IntroductionRecent studies have established that chromatin structure isimportant for transcriptional regulation. Starting from analysesof DNase l hypersensitivity (Elgin, 1981), lines of evidencehave been accumulated for changes in the chromatin structure,such as DNA methylation (Bird and Wolffe, 1999), chemicalmodifications of histones (Jenuwein and Allis, 2001) andchromatin remodeling (Vignari et al., 2000; Narliker et al.,2002) during regulation of gene expression. However, besidestranscription-driven supercoiling of DNA (Giaever and Wang,1988), our knowledge on conformation of chromatin DNA iselusive because of the lack of proper probes for analyses ofDNA topology in vivo (Freeman and Garrard, 1992).

Psoralen is a planar, aromatic compound that intercalatesinto DNA. On exposure to 365 nm light the intercalatedpsoralen mediates crosslinking of opposite DNA strands viathe formation of covalent bonds at each end of the molecule(for a review, see Sinden et al., 1992; Ussery et al., 1992). Ithas been shown that the rate of psoralen photocrosslink todouble-stranded DNA is linearly related to its level of negativesuperhelicity (Sinden et al., 1980). Detection of negativesupercoiling in living cells can be accomplished by comparingrates of crosslinking in intact cells with those in cells wherepotential torsional tension has been relaxed by DNA strandnicking. In prokaryotes, measurements averaged globallyacross the Escheridia coligenome have detected unconstrainednegative supercoiling with a superhelical density of –0.05(Sinden et al., 1980). In eukaryotes, similar global assayson human HeLa and Drosophila Schneider cell lines haveshown that bulk DNA within the genome is torsionarilyrelaxed (Sinden et al., 1980). However, it does not necessarilyexclude a possibility that there are negatively supercoiled

microdomains within the genome. Indeed Jupe et al. (Jupe etal., 1993) have shown the presence of unconstrained negativesupercoils in the hsp70and the 18S-ribosomal RNA genesof the Schneider cell line. They observed a high level ofunconstrained negative supercoils within the hsp70transcription units, whereas downstream regions of thedivergent hsp70 genes at 87A do not contain a significantlevel of supercoiling. Quantitative analyses of psoralenphotocrosslinking have also shown an increase in the level ofnegative supercoils in the coding region of hsp70upon heatshock. Accessibility of psoralen has been also reported forrDNA of growing Dictyostelium discoideumcells (Sogo et al.,1984), and the dihydrofolate reductase gene (Ljungman andHanawalt, 1992) and the hygromycin resistance transgenes(Kramer and Sinden, 1997) in cultured human cells. Becausethese psoralen-based studies on the conformation of chromatinDNA rely on Southern hybridization for detecting thecrosslinks, only limited regions of the genome can be analyzed.

In the larval polytene chromosome of Drosophila,approximately 1000 chromatids are laid down in juxtaposition,building up horizontally amplified chromosomes visible underthe light microscope. Alternating more and less tightlycondensed regions in precise register give rise to an alternatingpattern of bands and interbands. As perceived in the 1930s,these structures display an amplified interphase genomeon which specific genetic loci can be mapped and thetranscriptional state of genes can be observed as a puff. Forexample, puffs are induced on loci carrying heat-shock geneswith a brief heat shock. In heat-shock puffs, the active state ofthe hsp70gene continues for more than 10 minutes and a halfof the transcriptional activity is maintained even after 1 hour(Kroeger and Rowe, 1992). In addition, nascent RNA can be

3797

Bulk DNA within the eukaryotic genome is torsionarilyrelaxed. However, unconstrained negative supercoils ofDNA have been detected in few local domains of thegenome through preferential binding of psoralen. To makea genome-wide survey for such domains, we introducedbiotinylated psoralen into Drosophila salivary glands andvisualized it on polytene chromosomes with fluorescentstreptavidin. We observed bright psoralen signals on manytranscriptionally active interbands and puffs. Upon heatshock, the signals appeared on heat-shock puffs. The

signals were resistant to RNase treatment but disappearedor became faint by previous nicking of DNA or inhibitionof transcription with α-amanitin. These data show thattranscription-coupled, unconstrained negative supercoilsof DNA exist in approximately 150 loci within theinterphase genome.

Key words: Drosophila, Polytene Chromosome, Psoralen, Supercoil,Transcription

Summary

Visualization of unconstrained negative supercoils ofDNA on polytene chromosomes of DrosophilaKuniharu Matsumoto 1 and Susumu Hirose 1,2,*1Department of Developmental Genetics, National Institute of Genetics, and 2Department of Genetics, SOKENDAI, 1111 Yata, Mishima, Shizuoka-ken 411-8540, Japan*Author for correspondence (e-mail: shirose@lab.nig.ac.jp)

Accepted 17 March 2004Journal of Cell Science 117, 3797-3805 Published by The Company of Biologists 2004doi:10.1242/jcs.01225

Research Article

JCS ePress online publication date 13 July 2004

3798

labeled with ribonucleotide analogues on puffs (Chang et al.,2000). All these characters are advantageous for in situanalyses of DNA conformation and transcription.

In this work, we photocrosslinked polytene chromosomes insalivary glands of Drosophila with biotinylated psoralen anddetected it with fluorescent streptavidin to achieve genome-wide survey for the negatively supercoiled domains of DNA.We anticipated that we would detect a small number ofpsoralen signals because so far, only two highly transcribedgenes have been known to harbor unconstrained negativesupercoils (Jupe et al., 1993). To our surprise, we observedmany signals of psoralen on the polytene chromosomes. Thesesignals were detected in many but not all interbands and puffsthat are the sites of active transcription, and disappeared onprevious nicking of chromatin DNA or inhibition oftranscription. This is the first visualization of unconstrainednegatively supercoiled domains of DNA within the interphasegenome.

Materials and MethodsHeat shock and X-ray irradiationDrosophila melanogaster Oregon R was grown at 18°C. To obtainheat-shocked salivary glands, third-instar larvae were collected in apolypropylene tube and submerged in a 37°C water bath for 10minutes. Salivary glands were dissected in a dissection buffer (10 mMHEPES-KOH pH 7.6, 5 mM MgCl2, 5 mM KCl, 130 mM NaCl, 1%polyethylene glycol 6000). When necessary the dissected salivaryglands were X-ray irradiated for 60 minutes at a dose rate of about3 Gy/minute (TORREX CABINET X-RAY SYSTEM ModelTRX2800, Faxitron).

Staining of Drosophila polytene chromosomesFor uptake of biotinylated psoralen, 4-5 pairs of salivary glands weretreated with 0.01% digitionin (Calbiochem) in 40 µl of dissectionbuffer for 10 minutes, then rinsed with dissection buffer withoutdigitonin and soaked in dissection buffer containing 0.2 ng/mlbiotinylated psoralen (Ambion) for 10 minutes. Then, the salivaryglands were illuminated with a long wave (365 nm) UV lamp (UVPmodel UVL-21) for 10 minutes to crosslink the psoralen. For RNAlabeling, 2 mM Br-UTP (Sigma) was added in dissection buffer duringthe digitonin treatment. To inhibit transcription by RNA polymeraseII, 3 µg/ml of α-amanitin (Sigma) was added in dissection bufferduring the digitonin treatment. The bulk of the Br-UTP incorporation,except in nucleoli, was abolished by the amanitin treatment. After thelight exposure salivary glands were fixed with 40% acetic acid andsquashed. For RNase treatment, the squashed samples were incubatedwith 100 µg/ml of RNase A (Sigma) for 3 hours in 10 mM sodiumphosphate pH 7.0/150 mM NaCl. Br-UTP was detected with anti-BrUmonoclonal antibody (Roche) and Rhodamin-labeled anti-mouse IgGantibody. Biotinylated psoralen was detected with Alexa Fluor 488-labeled streptavidin (Molecular Probes). DNA was stained with DAPI.Fluoroimages were analyzed with Carl Zeiss Axioplan 2 microscopeand IP lab software.

Southern analysis of photocrosslinked DNAHeat shock of larvae and dissection were done as described forpolytene chromosome staining. Twenty pairs of salivary glands weresoaked in dissection buffer containing 4,5′,8-trimethyl psoralen(Sigma), then exposed to 365 nm light for photocrosslinking.Crosslinked DNA was isolated by Proteinase K treatment in lysisbuffer (10 mM Tris-Cl pH 8.2, 100 mM EDTA, 0.5% SDS) for 4 hoursat 55°C, followed by phenol/chloroform and chloroform extraction.

Purified DNA was digested with restriction enzymes (TAKARAshuzo) that produce fragments containing the regions of interest.Following precipitation with isopropanol, DNA pellets were dissolvedin glyoxal denaturation buffer (1 M glyoxal, 10 mM sodium phosphatepH 7.0, 50% dimethylsulfoxide) and denatured at 50°C for 1 hour.Glyoxylated non-crosslinked and crosslinked DNA fragments werethen separated by electrophoresis on a 1% agarose gel in 10 mMsodium phosphate buffer (pH 7.0) at 3.6 volt/cm for 100 minutes.After electrophoresis, the gel was incubated with denaturing solution(0.5 M NaOH, 1.5 M NaCl) at 65°C for 100 minutes to reversepsoralen crosslink. DNA fragments were transferred to a nylonmembrane (Hybond-N, Amersham) in 10×SSC and the membranewas hybridized with 32P-labeled probe DNA produced by randompriming, then washed and exposed to a X-ray film. The hsp70CR andDDS fragments were excised from a plasmid p56H8RIA (Moran etal., 1979).

ResultsSouthern analyses for detecting negative supercoiling ofDNA in salivary glands of DrosophilaJupe et al. (Jupe et al., 1993) have developed a protocol forquantification of psoralen crosslinking localized to specificregions in the Drosophilagenome using cultured cells of theSchneider line SL2. To detect negative supercoils of DNA onpolytene chromosomes, we first tested whether the protocolcan be applied to a tissue like the salivary gland. Briefly,salivary glands were dissected from heat-shocked or non-heat-shocked larvae. The dissected glands were soaked in a buffercontaining various concentration of psoralen, followed byexposure to 365 nm light for photocrosslinking. DNA waspurified, restriction digested and denatured with glyoxal.Crosslinked and non-crosslinked DNA fractions wereseparated on a neutral gel, treated with alkali at 65°C to reversecrosslink, blotted onto a nylon membrane and detected bySouthern hybridization using probes in the hsp70locus at 87A(Fig. 1A, CR and DDS). We included the hot alkalineincubation of the gel containing DNA because it was essentialfor efficient and reproducible probe hybridization to thecrosslinked fraction. Otherwise, rapid self annealing of thecrosslinked DNA prevented hybridization with the DNA probe.We observed crosslinking of DNA in the hsp70coding region(CR) in a psoralen concentration-dependent manner (Fig. 1B).The frequency of crosslinking increased on heat shock of thelarvae (Fig. 1B). When the glands isolated from heat-shockedlarvae were irradiated with 180 Gy of X-rays to introduce nicksin chromatin DNA and then incubated with psoralen, we wereunable to detect a significant level of crosslinking (Fig. 1C,lane 3 vs lane 4). As a control, X-ray irradiation after thephotocrosslinking step did not alter the result (Fig. 1C, lane 5vs lane 2). This dose of X-rays is estimated to induce aboutone single-strand break per 30 kb of cellular DNA (Arnströmand Edvardsson, 1974), suggesting that the target size ofrelaxation is greater than the size of the CR fragment (2 kb).Similar results were obtained with the glands isolated fromnon-heat-shocked larvae (data not shown). By contrast, thelevel of photocrosslinking was low in the hsp70 distaldownstream sequence (DDS) and was not enhanced on heatshock of larvae (Fig. 1C, lanes 8 and 9). These results are ingood agreement with those obtained using SL2 cells (Jupe etal., 1993), and indicated the presence of unconstrained negativesupercoils in the coding region of hsp70 and further increase

Journal of Cell Science 117 (17)

3799Visualization of negative supercoils of DNA

in their levels on heat shock. From these data, we conclude thatthe psoralen photocrosslinking protocol is applicable to thesalivary gland.

Interestingly, crosslinking was barely detectable whenthe glands were incubated with α-amanitin before thephotocrosslinking step, even though the glands were dissectedfrom heat-shocked larvae (Fig. 1C, lane 6 vs lane 7). Theseresults reveal that the negative supercoils are readily relaxedon inhibition of transcription by RNA polymerase II.

Visualization of unconstrained negative supercoils ofDNA on polytene chromosomesWe then extended the psoralen crosslinking protocol togenome-wide survey of negative supercoils throughvisualization of biotinylated psoralen on salivary glandpolytene chromosomes with fluorescent streptavidin. Thebiotinylated psoralen was less permeable to cells than freepsoralen, and fluorescent signals on polytene chromosomeswere hardly detectable when salivary glands were challengeddirectly (data not shown). Therefore, the glands were firsttreated with 0.01% digitonin. This mild treatment withthe nonionic detergent has been successfully used forpermeabilization of salivary gland cells and subsequentobservation of puff formation and transcription on polytenechromosomes (Myohara and Okada, 1987). The digitonin-treated salivary glands were soaked in a buffer containingbiotinylated psoralen and then exposed to 365 nm light, fixedand squashed. Biotinylated psoralen was detected with Alexa488-labeled streptavidin. We observed many bright bandsof psoralen signals over weak and rather homogeneoussignals along polytene chromosomes (Fig. 2A). There wereapproximately 150 bright signals within the genome. Suchsignals were not detected when the glands were incubated withAlexa 488-labeled streptavidin directly without previouspsoralen treatment (data not shown), or the exposure to 365 nmlight was omitted (Fig. 2D). When compared with the imageof DAPI, the bright signals were detected on many but notall interbands and puffs where DAPI signals were barelydetectable (e.g. arrows in Fig. 2B,C), while the vague signalsalmost coincided with those of DAPI (Fig. 2B,C). Some brightsignals were observed only on the side margins of interbandsor puffs (e.g. arrowheads in Fig. 2B,C). These bright signalswere not due to binding of psoralen to RNA because the signalswere clearly seen after RNase treatment (Fig. 2F), whereasstaining of RNA with YOYO dye was erased completely (datanot shown). The psoralen signals on interbands and puffsdisappeared or became faint when salivary glands wereirradiated with 180 Gy of X-rays to introduce nicks inchromatin DNA before incubation with biotinylated psoralen,whereas the vague psoralen signals were left unaffected (Fig.3A-C). Essentially similar disapperance of the bright psoralensignals was observed when salivary glands were treated withα-amanitin before photocrosslinking (Fig. 3D,E). As controls,mock-treated samples without X-ray or α-amanitin showedessentially the same patterns as in Fig. 2 (data not shown).These results indicate the presence of many local domains ofDNA that harbor transcription-coupled, unconstrained negativesupercoils within the interphase genome. The vague psoralensignals are most probably due to nonspecific binding ofbiotinylated psoralen to bulk relaxed DNA.

Unconstrained negative supercoils of DNA on heat-shock puffsHeat treatment of larvae induces puffs on heat-shock geneswhile transcription of other genes is declined. To examinewhether these drastic changes in the mode of transcriptionaffect the superhelical state of DNA within the genome,biotinylated psoralen was introduced into salivary glandsdissected from heat-shocked larvae. An impressive signal ofpsoralen was observed on a heat-shock puff at 87B (Fig. 4,

Fig. 1.Detection of negative supercoils of DNA on hsp70in salivarygland. (A) Genomic organization of the hsp70 genes at 87A. Arrowsdenote the orientation of transcription. The hsp70genes are flankedby specialized chromatin structures, scsand scs’ (Udvardy et al.,1985). Restriction sites used for Southern analysis are indicated: X,XbaI; E, EcoRI; B, BglII. DNA fragments CR and DDS were usedfor hybridization probes. (B) Psoralen photocrosslinking to the hsp70genes at 87A in salivary gland cells. Salivary glands from non-heat-shocked or heat-shocked larvae were incubated with indicatedconcentrations of psoralen. After photocrosslinking, genomic DNAwas isolated, digested with XbaI, denatured, electrophoresed andanalyzed by Southern hybridization using CR fragment as a probe.(C) Effect of previous nicking of DNA or inhibition of transcriptionon the psoralen photocrosslinking. Lanes 1-7: salivary glands fromheat-shocked larvae were incubated with (lanes 2-7) or without (lane1) psoralen and analyzed as in (B). Where indicated, the glands wereirradiated with X-rays before (lane 3) or after (lane 5)photocrosslinking. The sample in lane 4 was mock treated withoutX-ray. The glands were treated with (lane 6) or without (lane 7) α-amanitin before incubation with psoralen. Lanes 8 and 9: salivaryglands from non-heat-shocked (lane 8) or heat-shocked larvae (lane9) were incubated with psoralen. Genomic DNA was digested withBglII and EcoRI, and analyzed as in (B) using DDS fragment as aprobe.

3800

arrow 87B). The signals were detected on both side margins ofa heat-shock puff at 87A (Fig. 4, arrow 87A). The psoralensignal was also observed on another heat-shock gene locus at93D (Fig. 4, arrow 93D). Many psoralen signals detected on

interbands of polytene chromosomes from non-heat-shockedlarvae disappeared or became less prominent on polytenechromosomes from heat-shocked larvae (Fig. 4, arrowheads).Upon X-ray irradiation (Fig. 5A-C), or α-amanitin treatmentof salivary glands (Fig. 5D-F) before photocrosslinking, thepsoralen signals on heat-shock puffs disappeared or becamefaint, although heat-shock puffs were clearly seen after thesetreatments (Fig. 5, arrows). The X-ray irradiation did notsignificantly affect transcription on the heat-shock puffsas revealed by Br-UTP incorporation followed by itsimmunological detection (data not shown), suggesting that thedisappearance of psoralen signals after the X-ray irradiation isnot due to a secondary effect of transcription inhibition.Essentially the same results as shown in Fig. 4 were obtainedfrom mock-treated samples without X-ray or α-amanitin(data not shown). These data show transcription-coupled,unconstrained negative supercoils on heat-shock puffs. Thedata also reveal that the superhelical state of DNA within thegenome is not static but dynamic.

Correlation between unconstrained negative supercoilsand transcriptionJudging from α-amamtin sensitivity, the generation of negativesupercoils on interbands and puffs appears to be coupled withtranscription. To confirm this, dissected salivary glands wereincubated with Br-UTP to label nascent transcripts just beforephotocrosslinking (see Materials and Methods). The labeledRNA was detected with anti-BrU antibody in many interbandsand puffs (Fig. 6D,E). On a merged image, all signals ofpsoralen were highlighted in interbands and puffs together withthe nascent RNA (Fig. 6F). After heat shock, uptake of Br-UTPwas impressive on the heat-shock puffs at 87A and 87B. A less-prominent signal was also observed on the heat-shock genelocus at 93D (Fig. 7). This result is consistent with the previousreport that the C-terminal domain (CTD)-phosphorylatedRNA polymerase II localizes exclusively on heat-shock puffs(O’Brien et al., 1994). Strong signals of psoralen were seen onthe heat-shock puff at 87B and on both side margins of theheat-shock puff at 87A. For detailed inspection on thedistribution of negative supercoils and transcripts, the signalintensities of psoralen, nascent RNA and DNA were quantifiedalong the 87A and 87B puffs (Fig. 7B,C). The distribution ofsignals of psoralen and nascent RNA was not even within the87A and 87B puffs, but showed a similar pattern to each other.DNA signals were extremely low on these puffs. When theratios of signal intensities of psoralen to DNA were plotted,these exhibit a sharp peak at 87B puff and a broad peak at87A puff (Fig. 7C, yellow curve). Collectively, these resultsestablish a correlation between unconstrained negativesupercoils and transcription.

DiscussionUnconstrained negative supercoils within interphasegenomeIn this study we detected bright signals of biotinylated psoralenon many transcriptionally active interbands and puffs alongpolytene chromosomes of the salivary gland. Upon heat shock,the signals on the interbands and endogenous puffs disappearedor diminished, and strong signals appeared on newly induced

Journal of Cell Science 117 (17)

Fig. 2.Visualization of psoralen signals on polytene chromosomes.(A) Biotinylated psoralen signals were detected with Alexa 488-labeled streptavidin (light green color). (B) DAPI signals (bluecolor). (C) Merged image. Arrows indicate representative interbandsand puffs. In some loci, intense psoralen signals were observed onlyon the side margins of an interband or puff (arrowheads).(D) Biotinylated psoralen signals disappeared when the photo-crosslinking step was omitted. (E) DAPI image of (D).(F) Biotinylated psoralen signals were resistant to RNase treatment.(G) DAPI image of (F). Bars, 10 µm.

3801Visualization of negative supercoils of DNA

heat-shock puffs. The signals were resistant to RNase treatmentbut abolished by previous nicking of cellular DNA or inhibitionof transcription with α-amanitin. These data indicate thepresence of transcription-coupled, unconstrained negativesupercoils in many loci within the interphase genome.

A priori, we analyzed the binding of biotinylated psoralento polytene chromosomes in case there were other reasons forthe observation. For example, psoralen could be more easilyaccessible to open chromatin than tightly packed chromatin. Itis possible that the psoralen signals on interbands and puffsreflect its preferential binding to open chromatin. However, thebright psoralen signals disappeared upon nicking of chromatinDNA or α-amanitin treatment, although the interbands andpuffs were clearly seen after these treatments. These results cannot be explained by simple binding of psoralen to openchromatin. From the available data, we conclude that the brightpsoralen signals on interbads and puffs represent unconstrained

negative supercoils in chromatin DNA. This does notcontradict the previous conclusion that bulk DNA of theeukaryote genome is relaxed (Sinden et al., 1980). Althoughwe detected negative supercoils on many interbands and puffs,DNA contents of these loci are extremely low and account fora tiny portion compared with bulk DNA.

The negative supercoils should be in equilibrium with theirgeneration and relaxation. The heterogeneity in the ratio ofsignal intensities of psoralen to nascent RNA on variousinterbands and puffs (Figs 6, 7) suggests that the equilibriumis not even throughout the genome. It has been shown thattopoisomerase I is associated with transcribed regions of activegenes (Fleischmann et al., 1984; Gilmour et al., 1986; Gilmourand Elgin, 1987). Interestingly, the level of RNA polymeraseII present on the transcriptionally active heat-shock genesexceeds the level of topoisomerase I by twofold to fourfold,whereas twofold more topoisomerase I than RNA polymerase

Fig. 3.Bright psoralen signalsdisappear upon previous nickingof DNA or inhibition oftranscription. Salivary glandswere irradiated with X-rays forrelaxation of DNA by nicking(A-C) or treated with α-amanitinfor inhibition of transcription(D-F) before photocrosslinkingwith biotinylated psoralen.(A,D) Biotinylated psoralen.(B,E) DAPI. (C,F) Merged.Representative interbands andpuffs are indicated by arrows.Bars, 10 µm.

Fig. 4.Distribution of psoralen signals on polytenechromosomes from heat-shocked larvae.(A) Biotinylated psoralen. (B) DAPI. (C) Merged.Major heat-shock puffs at 87A, 87B and 93D areindicated by arrows. Arrowheads representinterbands with biotinylated psoralen signals thatbecame less prominent after heat shock. Bar,10µm.

3802

II occupies the modestly transcribed copia gene (Gilmour etal., 1986). The different ratios of topoisomerase I and RNApolymerase II on different genes would contribute to theheterogeneity in the equilibrium between generation andrelaxation of negative supercoils.

Given that the difference in psoralen binding is abouttwofold greater for supercoiled DNA with a negativesuperhelicity of 0.05 (Sinden et al., 1980), why does ourfluorescence system work so clearly? There may be tworeasons for this. The first one is related to the accessibility ofpsoralen. Bulk chromatin DNA is tightly packaged and may beless accessible to psoralen compared with DNA in interbandsand puffs. The second one is a threshold effect of fluorescencedetection. Fluorescence is barely detectable under amicroscope if the concentration of a fluorescent molecule isbelow certain threshold level. We have used the lowestconcentration of biotinylated psoralen (0.2 ng/ml) at which wereally detected the bright signals. When we dropped below it,the signals blended into background. At this concentration, thedifference in the fluorescent signals between supercoiled andrelaxed DNA may be exaggerated.

Mechanism of negative supercoiling of DNAWhat is the mechanism underlying the generation of the

negative supercoils that are coupled with transcription? Themost potent mechanism is transcription-driven supercoiling ofDNA (Liu and Wang, 1987). As transcription proceeds, apositively supercoiled domain is formed in front of thetranscription machinery, and a negatively supercoiled domainis formed behind it. This type of supercoiling has beendocumented in eukaryote using a yeast mutant defective intopoisomerases (Giaever and Wang, 1988). According tothe twin-supercoiled domain model, one could imagineaccumulation of negative supercoils in the intergenic region oftwo oppositely oriented hsp70transcription units at 87A (Fig.1A). Consistent with the expectation, the ratios of signalintensities of psoralen to DNA formed a broad peak at 87Apuff (Fig. 7C).

In addition, chromatin remodeling could release negativesupercoils that are constrained by histone-DNA interactions.Indeed, SWI/SNF-type remodeling complexes have beenshown to reduce the constrained superhelicity (Kwon et al.,1994; Gavin et al., 2001). Unconstrained negative superhelicitycould be also released by transcription-induced displacementof a single H2A·H2B dimer from a nucleosome (Kireeva et al.,2002). The possibility is supported by a recent finding thatFACT (facilitates chromatin transcription, a heterodimer ofSSRP1 and SPT16) facilitates the displacement of H2Aand H2B from nucleosomal DNA (Formosa et al., 2002;

Journal of Cell Science 117 (17)

Fig. 5.Effect of previous nicking of DNA or inhibition of transcription on psoralen signals on heat-shock puffs. Salivary glands dissected fromheat-shocked larvae were irradiated with X-ray (A-C) or treated with α-amanitin (D-F) and then photocrosslinked with biotinylated psoralen.(A,D) Biotinylated psoralen. (B,E) DAPI. (C,F) Merged. Arrows indicate heat-shock puffs at 87A and 87B. Bars, 10 µm.

3803Visualization of negative supercoils of DNA

Belotserkovaskaya et al., 2003; Shimojima et al., 2003).Although the release of histone-DNA interactions in a singlenucleosome has only a subtle effect on the superhelical stateof DNA, the sum of these effects over nucleosome arrayswould generate a detectable level of negative supercoils.However, the processes are at least one order of magnitude lessefficient in generating supercoils per unit length of DNA thanthe transcription-driven supercoiling.

Finally, negative supercoils could be generated by theaction of supercoiling factor (SCF) and topoisomerase II.SCF is a protein capable of introducing negative supercoilsinto DNA in conjuction with topoisomerase II (Ohta andHirose, 1990). Drosophila SCF localizes to interbands andpuffs on polytene chromosomes and hence, it is thought tobe involved in the formation of transcriptionally activechromatin (Koyayashi et al., 1998). The idea is supported byrecent genetic studies from this laboratory (H. Furuhashi andS.H., unpublished). All three mechanisms proposed forgenration of negative supercoils are not mutually exclusivebut could operate simultaneously.

Implication of negative supercoils in transcriptionalregulationThe present study illuminates a dynamic nature of thesuperhelical state of DNA in many local domains of the

eukaryotic genome. On the basis of the data we proposetranscriptional regulation through conformation of chromatinDNA. For example, transcription activities from variouspromoters have been shown to change markedly by the degreeof DNA supercoiling (Harland et al., 1983; Hirose and Suzuki,1988; Mizutani et al., 1991; Schultz et al., 1992; Dunaway andOstorander, 1993; Parvin and Sharp, 1993; Tabuchi et al.,1993). Dissection of transcription revealed that the binding ofTATA element-binding protein (TBP) to TATA element isfacilitated by negative supercoiling of DNA in most genesexamined (Mizutani et al., 1991; Tabuchi et al., 1993).Therefore, transcription-coupled, unconstrained negativesupercoils of DNA shown here, in turn, can affect transcription.Although relaxation of DNA by X-ray treatment did notsignificantly affect transcription under our conditions, it doesnot necessarily exclude the above idea because once TBP bindsto the TATA element, subsequent relaxation of DNA will notreduce transcription until the TBP dissociates from thepromoter. However, negative supercoils are not necessarilydetectable in all transcriptionally active regions. Studies fromother group have shown active genes without unconstrainednegative supercoils (Kramer and Sinden, 1997; Kramer et al.,1999). It is possible that local DNA domains should beencompassed by some special structures such as chromatinboundaries to accumulate detectable levels of supercoils duringtranscription.

Fig. 6.Colocalization of bright psoralen signals with those of nascent RNA on polytene chromosomes. (A) Biotinylated psoralen. (B) DAPI.(C) Merged image of A and B. (D). Nascent RNA labeled with Br-UTP. (E) Merged image of (B) and (D). (F) Merged image of (A) and (D).Bar, 10 µm.

3804

Formation of unusual DNA structures such as cruciform(Lilley, 1980; Panayatatos and Wells, 1981) and Z-form(Nordheim et al., 1982; Singleton et al., 1992) is significantlyfacilitated by negative supercoiling and hence, the observednegative supercoils can also affect conformation of DNA. Suchunusual DNA structures are likely to affect transcription.Indeed, Z-DNA in a promoter region has been suggested toparticipate in transcriptional activation in collaboration with achromatin remodeling complex BAF (Liu et al., 2001). Finally,it is possible that negative supercoiling of DNA has influenceon higher order chromatin structure which, in turn, can affecttranscription.

We thank R. R. Sinden, J. C. Wang and A. Travers for helpfulsuggestions, and H. R. Drew for encouragement. This work was

supported by Grants in aid for Scientific Research from the Ministryof Education, Science, Sports, Culture and Technology of Japan. K.M.was supported by a Center of Excellence Program of Japan.

ReferencesArnström, G. and Edvardsson, K. A. (1974). Radiation-induced single-

strand breaks in DNA determined by rate of alkaline strand separation andhydroxylapatite chromatography, an alternative to velocity sedimentation.Int. J. Radiat. Biol. 26, 493-497.

Belotserkovskaya, R., Oh, S., Bondarenko, V. A., Orphanides, G.,Studitsky, V. M. and Reinberg, D. (2003). FACT facilitates transcription-dependent nucleosome alteration. Science301, 1090-1093.

Bird, A. P. and Wolffe, A. P. (1999). Methylation-induced repression-belts,braces, and chromatin. Cell 99, 451-454.

Chang, W. Y., Winegarden, N. A., Paraiso, J. P., Stevens, M. L. andWestwood, J. T. (2000). Visualization of nascent transcripts on Drosophila

Journal of Cell Science 117 (17)

Fig. 7.Colocalization of bright psoralen and nascent RNA signals on heat-shock puffs. (A) Comparison of psoralen and nascent RNA signalson polytene chromosomes dissected from heat-shocked larvae. (B,C) Detailed inspection on the heat-shock puffs at 87A and 87B. (C) Signalintensities of biotinylated psoralen (green), nascent RNA (pink) and DAPI (blue), and ratio of signal intensity of biotinylated psoralen to DNA(yellow) quantified along the region indicated in (B). Bar, 10 µm.

3805Visualization of negative supercoils of DNA

polytene chromosomes using BrUTP incorporation. Biotechniques29, 934-936.

Dunaway, M. and Ostrander, E. A. (1993). Local domains of supercoilingactivate a eukaryotic promoter in vivo.Nature361, 746-748.

Elgin, S. C. R. (1981). DNAase 1-hypersensitive sites of chromatin. Cell 27,413-415.

Fleischman, G., Pflugelder, G., Steiner, E. K., Javaherian, K., Howard, G.C., Wang, J. C. and Elgin, S. C. R. (1984). DrosophilaDNA topoisomeraseI is associated with transcriptionally active regions of the genome. Proc.Natl. Acad. Sci. USA81, 6958-6962.

Formosa, T., Ruone, S., Adams, M. D., Olsen, A. E., Eriksson, P., Yu, Y.,Phoades, A. R., Kaufman, P. D. and Stillman, D. J. (2002). Defects inSPT16or POB3(y FACT) in Saccharomyces cerevisiaecause dependenceon the Hir/Hpc pathway, polymerase passage may degrade chromatinstructure. Genetics162, 1557-1571.

Freeman, L. A. and Garrard, W. T. (1992). DNA supercoiling in chromatinstructure and gene expression. Crit. Rev. Eukaryot. Gene Expr.2, 165-209.

Gavin, I., Horn, P. J. and Peterson, C. L. (2001). SWI/SNF chromatinremodeling requires changes in DNA topology. Mol. Cell 7, 97-104.

Giaever, G. N. and Wang, J. C. (1988). Supercoiling of intracellular DNAcan occur in eukaryotic cells. Cell 55, 849-856.

Gilmour, D. S. and Elgin, S. C. R. (1987). Localization of specifictopoisomease I interactions within the transcribed region of active heatshock genes by using the inhibitor camptothecin. Mol. Cell. Biol. 7, 141-148.

Gilmour, D. S., Pflugelder, G., Wang, J. C. and Lis, J. T. (1986).Topoisomerase I interacts with transcribed regions in Drosophilacells. Cell.44, 401-407.

Harland, R. M., Weintraub, H. and McKnight, S. L. (1983). Transcriptionof DNA injected into Xenopusoocytes is influenced by template topology.Nature302, 38-43.

Hirose, S. and Suzuki, Y. (1988). In vitro transcription of eukaryotic genesis affected differently by the degree of DNA supercoiling. Proc. Natl. Acad.Sci. USA85, 718-722.

Jenuwein, T. and Allis, C. D. (2001). Translating the histone code. Science.293, 1074-1080.

Jupe, E. R., Sinden, R. R. and Cartwright, I. L. (1993). Stably maintainedmicrodomain of localized unrestrained supercoiling at a Drosophila heatshock gene locus.EMBO J. 12, 1067-1075.

Kireeva, M. L., Walter, W., Tchernajenko, V., Bondarenko, V., Kashlev,M. and Studitsky, V. M. (2002). Nucleosome remodeling induced by RNApolymerase II, loss of the H2A/H2B dimer during transcription.Mol. Cell9, 541-552.

Kobayashi, M., Aita, N., Hayashi, S., Okada, K., Ohta, T. and Hirose, S.(1998). DNA supercoiling factor localizes to puffs on polytenechromosomes in Drosophila melanogaster. Mol. Cell. Biol. 18, 6737-6744.

Kramer, P. R. and Sinden, R. R. (1997). Measurement of unrestrainednegative supercoiling and topological domain size in living cells.Biochemistry36, 3151-3158.

Kramer, P. R., Fragoso, G., Pennie, W., Htun, H., Lager, H. G. and Sinden,R. R. (1999). Transcriptional state of the mouse mammary tumor viruspromoter can affect topological domain size in vivo. J. Biol. Chem. 274,28590-28597.

Kroeger, P. E. and Rowe, T. C. (1992). Analysis of topoisomerase I and IIcleavage sites on the Drosophila actin and Hsp70 heat shock genes.Biochemistry31, 2492-2501.

Kwon, H., Imbalzano, A. N., Khavari, P. A., Kingston, R. E. and Green,M. R. (1994). Nucleosome disruption and enhancement of activator bindingby a human SWI/SNF complex. Nature370, 477-481.

Lilley, D. M. (1980). The inverted repeat as a recognizable structural featurein supercoiled DNA molecules. Proc. Natl. Acad. Sci. USA77, 6468-6472.

Liu, L. F. and Wang, J. C. (1987). Supercoiling of the DNA template duringtranscription. Proc. Natl. Acad. Sci. USA84, 7024-7027.

Liu, R., Liu, H., Chen, X., Kirby, M., Brown, P. O. and Zhao, K. (2001).

Regulation of CSF1 promoter by the SWI/SNF-like BAF complex. Cell 106,309-318.

Ljungman, M. and Hanawalt, P. C. (1992). Localyzed torsional tension inthe DNA of human cells. Proc. Natl. Acad. Sci. USA89, 6055-6059.

Mizutani, M., Ohta, T., Watanabe, H., Handa, H. and Hirose, S. (1991).Negative supercoiling of DNA facilitates an interaction betweentranscription factor IID and the fibroin gene promoter. Proc. Natl. Acad. Sci.USA88, 718-722.

Moran, L., Mirault, M. E., Tissieres, A., Lis, J., Schedl, P., Artavanis-Tsakonas, S. and Gehring, W. J. (1979). Physical map of two D.melanogasterDNA segments containing sequences coding for the 70,000dalton heat shock protein. Cell 17, 1-8.

Myohara, M. and Okada, M. (1987). Puff induction in permeabilizedDrosophilasalivary glands in a chemically defined medium. Dev. Biol.125,462-465.

Narlikar, G. J., Fan, H. Y. and Kingston, R. E. (2002). Cooperation betweencomplexes that regulate chromatin structure and transcription.Cell 108,475-487.

Nordheim, A., Lafer, E. M., Peck, L. J., Wang, J. C., Stollar, B. D. andRich, A. (1982). Negatively supercoiled plasmids contain left-handed Z-DNA segments as detected by specific antibody binding.Cell 31, 309-318.

O’Brien, T., Hardin, S., Greenleaf, A. and Lis, J. T. (1994). Phosphorylationof RNA polymerase II C-terminal domain and transcriptional elongation.Nature370, 75-77.

Ohta, T. and Hirose, S. (1990). Purification of a DNA supercoiling factorfrom the posterior silk gland of Bombyx mori. Proc. Natl. Acad. Sci. USA87, 5307-5311.

Panayotatos, N. and Wells, R. D. (1981). Cruciform structures in supercoiledDNA. Nature289, 466-470.

Parvin, J. D. and Sharp, P. A. (1993). DNA topology and a minimal set ofbasal factors for transcription by RNA polymerase II.Cell 73, 533-540.

Shimojima, T., Okada, M., Nakayama, T., Okawa, K., Iwamatsu, A.,Handa, H. and Hirose, S. (2003). Drosophila FACT contributes to Hoxgene expression through physical and functional interactions with GAGAfactor. Genes Dev.17, 1605-1616.

Shultz, M. C., Brill, S. J., Ju, Q., Sternglanz, R. and Reeder, R. H. (1992).Topoisomerases and yeast rRNA transcription, negative supercoilingstimulates initiation and topoisomerase activity is required for elongation.Genes Dev. 6, 1332-1341.

Sinden, R. R. and Ussery, D. W. (1992). Analysis of DNA structure in vivousing psoralen photobinding, measurement of supercoiling, topologicaldomains, and DNA-protein interactions. Methods Enzymol. 212, 319-335.

Sinden, R. R., Carlson, J. O. and Pettijohn, D. E. (1980). Torsional tensionin the DNA double helix measured with trimethylpsoralen in living E. colicells, analogous measurements in insect and human cells.Cell 21, 773-783.

Singleton, C. K., Klysik, J., Stirdivant, S. M. and Wells, R. D. (1982). Left-handed Z-DNA is induced by supercoiling in physiological ionic conditions.Nature299, 312-316.

Sogo, J. M., Ness, P. J., Widmer, R. M., Parish, R. W. and Koller, T. (1984).Psoralen-crosslinking of DNA as a probe for the structure of active nucleolarchromatin. J. Mol. Biol. 178, 897-919.

Tabuchi, H., Handa, H. and Hirose, S. (1993). Underwinding of DNA onbinding of yeast TFIID to the TATA element. Biochem. Biophys. Res.Commun.192, 1432-1438.

Udvardy, A., Maine, E. and Schedl, P. (1985). The 87A7 chromomere.Identification of novel chromatin structures flanking the heat shock locusthat may define the boundaries of higher order domains. J. Mol. Biol.185,341-358.

Ussery, D. W., Hoepfner, R. W. and Sinden, R. R. (1992). Probing DNAstructure with psoralen in vitro. Methods Enzymol.212, 242-262.

Vignari, M., Hassan, A. H., Neely, K. E. and Workman, J. L. (2000). ATP-dependent chromatin-remodeling complexes. Mol. Cell. Biol. 20, 1899-1910.