Transcriptional State of the Mouse Mammary Tumor Virus Promoter Can Affect Topological Domain Size in Vivo *

Unrestrained DNA supercoiling and the number of topological domains were measured within a 1.8 megabase pair chromosomal region consisting of about 200 tandem repeats of a mouse mammary tumor virus promoter-driven ha-v-rasgene. When uninduced, unrestrained negative supercoiling was organized into 32-kilobase pair (kb) topological domains. Upon induction, DNA supercoiling throughout the region was completely relaxed. Supercoiling was detected, however, when elongation was blocked before or following induction. The formation of transcription initiation complexes upon addition of dexamethasone decreased the domain size to 16 kb. During transcription the domain size was 9 kb, the length of one repeat. These results suggest that topological domain boundaries can be “functional” in nature, being established by the formation of activated and elongating transcription complexes.

Topological domains, often referred to as chromosomal loops, have been detected in many organisms including Escherichia coli (1), Drosophila (2), and humans (3) (for review see Ref. 4). The formation of a topological domain requires restraint of the DNA helix such that rotation of one strand of the DNA double helix around the other is prevented. Consequently, DNA within a topological domain can contain a linking number deficit that is manifest as unrestrained superhelical energy (for review see Ref. 5). Domain boundaries may result from the attachment of the DNA at specialized sites (i.e. MARs 1 or SARs) (6, 7) onto the nuclear matrix. Recently a bipartite sequence element has been identified within matrix attachment regions that may be im-portant in chromosome organization or function via these sites (8). Alternatively, topological domain boundaries may be functional in nature resulting from the attachment of functional proteins such as RNA or DNA polymerase complexes to the nuclear membrane (9). For example, in Salmonella typhimurium membrane attachment of TetA leads to the formation of a topological domain boundary defined by the transcriptional complex where the RNA is being translated (10). In addition, in yeast, telomeric sequences can act as functional anchor points for chromosome organization to provide blocks to the transmission of supercoils across the block (11). Other DNA⅐enzyme complexes in bacteria, such as that created by UvrAB can also act as a boundary to supercoiling (12).
DNA in living bacterial cells is organized with a linking number deficit leading to a state of unrestrained negative supercoiling (1,13). Although, DNA is negatively supercoiled in bacterial cells, not all supercoiling is unrestrained. About half the supercoils in DNA in Escherichia coli are restrained, possibly by the wrapping of DNA around histone-like proteins (14 -18). Consequently, in vivo measurements of levels of supercoiling are about half that measured for the purified chromosome (19 -22). Initial studies of supercoiling in eukaryotic cells failed to detect unrestrained supercoiling averaged over the entire chromosome (13), presumably due to the restraint of supercoils through the organization of DNA into nucleosomes. However, analyses of individual genes in Drosophila, mouse, and human cells have revealed the presence of unrestrained supercoiling associated with gene regions (2,(23)(24)(25)(26), whereas DNA outside the functional hsp70 domain at locus 87A Drosophila was completely relaxed (2). However, a relationship between transcription or transcriptional activation and unrestrained supercoiling remains to be clearly established. For example, the level of supercoiling within a transcriptionally active hygromycin resistance gene introduced into different regions of the human genome can vary from highly supercoiled to relaxed (26).
The mouse mammary tumor virus promoter provides a model system in which the nucleosomal organization and nucleosomal reorganization upon transcription activation are extremely well characterized (27). Transcription rapidly follows the addition of the glucocorticoid hormone dexamethasone, allowing analysis of the chromatin under conditions of gene repression or activation. When stably introduced into the genome, the mouse mammary tumor virus (MMTV) promoter acquires six positioned nucleosome families (28), positioned over the binding sites for the glucocorticoid receptor and associated factors involved in promoter activation. Steroid hormone induction of the MMTV promoter with dexamethasone renders the DNA sequences associated with the B nucleosome family accessible to restriction endonuclease digestion (28). The hormone-dependent increased nucleosome accessibility is believed to be mechanistically responsible for the loading of transcription factors and the subsequent induction of transcriptional activation (29,30). The increase in enzyme accessibility of nucleosome B in a stably integrated MMTV promoter is 15-20% in one mouse cell line, suggesting that transcriptional activation of the promoter occurs at similar levels (31).
The well characterized MMTV promoter and the availability of a cell line containing multiple tandem copies of this promoter afford an excellent system in which to study the effects of gene activation and transcriptional elongation on the level of DNA supercoiling and the number of topological domains. We have studied the change in supercoiling and topological domain size in a DNA region containing approximately 200 copies of the MMTV promoter 5Ј of the ha-v-ras gene to understand the role of chromatin organization and supercoiling on gene expression. The transcriptional state of the gene influences the topological domain size and the level of unrestrained supercoiling. The average domain size within the tandem array of MMTV promoter-driven genes changes upon transcriptional activation and elongation suggesting that transcription complexes are functional topological domain boundaries.

EXPERIMENTAL PROCEDURES
Cells, Cell Growth, and DNA Purification-Cell line 3134 contains approximately 200 copies of the 9-kb fragment (pM18D) of the plasmid pM18 (28) in a tandem array in a single chromosomal location. Cells were maintained as a monolayer in Dulbecco's modified Eagle's medium containing 10% fetal bovine serum at 37°C in a 5% CO 2 atmosphere. To incorporate BrdUrd into the DNA, cells were grown in Dulbecco's modified Eagle's medium ϩ 10% fetal bovine serum containing 10 mM bromo-2-deoxyuridine. At this concentration, 45% of the thymidines were substituted with bromo-2-deoxyuridine in hamster cells (32). 24 h before use, cells were incubated in Dulbecco's modified Eagle's medium lacking phenol red and containing 10% charcoal-stripped serum (Hyclone). For transcriptional activation, cells were treated with dexamethasone at a concentration of 0.5 M for 1 h, either before or after a 60-min treatment with 50 g/ml of ␣-amanitin (Roche Molecular Biochemicals).
Genomic DNA Analysis-Genomic DNA from approximately 1.2 ϫ 10 6 3134 cells, lysed in agarose blocks, was digested overnight at 37°C with various restriction enzymes. For clamped homogeneous electric field (CHEF) gel analysis, the digested DNAs were separated for 83 h in a 0.8% agarose gel equilibrated with 1X TAE buffer (0.04 M Tris base, 0.04 M glacial acetic acid, 0.001 M EDTA) in a Bio-Rad CHEF Mapper XA pulsed field electrophoresis system with the auto algorithm supplied by the manufacturer for optimal separation of 1 to 3 mega base pair DNAs. Following electrophoresis, location of the Hansenula wingei chromosomal marker was determined by ethidium bromide staining. Genomic DNA was transferred to a nylon membrane (Amersham Hybond-Nϩ) by capillary action under alkali conditions. This blot was hybridized for 48 h under moderate stringency (2ϫ SSC (0.3 M NaCl, 0.03 M sodium citrate, pH 7.0); 5ϫ Denhardt's solution (0.1% Ficoll (M r 400,000), 0.1% polyvinyl pyrrolidone (M r 400,000), and 0.1% bovine serum albumin); 50% deionized formamide; 1% SDS; 100 g/ml denatured sheared salmon sperm; 55°C) with a radioactive RNA probe derived from the 69% transforming portion of bovine papilloma virus (BPV; sense strand position 1234 -3224 of BPV-1 DNA (GenBank TM accession number X02346)). The blot was washed at 42°C in 0.2ϫ SSC, 0.1% SDS following treatment with RNase A (25 g/ml) and RNase T1 (10 g/ml) at room temperature in 2ϫ SSC. Subsequently, the blot was placed on a PhosphorImager plate to collect the radioactive signal. For field inversion gel electrophoresis, the genomic DNAs digested in plugs as described above were separated in a 1% agarose gel in 0.5ϫ TBE buffer (0.045 M Tris base, 0.045 M boric acid, 0.001 M EDTA) using the auto algorithm feature optimal for separation of 2-to 50-kilobase pair DNA. Reference marker DNA in this case was HindIII-digested lambda DNA (range of fragments, 8.3 to 48.5 kb). The separated genomic DNA was analyzed by hybridizing to the BPV RNA probe. S1 Analysis-The S1 analyses were performed as described previously (33). Briefly, total cellular RNA was extracted by phenol-chloroform, precipitated with ethanol, and dissolved in H 2 O. Single-stranded MMTV probes were synthesized by primer extension in the presence of [␣-32 P]ATP, using an oligonucleotide priming from ϩ85 in the long terminal repeat (LTR) (5Ј-TCTGGAAAGTGAAGGATAAGTGACGA), and SacI-cut pM18 as a template (end point at Ϫ105) (34). The probes were purified by electrophoresis in denaturing gels, recovered by the "crush-and-soak" method, and 10 5 cpm were hybridized to 10 g of RNA in 80% formamide, 0.2 M NaCl, 1 mM EDTA, 40 mM PIPES, pH 6.4, for 6 -12 h at 37°C. Hybrids were treated with 100 units of S1 nuclease for 1 h at room temperature, in 50 mM NaCl, 1 mM zinc acetate, 30 mM sodium acetate, pH 4.6, and then extracted with phenol-chloroform and precipitated with ethanol. Digestion products were resolved in denaturing 8% acrylamide sequencing gels. Visualization of the separated fragment was achieved with a PhosphorImager, and quantitation performed with the ImageQuant program (Molecular Dynamics).
Isolation and Restriction of Nuclei-Preparation of nuclei, restriction enzyme treatment, and analysis by primer extension was performed as described previously (31). Briefly, treated or untreated cells were scraped into cold phosphate-buffered saline and homogenized with a dounce ("A" pestle) in 0.3 M sucrose, 3 mM CaCl 2 , 2 mM magnesium acetate, 10 mM HEPES, pH 7.8, 0.5 mM dithiothreitol, 0.5 mM phenylmethylsulfonyl fluoride, 1% Triton X-100. The homogenate was diluted 1:1 with digestion buffer (25% glycerol, 5 mM magnesium acetate, 10 mM HEPES, pH 7.8, 0.5 mM dithiothreitol, 0.1 mM EDTA), and centrifuged through a pad containing dilution buffer for 15 min at 1,000 ϫ g and 4°C. Nuclear pellets were resuspended in 25% glycerol, 5 mM HEPES, pH 7.8, 0.1 mM EDTA, 0.5 mM dithiothreitol. 10 g of DNA equivalents of nuclei was restricted with 100 units of SacI, for 15 min at 37°C in 50 mM NaCl, 50 mM Tris-Cl, pH 8, 0.5 mM MgCl 2 , 1 mM ␤-mercaptoethanol (35) extracted with phenol-chloroform, and precipitated with ethanol. The extracted DNA was cut to completion with DpnII before primer extension to determine the fractional cleavage. To obtain a linear amplification of the signal, primer extensions with Taq polymerase were thermally cycled in 50 mM KCl, 50 mM Tris-Cl, pH 8.8, 200 M dNTP, 3.5 mM MgCl 2 , 0.1% Triton X-100, using an oligonucleotide priming from ϩ27 in the LTR (5Ј-ACAAGAGGTGAATGTTAG-GACTGTTGC). Extension products were extracted with phenol-chloroform and precipitated with ethanol before electrophoresis in sequencing gels and analysis in the PhosphorImager.
Nicking, Psoralen Cross-linking, and Southern Protocols-Following growth in the presence of bromo-2-deoxyuridine, the DNA nicking was accomplished by exposure of the cells to 313-nm light using a mercury vapor lamp with a K 2 CrO 4 /NaOH filter as described previously (26, 36 -39). Subsequent treatment with psoralen and 360-nm light, and DNA purification protocols were described previously (26).
For Southern analysis, 7 g of chromosomal DNA were digested to completion at 37°C with 100 units of SpeI in 60 l of restriction buffer. After digestion, DNA was purified and treated with glyoxal aldehyde and dimethyl sulfoxide (Me 2 SO) as described previously (26). DNA samples were separated on a 1% agarose gel in 10 mM sodium phosphate (pH 7.0) as described (40), the gel was denatured, and the DNA was transferred to a nylon membrane (26). For hybridization analysis of the MMTV promoter, a 3.0-kb SpeI fragment was isolated from plasmid pM18 (28) and labeled with [␣-32 P]CTP using the random priming method (Roche Molecular Biochemicals). The membrane was hybridized at 65°C for 12 h, washed twice with 2ϫ SSC (300 mM NaCl, 30 mM sodium citrate, pH 7.0) containing 0.1% SDS at 65°C for 60 min. Quantitation was completed using a Molecular Dynamics PhosphorImager using ImageQuant software. The fraction of DNA cross-linked (Fx) was calculated by dividing the area of the cross-linked peak by the sum of the area for the cross-linked and noncross-linked peaks. The crosslinks per kilobase (Xl/kb) were calculated using the formula Xl/kb ϭ Ϫln(1Ϫ Fx)/S, as described (24), where S is the size of the restriction fragment (S ϭ 3.0 kb for the SpeI fragment of pM18). R I/N values represent an average of two Southern blots each from a minimum of three separate experiments. The ratio of the mean cross-linking rate in intact verses relaxed domains (R I/N ϭ Xl/kb I /Xl/kb N ) reflects the level of unrestrained supercoiling (24). An unpaired, two-tailed, t test was performed with these R I/N values, and a significant difference is indicated if p Յ 0.05.
Measurement of Single Strand Breaks-The frequency of single strand breaks introduced by BrdUrd photolysis was determined as described previously (26) with the following modifications: 5 ml of 5-20% alkaline sucrose gradients were centrifuged at 35,000 rpm in a Beckman SW55Ti rotor for 4 h. Thirty equal volume fractions were collected from the bottom of the tube onto 3-cm square sections of a nylon membrane. After baking at 80°C for 2 h, the membrane was hybridized with the 3.0-kb SpeI fragment 32 P-labeled probe of pM18D containing the MMTV promoter and v-ras gene (Fig. 1). The membrane was washed twice with 2ϫ SSC at 65°C for 60 min. The membrane was dried then the 3-cm sections were placed into vials with scintillation fluid and counted for radioactivity. Molecular weights of the single strand DNA fragments (Z) were calculated (41), and the number of single strand breaks per tandem array was calculated using a molecular weight of 338 g/mol bp and 8 ϫ 10 6 bp per tandem array.

RESULTS
Characterization of Cell Line 3134 -Cell line 904.13 was originally established by transformation of C127 cells with a bovine papilloma virus vector containing the MMTV promoter driving expression of the ha-v-ras gene (Fig. 1C). Cell line 3134 was derived from the 904.13 line after a spontaneous event resulted in the integration and amplification of the episome. CHEF gel analysis (Fig. 1A) of DNA from this cell line with three different restriction enzymes, which do not cut the circular DNA construct, resulted in fragments ranging in size from 1.8 to 2.2 ϫ 10 6 base pairs. Partial digestion with a single-cut enzyme produced a ladder of fragments with a repeat size of 9 kb (Fig. 1B). These data indicate that the cell contains a large tandem array with approximately 200 head to tail copies of the original 9-kb fragment (Fig. 1C). Details of the 9-kb repeat unit are diagrammed in Fig. 1D.
A 3,026-bp SpeI fragment encompassing the MMTV promoter was used to assess the level of unrestrained supercoiling within the inserts of the tandem array (Fig. 1D). This fragment contains the entire MMTV LTR, and most of the transcribed region of the ras sequence. The percentage of MMTV templates activated by dexamethasone in this cell line was assessed by hypersensitivity to SacI cleavage, which cuts in the nucleosome B region of the promoter (Fig. 1D) (29,31,42). In a representative experiment, about 7-8% of the SacI sites were accessible in control cells or in cells treated with ␣-amanitin ( Fig. 2; Control and Amanitin). Treatment with dexamethasone resulted in an increase in the accessibility of this region, approximately 23% of the promoters were cut ( Fig. 2; Dex). This represents an increase of 0.15-0.16 in the fractional cleavage, in agreement with previous studies (31,42) and suggesting that 15-20% of the promoters are active at the time of the assay. In addition, treatment with dexamethasone following a 1-h pretreatment of the cells with ␣-amanitin resulted in a similar increase in fractional cleavage ( Fig. 2; Dex ϩ Amanitin), indicating that remodeling of the Nuc-B region occurs independently of transcription, as suggested by others (43).
␣-Amanitin inhibits MMTV-driven transcription under our incubation conditions as illustrated by a representative experiment in Fig. 3. Specific transcription, or more properly RNA accumulation, was examined with an S1 protection assay. 3134 cells treated with dexamethasone for 1 h displayed a 7-8-fold increase in RNA detected relative to uninduced cells ( Fig. 3; Control and Dex), comparable with previous determinations in the parental 904.13 cell line (44). Incubation of the cells with 50 g/ml ␣-amanitin for 1 h before dexamethasone treatment resulted in approximately a 1.5-1.7-fold increase in transcript levels relative to control cells ( Fig. 3; Amanitin ϩ Dex). This represents a substantial 90 -95% decrease in the steady state RNA level induced by dexamethasone after 1 h. A 50 -70% decrease in the basal transcription of uninduced templates was also observed in cells treated with ␣-amanitin without dexamethasone ( Fig. 3; Amanitin); however, the quantitation in this case is not as accurate because of the low RNA levels.
Psoralen Accessibility and Unrestrained Supercoiling at the MMTV Promoter-The rate of Me 3 -psoralen cross-linking to DNA depends on two parameters: the accessibility of the DNA and the level of unrestrained supercoiling in the DNA. Accessibility is dependent upon the extent of association of DNA with nucleosomes or other proteins that prevent Me 3 -psoralen binding (37). Psoralen accessibility of the MMTV promoter in this mouse cell was analyzed in four sets of cells: untreated (control) cells where the ha-v-ras gene was inactive; dexamethasone-treated cells that were transcriptionally active; cells treated with dexamethasone followed by ␣-amanitin, in which transcription elongation was allowed to proceed and then was blocked; and cells treated with ␣-amanitin followed by dexamethasone, in which the promoter was activated but transcription elongation was prevented. Each set of cells was treated with psoralen and exposed to increasing doses of 360-nm light. The DNA was irreversibly denatured as described under "Experimental Procedures," and the cross-linked and noncrosslinked molecules were separated on a native agarose gel (Fig.  4A). The cross-links per kilobase (Xl/kb), calculated from the intensities of the cross-linked (Xl) and noncross-linked (non-Xl) SpeI fragments (Fig. 4A), showed a linear relationship as a function of light exposure for each set of cells (Fig. 4B). As shown in Fig. 4B, the cross-linking rates for the cell lines were linear to at least 12 kJ/m 2 . The molecular bases for the slight decrease in rate of binding in cells treated with dexamethasone (permeability or accessibility changes) was not investigated. In the experiments to determine the level of unrestrained negative superhelical energy, a dose of 6 kJ/m 2 of 360-nm light was chosen because it is within the linear range of the cross-linking reaction.
Unrestrained supercoiling was determined by comparing the Xl/kb at a constant dose of psoralen and light within the SpeI fragment before and after the chromosomal DNA was nicked by BrdUrd photolysis. In Fig. 5A the intensity of both the Xl and non-Xl species decreased at higher doses of nicking but proportionally the Xl band decreases faster (determined by Phosphor-Imager analysis). Thus, the rate of cross-linking (or Xl/kb ver-

FIG. 3. Incubation of 3134 cells with ␣-amanitin inhibits MMTV-driven RNA synthesis.
Total RNA was extracted from treated or untreated 3134 cells, and MMTV LTR reporter-specific RNA assayed by S1 hybrid protection. The single-stranded probe spanned from ϩ85 to Ϫ105. The signal from the protected hybrid was quantitated and normalized to the signal in untreated cells. The chart shows the fold-induction of cells treated as follows: untreated (control); ␣-amanitin for 1 h (amanitin); dexamethasone for 1 h (Dex); ␣-amanitin for 1 h followed by dexamethasone and additional incubation for 1 h (Dex ϩ amanitin). Results are for a representative experiment. Results from duplicate experiments were similar.

FIG. 4. Rate of psoralen cross-linking within the 1.8-megabase pair array.
A, Southern analysis on DNA from dexamethasone-treated cells cross-linked for various times after exposure to psoralen. B, a plot of the Xl/kb as a function of light dose. Cells were treated with 360-nm light at an incident intensity of 1.2 kJ/m 2 /min. Samples were removed after various times and analyzed by Southern hybridization to determine the Xl/kb as described under "Experimental Procedures." E, untreated; Ⅺ, dexamethasone followed by ␣-amanitin treatment; ƒ, dexamethasone; 224 , ␣-amanitin followed by dexamethasone treatment. Lines represent the linear regression, best fit to the data. Lines for the dexamethasone and the ␣-amanitin followed by dexamethasone treatment data are identical.
sus dose) decreased at higher doses of 313-nm light indicating the presence of unrestrained supercoiling. The measurement of supercoiling is presented as R I/N as described previously (24,26)). A value of R I/N ϭ 1 indicates that no unrestrained supercoiling was detected, whereas an R I/N Ͼ 1 indicates the presence of negative supercoiling. All R I/N values in Fig. 5B represent the average from at least six independent determinations from three separate experiments. A moderately high level of unrestrained superhelical tension (R I/N ϭ 1.55) was present in the promoter region in cells with an uninduced basal level of transcription of the MMTV promoter (Fig. 5B, Control). This level of supercoiling was lower than that observed in the Drosophila hsp70 or rRNA genes (24), but higher than that seen in a hygromycin gene in human cells (26). After dexamethasone treatment for 1 h, the overall level of negative superhelical energy dropped significantly (about 80%), to R I/N ϭ 1.11 (Fig.  5B, Dex). The level of supercoiling in the MMTV promoter in cells treated with dexamethasone and then ␣-amanitin (for 1 h), R I/N ϭ 1.43, was significantly higher than in the dexamethasone-treated cells (R I/N ϭ 1.11). However, the level of supercoiling in these cells was 15-20% lower than in control (uninduced) cells with a basal level of expression (Fig. 5B, Dex ϩ ␣-amanitin). Furthermore, cells treated with ␣-amanitin for 1 h and then dexamethasone for 1 h had a level of supercoiling of R I/N ϭ 1.48, which is significantly higher than in the dexamethasone-treated cells (R I/N ϭ 1.11) and about a 10 -15% lower than in control cells (Fig. 5B, ␣-amanitin ϩ Dex).
Measurement of Topological Domains in 3134 Cells-Different times of irradiation with 313-nm light that introduced nicks into the DNA were required to relax all supercoils in cells following the different treatments shown in Fig. 5B. In untreated cells, the relaxation of most supercoiling (Ͼ95%) within the SpeI fragment occurred after 5 min of nicking (Fig. 5B,  Control), but relaxation required greater than 15-18 min of nicking in cells treated with dexamethasone and then ␣-amanitin or in cells treated with ␣-amanitin and then dexamethasone (Fig. 5B, Dex ϩ ␣-amanitin, ␣-amanitin ϩ Dex). The different doses of 313-nm light required for nicking suggest that the domain size in the tandem array may be different in the nontreated and treated mouse cells.
The psoralen assay for measurement of a topological domain size requires that the DNA contain unrestrained negative supercoils, and the assay measures the loss of superhelical tension as a function of the number of nicks introduced into DNA (37-39). The number of nicks introduced into the DNA by photolysis of BrdUrd-labeled DNA was measured by alkaline sucrose gradient sedimentation analysis. Alkaline sucrose gradient fractions were collected onto a nylon membrane and probed using Southern hybridization to determine the average molecular weight of the DNA containing the MMTV inserts within the tandem array. From this analysis, the number of strand breaks introduced into the MMTV tandem array as a function of the 313-nm light dose could be calculated.
The nicking rates were not significantly different statistically between the nontreated cells and cells treated with ␣-amanitin, dexamethasone, or any combination of the two. However a statistically significant 2-fold difference was observed between the cells that were treated with dexamethasone then ␣-amanitin and the cells that were treated with ␣-amanitin then dexamethasone (Fig. 6, insert, top and bottom lines). Treatment with dexamethasone followed by ␣-amanitin yielded the highest number of strand breaks, and this treatment may have resulted in the most open chromatin configuration. Treatment with ␣-amanitin would be expected to keep the chromosome more organized into nucleosomes and thus more protected from free radical attack during BrdUrd photolysis.
To estimate the number of topological domains, it is assumed that the nicks are introduced in a Poisson distribution (1,38,39). Using the Poisson formula, theoretical curves are calculated to determine the best fit with the experimental data points (described in the legend to Fig. 6). In Fig. 6, three sets of experimental data are plotted against five theoretical curves each representing a different topological domain size. The theoretical curve that best fits the experimental data represents the domain size. The average number of topological domains/ tandem array for untreated cells was 55/tandem array, or 32,000 bp/domain. For cells treated with ␣-amanitin and then dexamethasone, the number of domains was approximately 110/tandem array, or 16,000 bp/domain. For cells treated with dexamethasone and then ␣-amanitin approximately 215 domains/tandem array were present, with 8,000 -9,000 bp/domain. Treatment with dexamethasone or ␣-amanitin should not introduce major changes in chromosome structure that would influence the results obtained. The effects of dexamethasone have been extensively studied in the MMTV promoter (28,45) and the TAT promoter (46). In both of these genes, dexamethasone is known to have very local effects on chromatin structure. Therefore, treatment with dexamethasone would not be expected to have global chromatin effects. In addition, no significant effect of dexamethasone treatment on the rate of nicking was observed when looking at the DNA size averaged over the entire chromosome (data not shown). In a previous analysis of a hyg gene in five random chromosomal locations in human cells, no unexpected differences in the rate of cross-linking were observed following ␣-amanitin treatment, other than those expected for a loss of supercoiling (26). Moreover, in two cell lines, no supercoiling was observed, and ␣-amanitin had no effect of cross-linking rates. In Drosophila experiments, using heat shock to induce the HSP70 genes, very little change was observed in the rRNA gene response to nicking or heat shock. In addition, differential changes were observed throughout locus 87A7, which were specific for the hsp70 transcription units, the flanking nontranscribed regions, and scs and scsЈ sequences (2). These total experiments argue against any global genome wide effects of induction of specific genes or ␣-amanitin treatment that would complicate the interpretation of the results. DISCUSSION In this study the effect of transcriptional activation/elongation on unrestrained supercoiling and its relation to topological domain size was examined. Studies were performed on the MMTV promoter-driven ha-v-ras gene inserted in the chromosome of a mouse cell line. Analysis was completed using a psoralen-based assay for unrestrained negative supercoiling in combination with measuring the rate of loss of supercoils following the introduction of nicks into DNA by BrdUrd photolysis. Our results demonstrate that this MMTV promoter-driven ha-v-ras gene is maintained with a moderate level of unrestrained torsional tension in the uninduced state. Transcription elongation decreased negative supercoiling. This is the first example of a gene that becomes torsionally relaxed upon transcription. Transcriptional activation had little or no effect on supercoiling, and blocked elongating complexes also had little or no effect on negative unrestrained supercoiling. However, the activation of transcriptional complexes and the presence of elongating transcriptional complexes increased the number of topological domains within the MMTV promotergene region.
Unrestrained negative supercoiling in the tandem ha-v-ras genes was relaxed following the activation of transcription by addition of dexamethasone. This result is in contrast with certain models in which supercoiling has been thought to be a prerequisite for transcription and possibly a consequence of RNA polymerase movement during elongation (for example, see Ref. 4). In fact, transcription in the Drosophila hsp70 locus results in a slight increase in an already high level of unrestrained supercoiling (24). In the case of the ha-v-ras gene, the release of unrestrained supercoiling during active transcription may be due to the active recruitment or activation of topoisomerases within the locus.
Unrestrained supercoiling was maintained in cells treated with any combination of dexamethasone and ␣-amanitin; i.e. cells induced for transcription but in which active transcriptional elongation was inhibited. The level of unrestrained supercoiling in these cells (R I/N ϭ 1.43, 1.48) was similar to that in uninduced cells (R I/N ϭ 1.55). The slight decrease in supercoiling in cells treated with ␣-amanitin may be due to residual active transcription (90 -95% of transcriptional elongation has been blocked, Fig. 3). Supercoiling was reestablished in cells treated with dexamethasone and then ␣-amanitin upon the inhibition of transcription elongation. The reestablishment of supercoiling occurred within the 1 h between ␣-amanitin treatment and the psoralen photobinding. If active transcriptional elongation is tightly coupled with the active topoisomerase(s) it is understandable that negative supercoiling could be rapidly relaxed. It is known that topoisomerase I is associated with regions of active transcription (47)(48)(49), and that it is also recruited to promoters by transcription factor IID (50,51). That the inhibition of elongation leads to reestablishment of unrestrained supercoiling implies a mechanism exists for the regeneration, or loss of restraint, of supercoils. This mechanism does not appear to be dependent upon tracking by RNA polymerase. Furthermore, current evidence indicates that no loss of nucleosomes occurs from the LTR during activation (31,52).
Changes in Topological Domain Organization Occur during Transcription of the ha-v-ras Gene-The topological domain size changed dramatically upon transcription. In the mouse cell line 3134 there are approximately 200 tandem inserts containing the MMTV promoter, each insert is 9,000 bp in length, and the entire region consists of 1.8 ϫ 10 6 bp. In cells in which the ha-v-ras gene was inactive approximately 55 topological domains existed, such that each domain was, on average, 32,000 bp in length. This is similar to the size found in the rDNA genes in humans fibrosarcoma cells (26). These domain boundaries may be defined by some structural attachment to a nuclear structure. Cells treated with ␣-amanitin and then dexametha- Cells treated with dexamethasone and then ␣-amanitin had 9,000 bp/domain, a smaller domain size than in cells with simply activated transcription complexes (16,000 bp/domain). Under these conditions, the maximum number of transcription complexes are presumably formed before elongation is inhibited by ␣-amanitin. The number of domain boundaries created under these conditions of maximal gene induction is similar to the suggested number of MMTV promoters activated by dexamethasone treatment. These results strongly suggest that transcription complexes constitute transient (and mobile) topological domain boundaries and, thus, support the hypothesis that domain boundaries can be "functional" in nature (9).
These domain measurements represent approximate values, because it is uncertain if there are damaged sites from BruUrd photolysis that are alkali labile but do not represent breaks in vivo, as discussed previously (26). This may influence domain sizes by as much as 10 -20%. Moreover, it is formally possible that not all torsional tension would be relaxed at a nick if DNA repair proteins bound to the site and prevented rotation about the nick. However, experiments are performed in phosphatebuffered saline buffer at 0 -4°C to prevent any such repair activity. In small bacterial chromosomes most supercoils ap-pear to be lost, and only restrained supercoils remain following nicking in vivo (14,22).
The observations that transcription can influence a topological domain size suggests several mechanisms for domain organization. In Fig. 7, top panel, the polymerase complex could establish a topological domain by simply preventing rotation of the DNA double helix. Polymerase complexes bound to the promoter sequence and polymerase complexes transcribing the gene can establish a topological domain. In this case rotation is prevented by restraints to rotational diffusion. Alternatively, in Fig. 7, bottom panel, the polymerase complex is recruited to a nuclear structure upon gene activation. Alternatively, the transcriptional machinery is localized at a nuclear structure and activation involves the association of the DNA with the localized, transcription complexes. In this case the DNA template feeds through the attached complex during elongation, as suggested for the tetA gene in E. coli (53). Attachment to a nuclear structure would clearly prevent rotation of the DNA and establish a topological domain.
In conclusion, our results establish that unrestrained supercoiling is present in the tandem array of MMTV inserts within a mouse cell. Active transcription decreases the level of unrestrained supercoiling within this gene. Remarkably, the topological domain size decreases when the MMTV promoter is activated by dexamethasone, and a further reduction is observed if an actively transcribed ha-v-ras gene is inhibited by ␣-amanitin. These results demonstrate that the transcriptional state of a gene can influence both the level of unrestrained supercoiling and the topological domain size. Although certain topological domains may be defined by the interaction of MARs with a nuclear structure, clearly some topological domains in eukaryotic cells appear to be defined, in some way, by active transcription complexes.

FIG. 7. Models for the generation of new domain boundaries.
A, loading a large initiation complex, including RNA polymerase II, on the promoter provides a constraint to DNA rotation that effectively creates a new topological domain. B, receptor activation recruits the active promoter units to a nuclear structure that serves as an active center for transcription. These centers represent the domain boundaries.