MEK Inhibition Leads to PI3K/AKT Activation by Relieving a Negative Feedback on ERBB Receptors

The phosphoinositide 3-kinase (PI3K), RAF/MEK/ERK mitogen-activated protein kinase (MAPK), and mTORC1 pathways transmit signals from receptor tyrosine kinases (RTKs) to downstream effector networks regulating cell growth, metabolism, survival, and proliferation (1–3). Numerous feedback systems regulating these oncogenic pathways have been described, and can potentially impact the sensitivity of cancers to kinase inhibitors. For example, inhibition of mTORC1 relieves proteasomal degradation of IRS-1 leading to feedback up-regulation of IRS-1/PI3K/AKT, reducing the efficacy of mTORC1 inhibitors as single agents and prompting the use of combination therapies (4–6). PI3K and AKT inhibitors relieve a negative feedback on ERBB receptors and other RTKs leading to partial reactivation of PI3K/AKT signaling, MEK/ERK signaling, and other downstream pathways, potentially limiting the use of PI3K inhibitors as single agents (7–9).

Targeted therapies, such as the EGFR inhibitors gefitinib and erlotinib, are highly effective when cells are “addicted,” and inhibition of the target leads to downregulation of critical growth and survival signaling pathways, especially PI3K/AKT and MEK/ERK (10–12). We recently found that treatment with a combination of a MEK inhibitor and a PI3K inhibitor led to significant apoptosis in EGFR-driven cancers, similar to that induced by an EGFR TKI, whereas treatment with either pathway inhibitor alone did not induce marked cell death (11). In those studies, treatment with a single-agent MEK inhibitor led to increased AKT phosphorylation. Indeed, several other studies have shown that MEK inhibition leads to increased AKT activation, often resulting in reduced efficacy of MEK inhibitors as single agents (11, 13–15). However, the molecular mechanisms underlying this feedback remain unknown.

Several mechanisms for MEK feedback regulation of AKT signaling have been suggested. For example, ERK-mediated serine phosphorylation of the GAB1 adaptor has been shown to negatively regulate GAB1–PI3K binding and downstream AKT signaling (16–18). MEK inhibition can also downregulate mTORC1 signaling, relieving negative feedback on IGF-IR/IRS-1 and activating PI3K/AKT signaling (19). ERK has also been shown to directly regulate ERBB tyrosine phosphorylation (20, 21). However, it remains unclear which mechanisms, if any, are dominant in MEK inhibitor–induced activation of AKT signaling in EGFR or HER2-driven cancers. As multiple MEK and BRAF inhibitors, including the highly selective allosteric MEK1/2 inhibitor, AZD6244 (22), are being developed, understanding the signaling feedbacks induced by MEK inhibitors that may ultimately impact their use will become increasingly important.

In this study, we examined the molecular mechanism by which MEK inhibition leads to increased AKT phosphorylation in EGFR and HER2-driven cancers. We provide evidence suggesting that this feedback occurs at the level of increased phosphatidylinositol 3,4,5-trisphosphate (PIP₃) induced by an increased association between ERBB3 and PI3K. Increased ERBB3 activation results from loss of an inhibitory ERK-dependent threonine phosphorylation in the conserved juxtamembrane domains of EGFR and HER2, previously found to regulate to EGFR auto-phosphorylation (21). Elucidation of this mechanism provides a greater understanding of the feedback systems regulating key pathways that drive human cancers.

Cell lines, inhibitors, and growth conditions are described in Supplementary Materials and Methods. Cells were lysed in an NP-40 containing buffer, separated by SDS/PAGE, and transferred to PVDF membranes. Antibody binding was detected using enhanced chemiluminescence (Perkin-Elmer).

HCC827 cells were washed with PBS and labeled for 1 hour at 4°C in 0.5 μg/mL Sulfo-NHS-LC-Biotin (Thermo Scientific) resuspended in PBS ± AZD6244. Labeling was quenched with 100 mmol/L glycine. Cells were then returned to media at 37°C before lysis. Biotin-labeled cell surface proteins were immunoprecipitated with NeutrAvidin Agarose Resins (Thermo Scientific), separated by SDS-PAGE, and immunoblotted to detect the indicated proteins. Transferrin receptor was used as a loading control.

Xenograft studies were conducted in accordance with the standards of the Institutional Animal Care and Use Committee (IACUC) under a protocol approved by the Animal Care and Use Committee of Massachusetts General Hospital. Nude mice (nu/nu; 6–8 weeks old; Charles River Laboratories) were injected with a suspension of 5 × 10⁶ H1975 cells subcutaneously into the flank of each mouse. Once the mean tumor volume reached ∼500 mm³, AZD6244 was administered by oral gavage in 3 doses of 25 mg/kg over 30 hours.

Phospholipids were isolated from cells and PIP₃ and PI(4,5)P₂ levels were measured using ELISA kits (Echelon, K-2500s and K4500) according to the manufacturer's instructions. Statistical significance was calculated using a t test.

RNA was isolated using the RNEasy Kit (Qiagen) and cDNA was transcribed with Superscript II Reverse Transcriptase (Invitrogen) and used as a template for PCR amplifications. The relative copy number for ERBB3 and HRG was determined using quantitative reverse transcriptase-polymerase chain reaction (qRT-PCR) using a light-cycler 480 (Roche) as previously described (11). The PCR primers and conditions are available upon request.

HCC827 and BT-474 cells were transfected with 50 nmol/L ERBB3s4779 silencer select validated siRNA or negative control (Ambion) with HiPerFect Transfection Reagent (Qiagen) according to manufacturer's instructions.

Transient transfections of CHO-KI cells were conducted with TransIT-LT1 Transfection Reagent (MirusBio LLC) according to the manufacturer's recommendations. Wild-type ERBB3 was cotransfected with an equal ratio of GFP or wild-type or mutant EGFR or HER2.

HCC827 cells were infected with shERBB3 as previously described (23, 24) with tet-on PLKO shERBB3 or scramble shRNA knockdown vectors (Addgene) and selected in puromycin. shRNA hairpin sequences are provided in the Supplementary Materials and Methods.

Human EGFR (wild-type and exon 19del) and HER2 cDNA coding regions were cloned into the pENTR/D-TOPO vector (Invitrogen) and mutants were constructed with Quick Change Site-Directed Mutagenesis Kit (Stratagene) according to the manufacturer's instructions. All constructs were confirmed by DNA sequencing. Constructs were cloned into the plenti-IRES-GFP lentiviral vector (Addgene) and infections were conducted as described previously (25).

BT-474 cells were transfected with ERBB3 siRNA (Ambion) for 48 hours, then treated with AZD6244 or GDC-0941 for 72 hours. Cells were collected and stained with propidium iodide (PI) and Annexin V as described previously (26). Cells were analyzed using a BD LSR3 analytical flow cytometer (BD Biosciences). Apoptosis was calculated using the sum of Annexin V–positive and PI/Annexin V double-positive cells.

EGFR or HER2 was immunoprecipitated from cells treated with AZD6244 using anti-EGFR antibody (Santa Cruz) or an anti-HER2 antibody, separated by SDS/PAGE, stained with Coomassie blue. Bands were excised and samples were prepared and analyzed by reversed-phase microcapillary/tandem mass spectrometry (LC/MS-MS) as described previously (27–29) and further detailed in the Supplementary Materials and Methods.

We previously observed that AKT phosphorylation increased in response to MEK inhibition in HER2-amplified and EGFR-mutant cancer cells (11). To determine whether this potential feedback is observed in multiple EGFR- or HER2-addicted cancer models, we treated HER2-amplified or EGFR-mutant cell lines with the highly selective allosteric MEK1/2 inhibitor, AZD6244. This MEK inhibitor was used at a concentration of 2 μmol/L, which sufficiently inhibited ERK1/2 phosphorylation in the HCC827 cell line (Supplementary Fig. S1). Similar results were observed using 2 distinct allosteric MEK inhibitors, GSK212 and PD0325901 (Supplementary Fig. S1). In each cell line, we observed increased AKT phosphorylation at both S473 and T308 after AZD6244 treatment, as well as increased phosphorylation of several AKT targets including GSK3α/β, ATP citrate lyase, and PRAS40 (Fig. 1A). We confirmed that these proteins were AKT substrates, as cotreatment with an allosteric AKT inhibitor blocked their phosphorylation (Supplementary Fig. S2). MEK inhibition also led to upregulation of phospho-CRAF and phospho-MEK (Fig. 1A), suggesting activation of a common upstream signaling molecule. This feedback also occurred in vivo, as we observed increased phospho-AKT in an EGFR-mutant H1975 (L858R/T790M) xenograft model treated with AZD6244 (Fig. 1B).

Increased AKT phosphorylation suggested a potential increase in the abundance of PIP₃ (the lipid product of PI3K). Therefore, EGFR-driven HCC827 and HER2-driven MDA-MB-453 cells were treated with a MEK inhibitor, lipids were isolated, and PIP₃ levels were quantified. In both cell lines, AZD6244 induced significant increases in PIP₃ (Fig. 1C). We did not observe any change in expression of the PTEN phosphatase responsible for de-phosphorylating PIP₃ after MEK inhibition (Fig. 1A). To determine if MEK inhibition led to activation of PI3K, we immunoprecipitated the p85 regulatory subunit of PI3K and assessed the abundance of bound adaptors. PI3K consists of a p110 catalytic subunit and a p85 regulatory subunit, and is activated when p85 SH2 domains bind to tyrosine-phosphorylated proteins with YXXM motifs. Treatment with AZD6244 increased the association between PI3K and tyrosine-phosphorylated adaptors, including ERBB3 and GAB1 (Fig. 1D). These results suggest that MEK inhibition leads to an increase in the phospho-tyrosine signaling cascades that directly activate PI3K.

In EGFR- and HER2-driven cancers, ERBB3 is a primary activator of PI3K/AKT (24, 30). We observed increased ERBB3 binding to PI3K after MEK inhibition (Fig. 1D), and accordingly, MEK inhibition substantially increased tyrosine-phosphorylated ERBB3 levels (Fig. 1A). In some cell lines, we observed an increase in total ERBB3 along with phospho-ERBB3 (Fig. 1A). Of note, we did not observe a change in expression of the E3-ubiquitin ligase, neuregulin receptor degradation protein 1 (NRDP1), which can control the steady-state levels of ERBB3 (refs. 31, 32; Fig. 1A). There was also no increase in ERBB3 mRNA levels after AZD6244 treatment (Supplementary Fig. S3), suggesting that any increase in ERBB3 protein levels is posttranscriptional.

To assess the kinetics of this feedback response, we treated the HCC827 cells with AZD6244 over a time course. Phoshosphorylation of ERBB3 and AKT, as well as downstream substrates, increased after just 1 hour of MEK inhibition and continued to accumulate for 24 hours (Fig. 2A). To determine if the feedback activation of ERBB3 occurs on the plasma membrane, we biotin-labeled the surface of HCC827 cells in the presence or absence of AZD6244 and immunoprecipitated the labeled proteins. After just 1 hour of MEK inhibition during biotin labeling, surface levels of the activated receptor were substantially elevated (Fig. 2B). Total ERBB3 on the cell surface also increased after AZD6244 treatment. MEK inhibition did not seem to significantly affect the kinetics of loss of ERBB3 on the cell surface (Fig. 2B), suggesting that receptor internalization or cycling was not significantly affected. These data show that feedback activation of ERBB3 occurs rapidly on the plasma membrane.

To determine if increased ERBB3 phosphorylation caused the increase in AKT phosphorylation after MEK inhibition, we suppressed expression of ERBB3 using a Tet-inducible shERBB3 hairpin construct. After treatment with doxycycline there was effective knockdown of ERBB3, and this abrogated the increase in AKT signaling normally observed after MEK inhibition (Fig. 3A). In HER2-amplified BT-474 cells with abrogated ERBB3 expression, the increase in AKT signaling after MEK inhibition was also attenuated (Fig. 3B). In contrast to HCC827 cells, we observed significant downregulation of basal AKT phosphorylation in BT-474 cells after ERBB3 knockdown (Fig. 3B), indicating the sole reliance on ERBB3 for PI3K activation in this HER2-amplified cancer. In contrast, EGFR-mutant cancers also use GAB1 to activate PI3K (Fig. 1D; refs. 24, 33).

We suspected that knockdown of ERBB3 may increase the efficacy of MEK inhibition by suppressing PI3K/AKT signaling. Treatment with ERBB3 siRNA induced similar levels of cell death compared with treatment with a PI3K inhibitor, GDC-0941 (Fig. 3C). Indeed, combining ERBB3 siRNA with AZD6244 increased the cell death response, approaching the level of apoptosis achieved with GDC-0941 in combination with AZD6244 (Fig. 3C). These data indicate that ERBB3 plays a significant role in MEK feedback on PI3K/AKT signaling in EGFR- and HER2-driven cell lines, suggesting that combination therapies targeting MEK and ERBB3 or MEK and PI3K may block feedback activation of ERBB3/PI3K/AKT signaling and thus be more effective than treatment with a MEK inhibitor alone.

We next determined whether MEK feedback on ERBB3 also occurs in cancers not addicted to EGFR or HER2. We treated a panel of KRAS-mutant cell lines, which have low basal levels of phospho-ERBB3 with AZD6244. Surprisingly, MEK inhibition led to significant activation of ERBB3, but in contrast to EGFR-mutant and HER2-amplified cancers, the increased ERBB3 activation did not translate to increased phospho-AKT (Fig. 4A). Similar to the EGFR- and HER2-driven models, we also observed upregulation of phospho-CRAF and phospho-MEK after MEK inhibition. We suspect that increased ERBB3 phosphorylation did not drive PI3K in these KRAS-mutant cell lines because they express significantly less EGFR and HER2, resulting in markedly lower levels of phospho-ERBB3 compared with those observed in EGFR- and HER2-driven models (Supplementary Fig. S4). Indeed, we recently reported that IGF-IR/IRS signaling is the major PI3K input in these cells (19). Thus, the feedback from MEK inhibition to activation of ERBB3 seems to be conserved in all 3 of the models we tested, including EGFR-mutant, HER2-amplified, and KRAS-mutant cancers, but results in increased PI3K/AKT signaling only in cells that express sufficient absolute levels of phospho-ERBB3.

The feedback observed in EGFR- and HER2-driven cancers is distinct from a well-described feedback mechanism in which mTORC1 inhibition leads to increased IRS-1 expression and upregulation of IGF-IR/IRS signaling (5, 6). In the KRAS-mutant cell lines that we analyzed, which primarily use IGF-1R/IRS to activate PI3K (19), treatment with the mTORC1 inhibitor rapamycin led to feedback activation of AKT signaling that was blocked by cotreatment with the IGF-IR/IR inhibitor, NVP-AEW541 (Fig. 4A and B). In contrast, MEK inhibitor–induced activation of ERBB3 in the KRAS-mutant cancers was blocked by gefitinib, but not by NVP-AEW541 (Fig. 4B). Accordingly, NVP-AEW541 failed to abrogate AZD6244-induced activation of phospho-AKT in EGFR- and HER2-driven cell lines (Supplementary Fig. S5). Of note, we have also previously observed cancers in which MEK inhibition leads to inhibition of downstream phospho-S6, resulting in feedback activation of IGF-IR/IRS-1/AKT signaling independent of ERBB3 in both KRAS wild-type and mutant cancers (19), suggesting that cancers not driven by EGFR or HER2 may have alternate, ERBB3-independent, mechanisms of MEK inhibitor–induced feedback activation of AKT. Our data suggest that the effect of MEK inhibition on ERBB3 is a novel feedback mechanism, distinct from mTORC1 feedback on IGF-IR/IRS-1. A model describing these findings is shown in Fig. 4C.

We investigated the mechanism leading to increased ERBB3 phosphorylation after MEK inhibition. HRG ligand expression was not increased with AZD6244 (Supplementary Fig. S6); however, MEK inhibitor–induced feedback activation of AKT required EGFR or HER2 kinase activity (Supplementary Fig. S7). Indeed, even in KRAS-mutant SW1463 cells, MEK feedback on ERBB3 was still dependent on EGFR kinase activity (Fig. 4B).

Because EGFR and HER2 inhibition blocked MEK feedback activation of ERBB3/PI3K/AKT, we investigated whether MEK inhibition affected the activation of these receptors. Treatment of HCC827 and BT-474 cells with AZD6244 resulted in increased tyrosine phosphorylation of both EGFR and HER2, indicative of receptor activation (Fig. 5A). It has been reported that activation of EGFR involves the formation of an asymmetric dimer (34). Formation of the active RTK dimer is facilitated by stabilizing contacts made between the juxtamembrane domain of the “receiver/acceptor” kinase and the C-terminal lobe of the “activator/donor” kinase (21, 34, 35). Threonine 669 of EGFR, a putative MAPK target site, is conserved within the juxtamembrane domains of EGFR, HER2, and ERBB4 (21). When overexpressed in CHO-KI cells, mutation of this threonine site has been shown to augment EGFR tyrosine phosphorylation (20, 21, 36, 37). However, the physiological consequences of EGFR T669 phosphorylation in cancer models and on ERBB3/PI3K/AKT signaling remained unknown.

We hypothesized that the MEK/ERK pathway may suppress trans-phosphorylation of ERBB3 by directly phosphorylating the juxtamembrane domains of EGFR and HER2, and that this could be a dominant MEK inhibitor–induced feedback leading to AKT activation in these cancers. We used tandem mass spectrometry to measure the effects of AZD6244 on phosphorylation of this juxtamembrane domain threonine residue in both EGFR-mutant and HER2-amplified cancer models. Targeting both the phosphorylated and nonphosphorylated peptide forms, we detected a 66% average decrease in EGFR T669 phosphorylation and a 75% decrease in HER2 T677 phosphorylation upon treatment with AZD6244 (Fig. 5B; Supplementary Fig. S8). Phospho-specific antibodies confirmed that treatment with AZD6244 inhibited phosphorylation of T669 of EGFR and the analogous T677 of HER2 (Fig. 5A). Together these data indicate that loss of this inhibitory threonine phosphorylation on the juxtamembrane domains of EGFR and HER2 occurs in cancer cell lines after MEK inhibition.

We hypothesized that MEK inhibition activates AKT by inhibiting ERK activity, which blocks an inhibitory threonine phosphorylation on the juxtamembrane domains of EGFR and HER2, thereby increasing ERBB3 phosphorylation. To test this hypothesis, we transiently transfected CHO-KI cells, which do not express ERBB receptors endogenously, with wild-type ERBB3 with either wild-type EGFR or EGFR T669A. In cells transfected with wild-type EGFR, MEK inhibition led to feedback activation of phospho-ERBB3 and phosho-EGFR, recapitulating the results we had observed in our panel of cancer cell lines (Fig. 6A). In contrast, the EGFR T669A mutant increased both basal EGFR and ERBB3 tyrosine phosphorylation that was not augmented by MEK inhibition. As a control, we treated CHO-KI cells expressing EGFR T669A with HRG ligand to induce maximal ERBB3 phosphorylation (Fig. 6A), indicating that the lack of induction of phospho-ERBB3 in EGFR T669A expressing cells after MEK inhibition was not simply because of the saturation of the system with phospho-ERBB3. We observed analogous results in CHO-KI cells expressing wild-type ERBB3 in combination with wild-type or T677A mutant HER2 (Fig. 6B). Together these results support the hypothesis that inhibition of ERK-mediated phosphorylation of a conserved juxtamembrane domain threonine residue leads to feedback activation of EGFR, HER2, and ERBB3 (Fig. 7).

To determine if this feedback model explains the activation of PI3K signaling in EGFR-mutant cancers, we used shRNA to knockdown endogenous EGFR (which carries an exon 19 deletion) in the HCC827 NSCLC cell line and replaced with either EGFR (exon 19del) wild-type at T669, or EGFR (exon 19del) carrying a T669A mutation. Of note, this is the same EGFR-mutant cell line in which we observed that EGFR T669 is phosphorylated in MEK-dependent manner (Fig. 5; Supplementary Fig. S8A). When endogenous EGFR was replaced with EGFR (exon19del) wild-type at T669, MEK inhibition led to significant feedback activation of ERBB3/PI3K/AKT signaling (Fig. 6C). However, replacement with the EGFR (exon19 del) T669A mutant led to increased tyrosine phosphorylation of both EGFR and ERBB3, and activation of PI3K/AKT signaling, mimicking the effect of MEK inhibition (Fig. 6C). As expected, addition of AZD6244 failed to further augment ERBB3 and AKT phosphorylation in cells expressing the 669A mutant. These results show that EGFR T669 phosphorylation is necessary for MEK/ERK to suppress EGFR-mediated activation of ERBB3. This supports the hypothesis that a dominant ERK feedback on ERBB3/PI3K/AKT is mediated through phosphorylation of T669 on EGFR (or T677 HER2).

RAF and MEK inhibitors are being developed as treatments for cancers with activation of RAF/MEK/ERK signaling. However, with the exception of BRAF-mutant melanomas, the efficacy of these drugs as single agents has been underwhelming to date. Although there are several potential reasons for this lack of efficacy, feedback activation of parallel oncogenic pathways including PI3K/AKT has been invoked (11, 13–15). This idea is analogous to findings that mTORC1 inhibitors are limited by feedback activation of PI3K signaling (4, 6). In this study, we observe that MEK-inhibitor–induced activation of PI3K/AKT occurs in multiple ERBB-driven cancer models via loss of an inhibitory threonine phosphorylation in the conserved juxtamembrane domains of EGFR and HER2. Phosphorylation of this threonine residue has been shown to impair EGFR activation, likely through disruption of receptor dimerization (21).

Our findings suggest that direct ERK-mediated phosphorylation of EGFR T669 and HER2 T677 suppresses activation of ERBB3. These findings agree with those by Li and colleagues who observed that MEK inhibition failed to increase phosphorylation of EGFR T669A homodimers expressed in CHO-KI cells (20). In this study, we extend previous findings by directly showing the effects of EGFR T669A on ERBB3/PI3K/AKT signaling in an EGFR-mutant cancer cell line. Furthermore, we show that although multiple mechanisms for MAPK feedback regulation of AKT signaling have been proposed, T669A mutation of EGFR is sufficient to block MEK inhibitor-induced feedback activation of PI3K/AKT, suggesting that the feedback we describe herein is one of the dominant mechanisms regulating AKT activation in EGFR- and HER-driven cancers.

In addition to increased ERBB3 tyrosine phosphorylation, we also observe increased expression of total ERBB3 protein after MEK inhibition. This increase seems to be posttranscriptional as no change in ERBB3 mRNA levels was observed with AZD6244 (Supplementary Fig. S3). We were unable to definitively determine the mechanism for increased expression of total ERBB3. However, we observed that increased ERBB3 expression was not solely a result of increased tyrosine phosphorylation of ERBB3 (Fig. 4B). Interestingly, inhibition of ERK-mediated phosphorylation of the threonine juxtamembrane domain sites seems to be necessary for both increased phospho- and total ERBB3 levels. For example, expression of T669A EGFR in CHO-KI cells and HCC827 cells led to increased basal ERBB3 expression and phosphorylation, which was not further augmented by AZD6244 (Fig. 6). This suggests that the increases in both phosho- and total ERBB3 are the result of increased dimer formation between EGFR and ERBB3, which results from loss of inhibitory threonine phosphorylation within the juxtamembrane domain of EGFR. Although we believe that the data support such a model, it remains possible that phosphorylation of the EGFR juxtamembrane domain affects tyrosine-phosphorylated and total ERBB3 levels via a mechanism not linked to heterodimer formation.

Overall, this study suggests that combining MEK inhibitors with either ERBB or PI3K inhibitors may be effective strategies in the clinic. Although there are currently no approved therapies targeting ERBB3, development of anti-ERBB3 antibodies is underway and our data suggests the possible use of combining these antibodies with MEK inhibitors to block feedback activation of AKT in multiple cancer models. Interestingly, we also observed feedback activation of ERBB3 after MEK inhibition in KRAS-mutant cancers that express low basal levels of phospho-ERBB3 and therefore do not use ERBB3 to activate PI3K. This observation suggests that MEK feedback on ERBB3 occurs in a range of cancers, regardless of dependence on ERBB signaling, and highlights another potential complication for patients treated exclusively with inhibitors of the RAF/MEK/ERK pathway. For example, in KRAS-mutant cancers that initially respond to single agent RAF/MEK inhibitors, chronic inhibition of this pathway may lead to persistent activation of EGFR or HER2. Therefore, these data suggest that activation of ERBB signaling may lead to resistance to single-agent RAF or MEK inhibitors.

J.A. Engelman receives research support from AstraZeneca and GalaxoSmithKline and consults for AstraZeneca, GalaxoSmithKline, Novartis, and Genentech. No potential conflicts of interest were disclosed by the other authors.

Conception and design: A.B. Turke, J.A. Engelman

Development of methodology: A.B. Turke, Y. Song, C. Costa, J.A. Engelman

Acquisition of data (provided animals, acquired and managed patients, provided facilities, etc.): A.B. Turke, Y. Song, R. Cook, C.L. Arteaga, J.M. Asara

Analysis and interpretation of data (e.g., statistical analysis, biostatistics, computational analysis): A.B. Turke, C. Costa, J.M. Asara, J.A. Engelman

Writing, review, and/or revision of the manuscript: A.B. Turke, Y. Song, C.L. Arteaga, J.M. Asara, J.A. Engelman

Study supervision: J.A. Engelman

The authors thank Min Yuan and Susanne Breitkopf for help with mass spectrometry, Mike Rothernberg and Cyril Benes for helpful discussions, and Zachary Morris and Andrea McClatchey for assistance with biotin labeling experiments.

This study is supported by the NIH Gastrointestinal Cancer SPORE P50 CA127003, K08 grant CA120060-01, R01CA137008-01, R01CA140594, 1U01CA141457-01, Lung SPORE P50CA090578, the V Foundation, American Cancer Society RSG-06-102-01-CCE, the Ellison Foundation Scholar Award, and the American Lung Cancer Association Lung Cancer Discovery Award (all to J.A. Engelman); NIH 5P01CA120964-04 and NIH DF/HCC Cancer Center Support Grant 5P30CA006516-46 (J.M. Asara); and R01CA80195 and Vanderbilt Breast Cancer SPORE P50 CA98131 (C.L. Arteaga).

The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.