Patients

Cell lines

Human SCLC cell lines (H1688 and H446), a human squamous cell line (SK-MES-1), and a human normal fibroblast epithelial cell line (HFL-1) were purchased from Shanghai Cell Library of Chinese Academy of Science. Human adenocarcinoma cell lines (A549, H292 and H1299) and a human normal bronchial epithelial cell line (HBE) are stored in our lab.

Immunohistochemistry

Immunohistochemistry (IHC) was performed according to our previous report [22, 23]. The dilutions of antibodies were 1:50 for E2F1 (Merk Millipore, USA), MMP-7, MMP-9, MMP-16 (Abgent, China), MMP-2, Sp1, p65 (Santa Cruz Biotechnology, USA) and VEGFR (Cell Signaling Technology, USA). The staining samples were scored by two pathologists without any knowledge of the clinical pathological outcomes. Staining intensity was divided into four grades: 0 as negative; 1 as weak intensity (less than 10% positive); 2 as moderate intensity (more than 10% and less than 60% positive); and 3 as strong intensity (more than 60% positive). Grade 0 was considered as negative expression, and grades 1, 2, and 3 were considered as positive staining.

siRNA transfection

Real time PCR

Western blotting

Cells were lysed in RIPA lysis buffer. A total of 40 μg protein was separated by SDS-PAGE and samples were electrophoretically transferred onto nitrocellulose membranes. The membranes were blocked with 5% fat-free dry milk and incubated with primary antibodies against E2F1 (1:100, Merk Millipore), Sp1, p65 (1:100, Santa Cruz Biotechnology), MMP-3, −7, −9, −14, −15 and −16 (1:200, Abgent), Vascular endothelial growth factor receptor (VEGFR, 1:1000, Cell Signaling Technology), and Glyceraldehyde 3-phosphate dehydrogenase (GAPDH, 1:2000, Santa Cruz Biotechnology) at 4°C overnight. The membrane was washed and incubated with HRP-conjugated secondary antibodies for 45 min. The immunoblot bands were detected by an ECL system, and membranes were exposed to X-ray films [22].

Wound healing analysis and transwell experiments

The wound healing experiment was performed according to a previous report [9]. The cells were scratched by a 10 μl pipette tip and photographed by microscopy at 0, 12, and 24 h. The transwell experiment was conducted according to the manufacturer’s instruction (BD Company). A total of 60 μl of matrigel was placed into the upper chamber and plates were incubated for 3 h at 37°C. After the matrigel solidified, 1 × 10⁴ cells were plated into the upper chamber with media containing 1% fetal bovine serum. Media containing 10% fetal bovine serum was placed into the lower well. After 72 h, the matrigel was cleaned and the cells were stained by Gimsa Dye. The cells that invaded through the chamber were quantified by counting three fields.

ChIP-to-sequence

Chromatin immunoprecipitation (ChIP) was conducted according to the manual supplied by Merck Millipore Company (ChIP Assay kit, Cat. No.: 17–295). Cells (5 × 10⁷) were prepared and cross-linked by 1% final concentration of formaldehyde at 37°C for 10 min. Cells were centrifuged at 2,000 rpm for 4 min at 4°C, and then collected and incubated in SDS Lysis Buffer on ice for 10 min. The genomic DNA was sheared with Sonicate (36% strengthen, 25 sec and 30 cycle) and the average length of the fragments generated was 200 bp. Protein A agarose beads were added to the samples for 30 min at 4°C with agitation. Next, E2F1 antibody (4 μg) or equal amount of normal mouse IgG was added into the samples, and samples were incubated at 4°C with rotation overnight. The agarose beads were collected by gentle centrifugation (800 rpm) for 5 min and washed five times. Reverse cross-linking was performed with high salt solution (5 M NaCl) and the DNA fragments were obtained. The cyclin D1 primer was used as a positive control in real-time PCR. The DNA fragments were sequenced by BGI Company (http://www.genomics.cn). The Hiseq2000 50SE sequencing platform was used and the data analysis algorithm included SOAP2.20 comparison and MACS peak calling. Clean data was obtained by filtering the low quality data according to a certain criteria: the sequences not containing adapter, N less than 10%, quality values less than 20 and ratio less than 50%. For the peak value, the filtering was conducted according to the p value obtained by MACS analysis. The data was discarded when the p value was higher than 1^e-5, which ensured the fidelity of the data and exclusion of false positives.

Construction of the MMP-9, MMP-16, Sp1 and p65 luciferase reporter constructs

Genomic DNA was extracted from H1688 cells, and MMP-9, MMP-16, Sp1 and p65 were amplified by PCR using primer sequences shown in Additional file 1: Table S1. The PCR DNA fragments were extracted by a Gel Extraction kit (Invitrogen, USA). The PCR fragments and pGL3-basic luciferase reporter vector (Promega, USA) were digested with FastDigest SacI, NheI or XhoI (Thermo, USA), extracted and ligated with T4 DNA Ligase (TakaRa, Japan) to generate the four luciferase reporter constructs. The binding site mutants were constructed by overlap PCR and nested PCR, and the primers were listed in Additional file 1: Table S1. The constructs were confirmed through sequencing by BioSune Company.

Transient transfections and luciferase assays

Cells were transiently transfected with 0.5 μg of luciferase reporters and 0.3 μg of E2F1, Sp1, or p65 expression vector with Lipofectamine 2000 (Invitrogen). Cotransfection with 0.02 μg of the pRL-TK Renilla reniformis luciferase served as a normalizing control. Luciferase assays were performed using the Dual Luciferase Assay System (Promega).

Statistical analysis

SPSS 17.0 was used as the statistical software. The immunohistochemistry samples were treated with Chi Square test. The association and statistical difference between E2F1 lower, moderate, and higher and clinicopathological variables was analyzed by Spearman’s analysis and χ² test. Univariate survival rate was analyzed by the Kaplan-Meier method, and the significant were tested by Log-Rank test. Multivariate survival analysis was performed by using Cox’s regression. The expression differences among target genes were analyzed using paired t test. P < 0.05 was considered to be statistically significant.

E2F1 was highly expressed in SCLC

Although expression of E2F1 had been detected in lung cancer tissue [21, 26–30], its expression was inconsistent among different populations, especially in NSCLC. Therefore, we firstly examined E2F1 levels in human lung cancer tissues in a Chinese Han population. E2F1 expression was positive in 95.56% (86/90) of SCLC, 50% (5/10) of large lung cancer cell (LCLC), and 10% (2/20) of adenocarcinoma samples compared with the normal alveolar sections. However, it was not detected in squamous tissues (0/20). The normal bronchial epithelial tissues with exclusive E2F1 expression served as positive controls (Figure 1A) [31]. In 90 SCLC samples, the numbers of negative, weak, moderate, and strong positive E2F1 staining cases were 4, 11, 23, and 52, respectively. In adenocarcinoma samples, only two weak positive staining cases were found. In LCLC samples, two weak and three strong positive staining cases were found (Table 4, Additional file 2: Figure S1).

Consistent with these observations, E2F1 was positively expressed in H1688 and H446 cell lines as well as HBE cells, which served as the positive control. However, weak expressions were detected in A549, H1299 and H292 cell lines compared with SCLC cells. In addition, E2F1 was not detected in SK-MES-1 and HFL-1 cell lines (Figure 1B). Therefore, E2F1 expression was predominantly elevated in SCLC tissues and cell lines, suggesting the importance of E2F1 in SCLC development and progression.

E2F1 was an independent and adverse prognostic factor for SCLC patients

Patient survival time was collected by follow-up and data showed that the median survival period of patients displaying lower E2F1 (including negative staining) was 15.67 months, and the moderate E2F1 and higher E2F1 expression groups were 13.74, and 10.21 months, respectively. These results suggested that high level of E2F1 was correlated with poor survival in SCLC. Moreover, univariate survival analysis revealed that E2F1 (P < 0.01, Figure 2A) and clinical stage (P < 0.01 Figure 2B) were prognosis factors in SCLC patients, while other factors including gender (P = 0.768), age (P = 0.818), smoking (P = 0.827), tumor size (P = 0.411) were not significant. Multivariate analysis provided additional evidence that higher E2F1 expression proved to be an independent and adverse prognosis factor in SCLC (HR = 0.461, 95% CI: 0.230–0.925, P = 0.029) (Additional file 3: Table S2).

Depletion of E2F1 inhibited cell migration and invasion in SCLC cells

Clinical data analysis showed that highly expressed E2F1 was associated with clinical stage and was an independent and adverse prognosis factor. Thus, we next examined whether E2F1 knockdown led to suppression of SCLC cell migration. As shown in Figure 3A, transfection of siRNA targeting E2F1 significantly abolished E2F1 expression in both H1688 and H446 cells. Serum-induced invasion through matrigel-coated transwell filters was significantly reduced in cells depleted for E2F1. Cells transfected with scrambled siRNA displayed similar migration compared with that of the untreated control cells (Figure 3B). Consistent with the transwell results, wound healing assays showed that E2F1 knockdown significantly blocked H1688 and H446 cell migration into the wound areas compared with cells transfected with scrambled siRNA or untreated control cells (Figure 3C). To further verify the ability of E2F1 to promote invasion and metastasis, A549 cells (with lower E2F1 expression compared with H1688 and H446 cells, Figure 1B) were transfected with an E2F1 expression vector and assayed as described above. The results showed that enforced expression of E2F1 could promote A549 cell invasion and metastasis (Additional file 4: Figure S2). These results suggested the importance of E2F1 in cell invasion and migration.

E2F1 knockdown significantly inhibited the expression of MMP-9 and −16 in SCLC

Because E2F1 was closely associated with invasion and metastasis (Figure 3), we next determined whether E2F1 affected invasion and metastasis by regulating the expression of MMPs. Firstly, we examined the expression of MMP-2, −7, −9 and −16 in 90 SCLC samples by IHC staining. As shown in Figure 4A, expression of MMP-7 was detected in 73.33% (66/90) of the specimens, MMP-9 in 86.67% (78/90) of the cases, and MMP-16 in all samples (90/90). However, MMP-2 was not shown in any of the SCLC samples. Because VEGFR expression was previously identified in SCLC [17, 32], we included VEGFR as a positive control. As expected, VEGFR positive staining in SCLC samples was observed in 95.56% (86/90) of the cases. These results suggested that MMP-9 and MMP-16 might play an important role in the process of invasion and metastasis of SCLC.

Next, we determined whether expressions of MMPs were affected by E2F1 in SCLC cells. Real time PCR and western blotting results showed that expression levels of MMP-3, −7, −14, and −15 were only slightly reduced when E2F1 was depleted in both H1688 and H446 cells, but the expression of MMP-9 and −16 were significantly decreased upon E2F1 depletion (Figure 4B). Based on the observations that MMP-9 and −16 were expressed at higher levels in SCLC tumors (Figure 4A), it suggested that E2F1 may be involved in the invasive potential of SCLC by regulating the expression of MMP-9 and −16.

E2F1 controlled MMP-16 expression via E2F1 binding sites in SCLC cells

Since E2F1 could regulate expressions of MMP-9, −14 and −15 in NSCLC [9], together with the observations mentioned above, we performed ChIP-to-sequence to identify E2F1 target genes in H1688 cell line. As summarized in Additional file 5: Table S3 and Additional file 6: Table S4, many MMP genes including MMP-1, −14, −16, −17, −19, −24, and −25 were found to be regulated by E2F1. MMP-16 was selected for further study due to its higher expression in SCLC tumor. IHC results revealed that E2F1 was strongly positive in SCLC tumor where MMP-16 was highly expressed (Figure 5A), indicating that E2F1 was associated with the expression of MMP-16.

We used MatInspector analysis to identify two putative E2F1 binding sites in the MMP-16 promoter. H1688, H446, and A549 cells (with lower E2F1, Figure 1B and Figure 5B) were transfected with luciferase constructs driven by the wild-type MMP-16 promoter or the MMP-16 promoter with mutated E2F1 binding sites (Additional file 7: Figure S3A) and an E2F1 expression plasmid. As shown in Figure 5C, overexpression of E2F1 increased the activity of the MMP-16 promoter. Furthermore, E2F1 could still activate MMP-16 promoter containing the mutated E2F1 binding site (mutant 1), but not mutant 2, indicating that E2F1 could enhance the expression of MMP-16 and that the sequence (ggtgGGCGggaagaaag, binding site 2) was required for E2F1-mediated stimulation of MMP-16 promoter activity. These results indicated that E2F1 stimulated the expression of MMP-16 by binding the binding site 2 sequence in the MMP-16 promoter.

Sp1 and p65 regulated MMP-9 expression in SCLC cells

Previous studies [8] and our IHC results (Figure 4A) showed that MMP-9 expression was higher in SCLC and was significantly affected by E2F1 knockdown (Figure 4B). Johnson et al. reported that one E2F1 binding site (tcagggaggGAAAaaga) was predicted in the MMP-9 promoter (Additional file 7: Figure S3B) [9]. Then, we first tested the activity of MMP-9 promoter containing the E2F1 binding site mutant by a luciferase reporter experiment, and found that this site was not functional (Figure 6A). However, our results showed that MMP-9 promoter activity was significantly enhanced when E2F1 was co-expressed in H1688, H446, and A549 cells (Figure 6A), indicating that E2F1 regulated MMP-9 expression via an indirect pathway.

Previous studies suggested that Sp1 and/or p65 might be involved in the regulation of MMP transcription [33–35]. MatInspector analysis identified four putative Sp1 and two putative NF-kappa B binding motifs in the MMP-9 promoter (Additional file 7: Figure S3B). Activation of the wild-type MMP-9 promoter was significantly increased when the cells were cotransfected with a Sp1 expression plasmid. The activity of the MMP-9 promoter Sp1 binding site mutant 1 and 2 constructs was unchanged, but the MMP-9 promoter Sp1 binding site mutant 3 and 4 constructs showed significantly decreased activity. Together this data indicated that Sp1 upregulated the expression of MMP-9 by binding to sites 3 and 4, but not 1 or 2 in the MMP-9 promoter (Figure 6B).

Additionally, overexpression of p65 induced activity of the wild-type MMP-9 promoter, but not the promoter reporter with a mutated NF-kappa B binding site, indicating that p65 also plays a role in regulating MMP-9 expression (Figure 6C). Real time PCR and western blotting results validated that MMP-9 expression was significantly decreased after transfection with siRNA specifically targeting Sp1 and p65 (Figure 6D), indicating that Sp1 and/or p65 could simulate the expression of MMP-9 in SCLC cells.

E2F1 modulated Sp1 and p65 expressions in SCLC cells

The results described above suggested that Sp1 and p65 could regulate the expression of MMP-9 (Figure 6), and our ChIP-seq data showed that E2F1 was recruited to the sequences of Sp1 and p65 (Additional file 5: Table S3 and Additional file 6: Table S4). Therefore, we speculated that E2F1 regulated the expression of MMP-9 mediated by Sp1 or p65. We next explored the correlation among MMP-9, Sp1, p65, and E2F1. In addition to high levels of E2F1 (95.96%) and MMP-9 (86.67%) in SCLC, IHC staining in 90 SCLC samples revealed positive Sp1 and p65 expression in 93.33% (84/90) and 98.89% (89/90) of the cases, respectively (Figure 7A). The expression of p65 was consistent with a previous report [36], but ours is the first report revealing Sp1 expression in SCLC. More importantly, E2F1, Sp1 and p65 were highly expressed in SCLC samples where MMP-9 staining was also strong, indicating a positive correlation among E2F1, Sp1, p65 and MMP-9 in SCLC tissues (Figure 7A). This observation supported the notion that E2F1 upregulated MMP-9 expression, mainly via Sp1 and/or NF-kappa B.

We next want to examine whether E2F1 contributes to the overexpressions of Sp1 and p65 in SCLC. Depletion of E2F1 led to significantly reduced Sp1 and p65 expression in two SCLC cell lines (Figure 7B). VEGFR, which was transcriptionally regulated by E2F1 [17], was used as a positive control. Enforced expression of E2F1 in H1688, H446 and A549 cells led to a significant induction of the luciferase reporter driven by the wild-type Sp1 promoter compared with controls (Figure 7C). Furthermore, E2F1 overexpression could also stimulate activities of Sp1 mutated 1 and mutated 2 promoters. However, the activity of Sp1 mutant 3 promoter (Additional file 7: Figure S3C) was dramatically reduced, suggesting that the actgcGCGCcgaatgcc motif in Sp1 promoter was functional and essential for E2F1-mediated induction (Figure 7C). E2F1 also notably induced p65 promoter activity, while mutation of each E2F1 binding motif in the p65 promoter (Additional file 7: Figure S3D) resulted in decreased luciferase activity, even with E2F1 expression (Figure 7D). Together these results showed that E2F1 regulated Sp1 and p65 expressions at the transcriptional level, which subsequently led to enhancement of target gene expression, such as MMP-9.