Cancer of the ovary is the leading cause of death for women with gynecologic malignancies. Over 90% of ovarian cancers derive from malignant transformation of the ovarian surface epithelium (OSE), from which cells disseminate into the peritoneal cavity, invade locally or spread via lymphatics, resulting in ascites formation and abdominal distension. Whereas early-stage ovarian cancer is curable in >90% of patients, it is asymptomatic, and most patients are diagnosed with advanced disease that carries a poor prognosis. Current tools for early detection of ovarian cancer are limited: neither ovarian palpation, transvaginal ultrasonography nor serum CA125 levels have sufficient sensitivity or specificity for general screening. Among candidate biomarkers for ovarian cancer are MUC1, claudin-3, or lysophosphatidic acid (LPA), which may be elevated in malignant ascites or even serum of ovarian cancer patients. Experimentally, LPA has been shown to stimulate mitogenic signaling cascades (Fig. 1), which may be mediated, at least in part, through metalloprotease-induced cleavage of EGF-like growth factors, notably heparin-binding epidermal growth factor–like growth factor (HB-EGF), that activate the epidermal growth factor receptor (EGFR) and downstream signaling pathways, a phenomenon termed transactivation (ref. 1; Fig. 2). The pathophysiologic role of LPA or EGF-like ligands and their receptors in ovarian carcinogenesis is poorly understood. In this issue of Clinical Cancer Research, Tanaka et al. have extended their previous studies on the expression and clinical significance of molecules involved in signaling through the LPA and EGFR axes (2–4) in ovarian cancer. Together, their findings suggest a pathophysiologic role for LPA-induced a disintegrin and metalloprotease-17 (ADAM-17/TACE)–mediated HB-EGF cleavage and EGFR transactivation in ovarian cancer, identifying HB-EGF as a prognostic marker and promising therapeutic target (5). This commentary will discuss the biology of both LPA and EGFR-associated signaling in ovarian physiology and cancer from the perspective of EGFR transactivation.
LPA has emerged as an important intercellular signaling molecule implicated not only in physiologic processes such as brain development or angiogenesis but also in the pathophysiology of cancer promotion (6, 7). Being generated through hydrolysis of lysophosphatidyl choline by lysophospholipase D/autotaxin or via hydrolysis of phosphatidic acid by phospholipase A2 or A1, LPA is present and bioactive in numerous extracellular fluids, including serum, ascites or malignant effusions. Binding of LPA to one of at least four different heptahelical transmembrane G protein–coupled receptors (GPCR; LPA1/endothelial differentiation gene (Edg)2, LPA2/Edg4, LPA3/Edg7, and LPA4/GPR23/P2Y9) results in activation of at least three distinct G protein subfamilies (Gq, Gi, G12/13) and initiation of multiple signaling pathways, including Ras/Raf/mitogen-activated protein kinase, phosphoinositide-3-kinase/Akt, phospholipase C/protein kinase C, or RhoA small GTPase signaling (Fig. 1). Subsequent activation of cell surface metalloproteases such as members of the ADAM family may induce cleavage of EGF-like ligand precursors and autocrine or paracrine stimulation of the human EGFR (HER/erbB) family of transmembrane tyrosine kinases (Fig. 2). This transactivation process seems to involve multiple signaling pathways, including mitogen-activated protein kinase (p38 and p44/42), protein kinase C or c-Src, and may be amplified through additional mechanisms, such as LPA-induced production of interleukin-8. Moreover, there seems to exist a signal amplification loop through activation of protein kinase C, which up-regulates LPA production. Thus, LPA induces cancer cell proliferation, survival, drug resistance, invasion, opening of intercellular tight junctions and gap junction closure, cell migration or metastasis (Fig. 1). Membrane-bound lipid ectophosphatases of the LPP family rapidly inactivate LPA by converting it to monoacylglycerol. LPA signaling may further be attenuated through receptor desensitization and ligand-induced receptor internalization, although the fate of internalized LPA receptors—degradation versus recycling—remains to be elucidated (7). Recently, LPA has also been suggested as an intracellular messenger. Although glycerolipid synthesis at the endoplasmatic reticulum or the outer mitochondrial membrane continuously involves LPA production, the kinetics of further acylation to phosphatidic acid are high, and LPA accumulation seems to be negligible at these sites. More interestingly, LPA has been shown to bind the lipid-inducible transcription factor PPARγ and induce vascular remodeling, suggesting that LPA actions may in part be independent of classical LPA receptors (Fig. 2). However, it remains unclear how the charged phospholipid LPA may traverse the plasma membrane and translocate into the nucleus in vivo, or conversely if PPARγ is activated in an indirect manner.
Epidermal Growth Factor Receptors and their Cognate Ligands
EGFR signaling is involved in the regulation of a myriad of biological and pathophysiologic processes including proliferation, differentiation, apoptosis, angiogenesis, or metastasis and is a convergence point for diverse signaling pathways (8). Overactivity of the receptor tyrosine kinase is considered a hallmark of malignancy and has been associated with poor prognosis for cancer patients. Ligands of the EGF superfamily are synthesized as transmembrane precursors, from which mature growth factors are released through metalloprotease-mediated cleavage of the ectodomain (9). These ligands bind to and activate one or more of the four human EGFRs designated ErbB-1 (EGFR), ErbB-2 (HER-2/neu), ErbB-3, and ErbB-4, resulting in receptor homo- and heterodimerization, a necessary prelude to receptor phosphorylation and activation of three major intracellular pathways: Ras/Raf/mitogen-activated protein kinase, phosphoinositide-3-kinase/Akt, and phospholipase C/protein kinase C. Additionally, multiple other signaling molecules are stimulated; these include c-Src, the STAT family of transcription factors as well as Rho, Rac and Cdc42 small GTPases (ref. 10; Fig. 2).
Heparin-binding Epidermal Growth Factor
The EGF-like ligand HB-EGF is expressed in a wide variety of cells and has been implicated in numerous physiologic or pathologic processes, including malignant tumor growth, where its expression has been correlated with survival in various cancers of epithelial origin. Although membrane-bound ligands can activate EGFR in adjacent cells and induce proliferation in a process called juxtacrine stimulation, metalloprotease-induced shedding seems to be critical for the majority of EGFR-related processes, including GPCR-mediated EGFR transactivation (11). Recent results confirm distinct roles for soluble versus membrane-associated HB-EGF. Whereas the membrane-anchored proform promotes cell interactions and decreases migration, soluble HB-EGF induces opposite effects, promoting a transformed phenotype in terms of proliferative rates, colony-forming ability, activation of the cyclin D1 promoter or induction of matrix metalloproteases (MMP). Likewise, LPA-induced ectodomain shedding of HB-EGF has been shown to promote tumor formation of ovarian cancer cells in nude mice (5). The regulated proteolytic cleavage of proHB-EGF yields at least two fragments: the amino-terminal soluble ligand for EGFR as well as the carboxyl-terminal cell-associated remnant, HB-EGF-c. The latter has recently been shown to translocate to the nucleus, where it interacts with the promyelocytic leukemia zinc finger (PLZF) transcriptional repressor, resulting in CRM1-dependent nuclear export and functional loss of PLZF. Thus, HB-EGF shedding results in dual intracellular signaling that is in part independent of EGFR activation (Fig. 2).
Heparin-binding Epidermal Growth Factor Cleaving Enzymes
Proteolytic processing of EGFR ligands and their receptors is a key regulatory switch in EGFR signaling (12). Enzymes implicated in HB-EGF shedding seem to be cell/tissue type-, localization- and/or stimulus-specific; they include MMP-3/stromelysin-1, MMP-7/matrilysin, and the ADAM family members ADAM-9/meltrin-γ, ADAM-10/kuzbanian, ADAM-12/meltrin-α, and ADAM-17/TACE. ADAMs may be activated through removal of their pro-domain by a furin-type pro-protein convertase or through autocatalysis in the trans-Golgi network; additionally, proteins interacting with the cytoplasmic tails of ADAMs may influence their activity. To date, bona fide enzymes implicated in LPA-induced HB-EGF cleavage are ADAM-10 and ADAM-17 (13, 14). Importantly, tissue inhibitors of metalloproteinases can counteract the catalytic activity of various ADAMs, and the balance between the level of active metalloproteases and tissue inhibitors of metalloproteinases may determine the net metalloprotease activity in a cellular microenvironment (15).
Epidermal Growth Factor Receptor Transactivation
Metalloprotease-dependent release of EGF-like ligands may transactivate the EGFR in response to multiple stimuli, including cytokines, osmotic stressors, phorbol 12-myristate 13-acetate or GPCR activation (1). Besides LPA, numerous other GPCR agonists have been implicated in this process (including angiotensin-II or insulin-like growth factor I), and a role for multiple metalloproteases (including ADAM-9, -10, -12, -15, -17, or MMP-9) and EGF-like ligands (predominantly HB-EGF but also transforming growth factor-α or amphiregulin) has been suggested (16). Multiple signaling pathways seem to control ligand shedding in response to diverse stimuli, including p38 and p44/42 mitogen-activated protein kinase, protein kinase C or c-Src. A major signaling route of LPA is Gi-mediated activation of Ras and the downstream Raf/mitogen-activated protein kinase cascade in a tyrosine kinase-dependent fashion (Fig. 1). The precise role of receptor tyrosine kinases in LPA receptor-mediated Ras activation, however, is controversial with recent evidence suggesting that these receptor types may independently control distinct signaling cascades leading to Ras activation (17). Indeed, EGFR downstream pathways overlap with prototypical LPA signal transduction, and cross-talk between these cascades may occur at various levels, which may or may not involve ligand-induced receptor transactivation (Fig. 2).
Lysophosphatidic Acid and Epidermal Growth Factor Receptor Signaling Axes in Ovarian Cancer—Therapeutic Targets?
LPA and LPA receptors. In the normal ovary, LPA is present and active in follicular fluid, suggesting its local synthesis and physiologic function. LPA had first been related to carcinomatous growth in the mid-1990s, when excess concentrations were found in ascites of ovarian cancer patients. Although the mechanism of LPA production in ovarian cancer has not been fully identified, both lysophospholipase D/autotaxin and phospholipase A2 seem to play a role. Autotaxin and its major substrate LPC are widely produced in the human body and are abundantly present in plasma. Both are also generated in excess by a variety of human cancers and have been found in the supernatant of transformed cells, including ovary-derived cells, most likely as part of secreted microvesicles. Thus, not surprisingly, mammalian serum contains high levels of bioactive LPA. Several studies have reported 10-fold increases of serum-LPA in ovarian cancer patients compared with healthy subjects or other cancer conditions, suggesting its use as a biomarker for ovarian cancer screening, early detection, or prognosis prediction. Discrepancies between these and opposite findings that fail to establish a correlation between LPA plasma levels and carcinomatous growth may be attributable in part to differences in sample handling and/or the method of detection employed. Whether accumulation of LPA in the tumor microenvironment constitutes the major mechanism through which autotaxin might promote tumor growth and angiogenesis or whether autotaxin might exert LPA-independent effects in vivo remains to be determined. LPA alone exerts similar trophic effects as total ascites fluid from ovarian cancer patients when added to tumor cells, inducing proliferation, survival, metastasis, or cisplatinum resistance. Conversely, expression of a lipid phosphatase that hydrolyzes LPA may reduce survival, growth and tumorigenesis of ovarian cancer cells (18). Interestingly, whereas normal OSE cells only express significant amounts of the LPA receptors LPA1/Edg2 and LPA4/GPR23, LPA2/Edg4 and LPA3/Edg7 are abnormally expressed in ovarian cancer cells, suggesting differential functions for these receptors—alone or in combination—in ovarian physiology and pathology. Recently, LPA has been shown to induce ectodomain shedding of HB-EGF essential for tumor formation of ovarian cancer cells in nude mice (5), underscoring the role of this lysophospholipid in ovarian carcinogenesis.
Metalloproteases. ADAMs and MMPs are widely expressed in the human body and have been found up-regulated or decreased in various cancers; both metalloprotease families may be induced by LPA. Dynamic equilibrium between metalloproteases and their inhibitors (tissue inhibitors of metalloproteinases) plays an important role in ovarian physiology (follicular growth and ovulation), and ovarian cancer cells may exploit that proteolytic potential to promote degradation of extracellular matrix components and release membrane-anchored growth factors, their receptors or adhesion molecules. Metalloproteases may thus play a role in both, initiation or termination of mitogenic signaling (19). Interestingly, metalloproteinases may be up-regulated by expression of their substrates or related enzymes (e.g., overexpression of HB-EGF or MMP-7 may induce MMP-3; ref. 20). The concomitant up-regulation of ADAM-17 and its substrate HB-EGF in ovarian cancer observed by Nakano's group in this issue of Clinical Cancer Research (2) is intriguing and warrants further validation at a cell-biological level.
EGF-like ligands and their receptors. Numerous tumors overexpress the EGFR, its heterodimeric partner ErbB-2 and/or its ligands, including up to 75% of ovarian cancers, and accumulating evidence underscores the importance of EGF-like ligands for malignant transformation of ovarian epithelial cells. The prognostic impact of EGFR expression in ovarian cancer, however, remains controversial. As for HB-EGF, previous studies have reported this ligand in the stroma but not in the surface epithelium of the normal ovary (21). The fact that Tanaka et al. did not detect HB-EGF in the normal ovary might be attributable to differences in antibody sensitivity and/or specificity. Nakano's group had previously reported increased HB-EGF concentrations in the peritoneal fluid from ovarian cancer patients (4). They now relate these findings to abnormal expression of HB-EGF in malignant OSE cells (mRNA/fluorescence in situ hybridization) that express ADAM-17 (2) and in the surrounding interstitial tissue (protein/immunohistochemistry), indicating that HB-EGF produced from malignant OSE cells is constitutively secreted into the microenvironment. Interestingly, the tetraspanin CD9 has previously been detected on OSE cells only, and association of proHB-EGF with CD9 has been shown to increase HB-EGF's juxtacrine activity and cytoprotective capacity in epithelial cell models. Thus, aberrant expression of HB-EGF in OSE cells might entail interactions with different cell surface proteins than those present in stromal cells, resulting in distinct biological functions. Increasing evidence suggests the organization of metalloproteases, ligand growth factors and/or their receptors into preformed complexes within membrane microdomains and shed membrane vesicles (22). Examples include the interaction of CD9 with HB-EGF and ADAM-10 in complexes that mediate EGFR transactivation by GPCRs (14) or the association of CD44 with HB-EGF, its sheddase MMP-7 and its receptor erbB4 on the surface of tumor cell lines (23). A detailed study of the colocalization of HB-EGF and distinct ADAM proteins, especially ADAM-10, in ovarian cancer tissue is therefore warranted to judge the physiologic relevance of the presumed interaction. Furthermore, important insights into the (patho-) physiology of both GPCR and EGFR signaling axes may be gained through studies of their polarized distribution and function in normal epithelia and during loss of polarity characteristic of malignant transformation.
Drugability. The link between LPA and ovarian carcinogenesis has been mostly correlative, such as elevated levels of LPA in serum or ascites fluid and in vitro studies with exogenous LPA. The study by Tanaka et al. in this issue of Clinical Cancer Research complements their previous observations, now suggesting a functional role for LPA-induced, ADAM-17-mediated HB-EGF cleavage in ovarian cancer. This first laudable dispatch raises several questions: what is the role of other membrane-anchored or soluble metalloproteases (in particular ADAM-10, ADAM-15, or MMPs) in ovarian cancer? Will cell-permeable metalloprotease inhibitors be required to block metalloprotease action in preformed intracellular complexes with their substrates? Is loss of polarity and aberrant access of LPA to its receptors of importance? Will specific inhibitors of LPA and/or erbB receptors and/or other components of the GPCR/EGFR signaling axes (e.g., metalloproteases or HB-EGF) stand the test of clinical application? Thus far, metalloprotease- or erbB receptor-targeted agents have exhibited only limited activity in ovarian cancer patients. Now, in the light of functional redundancy and incomplete overlap of LPA and EGFR signaling, the additional exploration of LPA receptors as novel drug targets seems promising. What about other signaling events? Abundant crosstalk between diverse mitogenic pathways such as EGFR, insulin-like growth factor receptor or estrogen receptor signaling has been reported, and these interactions will have to be considered in our efforts to combat the disease. The stage is set for more mechanistic studies into the complex pathophysiology of ovarian cancer, which, hopefully, will advance to effective treatment strategies that are urgently needed in the clinics.
Grant support: West German Cancer Center Grant, Vanderbilt Physician Scientist Development Award (to A.H. Braun) Vanderbilt GI Specialized Program of Research Excellence (P50 CA95103; to A.H. Braun and R.J. Coffey) as well as NCI CA46413 grant (to R.J. Coffey).
The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
We apologize to all colleagues, whose original work could not be cited due to strict space limitations. Interested readers are referred to the references indicated in the mentioned reviews. We thank Jeffrey L. Franklin for critical review of the manuscript.