Mining the cancer genome uncovers therapeutic activity of EphA7 against lymphoma

Abstract

The functional annotation of the cancer genome can reveal new opportunities for cancer therapies. The wealth of genomic data on various cancers has not yet been mined for clinically and therapeutically useful information. We use cross-comparisons of genomic data with the results of unbiased genetic screens to prioritize genomic changes for further study. In this manner, we have identified a soluble variant of the ephrin receptor A7 (EPHA7^TR) as a tumor suppressor that is lost in lymphoma. We also developed antibody-based delivery to restore this tumor suppressor to the cancer cells in situ. We will discuss our strategy of screening genomic data, specific findings concerning EPHA7 and the potential for future discoveries.

Key words:

• genomic data
• lymphoma
• tumor suppressor
• EPHA7

Cancer is a heterogeneous disease characterized by countless genomic alterations. While it has become relatively easy to describe changes in the cancer genomes in great detail, the mining of genomic data for clinically relevant information is only just beginning. We can expect to gain new insight into the molecular origins of cancer and to identify new avenues of cancer therapy. To this end, functional analyses are needed, but—given the wealth of genomic data—where should we begin?

Our laboratory is especially interested in lymphoid cancers, Non-Hodgkin lymphoma and acute lymphatic leukemias (ALL). Great progress has been made in the treatment of childhood ALL, and, similarly, many aggressive forms of Non-Hodgkin lymphoma are largely curable today. This is in contrast to ALL arising in adults (> 35 y of age) and also to so-called indolent lymphomas. These cancers remain incurable except for aggressive transplantation regimens that only a subset of patients are eligible for. Hence, within the category of lymphoid cancers, we focus on these problematic diseases.

Follicular lymphoma (FL) is a common and deadly type of lymphoma. FL is diagnosed in ∼18,000 Americans per year. Clinically, it is characterized by an indolent growth pattern of slow and persistent growth. FLs respond to conventional chemotherapy, but incessant relapses progressively limit marrow and organ function, and frequently, transformation toward a more aggressive cancer leads to patients' demise. Genetically, FLs are characterized by a translocation t(14:18) that activates Bcl2 expression.3,4 Bcl2 blocks cell death and delays cell cycle entry; clearly, additional genetic changes are required for lymphoma development. Unfortunately, a lack of cell lines and murine models of this cancer have hampered research into FL. On the other hand, many excellent cytogenetic studies are available,5–7 and recent sequencing studies have revealed recurrent mutations in several epigenetic regulators, such as MLL2, EZH2, EP300 and CREBBP.8,9 However, the functional consequences of these changes remain to be explored.

Progress in cancer genomics has been immense, and an avalanche of genomic information is published and available. In large part, this reflects advances in sequencing technology but also the ready availability of array comparative genomic hybridization (array CGH), genome-wide methylation profiling and coding and expression data concerning coding and non-coding RNAs. Somatic mutations may be the most readily understood way gene function is affected in cancers. However, alterations in genomic integrity, promoter methylation and microRNA expression affect gene expression and cancer phenotypes. In fact, determining the actual gene dosage in a given cancer is not an easy task and likely requires combining data from different technologies.

Genomic lesions in cancer are increasingly cataloged. As indicated, many different platforms are available to analyze alterations in cancer genomes at high resolution and compare with normal counterparts or similar cancers at different stages. However, these descriptive data do not immediately reveal function, contribution and requirement in cancer cells. Hence, the next challenge is to provide functional annotation of genomic alterations in cancer.

Changes in the cancer genome are often complex and reflect complicated process of malignant transformation. For example, array CGH analyses typically reveal large regions of gains and losses and significant variation between individual cases. Extracting the common patterns is one way to narrow down the list of potential gene targets. Regions of bi-allelic loss provide even more focal information. However, in most cases, even the minimal overlapping regions of change are large, and areas of loss are frequently hemizygous. Hence, often the genomic data alone do not unequivocally identify genetic targets of large-scale changes.

In parallel with advances in genomic technologies, we now have genetic tools to perform large-scale gain- and loss-of-function screens. While long available in simple model organisms like yeast or bacteria, advances in RNAi technology now allows similar genetic approaches in mammalian cells and even in vivo.10–16 Libraries of RNAi constructs, microRNAs or cDNAs are now readily available for gene-by-gene or pooled screening approaches. Clearly, these screening approaches have limitations. For example, RNAi constructs can produce off-target effects, and cDNA expression may cause unphysiological protein levels. Moreover, a screen result—a “hit”—implies biological potential; it does not demonstrate actual relevance to the cancer problem. Hence, results of genetic screens need to be carefully validated and interpreted in the disease context to determine their relevance.

The approach our lab has pursued is to cross-compare cancer genome data with the results of genetic screens (Fig. 1). The rationale is that these are complementary tools: the genetic screen provides a functional filter for complex genome data and indicates potential biological roles. Conversely, evidence that a screen “hit” is also a target of genomic change in cancer provides a first indication of disease relevance. This “functional genomics” approach provides a rapid means to prioritize candidate genes form complex genomic data sets.

In our recent work, we applied a functional genomics approach to follicular lymphoma. Briefly, arrayCGH data revealed large and hemizygous chromosomal losses affecting chromosome 6q12–27 in ∼25% of FL cases. Moreover, patients carrying 6q deletions have a bad prognosis, indicating that this region contributes to lymphoma development and clinical outcome. To identify target genes of 6q loss, we built an RNAi library and performed in vitro screens followed by validation in a new mosaic mouse model of FL based on the vavPBcl2 transgene.17 In short, we confirmed TNFAIP3/A20 as a tumor suppressor in lymphoma; we also identified several additional candidates that localized to hotspots within the 6q region. The presence of multiple target genes within this region had been anticipated by prior cytogenetic studies in reference 7 and may illustrate a biological principle behind large-scale chromosomal changes. We were especially intrigued by the ephrin receptor A7 gene (EPHA7) that produced only a short splice variant, and that can be shed form normal B cells but is lost in the tumors.

Ephrin receptors and their ligands form a large family of receptor tyrosine kinases that are involved in several physiological and pathological processes.18,19 Briefly, members of the ephrin A ligand family bind EPH-A receptors, while ephrin B ligands interact with EPH-B receptors. There is a great diversity of ligands and receptors that have tissue-specific expression patterns and overlapping ligand-receptor specificities. A key feature of these interactions is that their bi-directional signaling is triggered by ligand-receptor interactions and consequent receptor dimerization and the formation of higher-order clusters of activated receptors. Physiological roles of this cell-cell signaling pathway are established in axon guidance, angiogenesis and in developmental contexts, such as neural tube closure.20–22 One may speculate that bi-directional EPH signals have even broader effects in cell-cell contact and tissue organization. EPHA7 is remarkable for its expression of short splice variants that have been reported to act as dominant-negative inhibitors of the full-length receptors.23,24 The role of EPH signaling in cancer is quite unclear, and some reports suggest oncogenic activity for certain receptors, while others may have tumor-suppressive properties.19 Methylation studies have revealed silencing of EPHA7 in B-ALL and also colon and prostate cancers.25–27 Other EPHA receptors, e.g., EPHA2 and EPHA3, have emerged as mutational targets lung cancer and melanoma.28,29 The functional consequence of these mutations has not been explored, although it is noteworthy that overexpression of EPHA2 has been observed in ovarian and breast cancer.30,31 Hence, EPH receptors are recurrent targets in cancer, and both oncogenic and tumor-suppressive functions have been discussed.

The truncated form of EPHA7 (EPHA7^TR) is a soluble tumor suppressor against lymphoma.1 Our detailed studies on EPHA7^TR revealed that its knockdown accelerated lymphoma development in murine model of FL and also Burkitt lymphoma. Concordantly, re-expression or administration of purified EPHA7^TR protein caused regression of xenografted human lymphomas via binding to EPHA2, inhibition of EPHA2 phosphorylation and its downstream ERK and SRC signals. These findings indicate a potential therapeutic application for EPHA7^TR against lymphoma.

Studies in genetically engineered models indicate that restoration of a tumor suppressor can have dramatic therapeutic effects.32,33 The concept has been named tumor suppressor hypersensitivity in analogy to dependence of cancer cells on the continued activity of an oncogene.34,35 While several drugs that can block oncogenic kinases are now available and show great promise in the clinic, the restoration of a tumor suppressor that has been inactivated in cancer cells remains a challenge. Soluble tumor suppressors offer a unique opportunity, because their auto- and paracrine activities can be achieved, at least in principle, by exogenous administration.

The antitumor and signaling properties imply that soluble EPHA7^TR has drug-like and anti-lymphoma activities.1 These effects were readily detected upon local injection into xenografted tumors and could even be elicited by systemic (intravenous administration). To enhance its antitumor activity even further, we fused EPHA7^TR to the anti-CD20 (Rituximab^®) antibody that is already used in lymphoma therapy. This fusion readily localized to CD20-expressing lymphoma cells in vivo, caused a block in EPHA2 and ERK signaling with subsequent growth arrest and cell death induction. Overall, the fusion had superior activity compared with anti-CD20 alone.1 Finally, we observed no frank toxicity either with the pure EPHA7^TR protein or the fusion antibody. Hence, EPHA7^TR-anti-CD20 is an example of a bi-functional therapeutic based on the antibody-guided delivery of a soluble tumor suppressor.

Ephrin signaling is an emerging area in cancer, and many important questions remain. EPHA7^TR acts as a dominant inhibitor by binding to EPHA2 and likely prevents receptor dimerization and formation of active clusters of EPH receptors.1 Hence, soluble EPHA7^TR is a valuable reagent to explore signals emanating from EPH receptors in healthy and cancerous tissues. Ongoing mass spectrometric studies are designed to identify potential additional receptors of EPHA7^TR and shed light on the downstream events triggered by EPHA7^TR binding. The structural details of EPHA7^TR binding to other EPH receptors are unclear, and reconstitution of EPHA7^TR-EPHA2 complexes with rationally designed mutant and truncated alleles will provide new insight. These mechanistic studies will enhance the development of EPHA7^TR-based peptide therapeutics, for example, by defining the minimum peptide sequence required and optimizing structural assembly of antibody fusions.

Does EPHA7^TR have activity against other cancers? EPHA7^TR is a dominant inhibitor of EPHA2 and -A3 receptors. This implies that EPHA7^TR could produce therapeutic effects against tumors that inactivate EPHA7 and those that have increased expression/activity of EPHA2 and -A3. In lymphoma we observed loss of EPHA7 expression by deletion and/or epigenetic silencing in ∼70% of cases. Specifically, EPHA7 is inactivated in FL, in in marginal zone B-cell lymphomas,24 in diffuse large B-cell (DLBCL) and in HIV-associated Burkitt lymphoma (BL).1 Epigenetic silencing of EPHA7 has been reported in adult B-ALL and in solid tumors, including colorectal and prostate carcinoma.25–27 Further, EPHA2 and EPHA3 are targets of genomic change in lung and ovarian cancer28,29 and over-expressed in breast and ovarian cancers. These data provide a rationale to test EPHA7^TR against a spectrum of cancers and derived cell lines.

Outlook—Beyond EPHA7

What determines the actual gene dosage of a tumor suppressor in cancer? Deletions of chromosome 6q (and most other genomic losses) are heterozygous and contain small or no regions of bi-allelic loss. Accordingly, we find that EPHA7 is hemizygously deleted in ∼20% of FLs. However, epigenetic silencing contributes to loss of EPHA7 expression in a cooperative manner in cases with deletions and also occurs in non-6q-deleted cases. The end result is low and near-absent expression of EPHA7 in most FL samples. The genetic concepts of hemi- and homozygosity do not adequately capture these quantitative aspects and suggest that tumor suppressor inactivation comes in three flavors: all, half and zero. Epigenetic silencing and genomic loss are non-mutually exclusive mechanisms of gene inactivation that contribute quantitatively to the final dosage of a tumor suppressor.

What shapes the organization of chromosomal aberrations in cancer? In distinction from somatic mutations the loss of larger genomic regions allows for the disruption of multiple target genes. We identify several known and candidate tumor suppressors that are targeted in a variably overlapping and hemizygous pattern in FL patients. These genes include TNFAIP3, PRDM1 and EPHA7 and others.36–38 Each loss is typically hemizygous, and none is absolutely required for lymphoma development. However, initial studies indicate cooperation between multiple tumor suppressors that are often jointly affected by deletions. Apparently, genes encoded in coherent genomic regions encode facultatively cooperating tumor suppressors. Conversely, the pattern of genomic loss along chromosome 6q reveals areas that are relatively spared from deletion. In part, these regions encode genes essential for cell survival. Hence, in addition to structural features encoded in the DNA sequence,39 the patterns of genomic gains/losses in cancer likely reflect selective pressures for and against specific gene products.

Is there more to be discovered? With over 30,000 publications written about some tumor suppressor genes, one might wonder if there is anything new to be found. The data on various cancer genomes indicate a complex landscape of changes that has occurred in tumor evolution. Functional annotation has, in this case, revealed a surprising new soluble tumor suppressor. Importantly, while the deletion affects multiple tumor suppressor genes, restoration of EPHA7^TR is sufficient to produce therapeutic responses. This indicates that important genetic modifiers of the malignant phenotype are still hidden in the genomic data. Hence, mining the cancer genome has only just begun, and there is great potential for actionable insight and new therapeutic reagents.

Figures and Tables

Figure 1 Cancer genomic analysis and unbiased genetic screens are complimentary tools in cancer gene discovery. Genomic analyses reveal many differences between cancer cells and their normal counterparts. Defining the key “drivers” remains a challenge. On the other hand, genetic screens identify modifiers of biological processes; however, their relevance to cancer is often unclear. Cross-referencing the results from these studies provides a functional filter for genomic data and a first indication of cancer relevance.

Acknowledgements

This work is supported by grants from the NCI (R01-CA142798-01), and a P30 supplemental award (H.G.W.), the Leukemia Research Foundation (H.G.W.), the Louis V. Gerstner Foundation (H.G.W.), the WLBH Foundation (H.G.W.), the Society of MSKCC (H.G.W.), the Starr Cancer Consortium grant I4-A410 (H.G.W.), Lymphoma Research Foundation Fellowship (E.O.).