Suppression of FOXM1 Transcriptional Activities via a Single-Stranded DNA Aptamer Generated by SELEX

The transcription factor FOXM1 binds to its consensus sequence at promoters through its DNA binding domain (DBD) and activates proliferation-associated genes. The aberrant overexpression of FOXM1 correlates with tumorigenesis and progression of many cancers. Inhibiting FOXM1 transcriptional activities is proposed as a potential therapeutic strategy for cancer treatment. In this study, we obtained a FOXM1-specific single stranded DNA aptamer (FOXM1 Apt) by SELEX with a recombinant FOXM1 DBD protein as the target of selection. The binding of FOXM1 Apt to FOXM1 proteins were confirmed with electrophoretic mobility shift assays (EMSAs) and fluorescence polarization (FP) assays. Phosphorthioate-modified FOXM1 Apt (M-FOXM1 Apt) bound to FOXM1 as wild type FOXM1 Apt, and co-localized with FOXM1 in nucleus. M-FOXM1-Apt abolished the binding of FOXM1 on its consensus binding sites and suppressed FOXM1 transcriptional activities. Compared with the RNA interference of FOXM1 in cancer cells, M-FOXM1 Apt repressed cell proliferation and the expression of FOXM1 target genes without changing FOXM1 levels. Our results suggest that the obtained FOXM1 Apt could be used as a probe for FOXM1 detection and an inhibitor of FOXM1 transcriptional functions in cancer cells at the same time, providing a potential reagent for cancer diagnosis and treatment in the future.

The phosphorthioate-modified FOXM1 Apt (M-FOXM1 Apt) maintained the binding activities to FOXM1 protein. Wild-type DNA aptamers are too susceptible to nuclease-mediated degradation to be useful for in vivo applications 20 . Among multiple approaches for modifying oligonucleotides, we chose internucleotide phosphorothioate linkage modification 22 for FOXM1 Apt. The internucleotide linkages of FOXM1 Apt were modified by the sulfur substitution in the phosphate ester backbone to form phosphorothioate linkages ( Fig. 2A). As the consequence of phosphorothioate modification, the molecular mass of the modified FOXM1 Apt (M-FOXM1 Apt) was elevated in the measurement of mass spectrometry (Fig. S4A). The M-FOXM1 Apt possessed a similar spatial structure as the FOXM1 Apt, evidenced by the Circular Dichroism (CD) measurement in the parallel structure formation for the FOXM1 Apt and M-FOXM1 Apt, although M-FOXM1 Apt showed enhanced negative peaks at 211 nm and 248 nm and an elevated positive peak at 277 nm (Fig. S4B). M-FOXM1 Apt obtained an enhanced stability in the DNase I treatment compared to FOXM1 Apt (Fig. S4C). The EMSA  TCT CTA  CGT CCG GTT GCG CTT TCC TTT  17  42.5%   C2  TCC CAG TCA CGA CGT TGT AAA  ACG ACG GCC AGT GAA TTG TAG  8  20%   C3  CTG AAA GCG CAA CCG GAC GTA  TAC AAG CGG GGG TCC TTG ATT  6  15%   C4  TTA CTG GAG CCC CGC ATC ACT  GCG ATC CGG GTG CGG TTC CCT  6  15% Other sequences 7.5% Table 1. The sequences of selected colonies (n = 40) from the sixth round of SELEX.
Scientific RepoRts | 7:45377 | DOI: 10.1038/srep45377  experiments were performed with ROX-labeled M-FOXM1 Apt (M-FOXM1 Apt-F) and FOXM1 DBD or FOXM1 full-length protein, and confirmed that M-FOXM1 Apt maintained the binding activities to FOXM1 protein ( Fig. 2B,C). The FP assays determined that the K D of M-FOXM1 Apt-F to FOXM1 full-length protein was 63.86 ± 6.6 nM (Fig. 2D), implicating that the phosphorthioate modification in FOXM1 Apt even improved its binding affinity to FOXM1 proteins compared to wild-type FOXM1 Apt (FOXM1 Apt to FOXM1 protein: K D = 172.1 ± 26.55 nM, see above). Therefore, we used M-FOXM1 Apt-F for further experiments.
M-FOXM1 Apt bound to FOXM1 protein specifically in cancer cells. There are more than 50 members identified in FOX transcription factor family and all of them contain a highly conserved DBD. Because FOXM1 Apt was selected by using FOXM1 DBD as the target molecule in SELEX, we had to determine whether the binding of FOXM1 Apt to FOXM1 was highly specific. We tested whether or not M-FOXM1 Apt was able to bind to other FOX proteins such as FOXA2, which is the first cloned mammalian FOX transcription factor containing the typical conserved DBD 23 . A His-FOXA2 full-length protein was purified (Fig. S5) and used for EMSAs with M-FOXM1 Apt. In the same experimental condition, M-FOXM1 Apt bound to FOXM1 but not FOXA2 (Fig. 3A). To further test that FOXM1 Apt could recognize FOXM1 specifically from a mixture of cellular proteins, we first prepared whole-cell lysates from the FOXM1-overexpressing cells in which FOXM1 expression plasmid vectors were transfected (Fig. S6A). As a control, the lysate of the FOXA2-overexpressing cells was also prepared (Fig. S6B). Biotinylated M-FOXM1 Apt (M-FOXM1 Apt-biotin) was generated and incubated with FOXM1-overexpressing or FOXA2-overexpressing cell lysates. Streptavidin agarose beads were added to the reactions to pull down M-FOXM1 Apt-biotin/protein complexes, which were detected by Western Blotting only in the reaction of M-FOXM1 Apt-biotin plus FOXM1-overexpressing lysates (Fig. 3B), suggesting a specific interaction between FOXM1 Apt and FOXM1 proteins in the complicated conditions. We then prepared lysates from MDA-MB-436 breast cancer cells endogenously expressing both FOXM1 and FOXA2, and repeated the experiments described above to confirm the specific interaction between FOXM1 Apt and FOXM1 proteins in the cancer cells (Fig. 3C). The specificity of the binding of FOXM1 Apt to FOXM1 proteins was further determined by Pull-down experiments with MDA-MB-436 cell lysates in which the endogenous FOXM1 expression was knocked down by the transfection of pFOXM1-shRNA vectors to the cells (Fig. 3C). Furthermore, to confirm the recognition of FOXM1 Apt to FOXM1 proteins in cells, we transfected HEK293T cells with pCMV-FOXM1-GFP plasmids expressing FOXM1-GFP fusion proteins, followed by the transfection of M-FOXM1 Apt-F one day later. Observed under a fluorescent Con-focal microscope, FOXM1-GFP fusion proteins localized in nuclei of cells (Fig. 3D). Compared to Control Apt that stayed in the cytosol of cells, FOXM1 Apt was found to co-localize with FOXM1 proteins in nuclei of cells (Fig. 3D), suggesting that FOXM1 Apt could specifically bind to FOXM1 in cells. As a control experiment, HEK293T cells were also transfected with pCMV-FOXA2-GFP plasmids expressing FOXA2-GFP fusion proteins, followed by the transfection of M-FOXM1 Apt-F. We found that the FOXA2-GFP fusion proteins localized in nuclei but FOXM1 Apt stayed in cytoplasm, confirming no interactions between FOXM1 Apt and FOXA2 (Fig. 3D).

FOXM1 Apt suppressed the transcriptional activities of FOXM1. FOXM1 is linked to various types
of human malignancies through stimulating the expression of cancer cell-related target genes 3 . The transcriptional activation of FOXM1′ s target genes requires FOXM1 binding to its target gene promoters 15 . To test whether FOXM1 Apt could affect FOXM1 transcriptional activities, we first performed EMSAs and determined that M-FOXM1 Apt abolished the binding ability of FOXM1 to its putative DNA binding sequence dramatically in vitro, at a dose-dependent manner (Fig. 4A). The wild type FOXM1 Apt also showed similar inhibitory effects on the FOXM1-DNA binding ability as M-FOXM1 Apt (Fig. S7). Based on previous studies 15,24 , the cotransfection of FOXM1 expression vector in cells could stimulate the promoter activities of the luciferase reporter plasmids, whose luciferase expression was controlled by either an artificial 6x FOXM1 binding sequence-containing promoter or an endogenous − 2.3 kb promoter of Cdc25B (FOXM1′ s target gene) (Fig. 4B). The addition of FOXM1 Apt or M-FOXM1 Apt to the cotransfection inhibited the FOXM1-mediated stimulation on the promoter activities significantly while M-FOXM1 Apt appeared better effects than the wild type FOXM1 Apt (Fig. 4B), demonstrating that FOXM1 Apt was able to prevent FOXM1 transcriptional activities in vivo. The control aptamer M-Control Apt did not affect FOXM1 transcriptional activities in the cells (Fig. 4B).  (Fig. 5E,F). On the other hand, the treatment of M-FOXM1 Apt at the high dosage in 293T-FOXM1 cells also caused a dramatic down-regulation of the expression of Cdc25B and Cyclin B1 without changing the levels of FOXM1 mRNA and protein in the cells (Fig. 5E,F), implicating that the inhibitory effects of FOXM1 Apt on FOXM1 resulted from disturbing FOXM1 transcriptional activation function but not from affecting FOXM1 expression. As a control, we also tested whether the proliferation of wild type 293T cells with a very low endogenous FOXM1 expression was affected by FOXM1 Apt. The M-FOXM1 Apt treatment at the dosage (500 nM) that abolished the proliferation of 293T-FOXM1 cells did not show dramatic effects on 293T cells (Fig. S8). In addition, we treated MDA-MB-436 breast cancer cells with M-FOXM1 Apt and measured the levels of FOXM1, Cdc25B, and Cyclin B1. The treatment of M-FOXM1 Apt caused a dramatic down-regulation of the expression of Cdc25B and Cyclin B1 without obvious changing of FOXM1 levels in MDA-MB-436 cells (Fig. 5G,H). Taken together, these results determined that FOXM1 Apt could suppress the cellular proliferation of cancer cells by abolishing the functions of FOXM1 in vivo.

Discussion
Aptamers have been suggested as diagnostic or therapeutics reagents because of their high affinity and specificity towards selected targets and abolishing the functions of their targets without obvious side-effects 20 . In 2005, FDA approved Macugen, a RNA aptamer targeting VEGF, as the first aptamer therapeutic for age-related macular degeneration or diabetic retinopathy 21 . A number of aptamer-based therapeutics are currently undergoing clinical trials, including therapeutic aptamers for cancer treatment such as AS1411 targeting nucleolin 26 and NOX-A12 targeting stroma cell-derived factor-1 (SDF-1) 27 . An aptamer specific for modulating the function of intracellular transcription factor NF-κ B has been developed and shows effective inhibition of NF-κ B in vitro and in vivo 28,29 , suppressing non-small cell lung cancer resistance to Doxorubicin 30 . An aptamer that inhibits the function of the E2F family of transcription factors has also been obtained and can be potentially used for preventing tumor development 31 . In this study, we obtained the FOXM1-specific single stranded DNA aptamer FOXM1 Apt to target FOXM1 DBD and full-length protein specifically. FOXM1 Apt was confirmed to prevent the binding of FOXM1 to its consensus binding sites in promoters and consequently suppress FOXM1 transcriptional activities, resulting in the downregulation of the expression of FOXM1 target genes and the inhibition of cancer cell proliferation. FOXM1 is ubiquitously expressed in proliferating and regenerating mammalian cells 1 and is a key cell cycle regulator of both the transition from G1 to S phase and the progression to mitosis by regulating transcription of cell cycle genes 2 . Loss of FOXM1 expression causes diminished DNA replication 32 , mitotic spindle defects 33 , and mitotic catastrophe 34 . Furthermore, we and others have shown that FOXM1 is involved in contra-acting stresses induced by cytotoxic or genotoxic signals, such as oxidative stress or DNA damage, and enhances the drug resistance of cancer cells 35 . Moreover, we have characterized that FOXM1 plays an essential role in maintenance of cell stemness and its expression is absent from differentiated cells 36,37 . These observations suggest that FOXM1 is associated with cancer initiation and progression through its critical roles in cell proliferation, drug resistance, and Scientific RepoRts | 7:45377 | DOI: 10.1038/srep45377 malignant transformation of undifferentiated cells. This notion is apparently supported by the fact that FOXM1 is highly expressed in various types of human malignancies, such as lung cancer 38 , prostate cancer 39 , basal cell carcinomas 40 , hepatocellular carcinoma 41 , gastric cancer 42 , and squamous cell carcinoma 43 , demonstrating FOXM1 as a diagnostic marker for monitoring the initiation and progression of multiple human cancers. The inactivation of FOXM1 leads to inhibition of progression and/or invasion of these cancers, suggesting FOXM1 as an attractive target for the development of novel anti-cancer therapies. Small chemical compounds and FOXM1-specific RNA interference adenovirus vectors has been developed to inhibit FOXM1 functions for cancer treatment [9][10][11][12][13][14] . It is necessary to explore more FOXM1-targeting therapeutic strategies that would finally reach the stage of clinical usage. Therefore, FOXM1 Apt obtained from this study provided a potential reagent to detect and repress FOXM1 at the same time, good for cancer diagnosis and treatment in the future.
We confirmed the specificity of FOXM1 Apt binding to FOXM1 by showing that the aptamer was not able to bind to FOXA2, which also contains the typical conserved DBD of FOX transcription factors. In addition, we found that FOXM1 Apt could not interact with FOXP2, another member of FOX transcription factors, in the pull-down experiments with FOXM1 Apt and MDA-MB-436 cell lysates (data not shown). To understand how FOXM1 Apt recognizes FOXM1, we predicted the secondary structure of the aptamer that comprises a primarily stem-loop-loop structure (Fig. S9A). We generated three truncated aptamers, FOXM1 Apt (1-15 nt), FOXM1 Apt (11-34 nt) and FOXM1 Apt (28-42 nt), based on the structural motifs of FOXM1 Apt (Fig. S9B). Even though the CD measurement could not tell obvious differences in spatial structures among these truncated aptamers (Fig. S9C), EMSA experiments with FOXM1 full-length protein confirmed that only the truncated FOXM1 Apt (11-34 nt) possessed FOXM1-binding activities but much weaker than that of wild-type FOXM1 Apt (Fig. S9D). This finding suggested that the 11-34 nt sequence of FOXM1 Apt was the core binding sequence mediating the interaction between FOXM1 Apt and FOXM1 protein. Aptamers often bind to functionally important parts of target proteins and affect protein-protein or protein-nucleic acid interaction of the target proteins 20 . It is still a challenge to predict the binding structure of an aptamer recognizing its target protein according to its nucleotide sequence, resulting in that pharmacokinetic and other systemic properties of the aptamer are variable and often hard to be predicted in practice. With the solved crystal structure of FOXM1 DBD and FOXM1 protein 15 , our study provided a suitable pair of molecules FOXM1 Apt and FOXM1 protein to learn the detail mechanisms how FOXM1 Apt interacting with FOXM1. Further research will concentrate on the analysis of FOXM1 Apt-FOXM1 DBD crystal structure, identify the key nucleotide(s) and conformations of FOXM1 Apt to mediate the interaction between the two molecules, and hopefully provide clues to predict or even design aptamers for target proteins in the future.
In general, native ssDNA aptamer molecules are too susceptible to nuclease-mediated degradation to be useful for most therapeutic applications. When unmodified aptamers enter the cell or are administered in vivo, they are rapidly degraded by nucleases 20 . Therefore, chemical modifications of the oligonucleotides are often required to increase resistance to degradation by nucleases. Several strategies have been developed to increase the stability of aptamers without compromising the binding affinity and specificity towards their targets. The nucleotides of aptamers can be partially or completely substituted with one or more modifications, including 2′ -amino pyrimidines 44,45 , 2′ -fluoro pyrimidines 46 , and 2′ -O-methyl ribose purines and pyrimidines 47 . Internucleotide linkages can also be modified to phosphorothioate linkages 22 and high molecular mass polyethylene glycol (PEG) can be conjugated to the 5′ -terminus 48 . In this study, we chose to modify the internucleotide linkages of FOXM1 Apt to phosphorothioate linkages. The modified M-FOXM1 Apt possessed a similar spatial structure as FOXM1 Apt and obtained an enhanced resistance to nuclease-mediated degradation. M-FOXM1 Apt imparted greater affinities to FOXM1 protein compared to wild-type FOXM1 Apt, consistent to other findings that sulfur substitution of the phosphate ester backbone in aptamers often results in enhanced binding to their target proteins 22 . Whether M-FOXM1 Apt shows better performance to detect and/or repress FOXM1 in animal models than FOXM1 Apt needs be further evaluated.
In conclusion, the obtained FOXM1 Apt from this study could selectively bind to FOXM1 protein, co-localize with FOXM1 in nucleus after entering cells, and potentially abolish the binding abilities of FOXM1 to its target gene promoters. By this way, the transfection of FOXM1 Apt results in the downregulation of the expression of FOXM1 target genes in cells and consequently inhibits cancer cell proliferation. FOXM1 Apt is a novel and specific FOXM1 inhibitor, providing a potential reagent for cancer diagnosis and therapy in the future.

Expression and Purification of Recombinant Proteins. Certain plasmids were transformed into
Rosetta/DE3 E. Coli cells and positive colonies were confirmed by colony PCR screening. Cells were grown at 37 °C in LB media until reaching an optical density (OD 600 ) of 0.8 and protein expression was induced by addition of 1 mM IPTG for additional 6 hr at 37 °C. The GST protein and GST-FOXM1 DBD protein were purified by Glutathione Sepharose TM 4B (GE, USA) following the manufacturer's instructions. The FOXM1 FL protein and FOXA2 FL protein were purified by Ni-Sepharose TM 6 Fast Flow (GE, USA) following the manufacturer's instructions.
Aptamer Selection by SELEX. Single-stranded DNA (ssDNA) aptamers recognizing FOXM1 DBD were identified using the PCR-based SELEX method 17 . Briefly, a synthetic ssDNA library, containing random 42 nucleotides flanked by 20 nt at each end (5′ -AG CAA TGG TAC GGT ACT TCC-42N-GTG CCA CGC TAC TTT GCT AA-3′ ), was synthesized by Sangon (Shanghai) Co., Ltd, China. The ssDNA library (100 μ M, 20 μ l, ~2.14 × 10 15 sequences) was mixed with 500 μ l binding buffer (20 mM Hepes, 120 mM NaCl, 5 mM KCl, 1 mM CaCl 2 , 1 mM MgCl 2 , pH 7.35) and heated at 95 °C for 5 min and snap-cooled on ice. GST-FOXM1 DBD protein (10 μ g) and Glutathione SepharoseTM 4B beads (50 μ l, GE, USA) were added and mixed thoroughly and incubated on ice for 1 hr in a rotary shaker. The mixture was centrifuged at 8000 rpm for 5 min at 4 °C and the pellets was washed twice with binding buffer. The pellets were resuspended with 500 μ l binding buffer, heated at 95 °C for 10 min, centrifuged at 13,100 g for 5 min. The collected supernatant was mixed with GST protein (10 μ g) and Glutathione SepharoseTM 4B beads (50 μ l, GE, USA) and incubated on ice for 1 hr in a rotary shaker. The mixture was centrifuged at 8000 rpm for 5 min at 4 °C to remove the pellet containing non-specific binding ssDNAs. The supernatant was collected and amplified by PCR with the primers (forward: 5′ -ROX-A GCA ATG GTA CGG TAC TTC C-3′ and reverse: 5′ -Biotin-T TAG CAA AGT AGC GTG GCA C-3′ ). The PCR products were incubated with streptavidin agarose beads (GE, USA) for 30 min at room temperature and washed 3 times with PBS. The biotin-labeled reverse strand of DNA products were eluted by adding NaOH (1.5 M) and considered as the selected ssDNA pool for the next round of SELEX. After the 6 rounds of selection, the enriched PCR products were cloned with TOPO TA cloning kit (Invitrogen, USA) and the colonies were selected for DNA sequencing. Statistical Analysis. The experimental data are expressed as the mean ± standard deviation. The comparison among groups was performed using one-way analysis of variance and Dunnett-t tests were used for comparison between experimental groups and control groups. P < 0.05 was considered to indicate a statistically significant difference.