|Home | About | Journals | Submit | Contact Us | Français|
ErbB3 is a critical activator of PI3K signaling in EGFR (ErbB1), ErbB2 (HER2), and MET addicted cancers, and reactivation of ErbB3 is a prominent method for cancers to become resistant to ErbB inhibitors. In this study, we evaluated the in vivo efficacy of a therapeutic anti-ErbB3 antibody, MM-121. We found that MM-121 effectively blocked ligand-dependent activation of ErbB3 induced by either EGFR, HER2, or MET. Assessment of several cancer cell lines revealed that MM-121 reduced basal ErbB3 phosphorylation most effectively in cancers possessing ligand-dependent activation of ErbB3. In those cancers, MM-121 treatment led to decreased ErbB3 phosphorylation, and in some instances, decreased ErbB3 expression. The efficacy of single-agent MM-121 was also examined in xenograft models. A computational learning algorithm found that MM-121 was most effective against xenografts with evidence of ligand-dependent activation of ErbB3. We subsequently investigated whether MM-121 treatment could abrogate resistance to anti-EGFR therapies by preventing reactivation of ErbB3. We observed that an EGFR mutant lung cancer cell line (HCC827), made resistant to gefitinib by exogenous heregulin, was re-sensitized by MM-121. In addition, we found that a de novo lung cancer mouse model induced by EGFR T790M-L858R rapidly became resistant to cetuximab. Resistance was associated with an increase in heregulin expression and ErbB3 activation. However, concomitant cetuximab treatment with MM-121 blocked reactivation of ErbB3 and resulted in a sustained and durable response. Thus, these results suggest that targeting ErbB3 with MM-121 can be an effective therapeutic strategy for cancers with ligand-dependent activation of ErbB3.
The ErbB family of receptor tyrosine kinases includes EGFR, ErbB2 (HER2), ErbB3 (HER3), and ErbB4 (HER4). Over the past ten years, it has become evident that many epithelial cancers require EGFR or HER2 signaling for their growth and survival. Agents targeting EGFR have become widely used for the treatment of lung, colon, and head and neck cancers, whereas agents targeting HER2 are commonly used to treat HER2 amplified breast cancers. Inhibitors of EGFR and HER2 come in the form of small molecule tyrosine kinase inhibitors (TKIs) and targeted antibodies.
Several recent studies have found that those cancers that are sensitive to EGFR or HER2 inhibitors are unique in that phosphoinositide 3-Kinase (PI3K) signaling is under the sole control of either EGFR or HER2, respectively. For these inhibitors to be effective, they must lead to downregulation of the PI3K/AKT pathway (1–4). Prior studies have identified ErbB3, a kinase dead member of the ErbB family, as the key activator of PI3K/AKT signaling in EGFR addicted cancers (2, 5). In these cells, ErbB3 is tyrosine phosphorylated in an EGFR-dependent manner, and then directly binds PI3K. Upon inhibition of EGFR, ErbB3 phosphorylation is abrogated, it no longer binds PI3K, and there is loss of PI3K/AKT signaling (2, 5). Furthermore, downregulation of ErbB3 using short hairpin RNA leads to a decrease in AKT phosphorylation in EGFR addicted cancers (2). Similarly, ErbB3 is the major activator of PI3K in HER2 amplified breast cancers (reviewed in (6)), and trastuzumab treatment leads to loss of ErbB3 phosphorylation, dissociation between ErbB3 and PI3K, and loss of AKT phosphorylation in these cancers (4). Thus, signaling through ErbB3 is the major mechanism of PI3K/AKT activation in both EGFR and HER2 driven cancers.
Although EGFR and HER2 driven cancers often respond to anti-ErbB therapies, these cancers invariably become resistant. We and others have learned that some cancers become resistant when they re-activate ErbB3 signaling. There are examples of resistance that implicate EGFR, HER2, and MET in reactivating ErbB3 (5, 7–9). In addition, heregulin-induced activation of HER2-ErbB3 heterodimers has also been associated with resistance to EGFR inhibitors (10). Because ErbB3 is a focal point for both the initial effectiveness of EGFR and HER2 therapies as well as the development of drug resistance, there is considerable effort to develop methods to directly target ErbB3 with therapeutics. Unlike other ErbB family members, ErbB3 is characterized by the lack of kinase activity (11). Thus, antibodies directed against ErbB3 may be the most effective method to disrupt its function. In this study, we provide the first evaluation of this class of therapeutics by examining the efficacy of the anti-ErbB3 antibody, MM-121, which is currently in clinical development.
Using a systems biology approach, we previously identified ErbB3 to be a key node in the ErbB signaling network (12). The fully human anti-ErbB3 monoclonal antibody, MM-121, was identified from a phage display library screen based on computationally driven selection criteria (12). MM-121 binds with high affinity to ErbB3 and blocks the binding of its ligand, heregulin, to ErbB3 and inhibits betacellulin(BTC) induced phosphorylation of ErbB3. ErbB3 is known to form heterodimers with a variety of receptors within the ErbB family like EGFR (13) and ErbB2/Her2, and it also associates with MET (5, 14). To assess if MM-121 could inhibit ligand-induced activation of ErbB3 by different receptors, ErbB3 was co-transfected with GFP (control), EGFR, MET, or ErbB2/HER2 in CHO cells. The transfected cells were then treated with the indicated ligands in the absence or presence of MM-121 as shown in Figure 1A. ErbB3 phosphorylation was measured using an antibody that specifically recognizes ErbB3 that is phosphorylated on tyrosine Y1289. In addition, to more globally assess ErbB3 tyrosine phosphorylation, total ErbB3 was immunoprecipitated from the cells and probed with an anti-phosphotyrosine antibody. MM-121 effectively blocked ligand-induced phosphorylation of ErbB3 by all of the co-receptors as evidenced by reduction at the Y1289 site and total ErbB3 phosphotyrosine levels (Fig. 1A). However, the effect on total ErbB3 levels following the different co-transfections was more variable (please see quantification of phospho-ErbB3/total ErbB3 ratios in Fig. S1). Of note, MM-121 decreased ErbB3 expression in CHO cells transfected with EGFR in the presence and absence of the ligands EGF or BTC (Fig. 1A) but not with HER2 or MET. In CHO cells co-transfected with MET and ErbB3, stimulation with heregulin and HGF resulted in a potent phosphorylation of ErbB3 that was inhibited by MM-121. In CHO cells co-transfected with HER2 and ErbB3, MM-121 potently blocked heregulin-induced phosphorylation of ErbB3. Of note, even when ErbB3 internalization in MALME 3M cells was disrupted with cold temperature or a transglutaminase inhibitor, MM-121 inhibited ErbB3 phosphorylation similarly (Fig. S2). Thus, these studies suggest that MM-121 can potently inhibit ligand-induced activation of ErbB3. However, MM-121 may be less effective when ErbB3 phosphorylation does not require ligand.
To further address the effects of MM-121 on both ErbB3 phosphorylation and expression, we treated three cancer cell lines, ACHN, NCI-N87 and BT-474, with MM-121 (Figs. 1B,C). As will be described below, these cell lines are respectively sensitive, moderately sensitive and resistant to MM-121 in vivo. Interestingly, we observed that MM-121 impacts ErbB3 through at least two potential mechanisms. In ACHN cells, there was clear loss of ErbB3 phosphorylation prior to any effect on ErbB3 protein expression, presumably due to inhibition of ligand binding. Contrastingly, in the NCI-N87 cells, the trajectory of ErbB phosphorylation mirrored that of ErbB3 expression. This result implies that receptor downregulation in NCI-N87 cells may contribute to the loss of ErbB3 phosphorylation. Lastly, in BT-474 cells, MM-121 had a minimal effect on either ErbB3 phosphorylation or protein expression levels. Of note, these cells have marked HER2 amplification and ligand-independent activation of ErbB3 (Table S1).
In addition, fluorescent activated cell sorting (FACS) and western blot analyses determined that the effect of MM-121 on ErbB3 expression varied across other cell lines as well. For example, MM-121 effectively downregulated expression of ErbB3 in MALME 3M cells, but not in DU145 cells (Fig. S3).
In order to understand the effect of MM-121 on downstream signaling, we studied the ability of MM-121 to inhibit signaling in vitro in ACHN (renal), Du145 (prostate), OvCAR8 (ovarian) and ADRr (ovarian) cells in the presence of heregulin and BTC (Fig. 2A and Fig. S4). We observed that MM-121 blocked the capacity of heregulin to stimulate ErbB3 and downstream AKT and ERK phosphorylations in all cell lines. In cell lines that MM-121 reduced signal to levels equal to or below the unstimulated cells (Fig. 2A), MM-121 demonstrated efficacy in subsequent in vivo studies (Fig. 2B). In the NCI-ADRr cell line where only partial inhibition of AKT and ERK phosphorylation in response to MM-121 was observed in vitro, MM-121 did not show efficacy in vivo (Fig. 2B). These findings were confirmed in independent experiments assessing the ability of MM-121 to inhibit basal ErbB3 phosphorylation (Fig. S5). Similar experiments with BTC as the activating ligand were conducted; however, the signal strength induced by BTC is often weaker compared to heregulin (Fig. S3).
To more comprehensively evaluate the anti-tumor activity of MM-121 as a single-agent in vivo, we screened a panel of xenograft tumor models (Fig. 2B and Fig. S6). Of the nine tumor models studied, we observed substantial tumor growth arrest in three tumor models (DU145, OVCAR8, and ACHN), modest tumor growth delay in two tumor models (NCI-87, and SKOV3), and no significant anti-tumor activity in the other four tumor models (NCI- ADRr, BT-474, IGROV1 and MALME 3M). If MM-121 treatment led to tumor stasis, the cell line was considered a ‘responder’, and if there was no effect on tumor growth, the cell line was considered a ‘non-responder’. Importantly, we observed that short-term exposure to MM-121 blocked ErbB3 phosphorylation, induced apoptosis and reduced proliferation in one of the responding tumor models (Fig. S7). Upon investigating the expression of phospho-ErbB3 and ErbB3 at the end of the treatment study, we observed that MM-121 continued to inhibit phospho-ErbB3 in the responding ACHN cells, but ErbB3 total protein expression remained intact (Fig. 3A). However, MM-121 had no effect on P-ErbB3 or total ErbB3 in the non-responder BT474 cells (Fig. 3B). These results mirror those obtained from the in vitro studies (Fig. 1B).
In order to gain a better understanding of the responding vs. non-responding cell lines in vivo, we quantified receptor expression levels in the untreated tumors established from the cell lines of interest (Table S1). Since BTC and heregulin were both potent ligands for activating ErbB3 phosphorylation, these two ligands were assessed as well (12). ErbB4 expression levels were essentially undetectable, and were thus omitted in the subsequent analysis. From all the ligands and receptors quantified in the different tumor types, the heregulin expression levels appear to be the best single feature separating responding from non-responding cell lines (Fig. 3C) but no clear separation was obtained by assessment of a single feature.
To distinguish responders and non-responders using two or more variables, we utilized a computational model by using a support vector machine (SVM) algorithm (15, 16). The SVM uses the given set of training xenograft studies that are each characterized by a set of features (ligand and receptor expression levels shown in Table S1) to distinguish responders and non-responders. An SVM model maps the xenograft responses as points in the feature space, so that the responders and non-responders are divided by a clear gap, represented by a hyperplane. The SVM constructs a separating hyperplane in the feature space, which maximizes the margin between the two data sets (responder vs. non-responder). New xenograft studies are then mapped into that same space and predicted to belong to a category based on which side of the gap they fall on. Since parsimonious models are more robust to noisy training data (17), we applied the SVM algorithm with all possible combinations of input variables and then ranked the hyperplanes based on their margin and number of dependent variables. The hyperplane that separated the two classes with the largest margin and the fewest variables consisted of ErbB1, BTC, and HRG1-β1 (shown in Fig. 3D). Our rationale to use the protein expression level of HRG1-b1 and BTC as the two ligands to characterize the xenograft tumors in this modeling algorithm is based on our previous findings that from a panel of nine different ErbB ligands, HRG1-b1 and BTC are the two most potent inducers of ErbB3 phosphorylation (12).
Despite the relatively small number of xenograft studies used as a training set, the SVM model was able to accurately predict that the HCT116, HT29 and H1975 cells would be non-responders and the NCI-N87 cells would be partial-responders (Fig. 3D and Fig. S6). We note that algorithms such as SVM only identify trends in the data and generate hypotheses requiring continued interrogation as this antibody enters the clinic. In addition, we found that HER2 amplification is seemingly associated with resistance to MM-121 (Table S1), likely because these cancers probably are driven by ligand-independent activation of ErbB3 (see BT474 in Fig. S6 and Table S1). Based on the earlier data (Fig. S5), this would suggest that cancers in which HER2 is driving ErbB3 phosphorylation in a ligand-dependent manner may be the ones that are most sensitive to MM-121 as a single-agent.
Interestingly, MM-121 sensitive ACHN cells expressed about 500 times more heregulin and about 100 times more BTC than the MM-121 resistant MALME 3M cells (Fig. 3C) while both cell lines showed similar expression level of HER2 receptors. It thus appears that the ligand expression level is a key indicator of response to MM-121. Indeed, all of the responding cell lines expressed high levels of either heregulin or BTC confirming our earlier observation that MM-121 is likely active in cancers with ligand-dependent ErbB3 signaling.
Over the past few years, the critical role of ErbB3 in EGFR, HER2, and MET driven cancers has become clear. Indeed, recent evidence suggests that cancers can acquire resistance to EGFR and HER2 by adopting a novel way to activate ErbB3 (2, 5, 8, 9). Our studies demonstrate that MM-121 can potently block ligand-induced activation of ErbB3. Thus, we hypothesized that this antibody could overcome ligand-dependent, acquired resistance to EGFR inhibitors, a mechanism shown to cause gefitinib-resistance (10). Accordingly, heregulin promoted gefitinib resistance in HCC827 cells (EGFR exon 19 deletion mutation and EGFR amplified) overexpressing ErbB3 (Fig. 4A and C). The ErbB3 overexpressing cells stimulated with heregulin maintained ErbB3 and AKT phosphorylation in the presence of gefitinib (Fig. 4B). Indeed, HER2 appeared to promote heregulin-induced ErbB3 phosphorylation in the presence of gefitinib. Only the combination of gefitinib and lapatinib abolished ErbB3 and AKT phosphorylation (Fig. 4D, left). Similar to lapatinib, MM-121 reversed the resistance promoted by heregulin (Fig. 4C). Accordingly, MM-121 blocked heregulin-induced rescue of ErbB3 and AKT phosphorylation in the presence of gefitinib (Fig. 4D, right).
We next assessed the activity of MM-121 in a genetically engineered mouse model of in situ lung cancers that are induced by a doxycycline inducible human EGFR T790M-L858R transgene that is expressed specifically in the lung epithelium (18). This genetically engineered mouse model mimics patients with acquired resistance to EGFR TKIs that is caused by acquisition of the secondary T790M mutation (19). These mice develop aggressive lung adenocarcinomas upon continuous administration of doxycycline in their diets. Similar to patients with the T790M resistance mutation, these cancers are resistant to erlotinib (20). We were able to utilize MM-121 for these experiments because MM-121 cross-reacts potently with mouse ErbB3 (Fig. 5A and Fig. S8).
When we treated mice harboring established tumors with the EGFR antibody, cetuximab, there was a modest response at two weeks, but this response was transient, with clear resistance at four weeks (Fig. 5B and 5C). Interestingly, this resistance was associated with increased phospho-ErbB3 in the EGFR T790M-L858R tumors compared to untreated controls (Fig. 6A). The observed increase in phosphorylated ErbB3 levels in the presence of cetuximab correlates with the observed increased levels of heregulin (Fig. 6B). Addition of MM-121 to cetuximab downregulated ErbB3 and the increase in ErbB3 phosphorylation observed following single-agent cetuximab therapy. Importantly, the combination of MM-121 and cetuximab led to a more dramatic and sustained response (Fig. 5B).
In this study, we have evaluated the in vitro and in vivo efficacy of a new type of targeted therapy, an anti-ErbB3 antibody, MM-121. ErbB3 has emerged as a critical partner in EGFR, HER2, and MET oncogene addicted cancers. In particular, its many tyrosine phosphorylation sites serve as potent modules to activate intracellular signaling, especially the PI3K pathway (21, 22). In many EGFR and HER2 driven cancers, ErbB3 tyrosine phosphorylation is necessary for transmittal of these downstream signaling events, and treatment with the appropriate TKI leads to loss of ErbB3 phosphorylation and cessation of downstream signaling. Recent studies have revealed that reactivation of ErbB3/PI3K signaling is a major mechanism of acquired resistance to EGFR and HER2 inhibitors (5, 8, 9, 23). In particular heregulin-induced activation of ErbB3 may cause resistance to gefitinib (10). For these reasons, there has been great enthusiasm for targeting ErbB3 directly as a therapy. However, unlike other ErbB family members, ErbB3 is kinase dead (11). Thus, it seems that antibody therapies directed against the extracellular domain of ErbB3 appear to be the most effective method to disrupt its function.
The in vivo activity of MM-121, an anti-ErbB3 antibody, was assessed in this study. This antibody was chosen for clinical development because it potently blocks ligand-dependent activation of ErbB3 (Figs. 1, ,22 and Figs. S4, S5). In this study, we also provide evidence that MM-121 blocks ligand binding and leads to receptor internalization and degradation. Interestingly, although the antibody blocks ligand-dependent activation of ErbB3 phosphorylation in all cell lines examined, antibody mediated receptor internalization and downregulation appears to be cell line dependent. Our results also imply, that in the presence of a strong autocrine ligand activation of ErbB3 as in the ACHN or DU145 cells, MM-121 blocks ligand activation of ErbB3 without downregulating ErbB3 expression (Fig. 1B and and3A).3A). Yet, in other cell lines, such as the NCI-N87 cells, receptor internalization and degradation may contribute to MM-121 activity (Figure 1B). However, we currently do not know why MM-121 downregulates ErbB3 expression in some cells, but not others. Of note, for MM-121 to be effective in a ligand-independent manner (i.e. only by downregulation of ErbB3), treatment would need to result in elimination of most of the ErbB3 protein. It would appear that in cancers with ligand-independent activation of ErbB3 (e.g. HER2 amplified cancers) MM-121 does not sufficiently downregulate ErbB3 to low enough levels, and thus has minimal effect on cell viability.
MM-121 exerted substantial anti-tumor activity in three xenograft tumor cell lines (DU145, OVCAR8, and ACHN) as a single-agent therapy. However, it was ineffective in many other cancer models. Of particular note, this antibody did not block ErbB3 phosphorylation and was ineffective in the HER2 amplified cancer cell lines (Table S1, Figs. S5 and S6). We hypothesize that the high concentration of HER2 receptors on the membrane likely obviates the need for ligand-dependent activation of ErbB3. On the other hand, it is not surprising that cancers with high ligand expression in non-HER2 amplified cancers were the most responsive to single-agent therapy (Fig. 3). Although these biomarkers are useful towards understanding the biology behind responsiveness to this antibody, it is hard to predict how applicable they will be in the clinical development of this antibody, and more work needs to be done to elucidate possible biomarker-response relationships. Indeed, predicting response may require a panel of biomarkers that will likely include heregulin. Based on these findings, we hypothesize that MM-121 will be most effective when treating ligand-dependent tumors. Of note, since ovarian cancers, pancreatic cancers, papillary thyroid cancers and medulloblastomas express heregulin in more than 70% of analyzed primary tumors (24), therapy using MM-121 may demonstrate activity in these malignancies.
There are likely some cancers that express phosphorylated ErbB3, yet ErbB3 is not critical for the growth and survival. In these cancers, MM-121 may downregulate ErbB3 phosphorylation and/or total ErbB3 without therapeutic impact as a single-agent. In such cancers, PI3K-AKT signaling may not be solely dependent on ErbB3, and/or downregulation of PI3K-AKT may not be sufficient to induce growth arrest and/or apoptosis. In these cancers, there may be a benefit for combining MM-121 with other targeted therapies.
Recent studies have highlighted the central role of re-activation of ErbB3 signaling as a mechanism of acquired resistance to EGFR and HER2 based therapies. Therefore, an effective ErbB3 antibody may retard the development of resistance to these inhibitors. In this study, we observed that MM-121 potently prevented resistance to anti-EGFR based therapies in vitro and in vivo (Figs. 4, ,5,5, and and6).6). Importantly, we assessed the activity of MM-121 in the genetically engineered mouse model of lung cancers driven by EGFR T790M-L858R. The “gatekeeper” T790M mutation in EGFR is observed in 50% of lung cancers that become resistant to the EGFR TKIs, gefitinib and erlotinib (19, 20). This mouse model faithfully recapitulates human lung cancers that harbor this mutation in that tumors derived from these mice are resistant to gefitinib yet sensitive to irreversible EGFR inhibitors (18, 25). These cancers demonstrated a transient response to cetuximab (Fig. 5). However, resistance to cetuximab was associated with an activation of ErbB3 phosphorylation and increased expression of heregulin (Fig. 6). This suggests that activation of ErbB3 in a ligand-dependent manner caused resistance. Importantly, addition of MM-121 to cetuximab blocked the reactivation of ErbB3 and led to a greater and more durable response. Taken in its entirety, these data provide the clinical basis for the evaluation of MM-121 in combination with anti-EGFR and HER2 based therapies. This may lead to more impressive responses and improved times to progression for patients with EGFR or HER2 driven cancers.
Cell lines used in this study were obtained from the National Cancer Institute’s Developmental Therapeutics Program (http://www.dtp.nci.nih.gov) where these cell lines have been maintained in cryopreservation and in culture, and they have been subjected to strict quality controls, including adventitious agent testing, human isoenzyme analysis, karyology, morphological and immunocytochemical characterization, and DNA fingerprinting. The cell lines were kept below 10 passages or passaged for less than 6 months after receipt or resuscitation in our laboratories. All cell lines were maintained in RPMI-1640 M Lglutamineμmedia - supplemented with 10% fetal calf serum (FCS), 2 μM L-glutamine and Pen-Strep, and grown in a humidified atmosphere of 5% CO2, 95% air at 37°C, unless otherwise indicated. CHO cells were described previously (2).
ACHN, NCI-N87 and BT-474 cells were treated with 170 μg/ml of MM-121. Cells were lysed, blotted and analyzed for total and phospho-ErbB3 at 0 min, 10 min, 30 min, 3 h, 9 h and 18 h. Proteins from western blots were quantified using Syngene GeneTools software program (Frederick, MD). Phospho-ErbB3/ErbB3 ratios of MM-121 treated lysates were normalized to the corresponding control (same treatment, without MM-121). Please see supplementary materials for additional experimental details.
Mouse lung cancer model driven by doxycycline inducible EGFR TL expression was reported previously (18). mice were put on doxycycline diet (Research Diets, Inc., New Brunswick, NJ) at the age of six weeks and imaged by MRI to document tumor burden after another six to eight weeks. The mice were then randomly divided into placebo, MM-121, cetuximab (BMS pharmaceuticals, New York, NY) and MM121 plus cetuximab treatment groups. MM-121 and cetuximab groups were all given a 1mg/mouse drug dose through intraperitoneal injection every three days. Each antibody was given at different days in combination treatment group. Mice were imaged by MRI at two weeks and four weeks following initiation of treatment and then sacrificed for histological analysis. The protocol for the animal work was approved by the Dana-Farber Cancer Institute IACUC, and the mice were housed in a pathogen-free environment at the Harvard School of Public Health. Please see supplementary materials for further experimental details.
We attempted to characterize the differences between MM-121 xenograft responders and non-responders based on the measurements of ErbB1-3, BTC secretion, and heregulin (HRG1-β1) expression in three untreated tumors reported as means and standard deviation in Table S1. In order to be able to take the log10 of the zero entries in Table S1 we replaced zero with a very small number by using the smallest non-zero value of the data and dividing it by 10. The Support Vector Machine (SVM) algorithm (15, 16) was used to define a boundary that distinguished the xenograft responders and non-responders based on these five measurements. We used a Matlab (Mathworks, Natick, MA) toolbox that can be downloaded from http://www.isis.ecs.soton.ac.uk/resources/svminfo/. Briefly, our implementation of the algorithm computes a linear, multi-dimensional boundary, or hyperplane, that separates the two classes such that the distance (or “margin”) between them is maximized.
This work was supported by a Dana-Farber Harvard Cancer Center Lung Cancer Specialized Program of Research Excellence (SPORE) grant P50 CA090578 (J.A.E. and K.-K.W.), National Institutes of Health (NIH) grants K08 AG024004 (K.-K.W.), R01 AG2400401 (K.-K.W.), R01 CA122794 (K.-K.W) and R01 CA140594 (J.A.E. and K.-K.W), NIH K08 grant CA120060 (JAE), R01CA137008 (J.A.E.), R01CA140594 (J.A.E), DF/HCC Gastrointestinal Cancer SPORE P50 CA127003 (J.A.E.), the American Association for Cancer Research (J.A.E.), the V Foundation (J.A.E.), American Cancer Society RSG-06-102-01-CCE (J.A.E.), and the Ellison Foundation Scholar (J.A.E.).