The association of semaphorin 5A with lymph node metastasis and adverse prognosis in cervical cancer

Tissue specimens were obtained from the Tumor Hospital of Harbin Medical University and First Clinical College of Harbin Medical University. Two hundred and thirty-two cervical cancer patients who underwent cervical surgery and dissection of pelvic/aortic lymph nodes between July 2004 and December 2011 were included in the study. Pathological features and clinical data of patients were obtained retrospectively from the hospital database. The median follow-up period was 96 months (range 12–97 months). None of the patients had any other specified carcinomas within 5 years of their cervical cancer diagnosis. The median age at the time of surgery was 49 years (range 23–69 years). Surgical treatments included cone biopsy in 27 patients (11.6%) and radical hysterectomy in 205 patients (83.4%). After surgery, 215 patients were treated with adjuvant therapy. Seventeen patients (7.3%) did not receive adjuvant therapy because they were at minimal risk or for other reasons.

Tumor stage, tumor grade, and LNM presence were assessed using HE-stained sections. Of the 232 patients, 156 patients (67.2%) had squamous cell carcinoma, 61 patients (26.3%) had adenocarcinoma, and 15 patients (6.5%) had adenosquamous carcinoma. Lymph node involvement was present in 121 of the 232 patients. Low, intermediate, and high tumor grades were confirmed in 47, 104, and 81 patients, respectively. Furthermore, 12 cases were stage Ia, 79 cases were stage Ib, 87 cases were stage IIa, and 54 cases were stage IIb according to the International Federation of Gynecology and Obstetrics (FIGO) staging system. The mean tumor size was 2.75 cm, and the mean depth of invasion was 10.3 mm. The study procedures were in accordance with the guidelines of the National Research Council and approved by the Research Ethics Committee of Harbin Medical University.

The HeLa human cervical carcinoma cell line, Siha and Caski human papillomavirus 16-positive cervical carcinoma cell lines were used as in vitro cervical cancer models. The HeLa cell line was provided from the Cancer Institute of Harbin Medical University. Another two cervical carcinoma cell lines Siha and Caski were obtained from the American Type Culture Collection (ATCC, Manassas, USA). HeLa cell line was cultured in RPMI 1640 (Invitrogen, Carlsbad, CA, USA). The Siha and Caski cell lines were cultured in DMEM (Invitrogen) supplemented with 10% FBS in a 5% CO₂ at 37 °C.

Immunohistochemistry was used to evaluate SEMA5A and lymphatic vessel endothelial hyaluronan receptor-1 (LYVE-1) expression in cervical cancer tissues. For immunostaining, 4-μm-thick tissue sections were deparaffinized in xylene, rehydrated through graded alcohol, and treated with 5% H₂O₂ to quench endogenous peroxidase activity. After antigen retrieval in citrate buffer, the sections were incubated with polyclonal goat anti-human SEMA5A (1:400 dilution; sc-67953, Santa Cruz Biotechnology, CA, USA) and polyclonal rabbit anti-human LYVE-1 primary antibodies (1:200 dilution; sc-19316, Santa Cruz Biotechnology) at 4 °C overnight. Species-appropriate biotinylated secondary antibodies were used for antigen detection. All subsequent immunohistochemistry procedures were carried out as previously described [10]. The negative control was obtained by using PBS instead of the primary antibody (Additional file 1: Figure S1A). Specimens with known SEMA5A expression served as positive controls (Additional file 1: Figure S1B). Scoring of SEMA5A was determined according to Fanourakis et al. [11]. Immunoreactivity for SEMA5A was evaluated according to the extent and intensity of staining. For statistical analysis, we divided patients into two groups. Tumor tissue with a score of ≥ 5 for SEMA5A staining was defined as the high expression and tumor tissue with a score of < 5 was regarded as the low expression. To assess lymphangiogenesis, lymphatic microvessel density (LMVD) was determined by LYVE-1 immunostaining as previously reported [12]. The area containing the most LYVE-1-stained vessels (hot spot) was identified by scanning the sections at low magnification. LMVD was then measured by counting the number of LYVE-1-positive vessels from three areas of the highest vessel density/section at 200×. Each brown-stained lumen was regarded as a single countable microvessel. Three sections/tumors were analyzed. The analysis of SEMA5A and LYVE-1 immunostaining was conducted by two independent observers without knowledge of any other variables or clinical data. Cases of disagreement were reanalyzed until a consensus was reached.

Quantitative real-time reverse transcription-polymerase chain reaction (RT-PCR) analysis was performed using an ABI PRISM 7000 Sequence Detection System (Applied Biosystems/Life Technologies, Foster, CA, USA). The following primer sequences were used: SEMA5A, 5′-GATCTATGGCATCTTTACCACCAA-3′ and 5′-TGG CGCTCAGGTTGAAGAC-3′; VEGF-C, 5′-AAGGAGGCTGGCAACATA-3′ and 5′-TGGCAGGGAACGTCTAAT-3′; matrix metalloproteinase (MMP)-2, 5′- CAGGAGGAGAAGGCTGTGTT-3′ and 5′-AGGGTGCTGGCTGAGTAGAT-3′; MMP-9, 5′-AGAACCAATCTCACCGACAGG-3′ and 5′-CGACTCTCCACGCATCTCT-3′; and GAPDH, 5′-GAGTCAACGGATTTGGTCGTA-3′ and 5′-ATGGGATTTCCATTGATGACA-3′. SEMA5A, MMP-2, MMP-9, and GAPDH expression levels were assessed using a fluorescence-based real-time detection method. For quantitative analyses, SEMA5A, MMP-2, and MMP-9 mRNA expression were normalized to GAPDH mRNA expression using previously published protocols [10].

For immunofluorescent antibody staining, chamber slides were precoated with SEMA5A (100 ng/mL; R&D Systems, Minneapolis, MN, USA) overnight at 4 °C. HeLa human cervical cancer cells were seeded into SEMA5A-coated and uncoated chamber slides for 24 h. The slides were fixed with fresh 4% paraformaldehyde for 15 min at room temperature, blocked in PBS containing 10% normal serum, and incubated with vascular endothelial growth factor (VEGF)-C primary antibody (1:100 dilution; sc-9047, Santa Cruz Biotechnology) for 2 h. The slides were incubated with fluorescein isothiocyanate (FITC)-conjugated IgG (1:100 dilution; Sigma-Aldrich, Beijing, China) for 1 h, counterstained with the fluorescent nuclear stain PI (Sigma-Aldrich, Beijing, China) for 5 min, and examined under a Nikon fluorescence microscopy.

VEGF-C protein levels in cell culture supernatants were determined to us the VEGF-C ELISA Kit (R&D Systems), according to the manufacturer’s instructions.

Western blot was carried out as previously described [10]. The following antibodies were used for western blot: anti-SEMA5A (1:200 dilution), anti-VEGF-C (1:200 dilution), anti-Tubulin (1:500 dilution; MAB1637, Chemicon International), anti-phosphotyrosine (1:100 dilution; clone 4G10, 16-105, Upstate Biotech), anti-Met (1:100 dilution; SP260, sc-162, Santa Cruz Biotechnology), anti-phosphorylated AKT (1:500 dilution; #9271, Cell Signaling Technology, Danvers, MA), anti-MMP-2 (1:300 dilution; sc-53630, Santa Cruz Biotechnology), and anti-MMP-9 (1:300 dilution; sc-6840, Santa Cruz Biotechnology).

To block the kinase activity of Met, the competitive Met inhibitor SU11274 (Sigma-Aldrich, St. Louis, MO, USA) and a Met neutralizing antibody (R&D Systems) were used. Cells were treated with 100 ng/mL SEMA5A in the presence or absence of SU11274 (2.5 μM) or Met neutralizing antibody (2 μg/mL) for 24 h. The concentration of SU11274 was used at 2.5 μM because inhibitory effect on cell growth was apparent with over 2.5 µM of SU11274 [13‐15]. To block phosphoinositide 3-kinase (PI3K)/AKT signaling, the PI3K inhibitor LY294002 (Cell Signaling Technology, Danvers, MA) was used. Cells were treated with 50 μmol/L LY294002 for 24 h. Cells were treated with 50 μmol/L LY294002 for 24 h. The concentration of LY294002 was set at 50 μmol/L in accordance with previous study [16].

Recombinant hepatocyte growth factor (HGF) was purchased from CalBiochem (San Diego, CA). The human cervical carcinoma cells were treated with HGF (30 ng/mL) and then incubated in a humidified incubator at 37 °C for 24 h. To examine the downstream signaling pathways involved in HGF treatment, cells were pretreated with the Met neutralizing antibody (2 μg/mL) or transfected with Met siRNA for 24 before addition of HGF (30 ng/mL). The small interfering RNAs (siRNAs) against Met and control siRNA were purchased from Santa Cruz Biotechnology (Santa Cruz, CA). VEGF-C in the medium was assayed using by RT-PCR and ELISA.

Plexin-B3 expression was knocked down to use short hairpin RNA (shRNA) interference technology. HeLa cells were transfected with plexin-B3-specific shRNA (5′-AGCAGATGGTGGAGAGGTA-3′) or a scrambled shRNA control (5′-GGCTACGTCCAGGAGCGCA-3′) using Lipofectamine (Invitrogen, Carlsbad, CA, USA). Plexin-B3 knockdown was confirmed by RT-PCR analysis.

Invasion assays were performed as described previously [9, 17]. Briefly, HeLa, Siha, and Caski cells were treated with 100 ng/mL recombinant human SEMA5A in the presence or absence of 1 μM GM6001, a MMP inhibitor (Calbiochem, La Jolla, CA, USA). After 5 h, cells (1 × 10⁵) were seeded onto 6.5-mm Costar transwells coated with matrigel (Corning, Cambridge, MA). After incubation for 24 h at 37 °C, cells from the top of the transwell chambers were removed to use a cotton swab, and the percentage of invaded cells was calculated.

HeLa cells were transfected with a mammalian expression vector containing full-length mouse Sema5A tagged with the FLAG epitope (HeLa-Sema5A) or empty vector alone (HeLa-control) as described previously [8].

Gelatinolytic activity was determined to use zymography as previously described [18]. Briefly, culture supernatants were collected and centrifuged. The supernatants were added to sodium dodecyl sulfate (SDS) sample buffer without mercaptoethanol and electrophoresed on an 8% polyacrylamide gel containing 1.5 mg/mL gelatin. The gel was washed with 2.5% Triton X-100 to remove SDS and incubated with developing buffer (50 mmol/L Tris–HCl, 0.2 mol/L NaCl, 5 mmol/L CaCl₂, and 0.02% Brij-35) overnight at 37 °C. Gelatinolytic bands were visualized by staining with Coomassie blue R-250 and destaining with Coomassie blue R-250 destaining solution until all lytic bands became clear. The gelatinolytic bands were quantified to use Image Acquisition and Analysis Systems (Ultra-Violet Products, Biospectrum HR410, USA).

The constitutive expression of SEMA5A protein differed according to cervical cancer stage. Of the 232 cervical cancer patients, 123 patients had high SEMA5A expression. SEMA5A protein expression was significantly higher in tumor tissues from stage IIb patients than in tumors tissues from stage Ia, Ib, and IIa patients (P < 0.05). Immunostaining demonstrated that SEMA5A was highly expressed in cervical cancer cells from stage IIb tumor tissues, comparable to that seen in nonsmall cell lung carcinoma (NSCLC), which was used as the positive control (Additional file 1: Figure S1B), but was expressed only at very low levels in stage Ia, Ib, and IIa tumor tissues (Fig. 1). SEMA5A mRNA level was also significantly higher in stage IIb tumors than in stage Ia, Ib, and IIa tumors (P < 0.05) but was not significantly different among stage Ia, Ib, and IIa tumors (Table 1). Thus, the RT-PCR analysis shows that the data of SEMA5A mRNA is increased in stage IIb tumors and supports SEMA5A protein expression demonstrated by IHC analysis.

The relationship of SEMA5A expression to LNM and lymphangiogenesis is demonstrated in Table 2. High SEMA5A expression was detected at a significantly greater frequency in metastatic patients than in nonmetastatic patients (P < 0.001). In addition, tumor cells that metastasized to the pelvic/aortic lymph nodes also showed high SEMA5A protein expression (Fig. 2A). Peritumoral LYVE-1-positive vessels were commonly seen in patients with LNM (Fig. 2B) and infrequently detected in patients without LNM (Fig. 2C). LYVE-1-positive LMVD was significantly higher in patients with LNM compared with those without LNM (9.1 ± 0.5 microvessels/field vs. 6.7 ± 0.8 microvessels/field; P < 0.001). LYVE-1-positive lymphatic vessels invaded by tumor clusters were occasionally observed (Fig. 2D). Lymphovascular invasion was detected in 42 cervical cancer cases. The tumor cells within these lymphatic vessels were more often judged to have high SEMA5A expression. Among 42 lymphovascular invasions, high SEMA5A expression was evident in 33 (78.5%), in 30 with LNM and in 3 without LNM. SEMA5A high expression patients were more likely than low expression patients to have lymph node metastasis (P < 0.001).

The mean follow-up interval was 40.6 months (range 12–97 months). During the follow-up interval, 95 patients (42.8%) developed recurrent disease. SEMA5A expression was significantly correlated with DFS. DFS was significantly (P < 0.001) lesser in patients with high tumor SEMA5A expression levels than for those with low tumor SEMA5A expression levels (Fig. 3a). High SEMA5A expression was also associated with a significantly shorter OS (P < 0.001; Fig. 3b). The median OS was 61.0 months in patients with high SEMA5A-expressing tumors (n = 123) and 81.0 months in patients with low SEMA5A-expressing tumors (n = 109). In the metastatic group (n = 121), median OS was 47.0 months in patients with high SEMA5A-expressing tumors (n = 80) and 73.0 months in patients with low SEMA5A-expressing tumors. The 5-year survival rates were 12.5 and 34.6% for the high and low SEMA5A expression groups, respectively. Therefore, high tumor SEMA5A expression level was a significant predictor of poor prognosis.

The association between SEMA5A expression and various clinicopathological parameters is shown in Table 3. In the univariate analysis, tumor grade (P < 0.001), FIGO stage (P < 0.001), LNM (P < 0.001), presence of lymphatic invasion (P < 0.001), VEGF-C expression (P < 0.001), LMVD (P < 0.001), and SEMA5A expression (P = 0.004) were significant prognostic factors for OS. Multivariate analysis was performed to evaluate the independent prognostic role of SEMA5A after adjusting for other significant covariates. SEMA5A expression (P = 0.032), tumor grade (P < 0.001), FIGO stage (P < 0.001), LMN (P < 0.001), presence of lymphatic invasion (P < 0.001), VEGF-C expression (P < 0.001), and LMVD (P = 0.005) remained independent prognostic factors for OS in the multivariate analysis.

We performed western blots for SEMA5A using protein lysates collected from in Hela, Siha, and Caski cervical cancer cells. We observed bands at ~ 135 and ~ 110 kDa (Additional file 2: Figure S2). They are membrane-bound SEMA5A and equivalent to the molecular weight of its extracellular domain, respectively [19]. The results shown in Fig. 4a revealed that SEMA5A expression was similar between Caski cells, which are derived from an intestinal metastasis of cervical cancer, and Hela and Siha cells, which derived from primary cervical cancer. However, our findings from cervical cancer patients suggested that SEMA5A was likely involved in the lymphangiogenesis and capacity of cervical cancer cells to metastasize to the lymph nodes. In view of the similar structural features between lymph and blood vessels, we speculated that SEMA5A may also affect the lymphatic system in vitro. In support of this, recombinant human SEMA5A has been shown to increase VEGF-C expression, which plays a key role in lymphangiogenesis [20]. Therefore, we determined the effect of SEMA5A on VEGF-C expression using fluorescent antibody staining and western blot. FITC-labeled VEGF-C localized to the cytoplasm and cytoplasmic surface of the plasma membrane in Hela cells. We found that SEMA5A enhanced VEGF-C expression compared with the control (Fig. 4b). Similar results were obtained in the western blot analysis of HeLa, Siha, and Caski cells (Fig. 5a).

Recently, Met has been implicated in lymphangiogenesis through its association with VEGF-C [21]. Therefore, western blot analysis was used to determine whether SEMA5A induced VEGF-C expression through a Met-dependent signaling pathway. Inhibition of MET with the Met-specific inhibitor SU11274 or a Met neutralizing antibody decreased the SEMA5A-mediated enhancement of VEGF-C expression in Hela, Siha, and Caski cells (Fig. 5a). Qualitative measurement of VEGF-C levels by ELISA in the three cervical cancer cell lines was consistent with the western blot results (Fig. 5b). The known ligand for MET is HGF. Next, we examine whether Met receptor is involved in HGF-mediated VEGF-C production in Hela, Siha, and Caski cells. MET neutralizing antibody blocked HGF-stimulated VEGF-C mRNA and protein level. Furthermore, transfection with Met siRNA also inhibited HGF-increased VEGF-C production (Additional file 3: Figure S3A, B). In addition, SEMA5A treatment induced Met tyrosine phosphorylation (Fig. 6a).

SEMA5A has been found to interact with plexin-B3, a high-affinity SEMA5A-specific receptor [22, 23]. Moreover, plexin-B3 has also been proposed to interact with the receptor tyrosine kinase Met. Therefore, we tested whether SEMA5A-induced Met phosphorylation is dependent upon plexin-B3 using plexin-B3 RNA interference in HeLa cells. Western blot demonstrated Plexin-B3 knockdown inhibited SEMA5A-induced Met phosphorylation. Quantifying the bands and the statistical analysis showed that SEMA5A signals through MET (Fig. 6a). Moreover, we tested whether plexin-B3 was involved in the induction of VEGF-C expression by SEMA5A. RT-PCR analysis showed that plexin-B3 knockdown significantly decreased VEGF-C mRNA and protein expression compared with the control (Fig. 6b, c). Our results indicate that SEMA5A induces VEGF-C overexpression by activating Met signaling via plexin-B3.

To determine the proinvasive and prometastatic activities of SEMA5A, matrigel invasion assays were performed. SEMA5A treatment significantly increased the invasion of HeLa, Siha, and Caski cells relative to control cells. MMP-2 and MMP-9 can degrade type IV collagen, which contributes to tumor invasion and metastasis [24, 25]. Therefore, we determined whether SEMA5A-induced invasion involves MMP-2 and MMP-9. The MMP inhibitor GM6001 reduced the ability of SEMA5A to stimulate cervical cancer cell invasion (Fig. 7a). Next, we tested whether SEMA5A-induced MMP-2 and MMP-9 expression is dependent upon plexin-B3. Plexin-B knockdown strongly reduced SEMA5A-induced MMP-2 and MMP-9 in HeLa cells (Fig. 7b, c). Because the PI3K/AKT signaling pathway has been shown to increase tumor invasion and metastasis by regulating MMP-2 and MMP-9 [26, 27], we investigated the role of this pathway in SEMA5A-induced invasion. To do this, the protein expression and activity of MMP-2 and MMP-9 were determined in HeLa-Sema5A cells and HeLa-control cells treated with and without the PI3K inhibitor LY294002. MMP-2 and MMP-9 protein expression and activities were induced in HeLa-Sema5A cells but not in HeLa-control cells (Fig. 7d, e). Compared with the control, SEMA5A increased MMP-2 and MMP-9 activities by 52.7 and 38.2%, respectively. LY294002 reduced AKT phosphorylation to basal levels observed in control cells and nearly reversed the stimulatory action of SEMA5A on MMP-2 and MMP-9 expression and activities (Fig. 7d, e). Our findings indicate that the proinvasive activity of SEMA5A is associated with the induction of MMP-2 and MMP-9 in HeLa cervical cancer cells. Furthermore, the induction of MMP-2 and MMP-9 by SEMA5A is mediated by plexin-B3 and the PI3K/AKT pathway.