|Home | About | Journals | Submit | Contact Us | Français|
It has recently been shown that certain oligodeoxynucleotides (ODNs) designed as catalytic DNA molecules (DNAzymes) exhibit potent cytotoxicity independent of RNA-cleavage activity in a number of cell lines. These cytotoxic ODNs all featured a 5′ G-rich sequence and induced cell death by a TLR9-independent mechanism. In this study, we examined the sequence and length dependence of ODNs for cytotoxicity. A G-rich sequence at the 5′ terminus of the molecule was necessary for cytotoxicity and the potency of ODNs with active 5′ sequences was length dependent. Cytotoxicity appeared to be generally independent of 3′ sequence composition, although 3′ sequences totally lacking G-nucleotides were mostly inactive. Nucleolin, elongation factor 1-alpha (eEF1A) and vimentin were identified as binding to a cytotoxic ODN (Dz13) using protein pull-down assays and LC-MS/MS. Although these proteins have previously been described to bind G-rich ODNs, the binding of eEF1A correlated with cytotoxicity, whereas binding of nucleolin and vimentin did not. Quiescent non-proliferating cells were resistant to cytotoxicity, indicating cytotoxicity may be cell cycle dependent. Although the exact mechanism of cytotoxicity remains unknown, marked potency of the longer (25nt) ODNs in particular, indicates the potential of these molecules for treatment of diseases associated with abnormal cell proliferation.
Single-stranded oligodeoxynucleotides (ODNs) and in particular certain G-rich ODNs have been widely reported to have effects on cells ranging from the induction of senescence and aging (1) to inhibition of proliferation (2–4). In some cases, secondary structure formation of ODNs (e.g. G-quadruplexes) facilitates the recognition by cellular protein(s) thus leading to cytotoxicity (5). In other cases, anti-proliferative activity of ODNs is related to the ability of the ODN to bind specific cellular proteins independent of secondary structure (6,7). Known targets of G-rich oligodeoxynucleotides include the human ribosomal protein L7a (8), nucleolin (9,10), elongation factor 1A (7,11), STAT3 (12) and growth factors (13).
DNAzymes are single-stranded DNA molecules that are able to cleave RNA in a site-specific manner (14). The molecules consist of a 5′ and 3′ binding arm and a catalytically active core region. The stability and activity of the DNAzymes in vitro make them a useful tool for gene silencing, and a number of different therapeutic applications have been proposed (14). We recently showed that the cellular effects of some ODNs, originally designed as DNAzymes against the transcription factors c-jun and c-myc, were not due to the cleavage of the target mRNA (15). Indeed, DNAzymes such as Dz13 (c-jun) and Rs6 (c-myc) and their catalytically inactive controls were cytotoxic when transfected into several cultured cell lines at concentrations as low as 10–50nM (15). These cytotoxic ODNs all featured G-rich regions at the 5′ extremity suggesting that this sequence element is required for cytotoxicity. Although this cytotoxicity did not appear to be dependent on the formation of G-quadruplexes and other secondary structures (15), replacement of any of the three contiguous guanosines of Dz13 with 7-deaza-guanosine abolished cytotoxicity. We therefore postulated that the 5′ extremity sequence element is necessary for cytotoxicity and may be involved in mediating specific interactions with cellular proteins.
The aims of the present article were to clarify the sequence elements or ‘motifs’ required for cytotoxicity, to investigate the activity of various cytotoxic ODNs on different cell types, to determine the culture conditions required for cytotoxicity, and finally, to identify proteins that selectively bind to these cytotoxic molecules.
HPLC-purified oligodeoxynucleotides were purchased from Sigma and Trilink Biotechnologies. Stock solutions were made up to a concentration of 50μM in nuclease-free water and stored at −20°C. The oligodeoxynucleotides used are listed in Table 1. The 3′ biotinylated and 5′ Oregon-Green labeled congeners of Dz13 had the identical deoxyribonucleotide sequence to DT1549 (Table 1) in which the 3′-3′T inversion of Dz13 is missing.
Human dermal microvascular endothelial (HMEC-1) cells were maintained in MCDB131 medium (Gibco) containing 10% foetal bovine serum (FBS), 2mM l-glutamine, 1μg/ml hydrocortisone (Sigma) and 10ng/ml epidermal growth factor (EGF; Sigma) with or without 5U/ml penicillin–streptomycin. Rat vascular smooth muscle (RSMC), human embryonic kidney (HEK-293), HCT116 human colon cancer, NIH3T3 mouse fibroblast and human retinal pigmented epithelium (ARPE-19) cells were cultured in DMEM F12 medium (Gibco) supplemented with 10% FBS and 2mM l-glutamine with 5U/ml penicillin–streptomycin.
Cells were seeded in 96-well black MicroClear plates (Greiner) (4 × 103 cells/well) or 6-well plates (1.2 × 105 cells/well) in growth medium containing 5% FBS for HMEC-1, HCT116, HEK-293, NIH3T3 and ARPE-19 cells or 10% FBS for RSMC cells. Cells were transfected 24h after seeding with different concentrations of ODNs in triplicate using FuGene6 (Roche) as previously described (15). Cell survival was assessed 48h post-transfection in the 96-well plate format using the Cell Titer™-Blue cell viability assay (Promega) as previously described (15). Briefly, culture medium was replaced with 100μl OptiMEM and 20μl of Cell-titer™ blue reagent and incubated for 2h at 37°C. Fluorescence was measured at 544EX/590EM using FLUOstar OPTIMA (BMG Labtechnologies). For the cell density experiments, ARPE-19, NIH3T3 or HMEC-1 cells were seeded at densities of 4000 and 50000 cells per well in 96-well plates and transfected with 0–200nM ODN in triplicate. Transfection efficiencies at the 2 cell densities were determined in duplicate in 60mm dishes using a 5′-oregon green ODN (DT1565; Table 1) in a single experiment. The 60mm dishes were seeded with a similar number of cells/cm2 as for the 96-well plates for both seeding densities. For low density transfection, 2.6 × 105 cells/dish were seeded while for high density transfection 3.25 × 106 cells/dish were seeded. Fluorescence of mock and DT1565-transfected cells was measured 48h post-transfection using fluorescence-activated cell sorting (FACS).
Cells were washed twice with PBS and total proteins extracted using MPER (Pierce) or RIPA (150mM NaCl, 0.1% w/v sodium dodecyl sulphate, 1% v/v Nonidet P-40, 0.5% w/v sodium deoxycholate, 50mM Tris-HCl pH 8). Subcellular proteins were extracted using the ProteoExtract kit (Calbiochem). All extractions were performed in the presence of EDTA-free protease inhibitors (Roche). Cellular debris was removed by centrifugation at 10 000g for 20min at 4°C.
Protein pull-down assays were performed using oligodeoxynucleotides as ‘bait’ in order to identify Dz13-binding proteins. Dynabeads-Streptavidin (Dynal) were washed twice in 2× Buffer A (10mM Tris-HCl, 1mM EDTA, 2M NaCl, pH 7.4). Beads (0.5ml) were resuspended in 1ml 2× Buffer A to a final concentration of 5μg/μl beads. For in vitro assays, beads were incubated in an equal volume of 2μM 3′ biotinylated congeners of Dz13 or Dz13scr (DT1309h and DT1310c respectively, Table 1) made up in Buffer B (20mM HEPES, 100mM KCl, 0.2mM EDTA, 0.01% v/v NP-40, 10% v/v glycerol, pH 7.5) and the mixture incubated at room temperature for 10min with gentle mixing. The beads were washed with 3 × 1ml Buffer B prior to addition of protein. Protein extract (up to 1mg) was incubated with the bead–DNA mixture for 10min at room temperature with shaking. For the majority of protein pull-downs, the beads were washed with 20 × 1ml Buffer B following incubation with protein extract and proteins were eluted by two washes with 1μM Dz13 or Dz13scr at room temperature. In one experiment, the elutions were performed with 10μM Dz13 or Dz13scr. In between each elution, the beads were washed with 3 × 1ml Buffer B.
To test whether the same proteins eluted with other ODNs of interest an experiment was undertaken in which the bead preparation, protein incubation and washing steps were performed as described above. Following washing of the beads, non-specific proteins were eluted by 2 × 5min incubation with 1μM non-biotinylated Dz13scr and the beads washed again with 3 × 1ml Buffer B. This bead preparation was split into six identical samples and proteins eluted by 2 × 5min incubations with 1μM ODN (Dz13, NT36, DT1501, Rs6, DT1595 or GT). The resulting sample supernatants were collected and analysed and each of the bead samples was again washed with 3 × 1ml Buffer B and proteins eluted with 1μM Dz13. This experiment was performed twice.
For in situ protein pull-down assays, cells were transfected with 100nM biotinylated congeners of Dz13 or Dz13scr (i.e. DT1309h and DT1310c) for 24h. Beads were washed three times with 1ml Buffer B and resuspended at 5μg/μl prior to incubation for 10min at room temperature with protein extracts prepared from these cells using MPER. Following incubation, beads were washed with 20 × 1ml Buffer B and resuspended in 20μl Buffer B. Washing and elution steps were then performed as described above.
Fractions generated from pull-down assays were concentrated in a Centricon 10000 Mwt cutoff 0.5ml centrifugal device (Millipore) at 13 800g for 70min at 15°C. Proteins (from cell extracts or concentrated pull down fractions) were denatured at 70°C for 10min in 4× Loading Buffer (Invitrogen), loaded onto NuPAGE 4–12% bis-tris acrylamide gels (Invitrogen) and electrophoresed at 140V in MOPS-SDS running buffer. Following electrophoresis proteins were either silver-stained as described by Rabilloud et al. (16) or transferred to a PVDF membrane at 30V for 2h. Membranes were blocked in 3% BSA in TBST (10mM Tris pH 8, 30mM NaCl, 0.05% v/v Tween-20) for 1h at room temperature. Primary antibodies were incubated at concentrations recommended by the manufacturers for 1h at room temperature or overnight at 4°C in 5% skim milk-TBST. Membranes were washed 3 × 5min in TBST and incubated for 1h at room temperature with horseradish peroxidase conjugated antibodies (1:2000 in 5% skim milk-TBST; DakoCytomation). Membranes were washed 3 × 5min in TBST and visualized by chemiluminescence using ECL (Amersham) and Hypersensitive film (Amersham). The following commercially available antibodies were used: EF1A (CBK-KK1; Upstate Biotechnology), vimentin (V9; Sigma) and STAT3 (F-2; Santa-Cruz Biotechnology).
Total protein was digested by incubating 100 μl of sample, 25μl of 10mM NH4HCO3 and 1 μg trypsin at 37°C for 16h. The digested peptides were loaded onto a C18 precolumn (500µm × 2mm, Michrom Bioresources) using H2O:CH3CN (98:2, 0.1% formic acid, buffer A) at 20µl/min. After a 10min wash, the pre-column was switched (Switchos) in-line with an analytical column containing C18 RP silica (PEPMAP, 75µm × 15cm, LC-Packings) or a fritless C18 column (75µm × ~12cm). Peptides were eluted using a linear gradient of buffer A to H2O:CH3CN (40:60, 0.1% formic acid-buffer B) at 200nl/min over 60min. The column was connected via a fused silica capillary to a low volume tee (Upchurch Scientific) where high voltage (2300V) was applied and a nano electrospray needle (New Objective) or fritless column outlet was positioned ~1cm from the orifice of an API QStar Pulsar i hybrid tandem mass spectrometer (Applied Biosystems). The QStar was operated in an information-dependent acquisition mode. A TOF-MS survey scan was acquired (m/z 350–1700, 0.5s) and the two largest precursors (counts >10) sequentially selected by Q1 for MS/MS analysis (m/z 50–2000, 2.5s). A processing script generated data suitable for submission to the database search programs. CID spectra were analysed using Mascot MS/MS ion search (Matrix Science). The criteria were: trypsin digestion allowing up to 1 missed cleavage, oxidation of methionine, peptide tolerance of 1.0Da and MS/MS tolerance of 0.8Da. A Mascot score >18 indicated identity.
Variants of Dz13 and other selected DNAzymes, which reduce proliferation in HMEC-1 cells but do not act through cleavage of the RNA substrate, were designed in order to elucidate the sequence requirements for cytotoxicity. First, a set of oligodeoxynucleotides containing regions of the Dz13 sequence was designed in order to identify any requirement for an active motif. ODNs corresponding to the 5′ sequence of the first 9 bases of Dz13 (DT1530), this 5′sequence plus the 10–23 catalytic core (15bp) in the reverse orientation (DT1531) and the 5′sequence plus the catalytic core in the correct orientation (DT1532) were tested. All of these were substantially less active than Dz13 (data not shown), suggesting that either length or some other sequence requirement had not been met. Similar results were obtained with the corresponding ODN based on the 5′ Rs6 sequence (DT1533-5, Table 1; Figure S1a).
In order to elucidate the requirements of the tail for cytotoxicity, Dz13 analogues of the same length but with modified 3′ tail sequences were tested (Table 1). These included both CG dinucleotides in the 3′ tail changed to GC (DT1536), the first 16 bases of Dz13 plus scrambled tail (DT1537), a pool of ODNs all consisting of the first 10 bases of Dz13 followed by a random mix of nucleotides in tail sequence (DT1538) and the first 10 bases of Dz13 followed by a polyA tail (DT1539). The polyA-tailed Dz13 analogue (DT1539) exhibited no cytotoxicity in HMEC-1 cells whereas DT1536 demonstrated intermediate cytotoxicity (Figure 1A). Both the scrambled tail analogue (designed to remove the tail hairpin structure) and the mixture of random-tailed oligodeoxynucleotides had activity that was indistinguishable from Dz13, indicating minimal sequence requirements for the 3′ tail in these 33-mer oligodeoxynucleotides.
To elucidate the positional requirement for a 5′ G-rich element, three pools of ODNs featuring the 5′ motif from Dz13 (first 10 bases) were constructed with either the 5′ motif in the middle of a random sequence pool, at the 3′ end or at the 3′ end in reverse order (DT1600-1602, Table 1). None were significantly cytotoxic, although some activity was noted for the 3′ reverse motif ODN (Figure 1B). This suggests that the motif may also be recognized in the 3′ to 5′ direction or that part of the activity relates to the rest of the molecule. In any case, this experiment confirmed that the optimal placement for the G-rich sequence of Dz13 is at the 5′ end.
Given the lack of activity of the shorter variants (DT1530–DT1532), the next experiment sought to clarify the length requirement for the observed cytotoxicity. This was tested by using ODN pools containing the first 10 bases of Dz13 and 5, 10, 15, 20 or 25 random bases (DT1544-48, Table 1). Cytotoxicity was concentration and length dependent with only the 35-mer ODN pool (DT1548) matching the potency of Dz13 (Figure 1C), thus confirming the role of the tail length. A similar result was obtained with DT1572-6, which are based on a slightly different 5′ sequence (data not shown).
Given previous results indicating that potency was dependent on a relatively unbiased nucleotide composition of the 3′ tail, a new 35-mer ODN was tested in which the random tail was composed only of A, C and T bases (DT1552). This ODN was essentially devoid of cytotoxic activity over the concentration range tested (data not shown). These results indicate that the tail sequence has no particular sequence requirements but that potency is dependent on a relatively unbiased nucleotide representation, including some G content.
A set of 35-mer oligodeoxynucleotides were designed and synthesized in order to identify the minimal 5′ motif required for activity. This set comprised pools of oligodeoxynucleotides for which the first 10 bases of the Dz13 5′ G-rich sequence (plus a 25 random base tail; DT1548) was progressively reduced (DT1553–DT1554) to the first 4 bases of this sequence plus a 31 base tail (DT1555). Only DT1553 (first 8 + 27 random) and DT1548 (first 10 + 25 random) retained full activity (Figure 1D). The corresponding experiment with the Rs6 sequence (DT1557–60) produced gradated cytotoxicity profiles with activity increasing with increased retention of the 5’ Rs6 sequence (Figure S1b).
The Dz13 5′ G-rich sequence, when extended with tail sequences of TAG and TAGGAT repeat elements (DT1570 and DT1571, respectively) had reduced activity relative to Dz13 (data not shown). These tail sequences were specifically free of C to prevent the formation of GC clamps and possibly more complex secondary structures.
As reported earlier (15), other published G-rich oligodeoxynucleotides are capable of inducing cytotoxicity under the same conditions in HMEC-1 cells including the 20AG ODN described by Cogoi et al. (17). We examined whether extension of this sequence in a manner analogous to the Dz13 experiments would influence cytotoxic activity. Consistent with the Dz13 results, 16- (DT1591), 24- (DT1592), 28- (DT1593) and 32-mer (DT1594) variants of the 20-mer 20AG yielded length-dependent cytotoxic activities with the longest two molecules having similar profiles (data not shown).
Given their potent cytotoxic activity against proliferating cells, Dz13 and the other cytotoxic oligodeoxynucleotides could be of use in several disease states. Dz13 has been reported as being active in several preclinical models of disease including ocular angiogenesis (18), vascular intimal proliferation (19) and cancer (20), which are all diseases where inappropriate proliferation is present. To investigate any possible differential activity relative to non-proliferating cells, we examined the activity in two additional cell lines that display contact inhibition, namely ARPE-19 and NIH-3T3. In particular, ARPE-19 cells form differentiated, polarized monolayers in vitro (21) similar to that present in normal retinal epithelium. As shown in Figure 2A, proliferating ARPE-19 cells were sensitive to Dz13-induced cytotoxicity with an IC50 between 50 and 100 nM, whereas Dz13scr was inactive. Dz13 cytotoxicity in ARPE-19 cells was abrogated over that concentration range by increasing the cell seeding density to 50 000 cells/well, at which seeding density the cells grew to a uniform dense monolayer. Similar results were obtained for the murine fibroblast 3T3 cells, which are also contact-inhibited and for HMEC-1 cells, which form multilayered sheets at maximal confluence (data not shown). To rule out reduced transfection as the cause for this effect, repeat experiments were scaled up and transfection efficiency was determined using a fluorescent 5′-OregonGreen-488 congener of Dz13 (DT1565; Table 1), which was complexed with Fugene6 and transfected at a final concentration of 100 nM. DT1565 had reduced cytotoxicity compared to Dz13 (data not shown), which reduced the confounding effects of toxicity on analysis, and the transfection efficiency as determined by FACS analysis showed higher transfection at the higher cell density (Figure 2B). These preliminary experiments indicate selective cytotoxicity of Dz13 to proliferating cells as opposed to those that are contact-inhibited and non-proliferating.
In order to identify whether proteins bind the cytotoxic ODNs and if binding correlated with cytotoxic potency, protein pull-downs were performed using 3′ biotinylated congeners of Dz13 and Dz13scr (DT1309h and DT1310c, respectively; Table 1) as ‘bait’. In HMEC-1 cytotoxicity assays, these 3′-biotinylated ODNs had very similar cytotoxic properties to their untagged counterparts (data not shown). For the pull-down experiments, the biotinylated ODN was coupled to magnetic streptavidin beads and a crossover design employed in which beads coupled to the biotinylated ODN of interest were incubated with HMEC-1 protein, washed with the opposite ODN and eluted with the non-biotinylated bait ODN. That is, when the beads were prepared with the biotinylated congener of Dz13scr and loaded with cell protein extracts, they were subsequently washed with a Dz13 solution and proteins eluted with a Dz13scr solution. For the Dz13-coupled beads, a large number of non-specific proteins were washed off with Dz13scr and a number of strong staining bands were obtained by competitive elution with Dz13, the major one being a ~51kDa protein (Figure 3A). Elution with the non-cytotoxic Dz13scr yielded a large number of bands including a predominant band at ~39kDa, however unlike Dz13, none of these were specifically eluted by Dz13scr. Silver staining revealed a lack of protein remaining bound to unconjugated beads following extensive washing (data not shown).
In order to generate enough material to identify proteins eluting from the beads, the experiment was scaled up 5-fold and 1mg of protein was loaded onto beads coupled to the biotinylated ODNs. In these experiments, the number of washes was increased to 30 while the elution volume was kept constant. Following concentration of the eluted fractions, the proteins were digested with trypsin and analysed by LC-MS/MS, with liquid chromatography used to separate the reasonably complex mixture of peptides prior to mass spectrometric analysis. A number of proteins eluting with both Dz13 and Dz13scr were identified (Table 2). Proteins identified in the Dz13-eluting fraction (from Dz13-coupled beads) included microtubule-associated protein 4 (MAP4), nucleolin, vimentin, elongation factor 1A (eEF1A), plasminogen activator inhibitor 1 (PAI-1) RNA-binding protein, glyceraldehyde 3-phosphate dehydrogenase (GAPDH) and signal recognition particle 14kDa. Proteins identified in the Dz13scr-eluting fraction (from Dz13-coupled beads) included annexin II, GAPDH, heterogeneous nuclear ribonuclear proteins A1, Lamin A/C isoform 1 and nucleolin. The 51kDa band was competitively eluted by Dz13 in three independent experiments and eEF1A was identified in these fractions by LC-MS/MS each time. Overall, a total of 10 peptides from this protein were identified that together spanned most of the protein sequence (Table 2). Western blotting of pull-down proteins with an antibody directed to eEF1A confirmed that it was the major band identified by silver staining at 51kDa and that the protein bound to both Dz13 and Dz13scr beads but was only eluted by Dz13 (Figure 3B). The LC-MS/MS identification of vimentin was also confirmed by western blots in which vimentin was shown to bind to Dz-13 coupled beads as well as to Dz13scr-coupled beads (data not shown). Elution of vimentin from the beads with Dz13 or Dz13scr was not detected by western (data not shown). This may reflect the fact that more protein (5-fold) was loaded onto the beads for LC-MS/MS analysis. Vimentin distribution within the various cell compartments was assessed using cell fractionation and western blotting 24h post-transfection with Dz13 and Dz13scr. Vimentin was mostly localized to the cytoskeletal and nuclear fractions and this distribution did not appear to be modified 6 or 24h following transfection with 200 nM Dz13 (data not shown).
In order to further define the role of eEF1A in the cytotoxic mechanism of Dz13 and to determine whether cytotoxicity correlates with eEF1A binding, pull-down assays were performed where active and inactive oligodeoxynucleotides were used to elute proteins bound to Dz13. DT1501 (rat c-jun Dz13 homologue) and Dz13scr are non-cytotoxic to HMEC-1 cells under the conditions used here (15). Cytotoxicity experiments were performed with a GT-rich oligodeoxynucleotide (DT1605; Table 1) previously reported as binding to eEF1A and inducing cytotoxicity in human T-lymphoblasts (11). However, this GT ODN was not cytotoxic to HMEC-1 cells at concentrations of up to 200nM (data not shown). The oligodeoxynucleotides DT1595 (AGGG repeat), Rs6 (DT1556) and NT36 (DT1577; ATM-inducing) all resulted in significant cytotoxicity to HMEC-1 cells (15). Elution of eEF1A from Dz13-coupled beads occurred for all the cytotoxic oligodeoxynucleotides tested (Figure 4). In contrast, elution of eEF1A using non-cytotoxic oligodeoxynucleotides was very poor (Figure 4). Nevertheless, when Dz13 was subsequently used as an eluting solution, eEF1A was strongly released from beads that had been first eluted with the non-cytotoxic oligodeoxynucleotides Dz13scr and DT1501, and to a lesser extent with DT1605. eEF1A was only weakly released by Dz13 from beads where eEF1A had been strongly eluted using the cytotoxic oligodeoxynucleotides DT1595, DT1577 and DT1556 (Figure 4). This demonstrates that eEF1A initially captured by the biotinylated Dz13 beads and only weakly released by the non-cytotoxic ODN could still be released with Dz13 solutions.
To determine the cellular localization of the eluting eEF1A, the pull-down procedure was performed on the cytoplasmic, nuclear, membrane and cytoskeletal fractions of HMEC-1 cells. Whereas eEF1A from the cytoplasmic, nuclear and membrane/organelle fractions bound to Dz13 beads, the major source of eEF1A eluted by Dz13 was from the cytoplasm (Figure 5). Pull-downs were also performed with total protein lysates from a number of other cell lines including RSMC, HEK-293 and HCT-116, to which Dz13 is cytotoxic (15). Predominant elution by Dz13 of the 51kDa band was observed in all cases and this band was confirmed as being eEF1A by western blotting (data not shown). The results indicate that eEF1A binding also occurs in other cell lines in which Dz13 is cytotoxic. When examined directly by western blotting, the abundance of eEF1A was not dependent on the cell type and was not affected by treatment with Dz13 for 24h (data not shown).
The amount of eEF1A eluted with Dz13 was concentration dependent in that a concentration of 10μM Dz13 eluted more eEF1A than a concentration of 1μM (data not shown). Nevertheless, in all experiments elution was only partial, and residual eEF1A was found associated with beads post elution (Figures 3 and and4).4). This suggests that the interaction might not be canonical or that eEF1A undergoes a conformational change upon binding that inhibits its release. In addition to in vitro binding experiments, cells were transfected with 100 nM biotinylated Dz13 and Dz13scr and pull-downs performed directly on cell extracts. There was binding of eEF1A to both Dz13 and Dz13scr beads (data not shown), confirming the presence of the interaction of both oligodeoxynucleotides when the binding occurs within the cells. Furthermore, despite the high abundance of eEF1A, a possible interaction between Dz13 and eEF1A directly in cells was demonstrated using fluorescence colocalization experiments (Figure S2).
Whilst there are a number of reports describing the cytotoxic nature of certain oligodeoxynucleotides, in particular G-rich oligodeoxynucleotides (see Introduction Section), there is little known about the motif or sequence requirements for cytotoxicity. The current study attempted to define further the exact requirements for cytotoxicity and identify proteins involved in binding to these oligodeoxynucleotides, thereby unravelling the mechanisms involved in the eventual cytotoxicity of the molecules.
We determined that the 5′ sequence, the core or the tail of Dz13 alone are insufficient to regenerate the cytotoxic activity of Dz13 against HMEC-1 cells, thereby indicating a length and/or sequence requirement. We used the novel strategy of testing pools of oligodeoxynucleotides comprised of random sequences along with the 5′ G-rich elements of Dz13 and Rs6 in various position and length contexts to demonstrate that the G-rich element needs to be present at the 5′ extremity and be followed by a tail component that is preferably composed of mixed nucleotides. A comparison of the active oligodeoxynucleotides, in combination with the 5′-motif reduction experiment, leads us to conclude that the required 5′ sequence is G-rich, composed of 6–9nt with at least four consecutive purines. The presence of a triple G motif (G-G-G) provides for the greatest potency and the 5′ extremity of the triple G motif needs to be positioned no more than 3nt from the 5′ end of the ODN. We have previously shown by CD that Dz13 does not assemble into stable secondary structures in vitro (15). Nevertheless, substitution of any of the guanosines in the triple G motif of Dz13 abrogated its cytotoxic activity (15). This suggests that the hydrogen bonding activity of these guanosines does not lead to G-quadruplex formation, but is nevertheless required for cytotoxicity, perhaps by enhancing binding of the oligodeoxynucleotides to intracellular proteins. The tail sequence requirements were intriguing in that most sequences with some degree of ‘complexity’ supported cytotoxic activity, but some individual tail sequences, in particular those free of guanosines, had reduced activity. It is possible that the less complex tail sequences such as the polydA tail are being sequestered by polydA-binding proteins thereby abrogating cytotoxicity. Wu et al. (22) have previously demonstrated that tail sequence and complexity encourages the multimeric aggregation of the oligodeoxynucleotides, thereby activating TLR9 and cell uptake. However, we have previously demonstrated a need for transfection and ruled out endosomal TLR engagement in the mechanism of action of Dz13 and Rs6 against HMEC-1 cells (15). Therefore, although multimeric assembly remains a possibility, it is unlikely to relate to uptake and TLR engagement.
We previously compared the activity of Dz13, a prototypic G-rich oligodeoxynucleotide, with that of other published sequences including NT36, an ATM-inducing oligodeoxynucleotide (23) and 20AG, an oligodeoxynucleotide originally designed to be a triplex-forming inhibitor of K-ras (17). Both of these oligodeoxynucleotides have purine tracts in the 5′ sequence and NT36 is of comparable length to the active ODNs described in the present study. As shown for Dz13, the potency of the shorter 20AG ODN was length dependent. That is, the cytotoxicity of the AG molecule increased as the length of the molecule increased.
ODNs have been proposed to exert their cytotoxicity through specific interaction with cellular proteins such as nucleolin (24) and eEF1A (7). Nucleolin was identified as one of the proteins that bound to Dz13 and Dz13scr. The absence of an effect on HMEC-1 proliferation by the nucleolin-binding oligodeoxynucleotide GRO29A provided further evidence for the lack of direct involvement of nucleolin in Dz13-mediated cytotoxicity. Likewise, a representative from a class of GT-rich ODNs, which reportedly binds eEF1A and is cytotoxic to human T-lymphoblasts (11), was not cytotoxic to HMEC-1 cells. Furthermore in our experiments, the GT oligodeoxynucleotide did not displace eEF1A from Dz13 capture beads as potently as Dz13 or the other cytotoxic ODNs tested, indicating that it has less binding affinity to the eEF1A found in this cell line. The ODNs described in the present study therefore represent a novel class of potently cytotoxic molecules.
The ability of several of the proteins, including eEF1A to bind competitively to the cytotoxic oligodeoxynucleotides and not be eluted by the non-cytotoxic ODNs in vitro suggests that the ODNs may competitively bind to proteins within the cells and that this is mechanistically relevant to the cytotoxicity of the molecule. eEF1A is an extremely abundant protein with a multitude of roles including protein synthesis, stress-sensing, apoptosis and cellular proliferation (25,26). Scaggiante et al. (7) recently reported a correlation between eEF1A binding and cytotoxicity of G-rich oligodeoxynucleotides that is independent of secondary structure formation. They also reported that some non-cytotoxic G-rich oligodeoxynucleotides bind eEF1A to a small degree in the absence of competitor (7). The fact that both Dz13 and Dz13scr were able to bind eEF1A in vitro and in cells indicates that some of the captured eEF1A is also binding via a non-specific mechanism in HMEC-1 cells. However, the selective release of eEF1A from Dz13 capture-beads with those ODNs that are cytotoxic suggests that there is a relationship between eEF1A binding and cytotoxicity.
eEF1A has been proposed to act as a signalling molecule in proliferating cells through the binding of other nucleic-acid-binding proteins such as ZPR1 (27) followed by translocation to the nucleus. Once inside the nucleus eEF1A has the capacity to bind DNA, RNA and RNA polymerase and therefore potentially plays a role in transcriptional regulation. Binding of eEF1A to Dz13 was observed in nuclear, cytoplasmic, membrane/organelle and cytoskeletal fractions; cytoplasmic eEF1A eluted most strongly from Dz13 beads. This could reflect a greater concentration of DNA-binding proteins in the nuclear fraction and consequently a greater number of proteins eluting in general or that nuclear eEF1A has a higher affinity for Dz13. It is possible that the higher affinity cytotoxic oligodeoxynucleotides displace eEF1A from nuclear or cytoplasmic protein complexes such as the nuclear cytotoxicity-related complex (CRC; 7), thereby inhibiting the normal functioning of eEF1A in the cell, resulting in a reduction in cellular proliferation and eventually cell death.
Vimentin was also identified in the pull-down assay. Vimentin has been shown to bind to G-rich ODNs, causing the translocation of the vimentin–DNA complex to the nucleus (28). However, this was not evident in HMEC-1 cells transfected with Dz13. Collectively the protein pull-down experiments indicate that binding of cytotoxic ODNs to eEF1A and other proteins is potentially a key event in the mechanism of action of the molecules.
In conclusion, we have determined that ODNs such as Dz13 and Rs6 which were initially designed as catalytic DNAzymes, belong to a group of cytotoxic G-rich ODNs with novel sequence and length requirements. Although the mechanism is unclear, it is likely to be mediated through binding to eEF1A. These molecules are preferentially active against proliferating cells and therefore constitute part of a new class of potentially useful molecules for the treatment or investigation of diseases characterized by abnormal cell proliferation.
Supplementary Data are available at NAR Online.
The authors are grateful to Mark Raftery at the Bioanalytical Mass Spectrometry Facility, UNSW, Sydney for performing the Mass Spectrometry analyses. Funding to pay the Open Access publication charges for this article was provided by Johnson and Johnson Research.
Conflict of interest statement. None declared.