Semaphorin-3B (sema3B) and semaphorin-3F (sema3F) are secreted tumor suppressors of lung cancer. Sema3F functions as an antiangiogenic factor that repels endothelial cells and compromises their proliferation/survival. However, tumor cells expressing either endogenous or recombinant sema3B fail to repel endothelial cells efficiently. Sema3B found in the conditioned medium of such cells is almost completely cleaved by furin-like pro-protein convertases, generating inactive 61- and 22-kDa fragments. We have generated a sema3B variant that was point mutated at the cleavage site (sema3B-m), thereby conferring partial resistance to cleavage. Conditioned medium from HEK293 cells expressing sema3b-m and conditioned medium of HEK293 cells expressing sema3B contained similar concentrations of semaphorin but sema3B-m was cleaved much less than sema3B. In contrast to HEK293 cells expressing native sema3B, cells expressing sema3b-m strongly repel endothelial cells. Conditioned medium from sema3B-m–expressing cells rapidly caused disassembly of focal adhesions and a collapse of the actin cytoskeleton of endothelial cells, inhibited vascular endothelial growth factor–induced phosphorylation of extracellular signal-regulated kinase 1/2, induced apoptosis of endothelial cells, and inhibited the formation of tubes from endothelial cells in an in vitro angiogenesis assay more potently than conditioned medium from cells expressing sema3B. Furthermore, HEK293 cells expressing sema3B-m inhibited basic fibroblast growth factor–induced angiogenesis in vivo much more potently than cells expressing sema3B. Repulsion of human umbilical vascular endothelial cells by sema3B-m was mediated primarily by the neuropilin-1 (np1) receptor but sema3B-m was also able to transduce signals via neuropilin-2 (np2). These results suggest that up-regulation of furin-like pro-protein convertases in malignant cells may enable tumors to evade the antiangiogenic effects of sema3B. [Cancer Res 2008;68(17):6922–31]
- endothelial cells
- furin-like pro-protein convertases
The transition from a dormant nonangiogenic small tumor to an expanding angiogenesis-promoting tumor is known as the angiogenic switch; it is a crucial step in tumor progression and is believed to be induced by up-regulation of proangiogenic factors or by down-regulation of angiogenic inhibitors ( 1).
Semaphorins are axon guidance factors that induce localized collapse of neuronal growth cones ( 2). The seven class-3 semaphorins are the only secreted vertebrate semaphorins. Like all semaphorins, they contain an NH2-terminal sema domain required for bioactivity and are distinguished by the presence of a basic domain at their COOH termini. Except for sema3E, class 3 semaphorins bind to one or to both of the receptors belonging to the neuropilin family. Neuropilins do not transduce semaphorin signals on their own but associate with type A or type D plexins, which serve as the signal-transducing elements in neuropilin/plexin holoreceptors ( 3– 5). The neuropilins also function as receptors for several forms of the angiogenic factor vascular endothelial growth factor (VEGF) and are expressed in endothelial cells ( 6– 8). In addition to their indispensable role in axon guidance and central nervous system development, the neuropilins were also found to play important roles in developmental angiogenesis as revealed by gene-targeting experiments ( 9, 10). Recently, it has been found that the class-3 semaphorins (i.e., sema3A, sema3F, and sema3E) display antiangiogenic properties ( 11– 13). Experiments in which the native np1 receptor of mice was replaced by a np1 variant that binds VEGF but not sema3A indicate that the cardiovascular developmental abnormalities observed in mice lacking functional np1 receptors are probably caused by impaired VEGF signal transduction rather than by impaired semaphorin signaling ( 14, 15).
The genes encoding sema3B and sema3F were identified as tumor suppressor genes that are deleted or inactivated in lung cancer ( 16, 17). Expression of sema3B in HEY ovarian carcinoma cells inhibits tumor development from these cells ( 18). Recently, it was observed that hypermethylation of the sema3B promoter occurs in most cases of hepatocellular carcinomas and cholangiocarcinomas at a relatively early stage of tumor progression ( 19). Conditioned medium of recombinant sema3B-expressing cells inhibited the proliferation of several breast and lung cancer–derived cell lines, indicating that secreted sema3B can directly affect the proliferation of cancer cells ( 20). Studies conducted using MDA-MB-231 breast cancer cells that express primarily np1 suggest that sema3B uses np1 as its receptor ( 20); however, recent studies suggest that sema3B can also use np2 ( 21).
Pro-protein convertases (PPC) constitute a family of nine calcium-dependent serine endoproteases. The best studied member of the family is furin. PPCs cleave their substrates after the consensus motif RXK/RR. The minimal recognition site for PPCs is RXXR but the presence of a basic amino acid (K/R) in the third position greatly enhances processing ( 22). PPCs are often overexpressed in cancer cells and are known to contribute to tumorigenesis by proteolytic activation of precursors of protumorigenic factors such as insulin like growth factor-1 and its receptor, transforming growth factor-β; VEGF-C; and metalloproteases such as MT1-MMP, which plays an important role in the induction of tumor invasiveness and tumor metastasis ( 22).
All class-3 semaphorins contain PPC recognition sites. The major PPC cleavage site is a KRRXRR sequence that is conserved in all the class 3 semaphorins. Cleavage at this site generates an NH2-terminal fragment of ∼60 kDa. The cleaved NH2-terminal fragment of sema3A is inactive ( 23) but the corresponding NH2-terminal cleavage product of sema3E displays potent prometastatic activity ( 24). Two more potential processing sites are located near the COOH terminus of all class-3 semaphorins. Cleavage of sema3A at these COOH-terminal sites increased its repulsive activity for neuronal growth cones, indicating that cleavage at these sites may result in the potentiation of class 3 semaphorin activity ( 23).
We report that sema3B functions as a repellent of endothelial cells. However, sema3B is highly susceptible to degradation by PPCs and is inactivated by PPC-mediated cleavage at the major site of cleavage. We show that the conditioned medium of many types of malignant cells contains cleaved inactivated sema3B. We also show that a sema3B variant containing mutations at the major PPC cleavage site of sema3B displays increased resistance to cleavage and that cells expressing this mutated sema3B repel endothelial cells more effectively than cells expressing similar amounts of native sema3B. Furthermore, the mutated sema3B inhibits basic fibroblast growth factor (bFGF)–induced angiogenesis more efficiently than native sema3B despite the limited protection afforded by the mutation. Our results indicate that cleavage of sema3B by up-regulated PPCs of cancer cells is a possible mechanism by which tumors evade the antiangiogenic effects of sema3B, thereby contributing to tumor progression.
Materials and Methods
Materials. Antibodies directed against β-actin and the FLAG epitope tag, as well as chemicals, were from Sigma. Mediums and sera for cell culture were from Biological-Industries, Inc. Oligofectamine was from Invitrogen-Life Technologies. Antibodies directed against sema3B, Np1, Np2, phosphorylated extracellular signal-regulated kinase 1/2 (ERK1/2), and ERK2 (total ERK) were from Santa Cruz, Inc. The antibody against active caspase-3 was from MBL. The antibody against vinculin was from Chemicon International. The anti–VSV-G antibody was from ICL. The furin inhibitor decanoyl-RVKR-CMK was from Alexis Biochemicals. The fluorescent vital dye DiAsp and Alexa-conjugated phalloidin were from Molecular Probes. The QuickChange site-directed mutagenesis kit and Strataclean beads were from Stratagene. Cy2-conjugated donkey anti-mouse antibodies were from Jackson Immunoresearch Laboratories. bFGF was produced as previously described ( 25). Matrigel was from BD Biosciences. The cDNA encoding Sema3B was kindly donated by Dr. Susan Naylor (University of Texas Health Science Center, San Antonio, TX). The NSPI-CMV-MCS-myc-His lentiviral expression vector was constructed as follows: a linker containing the restriction enzymes NsiI-XbaI-BstBI-MluI-ClaI and SalI was inserted between the ClaI and SalI sites of pRRL SIN cPPT PGK GFP. A cassette containing SV40 promoter driving puromycin selection marker was digested from pBabe-puro using AccI and ClaI and cloned into the BstBI site. PGK-GFP cassette was then inserted into the ClaI and SalI sites to generate NSPI-PGK-GFP. CMV promoter with multiple cloning sites (MCS) was digested from pCDNA3.1+ Neo (Invitrogen) using MluI and XhoI, replacing the PGK-GFP to generate NSPI-CMV-MCS. Last, we replaced the original CMV promoter with a CMV promoter containing the MCS and myc His cassette from pCDNA3.1-myc/His (Invitrogen) using MluI and PmeI to generate NSPI-CMV-MCS-myc-His.
Expression plasmids. The VSV-G tag encoding sequence was inserted in frame just before the stop codon of sema3B (COOH-terminal tag). The modified sema3B cDNAs were cloned into the NSPI-CMV-MCS-myc-His lentiviral expression vector. The modification in the furin-processing site exchanging RFRR with KFKK was introduced into sema3B to generate sema3B-m using the QuikChange mutagenesis kit according to the instructions of the vendor. The primers used were 5′-CCAGTGCCAAGAGGAAGTTCAAGAAGCAAGACGTAAGGAATGG-3′ and the complementarily reverse primer. The truncated sema3B forms were generated by PCR amplification of appropriate sema3B cDNA fragments and a FLAG epitope tag was inserted in frame before the stop codon. The various cDNAs encoding the sema3B variants were all subcloned into the NSPI-CMV-MCS-myc-His lentiviral expression vector. Full-length sema3B-m containing the COOH-terminal FLAG tag inserted in frame just before the stop codon instead of the VSV-G tag was also cloned into the NSPI-CMV-MCS-myc-His and served as a control in experiments using the truncated sema3B variants.
Generation of recombinant lentiviruses. HEK293-T cells were seeded in 100-mm tissue culture dishes (2.5 × 106 cells per dish). A day after seeding, the cells were cotransfected by the calcium phosphate method with the lentiviral expression plasmid (15 μg), packaging vector pCMVdR8.91 (10 μg), and a plasmid encoding the vesicular stomatitis virus coat envelope pMD2-VSVG (5 μg). Conditioned medium containing infective lentiviral particles was collected 48 and 72 h posttransfection.
Cells. Human umbilical vascular endothelial (HUVEC), porcine aortic endothelial (PAE), and HEK293 cells were cultured as previously described ( 11). HUVEC were used between passages 3 to 7. DU145 prostate cancer cells were cultured in RPMI supplemented with 10% FCS. WW94, YU-ZAZ6, and SW1416 melanoma cells were grown in medium containing a mixture of F12 and low glucose DMEM supplemented with 10% FCS. All the other cell lines were grown in high glucose DMEM supplemented with 10% FCS.
Western blots. Cell lysates and Western blots were done as described ( 26).
Preparation of sema3B-containing conditioned medium. HEK293 cells expressing different forms of sema3B or control empty vector–infected HEK293 cells were grown to 70% confluence and incubated subsequently for 24 h with HUVEC growth medium ( 11) supplemented with 10% FCS (for ERK1/2 phosphorylation assays) or 20% FCS (for all the other assays). The proliferation rates of control, empty vector–infected HEK293 cells, and cells expressing either sema3B or sema3B-m were identical (data not shown).
Endothelial cell repulsion and proliferation/survival assays. Cell proliferation/survival assays were performed by counting surviving cells as described ( 26). The results of these assays are influenced on the one hand by effects on cell proliferation and on the other hand by proapoptotic effects and represent an integration of the two mechanisms. Repulsion assays were done as described ( 26, 27).
Immunofluorescence. Detection of vinculin and actin in HUVEC was performed essentially as described ( 26).
ERK1/2 phosphorylation. The effects of sema3B-m on VEGF121- and VEGF165-induced ERK1/2 phosphorylation were determined essentially as described ( 11).
Down-regulation of neuropilin expression in HUVEC using small interfering RNA. Inhibition of np1 and np2 expression in HUVEC was performed as previously described ( 26) using the np1-specific small interfering RNA (siRNA) r(AAGGAAACCUUGGUGGGAU)d(TT) and the np2-specific siRNA r(CCAGAAGAUUGUCCUC AAC)d(TT) or a control siRNA r(UUCUCCGAACGUGUCACGU)dTdT.
Caspase-3 apoptosis assay. The assay was performed as previously described ( 26).
Statistical analysis. Statistical analysis was performed using the upaired data with unequal variance Student's t test. Error bars represent the SE. Statistical significance is presented as *P < 0.05 and ***P < 0.001.
Native as well as recombinant Sema3B produced by various types of tumorigenic cells is cleaved by furin-like PPCs. Sema3F is an inhibitor of endothelial cell proliferation, a repellent of endothelial cells, and an inhibitor of tumor angiogenesis ( 11, 26, 27). To find out if sema3B also functions as a repellent of endothelial cells, we expressed the native sema3B cDNA or a sema3B cDNA containing an in-frame vesicular stomatitis virus (VSV-G) epitope tag at the COOH terminus (sema3B-VSV) in HEK293 cells using recombinant lentiviruses. However, although recombinant sema3B was efficiently expressed, we could only detect weak repulsion of HUVEC compared with the repulsion obtained using cells expressing sema3A ( Fig. 1A ). When we examined the conditioned medium of the cells using antibodies directed against the NH2 terminus of sema3B, we found that the sema3B was almost completely cleaved, yielding a single ∼60 kDa NH2-terminal fragment ( Fig. 1B, a), leaving very little intact full-length 83 kDa sema3B. A similar ∼60 kDa cleaved sema3B fragment was revealed by the same antibody in conditioned medium derived from HEK293 cells expressing native, untagged sema3B (data not shown). The antibody used to detect the NH2 terminus of sema3B does not detect the COOH-terminal cleavage product. This product was seen as a ∼20 kDa fragment in the conditioned medium of the cells using antibodies directed against the COOH-terminal VSV epitope tag ( Fig. 1B, b). A similar ∼60 kDa cleavage product was also observed when transfection rather than viral infection was used to express the recombinant sema3B (data not shown). Throughout this work, we have mainly used the COOH-terminal tagged sema3B. It did not differ in biological properties from untagged sema3B and we therefore refer to it from now on as sema3B.
All class-3 semaphorins contain a major conserved cleavage site for furin-like PPCs. Cleavage at this site ( Fig. 3B, site 1) generates an ∼60 kDa NH2-terminal fragment and an ∼20 kDa COOH-terminal fragment ( 23, 24), suggesting that the ∼60 kDa cleavage product observed in the conditioned medium of sema3B-expressing HEK293 cells is also the product of PPC activity. When the activity of furin-like PPC of HEK293 cells expressing recombinant sema3B was inhibited with the furin inhibitor decanoyl-RVKR-CMK ( 24, 28), the cleavage of sema3B was partially inhibited, leading to the secretion of a mixture of full-length sema3B and cleaved sema3B ( Fig. 1B, c). To find out if sema3B undergoes degradation in additional types of cancer cells, we expressed recombinant sema3B in several types of malignant cells that do not produce endogenous sema3B. These included MDA-MB-231 breast cancer cells, HA188 lung cancer cells, A549 lung cancer cells ( Fig. 1C, a), and H661 lung cancer cells ( Fig. 2A, b ). The conditioned medium of these cells contained cleaved sema3B, and we could not detect full-length sema3B in them. To make sure that the cleavage was indeed caused by PPCs, we inhibited the PPC activity of the MDA-MB-231 cells using decanoyl-RVKR-CMK. The inhibitor was somewhat toxic to these cells and caused a decrease in the total amount of recombinant sema3B that was secreted into the medium; however, it can be seen that cleavage was inhibited and that, in the presence of the inhibitor, full-length sema3B was secreted from the cells ( Fig. 1C, b). To find out if natural sema3B is cleaved by PPCs of cancer cells, we screened several types of malignant melanoma, colon cancer, and prostate cancer–derived cell lines. Several of these produced and secreted sema3B into their growth medium. However, the secreted sema3B was completely or almost completely cleaved in the conditioned medium of these cells ( Fig. 1D, a). To make sure that the cleavage of native sema3B was caused by furin-like enzymes, we inhibited the cleavage in one of the melanomas using the inhibitor. Indeed, the inhibition partially prevented the cleavage ( Fig. 1D, b).
Tumorigenic cells expressing a sema3B mutant that is partially resistant to degradation by PPCs strongly repel endothelial cells. Decanoyl-RVKR-CMK inhibited the activity of furin-like PPCs in the tumorigenic cells but compromised the proliferation/survival of HUVEC (data not shown). To enable the examination of the effects of sema3B on endothelial cells, we generated a sema3B mutant resistant to degradation by PPCs by the introduction of point mutations into the major PPC cleavage site ( Fig. 3B, site 1 ). We found that nonconservative mutations of arginine residues within the RFRR sequence found at this site resulted in the production of sema3B mutants that were almost completely resistant to cleavage but displayed strongly reduced biological activity (data not shown). We therefore mutated the amino acids of site 1 conservatively from RFRR to KFKK to generate the sema3B mutant sema3B-m ( Fig. 3B). Conditioned medium of sema3B-m–expressing HEK293 cells contained a higher concentration of full-length sema3B-m and much less of the ∼60 kDa cleavage product ( Fig. 2A, a), indicating that the point mutations confer partial resistance to cleavage by PPCs. Similarly, when sema3B-m was expressed in H661 lung cancer cells or in MDA-MB-231 cells that degrade sema3B more aggressively than HEK293 cells, we found that their conditioned medium now contained full-length sema3B-m, although most of the sema3B-m was still cleaved ( Fig. 2A, b and c).
The conditioned medium of the sema3B-m–expressing HEK293 cells also contained sema3B-derived fragments significantly larger than ∼60 kDa but of lower molecular weight than the full-length sema3B-m ( Fig. 2A, a). Similar fragments were also found in the conditioned medium of sema3B-m–expressing H661 lung cancer cells ( Fig. 2A, b). One of these fragments corresponds to the ∼75 kDa degradation product also seen in conditioned medium of sema3B-expressing HEK293 cells cultured in the presence of decanoyl-RVKR-CMK ( Fig. 1B and C). These fragments are likely to be generated by furin-like PPC-mediated cleavage at secondary putative PPC consensus cleavage sites located closer to the COOH terminus of sema3B ( Fig. 2B). To make sure that these fragments are not the products of unrelated proteases, we inhibited the activity of furin-like PPCs in HEK293 cells expressing sema3B-m by using decanoyl-RVKR-CMK. Under these conditions, the conditioned medium of the HEK293 cells contained only the full-length sema3B-m, indicating that all the sema3B-m–derived fragments found in the conditioned medium of the sema3B-m–producing HEK293 cells are generated by PPC-mediated cleavage ( Fig. 2B).
HEK293 cells expressing sema3B-m repelled HUVEC much more effectively than HEK293 cells expressing sema3B ( Fig. 2C), although the total amounts of sema3B-m and native sema3B found in the conditioned medium of the producing cells were similar ( Fig. 2A, a). We also compared the effect of conditioned medium derived from HEK293 cells expressing sema3B-m, sema3B, or control cells infected with empty expression vector on the proliferation/survival of HUVEC. Conditioned medium derived from sema3B-expressing HEK293 cells, which contains mainly the cleaved fragments, did not compromise the proliferation/survival of HUVEC. In contrast, conditioned medium of HEK293 cells expressing sema3B-m strongly inhibited the proliferation/survival of HUVEC ( Fig. 2D). These experiments indicate that full-length sema3B is active whereas the major PPC cleavage products of sema3B are substantially less active with respect to their effects on endothelial cells.
Determination of the biological activity of sema3B-m–derived NH2-terminal fragments generated by PPC cleavage. To make sure that the ∼60 kDa fragment generated by cleavage at site 1 of sema3B ( Fig. 1B, a) is indeed inactive, we generated a deletion mutant of sema3B in which we have inserted an in-frame FLAG epitope tag followed by a stop codon just before the site 1 cleavage site. This truncated sema3B fragment (s3bt1; Fig. 3B) was expressed and secreted by HEK293 cells ( Fig. 3A, lane 1). S3bt1-expressing cells did not repel endothelial cells ( Fig. 3C) and did not inhibit their proliferation/survival (data not shown), suggesting that the PPC-generated ∼60 kDa sema3B product lacks biological activity.
When examining conditioned medium obtained from cells expressing sema3B-m, we noticed the presence of sema3B-m fragments larger than 60 kDa but smaller than full-length sema3B-m ( Fig. 2A, a and b). These were apparently generated as a result of cleavage at secondary PPC cleavage sites, which are apparently used more extensively in sema3B-m due to the presence of the mutations at the main cleavage site. Some of these fragments are very close in molecular weight to full-length sema3B and may originate from cleavage at the cluster of potential sites located ∼2 kDa upstream of the COOH-terminus ( Fig. 3B, site 6), of which two are conserved in the class 3 semaphorins ( 23, 24). To determine the activity of NH2-terminal fragments generated by cleavage at these closely clustered sites, we produced a truncated sema3B-m variant (s3bt3) in which we inserted an in-frame FLAG epitope tag followed by a stop codon just before site 6 ( Fig. 3B), thereby removing all these putative cleavage sites as well as the whole basic box of sema3B. Although the concentration of s3bt3 in the conditioned medium of producing HEK293 cells was higher than the concentration of sema3B-m produced by expressing HEK293 cells (compare Fig. 3A, lanes 3 and 4), s3bt3-expressing cells repelled HUVEC very weakly if at all ( Fig. 3C). The ∼75 kDa fragment that is observed in the conditioned medium of sema3B-m–expressing cells ( Fig. 2B) is likely to be produced as a result of cleavage at site 4 of sema3B ( Fig. 3B). We therefore introduced into the sema3B-m cDNA an in-frame FLAG epitope tag followed by a stop codon just before the start of the expected cleavage site of site 4 ( Fig. 3B) and expressed this cDNA in HEK293 cells to generate sb3t2. S3bt2 was devoid of biological activity in the repulsion assay ( Fig. 3C) despite high expression levels ( Fig. 3A, lane 2).
To compare the biological activities of the truncated sema3B-m variants semiquantitatively, we compared the effects of conditioned medium containing sema3B-m or the three deletion mutants ( Fig. 3B) on the proliferation/survival of HUVEC. To assess the concentration of the deletion mutants in the conditioned medium of the producing HEK293 cells, we performed serial dilutions of serum-free conditioned medium and assessed the concentration of the fragments by Western blot analysis using antibodies directed against sema3B ( Fig. 3D, top left) or against the FLAG epitope tag inserted in frame before the stop codon of the mutants and the sema3B-m ( Fig. 3D, bottom left). Sb3t1 (the ∼60 kDa fragment) was devoid of bioactivity in this assay too (data not shown). The two other sema3B-m deletion mutants inhibited the proliferation/survival of HUVEC, but less effectively than sema3B-m ( Fig. 3D, right). The results of these experiments were normalized by taking into account the relative concentrations of the secreted sema3B-m and sema3B-m deletion mutants as determined in the Western blots using the anti-sema3B antibody and a Fuji Film LAS-3000 image reader to assess relative band densities ( Fig. 3D, top left). By these criteria, the relative EC50 of s3bt3 was 9.2 ± 2.8-fold higher than that of sema3B-m and the relative EC50 of s3bt2 was 57.5 ± 7.5-fold higher than that of sema3B-m ( Fig. 3D, right). This estimate may be conservative because cleavage at the site 6 cluster ( Fig. 3B) can generate sema3B-m fragments that the anti-sema3B antibody would not be able to distinguish from full-length sema3B-m because there is only an ∼2 kDa difference in mass. It is therefore possible that if such truncations take place, our estimate of the relative activity of full-length sema3B-m may be too low.
For this reason, we also assessed the relative activities of the truncated forms based on relative concentration measurements obtained using the anti-FLAG epitope tag ( Fig. 3D, bottom left). This antibody does not recognize cleaved sema3B-m or cleaved truncated sema3B-m forms because the epitope tag is lost in these forms. Using this antibody, we observed that the concentration of full-length sema3B-m is much lower in relation to the concentrations of full-length s3bt2 and s3bt3 than the estimate obtained with the sema3B antibody. Using the values obtained with this antibody, we estimated that s3bt3 is ∼40-fold less active than full-length sema3B-m, whereas s3bt2 is 1,200-fold less active (data not shown). However, regardless of the method used, we conclude that in contrast to sema3A, in which truncations near the COOH-terminal result in enhanced activity ( 23), in sema3B even a relatively minimal COOH-terminal truncation already results in substantial loss of biological activity.
Sema3B signaling in HUVEC is mediated primarily by the np1 receptor. Previous work suggested that sema3B can signal using both np1 and np2 ( 20, 21). To determine the relative contributions of np1 and np2 in sema3B signal transduction in HUVEC, we selectively inhibited the expression of these receptors using specific siRNAs. Inhibition of np1 expression by ∼80% ( Fig. 4A ) resulted in a substantial loss of responsiveness to sema3B-m in the endothelial cell repulsion assay ( Fig. 4B, b). In contrast, cells in which the expression of np2 was almost completely inhibited ( Fig. 4A) were still able to efficiently repulse HUVEC ( Fig. 4B, c), indicating that in HUVEC, the sema3B repulsive signals are primarily mediated by np1. To find out if np2 is able to transduce sema3B signals in endothelial cells, we determined if cells expressing sema3B-m are able to repulse PAE cells engineered to coexpress np1 or np2 and plexin-A1 ( Fig. 4C). The sema3B-m–expressing HEK293 cells were able to repel the np2/plexin-A1–expressing PAE cells as efficiently as they repelled PAE cells coexpressing np1 and plexin-A1 ( Fig. 4D), indicating that sema3B can indeed transduce signals via np2. This experiment was done in PAE cells because HUVEC contain relatively low concentrations of np2 and do not respond well to np2 agonists.
Additional effects of sema3B-m on cultured endothelial cells. Exposure of HUVEC to conditioned medium taken from sema3B-m–expressing 293 cells induced rapid disappearance of focal contacts, disappearance of actin stress fibers, and cell contraction within 20 minutes. In contrast, conditioned medium from sema3B-expressing HEK293 cells induced only slight changes ( Fig. 5A ).
VEGF is a proliferation/survival factor for endothelial cells ( 29, 30). In contrast to sema3B, sema3B-m strongly inhibited VEGF121 as well as VEGF165 induced activation of ERK1/2 ( Fig. 5B and C). It was reported that sema3B can induce apoptosis of responsive tumor cells and that this effect is competitively inhibited by VEGF165 but not by VEGF121 ( 20). Our results indicate that in the case of endothelial cells, sema3B-m does not act exclusively by competition with VEGF165 for binding to np1 but also seems to inhibit VEGF activity by induction of an independent inhibitory signal as previously reported for sema3A and sema3F ( 26). Prolonged stimulation with sema3B-m–containing conditioned medium, but not with sema3B–containing conditioned medium, induced apoptosis of HUVEC as revealed by a substantial increase in the concentration of activated caspase-3 ( Fig. 5D).
We also determined if sema3B-m can inhibit the formation of tubes from endothelial cells, a process that is essential in angiogenesis. HUVEC seeded on Matrigel organize into a capillary-like network ( 31, 32). Conditioned medium containing sema3B-m, but not conditioned medium containing a similar concentration of sema3B, strongly inhibited the formation of this capillary-like network ( Fig. 6A ).
HEK293 cells expressing sema3B-m inhibit bFGF-induced angiogenesis more potently than HEK293 cells expressing similar concentrations of native sema3B. The effect that sema3B-m has on the behavior of cultured endothelial cells suggested that sema3B may function as an inhibitor of angiogenesis and that cleavage by PPCs may inhibit its antiangiogenic activity. To test this hypothesis, we mixed control HEK293 cells as well as sema3B- and sema3B-m–producing HEK293 cells with Matrigel containing the angiogenic factor bFGF. The mixtures containing the different cell types were implanted s.c. in BALB/c nu/nu mice and the blood vessels that invaded the Matrigel plugs were quantified after 8 days. Although sema3B-expressing cells reduced the number of blood vessels that invaded the plugs, sema3B-m–expressing cells had a significantly stronger effect ( Fig. 6B). To compare the effect of sema3B and sema3B-m on tumor development, we also expressed sema3B and sema3B-m in MDA-MB-435 melanoma cells ( Fig. 6C; ref. 33). These cells express the sema3B receptor np2 (data not shown). The proliferation of the cells in cell culture was not affected by the expression of sema3B or sema3B-m (data not shown). However, both sema3B and sema3B-m inhibited the formation of these tumors relative to control MDA-MB-435 cells more or less similarly ( Fig. 6D).
Sema3B was identified, along with sema3F, as a tumor suppressor that inhibits the development of small-cell lung carcinoma ( 16, 17). Sema3F was subsequently characterized as a repellent of endothelial cells and as an antiangiogenic factor ( 11, 27). These findings suggested that part of the antitumorigenic behavior of sema3B could perhaps be attributed to antiangiogenic activity.
Surprisingly, HEK293 cells infected with lentiviral vectors directing high-level expression of sema3B did not repel endothelial cells efficiently. When we examined the conditioned medium of these cells, we found that sema3B was almost completely cleaved into two fragments of ∼60 and ∼20 kDa, suggesting that these fragments are inactive. All the class-3 semaphorins contain a conserved consensus PPC cleavage site that, when used, leads to the production of ∼60 and ∼20 kDa fragments. In the case of sema3A, cleavage at this site leads to loss of activity ( 23), whereas cleavage of sema3E at this site results in the generation of an active metastasis promoting ∼60 kDa fragment ( 24). However, the recombinant ∼60 kDa NH2-terminal fragment of sema3B was inactive. Furthermore, we have shown here that cleavage of sema3B at secondary PPC cleavage sites also results in substantial loss of activity. Interestingly, even the removal of a relatively small COOH-terminal peptide of 2 kDa, which in the case of sema3A potentiates sema3A activity, lead to a substantial loss of sema3B bioactivity.
Neither sema3A nor sema3F are efficiently cleaved by furin-like PPCs of cancer cells and the cleaved semaphorins found in the conditioned medium of sema3A- and sema3F-producing cancer cells usually represent in our experience no more than 10% to 20% of the total semaphorin found in the conditioned medium ( 11, 26, 27). In contrast, sema3B seems to be much more susceptible to cleavage by PPCs of cancer cells. We have found several melanoma- and colon cancer–derived cell lines that actively produce and secrete sema3B; however, in all these cells, the secreted sema3B was almost completely cleaved. This susceptibility suggests that in the case of sema3B, cleavage by up-regulated PPCs of malignant cells ( 22) may represent a major mechanism by which tumors may evade the antiangiogenic effects of sema3B in addition to down-regulation of expression and loss of heterozygosity ( 17, 34, 35). Indeed, HEK293 cells expressing the sema3B point mutant sema3B-m, which confers partial resistance to cleavage by PPCs, contain higher concentrations of full-length mutated sema3B in their conditioned medium. These cells repel endothelial cells more potently than HEK293 cells expressing native sema3B using a mechanism that depends on the presence of neuropilins. Furthermore, the full-length mutated sema3B found in their conditioned medium induced the contraction of endothelial cells, inhibited their proliferation/survival, induced apoptosis, inhibited VEGF-induced activation of ERK1/2, and inhibited tube formation by endothelial cells much more potently than conditioned medium containing a similar concentration of native sema3B. The proliferation/survival assays represent the balance between effects on cell proliferation and proapoptotic effects. It is possible that these effects are only secondary to the effects on the cytoskeleton and the adhesion and this will need to be examined in the future. Last, HEK293 cells secreting sema3B-m inhibited in vivo angiogenesis significantly better than HEK293 cells secreting similar amounts of native sema3B, suggesting that up-regulated furin-like PPCs produced in cancer cells ( 22) may contribute to tumor progression through inactivation of sema3B.
We also expressed sema3B and sema3B-m in MDA-MB-435 melanoma cells. In these cells, sema3B was cleaved by PPCs, but less efficiently than in the HEK293 cells. Expression of both sema3B and sema3B-m inhibited tumor formation similarly, although sema3B-m was marginally more efficient. The difference in the concentrations of full-length sema3B and sema3B-m found in the conditioned medium of these cells is not as pronounced as in the case of the HEK293 cells (compare Figs. 3A and 6C), which may be the reason for the relatively efficient inhibition seen with native sema3B. Furthermore, part of the inhibition may be due to a direct in vivo effect on these np2-expressing tumor cells because in the tumor, the local concentrations of sema3B to which the tumor cells are exposed may suffice to inhibit efficiently tumor development.
To conclude, our results suggest that sema3B is an inhibitor of angiogenesis. Our results further suggest that furin-like PPCs, previously shown to contribute to tumor progression through the activation of tumor-promoting enzymes and growth factors ( 22), may also contribute to tumor progression through the inactivation of angiogenesis inhibitors such as sema3B. Our results further suggest that Sema3B variants displaying higher resistance to PPC-mediated cleavage may perhaps find use as antiangiogenic agents.
Disclosure of Potential Conflicts of Interest
No potential conflicts of interest were disclosed.
Grant support: Israel Science Foundation, Komen Breast Cancer Foundation, International Union against Cancer, and Rappaport Family Institute for Research in the Medical Sciences of the Faculty of Medicine at the Technion, Israel Institute of Technology (G. Neufeld).
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 thank Dr. Susan Naylor (Johns Hopkins University, Baltimore, MD) for the sema3B cDNA and Dr. Stuart Aaronson (Mount Sinai School of Medicine, New York, NY) for helpful discussions and for the kind gift of lung cancer–derived cells and lentiviral expression vectors.
- Received September 15, 2007.
- Revision received June 17, 2008.
- Accepted June 25, 2008.
- ©2008 American Association for Cancer Research.