1(a). Analyses of Nedd9^-/- in the MMTV-PyVT model for mammary tumors.

Generation of mammary tumors in the mouse mammary tumor virus (MMTV)-polyoma virus middle T (PyVT) genetic model is rapid and highly penetrant, with average latency of tumor appearance ~ 51 days affecting 100% of mice.  We first showed the appearance of these tumors is delayed by a Nedd9^-/- genotype, and signaling by NEDD9 interacting partners considerably depressed (Izumchenko et al, Cancer Res 61:7198-206, 2009). Subsequently (Singh et al, Cancer Res, 2010 in press), we have now showed that despite the initial delay in tumor growth, cells derived from MMTV-PyVT;Nedd9^-/- tumors are characteristically hyperaggressive versus MMTV-PyVT;Nedd9^-/- cells in anchorage-independent growth in vitro, and in formation of tumors after mammary orthotopic reinjection, and of lung metastases after tail vein injection. MMTV-PyVT;Nedd9^-/- cells are also characterized by  increased cell cycle, centrosomal, and mitotic defects, accompanied by loss of expression of the Nedd9 partner and mitotic regulator Aurora-A. Intriguingly, in spite of their aggressive phenotype, MMTV-PyVT;Nedd9^-/- cells are hypersensitive to the Src kinase inhibitor dasatinib. Further efforts to examine Nedd9 as a candidate gene that modulates response to dasatinib and potentially other Src kinase inhibitors in clinical development, and unraveling mechanistic insights of such processes are ongoing (see work by Ratushny, below).

Top

1(b). Incidence of HER2-associated breast cancer is greatly reduced in mice lacking NEDD9.

Inappropriately high levels of ERBB2/HER2, a growth factor receptor, drive 30% of human breast cancers.  To further investigate the role of NEDD9 in breast cancer, we have mated mice lacking NEDD9 (Nedd9^-/-) with mice that transgenically express high levels of ERBB2/HER2/neu; the most physiological model for the human disease.  In this long-latency model of breast cancer, mice lacking NEDD9 are dramatically resistant to formation of tumors.  By 24 months, 88% of MMTV-neu;Nedd9^+/+ mice have developed mammary tumors, but only 29% of MMTV-neu;Nedd9^-/-mice.  We are further characterizing differences between MMTV-neu;Nedd9^+/+ and MMTV-neu;Nedd9^-/- tumors using primary cell lines derived directly from primary tumors for various phenotypic assays of metastasis, growth, and signaling. Thus far, analyses with MMTV-neu cell lines indicate little difference in these secondary tumor phenotypes between Nedd9 genotypes. These data combined indicate NEDD9 is required in initiation of neu-driven mammary tumors, but not required for downstream tumor progression or metastasis. Therefore, we are now focusing on identifying how Nedd9 status may affect the progenitor populations of mammary epithelial cells in MMTV-neu;Nedd9^+/+ and MMTV-neu;Nedd9^-/- animals.

Our hope is that current studies investigating a novel role for NEDD9 in mammary tumor initiation may reveal points of breast cancer cell vulnerability that can be therapeutically exploited. We are working to determine whether NEDD9 expression status conditions the ability of breast cancers to be treated by various chemotherapeutics so that, in the future, NEDD9 may be used as a biomarker for use in selecting patients for individualized treatment.

Top

1(c). Unraveling roles for Nedd9 and related proteins in organismal development.

NEDD9 is one of a 4-member family of proteins known collectively as the Cas proteins (Figure 1, Four evolutionary conserved branches of the Cas family). The study of the biological consequences of deleting individual Cas proteins in mammals is complicated by the fact that there are 3 remaining paralogous members of the family, which may partially mask phenotypes, particularly during development. However, Drosophila melanogaster has only one Cas gene, known as Dcas, which makes it a convenient genetic system to study developmental roles.

We have discovered (Tikmyanova et al, PLoS ONE, 2010) that loss of Dcas causes about 10% of flies to die as embryos from a prominent morphogenic defect.  Interestingly, we found that Dcas is an important modulator of the developmental phenotype severity of mutations affecting integrins (If and mew) and their downstream effectors Fak56D or Src42A.  The embryonically lethal Fak56D-Dcas double mutant embryos had extensive cell polarity defects, including mislocalization and reduced expression of E-cadherin. Further genetic analysis established that loss of Dcas modified the embryonal lethal phenotypes of embryos with mutations in E-cadherin (Shg) or its signaling partners p120- and β-catenin (Arm). We also found that Dcas specifically interacts with Shg, and loss of Dcas upregulates total levels of E-cadherin, but inhibits its ability to localize to cell-cell junctions, thereby influencing cell morphology and polarity. These signaling relations are summarized in Figure 2, Cas cooperates with Src/FAK to control cell-cell adhesion complex.

Human orthologs of integrins, Fak56D (FAK), and Src42A (SRC) are known to be important partners for Cas proteins in human cells and cancer progression, suggesting our Drosophila findings might be more broadly generally relevant in these contexts. Using the human breast adenocarcinoma cell line MCF7, we found two Cas proteins, BCAR1 and NEDD9, cooperate with Src to negatively regulate E-cadherin at the plasma membrane by degrading it through lysosomal pathway. Loss of E-cadherin from plasma membrane is a predictor of metastatic cancer progression and a poor prognosis factor. Our findings in Drosophila and human cells suggest a new mechanism that may help explain how upregulation of Cas proteins such as NEDD9 contributes to cancer progression.

Top

2(a). Roles for Aurora-A and NEDD9/HEF1 in Polycystic Kidney Disease and Calcium Signaling

Among its important partner proteins, NEDD9 interacts with the Aurora-A kinase (AurA).  In the past decade, AurA has attracted increasing attention because it is overexpressed and hyperactivated in many tumors arising in breast, colon, ovary, and other tissues. Abnormally high AurA activity causes defective cytokinesis and aneuploidy, which contribute to its tumor-inducing activity. In normal cells, AurA is most abundant at the centrosome in late G2 phase and early M phase, and activation of AurA helps time mitotic entry and progression. Activation of AurA at mitosis is supported by a complex set of interactions between the protein and a number of partners, including NEDD9. AurA is now being actively exploited as a target for development of new anti-cancer agents, based on this biological profile.  

Interestingly, a number of studies have emerged in recent years to challenge the idea that AurA is solely a mitotic kinase even in normal cells. One structure where our group identified a non-mitotic activity for AurA is at the cilium, a small antenna-like structure protruding from the cell surface that receives and interprets mechanical and soluble signals that influence cell growth. We have shown that serum induces AurA activation at the basal body of the cell cilium in non-cycling G0/G1 phase mammalian cells, causing AurA-dependent ciliary resorption (Pugacheva et al, Cell 129:1351-63, 2007), and hence indirectly impacting the functionality of cilia-dependent signaling cascades relevant to cancer and other diseases, including polycystic kidney disease (PKD) (Plotnikova et al, Cancer Res 68:2058-61, 2008).

We have continued to study involvement of AurA in control of cilia and cellular calcium homeostasis in interphase cells and in the development of PKD, common hereditary kidney disease characterized by progressive cyst formation and ultimate loss of renal function. This work has identified direct connections between AurA activity and activity of the polycystin-2 calcium channel, which is encoded by the PKD2 gene, and is often mutated in PKD.   This work has recently identified AurA itself as a direct target of calcium-dependent activation, and defined a calcium-calmodulin switch that controls AurA activity (Plotnikova et al, Nat Commun 1 : 64 doi: 10.1038/ncomms1061), and provides new insight into its activation in both interphase and mitotic cells, under normal and pathological growth conditions (Figure 3, Calcium activates aurora kinase). We have also determined that AurA and its partner and interactor Nedd9 are both extremely abundant in kidney tissue, and the level of AurA is elevated in cells lining PKD-associated renal cysts. We are now using a PKD1-floxed mouse model to investigate the effect of modulating the function of AurA and NEDD9 on renal cyst formation in PKD.

Top

2(b). Exploring therapeutic synergies by inhibiting the two NEDD9 partners, Aurora-A and Src.

Increased activity of SRC kinase promotes tumor invasion and metastasis, and overexpression of the mitotic regulator Aurora kinase A (AurA) drives tumor aneuploidy and chromosomal instability, nominating these proteins as valuable therapeutic targets for cancer and leading to the development of clinical inhibitors targeting these two kinases. Src and AurA share a common interaction partner, NEDD9, suggesting both may function in close proximity and reciprocally influence functionality. We have demonstrated potent synergy between inhibitors of AurA and inhibitors of Src in multiple ovarian and colorectal cancer cell lines but not in normal ovarian epithelial cell lines.  We have showed that the combination of AurA and Src inhibitors selectively kills cells that have undergone aberrant mitosis.  We are currently studying the mechanism underlying the drug synergy and the cell killing.  In doing so we have demonstrated that there is a physical interaction and mutual cross-phosphorylation between Src and AurA, in vitro, that leads to the hyperactivation of both kinases.  We are investigating whether this interaction exists in vivo and whether it is mediated by other binding partners such as NEDD9. We have also shown that the cell killing is associated with a post-mitotic reattachment defect, and that the combined inhibition of AurA and Src synergistically depresses Src activation.  We are interested in discovering whether the combination of AurA and Src inhibition affects other Src-related functions such as motility and invasion.

Top

3(a). Integrating informatics and RNAi to improve response to EGFR-targeting therapeutics

Drug resistance is a common problem in cancer treatment. Part of our work seeks to better understand the factors leading to drug resistance, with the goals of reversing these resistance mechanisms to improve cancer therapies, and of better stratifying patients as likely or unlikely to respond to specific treatments. For this work, we are applying bioinformatics approaches coupled with siRNA screening to identify resistance genes, focusing on the epidermal growth factor receptor (EGFR) signaling network (Figure 4).

In brief overview, hyperactivity of a central EGFR-Ras-MAPK signaling axis is an essential aspect of the growth, metastasis, and drug resistance of many types of tumor. Given the importance of increased EGFR signaling in many tumor types, a number of clinical agents have been developed that bind and block the activity of EGFR and its family members (e.g, ErbB2/HER2). These agents include antibodies (cetuximab, panitumumab), and small molecules (erlotinib, lapatinib). Although some of these agents are emerging as effective in a subset of patients, a larger group of patients with apparently comparable derangement of EGFR do not respond to treatment.

The primary overall goal of this project is to better understand why some tumors respond effectively to drugs targeting the EGFR pathway while others are resistant, and then to exploit this knowledge to improve breast cancer treatment. We propose that non-responsive cancers contain additional malignant changes (either genetic mutations or epigenetic modifications influencing gene expression) that provide alternative signaling routes to downstream essential EGFR signaling targets. Identifying these "rescue" routes would allow the design of combined therapeutic approaches in which both EGFR and the rescue pathway were simultaneously inhibited, improving clinical response.

We have based our project on several key observations. First, systems biology studies in model organisms have begun to establish that synthetic lethal relationships commonly involve genes that are involved in redundant, parallel pathways, or are vertically linked in the same pathway. Further, in the design of combination therapies in the clinic, the selection strategy for drugs to combine frequently involves common principles: that is, identifying two drugs that 1) inhibit the same target, 2) inhibit functionally linked and/or semi-redundant targets, or 3) inhibit vertically linked targets. Together, these observations suggest that generation of a mid-throughput siRNA library that is large enough to represent genes functionally linked to the drug target by no more than 2-3 degrees of separation, and small enough to be probed at low cost under multiple informative experimental conditions will greatly increase the useful "hit" rate for genes that chemosensitize EGFR family-targeted therapies.

We extensively mined public access databases containing information about protein-protein interactions and mRNA

expression profiles generated in humans (Figure 5, Design of a targeted library). We extracted a set of proteins that either directly bind EGFR and its proximal effectors, or are purified in complexes including EGFR; a set of genes transcriptionally upregulated by EGFR-pathway stimulation and downregulated by EGFR inhibition; and a set of genes otherwise involved in EGFR signaling. We also incorporated data generated from genetic interactions reported in model organisms, for strongly conserved evolutionary orthologs of genes in this pathway. We identified a core high value set of genes that fell into at least two of these linkage categories, and based on other weighting functions. We note, in spite of such linkage, the vast majority of these genes have never been tested for ability to modulate EGFR family-targeted therapies.

We have completed reiterative screens of a library representing such selected genes in a A431 cell line, and identified a set of 61 genes which when knocked down sensitized cells to EGFR-targeting agents (Astsaturov et al, Science Signaling 2010). The majority of the sensitizing genes (48/61) encoded proteins that physically interacted in a sub-network. We further investigated if these genes also influenced the sensitivity of other cancer cell lines to various drugs. siRNAs targeting 45 of these genes were profiled for the efficacy in sensitizing 7 other cell lines to erlotinib, cetuximab (an EGFR function-blocking antibody), or unrelated cytotoxic drug CPT11. Data were remapped to the network to reveal clusters that are particularly rich in sensitizing hits.

We have explored the potential for a clinical exploitation of our findings by combining the treatment with EGFR inhibitors with several of clinically available drugs inhibiting the candidate proteins. This study showed synergistic effect in both reducing tumor size in mouse xenografts and cell viability of multiple cancer cell lines, suggesting that a network-centered approach may be fruitful for development of rationally designed combination therapies. The further investigation and exploitation of the drug-sensitizing network is a major laboratory focus for 2010-2011.

Top

3(b). A Systems Biology Approach to Identifying the Molecular Determinants of Cell Survival in Breast Cancer

Estrogen receptor (ER) status is a key determinant in the treatment of breast cancer.  Currently, there are several effective hormonal therapies, targeting ER signaling, used in the clinic.  However, intrinsic and acquired resistance to the current therapies remains an issue.  Identification of novel drug targets for this cohort is needed.  A functional genomics approach to this problem will allow us to screen for potential drug targets by associating gene knockdown with growth inhibition.  

Based on our success using a similar approach for study of resistance to EGFR-targeted drugs, we hypothesized that screening a mid-throughput siRNA library functionally enriched for ER signaling will identify genes essential for breast cancer cellular viability.  Further, we believe that this approach will allow us to differentially identify genes important in resistance to current therapies through the use of breast cancer cell line models with varying drug sensitivities.  Our long term goal is to identify targets for drug development.
An siRNA library has been developed from 631 genes that make up the ER-focused network to screen breast cancer cell lines.  Our primary goal is to identify networks of candidate genes that differentially affect the intrinsic viability of these cell lines.  We have used MCF7 (E2 dependent/ tamoxifen sensitive/ fulvestrant sensitive) MCF7/LCC1 (E2 independent/ tamoxifen sensitive/ fulvestrant sensitive), MCF7/LCC2 (E2 independent/ tamoxifen resistant/ fulvestrant sensitive) and MCF7/LCC9 (E2 independent/ tamoxifen cross-resistant/ fulvestrant resistant) cell lines as syngeneic models with varying E2 dependence and drug sensitivity profiles for these experiments.  The results from these screens should illuminate novel targets for growth inhibition and mechanisms of resistance to hormonal therapy in breast cancer.  

Top