ABSTRACT
The rat [alpha]1(I) collagen promoter contains a unique polypurine-polypyrimidine sequence between -141 and -200 upstream of the transcription start site. The polypurine sequence from -171 to -200 (C2) is on the coding strand and the adjacent polypurine sequence from -141 to -170 (C1) is on the non-coding strand. Earlier we demonstrated triplex formation with a polypurine 30 nt parallel triplex-forming oligonucleotide (TFO) corresponding to C1 and inhibition of transcriptional activity of the rat [alpha]1(I) collagen promoter. In the present work we have tested triplex-forming abilities of shorter (18 nt) purine and pyrimidine TFOs in parallel and antiparallel orientation to the C1 purine sequence. Our results show that purine antiparallel TFOs formed triplexes with the highest binding affinities, while pyrimidine oligodeoxyribonucleotides (ODNs) did not show appreciable binding. Phosphorothioate modification of purine TFOs did not significantly reduce binding affinity. We also demonstrate that preformed triplexes are quite stable when precipitated with ethanol and resuspended in water. Further analysis was carried out using two purine phosphorothioate antiparallel TFOs, 158 APS and 164 APS, designed to bind to the promoter region from -141 to -158 and -147 to -164, respectively, which were found to form triplexes even under physiological conditions. DNase I footprinting experiments showed the ability of these TFOs to protect target sequences in the promoter region; both purine sequences (C1 and C2) were protected in the case of 158 APS. Transfection experiments using preformed triplexes with a reporter plasmid containing the collagen promoter sequence showed significant inhibition of transcription when compared with a control phosphorothioate ODN. The effect of 164 APS was greater than that of 158 APS. These results indicate that this triplex strategy could be used in the down-regulation of collagen synthesis in cultured cells and offer the potential to control fibrosis in vivo.
Diseases characterized by abnormal deposition of fibrous tissue in various organs, including the heart, contribute considerably to morbidity and mortality worldwide (1 ). The common feature of these disease states is an expansion of connective tissue and abnormal accumulation of extracellular matrix, of which type I collagen is a major component (2 -5 ). Fibrosis can occur following an inflammatory response triggered by cytokines or in response to humoral factors (6 ,7 ). Although various strategies to suppress cytokine effects are in progress (2 ), modulation of collagen gene expression, which is a major end point of the fibrotic cascade, if achieved in a controlled and specific manner should lead to therapeutic advances in the management of fibrotic diseases.
Several cellular genes contain potential triple helix (triplex)- forming homopurine/homopyrimidine sequences (8 ). These sequences can be targets for regulation of gene expression by triplex-forming oligonucleotides (TFOs). These TFOs occupy the major groove of duplex DNA forming purine motif (A:AT or G:GC) or pyrimidine motif (C+:GC or T:AT) triplex structures with the native purine strand (9 ). The most stable triplexes are formed when the third strand binds in antiparallel orientation to the homologous native strand (9 ). It has been shown that formation of triplexes with various promoters could result in suppression of transcription in cell culture systems (9 -12 ).
COL1A1 and COL1A2 genes code for the pro-[alpha]1(I) and pro-[alpha]2(I) chains of type I collagen (13 ). The promoter of the [alpha]1(I) collagen gene is unique in that two highly conserved contiguous polypurine-polypyrimidine sequences are present in the region from -141 to -200. The polypurine sequence at -141 to -170 (C1) is localized on the non-coding strand, whereas the polypurine sequence at -171 to -200 is present on the coding strand (C2). It has been shown that this polypurine-polypyrimidine sequence between -141 and -200 of the rat [alpha]1(I) collagen promoter contains elements that can regulate transcription in a positive or negative manner (14 ,15 ). Earlier we demonstrated that a TFO corresponding to the polypurine sequence between -138 and -172 (C1) in parallel orientation could form a stable triplex at high concentrations and inhibit transcriptional activity of the rat [alpha]1(I) collagen promoter (16 ). In the present work we have examined the ability of polypurine and polypyrimidine TFOs of varying chain lengths, with or without backbone modifications and in either parallel or antiparallel orientation to the native homopurine sequence, to form triplex structures with this region. We have also tested shorter, overlapping 18 nt phosphodiester and phosphorothioate oligonucleotides targeted to different elements of this promoter sequence. Our results demonstrate that the antiparallel TFOs form triplexes more efficiently than parallel TFOs and that formation of triplexes with phosphorothioate TFOs is comparable with the corresponding phosphodiester TFOs. Further, we show that triplex formation with two antiparallel purine phosphorothioate adenine/guanine-rich (AG) oligonucleotides targeted to different regions of the promoter is correlated with significant inhibition of transcriptional activity of the rat [alpha]1(I) collagen promoter.
Phosphodiester oligonucleotides were synthesized on an Applied Biosystems 381A DNA synthesizer at the DNA Core Laboratory of the University of Missouri (Columbia, MO) and phosphorothioate oligonucleotides were made available by the Kansas Research Laboratories (Lawrence, KS). The C1 and C2 regions and the sequence of oligonucleotides used in this study are schematically represented in Figure 1 . Double-stranded oligonucleotides were prepared by heating equal amounts of complementary single strands at 80oC for 5 min in 0.25 M NaCl, followed by slow cooling to room temperature. Annealed double-stranded oligonucleotides were purified by electrophoresis on a 6% polyacrylamide gel, electroeluted and ethanol precipitated.
Duplex oligonucleotides were end-labeled with [[gamma]-32P]ATP using T4 polynucleotide kinase and purified through a Sephadex G25 column. Approximately 10 000 c.p.m. (~2.0 nM) were incubated with control oligonucleotide or increasing concentrations of specific TFOs in a binding buffer (TFO-BB) containing 20 mM Tris-HCl, pH 7.4, 20 mM MgCl2, 2.5 mM spermidine, 10% sucrose, 0.25 mg/ml bovine serum albumin and incubated at 22oC for 60 min. The samples were analyzed on a 10% polyacrylamide-0.32% bisacrylamide gel in a buffer containing 89 mM Tris, 89 mM boric acid, pH 7.5, and 20 mM MgCl2 for 7 h at 10 V/cm2 at 6oC. Gels were dried, autoradiographed at -70oC and quantitated by densitometry. Kd values were determined by calculating the concentration of TFO at which 50% of target duplex was converted to the triplex form (17 ). Stability of triplexes was tested by ethanol precipitating the reaction mixture and resuspension in water, with subsequent gel electrophoresis (see also legend to Fig. 7 ).
A 320 bp fragment of the rat [alpha]1(I) collagen gene (-220 to +100) with BglII (-220) and XbaI (+100) sites was isolated from pCOLCAT 220 as described by Kovacs et al. (16 ). This fragment was end-labeled at the BglII site with [[alpha]-32P]dATP and an excess of cold dGTP using the Klenow fragment of Escherichia coli DNA polymerase I. Approximately 100 000 c.p.m. (4 ng) labeled fragment was incubated in TFO-BB with increasing concentrations of oligonucleotides at 22oC for 60 min. Twenty nanograms of DNase I was added to each reaction and incubated for 3 min. Reactions was stopped by adding 100 [mu]l stop buffer (200 mM NaCl, 20 mM EDTA, 1% SDS and 250 [mu]g/ml yeast RNA), followed by extraction with phenol:chloroform and precipitation with ethanol. The pellets were dissolved in 70% formamide buffer, denatured by boiling and loaded on a 10% acrylamide-7 M urea gel. G+A sequencing reactions of the same fragment were used as markers. Electrophoresis was carried out at 1800 V for 90 min. Gels were dried and autoradiographed at -70oC.
Fifty micrograms of plasmid pCOLCAT 220 containing the collagen promoter sequence from -220 (16 ) and 20 [mu]g pSVGAL (Promega, Madison, WI) were incubated with 50 [mu]g phosphorothioate oligonucleotides (increasing amounts of specific TFO, with total oligonucleotide amount being the same in each reaction) in TFO-BB at 22oC for 1 h and the mixture was ethanol precipitated. DNA was resuspended in water after precipitation and the amount of the plasmid was quantitated by agarose gel electrophoresis. Approximately 3 [mu]g pCOLCAT, which was co-precipitated with 1.2 [mu]g pSVGAL, was used for each transfection. Rat 2TK cells were plated at a density of 2 * 106 cells in 100 mm plates 15-20 h before transfection. Precipitated plasmids were incubated with 15 [mu]l LipofectAMINE (Life Technologies, Grand Island, NY) in serum-free Dulbecco's modified Eagle's medium (DMEM) for 45 min. The transfection mixture was diluted in DMEM containing 5% calf serum to a final volume of 3 ml and added to each plate and incubated for 5 h at 37oC. The medium was then changed to DMEM with 10% calf serum. After 24 h incubation, cells were harvested in 1 ml phosphate-buffered saline, sedimented by centrifugation and resuspended in 150 [mu]l 0.25 M Tris-HCl, pH 7.8. Cells were lysed by three cycles of freezing and thawing, centrifuged and protein concentration of the supernatant determined by the Bradford method (BioRad, Hercules, CA).
[beta]-Galactosidase activity was measured as described by Eustice et al. (18 ). The reactions were carried out by incubating 40 [mu]g cell extract in 80 mM sodium phosphate buffer, pH 7.4, 102 mM 2-mercaptoethanol, 9 mM MgCl2 and 8 mM chlorophenol red [beta]-D galactopyranoside (CPRG) at 37oC for 2 h. Optical density was measured at 570 nm. Chloramphenicol acetyltransferase (CAT) activity was assayed as described by Gorman et al. (19 ). Protein amounts used for CAT assay were normalized for transfection efficiency using [beta]-galactosidase activity. Cell extracts were incubated in 100 [mu]l containing 0.1 [mu]Ci D-threo(dichloroacetyl) [14C]chloramphenicol, 15 [mu]l 4 mM acetylcoenzyme A and 0.25 M Tris-HCl, pH 7.8, at 37oC for 90 min. Samples were extracted with ethyl acetate and subjected to thin layer chromatography in chloroform/methanol (95:5) for 90 min to separate chloramphenicol from its acetylated forms. The thin layer plates were dried and autoradiographed. The extent of conversion of [14C]chloramphenicol to the acetylated form was determined by scraping each spot and counting in a Beckman scintillation counter.
A unique feature of the [alpha]1(I) gene promoter is the presence of two contiguous stretches of polypurine/polypyrimidine sequences between -141 and -200. The polypurine sequence from -171 to -200 (C2) is on the coding strand and the adjacent polypurine sequence from -141 to -170 (C1) is on the non-coding strand (see Fig. 1 ). Several reports have demonstrated the importance of these two highly conserved contiguous stretches of polypurine/polypyrimidine sequences in the modulation of [alpha]1(I) collagen gene transcription (14 ,15 ).
We have earlier shown triplex formation with this target sequence using a polypurine 30 nt TFO (oligo C1) which was parallel in orientation to the native purine strand (C1 P in Fig. 1 ) (16 ). The capacity of the same C1 TFO, but in antiparallel orientation (Fig. 1 , C1 AP), to target the polypurine sequence to form a triplex was investigated. The results indicated that C1 AP formed an extraordinarily stable triplex, with a Kd of ~0.08-0.1 [mu]M (Fig. 2 ). In the previous work with the parallel TFO and in the experiment described in Figure 2 with the antiparallel TFO we used 30mer TFOs which span the entire C1. In order to define the influence of length on triplex formation with shorter TFOs, we used 18mer TFOs in parallel and antiparallel orientation. The results of these experiments are shown in Figure 2 . As expected, the 18mer parallel TFO formed a triplex only at very high concentration (Kd >~10 [mu]M; Fig. 2 ). In contrast, the antiparallel TFO formed a stable triplex at extremely low concentrations. More than 30% of the duplex was converted to triplex at 0.06 [mu]M and 50% conversion occurred at ~0.08-0.1 [mu]M. From these results we conclude that the antiparallel 18mer TFOs form triplexes at a Kd similar or identical to the 30mers and that the effective Kd is at least 100* lower for antiparallel compared with parallel TFOs. It is known that polypyrimidine TFOs can also form triplexes (9 ) and therefore the feasibility of triplex formation with a polypyrimidine motif was determined with 18mer TFOs targeted to the midregion of the C1 sequence, in parallel and antiparallel orientation to the homologous native sequence. There was no appreciable binding at physiological pH in the presence of polyvalent cations with TFOs of either orientation. Since polypyrimidine motif triplex formation is favored by acidic pH (20 ,21 ), we also carried out binding reactions at pH 6.8 and found no appreciable triplex formation (data not shown).
Both C2 parallel and antiparallel TFOs formed triplexes with C2 (Kd ~5 and 1.25 [mu]M for parallel and antiparallel 30mer C2 TFOs respectively). Similar results were obtained with an 18mer (177, APS, Fig. 5 ). Our previous results on the binding of nuclear factors to C1 and C2 indicated that at least one factor interacts with both C1 and C2, as demonstrated by competition experiments. C1 and C2 contain polypurine sequences that could potentially form triplexes with the TFOs described above. For example, 158 APS could enter into an intermolecular triplex with C2 duplex. The data presented in Figure 5 indeed support this notion, as 158 APS formed a stable triplex with the 30mer C2 duplex at concentrations comparable with 177 APS, strongly suggesting that a TFO corresponding to C1 could form a triplex with C2 (Fig. 5 ).
We have shown that radiolabeled phosphodiester TFO 164 AP is able to form a triplex structure with a plasmid containing the target collagen promoter sequence. These complexes were stable during 20 h of gel migration (data not shown). Triple helical complexes formed between the promoter sequence from -220 to +100 and 158 APS did not dissociate even after precipitation with ethanol and resuspension in water. Supporting data for this observation is provided in Figure 7 . When C1 duplex was incubated with increasing concentrations of 158 APS TFO, formation of triplexes could be demonstrated even after ethanol precipitation and resuspension in water prior to loading on the gel. It appears that some duplex unwinding occurred, but preformed triplexes remained stable. This is true for all the TFOs (C1 AP, 158 APS, 164 APS and 170 APS) examined (Fig. 7 , bottom). Stability of triplexes in the face of precipitation and subsequent removal from a favorable environment was not surprising, in view of the fact that triplexes once formed are very stable and slow to dissociate (30 ).
We chose to form the triple helical complexes with the plasmid and transfect after precipitation to reduce the amount of free oligonucleotide and ensure that the promoter region had a triple helical structure on entry into the cell. Transient transfection experiments were done to evaluate the effectiveness of phosphorothioate TFOs in inhibiting transcription of the [alpha]1(I) collagen promoter in rat fibroblasts in culture. Plasmid pCOLCAT 220 was incubated with varying concentrations of a control phosphorothioate oligonucleotide (unable to form a triplex with C1 or C2 sequences), 158 APS or 164 APS. The total amount of oligonucleotide remained constant in each triplex binding reaction. pSVGAL was added to the control to counter any possible variation which could occur during DNA precipitation, resuspension or transfection. In addition, plasmid amounts were quantitated by agarose gel electrophoresis and incubated under UV light before transfection to ensure that similar amounts of plasmid DNA were transfected. Since the amounts of oligonucleotide were the same in all reactions before preciptitation, any adverse effect on transfection efficiency or cell growth/metabolism due to precipitated oligonucleotides would have been similar in all groups of transfections. The data shown in Figure 8 indicate a significant effect of 158 APS and 164 APS in inhibiting transcription from the collagen promoter. The CAT activities shown were the average of three independent experiments done in duplicate. As compared with a control oligonucleotide, transfection of preformed triplexes of pCOLCAT 220 with 158 APS resulted in a significant decrease in promoter activity (~50% of control) at a TFO to plasmid ratio of 125:1. Increasing the ratio to 250:1 did not result in any further significant decrease in CAT activity, since, as shown by in vitro experiments, triplex formation with 158 APS reached saturation kinetics beyond a ratio of 125:1. 164 APS showed a more pronounced effect. CAT activity was reduced by 70% compared with control oligonucleotide at a TFO to plasmid ratio of 125:1, with a further reduction in activity (80%) on increasing the ratio to 250:1. These results clearly demonstrate significant inhibition of collagen promoter function on triplex formation of the target region with 158 APS and 164 APS, with 164 APS showing a more dramatic effect.
Earlier we showed relatively weak triple helix formation with the target C1 of the rat [alpha]1(I) collagen promoter using a 30mer purine TFO which was parallel in orientation to the native purine strand (16 ). Since polypurine antiparallel strands in general form stable triplex structures more efficiently than parallel strands, we tested and found that a 30mer antiparallel TFO had significantly higher binding affinity (Kd ~0.08-0.1 [mu]M). Shorter (18 bp) antiparallel TFOs also bind much better than a parallel TFO of similar length, indicating that antiparallel orientation was favored for purine motif triplex formation with this sequence.
Phosphorothioate-modified TFOs, which are resistant to nuclease digestion (29 ), were tested for their ability to form triplexes. We chose three 18mer phosphorothioate TFOs targeted to overlapping sequences of the C1 duplex (Fig. 1 ). Comparison of binding affinities of unmodified and modified oligonucleotides (164 AP versus 164 APS) showed only a slight decrease in the triplex-forming ability of 164 APS to target C1 (see Figs 2 and 3 ). The oligonucleotide targeted to the guanine-rich sequence on the right side (158 APS) showed the highest triplex-forming ability, with an approximate Kd in the nanomolar range. Svinarchuk et al. (31 ) have shown that AG oligonucleotides (5'-GGGGAGGGGGAGG-3') form extraordinarily stable triplexes with a duplex target in the c-pim-1 promoter. This TFO targets a cluster of guanine nucleotides, i.e. GGAGGGGGAGG, the central motif containing five G residues possibly contributing to the formation of these stable triplexes. Both C1 and C2 sequences contain a cluster of five G residues and, since the 158 APS has the highest affinity, suggests that this GGGGGG sequence of 158 APS is probably responsible for the formation of stable triplexes.
Triplex formation by guanine-rich oligonucleotides has been shown to be dramatically inhibited by physiological concentrations of Na+ and K+, due to the ability of these monovalent cations to stabilize guanine quartet structures (25 ). It has been reported that physiological concentrations of K+ (140 mM) and Mg2+ (1-10 mM) and spermine (5 mM) (24 ,26 ,27 ) actually inhibit triplex formation (3 ). Singleton and Dervan (32 ) showed that increasing the K+ concentration from 5 to 140 mM in the presence of 1 mM spermine4+ and 1 mM Mg2+ decreased triplex binding affinity by 100-fold. The negative effect of the physiological milieu is compounded in the case of guanine-rich TFOs by guanine quartet formation in the presence of K+. In view of these findings, it is possible that intracellular conditions may significantly hamper in vivo triplex formation by the C1 TFOs tested. Hence, we determined the triplex-forming ability of the most G-rich TFO among them, i.e. 158 APS, under ionic conditions mimicking that inside the cell (24 ,26 ,27 ). To our surprise, we found formation of stable triplexes even under these unfavorable conditions, albeit at somewhat higher concentrations of the TFO (Kd ~1-4 [mu]M). This, we feel, is further evidence that this triple helical strategy is feasible in vivo, where the presence of polyvalent cations, up to 5 mM spermine in the nucleus (26 ), could significantly stabilize the triplex structure. Interestingly, with increasing triplex fromation we found the appearance of a band which appears to be single-stranded DNA formed due to dissociation of double-stranded target. The explanation for this unusual finding may lie in the fact that the effect of various ions in promoting duplex stability parallels their effect on triplex stability (32 ). Hence, in the same unfavorable milieu containing excess monovalent cations, the di- and polyvalent cations required for duplex stabilization when diverted for triplex stabilization may result in duplex instability.
The polypurine sequence C2, which is just upstream of C1 (on the alternate strand), also formed triplex structures, albeit with somewhat lower affinity than the C1 sequence. Antiparallel 30mer TFO (oligo C2 AP) showed the highest binding affinity (Kd ~1.25 [mu]M) compared with the parallel oligonucleotide (oligo C2). Shortening the oligonucleotide to 18 nt and phosphorothioate modification (177 APS) did not seriously affect the triplex-forming ability (1.25 [mu]M for 30mer versus 4 [mu]M for 18mer). In view of significant sequence homology, experiments were also done to determine whether the purine sequences from the C1 region could form triplex structures with duplex target C2. The three C1 TFOs previously mentioned were used. 158 APS showed the highest binding affinity (Kd ~4 [mu]M), which was equivalent to 177 APS. There was no appreciable triplex formation with 164 APS and 170 APS. These results suggest that 158 APS could down-regulate collagen gene expression by binding to the dual target in the rat [alpha]1(I) collagen promoter. DNase I footprinting has been employed to explore formation of intermolecular triplexes at sequences containing purine tracts on alternate strands (33 ,34 ). Since the enzyme cleaves duplex DNA from the minor groove, inhibition of DNase I digestion is thought to arise from triplex-induced changes in DNA structure and/or flexibility (33 ). We employed this strategy to determine whether the TFO 158 APS was able to form a stable triplex structure with the purine sequences in the C1 and C2 regions simultaneously. As shown in Figure 6 , 158 APS protected not only the homologous sequence in C1, but also the C2 sequence from -193 to -176. Increased DNase I digestion in the intervening sequence between the areas of protection indicates that protection of both sequences is mediated by triplex formation.
The effect of triplex formation on transcriptional activity of the collagen promoter in vivo was assessed by using the phosphorothioate TFOs 158 APS and 164 APS. A significant inhibitory effect was seen when preformed triplexes were introduced into cells, which was maximum for 164 APS. This was a sequence-specific effect, as suggested by the lack of effect of a control oligonucleotide on transcription of the collagen promoter. The effect of 158 APS on transcriptional activity reached a peak at a TFO to plasmid ratio of 125:1 and did not significantly increase at higher ratios. These results are in keeping with the predicted kinetics of triple helix formation of 158 APS with C1 target duplex (Fig. 3 ). At ratios of ~60-120:1 (TFO to duplex) there was almost 80% conversion of duplex to triplex, which increases to >90% at ratios of 120-250:1. In the case of 164 APS, the triplex binding assays had shown that at ratios of TFO to target of 60-250:1 binding was still in the linear range. Hence, the increased inhibitory effect of 164 APS at a ratio of 250:1 as compared with 125:1 may be explained by increased triplex formation with the collagen promoter. These results offer preliminary evidence for down-regulation of collagen gene transcription in in vivo systems using two 18 nt phosphorothioate TFOs targeted to a critical region of the collagen promoter.
The differential effect of the two phosphorothioate TFOs on collagen gene transcription is very interesting. 158 APS, which binds more strongly to C1 and in addition binds to another purine sequence in C2, had a lower inhibitory effect in vivo on collagen promoter function than 164 APS. We would like to offer some potential explanations for this, based on the current understanding of mechanisms of biological effects of TFOs. Maher et al. (35 ) have proposed several mechanisms of action for promoter-specific repression by triple helical DNA complexes. These include: (i) elimination of activator binding; (ii) disruption of factor interactions with the promoter; (iii) induction of changes in the physical properties of promoter DNA (bending at duplex-triplex junctions and reduction of DNA flexibility) that inhibit interaction of different parts of the promoter to form a productive initiation complex. Hence, it is possible that although 158 APS binds more avidly to two separate regions of the promoter, it may not be displacing a strong positive activator of transcription. Furthermore, the changes in DNA conformation induced may not be critically placed as to significantly inhibit interaction of positive factors with the transcription initiation complex. One or both of these mechanisms may be strongly operational in the case of 164 APS, accounting for its more pronounced effect.
Controlled inhibition of collagen gene transcription using TFOs is a potentially powerful therapeutic tool, if fully realized with minimal toxicity. We have refined the triplex strategy targeted to a unique alternate strand polypurine sequence in the [alpha]1(I) collagen promoter and demonstrated the effect of two phosphorothioate-modified short 18mer TFOs in forming triplex structures with both polypurine sequences. Preliminary results indicate the potential for translation of in vitro efficacy into in vivo effect in cell culture systems. Further experiments to explore the effectiveness of this strategy in tissue culture and in animal models of fibrosis are in progress.
We thank David Pintel, Ashok Nambiar, S.K.Swamynathan, Masaru Nakanishi and Seth Ririe for their comments in the manuscript and Veni Kandala for technical assistance. This work was supported by a M.S.White grant from the School of Medicine, University of Missouri-Columbia and by the Hisamitsu Pharmaceutical Co. Inc. We thank CRC for their support.
REFERENCES