Background

Cancer metastasis is a complex process consisting of many critical steps, such as migration, invasion, adhesion, and metastatic colonization [1]. Gain of migrating and invading abilities are the most important steps during the development of metastasis. Cells experiencing these alterations undergo profound morphological changes, collectively referred to as the epithelial–mesenchymal transition (EMT) process. EMT is considered to be a critical mechanism in regulating cancer invasion and metastasis [1, 2]. It can be triggered by oncogenic activation or microenvironmental stimuli, such as hypoxia. Indeed, the hypoxia of the tumor microenvironment has been shown to be closely associated with metastasis [3]. Under hypoxic conditions, hypoxia-induced factor-1α (HIF-1α) becomes stabilized and up-regulates a number of EMT-related transcription factors (e.g., TWIST and SNAIL) that promote tumor metastasis [4, 5].

Cancer stem cells are a small population of cancer cells holding stemness properties known as cancer stemness (CS), which possess the ability to self-renew and contribute to unlimited cancer proliferation, tumor aggressiveness, drug treatment resistance, and metastasis [6, 7]. Cancer stem cells have been demonstrated to be regulated by several pluripotent transcription factors, such as OCT4, SOX2, and NANOG [8, 4g) in the SLUG promoter. We then conducted the quantitative ChIP-qPCR assay to test whether the FOXM1-STIL complex could bind to the SLUG promoter. Our result showed that both of STIL and FOXM1 bound to the SLUG promoter (Fig. 4g). Notably, the association of both STIL (Fig. 4h) and FOXM1 (Additional file 2: Fig. S6g) with the SLUG promoter was reduced in sh-FOXM1-treated cells. Reciprocally, the reduced SLUG promoter-binding activity of FOXM1 was observed in STIL-depleted CL1-5 cells, suggesting that STIL enhances the transcriptional activity of FOXM1 by increasing its promoter binding affinity (Additional file 2: Fig. S6h). Since FOXM1 directly interacts with the SLUG promoter [103] and STIL does not harbor any known DNA-binding domains, we hypothesize that STIL acts as a transcriptional coactivator of FOXM1 to promote SLUG expression.

Given that the CS core transcription factors (e.g., SOX2, OCT4/POU5F1, and NANOG) were reported to be directly regulated by FOXM1 [106], we tested whether STIL regulates the expression of these CS core genes in a FOXM1-dependent manner. As expected, overexpression of STIL significantly increased the mRNA expression levels of NANOG, SOX2, and POU5F1; however, these STIL-induced upregulations were suppressed upon FOXM1 depletion (Fig. 4i). Further studies demonstrated that STIL depletion affects the expression of these FOXM1-driven genes: overexpression of FOXM1 significantly increased these CS gene promoter-driven luciferase activities (e.g. SLUG, NANOG, SOX2 and POU5F1) in a dose-dependent manner, while knockdown of endogenous STIL reduced FOXM1-induced gene activation (Additional file 2: Fig. S6i). Together, our findings suggest that STIL associates with FOXM1 to enhance the FOXM1-modulated CS.

FOXM1 has also been reported to play an important role in cell cycle regulation that controls the expression of many genes required for G1/S and G2/M transition [107, 108]. To investigate whether STIL might modulate FOXM1-driven cell cycle related genes [109], the gene expression levels of the regulators for G1/S transition, such as CCND1, SKP2 (S-Phase Kinase Associated Protein 2) and CDC25A (Cell Division Cycle 25A), the components of G2/M phase progression including CCNB1 (Cyclin B1), CCNB2 (Cyclin B2), CDK1 (Cyclin Dependent Kinase 1) and PLK1 (Polo Like Kinase 1), and the activators of mitotic entry, such as AURKA (Aurora kinase A), AURKB (Aurora kinase B), CEBPB (CCAAT/enhancer-binding protein beta) and BUBR1/BUB1B (Budding uninhibited by benzimidazoles 1 homolog beta), were examined. Two CL1-5-based FOXM1 knockdown clones and their corresponding control cells overexpressing Flag vector or Flag-STIL were generated (Additional file 2: Fig. S7a). As expected, FOXM1 deletion resulted in the decreased mRNA expression including SKP2, CDC25A, CCNB1, CCNB2, CDK1, PLK1, AURKA, AURKB and BUBR1 (Additional file 2: Fig. S7b). Overexpression of STIL significantly increased the expression of the above genes; however, these STIL-induced upregulations were inhibited upon FOXM1 depletion. In contrast, FOXM1 knockdown or overexpression of STIL did not affect CCND1 and CEBPB mRNA (Additional file 2: Fig. S7c). The reason is not clear. Collectively, our findings suggest that the association of STIL with FOXM1 could upregulate some FOXM1-modulated genes involved in cell cycle regulation, such as SKP2, CDC25A, CCNB1, CCNB2, CDK1, PLK1, AURKA, AURKB, and BUBR1, but not CCND1 and CEBPB.

The interaction between FOXM1 and STIL is required for the STIL-FOXM1 axis-mediated tumorigenic abilities

To validate the importance of the association of STIL with FOXM1 in STIL-mediated tumorigenic functions, we have mapped the FOXM1-interacting region of STIL. Our co-IP results showed that HA-FOXM1 forms complexes with full-length wild type STIL (a.a. 1–1288) and STIL-M (a.a. 420–780), but not STIL-N (a.a. 1–628) or STIL-C (a.a. 780–1288), implying that the region of STIL between a.a. 628 to 780 is responsible for FOXM1-binding (Fig. 5a). Because the coiled-coil domain (a.a 726–748) of STIL [110] was reported to be located within this region, we examined whether the deleted coiled-coil domain mutant of STIL (ΔCC) impairs its FOXM1-binding ability. As shown in Fig. 5b, the full-length STIL and two STIL truncated mutants (a.a. 1–1061 and a.a. 437–1288) containing the coiled-coil domain are coimmunoprecipitated with FOXM1, while the STIL mutant missing the coiled-coli domain (ΔCC) does not. We next examined the tumorigenic effects of the STIL mutant (ΔCC) on migration/invasion and SLUG gene expression. Our results showed that the STIL mutant (ΔCC) impaired its abilities to enhance cellular migration/invasion (Fig. 5c) and SLUG gene activation (Fig. 5d), suggesting that the interaction of FOXM1 with STIL is important for STIL-mediated tumorigenic abilities. Together, our data support a model wherein STIL acts as a coactivator that complexes with FOXM1 to upregulate the FOXM1-mediated downstream genes involved in metastasis, CS, and cell cycle.

Fig. 5
figure 5

The interaction of FOXM1 with STIL is required for STIL-induced tumorigenic abilities. a HEK293T cells were transiently co-transfected HA-FOXM1 with full-length or various truncated Flag-STIL mutants including STIL-N (a.a. 1–628), STIL-M (a.a. 420–780) and STIL-C (a.a. 780–1288) as indicated. Protein complexes were immunoprecipitated using anti-Flag antibody and analyzed by Western blotting using indicated antibodies. b HEK293T cells were transiently co-transfected HA-FOXM1 with the full length or various truncated GFP-STIL mutants (a.a. 1–1061, a.a. 437–1288, and the deleted coiled-coil domain (ΔCC)). Protein complexes were immunoprecipitated using anti-HA antibody and analyzed by Western blotting. c, d NCI-H1299 cells were transiently transfected with GFP-STIL-WT or GFP-STIL mutant (ΔCC), and the expressed proteins were analyzed by Western blotting (c, upper panel). Cell migration and invasion abilities were analyzed by Boyden chamber assay (c, lower panel), while the SLUG promoter activity was measured by reporter assay (d, upper panel) and the SLUG mRNA level (d, lower panel) was measured by qPCR method. Data information: Statistical data in c, d represent the mean ± SD (n = 3 independent experiments). Significance is determined by t-test (NS, not significant; **p < 0.01; ***p < 0.001)

Full size image

STIL expression is induced by HIF1α under hypoxia

We next explored why STIL is up-regulated in lung cancer. We first examined the STIL DNA copy number using single-nucleotide polymorphism (SNP) array data (Additional file 2: Fig. S8a) and the DNA methylation status of STIL using 450K methylation array data (Additional file 2: Fig. S8b). No significant difference was observed among normal lungs, lung cancer cell lines, and lung cancer specimens in either datasets.

Hypoxia is an important micro-environmental characteristic that activates EMT-TFs and the HIF-mediated pathway during tumor metastasis [111]. The hypoxia pathway was also noted in our GSEA analysis of the STIL-regulated transcriptome (Fig. 3a), and four HIF1α DNA-binding sites ([A/G]CGTG) were identified in the STIL promoter (Fig. 6d). Interestingly, we found that an increase HIF1α level was accompanied by the elevated STIL expression (Fig. 2a) in CL1-0, CL1-3 and CL1-5 cells under the normoxic condition (Additional file 2: Fig. S8c). We thus investigated whether STIL is induced under the hypoxic condition. Figure 6a shows that the protein and mRNA levels of STIL were up-regulated in CL1-0, CL1-5, and NCI-H1229 cells under hypoxia, but that HIF1α depletion dramatically diminished the ability of hypoxia to induce STIL at the protein (Fig. 6a, upper panel) and mRNA (Fig. 6a, lower panel) levels in all three lung cancer lines. HIF1α, which is the major transcription factor in the cellular response to low oxygen, is easily degraded in normoxia. We thus examined whether overexpression of HIF1α (ΔODD), an HIF1α mutant that lacks the oxygen-degradation domain could induce STIL expression under normoxia. As shown in Fig. 6b, we observed increases of STIL at both the protein and mRNA levels in cells overexpressing HIF1α (ΔODD) under the normoxic condition. Consistent with this finding, HIF1α (ΔODD) overexpression could activate STIL promoter-driven luciferase activity in a dose-dependent manner under normoxia (Fig. 6c). Given that a small portion of STIL can be detected in nucleus (Fig. 3e) and STIL is upregulated by HIF1α (Fig. 6a), we further examined the effect of hypoxia on STIL nuclear localization. As shown in Additional file 2: Fig. S8d, the endogenous nuclear STIL was slightly increased under the hypoxic condition compared with that of normoxia (20% versus 11%). The ICC result further showed that HIF1α became stable and accumulated in the nucleus under hypoxia, and led to the increased FOXM1 in the nucleus (Additional file 2: Fig. S8e). Furthermore, hypoxia seems to partially promote the nuclear localization of GFP-STIL (Additional file 2: Fig. S8e), which may reflect with an increase of nuclear GFP-STIL protein detected by Western blotting under hypoxia (36% versus 29%) (Additional file 2: Fig. S8f). Together, our results indicate that HIF1α is an upstream factor that regulates STIL expression.

Fig. 6
figure 6

STIL expression is induced by HIF1α under hypoxia. a STIL protein (upper panel) and mRNA levels (lower panel) were analyzed by Western blotting and qPCR method, respectively, in the indicated cells under normoxic (20% O2) or hypoxic (1% O2) conditions, or in HIF1α-knockdown cells under hypoxia. b STIL protein (upper panel) and mRNA levels (lower panel) were analyzed in cells transiently transfected with the indicated constructs under normoxia. c STIL promoter activity was measured by reporter assay in HEK293T cells transiently transfected with the pGL3-STIL promoter-luciferase plasmid and the indicated constructs under the normoxic condition. d STIL promoter activity was measured by reporter assay in CL1-0 cells transiently transfected with pGL3-STIL promoter-driven luciferase constructs encoding wild-type HIF1α, HIF1α DNA-binding site mutants, and/or HA-HIF1α (ΔODD) under hypoxia. e The binding of HIF1α to the STIL promoter was analyzed by the ChIP-qPCR assay in CL1-5 cells under hypoxia. VEGF promoter was served as the positive control for HIF1α binding, and the region of 6.0 kb upstream of STIL transcriptional start site within STIL promoter was used as the unrelated control. Methylation of histone H3 (H3-K9Me3) on SAT2 gene was used as a positive control and IgG as a negative control for the ChIP-qPCR assay. f Clinical correlations between STIL, HIF1α, and SLUG in 200 lung cancer specimens were analyzed by IHC and the data were assessed using Pearson’s correlation. Scale bar: 50 μm. g Overall survival of 1927 lung cancer patients stratified by STIL and SLUG expression was determined by Kaplan–Meier analysis. The median survival of the four groups is shown. Significance is determined by the log-rank test (p < 0.001). Data information: Statistical data in a–e represent the mean ± SD (n = 3 independent experiments). Significance is determined by t-test (NS, not significant; **p < 0.01; ***p < 0.001)

Full size image

Since we identified four HIF1α consensus DNA-binding sites (Fig. 6d) in the STIL promoter (site1: nts − 195 to − 199; site2: − 894 to − 898; site3: − 1054 to − 1058; and site4: − 1235 to − 1239), we then generated mutations in these four sites ([A/G]CGTG mutated to [A/G]CTGT) and examined which site is responsible for HIF1α binding. Our results showed that the mutation within nts − 195 to − 199 (site1) dramatically inhibited HIF1α-induced luciferase activation in cells overexpressing HIF1α (ΔODD) under hypoxia (Fig. 6d). A similar effect was also observed in HIF1α (ΔODD)-overexpressing cells under normoxia (Additional file 2: Fig. S8g). To examine whether HIF1α directly binds to the STIL promoter, we performed ChIP-qPCR assay in CL1-5 cells that were pre-treated with hypoxia (to stabilize the HIF1α protein level). Our result showed that HIF1α binds to the STIL promoter under the hypoxic condition (Fig. 6e). Collectively, our findings suggest that HIF1α directly binds to the STIL promoter at the site1 region (nts − 195 to − 199) to drive STIL expression under hypoxia.

We next performed IHC analysis to assess the clinical relevance of STIL in relation to HIF1α and SLUG. Our results showed that the HIF1α intensity was positively correlated with the expression of STIL (Pearson’s co-efficient, R = 0.54) in lung cancer specimens (Fig. 6f). STIL expression was similarly associated with SLUG in the same patient cohort (R = 0.55, Fig. 6f). Intriguingly, the concordant expression of STIL with SLUG displayed an especially high correlation in patients with metastatic lymph nodes (R = 0.76, Additional file 2: Fig. S8h), suggesting that there is a strong association of the HIF1α-STIL-SLUG axis in lung cancer specimens. We thus evaluated the clinical application of the STIL-SLUG axis to predict survival among lung cancer patients. We found that patients with STILhigh and SLUGhigh were associated with poor patient survival (60.0 months) (Fig. 6g), whereas STILlow and SLUGlow patients exhibited prolonged survival (92.6 months). This suggests that the STIL-SLUG axis could be a useful prognostic marker for the survival rate of lung cancer patients.

Discussion

Tumorigenesis is a complex and dynamic process consisting of three major stages: initiation, progression, and metastasis. In the present studies, we identify a novel STIL-mediated mechanism that promotes tumor progression and metastasis. Our collective in vitro and in vivo results on cell proliferation, colony formation, and xenograft tumor assay (Additional file 2: Fig. S2) support a role of STIL in tumor progression. Furthermore, we demonstrated that STIL is associated with FOXM1 (Fig. 4f), and that this association promotes tumor metastasis by activating FOXM1-regulated downstream genes (e.g., SLUG, NANOG, SOX2, POU5F1, SKP2, CDC25A, CCNB1, CCNB2, CDK1, PLK1, AURKA, AURKB and BUBR1) that are involved in the EMT, CS, and cell cycle (Figs. 3b, 4i, and Additional file 2: Fig. S7b). Importantly, we demonstrate that hypoxia is a new factor contributing to STIL upregulation in cancers. HIF1-α directly binds the STIL promoter under hypoxia (Fig. 6e), consequently potentiating hypoxia-induced tumor metastasis. A model showing how STIL contributes to tumor development via the FOXM1-mediated transcriptional activation under hypoxia is shown in Fig. 7.

Fig. 7
figure 7

Schematic diagram illustrates the molecular mechanism how STIL contributes to tumor development via the FOXM1-mediated transcriptional activation of target genes under hypoxia. HIF1α becomes stable under hypoxia in cancer, and translocates to the nucleus to upregulate STIL expression. Furthermore, HIF1α was reported to drive FOXM1 expression [10, 11]. Thus, STIL can associate with FOXM1 to modulate the expression of downstream target genes involved in the EMT process and stemness, and consequently promotes tumor development, metastasis, and poor prognosis

Full size image

Centrosome abnormalities are commonly observed in human cancers and are correlated with aneuploidy and poor patient prognosis. Previous studies used mouse models to focus on PLK4, which is a key regulator of centrosome duplication [112, 113]. However, the studies in mice with high expression of PLK4 provided contradictory results on the contribution of centrosome amplification to tumor progression. For example, centrosome amplification in neural progenitor cells resulted in microcephaly but did not promote tumorigenesis [114]. Furthermore, Kulukian et al. [115] and Vitre et al. [116] reported that overexpression of PLK4 in the skin epidermis induced an increase in centrosome number but failed to initiate or promote tumorigenesis in skin. In contrast, Sercin et al. [39] showed that PLK4 overexpression accelerates skin tumor formation in mice lacking P53 and Levine et al. [37] demonstrated that supernumerary centrosomes are sufficient to drive tumorigenesis in multiple tissues of mice. Thus, the issue of whether direct associations exist between centrosome abnormalities and cancers remains unclear.

In this study, we examined the T (tumor)/N (non-malignant) ratio of 14 centriolar/centrosomal genes in lung cancer patients and their paired adjacent non-malignant lung tissues. STIL showed the highest T/N ratio (3.5) among the studied genes (Table 1); this was found in patients with lung cancer, and was even higher than the T/N ratio (1.5) of PLK4 in these patients. Further analysis also showed that the STIL mRNA level was significantly increased in many other types of cancers (Additional file 1: Table S1a) and its high expression is associated with poor prognosis in patients with many cancer types (Additional file 1: Table S1b). STIL was previously reported to be present in the cytosol, specifically in the centriole [43], and act as a master regulator of PLK4 in initiating centriole duplication [45,46,47,48,49]. Here, we reveal an unexpected novel role of STIL in the nucleus. In addition to centriolar STIL, a subset of STIL translocate into the nucleus and function as the coactivator to enhance the downstream FOXM1-drievn genes via the association between STIL and FOXM1, and consequently contributes to the metastasis. This finding may explain the elevated STIL expression in lung cancer patients with metastatic lymph nodes or brain metastases.

A remaining open question is: Do the extra centrosomes induced by excess STIL promote tumor initiation and drive spontaneous tumorigenesis? The answer is not yet clear. Using si-SASS6 to block excess STIL-induced centriole amplification, we found that the migration and invasion abilities were significantly reduced in the si-SASS6-treated cells harboring excess STIL (Additional file 2: Fig. S4a). However, the si-SASS6 treatment does not completely block the migration/invasion abilities of STIL overexpressing cells (Additional file 2: Fig. S4a). These findings suggest that in addition to the excess STIL-mediated cancer cell migration and invasion, the possibility of supernumerary centrosome aberration-triggered tumorigenesis (e.g. aneuploidy and/or tissue architecture disruption) [20] can’t be ruled out. Future experiments are needed to clarify this discrepancy.

Finally, it has been proposed that the small increases in centrosome number induced by a low-to-moderate level of PLK4 are permissive for tumor development, whereas high levels of PLK4 trigger larger number of centrosomes and are likely to be harmful for long-term cell survival [37]. Since STIL is a master regulator of PLK4, we speculate that a low-to-moderate level of STIL could promote tumor initiation, as seen for PLK4, via a yet-unknown mechanism. Future experiments by generating a transgenic mouse line with low to moderate expression level of STIL could be a way to test the role of STIL in the initial stage of tumorigenesis.

In summary, we herein show that STIL is significantly up-regulated in lung and many other types of cancers, and that its expression level is highly correlated with patient survival, implicating its potential application in cancer detection and as a prognostic marker. Importantly, we demonstrate that STIL plays a versatile role in multistage tumorigenesis through the HIF1α-STIL-FOXM1 axis, and therefore may serve as a promising target for cancer therapy.

Conclusion

Our findings indicate that the centriolar protein STIL functions not only as a key regulator in centriole duplication but also as a transcriptional coactivator that regulates EMT and stemness to promote tumor metastasis. Our findings show that a subset of STIL enter the nucleus, which interact with FOXM1 to activate its downstream target genes in metastasis. Furthermore, we provide evidence to show that HIF1α directly binds to STIL promoter and drives STIL gene expression under hypoxia. Thus, STIL can serve as a potential diagnostic marker for early lung cancer detection, and a promising therapeutic target for lung cancer treatment.