Selection of effective antisense oligodeoxynucleotides with a green fluorescent protein-based assay Discovery of selective and potent inhibitors of glutathione S -transferase Mu expression Peter A. C. ¢t Hoen 1,2 , Bram-Sieben Rosema 1 , Jan N. M. Commandeur 2 , Nico P. E. Vermeulen 2 , Muthiah Manoharan 3 , Theo J. C. van Berkel 1 , Eric A. L. Biessen 1 and Martin K. Bijsterbosch 1 1 Division of Biopharmaceutics, Leiden/Amsterdam Center for Drug Research, the Netherlands; 2 Division of Molecular Toxicology, Leiden/Amsterdam Center for Drug Research, the Netherlands; 3 ISIS Pharmaceuticals, Carlsbad, California, USA Antisense oligodeoxynucleotides (AS-ODNs) are frequently used for the down-regulation of protein expression. Because the majority of potential antisense sequences lacks effect- iveness, fast screening methods for the selection of effective AS-ODNs are needed. We describe a new cellular screening assay for the evaluation of the potency and specificity of new antisense sequences. Fusion constructs of the gene of interest and the gene encoding the enhanced green fluorescent pro- tein (EGFP) are cotransfected with AS-ODNs to COS-7 cells. Subsequently, cells are analysed for expression of the EGFP fusion protein by flow cytometry. With the assay, we tested the effectiveness of a set of 15 phosphorothioate ODNs against rat glutathione S-transferase Mu1 (GSTM1) and/or Mu2 (GSTM2). We found several AS-ODNs that demonstrated potent, sequence-specific, and concentration- dependent inhibition of fusion protein expression. At 0.5 l M , AS-6 and AS-8 inhibited EGFP–GSTM1 expression by 95 ± 4% and 81 ± 6%, respectively. AS-5 and AS-10 were selective for GSTM2 (82 ± 4% and 85 ± 0.4% decrease, respectively). AS-2 and AS-3, targeted at homologous regions in GSTM1 and GSTM2, inhibited both isoforms (77–95% decrease). Other AS-ODNs were not effective or displayed non-target-specific inhibition of protein expres- sion. The observed decrease in EGFP expression was accompanied by a decrease in GSTM enzyme activity. As isoform-selective, chemical inhibitors of GSTM and GSTM knock-out mice are presently unavailable, the selected AS-ODNs constitute important tools for the study of the role of GSTM in detoxification of xenobiotics and protec- tion against chemical-induced carcinogenesis. Keywords: antisense oligodeoxynucleotide; carcinogenesis; genetic polymorphism; glutathione S-transferase; green fluorescent protein. Antisense oligodeoxynucleotides (AS-ODNs) are frequently used for the down-regulation of gene expression, both in vitro and in vivo [1–4]. Due to the low stability of phosphodiester ODNs (PO-ODNs) in biological systems, more stable oligonucleotide analogues with a variety of chemical modifications have been developed [5,6]. ODNs with a phosphorothioate-modified backbone (PS-ODNs) are the most commonly used AS-ODNs. As AS-ODNs act via Watson–Crick base pairing with their target mRNAs, the nucleotide sequence of the target gene is in principle sufficient information for the design of AS-ODNs. It appears, however, that not all AS-ODNs are potent inhibitors of protein expression. In studies where large sets of PS-ODNs, directed against a single target gene, were tested for their ability to down-regulate their target mRNA and protein in cell culture [7,8], only 5–10% of the sequences tested appeared to be effective. Thus there is a need for rapid and accurate screening assays for the selection of effective and specifically acting AS-ODNs. Screening for effective antisense sequences is usually performed in cell-free systems or in cell culture. Several cell- free assay systems have been described [9]. These assays are fast, but not always reliable predictors for activity in biological systems. Use of differentiated cells generates more relevant information on the effectiveness of AS-ODNs in physiological systems. However, cellular assays are fre- quently hampered by low or irreproducible transfection of oligonucleotides. Furthermore, each new target requires the set-up and optimization of target-specific assays. Therefore, we developed a new assay that uitilizes fusion constructs of a particular gene with the gene encoding enhanced green fluorescent protein (EGFP) as reporter. Because the screening is based on flow cytometric detection of EGFP expression, there is no need for development of target- specific assays. Reproducible transfections are achieved by using an easy transfectable cell line and the cotransfection of antisense PS-ODNs directed against the gene of interest and plasmids encoding the chimeric gene. In the present assay, Correspondence to M. K. Bijsterbosch, Leiden/Amsterdam Center for Drug Research, Division of Biopharmaceutics, PO Box 9502, 2300 RA Leiden, the Netherlands. Fax: +31 71 5276032, Tel.: +31 71 5276038, E-mail: bijsterb@lacdr.leidenuniv.nl Abbreviations: ODN, oligodeoxynucleotide; AS, antisense; PO: phosphodiester; PS, phosphorothioate; EGFP: enhanced green fluorescent protein; GST, glutathione S-transferase; GSTM, gluta- thione S-transferase Mu; CDNB, 1-chloro-2,4-dinitrobenzene; GSH, glutathione; DOTAP, N-[1-(2,3-dioleoyloxy)propyl]-N,N,N- trimethylammonium; PMSF, phenylmethanesulfonyl fluoride; TRITC, tetramethylrhodamine isothiocyanate; DMEM, Dulbecco’s modified Eagle’s medium; FACS: fluorescence activated cell sorter. Enzymes: glutathione S-transferase (EC 2.5.1.18). (Received 13 December 2001, revised 4 April 2002, accepted 9 April 2002) Eur. J. Biochem. 269, 2574–2583 (2002) Ó FEBS 2002 doi:10.1046/j.1432-1033.2002.02924.x only transfected cells are analysed, thus eliminating the background of expression in untransfected cells. With this assay, we determined the effectiveness of a set of antisense PS-ODNs directed against the glutathione S-transferase Mu1 (GSTM1) and Mu2 (GSTM2) isoforms of the rat. GSTs play an important role in the detoxification of DNA- and/or protein-reactive compounds by catalyzing the conjugation of electrophilic groups with the tripeptide glutathione [10]. Approximately 50% of the Caucasian population is deficient for GSTM1, the human orthologue of rat GSTM1 and GSTM2 [11]. Meta-analyses of epide- miological studies reveal that this deficiency is associated with an increased risk of lung and colorectal cancer, especially when the GSTM-null genotype is combined with high-inducibility of cytochrome P450 1A1 [12–18]. It would, however, be very important to demonstrate directly the effect of differences in GSTM expression levels on the prevalence of cancer biomarkers in in vitro andinanimal models. As GSTM knock-out mice are still unavailable, temporal modulation of the expression of GSTM isoforms by AS-ODNs in relevant in vitro and in vivo models is an attractive possibility. In the current paper, several target- specific AS-ODNs are selected from a set of 15 PS-ODNs. These ODNs selectively inhibit the expression of GSTM1 and/or GSTM2 and can be used to study the influence of reduced GSTM expression on the detoxification of xeno- biotics and protection against chemical-induced carcino- genesis. MATERIALS AND METHODS Materials PCR primers were from Eurogentec, Seraing, Belgium. PS-ODNs were synthesized according to standard phos- phoramidite chemistry. The pEGFP-C1 plasmid and the rabbit anti-EGFP Living Colors Peptide antibody were from Clontech. VentÒ DNA Polymerase was from New England Biolabs. 1-chloro-2,4-dinitrobenzene (CDNB), glutathione (GSH), dithiothreitol, N-[1-(2,3-dioleoyl- oxy)propyl]-N,N,N-trimethylammonium salt (DOTAP), and propidium iodide were from Sigma. Phenyl- methanesulfonyl fluoride (PMSF) and Tween-20 were from Merck. Tetramethylrhodamine isothiocyanate 5,6-mixed isomers (TRITC) was from Molecular Probes. Cell culture agents were from BioWhittaker. Milkpowder was from Campina Melkunie (Eindhoven, the Netherlands). A horseradish peroxidase-conjugated antirabbit IgG antibody and an enhanced chemiluminescence assay were purchased from Amersham Pharmacia Biotech. All other chemicals were of analytical grade. Cloning of GSTM cDNAs into pEGFP-C1 Full-length cDNAs encoding GSTM1 (bases )21 to +1039; GenBank accession no. X04229, cloned in the PstI site of pBR322) and GSTM2 (bases )2 to +1036; GenBank accession no. J03914, derived mRNA sequence, cloned in the EcoRI site of pUC18) were kindly provided by D. Tu, Pennsylvania State University, PA, USA. The cDNA inserts were amplified and isolated by the sticky-end PCR method [19]. Briefly, the cDNAs were amplified with two sets of PCR primers (forA + revA and forB + revB) for each plasmid using Vent DNA polymerase (Table 1). The two PCR products were subjected to melting and cooling. Four different double-stranded products were obtained, one of which had the correct 5¢-EcoRI and 3¢-BamHI over- hanging ends. The PCR products were cloned into the EcoRI- and BamHI-digested pEGFP-C1 plasmid to gener- ate the C-terminal fusion constructs pEGFP-M1 and pEGFP-M2. Sequencing of the plasmids confirmed the in-frame ligation of the GSTM cDNAs and the absence of any PCR-induced mistakes in the inserts. TRITC-labelling of ODN A 24-mer PO-ODN, provided with three PS-linkages at the 5¢-end and a 3¢-end primary amino group (sequence: T*A*A*GCTGTCCCGGGGTCTACGGCC), was label- led with TRITC by incubating 15 nmol ODN in 500 lL 0.1 M Na-carbonate buffer (pH 9.0) with 10 molar equiv- alents of TRITC (dissolved in dimethylformamide at 2mgÆmL )1 ). The mixture was incubated overnight with shaking at room temperature. The TRITC-labelled ODN was separated from unreacted TRITC by gel filtration on a Sephadex G-25 column (20 · 0.4 cm), eluted with water. The TRITC-ODN was precipitated from the eluent by adding 0.01 vols 1 M MgCl 2 ,0.1vols3 M NaAc pH 5.2, and 3 vols cold ethanol. The precipitate was formed by overnight incubation at )20 °C and centrifugation for 30 min at 13000 g at 4 °C. The pellet was washed three times with 80% EtOH and subsequently dissolved in deionized water. The purity and identity of the TRITC-labelled ODN were checked by PAGE under denaturing conditions. Cell culture and transfection COS-7 cells (European Collection of Cell Cultures, Salis- bury, UK) were grown at 37 °Cina5%CO 2 atmosphere in Dulbecco’s modified Eagle’s medium (DMEM) containing 10% (v/v) fetal bovine serum, 2 m ML -glutamine, 100 UÆmL )1 penicillin and 100 lgÆmL )1 streptomycin. At 24 h before transfection, cells were seeded in 12-wells plates Table 1. Primers used for sticky-end PCR. Separate PCR reactions were carried out with the following primer sets: GSTM1-forA and GSTM1-revA; GSTM1-forB and GSTM1-revB; GSTM2-forA and GSTM2-revA; GSTM2-forB and GSTM2-revB. Products from the first two PCR reactions were combined, melted and reannealed to give four GSTM1 cDNA products, of which one has the right EcoRI/ BamHI overhanging ends (underlined) to enable ligation in EcoRI and BamHI restricted pEGFP-C1. Combining the last two PCR reactions results in the formation of GSTM2 cDNA with EcoRI/BamHI over- hanging ends (underlined) which was cloned in a similar way into pEGFP-C1. Primer Sequence GSTM1-forA 5¢-AATTCCATGCCTATGATACTGGGAT-3¢ GSTM1-forB 5¢- CCATGCCTATGATACTGGGAT-3¢ GSTM1-revA 5¢- CTAAAGATGAGACAGGCCTGG-3¢ GSTM1-revB 5¢-GATCCTAAAGATGAGACAGGCCTGG-3¢ GSTM2-forA 5¢-AATTCGATGCCTATGACACTGGGTTAC-3¢ GSTM2-forB 5¢- CGATGCCTATGACACTGGGTTAC-3¢ GSTM2-revA 5¢- CGTGGTTCACACTTTATTGCAAATC-3¢ GSTM2-revB 5¢-GATCCGTGGTTCACACTTTATTGCAAATC-3¢ Ó FEBS 2002 EGFP-based selection of antisense sequences (Eur. J. Biochem. 269) 2575 at a density of 1 · 10 5 cells/well,whichresultedincultures that were approximately 50% confluent at the day of transfection. For cotransfections of plasmid and ODN, the appropriate amount of plasmid [diluted to a concentration of 0.2 lgÆlL )1 in HBS (0.15 M NaCl in 20 m M Hepes, pH 7.4)] was mixed with the appropriate amount of ODN (diluted to a concentration of 0.2 lgÆlL )1 in HBS). Subse- quently, a transfection mixture was prepared by slowly adding the plasmid/ODN mixture to a solution of DOTAP in HBS (1 lgÆlL )1 ; charge ratio DNA : DOTAP ¼ 1:5). The DNA and DOTAP solutions were mixed by repeated pipetting. The transfection mixture was diluted with HBS to a total volume of 100 lL, and incubated for 15 min at room temperature. Then, the culture medium was taken from the cells and replaced by 400 lL of DMEM without serum or antibiotics, and 100 lL of the transfection mixture was slowly added. After 4 h, the transfection mixture was removed from the cells, and serum-containing medium was added. All analyses were performed after culture in the serum-containing medium for a further 18 h. Flow cytometry Cells were detached from the culture plates with trypsin, centrifuged for 5 min at 400 g, washed once with 1 mL NaCl/P i , and dispersed in 1 mL NaCl/P i . Immediately before FACS analysis, 3 lL1l M propidium iodide was added. Cellular fluorescence of approximately 3000 cells was determined in a Becton Dickinson FACS Calibur flow cytometer. The EGFP signal was detected in the FL-1 channel; TRITC and propidium iodide signals were detected in the FL-3 channel. Only single cells were gated in forward/ sideward scatter plots; dead cells were excluded from the analysis by gating of propidium iodide-positive cells. GST activity assay COS-7 cells were transfected with either pEGFP, pEGFP- M1 or pEGFP-M2 as described above. Then, the cells were washed twice with NaCl/P i and lysed in 300 lL10m M sodium phosphate buffer (pH 7.4) containing 2 m M dithio- threitol, 1 m M EDTA, and 50 l M PMSF. The lysates were homogenized by short sonication. Total GST activity was analysed in a CDNB conjugation assay, essentially as described before [20]. The assay makes use of the GST- catalyzed addition of GSH to CDNB. The CDNB–GSH conjugate formed can be measured spectrophotometrically. To this end, 50 lL of protein lysate ( 10 lgprotein, concentration determined with the Bradford protein assay [21]) was incubated with 150 lL of a solution of 1.67 m M CDNB in 0.1 M potassium phosphate buffer pH 6.5. Lysis buffer, instead of lysate, was used as a blank. The reaction was started by the addition of 50 lL5m M GSH dissolved in potassium phosphate buffer pH 6.5. CDNB–GSH con- jugate formation was monitored over time with a Perkin- Elmer HTS7000 bioassay plate reader at 340 nm and 37 °C. The rate of conjugate formation was constant 15–45 min after the addition of GSH. Western blotting Lysates of COS-7 cells transfected with pEGFP, pEGFP- M1 or pEGFP-M2, prepared as described above, were analysed for GFP expression by Western blotting. Five lg total cellular protein, dissolved in denaturing loading buffer (62 m M Tris/HCl pH 6.8, 12.5% v/v glycerol, 1.25% w/v SDS, 2.5% v/v 2-mercaptoethanol, and 0.25% w/v Bromo- phenol blue) were heated for 4 min at 96 °C, and subjected to gel electrophoresis in an SDS/15% polyacrylamide gel. Proteins were blotted overnight at 4 °C onto a nitrocellulose membrane at a current of 76 mA. Thereafter, the nitrocel- lulose membrane was incubated for 1 h in blocking buffer, consisting of 10 m M Tris/HCl pH 8.0, 150 m M NaCl, 0.5 m M CaCl 2 , 5% w/v milkpowder, 1% w/v BSA, 0.25% v/v Tween-20. Then, the membrane was incubated for 1 h at room temperature with the primary anti-EGFP antibody (100 · diluted in blocking buffer without milk powder, containing 0.5% v/v Tween-20). The membrane was washed 10 times with NaCl/P i containing 0.02% v/v Tween-20, and incubated with a horseradish peroxidase- conjugated donkey antirabbit IgG (10 · dilutedin10· diluted blocking buffer). EGFP was detected by an enhanced chemiluminescence assay, according to the manufacturer’s protocol. Statistical analysis Data were analysed statistically for significance with a one or two sample student t-test. GRAPHPAD INSTAT Software version 3.00, GraphPad Software Inc. (San Diego, CA, USA), was used for this purpose. RESULTS Cloning of EGFP–GSTM fusion constructs Cellular screening of antisense sequences for their potential to inhibit gene expression is often complicated by irrepro- ducible transfection procedures and lack of good quantita- tive assays for monitoring of gene expression. To circumvent these problems, we developed a screening assay, based on fusion proteins of the target protein with EGFP, that enables accurate determination of the effects of AS-ODNs by flow cytometry. C-terminal fusion constructs of GSTM1 and GSTM2 with EGFP (named pEGFP-M1 and pEGFP-M2, respectively) were made by ligating PCR- amplified cDNAs, coding for GSTM1 and GSTM2 (PCR primers in Table 1), into the multiple-cloning site of the pEGFP-C1 vector. Sequence analysis confirmed a correct in-frame ligation of the two cDNAs and the absence of any sequence errors in the inserts. By flow cytometric analysis, it was shown that transfection of COS-7 cells with pEGFP, pEGFP-M1, or pEGFP-M2 proceeded with equal efficien- cies (30 ± 2%, 31 ± 1%, and 29 ± 1%, respectively). The average intensity of the fluorescent signal of the EGFP– M1 and EGFP–M2 fusion proteins was only slightly lower than that of EGFP itself [2.1 ± 0.1, 1.7 ± 0.1, and 1.8 ± 0.1 (· 10 3 arbitrary units) for EGFP, EGFP–M1, and EGFP–M2, respectively], indicating that the EGFP moiety of the fusion proteins retained its activity. Fluores- cent microscopy revealed that EGFP and the EGFP–M1 and EGFP–M2 fusion proteins localized in the cytosol. GST activity in lysates of the transfected COS-7 cells was assayed by measuring GSH–CDNB conjugate formation. As shown in Fig. 1A, total GST activity in COS-7 cells was increased 3.4- and 1.9-fold after transfection with pEGFP– 2576 P. A. ’t Hoen et al. (Eur. J. Biochem. 269) Ó FEBS 2002 M1 and pEGFP–M2, respectively. The observed differences in CDNB conjugation between pEGFP–M1- and pEGFP– M2-transfected cells can be explained by the lower catalytic activity of the GSTM2-2 protein towards CDNB compared with the activity of the GSTM1-1 protein [22]. The size of the fusion proteins was  50 kDa, as determined by Western blotting with an EGFP-specific primary antibody (Fig. 1B). This value is in close agreement with the expected size, calculated by summation of the molecular weights of EGFP (25 kDa) and GSTM (27 kDa). Colocalization of ODN and pEGFP For proper evaluation of antisense effects, it is important that the AS-ODNs and the EGFP-expressing plasmids are transfected into the same cells. This was accomplished by the cotransfection of plasmid and AS-ODN. By FACS analysis, it was shown that after cotransfection of COS-7 cells with a fluorescently labelled ODN and pEGFP, ODN and plasmid colocalized in the same target cells as > 90% of the EGFP-positive cells were also positive for the TRITC- labelled ODN (Fig. 2). The observations suggest that the uptake of ODN is far more efficient than the uptake of the EGFP plasmid, because almost all cells were positive for TRITC-labelled ODNs, whereas only  33% of the cells were expressing EGFP. Screening of ODNs for their antisense activity To identify AS-ODNs that are potent and sequence-specific inhibitors of GSTM1 and/or GSTM2 expression, 15 PS- ODNs were screened for their ability to inhibit EGFP– GSTM fusion protein expression. The ODNs were targeted against different regions in the mRNA of GSTM1 and GSTM2 (Table 2). Some of the ODNs (i.e. AS-1, AS-6, AS-7, AS-8, AS-12 and AS-15) were designed to inhibit selectively GSTM1 expression, whereas others (i.e. AS-5, AS-10, AS-11, AS-13 and AS-14) were designed to inhibit selectively GSTM2 expression. A third group of ODNs (i.e. AS-2, AS-3, AS-4 and AS-9) was directed against homol- ogous regions in the GSTM1 and GSTM2 mRNAs, and should therefore inhibit the expression of both isoforms. An unrelated AS-ODN (AS-ctrl) with no sequence homology with EGFP, GSTM1 or GSTM2 was taken as a negative control. An EGFP-specific PO-ODN (AS-GFP), with two PS-linkages at either end for protection against nuclease activity, was taken as a positive control. It has been reported that this ODN inhibits GFP expression in HeLa cells transiently transfected with pEGFP [23]. Initially, COS-7 cells were transfected with 1.6 lgofthe AS-ODNs (final concentration in the medium: 0.5 l M )and 0.5 lg of pEGFP, pEGFP-M1 or pEGFP-M2. After a 4-h transfection period and culture for a further 18 h, cells were analysed for EGFP expression by flow cytometry. Propidium iodide was added to the cell suspensions to exclude nonviable cells from the analysis. The percentage of propidium iodide-positive cells increased from  3% in cell cultures that were transfected with plasmid only, to  9% in cell cultures cotransfected with ODN and plasmid. This is probably due to cytotoxicity of DOTAP, as a larger amount of DOTAP was used for cotransfection than for transfection of plasmid alone. The number of propidium-iodide positive cells was the same for all cotransfected PS-ODNs. Clear differences were found in the ability of the various AS-ODNs to inhibit EGFP expression. The control Fig. 1. Expression of EGFP–GSTM fusion proteins in COS-7 cells. COS-7 cells were transfected with pEGFP-M1, pEGFP-M2, or pEGFP (0.5 lg DNA per well). After a further 18 h of culture, the cells were lysed. (A) Total GST activity in 50 lL of protein lysate was measured by following CDNB–GSH conjugate formation over time. The increase in absorption at 340 nm was recorded with lysis buffer as a blank. The GST activity is expressed as a percentage of the activity in pEGFP-transfected cells (0.46 DA 340 Æmin )1 Æmg protein )1 ). Means of 12 determinations in three separate experiments ± SEM are shown. **P < 0.0001 (unpaired student t-test). (B) A Western blot was per- formed on 5 lg protein lysate of pEGFP-M1 (lane 1), pEGFP-M2 (lane 2) and pEGFP (lane 3) transfected cells. The samples were denatured, and separated by SDS/15% PAGE, together with a Bio-Rad prestained kaleidoscope protein marker. Subsequently, proteins were blotted onto a nitrocellulose membrane, and the blot was incubated consecutively with a rabbit anti-EGFP antibody and a peroxidase-labelled goat anti-rabbit secondary antibody. EGFP- containing proteins were visualized with enhanced chemiluminescence. The positions and molecular weights of the marker proteins are indi- cated in the left margin. In the right margin, the estimated sizes of the protein bands are shown. Ó FEBS 2002 EGFP-based selection of antisense sequences (Eur. J. Biochem. 269) 2577 AS-ODN did not have any effect on the expression of EGFP, EGFP–M1 or EGFP–M2, indicating that cotrans- fection of PS-ODNs per se does not influence EGFP expression. AS-6 and AS-8, directed against GSTM1, inhibited EGFP–M1 expression by 95 ± 1% and 81 ± 6%, respectively (Fig. 3A). The expression of EGFP and the other isoform, EGFP–M2, were also affected, but the inhibitory effect on expression of pEGFP–M1 was significantly greater (P < 0.05) than the effect on expres- sion of EGFP or EGFP–M2. AS-1, however, inhibited the expression of all three proteins, and the expression of EGFP even by > 95%. This is probably not caused by sequence- specific hybridization with the EGFP mRNA, because the maximal continuous homologous region with the EGFP sequence was eight nucleotides long. AS-7 also displayed some nonspecific inhibition of protein synthesis: its effect on EGFP–M1 expression, although greater, was not signifi- cantly different from its effect on the expression of EGFP or EGFP–M2. Two other AS-ODNs against GSTM1, AS-12 and AS-15, were completely ineffective in the down- regulation of protein synthesis. Similar results were found for AS-ODNs targeted at GSTM2. AS-5 and AS-10 inhibited EGFP–M2 expression by 82 ± 4% and 85 ± 0.4%, respectively, and affected the expression of the control proteins EGFP and EGFP– M1 by < 10% (Fig. 3B). For AS-10, the isoform-specif- icity was remarkably good, as this AS-ODN contains only three mismatches with respect to the sequence of GSTM1. Again, one AS-ODN, AS-11, displayed nonspecific effects on the expression of all three proteins, whereas two other ODNs, AS-13 and AS-14, were not able to affect protein expression. AS-2 and AS-3, targeted against the coding sequence of both GSTM1 and GSTM2, inhibited the expression of the EGFP–GSTM isoforms by  95% (AS-2) and  80% (AS-3), while expression of the control EGFP was inhibited by 45 ± 12% and 18 ± 13%, respectively (Fig. 3C). AS-9 demonstrated severe nonspecific effects on EGFP expres- sion, as the expression of EGFP, alone or in a fusion construct, was inhibited by > 95%. The effects on EGFP expression were not due to sequence-specific hybridization because a significant homology with the sequence of EGFP was not found. AS-4 showed less severe, but significant, nonspecific effects on EGFP expression. Surprisingly, the AS-GFP, which was reported to down-regulate EGFP expression [23], did not have any effect on the expression of either of the EGFP proteins at the tested concentration of 0.5 l M . Fig. 2. FACS analysis of COS-7 cells cotransfected with TRITC-ODN and pEGFP. Untransfected COS-7 cells (A), cells trans- fected with 0.2 l M TRITC-ODN (B), cells transfected with 0.5 lg pEGFP (C), and cells transfected with 0.2 l M TRITC-ODN and 0.5 lg pEGFP (D) were analysed by flow cytometry for EGFP expression (FL-1, x-axis) and TRITC-ODN uptake (FL-3, y-axis). Single cells were gated in the forward–side- ward scatter plot (gate R1, not shown). The following gates were applied: R2, nontrans- fected; R3, TRITC-positive; R4, GFP-posit- ive; R5, TRITC-positive and GFP-positive. (A–D) provide representive examples of multiple FACS analyses. The table gives the amounts of cells (expressed as percentage of the total amount of cells) counted in each gate under the different incubation conditions. 2578 P. A. ’t Hoen et al. (Eur. J. Biochem. 269) Ó FEBS 2002 Determination of the concentration/activity profile of the AS-ODNs To compare the potency and specificity of some of the effective AS-ODNs, COS-7 cells were transfected with different concentrations of AS-1, AS-2, AS-5, AS-6, and AS-9. Figure 4 shows the effects of cotransfection with 0.01, 0.1 and 0.5 l M AS-ODN on the expression of pEGFP, pEGFP–M1, and pEGFP–M2. Both the true antisense effects and the non-target-specific effects appeared to be highly concentration-dependent. AS-5 was a potent and selective inhibitor of EGFP–M2 (IC 50 value  0.2 l M ), and did not have any effect on the expression of EGFP–M1 or EGFP at the highest concentration tested. AS-2, directed against both GSTM isoforms, and AS-6, directed at GSTM1, potently inhibited the expression of their respect- ive targets with IC 50 values slightly above 0.1 l M . At 0.1 l M , the inhibition was specific. However, at the highest concentration tested also the expression of EGFP was affected, indicating that sequence-specific antisense effects occur at lower concentrations than non-target-specific effects. AS-9 was the most potent inhibitor of both EGFP–M1 and EGFP–M2 expression with estimated IC 50 values < 0.1 l M . However, the EGFP expression was also inhibited, although to a slightly lesser extent. AS-1, targeted to GSTM1, was another nonspecific inhibitor of EGFP expression as the IC 50 values for inhibition of EGFP expression and of EGFP–M1 were both in the same range. Analysis of the effect of AS-ODNs on GST activity To examine whether the inhibitory effects of the AS-ODNs on EGFP–GSTM fusion protein expression were associated with a decrease in GST activity, we determined the effect of cotransfection with AS-ODNs on the GST activity in lysates of COS-7 cells transfected with pEGFP, pEGFP- M1, or pEGFP-M2. In the GST activity assay, we tested AS-5 and AS-10, directed against GSTM2, and AS-6, directed against GSTM1, which were found to be specific inhibitors of either GSTM1 or GSTM2 expression in the EGFP assay. The results are shown in Table 3. None of these ODNs affected the CDNB conjugation in cells transfected with pEGFP, indicating that the AS-ODNs were specific for the rat GSTM isoforms, and did not inhibit the activity of endogenous GSTs present in lysates of COS-7 cells. The rates of CDNB conjugation in the lysates of pEGFP-M1 and pEGFP-M2 transfected cells were correc- ted for the endogenous GST activity, determined in COS-7 cells transfected with pEGFP. AS-6 appeared to be a very potent inhibitor of GSTM1-1 enzyme activity (95 ± 2% decrease). AS-5 and AS-10 were somewhat less potent inhibitors of GSTM2-2 enzyme activity (77 ± 6%, and 70 ± 6% decrease, respectively). However, conjugation by the nontargeted isoform was also affected by 40–50%. Nonetheless, the effects of the different AS-ODNs on the activity of the targeted GSTM isoform were significantly greater than on the nontargeted GSTM isoform (P < 0.002 for all tested AS-ODNs). DISCUSSION Most of currently available screening assays for the selection of effective AS-ODNs are based on cell-free assays, e.g. RNAse H digestion screens and oligonucleotide scanning arrays [9]. As activity in cell-free assay may not always correlate with activity in cellular systems, we developed in the present study a novel cellular screening assay for the selection of effective AS-ODNs with a sequence-specific Table 2. Antisense ODN sequences. Target site b Mismatch (number of bases) c Name Sequence a Region b GSTM1 GSTM2 AS-ctrl TGAGAGCTGAAAGCAGGTCCAT Unrelated – – AS-GFP G*A*GCTGCACGCTGCCG*T*C GFP–CDS – – AS-1 GGCGG ATCGGGTGTGTCAGC CDS 36–55 – 5 AS-2 CCACTGGCTTCTGTCATAGT CDS 119–138 119–138 0 AS-3 GAAGTCCAGGCCCAGTTTGA CDS 152–171 152–171 0 AS-4 TCAATTAAGTAGGGCAGATT CDS 175–194 175–194 0 AS-5 TCTCCA AAACGTCCACACGA CDS – 285–304 4 AS-6 ACAAAGCATGATGAGCTGCA CDS 326–345 – 8 AS-7 GAGTA GAGCTTCATCTTCTC CDS 397–426 – 1 AS-8 ACTGGTCAAGAATGTCATAA CDS 480–499 – 7 AS-9 CAGGTTTGGGAAGGCGTCCA CDS 524–543 524–543 0 AS-10 CAGGCCCTC AAACCGAGCCA CDS – 554–573 3 AS-11 GTCTGGACTTTGTGGTGCTA STOP – 655–674 13 AS-12 GGCATGACTGGGGTGAGGTT 3¢-UTR 786–805 – 5 AS-13 AA AATCAGTGAGGGAAGGGT 3¢-UTR – 870–889 8 AS-14 TCTAATCTCTCAGGCCAGGC 3¢-UTR – 921–940 10 AS-15 GCAGCTCCCCCACCAGGAAC 3¢-UTR 978–997 – 12 a All sequences were PS-ODNs except for AS-GFP. The sequence of AS-GFP is taken from literature [23]: it is a PO-ODN with PS-modified internucleotide linkages at the 3¢- and 5¢-ends, indicated by asterisks. b The region in the mRNA against which the ODNs are indicated as follows: CDS, coding sequence; STOP, STOP codon; 3¢-UTR, 3¢-untranslated region. The target sites in the GSTM1 or GSTM2 mRNAs are indicated, nucleotide 1 being the ATG start site. c The number of mismatches in the corresponding region of the nontargeted isoform are given. Mismatches are underlined in the sequence. Ó FEBS 2002 EGFP-based selection of antisense sequences (Eur. J. Biochem. 269) 2579 mode of action. In the present assay, antisense activity is directly correlated with EGFP-derived fluorescence by constructing fusion proteins of the target protein and EGFP. Unlike in conventional target-specific screens, in the current assay specific antibodies need not be available and isoform- specific assays for the determination of enzyme activity need not be developed. Furthermore, the measurement of EGFP-derived fluorescence by flow cytometry has excellent quantitative properties and offers good reproducibility. This is probably due to the elimination of variation in transfection efficiencies as a complicating factor in the assessment of antisense effectiveness. Our experiments in which an EGFP- containing plasmid was cotransfected with fluorescently labelled ODNs, suggest that all EGFP-positive cells had taken up ODNs. Therefore, in the present assay the antisense effects are determined in the whole population of cells that express the target gene. In other cellular assays, including a luciferase reporter gene-based assay [24], antisense effects may be underestimated because not all cells that express the gene of interest are transfected with AS-ODNs. An EGFP- based approach has been used previously for the selection of ribozymes against the c-erbB-2 oncogene [25]. However, in this earlier study the plasmid coding for the c-erb-B-2 EGFP fusion protein, was cotransfected with a ribozyme expressing plasmid and not with an exogenously added antisense molecule. Cotransfection with the ribozyme-expressing plasmid resulted in a reduction of EGFP expression to a maximum of 70%, whereas we observed a > 90% reduction with our most potent ODNs. Possibly, a significant part of the c-erbB-2-EGFP transfected cells had not taken up a ribozyme construct. A C-terminal fusion construct and not an N-terminal fusion construct, was used because AS-ODNs against the 3¢-untranslated region of the mRNA of the gene of interest, which has been shown to be a favourable region for antisense action [7], can only be tested in C-terminal fusion constructs. The newly developed screening assay was used for the selection of effective AS-ODNs against rat GSTM1 and GSTM2 out of a set of 15 PS-ODNs. Some ODNs were designed to specifically inhibit either GSTM1 or GSTM2 expression, which show a sequence identity of  80% at the DNA level. For these ODNs, the nontargeted isoform served as a mismatch target control with 1–13 mismatches. Other ODNs were targeted against homologous regions in both isoforms. As a control for the true antisense nature of the observed effects on protein expression, the effects of the ODNs on the expression of EGFP without a fusion construct were evaluated. Three ODNs (AS-3, AS-5 and AS-10) were found to inhibit gene expression with very high sequence specificity. These ODNs reduced at 0.5 l M the expression of their target isoforms by > 80%, whereas the nontargeted isoform and/or EGFP control were not affected significantly. Other ODNs (AS-2, AS-6, AS-7 and AS-8) displayed a combination of target sequence-specific and nontarget-specific inhibitory effects on EGFP levels. These ODNs inhibited the targeted isoform to a signifi- cantly greater extent than the nontargeted isoform and/or EGFP control, but also attenuated the expression of the controls by 40–60%. Three ODNs (AS-1, AS-9 and AS-11) had severe non-sequence-specific effects on EGFP expres- sion. In these cases, the expression of EGFP without a GSTM fusion was affected to a similar extent as the expression of the targeted proteins. Five ODNs (AS-4, AS-12, AS-13, AS-14 and AS-15) did not demonstrate major effects on EGFP–GSTM fusion protein expression. Our results indicate once again that, when evaluating antisense effects, identification of false-positives is common. As stated before by others [26], it is crucial to analyse the effects on the expression of target-related control proteins, which is easily accomplished in our screening assay. Fig. 3. Effects of AS-ODNs on EGFP and EGFP–GSTM fusion protein expression. COS-7 cells were transfected with 0.5 lg pEGFP (open bars), pEGFP-M1 (hatched bars) or pEGFP-M2 (closed bars), together with 0.5 l M of the indicated AS-ODNs. The ODNs were directed against GSTM1 (A), GSTM2 (B), or both GSTM isoforms (C). AS-ctrl is a control ODN without sequence homology with GSTM or EGFP. AS-GFP is an AS-ODN against EGFP, taken from [23]. At 22 h after transfection, cells were analysed for EGFP expres- sion (FL-1) and propidium iodide uptake (FL-3) by flow cytometry. The number of living (i.e. propidium iodide-negative), EGFP-positive cells was counted and is expressed as the percentage of EGFP-positive cells in cultures transfected with plasmid, but without AS-ODN. Means of three independent experiments ± SEM are shown. *P < 0.05; **P < 0.005 (one group student t-test compared to con- trol without AS-ODN). 2580 P. A. ’t Hoen et al. (Eur. J. Biochem. 269) Ó FEBS 2002 Table 3. Effects of AS-ODNs on GST activity. COS-7 cells were transfected with 0.5 lg pEGFP, pEGFP-M1 or pEGFP-M2, together with 0.1 l M of the indicated AS-ODNs. AS-5 and AS-10 are directed against GSTM2; AS-6 is directed against GSTM1. At 22 h after transfection, GST activity in the protein lysates was determined by assaying CDNB conjugation over time. Conjugation rates are expressed as percentages of EGFP controls (column 2, 3 and 5), or percentages of the additional EGFP-M1-dependent (column 4) or EGFP-M2-dependent (column 6) GST activity, calculated by subtraction of the endogenous GST activity, which was determined in pEGFP-transfected cultures. Means of 10–12 determinations in three separate experiments ± SEM are shown. Statistical significance of the difference between AS-ODN-treated and untreated cultures are indicated: a P < 0.005, b P < 0.0001. Statistical significance of the difference between the effect of the AS-ODNs on EGFP–M1 and EGFP–M2 expression are indicated: c P ¼ 0.0013 (AS-5), d P < 0.0001 (AS-6), e P ¼ 0.0002 (AS-10). CDNB conjugation rate EGFP EGFP-M1 EGFP-M2 AS-ODN % of EGFP control % of EGFP control % of EGFP-M1 control % of EGFP control % of EGFP-M2 control – 100 ± 6 340 ± 16 100 ± 6 190 ± 11 100 ± 7 AS-5 93 ± 7 207 ± 4 b 51 ± 3 c 118 ± 6 b 23 ± 6 c AS-6 99 ± 6 110 ± 4 b 5±2 d 146 ± 6 a 57 ± 6 d AS-10 95 ± 5 244 ± 9 b 61 ± 4 e 125 ± 7 a 30 ± 6 e Fig. 4. Concentration-dependent inhibition of EGFP and EGFP–GSTM fusion protein expression by AS-ODNs. COS-7 cells were transfected with 0.5 lgofpEGFP(n), pEG- FP-M1 (j)orpEGFP-M2(d), together with the indicated concentrations of AS-1 (A), AS-2 (B), AS-5 (C), AS-6 (D) or AS-9 (E). AS- 1 and AS-6 are directed against GSTM1, AS-5 is directed against GSTM2, whereas AS-2 and AS-9 are complementary to both GSTM1 and GSTM2. At 22 h after transfection, cells were analysed for GFP expression (FL-1) and propidium iodide uptake (FL-3) by flow cytometry. The number of living (i.e. propi- dium iodide-negative), EGFP-positive cells was counted and is expressed as the percentage of EGFP-positive cells in cultures transfected with plasmid, but without AS-ODN. Means of three independent experiments ± SEM are shown. An unpaired student t-test was used to determine whether the effect on EGFP–M1 or EGFP–M2 expression was significantly dif- ferent from the effect on EGFP expression: *P <0.05;**P < 0.005. Ó FEBS 2002 EGFP-based selection of antisense sequences (Eur. J. Biochem. 269) 2581 The sensitivity of the inhibition towards mismatches in the target sequences appeared to be high. One mismatch (AS-7) was not sufficient to achieve complete isoform- specificity (the expression of the targeted and nontargeted isoform was reduced by 77 ± 6% and 45 ± 15%, respect- ively). The presence of three mismatches (AS-10), however, resulted in isoform-specifc inhibition of EGFP–GSTM fusion protein expression (85 ± 0.4% and 10 ± 0.3% reduction of targeted and nontargeted isoform, respect- ively). Interestingly, an AS-ODN against EGFP, described to be effective in HeLa cells [23], was totally ineffective in inhibiting the expression of either of the EGFP-containing proteins in our study. This may be attributed to the fact that the ODN was a phosphorothioate-capped PO-ODN. The sensitivity of these chimeras towards nucleolytic degrada- tion is higher than that of PS-ODNs, and depends on the cell-type used [27,28]. True antisense effects and nonantisense effects elicited by the ODNs were both found to be concentration-dependent. The IC 50 values of the most potent, specifically acting AS-ODNs were  0.2 l M . It should be noted that for most AS-ODNs, with the exception of AS-5, the concentration window where sequence-specific antisense effects were observed, was narrow. This was also found in other studies where PS-ODNs were used, and may be explained by the relatively low affinity of PS-ODNs for their target mRNA sequences together with the high incidence of nonantisense effects [1,24]. It is therefore of highest importance to evaluate, in each antisense study, the concentration–activity profile. The nature of the nonspecific effects elicited by PS-ODNs remains to be clarified. With the possible exception of AS-11, which contained only five mismatches with respect to the EGFP sequence, neither of the AS-ODNs against GSTM showed significant sequence homology with EGFP. Thus, the observed effects on EGFP expression are probably not caused by partial hybridization of the AS-ODNs with the EGFP mRNA. We cannot exclude that some of the nonspecific AS-ODNs decrease the transfection efficiency of the EGFP plasmids. However, from earlier studies it became apparent that sequence-dependent variations in cationic lipid-mediated transfection efficiencies were small, unless homo-oligonucleotides, such as A 18 , were applied [29,30]. More likely, sequence-dependent aptameric effects play a role. The negative charge on the sulfur atom may cause avid binding of the ODNs to key cellular proteins, e.g. proteins involved in mRNA translation [31,32]. The inhibition of EGFP–GSTM fusion protein expres- sion was reflected by a decrease in GST enzyme activity, as determined in a CDNB conjugation assay. AS-6 displayed potent and specific inhibition of GSTM1-1 enzyme activity. AS-5 and AS-10, directed against GSTM2, inhibited GSTM2-2 enzyme activity but showed also some effect on GSTM1-1 enzyme activity. This was not expected because the effects of these AS-ODNs on EGFP fusion protein expression were highly isoform-specific. The inhibitory effects cannot be explained by a general inhibition of GST activity, because the ODNs did not affect endogenous GST activity in COS-7 cells. Possibly, the presence of four (AS-5) and three (AS-10) mismatches with respect to the GSTM1 sequence results in partial hybridization with the GSTM1 mRNA and in some hindrance of the synthesis of full-length EGFP–M1 fusion proteins without induction of RNAse H-mediated cleavage and subsequent degradation of EGFP–M1 mRNA. In that case, the formation of the EGFP moiety is not affected, whereas the formation of the GSTM1-1 is. This would explain the higher isoform specificity of AS-5 and AS-10 in the EGFP assay, compared to the GST assay. In summary, we selected several effective antisense ODNs againstratGSTM1andGSTM2fromasetof15PS-ODNs in a novel, sensitive screening assay. The assay discriminates between effective AS-ODNs and ODNs that are ineffective or inhibit protein expression by nonantisense mechanisms. The effectiveness of the selected AS-ODNs will be evaluated further in rat hepatocytes and in vivo, potentially allowing the study of the effect of decreased GSTM expression on the toxicity and carcinogenicity of xenobiotics. REFERENCES 1. Crooke, S.T. & Bennett, C.F. (1996) Progress in antisense oligonucleotide therapeutics. Annu. Rev. Pharmacol. Toxicol. 36, 107–129. 2. Flanagan, W.M. & Wagner, R.W. (1997) Potent and selective gene inhibition using antisense oligodeoxynucleotides. Mol. Cell. Bio- chem. 172, 213–225. 3. Galderisi, U., Cascino, A. & Giordano, A. (1999) Antisense oligonucleotides as therapeutic agents. J. Cell Physiol. 181, 251–257. 4. Phillips, M.I.E. (1999) Antisense Technology, part A and B, Methods in Enzymology, (Abelson, J.N. & Simon, M.I., eds), Vol 313–314. Academic Press, San Diego, CA. 5. Agrawal, S. & Iyer, R.P. (1995) Modified oligonucleotides as therapeutic and diagnostic agents. Curr. Opin. Biotechnol. 6, 12–19. 6. Herdewijn, P. (2000) Heterocyclic modifications of oligonucleo- tides and antisense technology. Antisense Nucleic Acid Drug. Dev. 10, 297–310. 7. Monia, B.P., Johnston, J.F., Geiger, T., Muller, M. & Fabbro, D. (1996) Antitumor activity of a phosphorothioate antisense oligodeoxynucleotide targeted against C-raf kinase. Nat. Med. 2, 668–675. 8. Dean,N.M.,McKay,R.,Condon,T.P.&Bennett,C.F.(1994) Inhibition of protein kinase C-alpha expression in human A549 cells by antisense oligonucleotides inhibits induction of inter- cellular adhesion molecule 1 (ICAM-1) mRNA by phorbol esters. J. Biol. Chem. 269, 16416–16424. 9. Sohail, M. & Southern, E.M. (2000) Selecting optimal antisense reagents. Adv. Drug Deliv. Rev. 44, 23–34. 10. Commandeur, J.N.M., Stijntjes, G.J. & Vermeulen, N.P.E. (1995) Enzymes and transport systems involved in the formation and disposition of glutathione S-conjugates. Role in bioactivation and detoxication mechanisms of xenobiotics. Pharmacol. Rev. 47, 271–330. 11. Seidegard, J., Vorachek, W.R., Pero, R.W. & Pearson, W.R. (1988) Hereditary differences in the expression of the human glutathione transferase active on trans-stilbene oxide are due to a gene deletion. Proc. Natl Acad. Sci. USA 85, 7293–7297. 12. McWilliams, J.E., Sanderson, B.J., Harris, E.L., Richert-Boe, K.E. & Henner, W.D. (1995) Glutathione S-transferase M1 (GSTM1) deficiency and lung cancer risk. Cancer Epidemiol. Biomarkers Prev. 4, 589–594. 13. Zhong, S., Wyllie, A.H., Barnes, D., Wolf, C.R. & Spurr, N.K. (1993) Relationship between the GSTM1 genetic polymorphism and susceptibility to bladder, breast and colon cancer. Carcino- genesis 14, 1821–1824. 14. Wormhoudt, L.W., Commandeur, J.N.M. & Vermeulen, N.P.E. (1999) Genetic polymorphisms of human N-acetyltransferase, cytochrome P450, glutathione-S-transferase, and epoxide 2582 P. A. ’t Hoen et al. (Eur. J. Biochem. 269) Ó FEBS 2002 hydrolase enzymes: relevance to xenobiotic metabolism and toxi- city. Crit. Rev. Toxicol. 29, 59–124. 15. Nakachi, K., Imai, K., Hayashi, S. & Kawajiri, K. (1993) Poly- morphisms of the CYP1A1 and glutathione S-transferase genes associated with susceptibility to lung cancer in relation to cigarette dose in a Japanese population. Cancer Res. 53, 2994–2999. 16. Kihara, M., Kihara, M. & Noda, K. (1995) Risk of smoking for squamous and small cell carcinomas of the lung modulated by combinations of CYP1A1 and GSTM1 gene polymorphisms in a Japanese population. Carcinogenesis 16, 2331–2336. 17. Kawajiri, K., Watanabe, J., Eguchi, H. & Hayashi, S. (1995) Genetic polymorphisms of drug-metabolizing enzymes and lung cancer susceptibility. Pharmacogenetics 5, S70–S73. 18. Stucker,I.,Jacquet,M.,deWaziers,I.,Cenee,S.,Beaune,P., Kremers, P. & Hemon, D. (2000) Relation between inducibility of CYP1A1, GSTM and lung cancer in a French population. Phar- macogenetics 20, 617–627. 19. Zeng, G. (1998) Sticky-end PCR: new method for subcloning. Biotechniques 25, 206–208. 20. Habig, W.H., Pabst, M.J. & Jakoby, W.B. (1974) Glutathione S-transferases. The first enzymatic step in mercapturic acid for- mation. J. Biol. Chem. 249, 7130–7139. 21. Bradford, M.M. (1976) A rapid and sensitive method for the quantitation of microgram quantities of protein utilizing the principle of protein-dye binding. Anal. Biochem. 72, 248–254. 22. Hayes, J.D. & Pulford, D.J. (1995) The glutathione S-transferase supergene family: regulation of GST and the contribution of the isoenzymes to cancer chemoprotection and drug resistance. Crit. Rev. Biochem. Mol. Biol. 30, 445–600. 23.Helin,V.,Gottikh,M.,Mishal,Z.,Subra,F.,Malvy,C.& Lavignon, M. (1999) Cell cycle-dependent distribution and specific inhibitory effect of vectorized antisense oligonucleotides in cell culture. Biochem. Pharmacol. 58, 95–107. 24. Monia, B.P., Johnston, J.F., Ecker, D.J., Zounes, M.A., Lima, W.F. & Freier, S.M. (1992) Selective inhibition of mutant Ha-ras mRNA expression by antisense oligonucleotides. J. Biol. Chem. 267, 19954–19962. 25. Wiechen, K., Zimmer, C. & Dietel, M. (1998) Selection of a high activity c-erbB-2 ribozyme using a fusion gene of c-erbB-2 and the enhanced green fluorescent protein. Cancer Gene Ther. 5, 45–51. 26. Stein, C.A. (2001) The experimental use of antisense oligonucleo- tides: a guide for the perplexed. J. Clin. Invest. 108, 641–644. 27. Hoke,G.D.,Draper,K.,Freier,S.M.,Gonzalez,C.,Driver,V.B., Zounes, M.C. & Ecker, D.J. (1991) Effects of phosphorothioate capping on antisense oligonucleotide stability, hybridization and antiviral efficacy versus herpes simplex virus infection. Nucleic Acids Res. 19, 5743–5748. 28. Monia, B.P., Johnston, J.F., Sasmor, H. & Cummins, L.L. (1996) Nuclease resistance and antisense activity of modified oligonu- cleotides targeted to Ha-ras. J. Biol. Chem. 271, 14533–14540. 29. Conrad, A.H., Behlke, M.A., Jaffredo, T. & Conrad, G.W. (1998) Optimal lipofection reagent varies with the molecular modifica- tions of the DNA. Antisense Nucleic Acid Drug Dev. 8, 427–434. 30. Meidan, V.M., Glezer, J., Amariglio, N., Cohen, J.S. & Barenholz, Y. (2001) Oligonucleotide lipoplexes: the influence of oligonu- cleotide composition on complexation. Biochim. Biophys. Acta 1568, 177–182. 31. Brown, D.A., Kang, S.H., Gryaznov, S.M., DeDionisio, L., Heidenreich, O., Sullivan, S., Xu, X. & Nerenberg, M.I. (1994) Effect of phosphorothioate modification of oligodeoxynucleotides on specific protein binding. J. Biol. Chem. 269, 26801–26805. 32. Bijsterbosch,M.K.,Rump,E.T.,DeVrueh,R.L.,Dorland,R., vanVeghel,R.,Tivel,K.L.,Biessen,E.A.,vanBerkel,T.J.& Manoharan, M. (2000) Modulation of plasma protein binding and in vivo liver cell uptake of phosphorothioate oligodeoxynucleo- tides by cholesterol conjugation. Nucleic Acids Res. 28, 2717–2725. Ó FEBS 2002 EGFP-based selection of antisense sequences (Eur. J. Biochem. 269) 2583 . CCATGCCTATGATACTGGGAT-3¢ GSTM1-revA 5¢- CTAAAGATGAGACAGGCCTGG-3¢ GSTM1-revB 5¢-GATCCTAAAGATGAGACAGGCCTGG-3¢ GSTM2-forA 5¢-AATTCGATGCCTATGACACTGGGTTAC-3¢ GSTM2-forB 5¢- CGATGCCTATGACACTGGGTTAC-3¢ GSTM2-revA. Selection of effective antisense oligodeoxynucleotides with a green fluorescent protein-based assay Discovery of selective and potent inhibitors of glutathione S -transferase Mu expression Peter. 4 AS-6 ACAAAGCATGATGAGCTGCA CDS 326–345 – 8 AS-7 GAGTA GAGCTTCATCTTCTC CDS 397–426 – 1 AS-8 ACTGGTCAAGAATGTCATAA CDS 480–499 – 7 AS-9 CAGGTTTGGGAAGGCGTCCA CDS 524–543 524–543 0 AS-10 CAGGCCCTC AAACCGAGCCA