibition alone increased expression of ERBB2 and ERBB3 in vitro, although the effect on ERBB2 phosphorylation was variable across the cell lines tested (Fig. 5C). Cotreatment with neratinib suppressed both expression and activity of ERBB2 and ERBB3 while also suppressing EGFR activity (Fig. 5C). Downstream of the RTKs, trametinib alone consistently suppressed expression of p-ERK and c-MYC, but increased phosphorylation of AKT and STAT3 (Fig. 5C). Suppression of increased AKT phosphorylation by neratinib was clear only in H358 cells, but the drug combination failed to reduce STAT3 phosphorylation. The growth inhibitory effects of the drug combination in vitro thus better correlate with suppression of the core RAS-ERK-MYC pathway than with the activity of ancillary signaling pathways.

Fig. 5 ERBB blockade enhances MEK inhibitor–driven apoptosis in vitro and therapeutic impact in vivo.

(A) Apoptosis induced in human NSCLC cells, measured 48 hours after treatment with the indicated doses of neratinib (nera) and/or trametinib (tram). Mean ± SEM of three independent experiments shown (ANOVA and Tukey test). (B) Clonogenic assay showing suppression of colony formation in A549 and H358 cells after 48 hours of treatment with the indicated inhibitors. Lower panels show quantification of colony area (% surface coverage) from five independent experiments. Significance was determined for drug combinations versus trametinib alone (ANOVA and Tukey test). (C) Lysates from the indicated cells treated for 24 hours with increasing doses of trametinib alone or the combination of trametinib and neratinib, immunoblotted with the indicated antibodies. Asterisks, where present, indicate the correct band. (D) Overall survival, measured from the first day of treatment, of tumor-bearing KM mice treated daily for 1 week (tan bar) with neratinib (80 mg/kg), trametinib (1 mg/kg), or the combination of both, then followed without further intervention. Treatment was commenced at 5 weeks PI. Cohorts shown are vehicle (n = 9), neratinib (n = 7), trametinib (n = 10), and trametinib + neratinib (n = 10). Log-rank hazard ratios (HR ± 95% confidence interval) and P values are shown for comparisons of trametinib + neratinib versus vehicle and trametinib + neratinib versus trametinib alone (dashed lines). (E) Lysates of individual tumors from mice treated with neratinib (80 mg/kg) and/or trametinib (1 mg/kg) for 3 days, blotted with the indicated antibodies.

The combination of MEK and multi-ERBB inhibition was previously shown to suppress the growth of subcutaneous NSCLC xenografts under continuous daily treatment (20); however, cell line xenografts have a poor track record of accurately predicting human patient responses. We therefore tested the therapeutic efficacy of transient inhibition of ERBB and/or MEK in our fully immunocompetent KM model of autochthonous LuAd. Tumors were induced in adult KM mice and allowed to develop for 6 weeks. The presence of an IRFP Cre-reporter allele (33) in a subset of such animals allowed us to confirm the presence of tumors by LI-COR Pearl fluorescence imaging before initiating therapy (fig. S5B). Mice were treated daily for 1 week with neratinib, alone or in combination with trametinib, and then left untreated to determine the effect on overall survival. Transient ERBB blockade alone showed little influence on overall survival, whereas MEK inhibition alone extended survival of KM mice. The combination of neratinib and trametinib further extended survival over that achieved by MEK inhibition alone (Fig. 5D). Fluorescence imaging of IRFP-positive KM mice showed pronounced suppression of lung tumor growth in individual mice treated with combination therapy (fig. S5, B and C). Immunoblotting of tumors from mice treated for 3 days with either drug alone or both combined confirmed the pathway-specific effects of these inhibitors in vivo (Fig. 5E). We conclude from these data that multi-ERBB inhibition may benefit LuAd patients with mutant KRAS-driven disease, if used in combination with other agents such as MEK inhibitors.


EGFR-selective inhibitors have failed to show clinical benefit in mutant KRAS-driven cancers. In contrast with targeted inhibition of EGFR in isolation, we show here that broad inhibition of the ERBB network, using the multi-ERBB inhibitor neratinib, almost completely suppresses formation of KRASG12D-driven lung tumors and enhances the benefit of MEK inhibition in established tumors. We show that ERBB activity enhances signaling through the core RAS→ERK pathway, establishing a feed-forward loop that is associated with progression to invasion and metastasis. We also show that brief, transient inhibition of ERBB signaling enhances the therapeutic benefit of MEK inhibition in the autochthonous setting. Although this latter result is broadly supportive of the clinical implications of an earlier study showing that increased ERBB activity underlies resistance to MEK inhibition (20), our data are fundamentally distinct in that we demonstrate a clear role for ERBB activity from the very outset of KRAS-driven tumor initiation, as opposed to in reaction to targeted inhibition of MEK.

Mechanistically, we show that nascent KRasG12D-driven lung tumors express multiple ERBB ligands and that production of these ligands increases markedly as tumors progress to more aggressive disease. In the cell culture setting, the release of these ligands drives autocrine signaling that supports the viability of cancer cells. In vivo, the release of such ligands into the extracellular milieu likely drives both autocrine and paracrine signal transduction in any local cell population that expresses ERBB RTKs. Hence, it is plausible to consider that some of the tumor-suppressive benefit of neratinib may derive from inhibition of ERBB signaling in such stromal populations. Nevertheless, our cell culture analysis argues that the primary effect of ERBB inhibition is autonomous to the tumor epithelium. Further work will be needed to determine whether paracrine activation of ERBB RTKs on attendant stromal cell populations plays any role in tumor maintenance or progression.

Previous work has revealed that threshold levels of KRAS signaling are required for tumor initiation and progression: In an inducible model of HRASG12V overexpression in mammary glands, focal tumors arising spontaneously from nontransformed epithelium that expressed low amounts of the transgene exhibited a pronounced increase in expression of the HRasG12V transgene and of multiple ERBB ligands, including Ereg, Tgfα, and Hbegf, again suggestive of feed-forward signal amplification (25). More recently, this same “activity threshold” principle was shown to apply to ERK signaling as the key oncogenic effector pathway of KRAS in the lung (34). Through pharmacological enhancement of ERK activity, this study showed that different airway cell types require distinct amounts of ERK signaling for oncogenic transformation by mutant KRAS, and that a second, higher, threshold signal was needed for progression to carcinoma. The existence of this higher threshold in KRAS-driven LuAd was already clear from the pronounced increase in ERK phosphorylation associated with tumor progression (25), which, in the KP model, is driven by spontaneous amplification of the G12D mutated allele (24). Our data suggest that ratcheting up ERBB signal transduction provides an alternative route to RAS pathway signal amplification, independent of KRAS gene amplification. It is important to note that, from the human data, there does not appear to be a preferred mechanism of ERBB signal enhancement—overexpression of ligands, amplification of RTKs, and/or accessory molecules could all plausibly achieve the same effect. What is clear from these data is that the potential to increase ERBB signaling appears to be widespread in KRAS-mutant human LuAd.

Perhaps the most surprising observation is that endogenously expressed KRASG12D requires basal signaling from ERBB RTKs to initiate lung tumors, even when accompanied by MYC overexpression. We previously showed that expression from the Rosa26 locus is refractory to growth factor signaling (35); thus, the requirement for ERBB activity does not reflect an artificial need for ERBB to sustain expression of the MYC transgene, whereas the observed effects on total MYC expression likely reflect the influence of RAS signaling on MYC protein stability (36). Instead, these data suggest that G12 mutant KRAS requires a push from upstream RTKs to breach the initial threshold needed for tumor initiation. Supporting this hypothesis, it was recently shown that G12 mutant KRAS, although no longer responsive to RAS-GAPs [guanosine triphosphatase (GTPase)–activating proteins], does retain some level of intrinsic GTPase activity and thus cycles slowly between on and off states (4), opening the possibility for upstream signaling to influence activity, either directly or by promoting activity of RAS guanonucleotide exchange factors (GEFs). Alternatively, tonic signal transduction through wild-type RAS isoforms may need to combine with that from mutant KRAS to likewise breach the threshold for transformation (37). Our data presently do not distinguish between these possibilities, and they may not be mutually exclusive.

In the course of preparing this manuscript, we became aware of an independent study that strongly complements our observations and underscores our conclusions: The work by Moll and colleagues similarly demonstrates a requirement for ERBB signaling to support progression of KRASG12D-driven lung cancer in vivo (38). Their study used an independent pan-ERBB inhibitor, afatinib, in the context of both KRASG12D-only and KRASG12D;p53Fl/Fl-driven tumor models. The striking similarities in the two studies attest to the on-target specificity of the two drugs used while also showing that the effects are independent of the genetic strategy used to accelerate KRAS-driven disease and are therefore likely to have broader potential application.

Activating mutations in KRAS are currently considered grounds for exclusion from clinical treatment with EGFR/ERBB inhibitors, and single-agent trials of the same drugs have only demonstrated efficacy in patient cohorts with activating mutations in EGFR or ERBB amplification (39). Our data concur that multi-ERBB inhibitors are unlikely to benefit KRAS-driven cancer patients if used in isolation. However, we show here the potential for such drugs to sensitize autochthonous KRAS-driven tumors to additional therapeutic agents, in this case to the MEK inhibitor trametinib. Our data argue that clinical use of multi-ERBB inhibitors as part of an inhibitor cocktail to treat KRAS-driven LuAd deserves reexamination.


Study design

Rationale for the project was to investigate the mechanistic basis of tumor progression in early-stage KRAS-driven lung cancer. Unbiased RNA-seq analysis identified increased expression of the ERBB network, prompting further mechanistic investigation. All mice were asymptomatic at the date of treatment initiation and were allocated to treatment groups randomly. For survival benefit analysis, the absence of previous data precluded a precise power calculation: Sample sizes were therefore estimated according to Lehr’s quick formula at 80% power {n=16(c.v.)2(ln[r.m.])2}, assuming a ratio of the means of 0.2 and a coefficient of variation of 0.3. For all other analyses, sample sizes were determined in accordance with the 3Rs. Animals were treated by facility staff without knowledge of anticipated outcomes. Humane end points were predefined as impaired breathing and/or hunching with 10% or more weight loss. Histology, IHC, and ISH were performed blindly by the Beatson Institute histology core facility and were scored without reference to genotype or treatment group. Reporting is consistent with the ARRIVE guidelines. In vitro experiments were all performed on three separate occasions, except where noted. In vitro analyses were not blinded.

Genetically engineered mice and mouse procedures

Procedures involving mice were performed in accordance with Home Office license numbers 60/4183 and 70/7950 [Cancer Research UK (CRUK), Beatson Institute for Cancer Research, UK]. LSL-KRasG12D (B6.129S4-Krastm4Tyj/J) mice (40) were obtained from the National Cancer Institute (NCI) mouse repository at Frederick, MD. Rosa26DM.lsl-MYC mice were generated as previously described (35) with the ATG-initiated human c-MYC (MYC2) complementary DNA (cDNA), devoid of 5′ and 3′ untranslated regions, replacing the MYC–estrogen receptor ligand-binding domain fusion cDNA. Targeted insertion into the Rosa26 locus was confirmed by Southern blotting, and genotyping was initially performed with the following primers: (i) CCCAAAGTCGCTCTGAGTTG (common), (ii) GCGAAGAGTTTGTCCTCAACC (targeted locus), and (iii) GGAGCGGGAGAAATGGATATGA (wild-type locus). All genotyping was subsequently performed by TransnetYX Inc. Detailed in vivo procedures are provided in the Supplementary Materials along with IHC, laser-capture microdissection, and RNA-seq methodology.

Cell culture and related assays

Human lung cancer cell lines (A549, H2009, and H358) were validated in-house and grown in RPMI 1640 with 10% fetal bovine serum (FBS) and penicillin (100 U/ml)/streptomycin (100 μg/ml). RAS-less MEFs reconstituted with wild-type or mutant KRAS isoforms were provided by the NCI RAS Initiative, Frederick National Laboratory for Cancer Research (Frederick, MD) and were cultured in Dulbecco’s modified Eagle’s medium supplemented with 10% FBS and penicillin (100 U/ml)/streptomycin (100 μg/ml). For immunoblotting, cells were lysed in RIPAHi buffer [150 mM NaCl, 50 mM tris (pH 7.5), 1% NP-40, 0.5% sodium deoxycholic acid, 1% SDS plus Complete protease/phosphatase inhibitor cocktails (Sigma-Aldrich)] and, after Western blotting, probed with the following primary antibodies: p-ERK Thr202;Tyr204 (E-4) (SC-7383, Santa Cruz Biotechnology), p-EGFR Tyr1068 (3777, Cell Signaling Technology), p-ERBB2 Tyr1248 (06-229, Millipore), p-ERBB3 Tyr1197 (4561, Cell Signaling Technology), ERK1/2 (4695, Cell Signaling Technology), EGFR (06-847, Millipore), ERBB2 (OP15L, Merck), and ERBB3 (05-390, Millipore). RAS:RAF binding assays were performed using a commercial RAS activity assay kit (Cytoskeleton) and probed with pan-RAS (AESA02, Cytoskeleton), KRAS (SC-30, Santa Cruz Biotechnology), or HRAS-specific (18295-1-AP, Proteintech) antibodies (41). Cell propagation and death were analyzed by IncuCyte time-lapse video microscopy in the presence of SYTOX Green. Cell death measurements were performed after 48 hours of drug treatment and corrected for confluence.

Statistical analysis

Raw data obtained from quantitative real-time polymerase chain reaction, fluorescence-activated cell sorting, and IncuCyte assays were copied into Excel (Microsoft) or Prism (GraphPad) spreadsheets. All mean, SD, and SEM values of biological replicates were calculated using the calculator function. Graphical representation of such data was also produced in Excel or in Prism. Statistical significance for pairwise data was determined by Student’s t test. For multiple comparisons, ANOVA was used with a post hoc Tukey test. *P < 0.05, **P < 0.01, and ***P < 0.001. For Kaplan-Meier plots, log-rank P values are presented.


Materials and Methods

Fig. S1. Characterization of KM lung tumors.

Fig. S2. Comparison of KM phenotype induced by Ad-CMV-CRE and Ad-SPC-CRE.

Fig. S3. Genomic alterations and expression of ERBB network genes in human KRAS-mutant LuAd.

Fig. S4. Sensitivity of KRAS-mutant cell lines to ERBB blockade.

Fig. S5. Longitudinal in vivo imaging of nascent lung tumors.

Table S1. Summary of Metacore GeneGO pathway analysis.

References (4247)


Acknowledgments: We express our gratitude to P. Adams, O. Sansom, D. Bryant, M. Eilers, J. O’Neil, and R. O’Hagan for helpful discussions; to B. Bauer, C. Kramer, S. Bell, D. Miller, and the staff of the CRUK Beatson biological services unit for animal husbandry and assistance with animal protocols; to P. Adams and N. Pchelintsev for help establishing RNA-seq; and to all members of the Murphy laboratory for assistance in manuscript preparation. Ad-SPC-CRE was donated to the University of Iowa vector core facility by A. Berns (Netherlands Cancer Institute). Funding: Funding was provided by a MINT collaborative grant from Merck Sharpe and Dohme, Deutsche Krebshilfe grant 109220, EC FP7 Marie Curie Actions CIG 618448 “SERPLUC,” and British Lung Foundation grants APHD13-5 and CSOBLF RG16-2 (all to D.J.M.). B.K. was supported by EC H2020 Marie Curie actions mobility fellowship 705190 “NuSiCC.” J.L.Q. was funded by Medical Research Council programme no. 5TS30. Author contributions: Experiments were performed and analyzed by B.K., T.M., S.N., A.H., E.K., S.L., and S.B.C. Histology was performed by C.N., and pathology was analyzed by J.L.Q. and C.D. RNA-seq was performed by W.C. with analysis by A.H. Statistical guidance was provided by A.H. Reagents and guidance were provided by K.H.V. and C.P.M. D.J.M. conceived the project, directed experiments, and wrote the manuscript. All authors read and approved the manuscript. Competing interests: K.H.V. is on the Board of Directors and is a shareholder of Bristol-Myers Squibb, on the Science Advisory Board and a shareholder of GRAIL Inc., and on the Science Advisory Board of PMV Pharma and RAZE Therapeutics. She has received research funding from Astex Pharmaceuticals and contributed to CRUK Cancer Research Technology filing of patent application WO/2017/144877. K.H.V. served on the Science Advisory Boards of Daiichi Sankyo and ITI Life Sciences and presented seminars at Nestle, Novartis, Genentech, GlaxoSmithKline, and Janssen Pharmaceuticals. All other authors declare that they have no competing interests. Data and materials availability: The following RNA-seq data sets are available from ArrayExpress: E-MTAB-6432, E-MTAB-6483, E-MTAB-6485, and E-MTAB-6486. The Rosa26-lsl-MYC and Hprt-lsl-IRFP mice are available upon request from D.J.M. and K.H.V., respectively.

Stay Connected to Science Translational Medicine

Navigate This Article