Semaphorin SEMA3F Affects Multiple Signaling Pathways in Lung Cancer Cells

The semaphorin SEMA3F gene was originally cloned from a recurrent homozygous deletion involving chromosomal region 3p21.3 (1–3). SEMA3F is abundantly expressed in normal lung and cultured bronchial epithelial cells. In contrast, it is markedly down-regulated in most human lung cancers and loss of SEMA3F protein staining is significantly correlated with advanced stage of disease and vascular endothelial growth factor-165 (VEGF₁₆₅) overexpression (4, 5). We previously reported that retroviral-mediated replacement of SEMA3F expression blocked tumor formation by H157 lung cancer cells when injected into the lungs of immunodeficient rats. This was associated with loss of activated α_Vβ₃ integrin, impaired cell adhesion to extracellular matrix (ECM), and down-regulation of phospho-extracellular signal-regulated kinase 1/2 (ERK1/2; ref. 6).

Class 3 semaphorins, including SEMA3F, are secreted proteins originally identified as mediators of growth cone repulsion (7). With the exception of SEMA3E, class 3 semaphorins use high-affinity neuropilin (NRP) 1 and NRP2 receptors and plexin coreceptors to mediate their biological activities (see ref. 8 for review). NRP1 was also independently identified as a coreceptor for VEGF₁₆₅ and placenta growth factor-2 (9, 10), and a subset of class 3 semaphorins has been shown to competitively inhibit VEGF₁₆₅ binding and function (11, 12). In addition to their effects on Rac activation (13, 14), class 3 semaphorins regulate integrin function (15, 16). This can be explained by the GTPase-activating protein activity of plexins for R-Ras (17, 18) and sequestration of talin by PIPKI-γ661 (14). Integrins, composed of transmembrane α and β heterodimers, bind ECM components and signal in a bidirectional manner (19). This suggested to us that integrin-linked kinase (ILK), which interacts with β-integrins and has multiple protumorigenic activities including AKT phosphorylation (20, 21), might be affected by SEMA3F. In the present report, we found that ILK activity was inhibited by SEMA3F in H157 lung cancer cells, along with down-regulation of adhesion to vitronectin. Exposure to SEMA3F, or inhibition of ILK by small interfering RNA (siRNA), led to decreased ERK1/2 signaling, which was observed in different lung cancer cell lines. Moreover, SEMA3F reduced AKT and signal transducer and activator of transcription 3 (STAT3) phosphorylation in H157 cells in a manner that was independent of ILK but consistent with Src inhibition. These changes were accompanied by downstream loss of hypoxia-inducible factor-1α (HIF-1α) protein and decreased VEGF mRNA levels in three of four SEMA3F stably transfected cell lines. Lastly, we show that these effects were mirrored by reductions in microvessel density and tumor growth following s.c. implantation of SEMA3F-expressing H157 cells in nude mice. These results show that SEMA3F inhibits diverse signaling pathways in responsive lung cancer cells, possibly contributing to its observed antitumor activity.

Cell lines and animals. All cultures were grown in regular RPMI 1640 supplemented with 10% FCS and appropriate selection drugs. Establishment and maintenance of NCI-H157 and NCI-H460 (hereafter H157 and H460) lung cancer cells stably transfected by retrovirus with SEMA3F has been described previously (6). Briefly, myc-tagged SEMA3F was cloned into pLXSN retroviral vector and cells were infected using retrovirus-containing culture medium from PT67 packaging cells. Stable SEMA3F (or empty vector)-expressing MCF7 and COS7 cell populations were obtained by retroviral transfection as described above and in ref. 6. Cells were selected and grown with 500 μg/mL G418 (Invitrogen). Establishment of H157 cells with stably integrated SEMA3F under the control of a tetracycline-inducible promoter was obtained following the Flp-In T-REx Core kit instructions (Invitrogen) after cloning 3′ myc-tagged SEMA3F in pcDNA5/FRT/TO (Invitrogen) at EcoRV site. Empty vector–transfected (“inducible H157EV”) and SEMA3F-transfected (“inducible H157S3F”) cell populations were selected and grown with 100 μg/mL hygromycin and 5 μg/mL blasticidin (Invitrogen). SEMA3F was induced for 5 days with doxycycline at 5 or 100 ng/mL. Microsatellite genotyping was done by the UC Cancer Center DNA Sequencing Core to confirm authenticity of cells. Supplementary information is available on SEMA3F mRNA expression and related transcripts (Supplementary Table S1). All animal experiments followed the guidelines of the Animal Committee of the Regienrungsprasidium Freiburg (Germany).

Proteins and reagents. Recombinant proteins were purchased from Chemicon International (Millipore), except VEGF and vitronectin (R&D Systems). LY294002, cobalt chloride (CoCl₂), and cycloheximide were from Sigma-Aldrich. PP2 and PP3 were from Calbiochem. UO126 was from Promega and doxycycline was from Sigma-Aldrich. SEMA3F-conditioned media were prepared as follows: 1 × 10⁶ H157S2 cells were plated in a 100-mm dish with 4 mL RPMI 1640 + 10% FCS. After 24 h of culture, media were collected and centrifuged briefly to eliminate cell residue. H157C-conditioned medium was used as a control.

RNA expression. Total RNA was extracted using the RNeasy Mini kit (Qiagen). Reverse transcription-PCR (RT-PCR) was done with SuperScript III (Invitrogen) using the procedure supplied by the manufacturer. Gene expression was assessed relative to glyceraldehyde-3-phosphate dehydrogenase (GAPDH) by quantitative real-time PCR with the GeneAmp 5700 Sequence Detection System and SYBR Green chemistry (Applied Biosystems; ref. 6). Primer sequences were as follows: GAPDH, 5′-TGCACCACCAACTGCTTAGC-3′ and 5′-GGCATGGACTGTGGTCATGAG-3′; HIF-1α, 5′-CATTACCCACCGCTGAAACG-3′ and 5′-TTCACTGGGACTATTAGGCTC-3′; and VEGF, 5′-CAAGACAAGAAAATCCCTGTGG-3′ and 5′-CCTCGGCTTGTCACATCTG-3′.

siRNA transfection. ILK and related scramble siRNAs were purchased from Ambion and transfected using LipofectAMINE 2000 as recommended (Invitrogen).

Immunoblotting. Immunoblots were done as reported (6). The primary antibodies rabbit anti-AKT, anti–phospho-Ser⁴⁷³-AKT, anti-STAT3, anti–phospho-Tyr⁷⁰⁵-STAT3, and anti–phospho-Thr²⁰²/Tyr²⁰⁴-ERK1/2 were from Cell Signaling Technology and used at 1:1,000 dilution. Anti-ERK1/2 (1:4,000) and phospho-Tyr³⁹⁷-focal adhesion kinase (FAK; 1:1,000) were from Promega and Chemicon, respectively. Anti-HIF-1α (1:300) and anti–phospho-serine/threonine (1:1,000) were from Transduction Laboratories. Anti-ILK (1:1,000) was from Upstate Biotechnology. Antiactin and antitubulin were from Sigma-Aldrich (1:2,000). Horseradish peroxidase (HRP)-conjugated antimouse and antirabbit secondary antibodies (Perkin-Elmer) were used at 1:5,000 dilution. Detection was done with enhanced chemiluminescence (Perkin-Elmer).

Immunoprecipitation and kinase assay for ILK. Ten million cells were lysed in 500 μL of lysis buffer [50 mmol/L HEPES (pH 7.4), 100 mmol/L NaCl, 1.5 mmol/L MgCl₂, 0.5% CHAPSO, 1 mmol/L NaF and Na₃VO₄, 5 μg/mL leupeptin and aprotinin, 1 μg/mL pepstatin A]. Protein (500 μg) was precleared by incubation with 50 μL protein G-agarose beads (Roche) and then centrifuged for 5 min at 900 rpm. The supernatant was incubated overnight at 4°C with 5 μg of anti-ILK. Protein G-agarose beads were added for 3 h and immune complexes were pelleted for 1 min at 3,000 rpm. Pellets were washed thrice in lysis buffer and resuspended in 100 μL of kinase assay buffer [50 mmol/L HEPES (pH 7.4), 2 mmol/L each of MnCl₂, MgCl₂, Na₃VO₄, and NaF]. In parallel, purity of the immunoprecipitate was checked by total protein staining using Silver Stain Plus (Bio-Rad). The kinase reaction was done by adding 2 μL of 10 mmol/L ATP and 2 μg of myelin basic protein (MBP; Upstate Biotechnology) to 50 μL of ILK immunoprecipitate resuspended in kinase assay buffer. The reaction was incubated for 30 min at 30°C and stopped by denaturing samples at 100°C in Laemmli electrophoresis loading buffer.

Immunofluorescence. Cells grown on glass coverslips were fixed for 15 min with 3.7% formalin. After rinsing with PBS, primary and secondary antibodies were each applied for 90 min at room temperature and then rinsed. Anti–HIF-1α was used at 1:100 dilution. Activated α_vβ₃ integrin was detected as described (6). Alexa⁴⁸⁸-conjugated antirat and Alexa⁵⁶⁴-conjugated antirabbit secondary antibodies (1:250) were from Molecular Probes. Stained slides were rinsed in PBS and mounted in Vectashield (Vector Laboratories), and images were captured with a confocal microscope (Olympus FluoView FV1000) at 100× immersion oil objective.

Adhesion assay. One thousand untransfected H157 cells in 100 μL of fresh medium were mixed to 100 μL of SEMA3F or control-conditioned medium and deposited on vitronectin-coated (0.5 μg/mL) wells. Cells were allowed to attach for 1 h and processed as detailed previously (6).

VEGF ELISA. Human VEGF dosage on culture supernatants was done following the manufacturer's instructions (R&D Systems).

Tumor xenograft experiments. Female athymic nude mice (NMRI-nu/nu) were purchased from Charles River Laboratories. S.c. injection of 2.5 × 10⁵ cells in Matrigel was done into the right flanks using a 29-gauge needle syringe. Tumor volume (mm³) was calculated by caliper measurements of the largest diameter (a) and its perpendicular (b) according to the following formula: volume = 0.5ab².

Immunohistochemistry and microvessel density quantification. Cryosections (8 μm), fixed with 4% paraformaldehyde, were stained for mouse CD31 using a rat primary antibody (1:100; BD PharMingen) and Alexa⁴⁸⁸-conjugated antirat secondary antibody (1:200). Slices were counterstained with 4′,6-diamidino-2-phenylindole and microvessel density was determined by counting the number of CD31⁺ blood vessels using the program Cell-P (Olympus).

Inhibition of ERK1/2, AKT, and STAT3 by SEMA3F. Previously, we reported that inhibition of in vivo tumor formation by SEMA3F is associated with decreased ERK1/2 activation (6). To extend these investigations, we asked if alternative signaling pathways might also be affected by SEMA3F. For this, we first used the H157 lung cancer cell line that has been previously modified by retrovirus-mediated transfection to constitutively reexpress SEMA3F (6). One control and two independent SEMA3F (S1 and S2) clones were selected. Expression of SEMA3F was verified at the mRNA and protein levels, and secretion of SEMA3F was shown by immunoblot analysis of culture medium (see Supplementary Fig. S1A and Supplementary Table S1). At the mRNA level, SEMA3F expression was approximately 2-fold (clone S1) to 8-fold (clone S2) greater than detected in normal lung, and there was no significant clonal variations (6). As shown in Fig. 1A, we found that phospho-AKT (Ser⁴⁷³) and phospho-STAT3 (Tyr⁷⁰⁵) were down-regulated in SEMA3F stably transfected H157 cells (H157S2) compared with controls (H157C). Similar results were also observed in the independent SEMA3F-expressing clone, H157S1 (data not shown). Furthermore, the addition of exogenous growth factors did not reverse the SEMA3F effects on phospho-AKT and phospho-STAT3. In contrast, stimulation with growth factors was able to overcome the inhibition of phospho-ERK1/2 (Fig. 1A).

To confirm the effects on AKT-STAT3, we generated stable isogenic H157 control (empty vector) and SEMA3F cells in which SEMA3F expression is directed by a tetracycline/doxycycline-inducible promoter (inducible H157; Fig. 1B). This approach eliminates clonal selection. SEMA3F induction following doxycycline treatment was verified at the protein level (Supplementary Fig. S1B). In agreement, after induction of SEMA3F with doxycycline, decreases in AKT and STAT3 phosphorylation were evident (Fig. 1B). Next, we screened a panel of eight lung cancer cell lines, including H157, to estimate the frequency of SEMA3F-responsive cell lines (Fig. 1C). Cells were exposed to control or SEMA3F-conditioned medium and signaling changes were analyzed by immunoblot. Among the pathways studied, ERK1/2 was found to be the most sensitive marker with four of eight cell lines responding. It is possible that transfection with SEMA3F would result in a higher frequency of STAT3 and AKT responses. We concluded that responses to SEMA3F are frequent among lung cancer cell lines and that phospho-ERK1/2 inhibition is the most sensitive biochemical indicator with the use of cultured supernatants.

Possible involvement of integrin outside-in signaling in SEMA3F effects. Beyond its effects on ERK1/2 phosphorylation, expression of SEMA3F leads to the loss of activated α_vβ₃ staining in H157 cells and reduced adhesion to ECM (6). Indeed, exposure of H157 cells to SEMA3F-conditioned medium reduced adhesion to vitronectin, and interestingly, growth factors stimulation did not rescue this inhibitory effect (Fig. 2A). We therefore explored the relationship between adhesion and the signaling effect of SEMA3F. We first asked whether AKT or ERK1/2 inactivation by specific inhibitors (LY294002 and UO126, respectively) would mimic the SEMA3F effect on activated α_vβ₃ integrin or adhesion to vitronectin. As shown in Fig. 2B, neither inhibitor affected α_vβ₃ activation (as determined with WOW-1 staining) or adhesion, although the drugs did inhibit their respective targets. Thus, AKT and ERK1/2 do not inhibit adhesion by inside-out signaling. On the other hand, adhesion inhibition by SEMA3F could mediate loss of phospho-ERK1/2 and phospho-AKT by outside-in signaling. Accordingly, brief treatment of H157 cells with EDTA to dissociate their membrane contacts resulted in down-regulation of phospho-AKT, and plating cells on vitronectin after cell detachment increased AKT activation (Fig. 2C). These results support the hypothesis that SEMA3F impairs the maintenance of active signaling pathways by inhibiting integrin-mediated outside-in signaling.

ILK inhibition by SEMA3F leads to phospho-ERK1/2 loss. Integrins connect various protein kinases, including ILK, which mediate intracellular signaling. To determine if ILK activity was affected by SEMA3F, we immunoprecipitated ILK from control (H157C) or SEMA3F (H157S2) cells. As shown in the silver-stained gel and total ILK immunoblot in Fig. 3A, equal amounts of ILK were precipitated. We found that ILK activity, as determined by an in vitro kinase assay using MBP as substrate, was reduced in H157S2 cells by 55% (Fig. 3A,, right). This finding was specific because phospho-FAK (Tyr³⁹⁷) was unchanged (data not shown). ILK activation has been shown to induce AKT phosphorylation on residue Ser⁴⁷³ (22), and ILK inhibition, at least in some cellular contexts, leads to down-regulation of phospho-STAT3 over time (23). Thus, we asked if ILK inhibition could account for the signaling changes observed with SEMA3F. Surprisingly, we found that ILK inhibition by siRNA in H157 cells caused loss of phospho-ERK1/2 after 72 h but not phospho-AKT (Fig. 3B). Moreover, STAT3 was unexpectedly activated probably due to increased total levels. These results suggested that other integrin-associated proteins control AKT-STAT3 in H157 cells. Accordingly, inhibition of Src family kinases by PP2 lowered AKT phosphorylation, and phospho-STAT3 was slightly affected (Fig. 3C). Together, these data indicate that ILK is a link between SEMA3F effects on adhesion and ERK1/2 signaling, and preliminary data suggest that the effects on AKT-STAT3 result from a Src family kinase inhibition.

Repercussion of AKT pathway inhibition by SEMA3F on HIF-1α/VEGF. HIF-1α is regulated at the level of protein initiation by activation of AKT and its downstream effector mammalian target of rapamycin (mTOR; ref. 24). HIF-1α protein levels are also regulated by von Hippel-Lindau (VHL)-mediated polyubiquitination and subsequent proteasome-dependent degradation (25). At the transcriptional level, both HIF and phospho-STAT3 regulate VEGF-A mRNA levels (26). Based on the above results, we compared HIF-1α and VEGF levels in SEMA3F-expressing H157 cells and controls. HIF-1α protein levels were reduced in SEMA3F-expressing clones both in normoxia and severe hypoxia (0.01% O₂; Fig. 4A,, left). This loss was further confirmed by exposure of parental H157 cells to SEMA3F-conditioned medium (data not shown). By immunofluorescence (Fig. 4A,, right), nuclear HIF-1α protein levels were markedly diminished in SEMA3F-expressing clones, in normoxia, and in response to the severe hypoxia mimetic CoCl₂. To determine the mechanisms underlying HIF-1α loss, we examined HIF-1α transcript levels by real-time RT-PCR and found them unchanged (Fig. 4B,, left). To examine HIF-1α protein stability, cells were first treated with the proteasome inhibitor MG132 to allow HIF-1α accumulation and then washed and treated with cycloheximide to inhibit new protein synthesis. As shown in Fig. 4B, HIF-1α levels disappeared at approximately the same rate in control and SEMA3F-expressing cells (S2). We also transfected H157C and S2 clones with a HIF-1α mutant insensitive to VHL-dependent degradation and found that SEMA3F-expressing cells had substantially lower levels than control cells (data not shown). Therefore, the reduction of HIF-1α in H157 cells by SEMA3F is not the result of impaired transcription or enhanced degradation. Thus, it is likely that HIF-1α protein translation, which is promoted by AKT-mTOR (24), is reduced by SEMA3F. Consistently, inhibition of AKT with LY294002 also led to loss of HIF-1α (Fig. 4C,, left). We examined HIF-1α protein levels in three other SEMA3F stably transfected cell lines from different origin (Fig. 4C , right). HIF-1α level was reduced in all SEMA3F-expressing cell lines except the MCF7 mammary adenocarcinoma cells in which it was increased; the reason of which is unknown. Furthermore, CoCl₂ treatment markedly induced HIF-1α in control cells but only weakly in SEMA3F-H460 and SEMA3F-COS7 cells. Based on these results, we conclude that SEMA3F down-regulates HIF-1α through inhibition of AKT and protein translation.

In agreement with the effects on HIF-1α, we found that both VEGF mRNA and protein levels were significantly reduced in the SEMA3F-expressing H157 and H460 cells (Fig. 4D). The same result was obtained with SEMA3F-transfected COS7 cells (VEGF mRNA reduced by 74%; Supplementary Table S1). In contrast, a modest increase of VEGF mRNA was seen in MCF7 (+18% of controls; Supplementary Table S1).

Inhibition of tumor growth by SEMA3F. VEGF is a well-established master regulator of angiogenesis in vivo. We consequently questioned whether SEMA3F up-regulation would interfere with tumor angiogenesis and tumor growth. A Matrigel suspension of H157 control or SEMA3F-expressing cells was injected s.c. into the right flanks of nude mice. Interestingly, there was no apparent inhibition of tumor growth by SEMA3F until day 14 (Fig. 5A). By day 24, however, tumor volumes (Fig. 5A) and microvessel density (Fig. 5B and C), as determined by CD31 (platelet/endothelial cell adhesion molecule-1) immunostaining, were significantly reduced by SEMA3F expression. H157S1- and H157S2-induced tumors had an average volume of 0.2 cm³, whereas control tumors had grown to 1.8 ± 0.3 cm³. Correspondingly, tumor weights were decreased by ∼94% (data not shown). Blood vessels were more homogeneously distributed in tumors derived from H157C compared with H157S1 and H157S2 tumors. Likewise, the vessels were also strikingly smaller in the SEMA3F-H157–induced tumors. Interestingly, SEMA3F also prevented development of H460-derived tumors, although the vascularization was not significantly different and the suppressive effect on tumor growth was not as dramatic as with H157 cells. These results indicate that angiogenesis inhibition by SEMA3F only partially contributes to the tumor prevention and that additional effects, such as signaling inhibition, might play an independent role.

The reexpression of SEMA3F in cancer cells was previously shown to impair in vivo tumorigenicity in three different s.c. models (27–29). More recently, using a lung orthotopic model, we reported that SEMA3F blocked H157 lung cancer cell-derived tumorigenesis and suggested that down-regulation of active α_vβ₃ integrin and phospho-ERK1/2 could explain this effect (6). In the present report, we have shown that ILK activity is down-regulated in SEMA3F-transfected H157 cells, and inhibition of ILK leads to reduced phospho-ERK1/2. In parallel, SEMA3F reexpression caused reduction in phospho-AKT (Ser⁴⁷³) and phospho-STAT3 (Tyr⁷⁰⁵), with downstream effects on HIF-1α and VEGF (see speculative model in Fig. 6). In contrast to phospho-ERK1/2, phospho-STAT3 and phospho-AKT were not inhibited by a knockdown of ILK. The AKT and STAT3 effects are likely mediated by a Src family kinase, which are known to constitutively associate with β-integrin tails. To our knowledge, this is the first report describing inhibition of ILK, STAT3, and HIF by semaphorins. Importantly, the effects of SEMA3F were not limited to H157 cells. Using transferred cultured supernatants, ERK1/2 inactivation by SEMA3F was found in four of eight lung cancer cell lines. In addition, HIF-1α and VEGF reduction were observed in three of four cell lines of different origins stably transfected with SEMA3F.

Recently, Ito et al. (30) reported that Sema4D suppressed R-Ras activity along with dephosphorylation of AKT in hippocampal neurons. This relationship was subsequently strengthened by evidence that R-Ras is upstream of ILK and glycogen synthase kinase-3β in this system (31). In addition, SEMA3A inhibited phospho-AKT in dorsal root ganglia growth cones and breast cancer cell lines through wild-type phosphatase and tensin homologue (PTEN; ref. 32). However, H157 cells have a mutant PTEN gene (33), which we confirmed by DNA sequence analysis and the absence of PTEN protein (Supplementary Fig. S1C). In addition, Ser⁴⁷³ of AKT has been directly shown to be a target of ILK (34–36), requiring α_vβ₃ integrin function that is independent of PTEN (35). Thus, PTEN is dispensable for SEMA3F signaling in some cellular contexts.

We found that HIF-1α is inhibited by SEMA3F at the level of protein translation initiation in H157 cells, with corresponding effects on VEGF mRNA and protein. Similar decreases in HIF-1α and VEGF were observed in additional (two of three) cell lines. This likely results from AKT inactivation by SEMA3F because inhibiting AKT with LY294002 induced loss of HIF-1α. In agreement, Vacca et al. (37) reported that SEMA3A reduced VEGF mRNA levels in endothelial cells isolated from patients with multiple myeloma, although the mechanism was not explored. Of interest, VEGF transcription is also regulated by a STAT3 binding site in the VEGF promoter (26), with an important contribution of Tyr⁷⁰⁵ STAT3 phosphorylation (38). Thus, SEMA3F-mediated inhibition of VEGF mRNA probably involves both HIF and STAT3 effects. Accordingly, both microvessel density and size were reduced in s.c. tumors induced by SEMA3F-H157 cells compared with control cells. Tumor size was also reduced after an initial period of indistinguishable growth, consistent with an angiogenic effect. Although VEGF may be one of the most important angiogenic factors affected by SEMA3F, our preliminary data indicate that other angiogenic mediators are also involved (data not shown).

Lastly, because SEMA3F inhibits several signaling components, it is tempting to hypothesize that up-regulation of SEMA3F would be more effective than an ILK, AKT, or ERK1/2 inhibition alone, although combined inhibitors may recapitulate the effects of SEMA3F. To up-regulate endogenous SEMA3F expression in tumor cells, only nonspecific histone deacetylase inhibitors can be presently used for this purpose (39); thus, further studies to understand SEMA3F gene regulation are necessary.

Grant support: University of Colorado Lung Cancer Specialized Program of Research Excellence grant CA58187 (R.M. Gemmill and H.A. Drabkin), La Ligue Contre le Cancer (V. Potiron, J.A. Clarhaut, and J. Roche), l'Association pour la Recherche sur le Cancer (V. Potiron, J.A Clarhaut, and J. Roche), Alexander Von Humboldt Foundation (P. Nasarre), and Deutsche Forschungsgemeinschaft SPP1190 Au83/9-1 (H.G. Augustin).

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.