Tumor Suppressor Function of the SEMA3B Gene in Human Lung and Renal Cancers

The SEMA3B gene is located in the 3p21.3 LUCA region, which is frequently affected in different types of cancer. The objective of our study was to expand our knowledge of the SEMA3B gene as a tumor suppressor and the mechanisms of its inactivation. In this study, several experimental approaches were used: tumor growth analyses and apoptosis assays in vitro and in SCID mice, expression and methylation assays and other. With the use of the small cell lung cancer cell line U2020 we confirmed the function of SEMA3B as a tumor suppressor, and showed that the suppression can be realized through the induction of apoptosis and, possibly, associated with the inhibition of angiogenesis. In addition, for the first time, high methylation frequencies have been observed in both intronic (32-39%) and promoter (44-52%) CpG-islands in 38 non-small cell lung carcinomas, including 16 squamous cell carcinomas (SCC) and 22 adenocarcinomas (ADC), and in 83 clear cell renal cell carcinomas (ccRCC). Correlations between the methylation frequencies of the promoter and the intronic CpG-islands of SEMA3B with tumor stage and grade have been revealed for SCC, ADC and ccRCC. The association between the decrease of the SEMA3B mRNA level and hypermethylation of the promoter and the intronic CpG-islands has been estimated in renal primary tumors (P < 0.01). Using qPCR, we observed on the average 10- and 14-fold decrease of the SEMA3B mRNA level in SCC and ADC, respectively, and a 4-fold decrease in ccRCC. The frequency of this effect was high in both lung (92-95%) and renal (84%) tumor samples. Moreover, we showed a clear difference (P < 0.05) of the SEMA3B relative mRNA levels in ADC with and without lymph node metastases. We conclude that aberrant expression and methylation of SEMA3B could be suggested as markers of lung and renal cancer progression.


Introduction
Semaphorins are negative mediators of axonal guidance in the central nervous system [1]. Semaphorins comprise a large family of glycoproteins (8 classes, including 5 vertebrate classes, of more than 30 members), but only class 3 (SEMA3) represents secreted soluble molecules. Members of the SEMA family are differentially expressed in cancer, and either promote or suppress cell proliferation, migration and angiogenesis, and the induction of drug resistance. Thus, the roles of separate members of semaphorin family can be quite different [2][3][4][5][6][7][8][9].
Class 3 semaphorins (SEMA3s, also known as collapsins) comprise one of five vertebrate families of semaphorins and play an important role in tumor biology, including regulating cellular processes, such as endothelial cell proliferation, apoptosis, migration and angiogenesis [10]. Recently, the involvement of this protein class in carcinogenesis has been intensively studied. SEMA3s are secreted by cells of multiple lineages, including epithelial cells, neurons, and specific tumor cells [10]. Neuropilins (NRP) represent the primary receptors of SEMA3s. The binding of SEMA3s to NRP1/2 initiates their downstream signaling but prevents the interaction between NRP1/2 and vascular endothelial growth factor (VEGF) and the subsequent induction of a pro-angiogenic transcriptional program. However, it is not clear whether SEMA3s inhibit tumor growth by competing with VEGF for neuropilins ligand-binding sites, by acting independently of VEGF, or by a combination of these effects [10][11][12][13].
Important evidence of tumor suppressor activity includes the identification of cell regulatory pathways and other mechanisms that are affected by SEMA3B. Using MDA-MB435 (breast carcinoma) and A549 (lung adenocarcinoma) cells it was previously shown that SEMA3B suppressed tumor growth but triggered a pro-metastatic program by releasing interleukin 8 [19,20]. Furthermore, it was found that the induction of apoptosis by SEMA3B in tumor cells was mediated by inactivation of the Akt signaling pathway [21]. Therefore, it was important to further elucidate particular aspects of SEMA3B tumor suppression.
Methylation is an important mechanism of SEMA3B gene inactivation [17,22]. However, the majority of previous research focused on methylation studies of the intronic CpG-island, that was incorrectly considered as located in the promoter region.
The objective of our study was to elucidate the distinct roles of SEMA3B in tumor suppression, particularly in apoptosis and angiogenesis. Furthermore we aimed to evaluate frequencies of promoter (hg38/chr3: 50,267,308-50,267,797) and intronic (hg38/chr3: 50,268,972-50,269,271) CpG-island hypermethylation correlations with SEMA3B expression, and tumor progression in lung and renal cancers.

Tissue samples
In total, 70 paired tumor/normal samples of NSCLC (29 ADC and 41 SCC) and 133 clear cell RCC (ccRCC) were obtained from the N.N. Blokhin Cancer Research Center, Russian Academy of Medical Sciences (Moscow, Russia). The set of 38 NSCLC (16 ADC and 22 SCC) and 83 ccRCC was used in the methylation studies and the expression or copy number studies by semi-quantitative RT-PCR. The additional set of 32 NSCLC (13 ADC and 19 SCC) and 50 ccRCC was used for validation by qPCR expression studies. The sample information is presented in Table 1 and S1 Table. The samples were collected in accordance with the guidelines issued by the Ethics Committee of N.N. Blokhin Cancer Research Center, Russian Academy of Medical Sciences (Moscow, Russia). All patients gave written informed consent (available upon request). The Ethics Committee of N.N. Blokhin Cancer Research Center, Russian Academy of Medical Sciences, specifically approved this study. The study was performed in accordance with the principles outlined in the Declaration of Helsinki. Tumor tissues and paired morphologically normal tissues were obtained from patients after surgical resection prior to radiation or chemotherapy and were stored in liquid nitrogen. The diagnosis was verified by histopathology, and only samples with 70-80% or more tumor cells were used in the study. "Normal" controls were obtained at a minimum of 2 cm from the tumor and were confirmed histologically as normal epithelial cells. Tumor specimens were characterized according to the  International System of Classification of Tumors, based on the tumor-node-metastasis (TNM)  and staging classification of the Union for International Cancer Control (UICC, version 2002) [26] and World Health Organization (WHO) criteria classification [27,28]. Blood samples from 15 healthy donors were also used in the study.

SCID mice
Twelve SCID mice were used in the experiments. The mice were obtained from Scanbur (Sollentuna, Sweden). Animal euthanasia was performed by CO 2 asphyxiation followed by cervical dislocation. Transfection and selection of stably transfected SEMA3B-U2020 cell clones The cDNA encoding the SEMA3B gene was cloned into an episomal tetracycline-regulated vector, pETE. The resulting plasmid was sequenced. To obtain stable cell clones expressing SEMA3B, U2020 cells (SCLC) were transfected with empty pETE or pETE/SEMA3B plasmid DNA (0.5 mg DNA per well) in 12-well plates using Lipofectamine and Plus Reagent (Invitrogen, CA, USA) according to the manufacturer's protocol. Transiently transfected pETE/SEMA3B-U2020 and pETE-U2020 cells were cultured for 2-3 weeks in IMDM medium containing Bsd (5 μg/ml) to select stable clones. The expression of SEMA3B was regulated by doxycycline.

Colony formation assay
Transiently transfected U2020 cells (with pETE and pETE/SEMA3B) were stripped 24-48 h after transfection and plated on 100 mm 2 cell culture dishes at a density of 500-1000 cells per plate. After selection with Bsd (5 μg/ml), Giemsa-stained colonies were photographed and counted using Quantity One software (version 4.4.0; Bio-Rad, Hercules, CA, USA). Cell viabilities were estimated by FACS analysis with propidium iodide (PI-FACS), following the manufacturer's guidelines (FACSCalibur, BD Bioscience).

Tumor growth in SCID mice
The tumorigenicity of pETE/SEMA3B-transfected U2020 and empty pETE-transfected U2020 cells (control) was assessed by subcutaneous injecting the cells into 6-8-week-old SCID mice as previously described [29]. The cells were collected by centrifugation at 800 rpm for 2 min and resuspended in serum-free IMDM medium at a concentration of 2-3×10 6 cells per 100 μl injection. The cells were embedded into a Matrigel matrix (BD Biosciences, Erembodegem, Belgium) according to the manufacturer's protocol. The mice were observed for tumor formation twice per week and the tumor size was measured using calipers. Then we used the multigene inactivation test (MGIT) for three genes (ZMYND10/BLU, TUSC2/FUS1 and SEMA3B) in SCID mice. MGIT is based on the monitoring of tumor suppressor candidate gene inactivation in cells and tumors. It was accomplished according to a published method [29,30]. Briefly, U2020 cells were transfected either with pETE/SEMA3B, pETE/ZMYND10, pETE/TUSC2 or empty pETE plasmids. Mixes of cell clones were subcutaneously injected in SCID mice. Subsequently, if a tumor is formed, the expression of SEMA3B, ZMYND10 and TUSC2 is evaluated. The knock-down of the genes suggests their importance for tumor suppression.
Immunohistochemical analysis of SCLC tumor cell line U2020 and tumor angiogenesis in SCID mice The plasmid pETE/SEMA3B was introduced into SCLC cell line U2020, which constitutively produced a tetracycline transactivator (tTA) [29]. The resulting sub-line was inoculated subcutaneously into SCID mice. The mice were given drinking water with doxycycline (+dox, gene is OFF) or without (−dox, gene is ON). The mice were euthanized after 1 month. The tumors or places of cell injections were excised and embedded with paraffin. The sections were 5 μm in thickness. Paraffin was dissolved in xylene (Sigma-Aldrich, St. Louis, MO, USA), and the tissue sections were treated sequentially with 99, 95, 75 and 30% ethanol. Epitopes were recovered by heating in a microwave oven for 5 min in citrate buffer. Anti-CD31 mouse antibody, together with rabbitanti-mouse FITC-conjugated secondary antibody (Dako, Karlstrup, Denmark), was used to stain microvessels. The TUNEL assay (In Situ Cell Death Detection Kit, Boehringer Mannheim, Germany) for the detection of apoptosis was performed according to the manufacturer's protocol.

DNA and total RNA isolation, reverse transcription-PCR
Nitrogen-frozen tissues were disrupted using a Mikro-Dismembrator (Sartorius, Germany). The DNA from human tissues and cell cultures was isolated by phenol extraction according to the standard protocols. Total RNA was isolated using the RNeasy mini kit as recommended by Qiagen (Netherlands). Purified RNA was quantified using NanoDrop-1000 (NanoDrop Technologies Inc., DE, USA). RNA quality was assessed using 28S and 18S rRNA bands after electrophoresis in a 1% denaturing agarose gel and analyzed using a Bioanalyzer 2100 (Agilent Technologies, CA, USA). The lack of DNA contamination was checked by semi-quantitative PCR with primers for the main histocompatibility complex I gene (MHCI, designed using Vector NTI, see S2 Table). All RNA samples were treated with RNase-free DNase I (Fermentas, Lithuania). RNA samples containing over 0.1% DNA were discarded. The cDNA was synthesized from 1 μg of total RNA using M-MuLV reverse transcriptase and random hexamers, according to the standard manufacturer's protocol (Fermentas, Lithuania).
Bisulfite sequencing of the promoter CpG-island of SEMA3B gene Bisulfite DNA conversion was conducted as described in [31,32] with the use of 1-2 μg DNA (lung and renal cell lines and tumor/normal tissues). The modified DNA was purified using a JETquick PCR Purification Spin Kit (Genomed, Sweden). Modified DNA was maintained at -20°C and used as a template for PCR with the designed primers (listed in S2 Table), whose product was sequenced. Amplification of the SEMA3B promoter CpG-island fragment was performed in a 50 μl reaction mixture containing PCR buffer (67 mM Tris-HCl pH 8.

Methylation specific PCR (MSP)
The bisulfite-treated DNA, dissolved in twice-distilled water, was also used as a template for MSP. The PCR conditions and primers for the methylated and unmethylated allele of intronic [17] and promoter (designed by DNASTAR Lasergene 2000 program) CpG-islands are given in S2 Table. In case of the promoter CpG-island, 6 CpG-dinucleotides were analyzed (2 by the forward primer and 4 by the reverse), and in case of the intronic-3 (1 by the forward primer and 2 by the reverse). PCR was performed on a DNA Engine Dyad Cycler amplifier (Bio-Rad, United States) using the following program: 95°C, 5 min; 35 cycles of 95°C, 10 s; T ann (see S2  Table), 20 s; 72°C, 30 s and 72°C, 3 min. The absence of PCR product on unconverted DNA was checked for each pair of primers. DNA of the human fibroblast cell line L-68 served as an unmethylated allele control; L-68 SssI DNA from L-68 fibroblasts treated with SssI methyltransferase (SibEnzyme, Russia) served as a positive control for 100% methylation.

Semi-quantitative RT-PCR
To control the reverse transcription, primers for the transcript of the beta2-microglobulin (B2M) gene were used [33]. Semi-quantitative RT-PCR was performed with equal quantities of cDNA using the primers and conditions listed in S2 Table. Multiplex PCR with primers to the SEMA3B [17] and B2M genes was performed under conditions optimized for the SEMA3B primers, and the concentration of B2M primers was 1.5 times lower. The products of the RT-PCR were separated on 2% agarose gels, and the band pattern was analyzed using GelImager software (DNA Technology, Russia). A semi-quantitative copy number assay for the markers of LUCA region was used as described elsewhere [14].

qPCR
Quantitative PCR was performed with the primers and TaqMan probes listed in S2 Table using a 7500 Real-Time PCR System (Applied Biosystems, CA, USA). Each reaction was repeated three times. The nucleotide sequences of the amplicons were verified by sequencing in a 3730 DNA Analyzer automated sequencer (Applied Biosystems, CA, USA). QPCR data were analyzed using the reference genes GAPDH, GUSB and RPN1 [34] and the relative quantification or ΔΔCt-method as described earlier [35]. At least 2-fold mRNA level changes were considered as significant because of mRNA level variability of the reference genes.

Statistical analysis
The nonparametric Wilcoxon test was used to compare mRNA expression differences in the target and reference genes in the ccRCC and NSCLC samples. Kruskal-Wallis and Mann-Whitney rank-sum tests, Fisher's exact test and χ 2 criteria were used for analysis of mRNA level and methylation status changes in ccRCC and NSCLC groups with different pathological and histological characteristics. Student's t-test was used to compare the data obtained for groups of replicates. P-values 0.05 were considered statistically significant. Spearman's rank correlation coefficient (r s ) was used for revealing correlations.

Results
In vitro growth suppression of SCLC cells U2020 by SEMA3B The small-cell lung carcinoma cell line U2020 was transfected with pETE/SEMA3B or pETE as the control. The transfected cells were cultured for 15 days. The growth rate of U2020 cells expressing SEMA3B was lower than the control (P < 0.01 since the day 5, see Fig 1A). The colony formation assay showed that the number of colonies of U2020 cells containing pETE/SEMA3B was lower after re-expression of the doxycycline-suppressed SEMA3B gene in comparison to the control cells (890 ± 60 vs 190 ± 40, P < 0.01, Fig 1B). Based on PI-FACS analysis, the abundance of apoptotic and necrotic cells expressing SEMA3B (without doxycycline) was increased significantly in comparison with SEMA3B-off cells (with doxycycline): from (7±2)×10 2 to (49 ±5)×10 2 , P < 0.01, see Fig 1C. Taken together, these data suggest that SEMA3B is the inhibitor of human SCLC cells growth via induction of apoptosis in vitro.

Multi-gene inactivation test in SCID mice
We have previously reported that the GIT technique allows for the efficient and controlled induction of various genes in cells [29,30]. For the MGIT experiment, we used the SCLC cell line U2020 for the conditional expression of three TSG-candidates located in 3p21.3: ZMYND10 (BLU), TUSC2 (FUS1) and SEMA3B. Mixes of cell clones carrying different genes were inoculated subcutaneously into six-week-old SCID mice using a Matrigel (basement membrane matrix) implantation technique in the absence of doxycycline (the genes were ON). A PCR-based comparison of SCID mice tumors and primary cell clones showed unequivocally that SEMA3B was selectively knocked-down in all three tumors grown in vivo (see samples 8-10 in Fig 2), whereas the expression of ZMYND10 and TUSC2 genes did not change under these conditions. These two genes most likely did not antagonize the tumor growth of U2020 cells in SCID mice (we leave this question open because the search for mutation through retained genes was not included in the MGIT). These data suggest that SEMA3B is a growth inhibitor of human SCLC cells in vivo. Effect of SEMA3B transgenes on tumor growth in SCID mice and angiogenesis The U2020 sub-line with conditional SEMA3B expression under doxycycline control was inoculated in SCID mice subcutaneously. Four mice received doxycycline in drinking water (+dox mice, control) and five were not administered doxycycline (-dox mice). The onset of solid tumors actively expressing SEMA3B was not observed in 4 out of 5 cases (Fig 3, red line). In the fifth-dox mouse active tumor growth (Fig 3, yellow line) began one week later compared to the control (Fig 3, blue line), in spite of the presence of SEMA3B in the construct. However, the  expression of the SEMA3B gene was not detected in this tumor according to the Northern blot analysis (data not shown) suggesting loss of SEMA3B in these cells. The tumors (observed in SEMA3B-OFF mice) had areas of active cell proliferation, in contrast to the tissues taken from sites of cell injections (in SEMA3B-ON mice), in which abundant fibrous and poorly differentiated cellular stroma and necrotic areas were observed. These data demonstrate the possibility of SEMA3B tumor suppression activity in SCID mice.
Tissues from two sites of U2020 cells inoculation were analyzed using CD31 staining and TUNEL assay: one SEMA3B-negative (Fig 4A and 4B) and one SEMA3B-positive (Fig 4C and  4D) mouse. Anti-CD31 mouse antibodies were used to stain blood microvessels. A very few number of microvessels was detected in SEMA3B-positive tissues. A fragment containing one of the microvessel-like objects is shown at Fig 4C. Microvessel signal (green channel) was co-localized with signal from erythrocytes (red channel) resulting to yellow colored areas in Fig 4A. However, SEMA3B-negative solid tumors demonstrated abundance of elongated, compressed blood ducts with fluorescence typical to epithelia. These tumor ducts were also co-localized with erythrocytes ( Fig 4A). This could support a role of SEMA3B as inhibitor of angiogenesis. However, further experiments on the extended sampling are needed to prove this suggestion.
We hypothesized that the SEMA3B-positive cells in the areas without blood vessels and erythrocytes were undergoing cell death. To test this hypothesis, sections of the same tissue fragments were analyzed by a TUNEL assay, a technique enabling detection of apoptotic cells. (Fig 4B, 4D). A huge area of apoptotic cells was observed in SEMA3B-positive tissue sample whereas no apoptotic area was seen in SEMA3B-negative tumors. In this case, we assumed that the expression of SEMA3B suppressed tumor growth in vivo, likely by the induction of apoptosis. Inhibition of angiogenesis could be suggested also.

Methylation of the promoter and intronic SEMA3B CpG-islands in lung and renal cell lines and primary tumors by bisulfite sequencing and methylation-specific PCR
The SEMA3B gene is comprised of 18 exons and contains two CpG-islands: one is located in the promoter region (1-st CpG-island, hg38/chr3: 50,267,308-50,267,797, 22 CpG-dinucleotides) and the other one in the first intron (2-nd CpG-island, hg38/chr3: 50,268,972-50,269,271, 12 CpG-dinucleotides). We analyzed the methylation profile of 16 CpG-dinucleotides of the promoter CpG-island in 5 lung cancer cell lines (3 SCLC and 2 NSCLC) and 12 NSCLC primary tumors using bisulfite sequencing (5 ADC and 7 SCC, see Fig 5A). Dense methylation (> 40% of the analyzed CpGs were methylated) of the promoter CpG-island of the SEMA3B gene was observed in 2 of 3 SCLC cell lines, but in none of the NSCLC cell lines (NCI-H157 and NCI-H647). However, in all 7 SCC primary tumors and in 2 of 5 ADC primary tumors, methylation was detected in 2-12 of the CpGs. In addition, we examined the methylation profile in 9 renal cancer cell lines and 25 ccRCC primary tumors (Fig 5B). Dense methylation of the promoter CpGisland was observed in 8 of 9 cell lines and 11 of 25 ccRCC primary tumors. Among matched histological normal tissues, methylated CpGs were detected only in 2 of 25 ccRCC cases and in none of 12 NSCLC cases (Fig 5A and 5B).
Next, we evaluated the methylation frequencies of both CpG-islands of the SEMA3B gene by the MSP method using a representative set of primary tumors. The methylation frequency of the promoter CpG-island was 44% (7/16) in ADC, 45% (10/22) in SCC and 52% (43/83) in ccRCC. The intronic CpG-island was methylated slightly less than the promoter island in both histological types of NSCLC (ADC −38%, 6/16; SCC −32%, 7/22) and in ccRCC (39%, 32/83; Table 2). The methylation frequencies of both islands were significantly higher in tumor tissues than in paired histologically normal tissues (P 0.04, see Table 2). Thus, the methylation of both CpG-islands of the SEMA3B gene is a hallmark of lung and renal cancer. As expected, the methylation of neither the 1-st nor the 2-nd SEMA3B CpG-island was detected in any DNA samples isolated from blood lymphocytes of healthy donors (n = 15). The MSP data were in agreement with the bisulfite sequencing results for each type of cancer investigated.
The use of a representative set of primary tumors allowed us to reveal possible correlations between the methylation frequency of the promoter and intronic CpG-islands with the pathological and histological parameters of the tumors. Advanced ccRCC and NSCLC tumors had a higher frequency of CpG-island methylation compared to the early stages (Tables 2 and 3). The strongest correlation was shown for SCC (Spearman's rank correlation coefficients were equal 0.37, P = 0.09, and 0.60, P < 0.01, for the 1-st and for the 2-nd CpG-island, respectively). A positive correlation was observed between tumor grade and the frequency of CpG-island methylation for NSCLC and ccRCC (Table 3). This correlation was stronger than the correlation between tumor stage and the frequency of CpG-island methylation for both histological types of NSCLC and almost equal to those for ccRCC ( Table 3). The strongest correlation was shown for SCC (0.49, P = 0.02, and 0.68, P < 0.01, for the 1-st and for the 2-nd CpG-island, respectively). These results suggest that methylation of both SEMA3B CpG-islands is a frequent event in renal and lung cell lines and primary tumors and contributes to the progression of tumors of these locations, especially in the lung.

Correlation between methylation status of two CpG-islands and SEMA3B gene expression level in ccRCC tumors
To investigate the consequences of CpG-island methylation, the mRNA level in 48 ccRCC samples was evaluated by semi-quantitative PCR. In ccRCC samples, the expression of SEMA3B was 5-1000 times lower than the matched normal samples in 24 of 48 cases (Fig 6A).
In 4 cases, a 5-300-fold up-regulation of expression was observed. No correlation between mRNA level changes and stage, grade or presence of metastases was observed.
Next, we compared the changes in mRNA levels of the SEMA3B gene in ccRCC tumors (semi-quantitative PCR data) with the methylation status of two CpG-islands (methylationspecific PCR data). Methylation of the promoter CpG-island was observed in 27 of 48 (56%) cases and methylation of the intronic CpG-island-in 19 of 48 (40%) (Fig 6B). As shown in Fig 6A and 6B, both islands were methylated primarily in the ccRCC samples with decreased  Table. The bold numbers of CpG-dinucleotides (3-4 and 9-12) indicate the location of the primers that were used for MSP method. Tumor Suppressor Function of the SEMA3B Gene in Lung and Renal Cancers SEMA3B expression. The Spearman's correlation coefficient between expression level downregulations and methylation status was equal to 0.50 (P < 0.01) for the 1-st CpG-island and 0.25 (P < 0.01) for the 2-nd CpG-island. These data suggest that inactivation of SEMA3B in carcinomas could be associated primarily with methylation of the promoter CpG-island.
Amplifications and deletions of the SEMA3B gene can contribute to mRNA level alterations in ccRCC A correlation between changes in SEMA3B mRNA expression and the methylation status of its promoter CpG-island was shown in ccRCC, but for several samples, significant aberrations in expression were not associated with methylation of either the 1-st or the 2-nd CpG-island. In one sample, a 100-fold up-regulation was observed when both CpG-islands were methylated. Consequently, we investigated an alternative mechanism of expression regulation, copy number alterations, by semi-quantitative PCR. The increase in SEMA3B mRNA level in ccRCC samples #9 and #36 (300-and 100-fold up-regulation, respectively) was associated with amplification of Tumor Suppressor Function of the SEMA3B Gene in Lung and Renal Cancers the 5'Sema5 marker [14] (Fig 6A and 6C). A five-fold increase in mRNA level in ccRCC sample #83 was associated with amplification of the D3S1573 marker (data not shown). Decreased SEMA3B expression (100-1000 times) in ccRCC samples #8, #14, #15, #44 and #71 was associated with hemi-or homozygous deletions of the 5'Sema5 marker (Fig 6A and 6C). A hemizygous deletion was observed in sample #34 with 5-fold down-regulation and no methylation. The Spearman's correlation coefficient between expression level alterations and copy number changes was equal to 0.61 (P < 0.01) Thus, in addition to the epigenetic modifications, genetic events (e.g., deletion or amplification of the SEMA3B gene locus) can also contribute to the alteration of SEMA3B gene expression in tumors. However, in samples #39 and #61, with 4-5-fold down-regulation, and sample #17, with a 10-fold up-regulation, no copy number changes were observed. We propose that other mechanisms may be responsible for these mRNA level alterations; for example, regulation via miRNA.

Quantitative evaluation of SEMA3B expression in lung and renal tumors
To confirm the mRNA level differences, we performed qPCR analysis on an additional set of lung and renal tumors (Table 1). We found a noticeable (up to 300 times) and frequent (94%, 30/32) down-regulation of SEMA3B gene expression in NSCLC primary tumors using qPCR (Fig 7A). The overall frequency and extent of the decrease in SEMA3B mRNA level was similar in the tumors of the two main histological types of NSCLC; ADC and SCC. An average 10-fold decrease and 92% (12/13) was observed in ADC, and a 14-fold and 95% (18/19) was observed in SCC (Table 4). Other differences were also observed. In ADC with lymph node metastases compared to ADC without metastases, the average mRNA level was considerably decreased (19-fold decrease vs. 3-fold, P < 0.05), but in SCC, significant expression down-regulation was observed in the early stages of tumor development. In ccRCC samples we also observed frequent (84%, 42/50) SEMA3B gene expression downregulation (4-fold on average, Fig 7B). However, these values were not as high as for NSCLC. No correlation was observed between mRNA level changes and stage (Table 4) or the presence of metastases (Fig 7B).
Based on these results, we conclude that significant and frequent decreases in SEMA3B mRNA level are in agreement with the semi-quantitative PCR and methylation data and confirm that SEMA3B down-regulation is a frequent event both in NSCLC and ccRCC. However, the frequency of down-regulation was higher than the frequency of CpG-island methylation (Tables 2 and 4). Thus, other mechanisms of inactivation are also likely contributors: for example, deletions in renal, lung and breast cancers, according to our results and the data of others [14,15,17]. . QPCR data, additional samplings. Light grey columns-samples without metastases, dark grey columns-samples with lymph node or distant metastases. The numbers of primary tumors correspond to those in S1 Table. Mean values ± standard deviations for 3 replicates are represented.

Discussion
The SEMA3B gene is located in the LUCA region (3p21.3) that harbors 19 genes, the majority of which are involved in carcinogenesis [16,[36][37][38]. Some of these genes, e.g., RASSF1, NPRL2, and SEMA3B, suppress the growth of tumor cells in vitro and in vivo [36][37][38]. However, SEMA3B may have both tumor-suppressive and pro-invasive properties. It has been reported that SEMA3B is dramatically down-regulated in the cell line H460-M, derived from the large cell lung tumor cell line NIH-H460, which induces spontaneous metastasis in nude mice [39]. On the other hand, it is highly expressed in many invasive and metastatic human cancers; for example, colorectal carcinoma, neuroblastoma, melanoma and acute myeloid leukemia [20,40]. SEMA3B induces apoptosis and the production of interleukin 8 by tumor cells by initiating the p38-mitogen-activated protein kinase pathway. In turn, the release of interleukin 8 induces the recruitment of tumor-associated macrophages, which may lead to metastatic dissemination of the tumor [20,41]. In this study, we confirmed the tumor-suppressive activity of SEMA3B both in vitro (U2020 SCLC cells) and in vivo (SCID mice models, by both tumor growth assay and multi-gene inactivation tests). Our studies in vitro are in agreement with previous research in NSCLC, breast and ovarian ADC cell lines [19,22,42]. However, in this study, the suppressive effect of SEMA3B in vivo has been shown for the first time.
Semaphorins 3 encode proteins with axonal guidance properties that generally affect the motility and migration of tumor and endothelial cells by inducing a collapse of the actin cytoskeleton via binding to neuropilins/plexins complexes [43,44]. Earlier, the class-3 semaphorins were implicated in tumor progression and metastasis [39,45,46]. However, tumor suppression properties of distinct semaphorins differ across various cancer types. For example SEMA3D and SEMA3E displayed strong inhibitory effect on the glioblastoma cells whereas SEMA3A and SEMA3B expression just caused persistent cell shape contraction [47]. On the other hand, SEMA3B was previously demonstrated to inhibit proliferation of breast and lung cancer cell lines in vitro [20,22]. Some studies suggest that the mechanism of the tumor-and angio-suppressive properties of SEMA3 proteins is competition between SEMA3 and VEGF for binding to neuropilin receptors [19,48]. Other studies reveal SEMA3 growth-inhibitory properties independent of competition with VEGF [10]. Several studies indicate that SEMA3A and SEMA3F are potent inhibitors of metastasis and angiogenesis [10,49,50]. Semaphorins 3 were suggested as potential angiogenesis inhibitory agents for triple negative breast cancer treatment  [53]. In this study, using PI-FACS we have shown that tumor growth suppression by SEMA3B is associated with induction of apoptosis in vitro. In addition, we have suggested that SEMA3B could be able to induce apoptosis and, possibly, inhibit angiogenesis in vivo, which was previously demonstrated only in vitro [20,22]. One of the hallmarks of tumor-suppressor genes is the down-regulation of their expression in various tumors at the mRNA or protein level. Inactivation of suppressor genes in tumors can be caused by methylation of their promoter regions. Hypermethylation often reduces the mRNA levels not only of SEMA3B but also of some other genes of the LUCA region in tumors; e.g., RASSF1A, BLU, and CACNA2D2 [36][37][38]. SEMA3B methylation has been observed in various types of cancer, including lung, liver, gallbladder, gastric, breast and oral carcinomas, and neuroblastoma [17,[54][55][56][57][58][59][60][61]. However, these studies included only the CpG-island, which, according to the NCBI database (http://www.ncbi.nlm.nih.gov/), belongs to the first intron (+1350..+1700 bp from SEMA3B 5'-end) of SEMA3B. However, there is a true promoter region, which includes CpG-dinucleotides with higher density (-300..+50 bp from SEMA3B 5'end) and is capable of binding transcription factors, such as Lyf−1, DeltaEF−1, Tcf−11, c−Myb, SP1, C/EBP, and AhR/Arnt (according to the recent ENCODE data, http://genome.ucsc.edu/ ENCODE/).
Here, for the first time, methylation of both intronic and promoter CpG islands has been analyzed in 14 lung and renal cancer cell lines and representative sets of primary tumors (38 NSCLC and 83 ccRCC cases). We observed frequent methylation of intronic and promoter CpG-islands in both lung (SCC and ADC) and renal primary tumors. In addition, the methylation frequency detected in the promoter CpG-island was higher than in the intronic island in SCC, ADC, and ccRCC (44-52% vs. 32-39%). Thus, methylation of both promoter and intronic SEMA3B CpG-islands is a feature of lung and renal tumors.
In addition, an association was observed between the decrease of the SEMA3B mRNA level and the methylation of the promoter and intronic CpG-islands in ccRCC primary tumors (the Spearman's correlation coefficient was equal to 0.50 (P < 0.01) for the promoter CpG-island and 0.25 (P < 0.01) for the intronic island). These data imply a primary contribution of promoter CpG-island methylation to down-regulation of the SEMA3B suppressor gene.
The frequency of SEMA3B promoter hypermethylation (44-52%) is approximately two times lower than the frequency of expression decreases (84-95% cases, according to semiquantitative PCR data) in NSCLC and ccRCC. Thus, other mechanisms could lead to SEMA3B down-regulation. For example, we have found hemi-or homozygous deletions of the 5'Sema5 marker in some ccRCC samples. There are several reports, including ours, of chromosome 3 allelic losses resulting in decreased SEMA3B mRNA levels in lung and renal tumors [14,55]. The impact of non-coding RNA, especially miRNA, should not be ignored; e.g., miR-137 or miR-193a, as predicted by miRWalk (http://www.ma.uni-heidelberg.de/apps/zmf/mirwalk/).
The use of a representative set of primary tumors allowed us to reveal the correlations between the methylation frequency of both the promoter and the intronic CpG-islands of SEMA3B with tumor stage and grade for histological types of NSCLC (SCC and ADC) and for ccRCC. The highest Spearman's rank correlation coefficients between mentioned features were observed for SCC (r s was localized over the range 0.37-0.68).
SEMA3B expression was evaluated by quantitative PCR using an additional set of NSCLC (SCC and ADC) and ccRCC samples. We observed frequent and significant SEMA3B down-regulation in SCC and ADC, and somewhat less in ccRCC. SEMA3B down-regulation has been shown in lung, liver, breast, ovarian, renal and colon primary tumors using different semiquantitative methods [2,62]. Down-regulation of SEMA3B gene was negatively correlated with tumor size and gastric cancer staging [61]. Here, a large decrease of SEMA3B mRNA level in primary NSCLC and ccRCC was shown for the first time using qPCR. We also observed an association of the frequency and extent of the decrease in SEMA3B mRNA level with the development of ADC metastases (P < 0.05). In SCC, we revealed 10-fold and above down-regulation of the expression of this gene already at early stages of tumor development.
The methylation analysis revealed a significant association of SEMA3B hypermethylation with tumor progression in terms of tumor stage and grade for both subtypes of NSCLC (SCC and ADC) and ccRCC. Moreover, our qPCR expression studies have established a strong reverse correlation between the SEMA3B mRNA level and the presence of metastases in lung ADC. Therefore, hypermethylation of SEMA3B CpG-islands could be suggested as a progression marker for ccRCC and NSCLC (especially SCC), and for the SEMA3B mRNA level as a marker of metastasis development in lung ADC.
Other members of class 3 semaphorins also reveal distinct features, such as expression level variations or tumor suppressive ability, depending on the tumor type or the model system. Mouse melanoma (B16F10) cells overexpressing Sema3A show a significant inhibition of cell motility, invasiveness and proliferation, as well as suppression of tumor growth in vivo, and angiogenesis and metastasis in mice models [50]. Sema3A/B/C/E are also involved in the lymph node metastasis of prostate cancer, but they are likely to modulate the behavior of prostate cancer with a pro-tumor or anti-tumor effect, depending on the subtype [63].
In conclusion, our in vitro and in vivo results reveal the tumor suppressor role of SEMA3B, which could acts by inducing apoptosis or, possibly, inhibiting angiogenesis. Our data also show that methylation of both the promoter and intronic CpG-islands of SEMA3B is a frequent event in ccRCC and two major histological types of NSCLC, but correlation of SEMA3B down-regulation and hypermethylation is stronger for the promoter CpG-island. Significant decreases in the level of SEMA3B mRNA in the majority of the lung and renal tumor samples were revealed by qPCR. Association of SEMA3B hypermethylation and down-regulation with tumor progression could serve as prognostic markers. An understanding of the role of SEMA3s, including SEMA3B, in carcinogenesis will help to use their tumor suppressive and anti-angiogenic properties for the development of new agents for cancer therapy in future.
Supporting Information S1