DNA Triplex Formation Selectively Inhibits Granulocyte-Macrophage Colony-stimulating Factor Gene Expression in Human T Cells*

Granulocyte-macrophage colony-stimulating factor (GM-CSF) is a hemopoietic growth factor that is expressed in activated T cells, fibroblasts, macrophages, and endothelial cells. Although GM-CSF does not appear to be essential for normal hemopoiesis, overexpression of GM-CSF has been implicated in the pathogenesis of some diseases such as myeloid leukemia and chronic inflammation. An NF- (cid:107) B/Rel binding site within the GM-CSF promoter, termed the (cid:107) B element appears to be important for controlling expression in reporter gene assays in response to a number of stimuli in T cells. We investigated oligonucleotide-directed triple helix for- mation across this regulatory sequence as a potential tool to inhibit GM-CSF gene transcription. A 15-base oligonucleotide, GM3, was targeted to a purine-rich region in the GM-CSF proximal promoter, which overlaps the (cid:107) B element. Gel mobility shift assays and DNase I footprinting demonstrated that GM3 formed a sequence- specific collinear triplex with its double-stranded DNA target. Triplex formation by GM3 blocked recombinant and nuclear NF- (cid:107) B proteins binding to the GM-CSF element. GM3 also caused selective inhibition of the human T-cell lymphotrophic virus-1 Tax transactivator-in-duced luciferase activity from a reporter construct driven by the GM-CSF promoter in Jurkat T cells. Fi-nally, GM3 greatly reduced the concentration of endog- enous GM-CSF mRNA induced by different stimuli in Jurkat T cells but did not affect interleukin 3 mRNA levels in the same

Granulocyte-macrophage colony-stimulating factor is a cytokine involved in the control of survival, proliferation, and differentiation of hemopoietic progenitor cells as well as in the functional activation of mature myeloid cells (reviewed in Ref. 2). GM-CSF 1 expression occurs in different cell types, including fibroblasts and endothelial cells in response to proinflammatory agents such as tumor necrosis factor and interleukin-1 (IL-1), monocytes in response to endotoxin, and activated T cells (reviewed in Ref. 3).
Despite the clinical application of GM-CSF as an activator of hemopoietic cell production after bone marrow transplantation, evidence has accumulated to suggest that GM-CSF may play a role in the pathogenesis of several disease conditions, in particular myeloid leukemia and chronic inflammation (4 -7). GM-CSF was found to be abnormally expressed in atopic patients and in patients with rheumatoid arthritis (4,5). In terms of leukemia, GM-CSF has been shown to be produced by some leukemic clones in an autocrine fashion in vitro, and the growth of the cells in vitro depends on the presence of GM-CSF (8). Recent in vivo experiments have also shown that GM-CSF increased the blast cell counts in 9 of 12 acute myeloid leukemia patients (9). Taken together these data suggest that it would be of significant therapeutic importance to be able to modulate GM-CSF gene expression.
Synthetic oligonucleotides are attracting increasing interest as tools for specific manipulation of gene expression and have potential therapeutic application. Several strategies that employ short nucleic acid fragments targeted to different intracellular components are currently under investigation with variable success (reviewed in Ref. 10). The "antigene" (11) strategy is based on inhibition of transcription of the selected gene by oligonucleotide-directed intermolecular DNA triple helix formation either through prevention of regulatory proteins binding to the control regions (12) or by acting as transcriptional repressors (13). Triplex formation may have certain advantages over the other oligonucleotide-mediated strategies, as there are fewer target molecules per cell and it is expected to be more selective. In vitro experiments demonstrated that the formation of such sequence-specific structures can inhibit DNA replication (14) or block transcription factors binding to the gene of interest and thereby directly affect transcription initiation (15). Triplex-forming oligonucleotides (TFOs) were also found to repress transcription in intact cells, resulting in highly specific inhibition of the synthesis of promoter reporter gene targets (16 -18).
Predominantly polypurine:polypyrimidine (Pur:Pyr) se-quences have been shown to be suitable targets for intermolecular triple helices. Different structures of base triplets have been described, forming two groups of DNA triple helices. Both have the third strand binding in the major groove of the duplex DNA through Hoogsteen (Y*RY motif) or reverse Hoogsteen (R*RY motif) base pairing. The best studied Y*RY type, formed by the addition of a polypyrimidine third strand, consists of C ϩ *G:C and T*A:T base triads with the third strand parallel to the purine-rich strand of the underlying duplex (19,20). These triplexes are stable only under acidic pH, but that condition can be partly overcome by methylation of the third strand cytosines (22). Stable R*RY type triplexes have been shown to form at physiological pH in the presence of divalent cations, where oligonucleotide interaction with double-stranded DNA involves G*G:C, A*A:T, or T*A:T triads with an antiparallel orientation of the third strand (21). TFO binding to the DNA duplex target appears to be highly sequence-specific, as triplex-forming oligos have been shown to recognize unique sites in the yeast chromosome (23). The human GM-CSF gene contains an enhancer located approximately 3 kilobases from the transcription start site (24) and a proximal promoter spanning at least the first 120 bp from the transcription initiation site (3,25). Activation of the human GM-CSF promoter in T cells appears to be tightly regulated by induction and binding of multiple transcription factors including NF-B/Rel, AP-1, and NFAT proteins (25)(26)(27). The GM-CSF promoter contains two binding sites for the NF-B/Rel family of transcription factors, the CK-1 element, which binds only specific members of the family such as p65 (Rel-A) and c-Rel (Rel), and the B element that binds mainly classical NF-B p65/p50 (NF-B1) complexes (26,27). The B element appears to play a central role in the response of the GM-CSF promoter to T cell receptor type signals as well as HTLV-1 Tax activation in T-cells (25,27) Here we attempted to modulate transcription from the GM-CSF gene by DNA triple helix formation within the promoter region. A triplex-forming oligodeoxynucleotide was targeted to a 15-bp purine-rich sequence that overlaps the B-element. Sequence specificity of the third strand coupling to its duplex target was investigated, as well as the ability of triplex formation to inhibit recombinant and nuclear NF-B proteins binding to the GM-CSF promoter. We also evaluated the effect of the triplex-forming oligomer on luciferase activity from GM-CSF promoter-driven reporter constructs and on endogenous GM-CSF mRNA and immunoreactive protein levels in activated Jukart T-cells and show that triplex formation can selectively suppress human GM-CSF gene expression.

MATERIALS AND METHODS
Oligonucleotides and Plasmids-Oligodeoxyribonucleotides were synthesized by the ␤-cyanoethylphosphoramidite method on an Applied Biosystems synthesizer (Applied Biosystems, Foster City, CA) using solid support for nonmodified oligomers or solid support coupled with 3Ј-amino-modifier (Glen Research, Sterling, VA) for oligomers with a propanolamine group at the 3Ј terminus. Deprotected oligonucleotides were purified by preparative 12-20% gel electrophoresis. Concentrations were determined by absorption measurements at 260 nm using molar extinction coefficients. Oligonucleotide sequences are presented in Fig. 1b.
Triplex Gel Shift Assays-Double-stranded oligodeoxyribonucleotides were end-labeled with [␥-32 P]ATP and T4 polynucleotide kinase. TFOs (specific and control) were heated at 65°C for 5 min to prevent self-association and placed on ice. Triplex formation was assessed by incubating for 2 h at 20°C 0.2 ng of double-stranded probe with the indicated concentrations of TFOs (see legend for Fig. 2) in 89 mM Tris borate (pH 8.0), 20 mM MgCl 2 , and 10% sucrose followed by electrophoresis through a 12% native acrylamide gel for 16 h at 4°C. Both gel and running buffer contained 89 mM Tris borate (pH 8.0) and 20 mM MgCl 2 .
DNase I Footprinting Analysis-The probe for DNase I footprinting was prepared by digesting the pGMluc plasmid with SalI, dephosphorylated with alkaline phosphatase, and radiolabeling with [␥-33 P]ATP using T4 polynucleotide kinase. After a second digest with MscI the 489-bp probe was purified on a 2% agarose gel. Triplex formation reactions were performed under the same conditions as described for the triplex band shift assays and were then digested with DNase I at a concentration of 1 unit/l for 30 s. Digestion was stopped by adding 20 mM EDTA and 0.2% SDS. A ϩ G chemical sequencing of the probe was carried out according to the Maxam-Gilbert method (29). Samples were then electrophoresed through 6% denaturing gels, dried, and visualized using a Molecular Dynamics PhosphorImager.
Protein Binding Assay-Triplex formation was performed as described above using 33 P-end-labeled double-stranded fragments of the GM-CSF gene promoter and immunoglobulin gene enhancer (see Fig.  1). After initial incubation for triplex formation of approximately 0.2 ng of the probe with specific or control single-stranded oligonucleotides as described above recombinant p65 (27) or nuclear extracts from transfected and stimulated Jurkat T cells (prepared according to Ref. 30) were added, and samples were incubated for a further 30 min at 20°C. The composition of the protein binding buffer was 25 mM Tris (pH 7.6), 0.5 mM dithiothreitol, 0.5 mM EDTA, 60 mM NaCl, and 1 g of poly(dI-dC) for the recombinant p65. Ten mM MgCl 2 was also present in the binding reaction for the nuclear proteins. Electrophoresis was carried out on 5% native polyacrylamide gels containing 0.25 ϫ TBE buffer at 190 V for 90 min for samples with the recombinant p65 and 0.5 ϫ TGE buffer at 150 V for 3 h using nuclear extracts.
Cell Culture and Transfection Procedure-The Jurkat T cell line was cultured in RPMI medium containing 10% fetal calf serum and supplemented with L-glutamine and penicillin/gentamycin antibiotics. Cells were transfected by electroporation at 270 V and a capacitance of 960 microfarads using a Bio-Rad gene pulser. 5 ϫ 10 6 cells in 350 l of RPMI containing 20% fetal calf serum were used per transfection. Ten g of HTLV-1 tax expression vector and 5 g of pGMluc or pIL3luc reporter plasmid was used per transfection. Twenty-four hours posttransfection, TFO GM3 or control oligonucleotide GMc was added (where indicated in the figure legends) to the culture medium. Cells were cultured for an additional 2 h, stimulated at a final concentration of 20 ng/ml PMA and 2 M calcium ionophore, and harvested 8 h later.
Luciferase Reporter Assay and RNA Analysis-Cells used for the luciferase reporter assay were activated and exposed to oligonucleotides as described above and lysed by three freeze/thaw rounds 8 h after stimulation with PMA/Ca 2ϩ ionophore. Three g of total protein from the cell lysates was used per luciferase assay according to a procedure previously described (31). Light emission was measured using a scintillation counter with the coincidence circuit switched off.
Jurkat T cells were transfected, treated with oligonucleotides (where indicated), and activated with PMA/Ca 2ϩ ionophore, as described above. Total RNA was purified using guanidine isothiocyanate 8 h after cell stimulation. The method for quantitating specific mRNA by RNase protection and the vectors for antisense RNA synthesis have been previously described (32). Ten g of total RNA was used for each assay with 33 P-labeled antisense probe. Specific bands were quantified using a Molecular Dynamics PhosphorImager and normalized against glyceraldehyde-3-phosphate dehydrogenase as internal control.
Assessment of the GM-CSF Protein Levels in the Culture Medium-The concentration of GM-CSF protein in the supernatants of the cell samples, used for RNase protection assay was quantified by enzymelinked immunosorbent assay (Quantikine R & O Systems Inc., Minneapolis, MN) according to the manufacturer's instructions. The detection limit for this assay was 1.5 pg/ml.

RESULTS
Oligonucleotide Design and Triplex Formation on the GM-CSF Promoter-It was previously shown that triple helix formation by oligonucleotides can inhibit DNA-protein interaction at specific regions by only overlapping and not spanning the entire protein binding site (16). The 15-bp sequence from Ϫ80 to Ϫ65, which overlaps the B element in the GM-CSF gene proximal promoter (Fig. 1a) was targeted to form an intermo-lecular triple helix. This sequence contains only one "mismatch" in an otherwise Pur:Pyr tract and is 72% G-rich, properties that have been demonstrated to be essential for stable triplex formation. A triplex-forming oligonucleotide GM3 (Fig.  1b) was designed according to the rules previously described in numerous studies (17,(33)(34)(35), i.e. T opposite A:T pairs and G opposite G:C pairs with its orientation antiparallel to the purine strand of the underlying duplex target. G was also put opposite the interrupting C:G pair, as it has been shown to be tolerated in this triplex motif (17,33). As a control, an oligomer GMc of random sequence of G and T residues with the same length as GM3 was used (Fig. 1b).
The ability of these oligonucleotides to form triplexes was analyzed by gel mobility shift assay and DNase I footprinting. A radiolabeled 30-bp fragment of the GM-CSF promoter from Ϫ90 to Ϫ60 (GM, Fig. 1b), which encompassed the TFO target and overlapping B element, was incubated with increasing concentrations of the GM3 oligonucleotide in the presence of Mg 2ϩ and subjected to electrophoresis on native polyacrylamide gels also containing Mg 2ϩ followed by autoradiography. Triplex formation was concentration-dependent ( Fig. 2) but occurred only at the higher concentrations of specific TFO and with apparent K D Ͼ10 M. However no triplex formation was observed when control GMc oligonucleotide was incubated with GM even at the highest concentration (Fig. 2). GM3 was also incubated with a control double-stranded oligonucleotide containing the B element from the immunoglobulin gene (Ig), and no triplex formation was detected (Fig. 2).
To further confirm the selectivity and specificity of the GM3 TFO, a DNase I footprinting assay was performed. Fig. 3 shows that triplex formation by GM3 protected from DNase I digestion, in a concentration-dependent manner, only the targeted sequence of 15 bp within a 489-bp SalI/MscI fragment of the GM-CSF gene promoter. The DNase I footprint occurred at 4 M concentration of the specific TFO, and full protection was obtained at 16 M (Fig. 3, lanes 3 and 4), which is consistent with the affinity of GM3 for its target duplex observed in the gel shift assay. The nonspecific oligomer GMc did not allow any protection of the target from DNase I-mediated degradation even at a concentration of 20 M. These data suggest that oligonucleotide GM3 forms a stable and sequence-selective triple helix with its double-stranded target from the GM-CSF gene proximal promoter.
Inhibition of Transcription Factor Binding by Triplexforming Oligonucleotides-The effect of triplex formation by GM3 on the ability of NF-B/Rel proteins to bind to the B element within the 30-bp fragment of the GM-CSF promoter was examined by gel mobility shift assay. The radiolabeled double-stranded oligonucleotide was incubated in the absence or presence of increasing concentration of GM3 followed by incubation with recombinant p65 protein or nuclear extracts from HTLV-1 Tax-expressing Jurkat T cells. The formation of the recombinant p65-DNA complex was inhibited by TFO GM3 in a concentration-dependent manner (Fig. 4a). The GMc un- related oligomer that could not form a triplex did not have any significant effect on p65 interaction with the GM-CSF promoter fragment (Fig. 4a, lane 9). p65 protein binding to the Ig control probe was not significantly affected by the GM3-specific oligonucleotide, used at the highest concentration (Fig. 4a, lane 10) Jurkat T cells transfected with the HTLV-1 tax expression plasmid contain NF-B/Rel complexes that can bind to the GM-CSF B element (Fig. 4b) (27). The specific NF-B complex was determined by competition with an excess of cold Ig oligomer containing the B consensus sequence (Fig. 4b, lane  3). Several constitutive DNA-protein complexes were also detected under the conditions used (Fig. 4b). Preincubation of the probe with increasing concentrations of GM3 inhibited the appearance of the B-DNA specific complex in a concentration-dependent manner (Fig. 4b, lanes 5-8), whereas no effect was detected with control GMc oligomer at the highest concentration (Fig. 4b, lane 9).
These results indicate that triplex formation specifically prevents binding of NF-B/Rel transcription factors (either recombinant or nuclear proteins) to the B element of the GM-CSF gene promoter. It is interesting to note that constitutive proteins binding to the GM fragment used as a probe in the gel retardation assay were also blocked by the related TFO but not by the control oligonucleotide (Fig. 4b). This fact could be viewed as additional evidence that the observed inhibition of NF-B protein binding to the DNA probe was due to triplex formation at the targeted site and not TFO-NF-B protein interactions.
Inhibition of Transcription from GM-CSF Promoter Reporter Constructs-It was shown previously (17, 36) that short singlestranded oligonucleotides, when applied to the culture medium, can internalize into cells and equilibrate the concentration inside and outside the cell membrane within a few hours. This phenomenon is not universal and varies depending on the cell type used in the experiment possibly as a consequence of different mechanisms of oligonucleotide uptake (reviewed in Ref. 37). When Jurkat T cells were incubated with an intrinsically 32 P-labeled oligonucleotide we also found that the intracellular concentration of the oligomer reached approximately 50% of that in the culture medium after 2 h and approximately 80% after 5 h of incubation (data not shown).
To investigate the effect of triplex formation on GM-CSF gene transcription in cell culture, 3Ј-amino-derivatized (specific and control) oligomers were directly added to the culture medium at the indicated concentrations (Fig. 5). 3Ј-modification with a primary amino group has been shown to significantly increase the resistance of natural phosphodiester oligode- oxynucleotides to nuclease degradation (17). Two luciferase reporter constructs were used to analyze the ability of the TFO to inhibit inducible transcriptional activity in Jurkat T cells. A 620-bp GM-CSF promoter reporter vector (pGMluc) containing the triplex target site and a plasmid with 670 bp of the IL-3 gene promoter (pIL3luc), which does not have a sequence capable of forming a DNA triplex with GM3, were transfected into Jurkat T cells together with an HTLV-1 tax expression vector. Both IL-3 and GM-CSF promoter activities are induced by HTLV-1 Tax (38). After 24 h, TFO GM3 and control oligomer GMc were added to the culture medium at the concentrations shown in Fig. 5, and cells were cultured for an additional 8 h and then harvested.
The specific triplex-forming oligonucleotide GM3 demonstrated a 65% inhibition of the reporter gene activity from pGMluc at a concentration of 1 M, whereas no effect was detected when cells transfected with pGMluc were treated with GMc at the same concentration of 1 M (Fig. 5). Also no significant differences in luciferase activity induced by the IL-3 promoter were detected when cells transfected with pIL3luc were exposed to medium alone, GM3, or GMc at the highest concentrations (Fig. 5) Triplex Formation Inhibits Transcription from the Endogenous GM-CSF Gene-We used RNase protection assays to determine whether the GM3 oligonucleotide, targeted to form a triple helix within the GM-CSF promoter, would selectively suppress the induction of mRNA from the endogenous GM-CSF gene. The effect on the response of the GM-CSF gene to PMA/ Ca 2ϩ ionophore treatment and HTLV-1 Tax expression in combination with PMA/Ca 2ϩ ionophore was measured. Jurkat T cells transiently transfected with HTLV-1 tax or untransfected were preincubated for 2 h with triplex-forming oligonucleotide GM3 or control oligomer GMc at the indicated concentrations (Fig. 6). After stimulation for a further 7 h with PMA/Ca 2ϩ ionophore, total RNA was extracted and assayed. GM-CSF mRNA levels in PMA/Ca 2ϩ ionophore-stimulated cells were significantly reduced when cells were treated with GM3 at 2.4 M concentration compared with samples exposed to the culture medium only or to the control GMc oligomer at the highest concentration. (Fig. 6). GM3 was also effective when the expression of GM-CSF mRNA was synergistically activated by HTLV-1 Tax and PMA/Ca 2ϩ ionophore 2 (Fig. 6). GM3 caused up to 70% reduction of the GM-CSF mRNA level induced by PMA/Ca 2ϩ ionophore and up to 50% decrease after activation with HTLV-1 Tax and PMA/Ca 2ϩ ionophore together (Fig. 6). Concentrations of immunoreactive GM-CSF in the supernatants of cells stimulated with both stimuli and treated with GM3, but not GMc, were also decreased by an average 45% with the highest concentration of GM3 (from an average 560 Ϯ 13 pg/ml (mean Ϯ S.D.) to an average 320 Ϯ 16 pg/ml in two experiments) in accordance with mRNA levels in the cells. In the same cell samples, concentrations of IL-3 mRNA were not affected either by TFO GM3 or by control GMc oligonucleotides regardless of the type of activation used (Fig. 6).
These results lead to the conclusion that a 3Ј-amino derivatized oligodeoxyribonucleotide, directed to form a triple helix with a 15-bp target sequence overlapping the NF-B binding site in the GM-CSF promoter, significantly inhibits transcription from the endogenous gene in human T cells, and this inhibition appears to be oligomer-and gene-specific. DISCUSSION We show here that a triplex-forming oligonucleotide can repress transcription from a reporter construct containing the human GM-CSF gene promoter and more importantly can reduce expression from the endogenous GM-CSF gene in Jurkat T cells. Moreover, we show that GM-CSF protein concentrations in cell supernatants were reduced in parallel with the GM-CSF mRNA levels. Although there are many reports of triplex formation inhibiting transcription from reporter gene constructs, only a few reports describe the inhibition of human endogenous gene expression in cell culture (17, 40 -42).
The TFO binds to a sequence overlapping a B element and blocks in vitro protein binding to this site. This is consistent with the previous mutation analysis of the GM-CSF promoter in our laboratory showing that this B element is required for the response of the promoter to PMA/Ca 2ϩ ionophore activation (25) and to transactivation by the HTLV-1 Tax protein (27) in reporter gene assays. The data presented here confirm these results for the endogenous gene. Activation of the GM-CSF gene by PMA/Ca 2ϩ ionophore is also mediated via NFAT/AP-1 binding sites in both the promoter and enhancer (25,43). HTLV-1 Tax, on the other hand, requires the B element as well as the adjacent CK-1 element for promoter induction but does not activate NFAT/AP-1 sites (27). 2 The fact that the GM3 TFO inhibits the response of the endogenous gene to both PMA/Ca 2ϩ ionophore and Tax/PMA/Ca 2ϩ ionophore costimulation suggests a central role for this B element in GM-CSF gene transcriptional activation. The effect of the TFO GM3 appears to be specific for the GM-CSF gene and has no effect on a reporter construct containing the IL-3 promoter or on mRNA levels from the endogenous IL-3 gene. The gene for IL-3 responds to the same stimuli as GM-CSF in T cells (38,44). This implies that the TFO is not blocking NF-B or other transcription factor function in general but is acting selectively through triplex formation on the GM-CSF promoter.
It is also worth noting that the GM3-targeted DNA sequence covers a G/C element (see Fig. 1) that represents a putative Sp1 binding site. The Sp1 protein has been implicated in the transcriptional activation of numerous genes, and the Sp1 site was found to be important in the expression of murine GM-CSF (45). The G/C region of the human gene has a 2-bp difference with the functional site in mouse and appears to be a weak Sp1 binding site. 3 However, blocking of Sp1 binding to the GM-CSF promoter by the TFO may also contribute to the observed inhibitory effect of GM3 on the expression of the gene.
We have also shown that the GM3 oligonucleotide caused a slightly greater decrease in the GM-CSF mRNA levels in Jurkat T cells induced by PMA/Ca 2ϩ ionophore than in cells activated by both PMA/Ca 2ϩ ionophore and the HTLV-1 transactivator Tax. Costimulation of GM-CSF expression by these two stimuli has been shown to have a strong synergistic effect leading to high levels of gene transcription. 2 The overexpression of Tax significantly increases binding of NF-B/Rel transcription factors to the B element in the GM-CSF promoter compared with that seen with PMA/Ca 2ϩ ionophore, 2 which may make it more difficult for the TFO to compete with proteins for binding to the target site. In addition, NFAT/AP-1 sites and the CK-1 element all contribute to this high level of synergism, and it may by more difficult to block gene activation via a single site. In order to achieve complete inhibition of GM-CSF transcription it may be necessary to use TFOs that block not only the B element but also NFAT/AP-1 sites in the gene enhancer or promoter since we have shown that all of these elements are required for full activation of the gene (25,27,43). 2 It may be also possible to increase the affinity of TFO binding to the target site on the GM-CSF promoter. This may be achieved by using, for instance, oligonucleotides coupled to an intercalating agent (46,47) or an oligomer that is linkersubstituted at the position of the "mismatch" in the duplex target (48).
The observed relatively high K D (Ͼ10 M) of the GM3 oligonucleotide binding to its duplex target could be due to the one mismatch in the otherwise homopurine-homopyrimidine sequence of the double-stranded DNA. Pyrimidine interruptions in the purine tract of the targeted DNA sequence have been shown to be significantly unfavorable for triplex formation because of loss of Hoogsteen hydrogen bonding (21) and can increase the value of the relative binding affinity of the third strand by approximately 1 order of magnitude (49). The central location of the interruption in the purine sequence of the duplex, as seen in the target on the GM-CSF promoter, has been also found to cause greater distortion in the stability of the DNA triplex than the same mismatch situated at the end of the targeted site (49,50). Another factor that may contribute to the relatively low affinity of GM3 for its target DNA is the stretch of eight consecutive guanines in the TFO, which makes this oligonucleotide highly susceptible to self-aggregation through the formation of G-tetrades (51,52). It also should be noted that some degradation of the triplex may occur during the course of electrophoresis, thus increasing the value of the K D observed.
In a number of previous studies it has been noted that the extracellular concentration of the TFO required for inhibition of target gene transcription in intact cells was in the range of 10 -25 M (17, 18, 40), which was 10 -50 times the dissociation constant for triplex formation in vitro. The data presented in this study are not consistent with these observations. A significant decrease of GM-CSF mRNA levels by GM3 was achieved at a concentration of 2.4 M of the TFO in the culture medium, which was below the K D value observed in the binding gel shift assays. This may be attributed to the intrinsic properties of the GM3 oligonucleotide sequence. But this finding may also be explained by the possible changes in the chromatin structure upon activation of the inducible GM-CSF gene that could make the target duplex in the gene promoter more easily accessible for the exogenous oligonucleotide and thus promote more efficient triplex formation by GM3. 50 -70% of cases of acute myeloid leukemia have been shown to depend on GM-CSF for proliferation both in vitro (8) and in vivo (9) and also for engraftment into SCID mice (53). A specific antagonist of GM-CSF/receptor interactions can block the growth of acute myeloid leukemia cells in vitro (54), as can antibodies against GM-CSF (55) and its receptor (56). It has been suggested, in fact, that paracrine production of GM-CSF is responsible for progression of the disease (57). Furthermore, 20 -30% of acute myeloid leukemia cases constitutively produce and respond to GM-CSF (39,58,59), suggesting that it can also act in an autocrine manner. The GM3 TFO may, therefore, have a potential therapeutic role in blocking either the paracrine or autocrine production of GM-CSF in cases of acute myeloid leukemia. The detection of GM-CSF mRNA in the bronchoalveolar lavage of atopic patients and GM-CSF protein in the synovial fluid of rheumatoid arthritis patients (4,5) suggests another role for the therapeutic use of TFOs that block GM-CSF transcription.