Slug belongs to the Snail superfamily of transcriptional repressors, which contains five zinc finger domains near its
C-terminus that recognize E-box elements, and a SNAG transrepressor domain at its
N-terminus that is responsible for regulating the transcriptional repression activity [
1]. Recent evidences suggest that Slug participates in epithelial-mesenchymal transition (EMT) during embryonic development and cancer metastasis by suppressing the expressions of its downstream target genes like E-cadherin, occludin, claudin-1, and integrin α3 [
8]. Previously, we showed that p53 induced ubiquitin-mediated proteasome degradation of Slug via p53–Mdm2–Slug complex formation and mutation of p53 might induce Slug accumulation by repressing E-cadherin expression and resultant poor clinical outcomes in NSCLC [
9]. All these findings suggest the crucial role of Slug modification in cancer progression.
In general, SUMOylation can facilitate protein stabilization, localization, and gene transcription [
10]. Small ubiquitin-like modifier (SUMO) family, consists of SUMO-1, 2 and − 3, are initially expressed as inactive precursors, which mediates to expose a
C-terminal double-glycine motif by SUMO-specific proteases [
12]. Then, activated SUMOs covalently attach to the ε-amino group of a lysine within the consensus sequence ψKXE via specific conjugation to a SUMO-activating enzyme heterodimer (E1), the unique SUMO-conjugating enzyme 9 (E2; Ubc9), and a SUMO protein ligase (E3) [
14]. Recent evidences indicated that non-consensus SUMO acceptor sites have also been found [
16]. Usually, E3 enzyme is not required for the addition of SUMO to target proteins in mammals, although the existence of specific E3 ligases can promote the conjugation of SUMO from E2 to its target protein [
17]. Known SUMO E3 enzymes, PIASs and Pc2 participate additionally in gene transcription [
19]. Besides, SUMOylation is a dynamic process, where deSUMOylation is catalyzed by sentrin-specific proteases (SENPs) [
20]. Recent reports show that SUMOylation plays important roles in tumorigenesis [
22]; however, the regulation mechanism remains as a mystery.
Recently, we utilized yeast two-hybrid system to identify potential Slug-interacting proteins; and found that components of the SUMOylation process were present in the pools isolated using Slug as prey. Herein, we detail dissect the regulation mechanism and the crucial role of SUMOylation in controlling Slug-mediated transcriptional repression; and also demonstrate that hypoxia could regulate Slug SUMOylation by reducing its deSUMOylation and result in cancer malignancy in NSCLC.
Cell line and culture conditions
The human embryonic kidney cell lines, HEK293 (ATCC® CRL-1573™) and HEK293T (ATCC® CRL-3216™) were purchased from American Type Culture Collection (ATCC, Manassas, VA). The human lung adenocarcinoma cell lines CL1–2 and CL1–5 are two sublines, selected from the CL1–0 cells via transwell invasion assays, with progressive invasiveness in a similar genotypic background [
23]. The NCI-60 NSCLC cell line, Hop62, was a kind gift from Dr. Ker-Chau Li (Institute of Statistical Sciences, Academia Sinica). The cells were cultured in DMEM or RPMI 1640 medium supplemented with 10% fetal bovine serum (all from Invitrogen, Eugene, OR), 1% penicillin and streptomycin (all form Sigma−Aldrich, St. Louis, MO) in a humidified atmosphere containing 5% CO
2 at 37 °C.
Human full-length Slug (GeneBank™ accession number NM_003068.4), SUMO-1 (NM_003352.4), SUMO-2 (NM_006937.3), SUMO-3 (NM_006936.2), PIAS1 (NM_016166.1), PIAS2 (NM_173206.2), PIAS3 (NM_006099.3), PIASy (NM_015897.2), and Pc2 (NM_003655.2) were PCR amplified and cloned into pBMT116, pCIneo-HA3, p3xFlag-CMV-7.1–2 and pFlag-CMV-2, pEGFP (Clontech, Mountain View, CA) and pcDNA3.1-HA vectors. Slug mutations were generated via PCR-directed mutagenesis according to the manufacturer’s instructions (QuickChange kit; Stratagene). The human active form and mutant construct of SUMO-1, Flag–Ubc9, wild-type and dominant negative mutant (C/S) SENP1/2 were kindly provided by Dr. Hsiu-Ming Shih (Institute of Biomedical Sciences, Academia Sinica), and the pcDNA3-HDAC1–Flag plasmid was kindly provided by Dr. Wen-Ming Yang (Institute of Molecular Biology, National Chung Hsing University).
Yeast two-hybrid assay
Yeast two-hybrid screening was performed as described previously [
24]. The DNA fragment encoding full-length human Slug was sub-cloned into the pBTM116 vector to produce LexA-Slug as bait. And it was used to screen against with human prostate cancer cDNA library. Approximately 10
5 transformants were selected and PCR amplified to check as the potential interacting candidates. Then, LexA-Slug/ MST3 and Gal-Ubc9/ SUMO-1 constructs were co-transformed into L40 yeast. After overnight incubation, the resulting yeast cells were cultured in medium lacking tryptophan and leucine for selection of diploid cells. Diploid cells were further transferred to plate lacking tryptophan, leucine, and histidine and X-gal for 5 days to confirm the protein-protein interactions.
In vitro pull-down assay
The GST–Ubc9 were expressed and purified as described previously [
25]. In vitro transcription and translation were performed using the TNT SP6/T7 Quick Coupled Transcription/Translation System (Promega, San Luis Obispo, CA) according to the manufacturer’s instructions. GST fusion proteins expressed by bacteria were coupled to beads, and incubated with in vitro-translated proteins. After washing with buffer D (20 mM HEPES, pH 8.0, 20% v/v glycerol, 0.2 mM EDTA, 100 mM KCl, freshly added 0.5 mM PMSF and 0.5 mM DTT), the bound proteins were analyzed via electrophoresis on SDS–PAGE.
Immunoprecipitation and immunoblotting assays
Immunoprecipitation and immunoblotting were performed as described previously [
26]. The cells were lysed on ice for 5–10 min in RIPA lysis buffer (0.5% sodium deoxycholate, 0.1% sodium dodecyl sulfate, and 1% Nonidet P-40 in phosphate-buffered saline, all from Sigma) containing a 25-fold dilution of a stock protease inhibitor solution (Roche Diagnostics, Basel, Switzerland). The cell lysates were passed several times through a 21-gauge needle and clarified via centrifugation at 12,000 rpm for 30 min at 4 °C. The supernatants were taken as total cell lysates and precipitated with specific antibodies and protein A Sepharose. Then, the precipitated proteins were separated via SDS–PAGE and transferred to polyvinylidene difluoride (PVDF) membranes (Millipore, Billerica, MA) for immunoblotting with anti-Flag M2, anti-GFP, anti-β-actin (Sigma), anti-HA (Covance Research Products), anti-SUMO-1, anti-Ubc9, anti-Slug (A7, G18), anti-CtBP1, anti-Myc, anti-Lamin B, anti-Sin3A, anti-HDAC1, anti-HDAC2, anti-ubiquitin, anti-CD71 (Santa Cruz). anti-His (Qiagen, Hilden, Germany), anti-HIF1α, anti-GST (BD Transduction Laboratories), anti-acetylated lysine (Cell Signaling Technology), anti-HSP90 (Enzo Life Sciences, Farmingdale, New York) primary antibodies, followed with appropriate secondary antibodies conjugated with horseradish peroxidase and detected signals by Chemiluminescent Substrates (PerkinElmer).
In vitro SUMOylation assays
The in vitro SUMOylation was performed in a 20-μl reaction mixture containing 2 mM ATP, 20 mM HEPES (pH 7.5), 5 mM MgCl
2, 150 ng of an E1 SUMO ligase (Aos1/Uba2), 500 ng of Ubc9, 1 μg of SUMO-1 (all from LAE Biotechnology Co. Ltd., Taipei, Taiwan), and 2.5 μg of recombinant Nus–His–Slug protein at 30 °C for 2 h. For detecting SUMOylation in cells, the transfected cells were washed with PBS containing 20 mM
N-ethylmaleimide (NEM, Sigma) and then lysed on ice with RIPA lysis buffer containing NEM and protease inhibitors. After incubation, the lysates were sonicated and centrifuged at 12,000 rpm at 4 °C for 30 min, and subsequently analyzed by immunoprecipitation and immunoblotting.
Prediction of 3D protein structure
The amino acid sequences of human Slug (NP_003059.1) and PIASy (NP_056981.2) were obtained from the NCBI database. The protein structures were predicted by homology modeling using Discovery Studio 2.5 software.
Reporter gene assays
The Slug binding region containing three tandem repeats of the Snail-binding site (SBS; 5’-AGC TTA GCA GGT GCA CGA TAT CAG CAG GTG CAC CAT ATG AGC AGG TGC AA-3′) and E-cadherin promoter sequences are described as previous [
27]. Cells cultured in 6-well plates were transfected using Lipofectamine 2000 according to the manufacturer’s protocol. Thirty-six hours after transfection, cell extracts were prepared using reporter lysis buffer, and luciferase activity was assessed using the dual luciferase reporter assay system (Promega Corp, Madison, WI) and a luminometer according to the manufacturer’s instructions. A control reporter expressing Renilla luciferase was used for normalization of the transfection efficiency.
Electrophoretic mobility shift assays (EMSAs)
EMSAs were performed as previously described [
28]. E-box C (5′-CGT CGG AAC TGC AAA GCA CCT GTG AGC TTG CGG AAG TC-3′) oligonucleotide probes were labeled with [γ-
32P] ATP using T
4 polynucleotide kinase. Binding reactions of in vitro-translated wild-type or mutant Slug protein and the labeled probes were performed according to the manufacturer’s instructions (Promega).
Chromatin immunoprecipitation (ChIP)
The assays were performed using a Magna ChIP™ A/G Chromatin Immunoprecipitation Kit (Millipore) according to the manufacturer’s instructions. Briefly, equal numbers of cells were treated with 1% formaldehyde and then quenched with 0.125 M glycine for protein–DNA cross-linking. After washing with cold PBS, the cells were scraped, and soluble chromatin lysates were extracted via sonication and centrifugation. Two percent of the diluted chromatin solution was reserved as the total input sample. The diluted chromatin solution was incubated with anti-HDAC1 and normal mouse IgG antibodies overnight at 4 °C with rotation. Then, the DNA/protein solution was eluted with proteinase K containing elution buffer at 65 °C for 2 h to break the formaldehyde cross-links. DNA solution was used as the template for 33 cycles of PCR amplification using E-cadherin gene-specific primers.
RNA was isolated from cells using TRIzol reagent (Invitrogen) according to the manufacturer’s instructions. Total RNA was reverse transcribed for 60 min at 50 °C using Super Script III Reverse Transcriptase (Invitrogen) and Random Hexamer primers (Applied Biosystems, Foster City, CA) in the presence of an RNase inhibitor. And the gene expressions were analyzed by PCR using the following primer sequences: Slug forward, 5′-CAT GCC ATT GAA GCT GAA AAG-3′, and reverse, 5′-GCA GTG AGG GCA AGA AAA AG-3′; Gβ-like forward, 5-GTA TGG AAC CTG GCT AAC TG-3′, and reverse, 5′-TAC TGA TAA CTT CTT GCT TC-3′; E-cadherin forward, 5′-GCT GGA GAT TAA TCC GGA CA-3′, and reverse, 5′-ACC CAC CTC TAA GGC CAT CT-3′; claudin 1 forward, 5′-CCG TTG GCA TGA AGT GTA TG-3′, and reverse, 5′-GTT TTG GAT AGG GCC TTG GT-3′; occludin forward, 5′-GAT GAG CTG GAG GAG GAC TG-3′, and reverse, 5′-GCT CAC AGA GGT TTG GCT TC-3′; and integrin α3 forward, 5′-GCC TGC CAA GCT AAT GAG AC-3′, and reverse, 5′-ATC TCC GTG GGA TAC AGC AG-3′.
Modified Boyden chamber invasion assay
Modified Boyden chambers with polycarbonate-membrane inserts (pore size, 8 μm; Falcon; Becton Dickinson, Franklin Lakes, NJ) coated with 12 μg of Matrigel (BD; San Jose, CA) were used for cell invasion assays. Stable transfectants were suspended in culture medium containing 10% NuSerum (Invitrogen). Cells (2.5 × 10
4) were placed in the upper chambers, and 1 ml of medium was placed in the lower chambers. After incubation for 24 h at 37 °C, the cells were fixed with methanol for 10 min at room temperature and then stained for 30 mins at room temperature with a 50 μg/ml solution of propidium iodide (PI) (Sigma). The number of cells on each membrane was counted under a microscope at a magnification of 50× using the Analytical Imaging Station software package (Imaging Research Inc., St. Catharines, ON, Canada).
Experimental metastasis assay in vivo
A single-cell suspension containing 10
6 cells in 0.1 ml of PBS was injected into the lateral tail vein of 6-week-old SCID mice (
n = 9 per group). Seventy-nine days after injection, the mice were sacrificed under carbon dioxide anesthesia, and their lungs were removed and fixed in 10% formalin. The lung tumor cell nodules were counted under a dissecting microscope. Embedded tissues were sliced into 4-μm-thick sections, and the sections were stained with hematoxylin and eosin for histological analysis. All mouse experiments were performed and approved by the Laboratory Animal Center, National Taiwan University College of Medicine.
Lung tumor tissue specimens were obtained from 104 and 85 patients who underwent surgical resection for histologically confirmed NSCLC at the Taichung Veterans General Hospital (TVGH, Taichung, Taiwan) from 2000 to 2004 and National Taiwan University Hospital (NTUH, Taipei, Taiwan) from 1996 to 2005, respectively. None of the patients had received preoperative adjuvant chemotherapy or radiation therapy; and the post-surgical pathologic stage of each tumor was classified according to the international TNM classification system [
29]. All the investigations were approved by the Institutional Review Boards of TVGH and NTUH.
The Slug and Ubc9 transcript levels were determined via Q-PCR using an ABI Prism 7900 sequence detection system (Applied Biosystems) according to the manufacturer’s instructions. The designed primers and probe sets for Slug (Hs00161904_m1) and Ubc9 (Hs00163336_m1) were purchased from Applied Biosystems; and DEAD box helicase 5 (DDX5; Hs00189323_m1) was used as the internal control. The cut-off value of 0.6 was used to divide patients into high- and low-expression groups.
Four-μm-thick paraffin-embedded tumor tissue sections were deparaffinized in Trilogy Solution (Cell Marque Corp., Rocklin, CA) at 121 °C for 10 min. Samples were then treated with 3% H
2-methanol and subsequently incubated with DakoCytomation Dual Endogenous Enzyme Block (DakoCytomation Inc., Carpinteria, CA) for 10 min, Ultra V Block (LAB VISION Corporation, Fremont, CA) for 10 min, antibody dilution buffer (Ventana Medical Systems Inc., Tucson, Arizona) for 10 min, and the Slug (Abgent, San Diego, CA) or Ubc9 antibodies overnight at 4 °C. Immunoreactivity was detected using the BioGenex Super Sensitive Link-Label IHC Detection System (BioGenex, San Ramon, CA) according to the manufacturer’s instructions.
The quantitative in vitro and in vivo data were analyzed using Student’s
t-tests. All statistical tests were 2-sided, and
P values of less than 0.05 were considered statistically significant. In addition, the curves obtained from the cell migration assay and from analyses of clinical patient samples were examined via one-way ANOVA and the log rank test.
Identification of Ubc9 and SUMO-1 as Slug-interacting proteins
As Slug is a key EMT regulator [
30], we used yeast two-hybrid system to identify specific Slug-interacting proteins and to elucidate how Slug functions in EMT progression. The results indicated that Ubc9 and SUMO-1, two SUMOylation pathway members, were present in the pools. Next, the interactions of Ubc9 and SUMO1 with Slug were confirmed by β-galactosidase activity on solid medium; serine/threonine-protein kinase 24 (MST3) was used as the negative control (Fig.
1a). Then, Slug-Ubc9 interaction was further examined by exogenous immunoprecipitation assay in HEK293T cells (Fig.
1b and c) and endogenous immunoprecipitation assay with cross-linking for Slug protein in CL1–5 cells (Fig.
1d). Therefore, we finally identified that Ubc9 and SUMO-1 are two Slug-interacting partners in human lung cancer cells.
Amino acids 130–212 of Slug are crucial for its interaction with Ubc9
To determine the region of Slug that interacts with Ubc9, we generated several deletion mutants of Slug for glutathione
S-transferase (GST) pull-down assays. Slug fragments consisting of amino acids 1–268 (full-length), 33–268, 130–268, or 130–212, but not fragments 1–129 and 213–268, were pulled down by GST–Ubc9 (Fig.
1e). Thus, amino acids 130–212 of Slug are crucial for interacting with Ubc9.
SUMO-1 modifies slug in vitro Because Slug interacted with both Ubc9 and SUMO-1, we explored whether Slug could be SUMOylated by SUMO-1. An in vitro SUMOylation assay was performed using purified Slug proteins as the substrate together with various combinations of E1, E2, and a mature form SUMO-1 (SUMO-1-GG). Immunoblotting revealed a shifted band only in the presence of E1, E2, and SUMO-1 (Fig.
2a). Subsequently, we examined whether Slug protein could be SUMOylated in cells; in this experiment, HEK293T cells were transfected with 3xFlag-tagged Slug and GFP-tagged SUMO-1 vectors. High-molecular-weight SUMOylated Slug was detected in cells co-expressing Slug and SUMO-1 but not in cells expressing Slug alone (Fig.
2b). To validate that the shifted band was SUMO-1 conjugated Slug, we utilized SUMO-1-AA, which cannot bind to substrates because it lacks a
C-terminal double-glycine motif, and the SUMO isopeptidase inhibitor
N-ethylmaleimide (NEM) to determine this event. The shifted band observed in samples of SUMO-1-GG and Slug was more obvious than that of SUMO-1-AA and Slug or SUMO-1-GG and Slug without NEM treatment (Fig.
2b). Moreover, endogenous Slug SUMOylation was also detected in CL1–5 cells (Fig.
2c, d). Therefore, Slug is conjugated with SUMO-1 in cells
As SUMO1–3 are three key members in SUMO family, whether SUMO-2 and SUMO-3 could like SUMO-1 to modify Slug protein should be further clarified. To address this issue, we transfected different SUMO isoform plasmids into cells and found that all SUMO isoforms can modify Slug with SUMO-1 showed predominantly Slug SUMOylation rather than SUMO-2 and SUMO-3 (Additional file
1: Figure S1).
Slug SUMOylation primarily occurs within its 213–268 amino acid region
To explore the target regions of Slug responsible for its SUMOylation, we generated several Slug truncated mutants and examined their SUMOylation patterns in cells. HEK293T cells were cotransfected with different Slug truncated mutants and the GFP–SUMO-1 plasmids. The data showed that the shifted band was detected in cells expressing Slug fragments 1–268 and 130–268 (Fig.
2e); this result implied that SUMOylation primarily occurs in Slug
C-terminal region. Further dissection the
C-terminus of Slug identified that fragments 213–268 exhibited high SUMOylation ability (Fig.
2f), suggesting the critical region for Slug SUMOylation.
Sequence analysis indicated that five lysine residues located within Slug fragment 213–268 (K239, K240, K244, K248, and K258), especially amino acids 257–260 (HKHE), matched well with the SUMOylation consensus sequence (ψKXE). Therefore, we established Slug5M, in which lysine 239, 240, 244, 248, and 258 residues were substituted with arginine, and determine its SUMOylation level. The results showed the intensity of the shifted band for Slug5M was decreased approximately 75% than that for the wild-type Slug (Slug5M/wild-type Slug = 27.48 ± 13.87%, Fig.
2g). In order to find out all the SUMOylated sites of Slug, we further mutated all lysine residues (Slug22M) and individually substituted them with lysine residue. The results showed that in addition to the 5 sites we identified, lysine 188 was also important for Slug SUMOylation (Additional file
2: Figure S2a). As such, we further created two Slug mutants, Slug6M (K188, 239, 240, 244, 248, and 258R) and Slug8M (K188, 192, 211, 239, 240, 244, 248, and 258R), to examine their effects on Slug SUMOylation levels, repression activities and their DNA binding abilities. Like our expected, the SUMOylation levels and transcriptional repression activities of Slug were significantly decreased with the increasing number of lysine mutations on Slug protein (Additional file
2: Figure S2b and 2c); but the DNA binding abilities of Slug6M and Slug8M were also simultaneously attenuated (Additional file
2: Figure S2d). As such, Slug5M is more suitable than the other two mutants for studying the functional events of Slug SUMOylation.
The E2 Ubc9 can promote Slug SUMOylation
Since Ubc9 is the unique E2 for SUMOylation and upregulated in cancers [
31], we examined whether Slug SUMOylation levels could be interfered through manipulating Ubc9 expressions in vitro. As Fig.
3a showed, the intensity of the shifted band became higher when we overexpressed Ubc9 in cells. Conversely, the intensity became lower when we silenced Ubc9 in cells (Fig.
3b). Moreover, stable Slug-overexpressing cells infected with virus of control vector or Ubc9 were subcutaneously injected into the mice and the resulting tumors were harvested for examining this event in vivo after 6 weeks by immunoblotting. The results showed that the SUMOylated level of Slug was higher in tumors from Slug- and Ubc9-overexpressing cells than those from the vector control (Additional file
3: Figure S3). All these suggest the existence of Ubc9 could enhance Slug SUMOylation both in vitro and in vivo; and the expressions of Slug and Ubc9 could be used as an indicator of SUMOylated Slug level.
The N-terminal region of Slug directly interacts with the E3 PIASy
The PIAS family proteins or Pc2 are well known E3 ligase for SUMOylation [
33]; this let us further investigated whether these proteins play a central role in Slug SUMOylation. The data indicated that Slug SUMOylation was facilitated by PIAS1, PIAS2, PIAS3, and PIASy but not by Pc2 (Fig.
As literature revealed that E3 proteins can facilitate SUMOylation via direct or indirect interactions [
34], we were interested in the relationship between Slug and PIASs. Using a GST pull-down assay, we examined the association between Slug and PIASs. The results showed that all PIASs, especially PIASy, directly interacted with GST–Slug rather than GST alone (Additional file
4: Figure S4). Then, we used PIASy as a material to further clarify the detail process of Slug SUMOylation. First, we tested whether the presence of Ubc9 could affect the Slug-PIASy interaction. The results indicated that the amount of pulled-down PIASy was greater in the presence of Ubc9 than in the absence of Ubc9 (Fig.
3d), indicating that Slug can form a complex with both Ubc9 and PIASy. Next, the specific binding domain of Slug to PIASy was determined via co-immunoprecipitation assays using HEK293T extracts from cells co-expressing 3xFlag–Slug truncated mutants and HA–PIASy. Our findings showed that PIASy interacted with Slug fragments 1–268, 1–129, 33–268, and 67–268 but not with the Slug
C-terminus (fragments 130–268) (Fig.
3e), suggesting that the Slug
N-terminus may be responsible for its interaction with PIASy.
Proposed three-dimensional (3D) interaction model of Slug, Ubc9, and PIASy
According to the results of Slug domain mapping, we found that the binding regions of Ubc9, PIASy and SUMO-1 were distinct in the linear structure (Additional file
5: Figure S5a). This raises the issue how the presence of Ubc9 enhanced the Slug-PIASy interaction (Fig.
3d) and mediated Slug SUMOylation. Recent studies indicated that the molecules E3, Ubc9, SUMO and their substrates might bring into close proximity, form complexes and subsequently facilitate SUMOylation [
35]. This let us try to produce a 3D structural model combining previously published data (Ubc9 and SUMO-1) and a chimeric structure (Slug and PIASy) to understand the stereochemistry of these Slug SUMOylation related components (Additional file
5: Figure S5b). The producing model supports our hypothesis that Slug can form a complex with Ubc9 and PIASy, and the closer three-dimensional space may help the occurrence of Slug SUMOylation.
SUMOylated Slug is present on chromatin
Furthermore, we were curious about where Slug SUMOylation may occur in cell. To solve this, cells were first separated into cytosolic and chromatin-bound fractions to identify the existence of SUMOylated Slug. The results showed that most SUMOylated Slug was presented in the chromatin-bound fraction but rarely found in the cytosolic fraction (Fig.
4a). Then, we determined the compartments in which Slug and Ubc9 interact via performing co-immunoprecipitation on each cellular fraction. The data showed that the DNA-bound form of Slug interacts with Ubc9 leading to Slug SUMOylation (Fig.
Slug SUMOylation enhances its transcriptional repression activity
Since Slug is a well-known transcriptional repressor, whether SUMOylation can affect its repression activity should be further clarified. As such, we analyzed the repression activities of Slug and Slug5M by Snail-binding site (SBS)–Gal4 promoter and E-cadherin promoter, respectively. First, HEK293T cells were co-transfected with the SBS–Gal4–luciferase reporter and Gal4–VP16 activator expression plasmids together with the Slug plasmid and various SUMO-1 mutants. As previous report, Slug could repress the expression of the reporter gene [
1]. We showed the repression activity of Slug was greater in cells co-expressing Slug and SUMO-1-GG than in cells expressing Slug alone or co-expressing Slug and SUMO-1-AA (Fig.
4c). Similar results were re-confirmed by E-cadherin promoter region, a well know downstream target of Slug [
3], that wild-type Slug suppressed E-cadherin promoter expression greater than Slug5M (Fig.
4d); and the immunoblotting data demonstrated the expression of the indicated proteins (Fig.
As the DNA-binding ability of Slug does not change with Slug5M (Additional file
6: Figure S6a), we then examined whether the decreasing transcriptional suppression activities of Slug5M is due to its’ protein instabilities. Through cycloheximide treatment, we found that the protein stabilities of wild-type Slug and Slug5M were not significantly different (Additional file
6: Figure S6b). Taken together, these findings suggest that Slug SUMOylation may enhance its transcriptional suppression activity.
Slug SUMOylation increases the recruitment of HDAC1
In the previous, HDAC1/2 corepressor complex had been identified could interact with Slug [
36]. Therefore, we verified whether SUMOylation could affect this complex formation. As Fig.
4e showed that HDAC1 were associated with both wild-type Slug and Slug5M, but only wild-type Slug overexpressed with Ubc9 and SUMO-1 displayed increasing recruitment of HDAC1. Moreover, wild-type Slug interacted with Sin3A, HDAC1, HDAC2 and CtBP1 more abundantly than Slug5M in the nucleus (Additional file
7: Figure S7).
To investigate whether the increasing Slug-HDAC1 complex can enhance the suppression activity of Slug, the occupation status of endogenous HDAC1 on the E-cadherin promoter was first examined by chromatin immunoprecipitation (ChIP) assays. As expected, HDAC1 more strongly associated with the E-cadherin promoter in cells overexpressing wild-type Slug and Ubc9 than in Slug5M and Ubc9 (Fig.
4f); and this result was also re-confirmed in human lung adenocarcinoma cell line, Hop62 (Additional file
8: Figure S8). Furthermore, we analyzed the RNA expressions of several downstream targets in cells infected with the empty vector (Ctrl), wild-type Slug, or Slug5M via reverse transcriptase polymerase chain reaction (RT-PCR). The data showed that the expressions of all target genes were suppressed when wild-type Slug was expressed in HEK293, CL1–2 and CL1–5 cells, and their expressions was partially restored when Slug5M was expressed (Fig.
4g and Additional file
9: Figure S9). Collectively, all these findings implied that Slug SUMOylation might increase HDAC1 recruited to the promoter region of its downstream target genes and enhance the repression activity of Slug.
Hypoxia induces the accumulation of SUMO-1 on Slug by decreasing deSUMOylation
Hypoxia, a characteristic of advanced solid tumors, could upregulate Slug or Ubc9 expression and could also increase global protein modification by SUMO-1 [
40]. These let us wonder whether hypoxia could enhance the event of Slug SUMOylation. First, we assessed the expressions of Slug and Ubc9 in xenograft tumor specimen; and also examined the expressions of transcription factor, hypoxia-inducible factor 1-alpha (HIF1α), to evaluate their hypoxic status. The results showed that whether Slug or Ubc9 proteins were co-localized with HIF1α (Fig.
5a) suggesting that Slug SUMOylation may occur in the hypoxic region in tumor. Then, Slug SUMOylation levels were detected under hypoxia and normoxia in vitro. Our data indicated that the intensity of the shifted band was stronger in cells exposed to hypoxia than that of normoxia (Fig.
5b); though ubiquitination showed no significant difference between the normoxic and hypoxic groups (Fig.
5c). Moreover, the luciferase reporter assay also indicated that hypoxia could induce higher degree of Slug downstream reporter gene suppression than normoxia (Fig.
5d). Collectively, all these suggested that hypoxia could regulate Slug SUMOylation in vitro and in vivo.
Surprisingly, manipulating HIF1α expression did not affect Slug SUMOylation (Fig.
5e); and the interactions of Slug with Ubc9 and PIASy were only slightly increased under hypoxia (Additional file
10: Figure S10a and b). Since SUMO modification is a dynamic process, this let us further speculate whether hypoxia-induced Slug SUMOylation is through interfering the process of Slug deSUMOylation. As literature pointed that SENP1 and SENP2 proteins participate in SUMO-1 de-conjugation [
20], we first examined which component was involved in Slug deSUMOylation. Using the expression of catalytically dead mutant (C/S) of HA–SENP1 or HA–SENP2, we found that both wild-type SENP1 and SENP2 rather than their C/S mutants could suppress Slug SUMOylation (Additional file
10: Figure S10c). Then, we detected the hypoxic effects on these interactions. Interestingly, we found that the Slug-precipitated SENP1 and SENP2 levels were dramatically reduced under hypoxia (Fig.
5f). These implied that hypoxia may enhance Slug SUMOylation through down regulation of the deSUMOylation rate in cells.
Slug SUMOylation promotes tumor migration, invasion and metastasis
Further to understand the impact of SUMOylation on Slug functions; the effects of Slug SUMOylation on cell migration and invasion should be determined. The results showed that the cells presented none significant difference in cell growth (Additional file
11: Figure S11), but the migratory ability of HEK293/Slug or Slug5M were greater than that of HEK293/Ctrl under normoxia; only the wild-type Slug reached significant (
P = 0.0006). Further, the migratory effects of Slug wild-type and Slug5M were re-confirmed by overexpressing the vector control, Slug and Slug5M into two additional human lung cancer cell lines, CL1–2 and Hop62 under normoxia (Additional file
12: Figure S12a and b). Moreover, we also found that hypoxia could significantly enhance the difference between wild-type Slug- and Slug5M-increased cell migratory abilities (Fig.
6a, Left panel); and the combined effect of Slug and hypoxia were calculated (Fig.
6a, Right panel).
Additionally, overexpression of wild-type Slug and Slug5M in HEK293 cells could increase cell invasiveness by 3.3-fold and 1.9-fold relative to the vector control under normoxia, respectively (both
P < 0.05; Fig.
6b, Left panel); an observed combined effect of wild-type Slug and hypoxia were also existed (Fig.
6b, Right panel). Similarity, the effects of Slug wild-type and Slug 5 M on cell invasiveness under normoxia were also observed in both CL1–2 and Hop62 lung cancer cells (Additional file
13: Figure S13a and b). Then, we extended the assessment on metastasis in vivo; cells were intravenously injected into the lateral tail vein of mice and the formation of pulmonary nodules was observed. HEK293/Slug developed more pulmonary nodules than HEK293/Ctrl and HEK293/Slug5M (mean number of pulmonary nodules: 2.1 ± 0.57 for HEK293/Ctrl, 13.0 ± 3.18 for HEK293/Slug, and 3.9 ± 0.91 for HEK293/Slug5M; HEK293/Ctrl vs. HEK293/Slug,
P = 0.0024; Fig.
6c) with morphology of metastatic lung nodules displayed; and similar results were also seen by mice tail vein injected with CL1–2 lung cancer cells (CL1–2/Ctrl vs. CL1–2/Slug,
P = 0.0324; Fig.
6d). Collectively, all these data indicated that SUMOylation may promote Slug-induced cancer metastasis in vitro and in vivo.
Overexpression of Slug and Ubc9 associates with poor overall survival among NSCLC patients
Although our results consistently suggested that Slug SUMOylation could promote cell metastatic abilities both in vitro and in vivo, such studies do not fully reflect clinical malignancy. Accordingly, we examined the mRNA expressions of Slug and Ubc9 in primary tumor specimens from 104 NSCLC patients, with baseline characteristics showed in Table
1, by real-time quantitative PCR (Q-PCR). Similar to previous findings [
41], patients exhibiting higher Slug or Ubc9 mRNA levels experienced poorer overall survival than those displaying lower levels (Fig.
6e, left and middle panel,
P = 0.0007 and 0.0016, respectively). Analysis of the combined effect of Slug and Ubc9 on patient prognosis revealed that patients displaying both lower Slug and Ubc9 mRNA levels had best overall survival than those showing either higher Slug and Ubc9 level, with those showing both higher Slug and Ubc9 level had worst outcomes (Fig.
6e, right panel,
P = 0.0002). Multivariate Cox survival regression showed that both higher Slug and Ubc9 mRNA levels (Hazard Ratio [HR] = 5.08, 95% confidence interval [CI] = 1.99–12.92;
P = 0.001) and more advanced tumor stage were independent prognostic factors for overall survival (Table
Characteristics of 104 NSCLC patients in real-time quantitative RT-PCR analysis
No. of patients
Low Slug (%)
High Slug (%)
Low Ubc9 (%)
Number of patients
Squamous cell carcinoma
≤ 3 cm
> 3 cm
aP values were calculated using a two-sided Pearson Chi-Squared test. Slug and Ubc9 expression was designated as ‘high’ or ‘low’ using 0.6 as the cut-off point
bOne patient’ tumor size state was missing
cTwo patients’ tumor stage states were missing
Hazard ratios for death (from any cause) among patients with NSCLC based on gene expression levels as determined via real-time quantitative RT-PCR and other parameters according to multivariate Cox regression analysis
Hazard ratio (95% CI)
High Slug and Ubc9 expression
5.08 (1.99 to 12.92)
Other expression profiles
2.20 (0.82 to 5.91)
0.54 (0.24 to 1.20)
0.66 (0.32 to 1.38)
1.04 (0.28 to 3.85)
2.22 (1.48 to 3.33)
aA stepwise method was used to select the optimal multivariate Cox proportional hazard regression model. Slug and Ubc9 expression was designated as ‘high’ or ‘low’ using 0.6 as the cut-off value (low Slug and low Ubc9 as the references), and the results were adjusted according to sex (female as the reference vs. male), histological type (squamous cell carcinoma as the reference vs. adenocarcinoma), tumor size (≤ 3 cm as the reference vs. > 3 cm), and tumor stage.
P values (two-sided) were calculated using a Pearson Chi-square test. Abbreviations:
CI confidence interval
Then, the studies were also re-confirmed by examining the protein expression levels of Slug and Ubc9 in tumor specimens from another NSCLC cohort (
n = 85, Table
3). As Fig.
6f shown, patients presented with both higher Slug and Ubc9 expression levels experienced similarly poorer overall survival than those displaying both lower levels (
P = 0.0177). Multivariate Cox proportional hazard regression also was confirmatory (Table
4, HR = 5.08, 95% CI = 1.55–16.67;
P = 0.007).
Characteristics of 85 NSCLC patients in Immunohistochemistry staining analysis
No. of patients
Low Slug (%)
High Slug (%)
Low Ubc9 (%)
High Ubc9 (%)
Number of patients
Squamous cell carcinoma
Lymph node metastasis
aP values were calculated using a two-sided Pearson Chi-Squared test. Slug and Ubc9 expression was designated as ‘high’ or ‘low’ using “50 and 60% immunoreactivity was shown in tumor sections” as the cut-off point respectively
bP value was calculated using a Student’s
cThree patients’ tumor stage states were missing
Hazard ratios for death among patients with NSCLC based on protein expression levels as determined via immunohistochemistry and other parameters according to multivariate Cox regression analysis
Hazard ratio (95% CI)
High Slug and Ubc9 expression
5.08 (1.55 to 16.67)
Other expression profiles
2.15 (0.72 to 6.39)
2.45 (1.13 to 5.31)
0.63 (0.23 to 1.70)
1.21 (0.96 to 1.51)
0.94 (0.60 to 1.49)
aA stepwise method was used to select the optimal multivariate Cox proportional hazard regression model. Slug and Ubc9 expression was designated as ‘high’ or ‘low’ using “50% and 60% immunoreactivity in the tumor sections” as the respective cut-off values (low Slug and low Ubc9 as the references). The results were adjusted according to sex (female as the reference vs. male), histological type (squamous cell carcinoma as the reference vs. adenocarcinoma), tumor size (≤ 3 cm as the reference vs. > 3 cm), and tumor stage.
P values (two-sided) were calculated using a Pearson Chi-square test. Abbreviations: CI, confidence interval
Collectively, all these data indicate the malignant role of Slug SUMOylation in lung cancer and also highlight the importance to dissect the regulation mechanism of Slug SUMOylation.
Slug participates in the regulation of cell migration, apoptosis, differentiation, and therefore can promote cancer invasion and metastasis [
43]. Our study demonstrates that Slug could be SUMOylated at its
C-terminus via direct binding to Ubc9 and PIASy on chromatin. Hypoxia could induce Slug SUMOylation through inhibiting the interaction between Slug and SENP1/2 with result in downregulating the process of deSUMOylation. Detail analyses indicated that SUMOylation could enhance the transcriptional repression activity of Slug by recruiting more corepressors (like Sin3A, CtBP, HDAC1 or HDAC2), decreasing its downstream target gene expressions (such as E-cadherin, claudin 1 and occludin) and promoting cancer cell migration, invasion, and metastasis (Fig.
7). Moreover, the expressions of Slug and Ubc9 could be used as effective diagnosis markers for Slug SUMOylation and the combined effect was associated with poor overall survival in NSCLC patients.
Previously, Slug protein had been reported could be stabilized by ARF and promote prostate tumorigenesis [
44]. In that report, Xie and his colleagues found that ARF could stabilize Slug through interacting with SUMO molecule at lysine residue 192 (K192), modulate the Slug/E-cadherin signaling, and augment prostate tumorigenesis in vivo
; the regulation mechanisms of Slug SUMOylation are different with our findings. In lung cancer, we found that in addition to the K192, the amino acids 213–268 of Slug are more critical for SUMOylation and responsible for the regulation of Slug transcriptional repression activity. In fact, we ever mutated all lysine residues of Slug (Slug22M) and individually substituted them with lysine residue to find out all the SUMOylation sites of Slug (Additional file
2: Figure S2a). The results showed that rather than K192, amino acids 213–268 contribute near 75% of Slug SUMOylation; and in which, only K258 is located on a known consensus sequence, while K239 and K248 are located on non-consensus regions of SUMO modification. In addition, we further discovered that hypoxia could enhance Slug SUMOylation through disrupting the interaction between Slug and SENP1/2, and result in lung cancer malignancy. All these suggest Slug protein could be SUMOylated under different regulation and affected cancer malignancy with different action mechanisms; while ARF overexpression, Slug was SUMOylated at K192 to accumulate Slug protein to promote prostate tumorigenesis in vivo; when hypoxia or Ubc9/PIASs overexpression, Slug was SUMOylated and enhanced transcriptional repression activity in lung cancer malignancy. We provided more detail molecular mechanism to state the regulation and its critical role of Slug SUMOylation in cancer.
Typically, PTMs are sophisticated and reversible; which confer the functional abilities on proteins at the appropriate time and place [
45]. Emerging evidences suggest that Snail and Slug could be post-translationally regulated by phosphorylation, ubiquitination, SUMOylation and acetylation [
49]; and the latter three are all occurred on lysine residue. Moreover, literature also indicates that the SUMOylation and ubiquitination pathways could intersect and communicate with each other [
50]. These let us curious about whether exist any relationship between ubiquitination, acetylation and Slug SUMOyaltion. To answer this question, we further examined and dissected the responsible regions of Slug ubiquitination and acetylation. The data showed that Slug could be ubiquitinated at K116 (Additional file
14: Figure S14a) that different from the sites for SUMOylation; and with nearly non-acetylation in the steady state (Additional file
14: Figure S14b). Previously, Vernon et al. reported that the Slug
N-terminus is responsible for its ubiquitin-mediated proteasomal degradation [
47]; but we found that Slug is ubiquitinated on K116. These imply that ubiquitination and SUMOylation may occur in different Slug regions. In addition, the data of none appreciable acetylation existed in Slug protein was different with the finding that acetylation of Snail proteins could determine its role as an activator or repressor in cells [
49]. Moreover, we also observed that SUMOylation predominantly occurred on Slug rather than on Snail (Additional file
15: Figure S15). Whether the processes of these PTMs communicate with each other should be further clarified.
As we know, SUMOylation participates in functional regulation of many proteins; impaired regulation may lead human diseases like cancer, neurodegenerative disorder, cardiac disease, and ocular pathology [
53]. Till now, several studies mentioned that SUMO E1, E2, and E3 enzymes such as Ubc9 and PIAS3 are upregulated in many cancer types [
57]. Though disturbing SUMOylation could interfere several protein functions that relevant to cancer progression [
59], the detail mechanism how SUMO promotes tumorigenesis remains unclear. Here, we provide a potential mechanism by which SUMO-1 could participate in cancer metastasis through binding to Slug. The investigation showed that Slug SUMOylation might increase corepressor, HDAC1, recruitment, inhibit downstream target genes expressions, and have end result in promoting cancer malignancy. Similar findings were observed in SUMOylated Elk-1 and p300 that SUMOylation might control transcriptional repression by recruitment of HDAC2 and HDAC6; and also mentioned in previous reports [
Hypoxia, a common characteristic in most malignant tumors, stimulates a complex cancer-related gene expression and is also involved in SUMO signaling [
68]. In this study, we found that hypoxia could induce Slug SUMOylation by repressing the interaction between Slug and SENP1/2, though the detailed mechanism should be further clarified. While Slug can be phosphorylated, ubiquitinated and SUMOylated in cells, we observed no significant difference in Slug ubiquitination between normoxia and hypoxia status (Fig.
5b). Our data also indicated that both serine and threonine phosphorylations of Slug were lower under hypoxia, suggesting Slug protein could be regulated by phosphorylation or SUMOylation under hypoxia. Moreover, the Slug interactome contained several subunits of protein phosphatase 2 (PP2A), such as PPP2CB, PPP2R1A, PPP2R2A and PPP2R2D [
37]; as such, we further discovered that the PP2 inhibitor okadaic acid exposure could reverse the hypoxia-induced increasing Slug SUMOylation (Additional file
16: Figure S16). All these findings implied that Slug phosphorylation and SUMOylation were related to PP2 activity under hypoxia analogue to prior report [
In sum, our study provided new insight into the modulation of the EMT regulator Slug via SUMOylation and disclosed novel mechanisms by which SUMOylated Slug promotes cancer invasion and metastasis under hypoxia. The identification of these findings may have clinical implications in targeting lung cancer treatment.
The authors thank Dr. Hsiu-Ming Shih (Institute of Biomedical Sciences, Academia Sinica) for providing the human active and mutant construct of SUMO-1, Flag–Ubc9, wild-type and dominant negative mutant (C/S) SENP1/2. Dr. Wen-Ming Yang (Institute of Molecular Biology, National Chung Hsing University, Taichung, Taiwan) for providing the pcDNA3-HDAC1-Flag plasmid, Dr. H. K. Sytwu (National Health Research Institute) for providing the plasmids for the lentivirus infection system, Dr. Ching-Wen Lin (Institute of Biomedical Sciences, Academia Sinica) for providing the pFlag-CMV-2-Snail plasmid and Dr. Ker-Chau Li (Institute of Statistical Sciences, Academia Sinica) for providing the Hop62 lung cancer cell line. In addition, the authors would like to thank the following individuals for providing technical support: Dr. Jin-Yuan Shih (Department of Internal Medicine, National Taiwan University Hospital), Dr. Hsuan-Yu Chen (Institute of Statistical Science, Academia Sinica), Dr. Tzu-Hung Hsiao (Department of Medical Research, Taichung Veterans General Hospital), Dr. Shu-Ping Wang, Dr. Wen-Lung Wang, Dr. Lu-Kai Wang (Institute of Biomedical Sciences, Academia Sinica) and Mr. Ying-Ren Wang (Graduate Institute of Biochemistry and Molecular Biology, National Taiwan University College of Medicine). The shRNA constructs were obtained from the National RNAi Core Facility at the Institute of Molecular Biology/Genomic Research Center, Academia Sinica, Taipei, Taiwan.
Availability of data and material
All data generated or analyzed during this study are included in this published article (and its additional file).
This work was supported by grants from the Ministry of Science and Technology of the Republic of China (MOST 102–2314-B-002-046-MY3, 105–2628-B-002-007-MY3, 106–0210-01-15-02 and 107–0210-01-19-01) and National Taiwan University (NTU-CDP-103R7879 and 104R7879).
Ethics approval and consent to participate
The human tissues experiments were approved by the Institutional Review Boards of Taichung Veterans General Hospital (TVGH, Taichung, Taiwan) and National Taiwan University Hospital (NTUH, Taipei, Taiwan). Written informed consent was obtained from all patients. All mouse experiments were performed in accordance with the animal care guidelines and with approval of the Laboratory Animal Center, National Taiwan University College of Medicine (Taipei, Taiwan).
Consent for publication
All authors have reviewed the manuscript and given consent for publication.
The authors declare that they have no competing interests.
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Open AccessThis article is distributed under the terms of the Creative Commons Attribution 4.0 International License (
http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The Creative Commons Public Domain Dedication waiver (
http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated.