Systems biology driving drug development: from design to the clinical testing of the anti-ErbB3 antibody seribantumab (MM-121).

The ErbB family of receptor tyrosine kinases comprises four members: epidermal growth factor receptor (EGFR/ErbB1), human EGFR 2 (HER2/ErbB2), ErbB3/HER3, and ErbB4/HER4. The first two members of this family, EGFR and HER2, have been implicated in tumorigenesis and cancer progression for several decades, and numerous drugs have now been approved that target these two proteins. Less attention, however, has been paid to the role of this family in mediating cancer cell survival and drug tolerance. To better understand the complex signal transduction network triggered by the ErbB receptor family, we built a computational model that quantitatively captures the dynamics of ErbB signaling. Sensitivity analysis identified ErbB3 as the most critical activator of phosphoinositide 3-kinase (PI3K) and Akt signaling, a key pro-survival pathway in cancer cells. Based on this insight, we designed a fully human monoclonal antibody, seribantumab (MM-121), that binds to ErbB3 and blocks signaling induced by the extracellular growth factors heregulin (HRG) and betacellulin (BTC). In this article, we present some of the key preclinical simulations and experimental data that formed the scientific foundation for three Phase 2 clinical trials in metastatic cancer. These trials were designed to determine if patients with advanced malignancies would derive benefit from the addition of seribantumab to standard-of-care drugs in platinum-resistant/refractory ovarian cancer, hormone receptor-positive HER2-negative breast cancer, and EGFR wild-type non-small cell lung cancer (NSCLC). From preclinical studies we learned that basal levels of ErbB3 phosphorylation correlate with response to seribantumab monotherapy in mouse xenograft models. As ErbB3 is rapidly dephosphorylated and hence difficult to measure clinically, we used the computational model to identify a set of five surrogate biomarkers that most directly affect the levels of p-ErbB3: HRG, BTC, EGFR, HER2, and ErbB3. Preclinically, the combined information from these five markers was sufficient to accurately predict which xenograft models would respond to seribantumab, and the single-most accurate predictor was HRG. When tested clinically in ovarian, breast and lung cancer, HRG mRNA expression was found to be both potentially prognostic of insensitivity to standard therapy and potentially predictive of benefit from the addition of seribantumab to standard of care therapy in all three indications. In addition, it was found that seribantumab was most active in cancers with low levels of HER2, consistent with preclinical predictions. Overall, our clinical studies and studies of others suggest that HRG expression defines a drug-tolerant cancer cell phenotype that persists in most solid tumor indications and may contribute to rapid clinical progression. To our knowledge, this is the first example of a drug designed and clinically tested using the principles of Systems Biology.

A systematic family-wide investigation reveals that ~30% of mammalian PDZ domains engage in PDZ-PDZ interactions.

PDZ domains are independently folded modules that typically mediate protein-protein interactions by binding to the C termini of their target proteins. However, in a few instances, PDZ domains have been reported to dimerize with other PDZ domains. To investigate this noncanonical-binding mode further, we used protein microarrays comprising virtually every mouse PDZ domain to systematically query all possible PDZ-PDZ pairs. We then used fluorescence polarization to retest and quantify interactions and coaffinity purification to test biophysically validated interactions in the context of their full-length proteins. Overall, we discovered 37 PDZ-PDZ interactions involving 46 PDZ domains (~30% of all PDZ domains tested), revealing that dimerization is a more frequently used binding mode than was previously appreciated. This suggests that many PDZ domains evolved to form multiprotein complexes by simultaneously interacting with more than one ligand.

Therapeutically targeting ErbB3: a key node in ligand-induced activation of the ErbB receptor-PI3K axis.

The signaling network downstream of the ErbB family of receptors has been extensively targeted by cancer therapeutics; however, understanding the relative importance of the different components of the ErbB network is nontrivial. To explore the optimal way to therapeutically inhibit combinatorial, ligand-induced activation of the ErbB-phosphatidylinositol 3-kinase (PI3K) axis, we built a computational model of the ErbB signaling network that describes the most effective ErbB ligands, as well as known and previously unidentified ErbB inhibitors. Sensitivity analysis identified ErbB3 as the key node in response to ligands that can bind either ErbB3 or EGFR (epidermal growth factor receptor). We describe MM-121, a human monoclonal antibody that halts the growth of tumor xenografts in mice and, consistent with model-simulated inhibitor data, potently inhibits ErbB3 phosphorylation in a manner distinct from that of other ErbB-targeted therapies. MM-121, a previously unidentified anticancer therapeutic designed using a systems approach, promises to benefit patients with combinatorial, ligand-induced activation of the ErbB signaling network that are not effectively treated by current therapies targeting overexpressed or mutated oncogenes.

PDZ domain binding selectivity is optimized across the mouse proteome.

PDZ domains have long been thought to cluster into discrete functional classes defined by their peptide-binding preferences. We used protein microarrays and quantitative fluorescence polarization to characterize the binding selectivity of 157 mouse PDZ domains with respect to 217 genome-encoded peptides. We then trained a multidomain selectivity model to predict PDZ domain-peptide interactions across the mouse proteome with an accuracy that exceeds many large-scale, experimental investigations of protein-protein interactions. Contrary to the current paradigm, PDZ domains do not fall into discrete classes; instead, they are evenly distributed throughout selectivity space, which suggests that they have been optimized across the proteome to minimize cross-reactivity. We predict that focusing on families of interaction domains, which facilitates the integration of experimentation and modeling, will play an increasingly important role in future investigations of protein function.

Effects of HER2 overexpression on cell signaling networks governing proliferation and migration.

Although human epidermal growth factor receptor 2 (HER2) overexpression is implicated in tumor progression for a variety of cancer types, how it dysregulates signaling networks governing cell behavioral functions is poorly understood. To address this problem, we use quantitative mass spectrometry to analyze dynamic effects of HER2 overexpression on phosphotyrosine signaling in human mammary epithelial cells stimulated by epidermal growth factor (EGF) or heregulin (HRG). Data generated from this analysis reveal that EGF stimulation of HER2-overexpressing cells activates multiple signaling pathways to stimulate migration, whereas HRG stimulation of these cells results in amplification of a specific subset of the migration signaling network. Self-organizing map analysis of the phosphoproteomic data set permitted elucidation of network modules differentially regulated in HER2-overexpressing cells in comparison with parental cells for EGF and HRG treatment. Partial least-squares regression analysis of the same data set identified quantitative combinations of signals within the networks that strongly correlate with cell proliferation and migration measured under the same battery of conditions. Combining these modeling approaches enabled association of epidermal growth factor receptor family dimerization to activation of specific phosphorylation sites, which appear to most critically regulate proliferation and/or migration.

Uncovering quantitative protein interaction networks for mouse PDZ domains using protein microarrays.

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.