G4-DNA Formation in the HRAS Promoter and Rational Design of Decoy Oligonucleotides for Cancer Therapy

HRAS is a proto-oncogene involved in the tumorigenesis of urinary bladder cancer. In the HRAS promoter we identified two G-rich elements, hras-1 and hras-2, that fold, respectively, into an antiparallel and a parallel quadruplex (qhras-1, qhras-2). When we introduced in sequence hras-1 or hras-2 two point mutations that block quadruplex formation, transcription increased 5-fold, but when we stabilized the G-quadruplexes by guanidinium phthalocyanines, transcription decreased to 20% of control. By ChIP we found that sequence hras-1 is bound only by MAZ, while hras-2 is bound by MAZ and Sp1: two transcription factors recognizing guanine boxes. We also discovered by EMSA that recombinant MAZ-GST binds to both HRAS quadruplexes, while Sp1-GST only binds to qhras-1. The over-expression of MAZ and Sp1 synergistically activates HRAS transcription, while silencing each gene by RNAi results in a strong down-regulation of transcription. All these data indicate that the HRAS G-quadruplexes behave as transcription repressors. Finally, we designed decoy oligonucleotides mimicking the HRAS quadruplexes, bearing (R)-1-O-[4-(1-Pyrenylethynyl) phenylmethyl] glycerol and LNA modifications to increase their stability and nuclease resistance (G4-decoys). The G4-decoys repressed HRAS transcription and caused a strong antiproliferative effect, mediated by apoptosis, in T24 bladder cancer cells where HRAS is mutated.


Introduction
The ras gene family consists of three functional proto-oncogenes (HRAS, NRAS and KRAS) that encode for guanine-binding proteins sharing a high homology (p21 RAS ) [1]. These proteins, located on the inner cell membrane through a farnesyl group [2], are active when they are bound to GTP, and inactive when GTP is hydrolyzed to GDP [3]. Ras proteins regulate cellular responses to many extracellular stimuli, including mitogens and differentiation factors [4]. The ras genes are expressed in a tissue-specific fashion: HRAS is highly expressed in skin and skeletal muscles, KRAS in colon and thymus and NRAS in male germinal tissue [1]. The ras genes have similar structures and sequences with five exons, the first of which is noncoding, and conserved splicing sites. The introns, instead, have different lengths and sequences [1]. The ras proto-oncogenes are converted to oncogenes by point mutations that decrease the capacity of the encoded protein to hydrolyze GTP to GDP, with the result that p21 RAS remains constitutively active. Hyperactivated ras proteins stimulate phosphorylation cascades including the Raf/MEK/ERK pathway which leads to uncontrolled cell proliferation [5,6]. Mutations in the ras genes are frequently found in many human tumors [7,8]. HRAS mutations are less common, but they have a high prevalence in skin papillomas and urinary bladder tumors [9]. As 80% of bladder tumors harbor HRAS mutations [10] and more than half of bladder tumors overexpress HRAS [11], both mutation and overexpression are important factors in the tumorigenesis of bladder cancer [12]. Actually, it has been recently shown that low-level expressions of constitutively active HRAS induced simple urothelial hyperplasia, while the doubling of the activated HRAS oncogene triggered rapidly growing and penetrating tumors throughout the urinary tract. Given the crucial role of HRAS overexpression and mutations in the tumorigenesis of bladder cancer, one attractive therapeutic strategy could be to inhibit HRAS transcription with molecules that are able to impair the activity of the gene promoter. For this aim we asked how HRAS transcription is regulated. We observed that the promoter of HRAS contains numerous copies of the GGGCGGG element or its complement. This G-box has been shown to interact with the Sp1 transcription factor [13,14]. Upstream of the transcription start site (TSS) there are runs of guanines spanning over three Sp1 sites, which are potential sites for G-quadruplex formation. We thus hypothesized that the G-rich elements might play a role in transcription regulation. G4-DNA are unusual structures stabilized by planar arrays of four guanines (G-quartet) linked one to the other by Hoogsteen hydrogen bonds [15]. The edges of the terminal G-quartets are connected by loops that can vary both in length and topology, giving rise to a variety of different conformations [16]. Genome-wide analyses have revealed that runs of guanines are abundant in gene promoter regions surrounding TSS [17][18][19][20]. It has been theorized therefore that G4-DNA might be involved in transcription regulation [21][22][23][24][25].
Our study provides compelling evidence that HRAS transcription is regulated by the interplay between Sp1, MAZ and G4-DNA, which acts as a transcription repressor. On the basis of this discovery we have designed G-rich oligonucleotides specific for HRAS which have a strong antiproliferative effect in urinary cancer cells bearing a mutant HRAS. Although the cytotoxicity of certain G-rich oligonucleotides has been previously reported, their mechanism of action is not yet fully understood [26,27]. Our study shows that the designed quadruplex-forming oligonucleotides may act by sequestering MAZ and thus impairing HRAS transcription. For their potent antiproliferative effect in T24 urinary bladder cancer cells, G4-decoys seem to be very promising effector drugs for urinary bladder cancer therapy.

Results
The HRAS promoter is structurally polymorphic The promoter of the human HRAS gene lacks typical TATA and CAAT boxes, contains a high G+C content (80%) and multiple copies of GGGCGGG (G-box), recognized by the transcription factor Sp1 [13,14]. The three G-boxes closest to the RNA start sites overlap quadruplex-forming sequences, namely hras-1 (435-462, accession number J00277) and hras-2 (506-530, J00277) ( Figure 1). According to a recent study, quadruplex-forming sequences covering Sp1 binding elements are present in several genes [28]. We have obtained a first hint that the HRAS promoter is structurally polymorphic while sequencing the expression vectors specially constructed for this study. When sequencing primer-extension reactions were performed with primers complementary to the G-rich strand, Taq polymerase unexpectedly arrested at the hras-2 or hras-1 G-rich elements. In contrast, with primers complementary to the C-rich strand we did not observe any impediment. This suggested that both hras-1 and hras-2 sequences formed unusual structures. Our hypothesis was confirmed by polymerase-stop assays. We designed two linear wild-type templates containing hras-1 or hras-2 and one mutant template in which four GRT mutations were introduced into hras-2 to prevent quadruplex formation. Primer-extension reactions showed that Taq polymerase in the presence of potassium arrested at the 39 end of hras-2 or hras-1, just before the first run of guanines, in keeping with the formation of a G-quadruplex structure by each G-rich element ( Figure S1).
Promoter sequences hras-1 and hras-2 form stable G4-DNA structures in vitro An insight into the G-quadruplexes formed by the HRAS G-rich elements was obtained by DMS-footprinting, circular dichroism (CD) and fluorescence resonance energy transfer (FRET) experiments. Figure 2 shows the results obtained with sequence hras-1. The DMS-footprints of 27-mer hras-1 in water or buffer containing 100 mM Li + or Cs + (lanes 1,4,5) show that all guanines react with DMS. Instead, in the presence of 50, 100 or 140 mM KCl (lanes 2, 3,9), the guanines of the G-runs A-D are progressively protected, while guanines G8 and G14 in the intervening ''TTGC'' and ''CGCA'' sequences are not. This cleavage pattern suggests that hras-1 folds into a G-quadruplex (qhras-1). In the presence of 50 nM TMPyP4, hras-1 gives a strong footprint either at low (1 mM) or high (100 mM) KCl concentration (lanes 7, 8) [29]. As expected, the mutant sequence h1-mut, bearing hras-1 with four GRT mutations that abolish the folding, does not give any footprint. To determine the strand orientation of qhras-1 we used CD (Figure 2c,d). The CD spectrum of qhras-1 in 100 mM KCl is characterized by positive and negative ellipticities at 287 and 260 nm, respectively, typical of an antiparallel conformation [30]. Heating the sample we obtained a melting curve (287 nm ellipticity versus temperature) that was not perfectly superimposable with the cooling curve, as the first was slightly biphasic while the second was monophasic. A more sensitive method of analysis was obtained by FRET-melting experiments with sequence hras-1 endlabelled with FAM and TAMRA at the 5 and 39 end, respectively. We obtained a well resolved biphasic melting curve with T M 's at 53 and 67uC indicating that hras-1 folds in at least two quadruplexes ( Figure 2e). When the concentration of hras-1 was Figure 1. Sequence of the human HRAS promoter upstream of TSS. Two G-rich elements (hras-1 and hras-2) located upstream of transcription start site can potentially fold into G-quadruplex structures. The quadruplex-forming G-rich elements contain the binding sites for the transcription factors MAZ and Sp1. doi:10.1371/journal.pone.0024421.g001 increased one order of magnitude from 200 nM to 2 mM, the T M 's did not change: a behavior typical of an intramolecular structure (not shown). Together, these data suggest that hras-1 adopts an antiparallel quadruplex which could assume different topologies: the one with lateral loops is shown in Figure 2f [16]. Although we know by footprinting and CD data the guanines that are involved in the formation of antiparallel qhras-1, an insight into structure of this quadruplex will only be obtained by NMR. Figure 3 shows the results obtained with sequence hras-2. The DMS-footprints of 24mer hras-2 shows that in 10, 50, 100 and 140 mM KCl (lanes 2-5), the G-runs A-D are protected from DMS, while G10, G12, G13, G21 and G22 react with DMS. This indicates that the guanine triads forming the quadruplex should be G3-G4-G5, G7-G8-G9, G14-G15-G16, G18-G19-G20. The fact that G16 shows some reactivity to DMS suggests that the pentaguanine G12-G16 stretch can participate to the quadruplex either with G14-G15-G16 or with G13-G14-G15. The mutant h2-mut sequence and wild type hras-2 in 100 mM Li + or Cs + did not give any footprint (lanes [6][7][8]. The footprinting in the presence of 50 nM TMPyP4 is very strong (lanes 9, 10). The CD spectrum of hras-2 shows a strong ellipticity at 260 nm indicative of a G-quadruplex with a parallel conformation ( Figure 3c) [30]. This structure is so stable that the CD at 85uC still shows an intense 260 nm ellipticity. The stability at various KCl concentrations was determined with sequence hras-2 labeled with FAM and TAMRA. We performed FRET-melting experiments at 20, 40, 60, 100 and 140 mM KCl and obtained T M values of 78 to 82, 85, .90 and .92uC, respectively ( Figure 3d). In this case, too, the T M 's did not change when the concentration was increased one order of magnitude (from 200 nM to 2 mM) (not shown). Overall, our data demonstrate that hras-2 forms a parallel G-quadruplex (qhras-2) whose putative structure inferred by DMS-footprinting and CD should have either 1/1/4 or 1/2/3 loops (Figure 3e). A definitive structure assignment will only be possible on the basis of NMR data.
G4-DNA destabilizing point mutations in hras-1 and hras-2 strongly upregulate HRAS transcription As sequences hras-1 and hras-2 are located immediately upstream of the transcription start site and form in vitro stable G-quadruplexes, we asked what happens to HRAS transcription when the capacity of quadruplex formation by sequences hras-1 and hras-2 is abolished. To address this point we constructed a plasmid, pHRAS-luc, bearing firefly luciferase driven by the HRAS promoter. From pHRAS-luc we obtained by site-directed mutagenesis two mutant plasmids: pHRAS- mut1 and pHRAS-mut2, where two guanines in the second and third G-tetrads of the putative quadruplexes were replaced with thymine/ cytosine or thymines (Figure 4a). These mutations abrogated the capacity of sequences hras-1 and hras-2 to fold into a quadruplex ( Figure S2). The wild-type and mutant plasmids were co-transfected in HeLa cells with pRL-CMV, a vector encoding for Renilla luciferase. Forty-eight hours after transfection we measured firefly and Renilla luciferases activity in the lysates of untreated and treated cells. The results reported in Figure 4b show that blocking quadruplex formation causes a dramatic increase of firefly luciferase expression, up to 5-fold compared to control. This strongly indicates that the Gquadruplexes formed by sequences hras-1 and hras-2 are structural elements of the HRAS promoter that behave as repressors, as observed for the CMYC gene [21].

G4-DNA stabilizing guanidinium phthalocyanines repress HRAS transcription
Given that quadruplex DNA behaves as a repressor element for HRAS, we examined the effect on transcription of G4-DNA stabilizing ligands. Due to their specificity for G4-DNA, in our experiments we used as ligands modified phthalocyanines: tetrakis-(diisopropyl-guanidine) phthalocyanine (DIGP) and its Zn-containing derivative Zn-DIGP ( Figure 5a) [31][32][33]. HeLa cells were first treated with 1 or 5 mM DIGP or Zn-DIGP for 24 h and then co-transfected with a mixture of pHRAS-luc and pRL-CMV. After transfection, the cells were let to grow for 48 h before firefly and Renilla luciferases activity were measured with a luminometer. As a control (i) we treated the cells with TMPyP2, a porphyrin that does not bind to G4-DNA [34]; (ii) we used as a reporter vector pHRAS-mut1 or pHRAS-mut2 bearing a mutated G-element unable to form a quadruplex structure. Figure 5b, c, d shows that DIGP and Zn-DIGP strongly inhibit luciferase from wild-type pHRAS-luc, while TMPyP2 does not. Moreover, the phthalocyanines do not inhibit luciferase in the cells treated with the mutant plasmids pHRAS-mut1 or pHRAS-mut2, in keeping with the fact that the G-quadruplexes cannot be formed by these vectors. These data provide further evidence that quadruplexes qhras-1 and qhras-2 act as transcription repressors.
Transcription factors Sp1 and MAZ bind to quadruplex forming sequences (QFS) hras-1 and hras-2 As our transcription data suggest that both sequences hras-1 and hras-2 are critical for transcription, we asked if they are recognized by nuclear proteins. An answer to this question was obtained by mobility-shift assays with a HeLa nuclear extract. Figure 6a shows that both HRAS duplexes form with a HeLa nuclear extract distinct DNA-protein complexes, thus pointing out the relevance of these sequences for HRAS transcription. Ishii et al. showed that sequence hras-2, which contains two copies of GGGCGGG, is bound by Sp1 [13,14]. However, sequences hras-1 and hras-2 should also interact with the myc associated zinc-finger transcription factor (MAZ), its binding site being (G/C)GG(C/A)GGG [35,36]. In fact, it has been reported that MAZ and Sp1 often regulate transcription in a cooperative way [37,38].
To prove that Sp1 and MAZ bind to sequences hras-1 and hras-2 under in vivo conditions, we performed chromatin immunoprecipitation (ChIP) experiments. Living HeLa cells were treated with formaldehyde to crosslink the DNA-protein complexes, chromatin was sheared into fragments and then immunoprecipitated with anti-MAZ and anti-Sp1 antibodies (Abs). The abundance of HRAS promoter sequences in the immunoprecipitates was measured by PCR using primers specific for sequences hras-1 and hras-2. The results reported in Figure 6b show that: IgG Ab did not immunoprecipitate DNA-protein complexes containing sequence hras-1 or hras-2 (negative control); anti-MAZ Ab did immunoprecipitate a DNA-protein complex containing hras-1 and hras-2; anti-Sp1 Ab pulled down a complex with hras-2. By measuring the intensity of the bands, we found that the HRAS sequences were more abundant in the immunoprecipitates with anti-MAZ and anti-Sp1 Abs than in the IgG Ab immunoprecipitate. We thus concluded that under in vivo conditions MAZ is associated to sequences hras-1 and hras-2, while Sp1 is associated to sequence hras-2. This was confirmed by EMSA with recombinant MAZ and Sp1 ( Figure S3). As previous studies have shown that MAZ binds to G4-DNA from the murine KRAS [24] and c-myb [39] promoters, we explored whether it also binds to the HRAS quadruplexes. We performed EMSA with recombinant MAZ and Sp1, which were expressed in E. coli as GST fusion proteins ( Figure S4). Figure 6c shows that at pH 7.4, 50 mM KCl, qhras-1 and qhras-2 with increasing amounts of MAZ-GST form -in the presence of 100fold excess nonradiolabelled poly d(I-C) -two retarded bands due to the formation of two DNA-protein complexes, most likely with 1:1 and 1:2 stoichiometry. It is important to note that in 50 mM CsCl, where the HRAS sequences are unstructured (see DMS footprinting), both complexes are destabilized indicating that the DNA-protein interaction is mediated by the quadruplex structure. In a buffer at pH 9 where MAZ probably modifies its own folding, the DNA-protein interaction is also inhibited. We also tested the binding of Sp1-GST to the HRAS quadruplexes and found that Sp1-GST interacts with the antiparallel qhras-1 quadruplex, but in a weaker way compared to MAZ.

MAZ and Sp1 synergistically activate HRAS transcription
To prove that MAZ and Sp1 are involved in HRAS transcription, we co-transfected plasmid pHRAS-luc in HeLa cells either with pMAZ (encoding for MAZ) or pSp1 (encoding for Sp1) or with a mixture of both plasmids (Figure 7a). When pMAZ or pSp1 was cotransfected with the reporter vector, the level of luciferase expression increased by 50% compared to control. When both pMAZ and pSp1 were cotransfected with the reporter vector, transcription increased nearly 3-fold over control, indicating that both proteins synergistically activated HRAS promoter. The synergy was stronger (7-fold compared to control) when pMAZ and pSp1 were cotransfected with the mutant vector pHRAS-mut1 or pHRAS-mut2, bearing mutated hras-1 or hras-2 sequences which are unable to form quadruplex structures. This suggests that transcription is activated when sequences hras-1 and hras-2 are in the unfolded duplex conformation. Furthermore, as a proof that transcription requires both MAZ and Sp1, we knocked down with validated shRNA each of the two transcription factors separately ( Figure S5). HeLa cells were treated with shRNA specific for MAZ or Sp1 and the levels of HRAS transcripts were measured by real-time PCR, at 48 and 72 h following treatment. It can be seen that HRAS transcription was reduced to 30 and 20% of control, 48 h following treatment with anti MAZ and anti Sp1 shRNA, respectively (Figure 7b). Taken together our data demonstrate that HRAS transcription is activated synergistically by MAZ and Sp1.

MAZ destabilizes the HRAS G-quadruplexes
Considering that MAZ is essential for HRAS transcription and recognizes qhras-1 and qhras-2, we asked whether these quadruplexes are unfolded by MAZ. To answer this question, we performed FRET-melting experiments with recombinant MAZ-GST. Quadruplex qhras-1 end-labelled with FAM and TAMRA was incubated in 50 mM KCl, for 30 min in the presence of increasing amounts of MAZ-GST or control protein (BSA or trypsinogen). The melting curves (2dF/dT) showed that qhras-1 alone melts with its typical biphasic profile with T M at 53 and 67uC (Figure 8a curve 1). But when qhras-1 was incubated with MAZ-GST at (moles of protein)/(moles of DNA) ratio r = 2.5, 5, 10 (curves 4, 5 and 6), the melting profile changed significantly as the T M 's fell to ,46uC. This indicates that both qhras-1 quadruplexes are destabilized by MAZ-GST. As expected, unspecific proteins such as BSA or trypsinogen at r = 10 did not affect the melting of qhras-1 (curves 2,3). Due to its high stability in potassium (T M = 78uC in 10 mM KCl), qhras-2 was analyzed in 100 mM NaCl where the T M was 65uC. Incubated with MAZ- Figure 5. Effect of quadruplex-stabilizing ligands on HRAS transcription. (a) Structure of tetrakis-(diisopropyl-guanidine) phthalocyanine (DIGP) used, with its Zn-containing analogue Zn-DIGP, for luciferase experiments; (b) Dual luciferase assays showing that guanidinium phthalocyanines DIGP and Zn-DIGP, stabilizing hras-1 and hras-2 quadruplexes, repress HRAS promoter activity. This effect is not observed with the mutant plasmids pHRAS-mut1 and pHRAS-mut2 (control) (panels c and d). Relative luciferase is given by R T /R NT 6100, where R T and R NT are (firefly luciferase)/(Renilla luciferase) in T24 cells treated with phthalocyanines and untreated cells. Differences from the control are supported by Student's t test, P,0.05 (one asterisk), P,0.01 (two asterisks). doi:10.1371/journal.pone.0024421.g005 GST at r = 2.5, 5, 10, the quadruplex melted at 42uC, indicating that also qhras-2 was destabilized by MAZ-GST, but not by BSA or trypsinogen. We also ascertained that GST had no influence on the melting of the HRAS quadruplexes ( Figure S6b). An additional control that we performed to rule out that bacterial proteins did not contribute to quadruplex destabilization was to mix Glutathione Sepharose 4B resin with an extract obtained from nontransformed BL21 DE3 plysS bacteria. An SDS-PAGE analysis of the eluate with 10 mM reduced glutathione showed that no bacterial proteins bound non-specifically to the resin ( Figure S6a).
We were surprised about the unfolding activity of MAZ, as we recently reported that MAZ stabilized the murine KRAS quadruplex [24]. However, it should be borne in mind that MAZ having six zinc fingers can interact with DNA in a complex way, as observed with qTBP42 and CBF-A [40,41]. These proteins disrupt the dimeric quadruplex formed by the FMR1 d(CGG) n repeat but they also stabilize the telomeric quadruplex d(TTAGGG) n . CBF-4 and qTBP42 have four RNP domains but only two are involved in G4-DNA binding. It has been demonstrated that different RNP combinations are responsible for either the stabilizing or the destabilizing activity.

G4-DNA decoy ODNs mimicking HRAS quadruplexes inhibit transcription and cell growth
Considering that HRAS transcription is activated by a combined action of MAZ and Sp1 and that qhras-1 and qhras-2 bind to MAZ (qhras-1 also binds to Sp1), we designed a decoy strategy against HRAS oncogene. We reasoned that these oligonucleotides mimicking quadruplexes qhras-1 or qhras-2 (G4-decoys) should sequester MAZ and inhibit HRAS transcription as well as cell growth (Figure 8c). To increase the stability of the G4-decoys we inserted in their sequence one or two units of (R)-1-O-[4-(1-Pyrenylethynyl) phenylmethyl] glycerol (P) to cap the quadruplex ends [42]. To increase their resistance against endogenous nucleases, we introduced LNA residues at the 39 end and in one loop. We designed three G4-decoys mimicking quadruplex qhras-1 (3, 4 and 5) and three mimicking quadruplex qhras-2 (6, 7 and 8) ( Figure S7) ( Table 2). The latter show CD spectra similar to that of wild type qhras-2, with a strong ellipticity at 260 nm typical of a parallel quadruplex conformation, while the former show a CD slightly different from that of qhras-1, with two positive ellipticities at 260 and 287 nm, suggesting that they should form a mixed parallel-antiparallel quadruplex [43] (Figure 9a). The antiproliferative activity of the designed G4-decoys was tested in two types of cells: HeLa and T24 urinary bladder cancer cells that harbour a mutant HRAS [codon 12 GGC (Gly) is changed in GTC (Val)], expressing a hyperactivated HRAS protein [44]. In a first set of experiments we delivered the G4-decoys (5 mM) without any transfectant agent and did not observe any effect on cell proliferation. To see if this was due to a poor oligonucleotide uptake, we analyzed by confocal microscopy T24 bladder cells treated for 24 h with 5 mM decoy 3 covalently labelled with fluorescein (3-F). Figure 9b shows typical images: the nuclei stained with propidium iodide (red fluorescence), the intracellular distribution of the oligonucleotide (green fluorescence) and the overlay of both emissions. It appears clear that 3-F localizes basically in the cytoplasm, and this is the reason why the decoys are not active when they are delivered as free molecules. We then treated the cells with 3-F (350 nM) mixed to polyethylenimine (PEI). Figure 9b shows that despite the low concentration, 3-F was efficiently taken up by the cells and uniformly distributed in the nucleus. We therefore decided to perform our proliferation experiments using PEI as a transfectant agent. We did a doseresponse experiment by delivering the G4-decoys in two doses to T24 cells, one 48 h after the other. Three days after the first delivery, we performed a resazurin proliferation assay. It can be seen that the three G4-decoys specific for hras-1 (3, 4 and 5) promoted a dramatic inhibition of cell growth with IC 50 of about 700 nM (Figure 9c). In contrast, only compound 6 specific for qhras-2 showed some antiproliferative effect in T24 cells. Remarkably, the oligonucleotides with the sequence of hras-1 (H1) or hras-2 (H2) and control oligonucleotide 1450 (not folding into a quadruplex) did not show any antiproliferative effect (H1 and H2 are probably degraded by nucleases). A proliferation assay performed as a function of time showed that the growth inhibition promoted by the G4-decoys did not weaken over a period of 144 h ( Figure S8). According to the postulated mechanism of action, the G4decoys 3, 4 and 5 should repress HRAS transcription by taking MAZ (and Sp1) away from the promoter. Actually, we found that the level of HRAS transcript, determined by real-time PCR 24 h after decoy treatment, was reduced up to 30% of control, in T24 cells treated with the active decoys 3, 4 and 5 (specific for qhras-1). In keeping with proliferation data, the decoys 7 and 8 that are not active (specific for qhras-2) do not repress HRAS transcription (Figure 10a). This result correlates with the finding that decoys 3, 4 and 5 strongly compete with the binding of MAZ to the hras-1 quadruplex, suggesting that these active decoys bind to MAZ (Figure 10b).
Finally, an insight into the killing mechanism caused by the designed G4-decoys was obtained by measuring the activity of caspases 3 and 7 in untreated and decoy-treated T24 cells. Figure 10c shows that decoys with the strongest antiproliferative activity (3, 4 and 5) considerably activate caspases 3/7 (24 h after treatment), suggesting that they promote cell death mediated by apoptosis.

Discussion
The data of this study show that HRAS transcription is activated by Sp1 and MAZ and repressed when the binding sites of these proteins closest to TSS assume a folded G-quadruplex structure. This is a new and compelling piece of evidence pointing to a transcription mechanism which involves a simple on-off switch in a gene regulatory region where unusual G-quadruplexes behave as repressors. This was first proposed by Hurley and co-workers for the CMYC gene [21]. Following this transcription model we designed several decoy oligonucleotides in quadruplex conformation eliciting a potent antiproliferative effect in T24 urinary bladder cancer cells bearing mutant HRAS.
ChIP assays showed that the G-rich elements called hras-1 and hras-2, located upstream of TSS, are bound by the zinc-finger proteins MAZ and Sp1. MAZ interacts with these promoter sequences in a complex way, as it recognizes both duplex and quadruplex conformations of hras-1 and hras-2. The binding to the quadruplexes is catalytic in nature as qhras-1 and qhras-2 bound to MAZ go through a destabilization process that decreases the stability of the DNA-protein complexes (see EMSA). This means that PAGE gives here only an apparent affinity between MAZ and the HRAS quadruplexes, for which we estimated a K D of about 1.5 mM. One should also consider that by using recombinant bacterially expressed proteins the binding data might be underestimated, as recombinant proteins do not undergo the post-translation modifications which occur in eukaryotic cells necessary for optimal binding. For instance, MAZ shows optimal binding to DNA when it is phosphorilated [45]. As a comparison, it has been estimated by a filter binding assay that recombinant . BSA and TR (trypsinogen) at r = 10 do not have any effect on qhras-1 and qhras-2 quadruplexes) (curves 2 and 3, panels a and b); (c) Schematic representation of HRAS transcription regulation proposed by this study. When the critical promoter elements hras-1 and hras-2 are folded, they behave as transcription repressors. This is suggested by the fact that quadruplex-destabilizing point mutations in hras-1 and hras-2 result in a significant increase of transcription, while quadruplex-stabilizing phthalocyanines are found to repress transcription. MAZ, through its capacity to recognize the quadruplex structures, should be recruited to the promoter critical region. The MAZ-DNA interaction destabilizes the G-quadruplexes, the critical hras-1 and hras-2 elements assume a duplex conformation that favors the binding of Sp1 and other proteins of the transcription machinery. These events result in the activation of transcription. doi:10.1371/journal.pone.0024421.g008 nucleolin binds to parallel quadruplexes with a K D between 79 to 367 nM and it binds to mixed parallel antiparallel quadruplexes with a K D of 0.45-2.5 mM [46]. Instead, by surface plasmon resonance assay it has been found that recombinant nucleophosmin binds to the CMYC quadruplex with a K D of 1.9 mM [47].
MAZ was identified as a G-box binding transcription factor for CMYC [48]. Previous studies have shown that MAZ can either activate [35][36][37][49][50][51] or repress [39,52] transcription. Furthermore, in certain genes MAZ regulates transcription together with Sp1, as the two proteins can bind to the same guanine blocks: the consensus sequence for MAZ is G(g/c)GGc/a GGGG(c/a)(g/t) while that of Sp1 is (g/t)GGGCGG(g/a)(g/a)(c/t) [35]. When MAZ or Sp1 was separately knocked-down with specific shRNAs, HRAS transcription dropped respectively to 30% or 20% of control. Conversely, when MAZ and Sp1 were over-expressed, a synergistic effect was observed and HRAS transcription increased 3-fold compared to control.
To explore the role of G4-DNA on HRAS transcription we introduced two quadruplex destabilizing GRT(C) point mutations in sequences hras-1 or hras-2 and found that transcription increased 5-fold compared to control. A similar behavior was previously reported for CMYC and CMYB genes [21,22,39]. Our conclusion is that both HRAS G-quadruplexes behave as a molecular on-off switch that either provides the binding sites to MAZ and Sp1 or subtracts them. This is also supported by the fact that guanidinium phthalocyanines stabilize the HRAS G-quadruplexes and repress luciferase from pHRAS-luc to 20-30% of control, but not from mutant vectors pHRAS-mut1 and pHRAS-mut2. In addition to function as a transcription factor, MAZ is also able to remove the structural blocking of transcription by the quadruplex structures, as indicated by FRET-melting data.
In the light of all these findings, we designed G4-decoys mimicking HRAS G-quadruplexes, which show a strong antiproliferative activity in T24 urinary bladder cancer cells harboring mutant HRAS. We hypothesized that the G4-decoys should take MAZ away from the promoter and inhibit HRAS transcription. We found that in T24 bladder cells (but also in HeLa cells, not shown) the decoys specific for quadruplex qhras-1 (3, 4, and 5) displayed a dramatic inhibitory activity on cell growth, at a concentration as low as 700 nM. Instead, only decoy 6 mimicking quadruplex qhras-2 showed some activity. In keeping with previous observations, our data suggest that quadruplex formation per se is not sufficient to give rise to a bioactivity, as decoys 7 and 8 though forming a stable quadruplex, are not active. It is well known that certain G-rich oligonucleotides show a clear antiproliferative effect in cancer cells which is not due to a true antisense effect, but to their propensity to fold into a Gquadruplex [26]. How these oligonucleotides precisely work, is not yet clear, but Bates and co-workers proposed that the antiproliferative activity of certain G-rich oligonucleotides requires: nuclease resistance; efficient cellular uptake; binding to a specific protein [53]. Our G4-decoys fulfill these requirements as: their compact structure and LNA residues make them resistant to nucleases; they efficiently penetrate cell membranes and internalize in the nucleus when complexed with PEI; they interact with MAZ, an essential protein for HRAS transcription.
That our G4-decoys act through their binding to a nuclear protein (MAZ) is suggested by the fact that when they are delivered without a transfectant agent, they localize in the cytoplasm and are not active. In contrast, when they are delivered with PEI, they reach the nucleus and show a strong antiproliferative activity.
In accord with the proposed mechanism of action, the decoys eliciting the highest inhibition of cell growth (3,4,5) caused in T24 cancer cells a strong decrease of HRAS transcript and activation of caspases 3/7.
In summary, this work shows that: G4-DNA near the transcription start sites of HRAS behaves as a transcription repressor; MAZ and Sp1 bind to G-elements that can fold into quadruplex forming sequences; transcription is activated by MAZ and Sp1; MAZ destabilized the HRAS G-quadruplexes; G4-decoys mimicking the HRAS quadruplexes behave as decoy molecules against MAZ and cause a potent antiproliferative effect in T24 bladder cancer cells bearing mutant HRAS; the decoy strategy could provide a new therapeutic approach to treat bladder cancer.

Materials and Methods
Plasmids and G4-decoys synthesis pHRAS-luc was obtained by standard cloning, while mutant plasmids pHRAS-mut1, pHRAS-mut2 were obtained by site directed mutagenesis with the gene tailor kit (Invitrogen). A 838bp Sac I-Sac I fragment, obtained from pEJ 6.6 plasmid bearing the human HRAS promoter, was cut with Xma I restriction enzyme and the resulting 345 bp fragment was subcloned in pGL3-E1 basic plasmid in Sac I-Xma I upstream of firefly luciferase. In the resulting construct luciferase was driven by wildtype HRAS promoter (pHRAS-luc). By site directed mutagenesis we introduced in pHRAS-luc two point mutations either in hras-1 or hras-2 sequence. The primers used were 59-CGG GGC CGA GGC CGG TGC GGT GCG TGT GCG TGT GC-39 (for) and 59-CCG GCC TCG GCC CCG GCC CTG GCC C-39 (rev). PCR was performed with 3 ng/ml DNA template, 0.1 mM each primer, 0.05 units/ml AccuPrime pfx DNA polymerase (Invitrogen) in 16 AccuPrime pfx reaction mix for 3 min at 95uC, 30 cycles 1 min at 95uC, 30 s at 81uC, and 5 min at 68uC. Bacteria DIH 101 5a were transformed with PCR product, and plasmid DNA was extracted and sequenced (primer pGL3bpr 2 59-CTT  TAT GTT TTT GGC GTC TTC CA-39).
G4-decoys were synthesized on an Expedite TM Nucleic Acid Synthesis System Model 8909 from Applied Biosystems. Purification of oligonucleotides was accomplished using a reverse-phase semipreparative HPLC on Waters Xterra TM MS C 18 column. Oligonucleotide concentrations were determined by UV-absorbance at 260 nm, 90uC and the calculated single-stranded extinction coefficients were based on a nearest neighbour model (extinction coefficient for monomers is 22400 at 260 nm). The DNA samples were resuspended in 90% formamide and 50 mM EDTA, denatured at 90uC and run for 2 h on a denaturing 20% polyacrylamide gel, prepared in TBE and 8 M urea, pre-equilibrated at 55uC in a Sequi-Gen GT Nucleic Acids Electrophoresis Apparatus (Bio-Rad, CA), which was equipped with a thermocouple that allows a precise temperature control. After running, the gel was fixed in a solution containing 10% acetic acid and 10% methanol, dried at 80uC and exposed to Hyperfilm MP (GE Healthcare) for autoradiography.

Cell culture and proliferation assay
HeLa and T24 urinary bladder cancer cells were maintained in exponential growth in Dulbecco's Modified Eagle's Medium (DMEM) containing 100 U/ml penicillin, 100 mg/ml streptomycin, 20 mM L-glutamine and 10% foetal bovine serum (Euroclone, Milan, Italy). For proliferative assays T24 cells were seeded (1250 cells/well) the day before decoy treatment in a 96well plate. The G4-decoys, mixed to jetPEI (Polyplus transfection), were delivered to the cells at increasing concentrations up to 800 nM. 48 h after the first delivery a second dose of oligonucleotides was given to the cells. 72 h after the second delivery, the cell viability was measured by resazurin assays following standard procedures.

Dual luciferase assays
Firefly luciferase activity in cell lysates was measured and normalized for Renilla luciferase activity using the Dual-Glo Luciferase Assay System (Promega) following vendor's instructions. Transfection was performed by mixing each vector 250 ng/ well with control plasmid pRL-CMV expressing Renilla luciferase under control of CMV promoter 10 ng/well using Metafectene transfection reagent (Biontex Laboratories, GmbH) following manufacturer instruction. For cotransfection with pMAZ and/or pSp1 100 ng of pHRAS-luc or mutant, with 10 ng of pRL-CMV, were transfected with either 100 ng of pMAZ/pSp1 or pcDNA3 plasmid (empty vector) as mass for control transfections. Each transfection was performed in triplicate. Luciferase assays were performed 48 h after transfection following instructions. Samples were read with Turner Luminometer and expressed as Relative luminescence (see figure captions).

Recombinant MAZ and Sp1: purification and EMSA
Recombinant MAZ and Sp1 tagged to GST were expressed in Escherichia coli BL21 DE 3 plys using plasmid pGEX-MAZ and  pGEX-Sp1. The bacteria were grown for 1-2 h at 37uC to an optical density at 600 nm between 0.5-2.0 prior to induction with IPTG (0.2 mM final concentration). Cells were allowed to grow overnight at 30uC before harvesting. The cells were centrifuged at 5000 rpm, 4uC, the supernatant removed and the cells washed twice with PBS. The pellet was resuspended in a solution of PBS with 1 mM PMSF, 1 mM DTT and Protease Inhibitor Cocktails (only for GST-tagged proteins) (Sigma-Aldrich). The bacteria were lysed by sonication and centrifuged for 30 min at 4uC, 16000 rpm. Glutathione Sepharose 4B (GE Healthcare) (50% slurry in PBS) was added to the supernatant from the previous step and incubated for 30 min at 4uC while shaking. The mix was centrifuged for 5 min at 500 g and the resin was washed 3 times with PBS. Proteins were eluted from the resin with a buffer composed by 50 mM Tris-HCl pH 8 and 10 mM reduced glutathione.

FRET-melting experiments
FRET melting experiments were performed on a real-time PCR apparatus (CFX 96, BioRad, Hercules, CA), using a 96-well plate filled with 50 ml solutions of dual-labelled qhras-1 and qhras-2, in 50 mM Tris-HCl, pH 7.4, 50 mM KCl (qhras-1) or 100 mM NaCl (qhras-2). The protocol, used for the melting experiments, is the following: (i) equilibration step of 5 min at low temperature (20uC); (ii) stepwise increase of the temperature of 1uC/min for 76 cycles to reach 95uC. All the samples in the wells were melted in 76 min.

Caspase assays
We performed Caspase activity assays using Apo-ONE TM Homogeneous Caspase-3/7 Assay (Promega), according to the manufacturer's protocol using a Microplate Spectrofluorometer System (Molecular Devices, Concorde, Canada). Figure S1 Polymerase stop assays with a wild-type DNA template containing hras-1 or hras-2 and a mutant template in which four GRT point mutations were introduced in sequence hras-2 to abrogate quadruplex formation. Taq polymerase is arrested at the 39 end of hras-2, before the first run of guanines, in keeping with the formation of a quadruplex structure by hras-2 (experimental conditions: 37uC, 140 mM KCl, 50 mM Tris-HCl pH 7.4) When the DNA template is incubated with G4-DNA ligands that stabilize quadruplex DNA, Taq polymerase is completely arrested and only the truncated product is produced. This is observed with porphyrin TMPyP 4 and guanidine phthalocyanines DIGP and Zn-DIGP at r = 4 (r = [ligand]/ [template]). Instead, TMPyP 2 , which does not bind to G4-DNA does not affect the processivity of Taq polymerase. When the experiment is performed with the mutated template, Taq polymerase does not stop at the G-element and full product is observed. A longer truncated product is observed with the mutated template, probably due to a hairpin structure stabilized by CG and GT base pairs. Polymerase stop assays with a DNA template containing hras-1show that Taq polymerase is arrested in the presence of phthalocyanines, indicating that the G-quadruplex formed by hras-1 is less stable than that formed by hras-2.  Figure S8 Proliferation assay with T24 cells untreated and treated with 800 nM G4-decoys 3,4,5 (mimicking qhras-1) and G4decoys 6,7,8 (mimicking qhras-2). Decoy 637 is a random sequence containing one P unit. Two doses, one 48 h after the other, of 800 nM G4-decoys mixed with polyethylenimine have been delivered to T24 cells. Viable cells, measured by a resazurin assay, have been performed at increasing times from 1 st treatment. (TIF)