Search

Supplementary Methods, Figures S1-S11, and Table Legends. Supplementary Methods: description of methods used in the Supplementary figures, expanded description of ELISA assays, and description of the computational model of multiple EGFR ligand antagonists. Supplementary Figure S1: Monoclonal EGFR antibodies poorly block signal amplification by high-affinity EGFR ligands. Supplementary Figure S2: Hierarchical clustering of EGFR ligand expression in primary colon, head and neck, and lung tumors. Supplementary Figure S3: Trends observed in EGFR ligand expression in TCGA data set are similarly apparent in primary colorectal samples analyzed by RT-qPCR. Supplementary Figure S4: Relative expression of EGFR ligands varies across cancer indications. Supplementary Figure S5: Simulated activity of a combination of two ligand antagonists versus single antagonists as a function of ligand affinity. Supplementary Figure S6: MM-151 antibodies have non-overlapping epitopes and can simultaneously engage the same EGF receptor molecule. Supplementary Figure S7: MM-151 elicits superior ligand antagonism. Supplementary Figure S8: MM-151 promotes potent down-regulation of EGFR that is dependent upon the oligoclonal combination of bivalent antibodies. Supplementary Figure S9: MM-151 promotes both antibody-dependent cell-mediated (ADCC) and complement-dependent (CDC) cytotoxicity. Supplementary Figure S10: Cell proliferation stimulated by high-affinity ligands is partially or completely resistant to cetuximab treatment but still sensitive to MM-151 treatment. Supplementary Figure S11: MM-151 overcomes signal amplification driven by high-affinity ligand. Legends for Supplementary Tables S1-S8.

Supplementary Tables S2, S3, S4. Structure of the computational model of multiple ligand antagonists: species (Table S2), parameters (Table S3), and Reactions (Table S4).

Supplementary Table S1. Cell lines and cell culture conditions

Supplementary Tables S5, S6, S7, S8. mRNA expression data for EGFR ligands as measured by RNAseq from The Cancer Genome Atlas (TCGA) across 14 cancer indications or by RT-qPCR of colorectal cancer FFPE samples. Table S5: Summary of TCGA datasets included in the analysis. Table S6: Raw and normalized expression data from TCGA. Table S7: EGFR ligand expression measured by RT-qPCR on commercially-obtained colorectal cancer FFPE samples. Table S8: Pairwise correlations of EGFR ligands across four TCGA indications.

Targeted therapies have shown significant patient benefit in about 5–10% of solid tumors that are addicted to a single oncogene. Here, we explore the idea of ligand addiction as a driver of tumor growth. High ligand levels in tumors have been shown to be associated with impaired patient survival, but targeted therapies have not yet shown great benefit in unselected patient populations. Using an approach of applying Bagged Decision Trees (BDT) to high-dimensional signaling features derived from a computational model, we can predict ligand dependent proliferation across a set of 58 cell lines. This mechanistic, multi-pathway model that features receptor heterodimerization, was trained on seven cancer cell lines and can predict signaling across two independent cell lines by adjusting only the receptor expression levels for each cell line. Interestingly, for patient samples the predicted tumor growth response correlates with high growth factor expression in the tumor microenvironment, which argues for a co-evolution of both factors in vivo., npj Systems Biology and Applications, 3, ISSN:2056-7189

Although EGFR is a validated therapeutic target across multiple cancer indications, the often modest clinical responses to current anti-EGFR agents suggest the need for improved therapeutics. Here, we demonstrate that signal amplification driven by high-affinity EGFR ligands limits the capacity of monoclonal anti-EGFR antibodies to block pathway signaling and cell proliferation and that these ligands are commonly coexpressed with low-affinity EGFR ligands in epithelial tumors. To develop an improved antibody therapeutic capable of overcoming high-affinity ligand-mediated signal amplification, we used a network biology approach comprised of signaling studies and computational modeling of receptor–antagonist interactions. Model simulations suggested that an oligoclonal antibody combination may overcome signal amplification within the EGFR:ERK pathway driven by all EGFR ligands. Based on this, we designed MM-151, a combination of three fully human IgG1 monoclonal antibodies that can simultaneously engage distinct, nonoverlapping epitopes on EGFR with subnanomolar affinities. In signaling studies, MM-151 antagonized high-affinity EGFR ligands more effectively than cetuximab, leading to an approximately 65-fold greater decrease in signal amplification to ERK. In cell viability studies, MM-151 demonstrated antiproliferative activity against high-affinity EGFR ligands, either singly or in combination, while cetuximab activity was largely abrogated under these conditions. We confirmed this finding both in vitro and in vivo in a cell line model of autocrine high-affinity ligand expression. Together, these preclinical studies provide rationale for the clinical study of MM-151 and suggest that high-affinity EGFR ligand expression may be a predictive response marker that distinguishes MM-151 from other anti-EGFR therapeutics. Mol Cancer Ther; 14(7); 1625–36. ©2015 AACR.

Receptor tyrosine kinases (RTKs) are high-affinity cell surface receptors for growth factors that are frequently deregulated in cancer. Signaling through these receptors has been associated with increased cancer cell proliferation and resistance to cytotoxic therapies. To block this detrimental signaling, many companies are developing inhibitory antibodies against various RTKs. A key challenge in clinical studies is the optimal stratification of patients who may benefit from these therapies. For an RTK targeted antibody, the detection of the respective growth factor in the tumor microenvironment may be an important bio-marker. Beyond the physical presence of the growth factor, the decision whether a cancer cell will respond to growth factor-induced signals is governed by complex intra-cellular signaling networks. We compared different approaches to predict cellular responses and will highlight a hybrid approach that combines mechanistic modeling based on ordinary differential equations with a machine learning algorithm. The models are trained on in vitro drug response screens and then applied to predict response in patient samples. The mechanistic models are trained on quantitative data from signal transduction studies as well as RNAseq data for cellular characterization. Using the hybrid approach, a correlation between growth factor expression in the tumor microenvironment and its predicted response was identified. This supports the hypothesis of addiction of tumors to growth factors abundant in the tumor microenvironment, and might enable more robust patient stratification in the future. Citation Format: Helge Hass, Kristina Masson, Sibylle Wohlgemuth, Violette Paragas, John E. Allen, Mark Sevecka, Emily Pace, Jens Timmer, Joerg Stelling, Gavin MacBeath, Birgit Schoeberl, Andreas Raue. Predicting ligand-dependent tumors from multi-dimensional signaling features [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2018; 2018 Apr 14-18; Chicago, IL. Philadelphia (PA): AACR; Cancer Res 2018;78(13 Suppl):Abstract nr 1312.

Although many methods exist to study the recognition and signaling properties of proteins in isolation, it remains a challenge to perform these investigations on a system-wide or proteome-wide scale and within the context of biological complexity. Protein microarray technology provides a powerful tool to assess the selectivity of protein–protein interactions in high-throughput and to quantify the abundances and post-translational modification states of many different proteins in complex mixtures. Here, we provide an overview of the various applications of protein microarray technology and compare the strengths and technical challenges associated with each approach. Overall, we conclude that if this technology is to have a substantial impact on our understanding of cell biology and physiology, increased emphasis must be placed on obtaining rigorously controlled, quantitative data from protein function microarrays and on assessing the selectivity of reagents used in conjunction with protein-detecting microarrays.

As an alternative to conventional, target-oriented drug discovery, we report a strategy that identifies compounds on the basis of the state that they induce in a signaling network. Immortalized human cells are grown in microtiter plates and treated with compounds from a small-molecule library. The target network is then activated and lysates derived from each sample are arrayed onto glass-supported nitrocellulose pads. By probing these microarrays with antibodies that report on the abundance or phosphorylation state of selected proteins, a global picture of the target network is obtained. As proof of concept, we screened 84 kinase and phosphatase inhibitors for their ability to induce different states in the ErbB signaling network. We observed functional connections between proteins that match our understanding of ErbB signaling, indicating that state-based screens can be used to define the topology of signaling networks. Additionally, compounds sort according to the multidimensional phenotypes they induce, suggesting that state-based screens may inform efforts to identify the targets of biologically active small molecules.

Although many anticancer drugs that target receptor tyrosine kinases (RTKs) provide clinical benefit, their long-term use is limited by resistance that is often attributed to increased abundance or activation of another RTK that compensates for the inhibited receptor. To uncover common and unique features in the signaling networks of RTKs, we measured time-dependent signaling in six isogenic cell lines, each expressing a different RTK as downstream proteins were systematically perturbed by RNA interference. Network models inferred from the data revealed a conserved set of signaling pathways and RTK-specific features that grouped the RTKs into three distinct classes: (i) an EGFR/FGFR1/c-Met class constituting epidermal growth factor receptor, fibroblast growth factor receptor 1, and the hepatocyte growth factor receptor c-Met; (ii) an IGF-1R/NTRK2 class constituting insulin-like growth factor 1 receptor and neurotrophic tyrosine receptor kinase 2; and (iii) a PDGFRβ class constituting platelet-derived growth factor receptor β. Analysis of cancer cell line data showed that many RTKs of the same class were coexpressed and that increased abundance of an RTK or its cognate ligand frequently correlated with resistance to a drug targeting another RTK of the same class. In contrast, abundance of an RTK or ligand of one class generally did not affect sensitivity to a drug targeting an RTK of a different class. Thus, classifying RTKs by their inferred networks and then therapeutically targeting multiple receptors within a class may delay or prevent the onset of resistance., W. M. Keck Foundation, National Institutes of Health (U.S.) (R21 CA126720), National Institutes of Health (U.S.) (P50 GM068762), National Institutes of Health (U.S.) (RC1 HG005354), National Institutes of Health (U.S.) (U54-CA112967), National Institutes of Health (U.S.) (R01-CA096504), Alfred and Isabel Bader (Fellowship), Jacques-Emile Dubois (fellowship)

A system-wide analysis of cell signaling requires detecting and quantifying many different proteins and their posttranslational modification states in the same cellular sample. Here, we present Protocols for two miniaturized, array-based methods, one of which provides detailed information on a central signaling protein and the other of which provides a broad characterization of the surrounding signaling network. We describe a bead-based array and its use in characterizing the different forms and functions of β-catenin, as well as lysate microarrays (reverse-phase protein arrays) and their use in detecting and quantifying proteins involved in the canonical and noncanonical Wnt signaling pathways. As an application of this dual approach, we characterized the state of β-catenin signaling in cell lysates and linked these molecule-specific data with pathway-wide changes in signaling. The Protocols described here provide detailed instructions for cell culture methods, bead arrays, and lysate microarrays and outline how to use these complementary approaches to obtain insight into a complex network at a systems level.

Receptor tyrosine kinases (RTKs) process extracellular cues by activating a broad array of signaling proteins. Paradoxically, they often use the same proteins to elicit diverse and even opposing phenotypic responses. Binary, ‘on–off’ wiring diagrams are therefore inadequate to explain their differences. Here, we show that when six diverse RTKs are placed in the same cellular background, they activate many of the same proteins, but to different quantitative degrees. Additionally, we find that the relative phosphorylation levels of upstream signaling proteins can be accurately predicted using linear models that rely on combinations of receptor‐docking affinities and that the docking sites for phosphoinositide 3‐kinase (PI3K) and Shc1 provide much of the predictive information. In contrast, we find that the phosphorylation levels of downstream proteins cannot be predicted using linear models. Taken together, these results show that information processing by RTKs can be segmented into discrete upstream and downstream steps, suggesting that the challenging task of constructing mathematical models of RTK signaling can be parsed into separate and more manageable layers. ### Synopsis Receptor tyrosine kinases (RTKs) constitute a large family of single‐spanning transmembrane proteins found only in Metazoans ([Robinson et al , 2000][1]). Their primary role is to mediate intercellular communication by recognizing extracellular ligands and translating that information into an appropriate cellular response ([Schlessinger, 2000][2]). Ligand binding to the extracellular domain of an RTK induces receptor dimerization and activation of its intracellular kinase domain, which results in phosphorylation of a number of tyrosine residues in its carboxy‐terminal tail. Intracellular signaling proteins that contain Src homology 2 (SH2) or phosphotyrosine‐binding (PTB) domains dock at these sites of tyrosine phosphorylation and initiate a variety of signaling cascades within the cell ([Sadowski et al , 1986][3]; [Kavanaugh and Williams, 1994][4]). Paradoxically, RTKs often use the same signaling pathways to elicit diverse and even opposing phenotypic responses, ranging from adhesion to migration, proliferation to differentiation, and survival to apoptosis ([Fambrough et al , 1999][5]; [Simon, 2000][6]). The ability of RTKs to signal through common pathways, yet induce diverse phenotypic responses, has largely been attributed to differences in cellular context, as signaling proteins are differentially expressed in different cell types ([Jordan et al , 2000][7]; [Simon, 2000][6]). When expressed in the same cellular background, however, different RTKs have also been shown to elicit different phenotypic responses ([Pollock et al , 1990][8]; [Lin et al , 1996][9]). How, then, are intrinsic differences between RTKs manifested within the same cell type, where does the information reside that defines these differences, and how is that information processed? To address these questions, we expressed six diverse RTKs in the same cellular background ([Figure 1A][10]) and monitored their ability to activate downstream signaling pathways. Quantitative immunoblotting was used to measure the relative phosphorylation levels of a wide range of proteins that have previously been implicated in RTK signaling ([Figure 1B and C][10]). In total, we queried 65 sites of phosphorylation on 57 proteins and observed growth factor‐induced phosphorylation of 24 sites on 23 proteins. Each receptor induced a distinct pattern of phosphorylation, and for every site of phosphorylation, quantitative differences were observed across the six cell lines ([Figure 1C][10]). Thus, although these six RTKs activate many of the same signaling proteins, they do so to different quantitative degrees. As RTKs initiate signaling by recruiting proteins to sites of tyrosine phosphorylation ([Schlessinger, 2000][2]), we asked whether there was information in the recruitment properties of the pTyr sites on these receptors that could explain the observed differences in signaling. To address this question, we used protein microarrays to define a quantitative interaction map for each receptor by measuring the affinity of almost every human SH2 and PTB domain for phosphopeptides representing known sites of tyrosine phosphorylation on the six receptors ([Figure 1D][10]) ([Jones et al , 2006][11]). We found that although there is considerable qualitative overlap in the receptors’ recruitment properties, they differ substantially at the quantitative level ([Figure 1E and F][10]). We therefore hypothesized that quantitative differences in recruitment potential could explain the observed differences in signaling elicited by each receptor. Using partial least‐squares regression, we found that the relative phosphorylation levels of upstream signaling proteins can be accurately predicted using linear models that rely on combinations of receptor‐docking affinities, whereas the phosphorylation levels of downstream proteins cannot be predicted using linear models ([Figure 3][12]). Additionally, we found that much of the information needed to predict the relative phosphorylation levels of upstream signaling proteins resides in the number and affinity of PI3K‐ and Shc1‐docking sites on the receptor. Interestingly, when we examined the sequences surrounding all known sites of tyrosine phosphorylation on human RTKs as reported in the Phospho.ELM database ([Diella et al , 2008][13]), we observed a distinct and significant bias for sites that feature consensus binding sequences for the PTB domain of Shc1 and the SH2 domains of PI3K. This bias is not observed in sites of tyrosine phosphorylation derived from all other human proteins. Thus, we find that intrinsic differences between RTKs are manifested in the degree to which they activate upstream signaling proteins and that much of this information resides in the number and affinity of docking sites for PI3K and Shc1. Our results suggest that different RTKs may be able to elicit different phenotypic responses in the same cell type by activating a common set of signaling proteins, but to different quantitative degrees. We propose a model in which information processing by RTKs can be segmented into discrete upstream and downstream layers, and submit that the difficult task of constructing mathematical models of RTK signaling can be parsed into separate problems, with the greatest challenge lying in dissecting the nonlinear layer. Mol Syst Biol. 5: 235 [1]: #ref-16 [2]: #ref-18 [3]: #ref-17 [4]: #ref-10 [5]: #ref-4 [6]: #ref-20 [7]: #ref-8 [8]: #ref-15 [9]: #ref-11 [10]: #F1 [11]: #ref-7 [12]: #F3 [13]: #ref-2

One of the principal challenges in systems biology is to uncover the networks of protein-protein interactions that underlie most biological processes. To date, experimental efforts directed at this problem have largely produced only qualitative networks that are replete with false positives and false negatives. Here, we describe a domain-centered approach--compatible with genome-wide investigations--that enables us to measure the equilibrium dissociation constant (K(D)) of recombinant PDZ domains for fluorescently labeled peptides that represent physiologically relevant binding partners. Using a pilot set of 22 PDZ domains, 4 PDZ domain clusters, and 20 peptides, we define a gold standard dataset by determining the K(D) for all 520 PDZ-peptide combinations using fluorescence polarization. We then show that microarrays of PDZ domains identify interactions of moderate to high affinity (K(D)or = 10 microM) in a high-throughput format with a false positive rate of 14% and a false negative rate of 14%. By combining the throughput of protein microarrays with the fidelity of fluorescence polarization, our domain/peptide-based strategy yields a quantitative network that faithfully recapitulates 85% of previously reported interactions and uncovers new biophysical interactions, many of which occur between proteins that are co-expressed. From a broader perspective, the selectivity data produced by this effort reveal a strong concordance between protein sequence and protein function, supporting a model in which interaction networks evolve through small steps that do not involve dramatic rewiring of the network.

MM-121/SAR 256212 is a fully human anti ErbB3/HER3 antibody that blocks ligand-induced receptor activation. Formation of EGFR/ErbB3 (ErbB1-3) heterodimers has been implicated as a major driver of tumor growth and survival in non-small cell lung cancer (NSCLC). Although erlotinib remains the standard-of-care treatment for patients with EGFR-wild-type NSCLC who have failed platinum combination chemotherapy, clinical benefit is typically modest. To address this unmet medical need, we investigated the combination of erlotinib with MM-121 in pre-clinical models of NSCLC. We initially assembled a panel of 25 EGFR-WT NSCLC cell lines spanning the most common histological subtypes (adenocarcinoma, squamous and large-cell carcinoma). Clinical studies show that tumors harboring activating Ras mutations less commonly respond to ErbB-directed therapies. To explore the effect of Ras mutations on responsiveness to MM-121, we also selected our cell lines to include a variety of H-/K-/N-Ras genotypes. Using a carefully optimized in-vitro 3D culture system, we then measured cell viability in response to MM-121, erlotinib or a combination of the drugs in the presence or absence of exogenously added epidermal growth factor (EGF) and/or heregulin-β1 (HRG). Our results indicate that MM-121 inhibits HRG-driven cell proliferation in the studied cell lines. In the five cell lines exhibiting dual-EGF-HRG-driven cell proliferation, the combination of MM-121 with erlotinib demonstrated superior inhibition of cell viability over erlotinib alone.To assess the ability of MM-121 to inhibit tumor cell growth independent of exogenously supplied HRG, we established mouse xenografts from three of the cell lines and treated mice with erlotinib, MM-121, or a combination of the two. All three in-vivo models showed greatly enhanced inhibition of tumor cell growth compared to erlotinib alone in a manner consistent with our in-vitro results. Analyzing our in vitro data we found that mutations in Ras genes (HRAS, KRAS, NRAS) do not preclude response to MM-121 or incremental benefit from the combination of MM-121 with erlotinib over erlotinib alone. Non-adenocarcinoma origin likewise did not preclude response to MM-121, although a weak trend towards diminished activity in the ten non-adenocarcinoma NSCLC cell lines was apparent. Merrimack Pharmaceuticals and Sanofi are co-developing MM-121 and a Phase 1-2 study of MM-121 in combination with erlotinib is currently enrolling patients with EGFR-wild-type NSCLC. Citation Format: Marisa J. Wainszelbaum, Mark S. Sevecka, Olga Burenkova, Gabriela Garcia, William Kubasek, Gavin MacBeath. In-vitro studies of MM-121/SAR 256212, an anti-ErbB-3 antibody, in combination with erlotinib in EGFR-wild-type NSCLC. [abstract]. In: Proceedings of the 104th Annual Meeting of the American Association for Cancer Research; 2013 Apr 6-10; Washington, DC. Philadelphia (PA): AACR; Cancer Res 2013;73(8 Suppl):Abstract nr 5464. doi:10.1158/1538-7445.AM2013-5464

Lysate microarrays (reverse-phase protein arrays) hold great promise as a tool for systems-level investigations of signaling and multiplexed analyses of disease biomarkers. To date, however, widespread use of this technology has been limited by questions concerning data quality and the specificity of detection reagents. To address these concerns, we developed a strategy to identify high-quality reagents for use with lysate microarrays. In total, we tested 383 antibodies for their ability to quantify changes in protein abundance or modification in 20 biological contexts across 17 cell lines. Antibodies yielding significant differences in signal were further evaluated by immunoblotting and 82 passed our rigorous criteria. The large-scale data set from our screen revealed that cell fate decisions are encoded not just by the identities of proteins that are activated, but by differences in their signaling dynamics as well. Overall, our list of validated antibodies and associated protocols establish lysate microarrays as a robust tool for systems biology.