Background
STAT3 belongs to the signal transducers and activators of transcription (STATs) family of transcription factors (TFs) [
1]. STAT3 is activated in response to several cytokines and growth factors, including IL-6, IL-10, the epidermal growth factor (EGF), and interferon (IFN) α and is also weakly activated in response to other cytokines, including IFNγ in some cellular contexts [
2]. Activation of STAT3 involves phosphorylation of tyrosine 705 by cytokine receptor-associated Janus Kinases (JAK); the involvement of the Src and Abl tyrosine kinases as well as EGFR have also been reported [
3]. Tyrosine phosphorylation of STAT3 is followed by dimerization through phosphotyrosine-SH2 domain interaction; activated STAT3 enters the nucleus where it stimulates the transcription of its targets, including Cyclin-D1, Survivin, Vegf, C-Myc, Bcl-xL, and Bcl2 [
4,
5].
STAT3 is a key regulator of cell survival and proliferation [
6]. Its constitutive activation has been observed in many human tumors, including colon, breast, lung, pancreas and prostate cancers, melanoma, head and neck squamous carcinoma, multiple myeloma, mantle cell lymphoma, and glioma [
7,
8]. However, in certain cell types such as PTEN-deficient glioblastoma, STAT3 can become a tumor suppressor [
9,
10].
STAT1 is another member of the STAT family. It is activated mainly by IFNs α and γ, and plays a major role as a pro-inflammatory, anti-pathogen and anti-proliferative factor [
11,
12]. Its biological function is thus mostly antagonistic to that of STAT3. Despite their 50% amino acid sequence homology [
13], STAT1 and STAT3 are structurally very similar; yet some important differences have been noted in their DBD sequences [
14]. Despite its major role as a tumor antagonist, STAT1 can also have functions in cancer cells, as documented in mouse leukemia [
15].
Inhibition of STAT3 in tumor cells in which it is constitutively activated leads to cell death [
16‐
18]. This is achieved using either non-specific inhibitors such as curcumin, which also inhibits other transcription factors (NF-κB), or inhibitors specifically designed to inhibit STAT3 through non-covalent binding to the SH2 domain, such as Stattic [
19] or STA-21 [
20]. Interestingly, these compounds have little effect in cells in which STAT3 is not activated, pointing to STAT3 as a highly valid target to focus on for the design of anti-cancer compounds. However, such compounds are still poorly developed.
TFs activate transcription of their target genes by binding to distinct short DNA consensus motifs. Decoy oligonucleotides (dODNs) containing these consensus motifs can bind the DNA binding domains (DBD) of the TFs and block their activity. dODNs [
21] and hairpin dODNs (hpdODNs) [
22] have been shown to induce the death of cells in which STAT3 is activated, suggesting that the DBD is another potential target for specific inhibitory compounds. Similarly to double-stranded oligonucleotides that are used to detect active dimers in electrophoretic migration shift assays, STAT3 hpdODNs interact with activated, dimeric STAT3. This interaction impairs the binding of the dimer to importins, resulting in the sequestration of STAT3 in the cytoplasm [
23].
Yet, because of the high degree of similarity between STAT3 and STAT1 consensus DNA binding sites, STAT1 competes with activated STAT3 for dODN binding in interferon (IFN) γ-treated cells [
22,
23], thereby preventing inhibition of active STAT3. Under such conditions the dODN loses its ability to block cell proliferation. In addition, since STAT1 plays a key role in cell death processes [
24,
25], including caspases expression [
26,
27] and cooperation with p53 function [
28,
29], its inhibition by the dODN prevents cell death. Finally, IFNγ being a cell death inducer in several cell types [
22,
30‐
32], it is important to design reagents that do not interfere with STAT1, one of its key effectors. Thus, in order to elaborate target-specific anti-cancer compounds, the specificity of hpdODNs to STAT3 needs to be enhanced. It should be noted, however, that in certain cellular contexts STAT1 has been found to be a tumor promoter [
15].
The difficulty in designing dODNs recognized by STAT3 but not STAT1 lies in the striking similarity of the consensus DNA sequences of the two TFs [
33], in spite of their different cellular functions. Nevertheless, early studies on STAT3/STAT1-discriminating DNA motifs established some sequence preferences that differentiate these TFs [
34‐
39], suggesting possibilities for designing STAT3/STAT1-discriminating dODNs. The notion that discrete nucleotide modifications in target DNA sequences might alter their recognition by closely related TFs is supported by the observation that a single nucleotide change in the κB consensus motif modified NF-κB subunit specificity [
40]. Furthermore, DNA recognition by proteins relies in part on DNA shape, known to deviate from the ideal B-conformation. The nature of the nucleotides in the sequence influences conformation and dynamics: for instance, dG:dC stretches confer rigidity [
41], pyrimidine/purine steps (particularly T/dA steps) confer flexibility and may also introduce kinks [
42,
43], and dA:T stretches can have complex configurations [
41]. The coordinates from available crystal structures of both STAT1 [
44] and STAT3 [
45] (in complex with their DNA consensus sequence) were used to analyze their 3D structure using the UCSF Chimera program [
46]. Based on the differences found, new hpdODNs were designed and tested for their STAT3/STAT1 discrimination ability by measuring SW480 colon carcinoma cell death and absence of inhibition of STAT1-dependent IFNγ-induced cell death. SW480 cells offer a relevant model since these cells show constitutive activation of STAT3, which is essential for their survival, and they are susceptible to IFNγ-induced cell death, which is a STAT1-dependent process. The newly designed hpdODNs were also compared for their relative binding capacity to STAT1 and STAT3 by performing in-cell pull-downs, and for their ability to prevent nuclear transfer using immunofluorescence.
Discussion
A new hairpin decoy oligonucleotide (hpdODN) carrying STAT3's DNA binding consensus sequence was designed following 3D analysis of protein/DNA interaction and shown to induce the death of STAT3-dependent tumor cells without interfering with STAT1, a key effector of cell death. In this paper, 3D structural analyses of the protein/DNA interaction of STAT1 and STAT3 demonstrated their high similarity, confirming previous reports [
13]. These 3D analyses served as a basis for the design of new sequences with base substitutions. The new sequences were tested for their ability to induce cell death in an IFNγ-sensitive [
22,
23,
48‐
51], active-STAT3-dependent colon carcinoma cell line [
22]. This enabled the design of the STAT3-specific hpdODN labeled here as hpdODN B. The ability of hpdODN B to discriminate between STAT1 and STAT3 was assessed by: i) its ability to kill cells without interfering with IFNγ-induced cell death; ii) its ability to inhibit STAT3 targets, including cyclin D1, iii) the absence of inhibition of IFNγ-induced STAT1 phosphorylation and IRF1 expression, iv) its lack of interaction with STAT1 in pull-down assays and iv) its inability to inhibit IFNγ-induced STAT1 nuclear location. Indeed, hpdODN A treatment, but not hpdODN B treatment, reduced STAT1 phosphorylation, probably by impairing nucleocytoplasmic shuttling as previously suggested [
22,
23]. Nevertheless, despite its ability to discriminate between STAT1 and STAT3, hpdODN B probably has a residual affinity for STAT1, as shown by low detection of STAT1 in pull-down assays and the fact that cell death induction by hpdODN B and IFNγ are not additive.
The STAT3/STAT1-discriminating hpdODN was obtained by replacing key nucleotides that 3D analyses had shown to be in the vicinity of amino acids of the DBD that distinguish the two STATs; the similarity of their DNA consensus sequences, despite their different functions, has been recognized for some time [
33]. Examination of the nucleotide modifications that led to STAT1/STAT3-discriminating hpdODN B showed that they are compatible with previous in vitro DNA-binding studies, such as the preference for T at 1003 and 1005, dC at 1010 and dA at 1015 of STAT3 [
36‐
38]. The fact that T at 1003 does not favor STAT1 binding is also in agreement with the earlier suggestion that "selection for a dG:dC base pair at position 7
(here: 1017) is likely to involve Glu 421 (
of STAT1) which can accept hydrogen bonds from guanine in the minor groove" [
44]. This has also been noted by others [
39]. Finally, altered recognition by a TF following single nucleotide changes has been previously shown, for instance with NF-κB subunit recognition of κB [
40]. One notable property of the hpdODN B is its dissymmetry. A symmetric version (with dA replacing dG in 1017) was tested and is apparently not different from hpdODN B (not shown). Intriguingly, although the preference of hpdODN D for STAT1 was expected from previous data showing its STAT1-specific binding [
47], its basis is not clear and may rest upon properties beyond nucleotide sequence such as DNA shape. The shape and flexibility of DNA strands are known to be influenced by their nucleotide content; here the 8-pyrimidine stretch in hpdODN B may confer a higher flexibility than hpdODN A and may account for a differential interaction with STAT3 Arg 423 and STAT1 Glu 421. In fact, the molecular dynamics studies which describe a scissor-like molecular movement upon DNA binding for STAT3, but not for STAT1 [
13] suggest that the flexibility of the DNA target may play a role in binding and therefore underly the preference of hpdODN B for STAT3. It may also account for the greater sensitivity of STAT3 to an intact palindromic structure compared to STAT1, as previously stated [
38]. Protein binding itself can affect DNA bending, as shown with the high-affinity target of the papillomavirus E2 [
52]. Nevertheless, despite its efficiency, the precise mechanism whereby the hpdODN B discriminates between STAT1 and STAT3 in cells is not understood. Changes in DNA shape may play a role in the preferential recognition of hpdODN B by STAT3; co-factors may also be involved in DNA recognition by STAT3, and might associate more efficiently when hpdODN B is used. The process might also be more complex than mere differential DNA binding: STAT1 and STAT3 are reciprocally regulated [
2,
53‐
55] and the relative abundance of their active forms may itself play a key role in biological responses, as previously discussed [
56]. Another level of complexity arises from the fact that in cells in which STAT3 has been suppressed, IFNγ-activated STAT1 induces the expression of mitogenic STAT3-targets [
57,
58]. Furthermore, STAT1 and STAT3 form heterodimers, whose function has not been elucidated to date. In this respect, quantification of the relative amounts of STAT1 and STAT3 bound to the hpdODNs A and B may help understand the complex interaction of these TFs. Preliminary experiments that are underway suggest a difference in heterodimer content. Therefore, it is possible that hpdODN B functions in cells by tilting the active STAT1/active STAT3 balance toward STAT1, thereby inducing cell death.
Material and methods
Computer analysis of STAT3 and STAT1
The PDB (Protein Data Base) files for STAT1 (1BF5) [
44] and STAT3 (1BG1) [
45] were downloaded and analyzed using Chimera [
46]. The STAT1 and STAT3 crystals used in the X-ray diffraction studies were proteins complexed with oligonucleotide duplexes featuring a consensus DNA sequence [
34,
35,
59] (see Figure
3). To compare the STAT1 and STAT3 DBDs in a complex with their DNA consensus sequences, the missing complementary strand of the STAT3-bound oligonucleotide [
45] was reconstructed through crystal symmetry operations.
Decoy oligonucleotides
The STAT3 decoy ODNs used were
RHN-(CH
2)
6-CATTTCCCGTAAATCGAAGATTTACGGGAAATG-(CH
2)
6-NHR (STAT3 decoy hpdODN), derived from the serum-inducible element of the human c-fos promoter [
34] (hpdODN A) and previously used in the lab [
22,
23].
RHN-(CH2)6-TATTTCCCCTAAATGGAACATTTAGGGGAAATA-(CH2)6-NHR (hpdODN B); RHN-(CH2)6-CATTTCCCCTAAATCGAAGATTTAGGGGAAATG-(CH2)6-NHR (hpdODN C); RHN-(CH2)6-GCTTTCCCCTAAATGGAACATTTAGGGGAAAGC-(CH2)6-NHR (hpdODN D); and the following mutated hpdODN as a negative control:
RHN-(CH
2)
6-CGATTGCCACAATCGGAACGATTGTGGCAATCG-(CH
2)
6-NHR (hpdODN E) (where R was either: hydrogen, fluorescein or biotin). The addition of fluorescein or biotin, followed by high-performance liquid chromatography, were carried out by the manufacturer (Eurogentec, Seraing, Belgium) using in-house protocols. The hairpin sequence GAA, previously shown to confer stability and nuclease resistance [
60,
61], was included in the dODNs. In the hpdODNs, the hairpin motif was built and incorporated in the X-ray structure using the BCE (Biomolecular Chain Elasticity) approach [
62]; this showed that the hairpin did not interfere with the DBD-DNA interaction (not shown).
Cell culture and reagents
SW480 (colon adenocarcinoma) cells were grown in DMEM (Invitrogen, Lifetechnologies, Cergy-Pontoise, France), supplemented with 10% FCS (Lonza, Levallois-Perret, France), 100 U/ml penicillin, 10 μg/ml streptomycin (GibcoBRL), 1 mM sodium pyruvate (GibcoBRL), MEM vitamins (100x, Invitrogen) and 5 μg/ml plasmocin (Cayla InvivoGen, Toulouse, France). Sodium orthovanadate (100 μM) was from Fischer (Illkirch, France). Interferon γ (50 ng/ml) was from Promocell (GmbH, Heidelberg, Germany) or Sigma-Aldrich (Montigny-le-Bretonneux, France).
Transfections
Cells were grown in 4-well plates to a density of 0.25 × 10
6 cells/ml. When the cells reached 50-60% confluence, they were transfected with the different STAT3 hpdODNs or the control hpdODN (2 μg/ml, corresponding to 400 nM) into 150 μL of DMEM medium (without SVF) combined with polyethyleneimine (PEI, reference 408727, average MW: 25000, Sigma-Aldrich), with an hpdODN/PEI ratio of 1:1. For immunocytochemistry, liposomes prepared as previously described [
23] were used (at 2 μg of cationic lipid). After 6 h at 37°C in a humidified 5% CO
2 incubator, the cells were placed in fresh serum-containing medium. Cells were examined after 48 h in the humidified incubator.
Flow cytometry and cell viability
To measure cell death, cells were resuspended in annexin V-binding buffer, incubated with 5 μL of propidium iodide (PI, BD Pharmingen, Morangis, France) and subjected to flow cytometry analysis, using a FACS Canto II Flow Cytometer (BD). To enable selective analysis of the cells that had incorporated the various hpdODNs, fluorescein-labelled hpdODNs were used. Fluorescein-labelled cells were analyzed for PI incorporation (red channel) or annexin V labelling (APC, red channel). A cell death index was established through computation of averages.
Gel electrophoresis, western blotting
Cells were washed in Phosphate-Buffered Saline (PBS), lysed in sodium dodecyl sulfate (SDS) sample buffer (50 mM Tris-HCl pH 6.8 (Bio-Rad, Marnes-la-Coquette, France), 2% SDS (Sigma-Aldrich), 20% glycerol (Prolabo, Fontenay-sous-Bois, France), 1 mM sodium vanadate (Na3VO4, Labosi, Elancourt, France), 1 mM dithiothreitritol (DTT) (Merck, Fontenay-sous-Bois, France) and 0.01% bromophenol blue (Sigma-Aldrich)), sonicated and stored at -70°C. Proteins (50 μg) were separated on SDS polyacrylamide gels (PAGE) (10%) and transferred onto nitrocellulose membranes; membranes blocked with dry skimmed milk (5%) in Tris Buffered Saline (TBS) were incubated with antibody overnight at 4°C. Anti-phospho-STAT1, anti-STAT1 and anti-STAT3 (1:1000) (Cell Signaling, Ozyme, Montigny-le-Bretonneux, France), anti-cyclin D1 and anti-IRF1 (1:1000) (Santa Cruz, Tebu-bio, Le Perray en Yvelines, France) were used. Blots were washed in TBS with Tween (0.05%) (TBS-T), incubated with peroxidase-coupled goat anti-mouse (Santa Cruz) or goat anti-rabbit (Upstate, Ozyme) secondary antibody (1:20,000), washed in TBS-T and revealed by chemiluminescence (LumiGLO reagent and peroxide; Cell Signaling) and autoradiography (X-Omat R film; Kodak). When necessary, membranes were stripped with Blot Restore Kit (Chemicon International, Millipore, Saint-Quentin-en-Yvelines, France) and reprobed with anti-tubulin (Cell Signaling) or anti-actin (Santa Cruz) antibody to ensure equal loading of the gels. Prestained molecular weight standards (Fermentas, Saint-Rémy-lès-Chevreuse, France) were used.
Oligodeoxynucleotide pull-down
For in-cell hpdODN pull-down assays, cells were transfected with the biotinylated hpdODNs (1 μg/ml, 6 h), as described under oligonucleotide transfection, and then lysed in cell lysis buffer (1% NP40, 50 mM Hepes, pH 7.6, 140 mM NaCl) containing salmon sperm DNA (1 μg/assay). Protein concentration was measured in the samples. Extracts (250 μg of protein per tube) were recovered on avidin-sepharose beads (50 μL) (NeutrAvidin, Pierce); beads were incubated for 30 min at 4°C in binding buffer (1% NP40, 50 mM Hepes, pH 7.6, 140 mM NaCl). After washing with binding buffer, complexes were eluted in SDS sample buffer, separated on SDS-PAGE (8%), and subjected to immunoblotting using anti-STAT1 or anti-STAT3 antibodies (Cell Signaling) and processed as above.
Immunocytochemistry
Cells were grown at 50-60% confluence in 8-well plates (Lab-Tek, Nunc, Rochester, USA) to a density of 105 cells/ml. Cells were transfected with fluorescein-labelled hpdODNs, incubated (6 to 16 h), washed in PBS, fixed with 3.7% formaldehyde for 15 min, permeabilized in 0.1% Triton X-100 for 15 min and incubated in 5% FCS-0.1% Tween- PBS for 1 h. Cells were stained with anti-STAT3 or anti-STAT1 antibody (Cell Signaling) (1:100) for 2 h, then stained with an Alexa fluor 546-labeled secondary antibody (Invitrogen) (1:200) for 90 min. Cells, counterstained with 4', 6'-diamidino-2-phenylindole (DAPI), were mounted onto glass slides with Vectashield (Vectorlabs, Clinsciences, Montrouge, France). Fluorescence images were acquired using a Zeiss Axioplan 2 Deconvolution microscope (Carl Zeiss, Le Pecq, France) and analyzed with Metafer4 (Metasystems, Altlussheim, Germany).
Competing interests
The authors declare that they have no competing interests.
Authors' contributions
IS, PL, VM, SLC, JAHC and RF made substantial contributions to conception, design, and data acquisition, analysis and interpretation; NVB and FBM were involved in revising the manuscript for important intellectual content; and IS, PL, VM, JAHC, SLC, NVB, FBM and RF gave final approval of the version to be published. All authors read and approved the final manuscript.