Role of Class 3 Semaphorins and Their Receptors in Tumor Growth and Angiogenesis

Semaphorins were first described in 1993 as negative mediators of axonal guidance in the central nervous system (1). However, the discovery of more widespread expression of class 3 semaphorins (SEMA3) by various tumor cell types has suggested a role for these ligands in tumor biology. SEMA3 mediates its actions by binding to neuropilin, cell surface receptors that complex with plexins to mediate downstream signaling. Neuropilins additionally serve as receptors for vascular endothelial growth factor (VEGF; refs. 2, 3), whereby they serve to promote VEGF-induced tumor angiogenesis and growth. Because SEMA3s are understood to inhibit angiogenesis and tumor growth through neuropilins, it has been suggested that perhaps they act by merely competing with VEGF and thus inhibiting its proangiogenic and protumorigenic effects. Despite numerous studies investigating the interaction between SEMA3s, neuropilins, and VEGF, it is not clear whether SEMA3s inhibit tumor growth by simply competing with VEGF for neuropilin ligand-binding sites, whether they act independently of VEGF or both (4,6). Therefore, from a therapeutic standpoint, it is not clear whether semaphorins represent (a) simply another approach to blocking VEGF activity or (b) an antitumorigenic/antiangiogenic therapy that works through alternative pathways. The rationale for developing semaphorin-based anticancer therapy will ultimately depend on a better understanding of the precise roles of SEMA3s, neuropilins, and VEGF in tumor biology. In this review, we present evidence for a role for SEMA3s in tumor biology with a specific focus on the possible effects of SEMA3s on the VEGF/neuropilin system.

Semaphorins, also known as collapsins, were first identified as a family of genes encoding guidance molecules for the embryologic development of the nervous system (1). They were found to induce collapse of growth cones of neurons and to repel axons from growing sensory, sympathetic, and motor neurons. We now know that semaphorins can affect other cell lineages such as endothelial cells, inflammatory cells, vascular smooth muscle cells, and various tumor cells (reviewed in ref. 7).

SEMA3s comprise one of five vertebrate families of semaphorins and are known to play an important role in tumor biology (8). The SEMA3 class consists of seven soluble proteins of 100 kDa (designated by the letters A-G), which are secreted by cells of multiple lineages, including epithelial cells, neurons, and specific tumor cells. SEMA3s act in a paracrine fashion by binding to neuropilins via a highly conserved amino-terminal 500amino acid region in the SEMA3 protein called the Sema domain (9). After binding to SEMA3s, neuropilins complex with another family of proteins, the plexins, to mediate downstream signaling. Plexins are large 200-kDa transmembrane receptors that primarily play a role in cellular organization and migration, although the exact mechanism remains unclear (reviewed in ref. 10). The seven SEMA3s have different binding affinities to neuropilins; hence, each SEMA3 has a distinct biological function (Fig. 1A). Whereas SEMA3A preferentially binds to NRP1, SEMA3F and SEMA3G have a higher binding affinity to NRP2. SEMA3B, SEMA3C, and SEMA3D bind to both isoforms (7); SEMA3E binds to neither neuropilin but instead binds directly to plexin D1 (11). Additionally, each SEMA3-neuropilin complex recruits specific plexins to mediate downstream signaling (Fig. 1B).

In the vascular system, SEMA3s impair endothelial cell adhesion, migration, tube formation, sprouting, branching, survival, permeability, and angiogenesis (reviewed in ref. 12). In tumor biology, however, SEMA3s function as repellants, inhibitors of migration, and inducers of apoptosis for both endothelial cells and specific tumor cells [e.g., breast cancer (13, 14) and lung cancer (15)]. Therefore, SEMA3s play a dual inhibitory role of impeding tumor angiogenesis and tumor growth (specifically SEMA3B and SEMA3F). It should be noted, however, that the inhibitory effects of SEMA3s cannot be generalized to all members of the SEMA3 family. Although most members of the SEMA3 family have been shown to be inhibitory (SEMA3A, SEMA3B, SEMA3D, SEMA3F, and SEMA3G), some promote tumor angiogenesis, growth, and metastasis (SEMA3C and SEMA3E; refs. 16, 17).

Neuropilins (NRP1 and NRP2) are transmembrane glycoproteins with short cytoplasmic domains (42-44 amino acids) and no known signaling motifs of their own; however, some researchers have refuted this and suggested that the last three amino acids (S-E-A) of either neuropilin may participate in signaling via G-interacting proteininteracting proteins (18). Although neuropilins were originally identified as key mediators of axonal guidance (via semaphorins), they are now known to also play a major role in vasculogenesis, angiogenesis, and tumor growth (reviewed in ref. 19). In addition to serving as coreceptors for SEMA3s, neuropilins are also coreceptors for VEGF family members, which are structurally unrelated to SEMA3s. VEGF family members mediate their downstream effects by binding to neuropilins and forming complexes with VEGF receptors (VEGFR) analogous to the SEMA3-neuropilin-plexin complex (2, 20). Neuropilins are expressed on various cell lineages, including neurons, endothelial cells, inflammatory cells, vascular smooth muscle cells, and tumor cells. This expression profile allows VEGF and SEMA3s to have wide-ranging effects on multiple cell lineages. Neuropilins are differentially expressed in the vascular system: whereas NRP1 is mainly expressed on arterial endothelial cells, NRP2 is primarily expressed on venous and lymphatic endothelial cells (21,23). This specificity forms the basis for how different combinations of SEMA3 and VEGF family members play tightly regulated roles in lymphangiogenesis and vasculogenesis through their interactions with specific neuropilins (21, 24).

Neuropilins are 130-kDa single-spanning transmembrane glycoproteins composed of five distinct domains: three extracellular domains (a1a2, b1b2, and c), one transmembrane domain, and one intracellular domain (19, 25). The a1a2 domain typically binds to SEMA3s, whereas the b1b2 domain binds to VEGF family members. Just as neuropilins have different binding affinities for SEMA3 family members, neuropilins have varying affinities for VEGFs (Fig. 2). Both NRP1 and NRP2 can bind to several different isoforms of VEGF-A, particularly to those that contain the neuropilin-binding domain in exon 7 (26, 27). In addition, NRP1 binds to VEGF-B and placental growth factor-2 (PlGF-2) and complexes with either VEGFR-1 or VEGFR-2 after ligand binding. NRP2 can bind to VEGF-C, VEGF-D, and PlGF-2 and can complex with VEGFR-1, VEGFR-2, or VEGFR-3 (19). The result of these biophysical differences is that various VEGFR/VEGF/neuropilin and SEMA3/neuropilin/plexin complexes are generated that each may exert unique influences on not only physiologic processes but also on tumor angiogenesis and subsequent tumor growth.

The primary cellular consequence of semaphorin signaling is cytoskeletal collapse mediated by plexins, which are part of the SEMA3-neuropilin ligand-receptor complex. Plexins are large transmembrane receptors with highly conserved cytoplasmic domains (structure reviewed in ref. 7). The plexin family consists of nine members: four type A plexins (A1, A2, A3, and A4), three type B plexins (B1, B2, and B3), and plexins C1 and D1. The SEMA3s mediate their actions either via plexins type A or D1, which then employ a distinct GTPase-activating intracellular domain to affect actin depolymerization (5, 28). Miao et al. first showed that Sema3A binds to endothelial NRP1 and inhibits endothelial cell migration and capillary sprouting by causing retraction of endothelial cell lamellipodia (5). Despite extensive research in this field, there is no general consensus on the exact intracellular signaling pathway employed by plexins to cause actin depolymerization and cellular collapse (28,30). In addition, Serini et al. recently showed that SEMA3s are able to inhibit integrin activation in endothelial cells, a phenomenon required for endothelial cells to adhere to fibronectin and vitronectin (components of the extracellular matrix) to migrate (31). These inhibitory effects of SEMA3s on integrins also seem to be dependent on plexins (32). Although no clear mechanism has been elicited for plexin-mediated inhibition of tumor growth and angiogenesis, it is universally accepted that SEMA3s inhibit cellular motility and migration by mediating its actions via plexins (33).

As discussed above, the functions of SEMA3s in guiding endothelial cells during vascular development are dependent on receptor complexes that vary according to tissue-specific expression as well as coreceptor combinations. Yet another level of complexity is added when variability among SEMA3 family members is taken into account. Among the SEMA3s, whereas SEMA3A and SEMA3F have been shown to have antiangiogenic properties (31, 3437), SEMA3C is considered to be a proangiogenic factor (16). Numerous studies have shown that both SEMA3A and SEMA3F are able to inhibit in vitro endothelial cell proliferation, survival, and cord formation, although the exact mechanism remains unclear. Some studies have suggested that SEMA3A and SEMA3F inhibit migration of endothelial cells by inhibiting activation of integrins and thereby disrupting tumor angiogenesis (31, 37). In contrast, other studies have shown that SEMA3A and SEMA3F inhibit VEGF-induced and basic fibroblast growth factorinduced angiogenesis (36, 37). SEMA3C, on the other hand, enhances endothelial cell proliferation, migration, and adhesion in renal glomeruli, which can be attributed to its ability to activate integrins and alter cytoskeletal dynamics (16).

The inhibitory effects of SEMA3A and SEMA3F on tumor angiogenesis have been explored by several investigators as potential therapeutic agents for cancer. A recent study by Kigel et al. has shown successful inhibition of tumor growth by coexpressing SEMA3A in tumor cells expressing NRP1 (38). Furthermore, the results of that study showed that SEMA3A, SEMA3D, SEMA3E, and SEMA3G can function as potent inhibitors of tumor growth and affect vessel density. Another two seminal studies by Bielenberg et al. (35) and Kessler et al. (37) recently showed that melanoma cells and tumorigenic baby hamster kidney-21 cells transfected with SEMA3F, respectively, produced tumors with reduced microvessel density. These studies suggested that exogenous SEMA3A and SEMA3F can inhibit tumor angiogenesis and can potentially be used therapeutically to inhibit tumor angiogenesis.

Table 1 provides a summary of the function of the different members of the SEMA3 family in tumor biology. Of the SEMA3 family members, SEMA3A, SEMA3B, and SEMA3F have the best-described roles in tumor growth and metastasis. One of the observations suggesting that SEMA3s may be a type of tumor suppressor gene came from studies showing that that loss of heterozygosity of the 3p21.3 region correlated with the development of lung cancer (39, 40). This region was later discovered to correspond to a deletion of SEMA3B and SEMA3F genes that are now known to behave as tumor suppressor genes (41,44). To this end, loss of SEMA3F has been shown to correlate with advanced tumor stage in lung cancer and the development of metastasis in melanoma (35, 45).

Over the past decade, the role of SEMA3s in the pathogenesis of multiple malignancies has also been investigated in preclinical studies. SEMA3A has been implicated in the inhibition of tumor cell migration and chemotaxis in breast cancer cells (9). In prostate cancer, SEMA3A-transfected cells differentially regulate adhesion of cells together and have been shown to exhibit decreased invasion and adhesion (46). Similarly, SEMA3B transfection has been shown to inhibit in vitro tumor cell growth and to induce apoptosis in lung and breast cancer cells (15, 47). In ovarian cancer, transfection and overexpression of SEMA3B has been correlated with decreased anchorage-independent growth in vitro and with tumor formation in vivo (43). In lung cancer, overexpression of SEMA3F has been correlated with a decrease in tumorigenicity (48). These findings have been confirmed in two other studies in which lung cancer cell lines overexpressing SEMA3F produced smaller tumors than did control cells (41, 49). In breast cancer cells, exposure to SEMA3F has been shown to inhibit cell attachment and lamellipodia extension, the two processes necessary to mediate cellular migration (13, 14). Additionally, transfection of SEMA3F in fibrosarcoma cell lines has been shown to result in complete loss of tumorigenicity (44). Likewise, melanoma cells overexpressing SEMA3F exhibited decreased in vitro migration and resulted in a benign tumor phenotype (highly encapsulated and poorly vascularized tumors) in vivo (35). Taken together, these findings suggest that SEMA3A, SEMA3B, and SEMA3F inhibit tumor growth.

The prognostic role of SEMA3s has been further elucidated in several clinical studies. For example, Osada et al. showed that a high VEGF:SEMA3 ratio is an adverse prognostic factor for patients with ovarian cancer (50). Additionally, whereas high levels of SEMA3A have been correlated with poor prognosis in patients with pancreatic cancer, low levels of SEMA3F have been correlated with an advanced stage of melanoma (35, 51). However, melanoma clones overexpressing SEMA3F have been shown to show decreased potential to spontaneously metastasize to both lymph nodes and lungs in vivo (35). Numerous pathologic studies have reported a correlation with prognosis in cancer patients with variable expression levels of SEMA3s (summarized in Table 4 in ref. 52), suggesting that SEMA3s may play a major role in tumor growth and that they can potentially be used as therapeutic targets or agents depending on the specific SEMA3 ligand and tumor type.

As discussed previously, SEMA3 ligands have been hypothesized to exert their antineoplastic behavior by competing with VEGF for their binding site on neuropilin receptors (5). This raises the possibility that exogenously administered SEMA3s may simply serve as a form of anti-VEGF therapy. VEGF is the prototypical member of a family of angiogenic factors that mediate a variety of biological effects on endothelial cells and tumor cells via receptor tyrosine kinases (VEGFR; ref. 26). The mammalian VEGF family consists of five distinct glycoproteins (VEGF-A, VEGF-B, VEGF-C, VEGF-D, and PlGF-2; ref. 53), each of which has unique receptor-binding and action profiles (Fig. 2). VEGF-A, the best-characterized molecule in the VEGF family, binds to VEGFR-1, VEGFR-2, NRP1, and NRP2. These receptors, expressed on tumor cells and endothelial cells, mediate downstream signaling after binding to VEGF-A and thereby promote tumor growth (reviewed in ref. 54). VEGF-B is presumed to mediate effects similar to those of VEGF-A and is known to bind to VEGFR-1 and NRP1, but the precise role of VEGF-B in malignancy has yet to be elucidated. VEGF-C and VEGF-D primarily play roles in lymphangiogenesis and bind to VEGFR-2, VEGFR-3, and NRP2, although VEGF-C may also be angiogenic (53). PlGF is a VEGF homologue that stimulates angiogenesis, local inflammatory response, tumor growth, and stromal cell migration (55,57). Differential splicing of the human PlGF gene generates two secreted PlGF proteins, PlGF-1 and PlGF-2. PlGF-2, the heparin-binding isoform, mediates its actions via binding to VEGFR-1 directly or by binding to the VEGFR-1 coreceptors NRP1 and NRP2 (58). Clinical and experimental data have suggested that the VEGF signaling pathways in tumor cells are important for their survival, migration, and invasion (59,61) as well as for the development of metastases (61,63). These observations have led to the development and validation of anti-VEGF therapies for advanced malignancies either as a single agent or combined with chemotherapy. However, although anti-VEGF therapy has been shown to be of benefit for patients with certain tumor types, resistance, either inherent or acquired remains an obstacle for long-term survival (26). Therefore, alternative approaches to anti-VEGF therapy are warranted that might provide alternative therapeutic regimens for patients who do not respond well to anti-VEGF therapy.

Semaphorins offer a novel approach in which the addition of a protein can inhibit tumor growth and angiogenesis. The mechanism(s) by which SEMA3s exert antagonistic effects on endothelial cells and tumor cells remains to be elucidated, and it is still not understood whether they compete with VEGF for neuropilin binding (Fig. 3). Miao et al. first showed that when SEMA3A complexes with the NRP1 a1a2 domain, the carboxyl-terminal immunoglobulin domain of SEMA3A, blocks the VEGF b1b2-binding domain of NRP1 (5). In terms of the consequence of this steric hindrance to function, Miao et al. suggested that SEMA3A competitively inhibits the binding of VEGF to NRP1, thus inhibiting endothelial cell motility (5). Similarly, Geretti et al. showed that VEGF can compete with SEMA3F and that SEMA3F can compete with VEGF for the binding of NRP2, suggesting that VEGF and SEMA3F are two competitive NRP2 ligands (64). In contrast, other investigators have shown that SEMA3F inhibits the VEGF-induced survival and proliferation of human umbilical vein endothelial cells in vitro and VEGF-induced angiogenesis in vivo; however, this inhibition is not mediated via competition between SEMA3F and VEGF but instead is regulated at an intracellular level that requires active SEMA3F-mediated signaling (37).

Despite other studies suggesting that the antitumor effects associated with SEMA3s are partially due to competition with VEGF (4, 65), structural studies have suggested that VEGF and SEMA3A have distinctly separate binding sites and therefore do not compete for NRP1 binding (25). According to this structural model, the NRP1 a1a2 domain is an exclusive binding region for SEMA3A, whereas the b1b2 domain is an exclusive binding region for VEGF-165. This structural model is further supported by evidence that SEMA3A-mediated collapse of axonal growth cones from dorsal root ganglia occurs even at increasing concentrations of VEGF-165 (25). Likewise, Pan et al. showed that saturating concentrations of SEMA3A do not seem to affect VEGF-induced endothelial cell migration (6). However, Miao et al. clearly showed that greater concentrations of SEMA3A are required to achieve the same level of growth cone collapse in the presence of VEGF, suggesting that VEGF competes with SEMA3A to induce the collapse of dorsal root ganglia (5). Given that experimental support exists for both of the competing hypotheses, further studies need to be done to better elucidate the true interaction between SEMA3s and various VEGF isoforms. The question of competitive inhibition versus independent antagonistic effects of SEMA3s and VEGF is important, because it offers investigators a novel approach to develop anticancer therapies.

Semaphorins, a secreted class of proteins originally described as guidance cues for developing neurons, have emerged as key regulators of vasculogenesis, angiogenesis, and tumorigenesis. Targeting endothelial cells and tumor cells with different members of the SEMA3 family may represent a promising new type of therapy for preventing tumor angiogenesis, growth, and metastasis. Additionally, if further investigation supports the hypothesis that SEMA3s compete with VEGF for binding to neuropilins, it might be possible to use SEMA3 overexpression to inhibit the proangiogenic and mitogenic effects of VEGF. Additionally, exogenous SEMA3 therapy may have additive or synergistic activity when added to anti-VEGF therapy (6). Further studies are warranted to elucidate the interactions between VEGF, neuropilins, and SEMA3s that inhibit tumor growth and progression.

L.M. Ellis, commercial research grant, Sanofi-Aventis, ImClone Systems; consultant, Schering-Plough, Roche, Amgen.

We thank Virginia M. Mohlere (Department of Scientific Publications) and Rita Hernandez (Department of Surgical Oncology) for editorial assistance and Kristin Johnson (Vascular Biology Program, Children's Hospital) for graphic design.