Invasion of the CAG triplet repeats by a complementary peptide nucleic acid inhibits transcription of the androgen receptor and TATA-binding protein genes and correlates with refolding of an active nucleosome containing a unique AR gene sequence.

The DNA sequence of the genes for the androgen receptor (AR) and TATA-binding protein (TBP), like many other genes encoding transcription factors, contains a series of tandem CAG repeats. Here we explore the capacity of complementary peptide nucleic acids (PNAs) to invade the CAG triplets of the AR and TBP genes in human prostatic cancer cells and show that the PNAs readily entered the nuclei of lysolecithin-permeabilized cells and effectively inhibited sense transcription of unique AR and TBP DNA sequences downstream of the site of PNA.DNA hybridization, but not upstream of that site. These PNAs had little or no effect on transcription of the c-myc gene, which lacks a CAG triplet domain. Conversely, a PNA complementary to a unique sequence of the c-myc gene did not inhibit transcription of the AR or TBP genes but did inhibit c-myc transcription. Comparisons of PNA effects on sense and antisense transcription of the AR, TBP, and c-myc genes confirm that progression of the RNA polymerase complex beyond the site of PNA.DNA hybridization is impaired in both directions. Suppression of the AR gene results in refolding of a transcriptionally active nucleosome containing a unique 17-mer AR DNA sequence.

The DNA sequence of the genes for the androgen receptor (AR) and TATA-binding protein (TBP), like many other genes encoding transcription factors, contains a series of tandem CAG repeats. Here we explore the capacity of complementary peptide nucleic acids (PNAs) to invade the CAG triplets of the AR and TBP genes in human prostatic cancer cells and show that the PNAs readily entered the nuclei of lysolecithin-permeabilized cells and effectively inhibited sense transcription of unique AR and TBP DNA sequences downstream of the site of PNA⅐DNA hybridization, but not upstream of that site. These PNAs had little or no effect on transcription of the c-myc gene, which lacks a CAG triplet domain. Conversely, a PNA complementary to a unique sequence of the c-myc gene did not inhibit transcription of the AR or TBP genes but did inhibit c-myc transcription. Comparisons of PNA effects on sense and antisense transcription of the AR, TBP, and c-myc genes confirm that progression of the RNA polymerase complex beyond the site of PNA⅐DNA hybridization is impaired in both directions. Suppression of the AR gene results in refolding of a transcriptionally active nucleosome containing a unique 17-mer AR DNA sequence.
Peptide nucleic acids (PNAs) 1 are synthetic structural homologues of DNA and RNA in which the entire phosphate-sugar backbone of the polynucleotide has been replaced by a flexible polyamide backbone consisting of 2-aminoethyl glycine units. Each unit is linked to an appropriate purine or pyrimidine base to create the sequence required for hybridization to the targeted nucleic acid (1)(2)(3). The absence of phosphate groups in the PNA molecule facilitates its invasion of negatively charged DNA duplexes containing the complementary base sequences (4 -6). Under appropriate conditions, PNAs show greater discrimination and form more stable hybrids with DNA than the corresponding DNA⅐DNA duplexes (1, 4 -8). When a PNA mol-ecule invades a targeted DNA sequence, one strand of the DNA is displaced (1,5,6,9), whereas the PNA binds quickly to its complementary DNA sequence by Watson-Crick base pairing (7). Studies of homopyrimidine and homopurine PNAs show that this rapid and highly specific association with the complementary DNA strand is followed by the addition of a second PNA molecule to the PNA⅐DNA duplex to form a very stable (PNA) 2 -DNA triplex, while the noncomplementary DNA strand is left in the single-stranded state (8 -10).
PNA invasion of DNA duplex strands to form specific and stable PNA⅐DNA hybrids has profound implications for both positive and negative control of transcription. For example, it has been shown that DNA loops displaced as a consequence of PNA binding act as artificial transcription promoters (11), but PNA binding to the transcribed strand of a targeted DNA sequence blocks transcript elongation beyond the site of PNA⅐DNA hybridization (5,(12)(13)(14).
We have tested the effects of PNAs complementary to the CAG triplet repeats that occur in many genes involved in transcriptional control. More than 33 transcription factors, including the human TATA-binding protein (TBP) (15)(16)(17) and the androgen receptor (18 -20), are characterized by the presence of polyglutamine tracts containing more than 20 residues encoded by tandem CAG triplet repeats. We have shown previously that a biotinylated PNA targeted to CAG repeats will strand invade those DNA sequences in their transcriptionally active states in intact chromatin (21). By combining the use of the biotinylated CAG-specific PNA with techniques that permit a clear separation of transcriptionally active and inactive chromatin restriction fragments (22), it became possible to capture the active chromatin fragments containing the stable PNA⅐DNA hybrids on streptavidin-agarose magnetic beads. Those chromatin fragments were shown to contain the transcriptionally active DNA for the TBP of human colonic cancer cells. Moreover, the selectivity of the PNA probe for the CAG triplet repeats was confirmed by the fact that the streptavidinbound chromatin fragments did not contain DNA for the c-myc protooncogene, which is amplified and actively transcribed in the same cells, but lacks a domain of tandem CAG triplets (21).
We have now tested the effects of CAG-specific and c-mycspecific PNAs on AR, TBP, and c-myc transcription in human prostatic cancer cell lines, and we now show that each PNA can selectively target its complementary DNA sequence in the intact chromatin of permeabilized cells. Detailed comparisons of PNA inhibitions of sense and antisense transcription of the AR, TBP, and c-myc genes confirm that progression of the RNA polymerase complex beyond the site of PNA⅐DNA hybrid formation is impaired in both directions.

MATERIALS AND METHODS
Cell Culture-Human prostatic carcinoma cell lines LNCaP and DU-145 were obtained from the American Type Culture Collection (ATCC, Rockville, MD). The LNCaP cells were cultured in RPMI 1640 medium (Life Technologies, Inc.) supplemented with 10% fetal bovine serum (JRH Biosciences, Lenexa, KS) and 20 ng/ml gentamycin (Life Technologies, Inc.) at 37°C in an atmosphere of 5% CO 2 in air. DU-145 prostatic cancer cells were cultured in Eagle's minimal essential medium (Sigma) supplemented with 10% fetal bovine serum and gentamycin. In all cases, cells in log phase growth were harvested at 5 ϫ 10 5 cells/ml.
PNA Effects on Transcription-An 18-mer PNA, 5Ј-CAGCAGCAG-CAGCAGCAG-Lys-CONH 2 -3Ј, complementary to the transcribed strand of the AR triplet repeat domain was synthesized and linked to biotin at its amino-terminal through two 8-amino-3,6-dioxaoctanoyl linkers to create a distance between the PNA and the biotin label. A second PNA, with the same sequence, but lacking the biotinylated linker was also tested, with identical results on transcription of the AR and TBP genes (custom synthesis by Per-Septive Biosystems, Framingham, MA). The purity of the biotinylated PNAs was confirmed by reverse phase chromatography and mass spectrographic analysis (M r 5,568 and 6,325. Ci of [␣-32 P]UTP (specific activity 3000 Ci/mmol; DuPont NEN) was added, and transcription was allowed to proceed for 30 min at 37°C. Following incubation the samples were digested with RNase-free DNase (Boehringer Mannheim) and proteinase K (Boehringer Mannheim), and total RNA was extracted and purified as described (23); or alternatively, the RNAs were precipitated by lysing the cells in a cationic detergent (Catrimox-14; Iowa Biotechnology Corp., Corville, IA) and converting the RNA to the ethanol-insoluble lithium salt (24). The 32 P-labeled RNA samples were dissolved in buffer D (1 mM TES, pH 7.4, 10 mM EDTA, 0.2% (w/v) SDS, 0.3 M NaCl) and hybridized to DNA probes for AR, TBP, and c-myc blotted on Nytran Plus filters, as described below.
Four DNA probes, specific for unique AR gene sequences, were employed in order to compare PNA effects on sense and antisense transcription in regions downstream and upstream of the targeted CAG triplet domain. The 23-mer oligonucleotide 5Ј-dCTCTTTTGAAGAA-GACCTTGCAG-3Ј was used to detect sense transcripts of the unique sequence (1890 -1912) of human AR cDNA (25), located downstream of the site of PNA⅐DNA hybridization. The probe for the corresponding antisense transcripts was 5Ј-dCTGCAAGGTCTTCTTCAAAAGAG-3Ј. The oligo-DNA probes for the sense and antisense transcripts of the unique AR sequence (163-185) upstream of the PNA⅐DNA hybridization site were 5Ј-dCCCAGCCCTAACTGCACTTCCAT-3Ј and 5Ј-dATG-GAAGTGCAGTTAGGGCTGGG-3Ј, respectively. Five probes were used for the detection of c-myc transcripts. The first was a 9-kilobase EcoRI-HindIII fragment in plasmid pHSR, which encodes the full human c-myc gene. The DNA oligonucleotide probes for sense and antisense transcripts of the unique c-myc sequence (7177-7213) downstream of the myc-targeting PNA hybridization site (4528 -4544) were 5Ј-dCG-CACAAGAGTTCCGTAGCTGTTCAAGTTTGTGTTTC-3Ј and 5Ј-dGAAACACAAACTTGAACAGCTACGGAACTCTTGTGCG-3Ј, respectively. The DNA probes for sense and antisense transcripts of the unique c-myc upstream sequence (2303-2333) were 5Ј-dGGGGGTCCT-CAGCCGTCCAGACCCTCGCATT-3Ј and 5Ј-dAATGCGAGGGTCTG-GACGGCTGAGGACCCCC-3Ј, respectively. In tests for the effects of PNAs targeting the CAG domain (270 -371) of the human TBP gene, the probes for sense and antisense transcripts of the unique downstream sequence (905-932) were 5Ј-dACTACTAAATTGTTGGTGGGTGAG-CACA-3Ј, and 5Ј-dTGTGCTCACCCACCAACAATTTAGTAGT-3Ј, respectively. The corresponding probes for transcripts of the unique TBP sequence (99 -126) upstream of the CAG domain were 5Ј-dGTAAGGT-GGCAGGCTGTTGTTCTGATCCAT-3Ј and 5Ј-dATGGATCAGAACAA-CAGCCTGCCACCTTAC-3Ј (15). In all experiments, 2 nmol of each DNA oligonucleotide were blotted onto Nytran Plus filters, pore size 45 m (Schleicher & Schuell) and exposed to the radioactive RNA samples for 36 h at 65°C, as described (23). After exposure for appropriate periods and development of the films in Kodak X-Omat M4 developer, the autoradiograms were scanned with a laser densitometer and peak areas corresponding to each slot-blot were integrated. In all experiments, the relationship between 32 P activity of the RNA in individual slot-blots and the intensity of the corresponding autoradiogram were determined by comparison with standard curves in which photographic density is plotted for a series of dilutions of a standard 32 P-labeled salmon sperm DNA solution applied to the same membrane as the experimental samples. This established the range of linear response and clearly indicated where reblotting or correction for overexposure was needed.
Isolation of Nuclei and Separation of Transcriptionally Active and Inactive Nucleosomes-All procedures were carried out at 4°C unless otherwise specified. Cells were collected by centrifugation at 500 ϫ g for 10 min and washed three times in calcium-free Dulbecco's phosphatebuffered saline (Life Technologies, Inc.) containing 5 mM Na ϩ butyrate to suppress histone deacetylase activity during cell lysis and isolation of the nuclei (26,27). To lyse the cells, they were suspended in buffer E (10 mM Tris-HCl, pH 7.4, 10 mM NaCl, 3 mM MgCl 2 , 0.5% Nonidet P-40, 5 mM Na ϩ butyrate, 0.1 mM PMSF, and 0.1 mM 1,2-epoxy-3-(p-nitrophenoxy)propane) (EPNP, Kodak, Rochester, NY) for 5 min at 0°C, with gentle mixing by pipetting. The cells were collected by centrifugation at 500 ϫ g for 10 min, resuspended in buffer E, and recentrifuged. They were then suspended in buffer F (10 mM Tris-HCl, pH 7.4, 25 mM NaCl, 25 mM KCl, 5 mM MgCl 2 , 5 mM Na ϩ butyrate, 0.35 M sucrose, 0.1 mM PMSF, 0.1 mM EPNP). After centrifugation as before, the cells were resuspended in buffer F and broken by shearing in a glass Dounce-type homogenizer with a type B pestle (Kontes, Inc., Vineland, NJ). The homogenate was centrifuged at 500 ϫ g for 10 min, and the nuclear pellet was washed three times by resuspension and centrifugation in buffer F. The nuclei were suspended in buffer F at a concentration of 1 mg DNA/ml and equilibrated at 37°C for 10 min. Micrococcal nuclease (Boehringer Mannheim) was added at 10 units/ml, and the reaction was started by the addition of CaCl 2 to 0.5 mM. After incubation at 37°C for 7 min, endonuclease action was halted by rapid chilling and addition of Na 2 EGTA to a final concentration of 2.5 mM. The suspension was centrifuged at 10,000 ϫ g for 20 min. The resulting supernatant, containing the nucleosomes released by endonuclease digestion, was analyzed for its DNA content by measuring the A 260 /ml and by the Hoechst dye-binding assay (28). Under these conditions, 11% of the total DNA of the LNCaP nuclei was released into the supernatant (average of 4 experiments).
To separate the transcriptionally active from inactive nucleosomes, the supernatant was brought to 5 mM in Na 2 EDTA, and applied to an organomercurial-agarose column (Affi-Gel 501, Bio-Rad). Elution of the unbound nucleosomes was carried out in buffer G (10 mM Tris-HCl, pH 7.4, 25 mM NaCl, 25 mM KCl, 5 mM Na ϩ butyrate, 5 mM Na 2 EDTA, 0.1 mM PMSF, 0.1 mM EPNP, 2% sucrose) until the A 260 of the eluate returned to base line. The remaining nucleosomes, which were bound to the Hg ϩ column through their accessible SH groups, were released by elution in buffer G containing 20 mM dithiothreitol. DNA was purified from the unbound and Hg ϩ -bound nucleosome fractions and analyzed for AR gene content by slot-blot hybridizations (29) to a 32 P-labeled oligonucleotide probe for the unique downstream AR sequence 5Ј-dCTTTTGAAGAAGACCTT as described (21).

Inhibition of AR Gene Transcription by PNA Invasion of CAG
Triplet Repeats-The mRNA of the androgen receptor gene is characterized by a series of tandem CAG triplets that encode a tract of glutamine residues in the receptor protein. To test whether a PNA targeted to the sense strand of the CAG domain (nucleotides 334 -390) (i.e. the tandem CTG sequence serving as template for the CAG repeats in AR mRNA) would inhibit transcription beyond the site of PNA⅐DNA hybridization, we selected a prostatic cancer cell line, LNCaP, which had been shown to contain both the androgen receptor protein and its mRNA (30). Run-on transcription experiments were carried out on cells that were gently permeabilized with lysolecithin under conditions shown to preserve nuclear RNA synthetic activity during nuclear transplantation experiments (31). The permeabilized cells were exposed to increasing concentrations of an 18-mer biotin-free PNA complementary to the CAG-encoding DNA strand. After incubation at 37°C for 30 min to permit PNA entry into nuclear chromatin (21), ATP, CTP, GTP, and [␣-32 P]UTP were added, and run-on transcription experiments were carried out as described (see "Materials and Methods"). Total RNA was purified from each sample and analyzed for its content of 32 P-labeled transcripts of a unique AR downstream sequence (nucleotides 1890 -1912) by slot-blotting to the complementary DNA oligonucleotide probe. The results for cells exposed to 0, 1, 5, 10, and 20 M PNA show a progressive inhibition of AR transcript elongation at PNA concentrations starting at 5 M, reaching 71% at 10 M, and 81% inhibition at 20 M PNA, as determined by densitometric analysis of the slot-blots (Fig. 1A). As expected, PNA invasion of the CAG domain of the AR gene strongly inhibited transcription at the binding site, reducing CAG-repeat transcripts by 93% at 20 M PNA (average of eight determinations).
Other genes containing CAG triplet repeats were also affected by the same PNA. For example, PNA invasion of the tandem CAG triplets at nucleotides 270 -371 of the human TATA-binding protein gene inhibited transcription of a unique TBP DNA sequence (nucleotides 905-932) downstream of the hybridization site by 80% at 10 M and by 88% at 20 M PNA (Fig. 1B).
The specificity of PNA targeting to different genes was confirmed in two ways. First, it was shown that sense transcription of the c-myc gene, which lacks a domain of CAG triplet repeats, was not inhibited in the same cells under identical conditions, only minor variations (Ϯ7%) were observed in four determinations of [␣-32 P]UMP incorporation into transcripts of two unique c-myc DNA sequences (nucleotides 2303-2333 and 7177-7213). In similar experiments, using both LNCaP and DU-145 prostatic cancer cells, a 21-mer PNA targeted to the antisense strand of the AR CAG-repeat domain was found to have no inhibitory effect on c-myc transcription (Fig. 3A). Conversely, exposure of the LNCaP cells to a PNA complementary to a unique sequence in the second exon of the c-myc gene had little or no effect on transcription of a unique sequence (nucleotides 1890 -1912) of the AR gene but did inhibit transcription of a unique DNA sequence in the third c-myc exon (Fig. 2C). Such specificity in PNA interactions with intact chromatin is in accord with earlier results on the selective binding and recovery of PNA⅐DNA hybrids of the TBP gene in colonic cancer cells (21) (see "Discussion").
Effects of PNA on Sense and Antisense Transcription of the AR Gene-The previous experiments were focused on the inhibitory effects of PNAs complementary to the sense strand of the AR CAG triplet domain on sense transcription of a unique sequence downstream of the PNA⅐DNA hybridization site (Fig.  1A). More detailed studies of PNA effects on AR transcription were initiated when run-on transcription experiments revealed the presence of 32 P-labeled transcripts of the AR antisense strand in cells that had never been exposed to PNA. The antisense transcripts were clearly evident in slot-blot hybridizations of the 32 P-labeled RNA samples to complementary DNA oligonucleotide probes ( Fig. 2A).
To determine whether PNA binding to the sense strand of the AR gene also affected antisense transcription, we compared the synthesis of sense and antisense transcripts at unique AR DNA sequences located downstream and upstream of the CAGrepeat domain. The results confirmed that 20 M PNA inhibited transcription of the downstream sense strand by 88%, but they also revealed that PNA binding to the sense strand led to an increase in 32 P-labeled transcripts of a unique AR sequence upstream of the PNA binding site (Fig. 2A). The opposite results were obtained for antisense transcription in which PNA binding to the CAG triplet domain inhibited transcription of the upstream AR sequence by 82-86% but stimulated downstream transcription (Fig. 2B). Simultaneous assays for transcripts of the CAG triplet domain show that 20 M PNA effectively inhibited both sense and antisense transcription of the tandem repeats by 93% (average of nine determinations). The contrasting effects of PNAs on sense and antisense transcription were also evident in the TBP gene, where PNA invasion of the CAG triplet domain inhibited downstream transcription of the sense strand by 88% (average of seven determinations) but stimulated transcription of a unique TBP sequence upstream of the PNA⅐DNA hybrid (Fig. 2B). These results are fully compatible with a model in which PNA binding to the CAG triplet domain impedes progression of the RNA polymerase complex in both directions. Further support for this view is provided by experiments using a 17-mer complementary PNA to target the sense strand of the unique sequence (nucleotides 4528 -4544) of the human c-myc gene, which is known to be transcribed in both sense and antisense directions (32). At a PNA concentration of 20 M, transcription of a unique c-myc downstream sequence (nucleotides 7177-7213) was inhibited by 87% in the sense direction and stimulated in the antisense direction. Conversely, sense transcription of a unique upstream sequence (nucleotides 2303-2333) was stimulated by the PNA, whereas antisense transcription was inhibited by 94% (Fig. 2C).
Distribution of a Unique Sequence of the AR Gene in Transcriptionally Active and Inactive Chromatin-The human prostatic cancer cell lines LNCaP and DU-145 have been shown to differ widely in their response to androgens. The difference can be attributed to the fact that the LNCaP cells contain the androgen receptor protein and its mRNA, whereas DU-145 cells have undetectable amounts of either the receptor or its mRNA (30). That the disparity in AR mRNA contents of LNCaP and DU-145 cells reflects differences in their rates of transcription of the AR gene was confirmed in run-on transcription experiments, measuring the intensity of the slot-blots of the 32 P-labeled RNA to DNA oligonucleotide probes for a unique AR sequence downstream of the CAG-repeat domain (Fig. 3A). Densitometric analyses indicated that the AR transcript content of the DU-145 cells was less than 3% of that observed in the RNA of LNCaP cells. This result cannot be attributed to a general loss of transcriptional activity of the DU-145 cells because the same RNA preparations gave a strong hybridization signal when probed with human c-myc DNA (Fig. 3A). Active transcription of the c-myc gene was also noted in the LNCaP cells (Fig. 3A). As both types of prostatic cancer cells actively transcribe the c-myc gene, which lacks a domain of CAG triplet repeats, a full-length c-myc cDNA probe was used as a negative control in hybridization experiments to test the effects of the CAG-specific PNA on AR transcription in the two cell lines. The results for LNCaP cells show an 83% reduction in AR unique transcripts at 10 M PNA and 89% reduction at 20 M (average of three experiments) (Fig. 3A). Under the same conditions, 10 M PNA inhibited AR gene transcription in the DU-145 cells to the same extent as in the LNCaP cells (Fig. 3A).
The disparity in the intensity of the hybridization signals of LNCaP and DU-145 cells also offers an opportunity to correlate the transcriptional activity with alterations in chromatin structure in the same defined region of the gene. To test whether this difference in AR mRNA content in the two cell lines is reflected in the nucleosome structure of their AR genes, we compared the distributions of the same unique sequence (nucleotides 1890 -1912) of human AR DNA in the transcription-

FIG. 3. Correlation between rates of transcription and the unfolded state of nucleosomes of the AR gene in LNCaP and DU-145 prostatic cancer cells.
A, slot-blot hybridizations of 32 P-labeled RNAs from control cells to a DNA-oligonucleotide probe for a unique AR gene sequence downstream of the CAG triplet domain. Note the contrast in levels of 32 P-labeled AR transcripts in these two cell lines. Exposure of each cell line to a PNA targeted to the CAG triplet repeats strongly inhibited sense transcription of the AR gene downstream of the PNA binding site but had no effect on c-myc transcription in the same cells. B, distribution of AR DNA in transcriptionally active and inactive nucleosomes of LNCaP and DU-145 cells. Nucleosomes released in a limited micrococcal nuclease digestion of isolated nuclei were fractionated by affinity chromatography on mercurated agarose columns. DNA was isolated from each nucleosome fraction and probed for its AR gene content by slot-blot hybridizations to a 32 P-labeled 17-mer DNA probe for the same unique AR sequence. Equal numbers of LNCaP cells were exposed to one of three PNAs. Two were 18-mer CAG PNAs with the same base sequence, but differing in their modification by biotin, are shown. Both were targeted to the CAG triplet domains of the AR gene (A) and TBP gene (B). The third PNA was complementary to a unique DNA sequence of the c-myc gene (C). After run-on transcription experiments, the 32 P-labeled RNA samples were isolated and slot-blotted to appropriate oligo-DNA probes for sense and antisense transcripts of unique sequences downstream and upstream of their PNA binding sites. Intensities of the hybridization signals are compared for control cells (solid bars), for cells treated with the unmodified 18-mer CAG PNA (lightly striped bars), or its biotinylated form (darkly striped bars), and for cells treated with the 17-mer c-myc PNA (stippled bars). Note that in every case, the PNA inhibited transcription of the sense strand downstream of its binding site on the gene, while transcription was stimulated upstream of the PNA block. The opposite effect was observed for antisense transcription, in which progression of the RNA polymerase complex was not impaired until it reached the PNA binding site, and upstream transcription was inhibited. ally active and inactive nucleosomes of each cell type.
The chromatographic separation of active and inactive nucleosomes is based on chemical and electron microscopic evidence that the nucleosome core unfolds during transcription to reveal the previously inaccessible cysteinyl-SH groups of histone H3 molecules located at the center of the core (33). Consequently, mixtures of active and inactive nucleosomes, released from isolated nuclei by micrococcal nuclease digestion, can be separated by entrapping the unfolded, SH-reactive nucleosomes on an organomercurial-agarose column (34). After washing the column to remove the unbound nucleosomes, the SH-reactive nucleosomes are rapidly released from the mercurated support by elution in dithiothreitol (35,36). The success of mercury-affinity chromatography in separating active from inactive nucleosomes has been demonstrated in organisms as diverse as humans, rodents, yeast, and Physarum (35)(36)(37)(38)(39). Of particular significance are observations showing that the nucleosome core is a dynamic structure and subject to rapid and reversible changes. For example, in the early response of quiescent 3T3 fibroblasts to growth factors, the capture of nucleosomes containing the c-fos and c-myc genes on mercuratedagarose columns reflects, with accuracy, both the timing and extent of transcription of each gene, as monitored by run-on transcription experiments (37). Other experiments have shown that the active nucleosomes refold within minutes and are no longer retained by the Hg ϩ column when transcription is blocked by ␣-amanitin (38,39).
The same chromatographic procedure was used to fractionate the nucleosomes of DU-145 cells. Nucleosomes released during a limited micrococcal nuclease digestion of the isolated nuclei were applied to an organomercurial-agarose column. The unbound nucleosomes were washed from the column and collected by centrifugation, whereas the Hg ϩ -bound nucleosomes were subsequently eluted in one step with 20 mM dithiothreitol (34,35). DNA analyses of the unbound and Hg ϩbound fractions indicated that the bound nucleosomes contained approximately 20% of the total nucleosomes applied to the column, in agreement with many earlier experiments on other cell types (34 -39). Total DNA was purified from each nucleosome fraction and analyzed for its content of AR DNA by Southern blot hybridizations to the 32 P-labeled oligonucleotide probe for the unique AR sequence, 5Ј-dCTTTTGAAGAAGAC-CTT (25). The results show only a very faint hybridization signal for the DNA of the dithiothreitol-eluted nucleosomes of DU-145 cells (Fig. 3B), amounting to less than 3% of the signal from the unbound nucleosomes. This paucity of AR DNA in the unfolded, transcriptionally active nucleosome fraction of DU-145 cells is in agreement with earlier observations on the deficiency of AR mRNA in this cell line (30), and with the results of run-on transcription experiments (Fig. 3A), all findings indicate that a block in transcription is the basis for the absence of the androgen receptor in DU-145 cells.
A similar fractionation of the nucleosomes of LNCaP cells by mercury-affinity chromatography also resulted in a distribution of 80% unbound versus 20% Hg ϩ -bound nucleosomes, but when equal amounts of DNA from each fraction were analyzed for AR DNA content by hybridization to the 32 P-labeled oligonucleotide probe for the unique AR sequence, a very substantial signal for the AR gene was seen in the Hg ϩ -bound nucleosomes (Fig. 3B). Densitometric comparisons of the slot-blots indicated a 154% enrichment of the AR gene in the Hg ϩ -bound nucleosomes of the LNCaP cells, as compared with the unbound nucleosomes. (The strong hybridization signal for the AR gene in the unbound fraction of the LNCaP nucleosomes is consistent with the fact that the AR gene is expressed at a relatively low level in the prostate (40).) The result for the Hg ϩ -bound nucleosomes of LNCaP cells contrasts with the 3% recovery observed in the Hg ϩ -bound nucleosomes of DU-145 cells (Fig. 3B). Thus, the differences in the recovery of this specific sequence of the AR gene in the thiol-reactive nucleosomes of these two cell types provide further evidence that unfolding of the nucleosome cores accurately reflects the transcriptional state of the gene. It should be pointed out that the recovery of the AR gene in the Hg ϩ -bound, transcriptionally active nucleosome fraction was detected by hybridization to a specific AR DNA sequence, and it follows that the unique 17-mer AR DNA sequence occurs within a nucleosome which unfolds during transcription to reveal the histone H3-thiol groups at the center of the core.

DISCUSSION
Over 33 transcription factors are characterized by domains of tandem glutamine residues encoded by CAG triplet repeats. Expansion of the CAG repeats of the androgen receptor gene has been correlated with the incidence and severity of the Kennedy disease (spinal and bulbar muscular atrophy). Recent studies have shown that changes in the size and position of the polyglutamine tracts of the androgen receptor strongly influence its transactivation functions. For example, progressive expansion of the CAG repeats in human AR DNA diminishes transcription from androgen-responsive reporter genes (41,42). Conversely, elimination of the CAG tracts from human and rat androgen receptor genes results in an elevation of transcriptional activation (41). How such differences in the number of glutamine repeats affect the functions of the androgen receptor and other transcription factors is not known, but recent x-ray and molecular modeling studies by Perutz et al. (43) point out the potential of paired polyglutamine tracts to act as "polar zippers," possibly joining transcription regulatory factors on separate DNA segments (43).
Mutations affecting the size of the CAG domains of the androgen receptor gene have been noted in prostatic cancer (44). The present experiments have focused on the AR gene in prostatic cancer cell lines as a model for further analysis of the potential role of PNAs in the control of gene expression in malignant cells. We have shown that CAG-specific PNAs can penetrate the chromatin and inhibit transcription of the AR and TBP genes beyond the site of PNA⅐DNA hybridization, without a corresponding effect on transcription of the c-myc gene, while PNA targeted to a unique sequence of the c-myc gene inhibits its transcription without affecting the synthesis of AR or TBP mRNAs. The PNA effect depends on the direction of movement of the RNA polymerase complex. Targeting of the sense strand of the AR, TBP, and c-myc genes inhibited transcription of downstream DNA sequences, but it had no inhibitory effect on RNA polymerase transit through DNA sequences upstream of the PNA binding site. Instead, upstream transcription of the sense strand was stimulated in all three genes (Fig. 2), possibly due to continued recruitment and accumulation of active RNA polymerase complexes ahead of the PNA⅐DNA block. This could also explain the opposite results obtained for antisense transcription, in which RNA synthesis was stimulated below the site of PNA⅐DNA hybridization and inhibited above it (Fig. 2).
All our results show that PNA binding to the sense strand of the AR, TBP, and c-myc genes can inhibit transcription equally well on both strands of the DNA template. Such parity was not observed in studies of PNA effects on transcript elongation in vitro, in which PNA bound to the nontranscribed strand of a plasmid was only half as effective in chain termination (12). We attribute the stronger PNA response in our experiments to the 20-fold higher PNA concentrations used in the permeabilized cells, to the probability that the RNA polymerase II elongation complex in vivo is much larger than that of a reconstituted Pol II complex and less likely to pass by the PNA block, and to the additional constraints imposed by the nucleosomal structure of the chromatin templates.
The present results on the specificity of PNA effects on transcription of the AR, TBP, and c-myc genes are in accord with the observation that a PNA targeted to the CAG triplet repeats can enter and specifically capture chromatin restriction fragments containing the TBP gene (21), and they lend further support to the view that appropriately designed PNAs have great potential as antisense chemotherapeutic agents, especially in view of the fact that PNAs are extremely refractory to digestion by nucleases and proteases (45). However, it should be noted that PNAs targeted to transcription factors with multiple transactivation functions, such as TBP, c-myc, and AR, could have deleterious effects throughout the genome.