DNA Length Is a Critical Parameter for Eukaryotic Transcription In Vivo

MOLECULAR AND CELLULAR BIOLOGY, Oct. 1996, p. 5821–5829 Vol. 16, No. 10

0270-7306/96/$04.0010

DNA Length Is a Critical Parameter for Eukaryotic

Transcription In Vivo

JOCELYN E. KREBS AND MARIETTA DUNAWAY*

Department of Molecular and Cell Biology, University of California, Berkeley, Berkeley, California 94720-3202

Received 8 May 1996/Returned for modiﬁcation 24 June 1996/Accepted 19 July 1996

The organization of eukaryotic chromosomes into topological domains has led to the assumption that DNA

topology and perhaps supercoiling are involved in eukaryotic nuclear processes. Xenopus oocytes provide a

model system for studying the role of DNA topology in transcription. Linear plasmid templates for RNA

polymerases (Pols) I and II are not transcribed in Xenopus oocytes, while circular templates are transcrip-

tionally active. Here we show that circularity is not required for transcription of Pol I or Pol II promoters if

the linear template is sufﬁciently long (>17 to 19 kb). The Xenopus rRNA (Pol I) promoter is active in central

positions on a long linear template but is not transcribed when located near an end. Because supercoils

generated by transcription could be retained by viscous drag against the long template, these results are

consistent with a supercoiling requirement for this promoter. Surprisingly, the herpes simplex virus thymidine

kinase (Pol II) promoter is active even 100 bp from the end of the long template, indicating that template length

fulﬁlls a critical parameter for transcription that is not consistent with a supercoiling requirement. These

results show that DNA length has unrecognized importance for transcription in vivo.

Eukaryotic chromosomes are organized into topological do-

mains, so that a linear eukaryotic chromosome behaves like a

series of isolated closed loops (1, 5, 24). Because closed loops,

like circular DNAs, can be supercoiled, these topological do-

mains could potentially provide supercoiling for processes such

as transcription. However, despite extensive investigation, the

role of supercoiling and topological domains in eukaryotic

transcription is not clear. The classical approach for demon-

strating that a process requires supercoiling is to show that the

activity varies as a function of superhelical density. This ap-

proach has demonstrated that DNA supercoiling is required in

prokaryotes for initiation of DNA replication, site-speciﬁc re-

combination, and transcription of some genes (reviewed in

reference 14). This kind of analysis is possible in prokaryotes

because of the speciﬁcities of the prokaryotic topoisomerases.

In prokaryotes, DNA gyrase introduces negative supercoils or

relaxes positive supercoils and topoisomerase I relaxes only

negative supercoils; thus, the entire genome is more negatively

supercoiled in a topoisomerase I mutant. In contrast, eukary-

otic topoisomerases relax positive and negative supercoils

equally well. An additional complicating factor is that when

supercoiled plasmids are introduced into cells, they are imme-

diately relaxed by endogenous topoisomerases and subse-

quently re-supercoiled by packaging into nucleosomes. One

approach to overcoming these difﬁculties is to study the de-

pendence of transcription on supercoiling in vitro. Some eu-

karyotic genes are transcribed more efﬁciently in cell extracts if

they are supercoiled (13, 20, 30), although the presence of

topoisomerases in the extracts compromises the conclusion

that supercoiling facilitates transcription. Furthermore, exper-

iments show that the basal transcription factors IIE and IIH

are not required in a minimal in vitro transcription reaction if

the transcription template is negatively supercoiled (11, 23).

However, in one in vitro study the transcriptional activity of the

Saccharomyces cerevisiae rRNA promoter varied as a function

of superhelical density in transcription extracts prepared from

top1 top2 mutants (29). This evidence strongly suggests that

supercoiling facilitates transcription of this promoter.

There have also been suggestive results in vivo for the par-

ticipation of supercoiling in eukaryotic transcription. Linear-

ized plasmid DNAs are poorly transcribed in mammalian tis-

sue culture cells compared with circular forms of the same

templates (34), and linear plasmid templates for RNA poly-

merase (Pol) I and II genes are not transcribed at all in Xeno-

pus laevis oocytes (12, 26). Even when circular templates in

oocytes are allowed to establish transcription and are subse-

quently linearized by injection of a restriction enzyme, tran-

scription is dramatically reduced or entirely abolished (12, 26).

Linear and circular templates are equally stable in oocytes, so

the failure of linear templates to be transcribed is not due to

degradation of injected templates. Thus, circular templates

fulﬁll a critical requirement for eukaryotic transcription in

vivo. Nevertheless, it is not clear whether the requirement for

circular templates reﬂects a supercoiling requirement or

whether circularity per se is somehow required for transcrip-

tion. For example, free DNA ends in linear templates could be

inhibitory, perhaps because they provide an entry site for in-

hibitory proteins or initiate cooperative binding of an inhibitor.

Indeed, capping free DNA ends has previously been shown to

increase transcription of linear templates in transient transfec-

tion assays (6). To show a requirement for DNA supercoiling

in vivo therefore requires that we address the issue of whether

the ends of linear DNA are inhibitory and that we distinguish

a requirement for template circularity from a requirement for

DNA supercoiling.

One way to separate circularity from supercoiling is to gen-

erate supercoiled domains on a linear template. Transcription

itself generates positive supercoils ahead of the Pol and nega-

tive supercoils behind it if two critical requirements are ful-

ﬁlled (18). First, the RNA Pol must be restricted from rotating

freely around the DNA helix. Second, the supercoils that are

generated must not diffuse off the ends; this requirement can

be fulﬁlled by tethering the DNA to restrict its free rotation.

Transcription-driven supercoiling clearly occurs both in vitro

and in vivo, but the effect of transcription-driven supercoiling

on other cellular processes in vivo is uncertain (4, 10, 29, 31,

* Corresponding author. Phone: (510) 643-6208. Fax: (510) 642-

7846. Electronic mail address: [email protected].

5821

37). Further, the importance of transcript length, DNA- and

RNA-binding proteins, chromosomal context, the competing

relaxation of supercoils by topoisomerases, and other aspects

of the nuclear environment in restricting rotation of RNA Pol

and in preventing diffusion of supercoils in vivo is not known.

We have previously separated template circularity from su-

percoiling by generating transcription-driven supercoils on a

linear template in the Xenopus oocyte (7). While endogenous

Xenopus Pol I or II cannot normally transcribe linear plasmid

templates, T7 RNA Pol is able to do so when injected into

oocytes. We showed that the Xenopus rRNA promoter can be

activated on a linear template by transcription from a divergent

upstream T7 RNA promoter. This and additional experiments

from that study suggest that transcription-driven supercoiling

can provide the minimum topological requirement for tran-

scription for this promoter (7). These results further suggest

that even a short linear DNA with no known tethers can, at

least transiently, retain supercoils at sufﬁcient density to acti-

vate a promoter.

In this study we investigate the importance of template to-

pology and template length on transcription. We show that we

cannot activate transcription on a linear plasmid by competing

for possible inhibitory end-binding factors with DNA frag-

ments nor by blocking the ends of linear plasmids with DNA

hairpins. Most importantly, we show that both Pol I and Pol II

promoters are transcribed in long linear templates provided by

phage l cloning vectors. We have varied the positions of the

reporter genes within the l templates by circular permutation

in order to determine whether DNA ﬂanking the promoters

was required. The rRNA promoter is transcribed only when it

is located far from the ends of the long linear template, sug-

gesting that the role of DNA length for this promoter is to

FIG. 1. Construction of hairpin-ended templates. (A) Diagram of method used to construct hairpin-ended templates. Hairpin-ended templates were made by

cutting plasmids with a restriction enzyme and then ligating them to hairpin-forming oligonucleotides. This is described in detail in Materials and Methods. (B)

Ethidium bromide-stained alkaline agarose gels of the hairpin-ended templates. The third lane of each gel shows the hairpin constructs immediately after the ligation

step. The fourth lane of each gel shows the result of exonuclease III digestion of the samples in the third lanes.

5822 KREBS AND DUNAWAY MOL.CELL.BIOL.

prevent or retard diffusion of transcription-driven supercoils.

However, the thymidine kinase (TK) promoter is active

whether it is centrally located or near an end, indicating that

the essential function of DNA length for this promoter is not,

or is not solely, to maintain supercoiling. Instead, DNA length

serves some other essential role in transcription. We discuss

several possibilities for the function of DNA length in tran-

scription, including the idea that DNA length overcomes a

kinetic barrier to assembly of the transcription complex by

facilitated diffusion of transcription factors to the promoter.

MATERIALS AND METHODS

Template construction. To make the l templates, the 760-bp rRNA minigene,

c40 (16), was cloned into the EcoRI site of lgt11. The 2-kb herpes simplex virus

TK gene (19) was cloned into the SalI and XbaI sites of lgt22A. To make the

multimerized plasmid template, two different plasmids containing the TK gene

(pBR322 and pBluescript) were linearized with the restriction enzyme XbaI and

then ligated to form concatamers. Ligation products were separated by agarose

gel electrophoresis, and high-molecular-weight products were isolated from the

gel. These gel-puriﬁed products were then used to transform DH5a cells. The

multimer plasmid was linearized with a restriction enzyme (XhoI) that cut in only

one of the two plasmids.

Hairpin-ended templates were made by cutting plasmids containing either the

TK gene or the rRNA minigene with the restriction enzyme XbaI and then

ligating the linearized plasmids to hairpin-forming oligonucleotides (59-CTAGC

GCTGCAGACTTTTTGTCTGCAGCG-39). The ligated samples were then di-

gested with exonuclease III to remove nicked and unligated material. Hairpin-

ended templates were analyzed on agarose gels under alkaline conditions (0.05

M NaOH, 1 mM EDTA), in which hairpin-ended templates run with the mobility

of single-stranded dimer circles. Nicked hairpin templates run as linear dimers

and are therefore separable on alkaline gels. All detectable nicked or unligated

material was digested by exonuclease III.

Analysis of oocyte injection experiments. All templates were microinjected

into the nuclei of X. laevis oocytes in a volume of 20 nl; 0.5 ng of linear and

circular plasmid templates and reference templates was injected unless otherwise

speciﬁed. Long linear templates were injected at approximately 5 ng in order to

achieve an equimolar ratio of experimental reporter gene to reference. rRNA

templates were coinjected with 50 mgofa-amanitin per ml. Otherwise, injections

and harvesting of oocyte nucleic acids were performed as described elsewhere

(16).

Two assays were employed to detect transcription from injected templates.

Transcription from the TK gene and its corresponding reference, ctk, was de-

tected by primer extension, as described previously (19). Transcription from the

rRNA minigene and reference was detected by S1 nuclease protection, as de-

scribed elsewhere (16). All transcription levels are normalized to the level of

transcription from the circular plasmid reference, which is given a value of one.

The normalized transcription levels also take into account any deviations from an

equal molar ratio in the injected templates (determined by Southern blotting [see

below]) and, in the case of the S1 protection assays, differences in the speciﬁc

activities of the S1 probes. Values for normalized transcription represent the

averages for at least three experiments, and the standard errors are given.

Transcription assays and Southern blots were quantitated with a Phosphor

Imager (Molecular Dynamics).

The molar ratios of the competing reporter genes and the stability of injected

templates were veriﬁed by Southern blotting of injected samples. All templates

used in this study were stable over the entire course of an injection experiment

(as long as 24 h at room temperature). In the cases in which the templates were

not injected at exact equal molar ratios, the differences in template levels were

taken into account in the ﬁnal values for normalized transcription. The com-

FIG. 2. Linear plasmid templates with hairpin ends are not transcribed. (A) Primer extension analysis of TK transcription by Pol II. The positions of the TK

(experimental) and the circular reference plasmid (reference) primer extension products are indicated by arrows. The band between the reference and experimental

signals is seen in the absence of experimental templates (see reference only lane) and is an artifact of the primer extension. (B) Southern blot of injected TK templates

after recovery from oocytes. (C) S1 nuclease analysis of rRNA transcription. The positions of the c40 (experimental) and the circular reference plasmid (reference)

S1-protected signals are indicated by arrows. (D) Southern blot of injected rRNA templates after recovery from oocytes. HRPN, hairpin; N/R, nicked/relaxed; SC, supercoiled.

VOL. 16, 1996 DNA TOPOLOGY AND LENGTH EFFECTS ON TRANSCRIPTION 5823

pleteness of digestion of l templates was determined by digesting the injected l

DNAs with HindIII or PvuI to produce multiple small fragments. These digested

samples were then Southern blotted and probed with HindIII or PvuI fragments

that contain the sites for the enzymes used for circular permutation or deletion

of l templates. Full-size HindIII or PvuI fragments would be seen only if diges-

tion by the ﬁrst enzyme(s) was incomplete. Injection samples in which these

full-size fragments were detectable were discarded.

RESULTS

Free DNA ends do not inhibit transcription. We ﬁrst tested

the possibility that linear templates are not transcribed because

of the presence of free DNA ends in these templates. Free

DNA ends could inhibit transcription because of binding or

entry of proteins at the ends. To test this possibility, we blocked

the free ends of linear templates by preparing linear plasmid

templates with hairpin ends. The construction of these tem-

plates is shown in Fig. 1. Plasmids were cut with a restriction

enzyme (XbaI) and then ligated to self-complementary oligo-

nucleotides. These oligonucleotides form hairpin structures

with a 59 single-strand extension that is complementary to the

restriction site. After ligation, the DNA is treated with exonu-

clease III to degrade any unligated or nicked templates (Fig.

1A; see Materials and Methods). The constructs were analyzed

on denaturing alkaline agarose gels, as shown in Fig. 1B. With

these gels it is possible to separate both monomer- and dimer-

length circles and linear DNAs. Dimer-length circles corre-

spond to hairpin-ended templates. Dimer-length linear DNAs

correspond to nicked hairpin templates or single-hairpin tem-

plates. These templates are sensitive to exonuclease III diges-

tion (Fig. 1B, 1EXO III). The identities of the bands on the

alkaline gels were conﬁrmed by end labelling the hairpin oli-

gonucleotides with

P and subsequently exposing the alkaline

gels to ﬁlm (data not shown).

In this and all other experiments reported here we have

tested two different transcription templates: the herpes simplex

virus TK gene, transcribed by RNA Pol II, and the X. laevis 40S

rRNA gene (rRNA) transcribed by RNA Pol I. In all cases, the

molar ratios and stability of injected templates were veriﬁed by

Southern blotting (Fig. 2B and D and data not shown; see

Materials and Methods).

Hairpin-ended templates for each promoter were coinjected

with the appropriate circular reference template, and tran-

scription was assayed (Fig. 2A and C). Neither the 6-kb TK

template nor the 4-kb rRNA template is detectably transcribed

when linearized (Fig. 2A, tk linear; Fig. 2C, rRNA linear). The

limit of detection for the rRNA S1 assay is approximately 1%

of the signal seen for circular templates. The limit of detection

for the TK primer extension assay is at least 0.1% of the signal

seen for the circular reference. Blocking the ends of these

templates with hairpins does not allow them to be transcribed

(Fig. 2A, tk HRPN; Fig. 2C, rRNA HRPN). As a control, the

same plasmids were linearized, self-ligated to recircularize,

and digested with exonuclease III to mimic the end-blocking

procedure. These control templates were transcribed normally,

showing that this procedure itself does not render templates

FIG. 3. Free DNA ends do not inhibit transcription. (A) Schematic of ex-

periment. Linear transcription template was coinjected with a circular reference

plasmid and various amounts of excess linear fragments (with EcoRI or SalI

termini). (B) Primer extension analysis of TK transcription with (tk linear 1 free

ends) and without (tk linear) excess free DNA ends. The triangle indicates

increasing molar excess of free ends of 2-fold, 20-fold, and 200-fold (left to

right).

FIG. 4. TK and rRNA promoters in linear l templates are transcribed. (A)

Templates used in the injection assay. The reporter genes used in previous

experiments were cloned into l vectors to create 45-kb linear templates. l

templates were injected at a molar ratio equal to that of the circular reference

plasmids. (B) Primer extension analysis of TK transcription. (C) S1 nuclease

analysis of rRNA transcription.

5824 KREBS AND DUNAWAY MOL.CELL.BIOL.

inactive (Fig. 2A, HRPN control). Southern blots of the in-

jected templates are also shown (TK templates, Fig. 2B; rRNA

templates, Fig. 2D). The hairpin-ended templates were still

resistant to exonuclease III digestion after reisolation from the

oocytes (data not shown).

We next attempted to compete for inhibitory end-binding

factors by coinjecting a linear template with increasing concen-

trations of free ends. No transcription was detected from linear

TK templates in the presence of a 2-, 20-, or 200-fold molar

excess of free ends (Fig. 3). The linear rRNA template also was

FIG. 5. Transcription of circular permutations of l templates. (A) Diagram of the procedure and consequences of circular permutation of the l templates. Circular

permutations were made by ligating the cos sites of l templates to form concatemers and then digesting the concatemers with the restriction enzymes indicated.

Linearization of injected templates was determined by Southern blotting. (B) Summary of transcription of lTK circular permutations. The direction of TK transcription

(arrow) and length of ﬂanking DNA in kilobases (numbers above templates) are indicated. Transcription of l templates was normalized to transcription of the circular

reference template. Values represent the averages from at least three experiments, and standard errors are given. (C) Summary of transcription of lrRNA circular

permutations.

VOL. 16, 1996 DNA TOPOLOGY AND LENGTH EFFECTS ON TRANSCRIPTION 5825

not activated by competing free ends (data not shown). Since

no transcription was detected even at the highest levels of

competitor free ends and since hairpin-ended linear plasmids

are not transcriptionally active, it is unlikely that the failure of

short linear templates to be transcribed is due to the binding of

inhibitory factors to their free ends.

Long linear templates are transcriptionally active. We next

tested whether template length was a factor in transcription of

linear templates. We reasoned that long DNA arms ﬂanking a

transcription template might act as a barrier to diffusion of

supercoils due to friction against the mass of DNA and pro-

teins. If this were the case, then negative supercoils generated

behind a Pol during transcription from a promoter in a long

template might be sustained long enough to facilitate further

initiation events.

Long linear templates were constructed by cloning either the

rRNA or the TK reporter gene into phage l vectors (Fig. 4A).

These templates are approximately 45 kb in length, and the

reporter genes are centrally located within the vector DNA

when linearized at the cos sites. The l templates were coin-

jected with an equimolar amount of circular reference plasmid,

and transcription was assayed (Fig. 4B and C). In dramatic

contrast to linear plasmids, both the rRNA and TK genes are

transcribed on long linear templates (Fig. 4B, lTK; Fig. 4C,

lrRNA).

The position of the rRNA promoter within a linear template

is critical for transcription. If transcription on the long linear

templates resulted from the formation of localized domains of

supercoiling due to a diffusion barrier provided by long DNA

arms, then transcription of a promoter in a long linear tem-

plate would be expected to depend on its position within the

template. Speciﬁcally, a promoter located near the end of a

template will not be transcribed as efﬁciently, since supercoils

diffuse rapidly off a short linear stretch of DNA. We tested the

importance of promoter position within the l templates by

making circular permutations of the l templates (Fig. 5A). The

cos ends of lTK or lrRNA were ligated in vitro to form a

mixture of concatemers and circular templates. Aliquots of

these mixtures were each digested with speciﬁc restriction en-

donucleases to create linear templates in which the position of

the reporter gene varies with respect to the ends of the tem-

plates. This circular permutation not only allowed us to test the

position dependence of transcription in these templates but

also allowed us to eliminate the possibility that the transcripts

detected from the l templates in Fig. 3 arose from templates in

which the l cos sites had annealed in vivo. The cos ends of the

l templates have much longer overhangs (12 bp) compared to

those generated by restriction enzyme cleavage of plasmid

DNAs. In the circularly permuted templates, the cos sites are

ligated and located internally, while the template ends are

generated by restriction enzymes like the linear plasmid tem-

plates. Permuted lrRNA templates were coinjected with cir-

cular reference plasmids, and transcription was assayed and

quantitated as described above (Fig. 5B). Templates in which

the rRNA promoter is still ﬂanked with long DNA arms but

that have restriction enzyme overhangs instead of long com-

plementary ends from the cos sites are transcribed with efﬁ-

ciency equal to that of the original l templates (Fig. 5B, cos

versus NheI). Therefore, the difference between transcription

on l templates and linear plasmid templates is not due to

FIG. 6. Transcription of shortened l templates. Shortened templates were made by digesting l templates with the restriction enzymes indicated, and the resultant

mixture of DNA fragments was coinjected with the reference plasmid into oocytes. (A) Summary of transcription of shortened lTK templates. (B) Summary of

transcription of shortened lrRNA templates. The direction of TK and rRNA transcription (arrows), the length of ﬂanking DNA in kilobases (numbers above

templates), and total template length are indicated.

5826 KREBS AND DUNAWAY MOL.CELL.BIOL.

reannealing of the ends. We have also prevented annealing of

cos sites directly by ﬁlling in the cos overhangs with Klenow

Pol, and we see that these templates are transcribed to the

same extent as those shown in Fig. 4 (data not shown).

The position of the rRNA promoter was critical for tran-

scription in the long template, however. This promoter is not

transcribed when located within 100 bp of the end of the

template (Fig. 5B, KpnI and Asp718), regardless of whether the

end has a 39 or 59 overhang. This observation is consistent with

our original idea that long linear templates might allow do-

mains of supercoiling to accumulate and that this makes tran-

scription possible on a linear template.

Transcription of the TK promoter is independent of posi-

tion. To our surprise, TK and rRNA promoters do not have the

same dependence on position within the l template. In con-

trast to the rRNA promoter, moving the TK promoter close to

the end of the l template had only modest effects on transcrip-

tion. Transcription is only decreased about twofold in TK tem-

plates in which the TK gene is between 100 bp and 1 kb from

an end (Fig. 5C, SalI, ScaI, and XbaI). Transcription was ob-

served whether the 59 or 39 end of the gene was very close to

an end, suggesting that transcription from this promoter is not

dependent on supercoiling. DNA length is apparently provid-

ing some other critical function for this promoter.

A minimum length is required for a linear template to be

transcriptionally active. Since 45-kb linear l templates are

transcribed but 6-kb linear plasmids are not, we tested tem-

plates of intermediate sizes for transcriptional activity. The ltk

and lrRNA templates were digested with restriction enzymes

to create shortened templates, and the mixture of DNA frag-

ments was injected into oocytes. The shortened templates and

a summary of the transcription data are shown in Fig. 6. Both

the TK and rRNA promoters are active in templates as small

as 20 kb (Fig. 6A, XhoI, HindIII, and StuI-XhoI; Fig. 6B, BglII,

SnaBI-BglII, and StuI). However, neither promoter is active in

templates in the 10- to 14-kb range (Fig. 6A, ScaI-XhoI and

PvuI; Fig. 6B, PvuI).

Once a critical minimum length is reached, the level of

transcription increases with increasing length for the shortened

l templates (Fig. 7). We have not tested templates longer than

45 kb, but we note that transcription of these long linear

templates is still less than for a circular template. The mini-

mum length is similar for ltk and lrRNA, and the shape of the

curve is also similar. Since transcription does increase with

length in these templates, it seems that length alone can de-

termine efﬁciency of transcription of a template.

Linear plasmid multimers of sufﬁcient size are transcribed.

To conﬁrm that length rather than a sequence context was the

critical parameter for transcription of the l templates, we

constructed a long TK template by multimerizing plasmids that

are not transcribed as linear monomers (Fig. 8A). Two differ-

ent TK gene-containing plasmids (pBR322 and pBluescript)

were linearized with XbaI and ligated to form concatemers.

Neither of these TK gene-containing plasmids is transcribed as

a linear monomer (TK gene in pBR322, Fig. 2; TK gene in

pBluescript, data not shown). The ligated material was trans-

formed into Escherichia coli, and multimers having only one

pBluescript vector were selected. Thus, the multimerized plas-

mid shown can be linearized by digestion with any restriction

enzyme having a site unique to pBluescript.

Transcription of the 17-kb multimer TK template is shown

in Fig. 8B. The linear plasmid multimer is transcribed at ap-

FIG. 7. Transcription increases with increasing length. Plot of transcription level versus template length for lTK and lrRNA templates. Units of transcription are

arbitrary. The plot includes data from circular permutations and shortened templates.

VOL. 16, 1996 DNA TOPOLOGY AND LENGTH EFFECTS ON TRANSCRIPTION 5827

proximately one-third of the level of the circular reference

plasmid (normalized with respect to the copy number of TK

genes in the construct). This level of transcription is similar to

that of the long TK template provided by the l vector. This

indicates that length alone provides an essential function for

transcription of linear templates.

DISCUSSION

We have shown that long linear templates are transcribed by

both Pol I and Pol II with an efﬁciency comparable to that of

circular plasmids. Transcription is not due to speciﬁc se-

quences in the l vector, because linear plasmid multimers of

sufﬁcient length are transcribed. Thus, a linear plasmid tem-

plate that is transcriptionally inactive can be activated simply

by increasing its length. The minimum length required for

transcription of either RNA Pol I or Pol II templates is be-

tween 14 and 17 kb, only a three- to ﬁvefold increase over the

length of typical plasmids. A similar effect is observed in tissue

culture transfection experiments, in which circular templates

are preferred to linear templates when the templates are at low

concentrations, but at high DNA concentrations, linear tem-

plates are transcribed. In these cases, the high DNA concen-

tration leads to concatemerization of the linear plasmids (25,

33, 34). Because there is no end-joining activity in Xenopus

oocytes, short linear templates do not form concatemers even

when they are injected at high concentrations in our assays.

It is clear from the experiments reported here that DNA

length facilitates transcription of RNA Pol I and II genes in

different ways. Transcription of the Xenopus rRNA promoter is

dependent on the position of the promoter with respect to the

ends of the DNA, consistent with the idea that a topological

domain can be created on a large linear template. The viscous

drag against a large DNA molecule is apparently sufﬁcient to

prevent transcription-generated supercoils from diffusing rap-

idly off the end of the DNA, so that DNA length creates a

topological domain. These results, especially when taken with

previous data showing that transcription-driven supercoiling

can activate the rRNA promoter (7), suggest that this pro-

moter requires supercoiling. The two studies taken together

are the best evidence to date that supercoiling is required for

any function in vivo in a eukaryote.

In contrast to rRNA transcription, the transcription of the

TK promoter is independent of its position within the l tem-

plate. Templates in which either the 59 or 39 end of the TK

gene is near the end are transcribed with efﬁciency similar to

the efﬁciency of those in which the TK promoter is centrally

located within the l DNA. These observations suggest that

long DNA serves different functions for the two genes tested

and separates transcription of the two genes mechanistically.

This distinction is further supported by the experiments using

T7 transcription to activate eukaryotic promoters. The TK

promoter could not be activated by transcription from a diver-

gent T7 promoter (6a).

If the results with TK are not consistent with a supercoiling

requirement, then what is the advantage to having a long

template for transcription? A classic problem in molecular

biology is how a sequence-speciﬁc DNA-binding protein ﬁnds

its target sequence within the vast excess of nonspeciﬁc DNA

(17, 32). Given the relative afﬁnities of sequence-speciﬁc

DNA-binding proteins for their target sequences and for non-

speciﬁc DNA, these proteins would never ﬁnd their targets if

the nonspeciﬁc DNA acted only as a competitor. Our experi-

ments show that the additional nonspeciﬁc DNA present in cis

on the long templates facilitates transcription and therefore

cannot be considered a competitor in the traditional sense.

There is previous evidence that DNA length can facilitate a

process. The rate at which site-speciﬁc DNA-binding proteins

ﬁnd their sites is increased (8, 15, 35, 36). For example, the

association of lac repressor with an operator sequence is ap-

proximately 10-fold faster in a l template than in an oligonu-

cleotide duplex (2, 27, 28). This rate is about 1,000 times faster

than one would expect for a diffusion-controlled macromolec-

ular association. The increased rate of association on long

DNA is proposed to occur by a process of facilitated diffusion

in which DNA-binding proteins associate with nonspeciﬁc

DNA and either slide or hop along the DNA contour or are

directly transferred from one segment of DNA to another (3,

32). This reduces a three-dimensional search to a one- or

two-dimensional search. Thus, the activation of the TK pro-

moter on long templates fulﬁlls one criterion discussed by von

FIG. 8. Linear plasmid multimers are transcribed. (A) Construction of the

TK multimer template. Two different plasmids containing the TK gene (pBR322

and pBluescript SK) were linearized with the restriction enzyme XbaI and then

ligated to form concatemers. Ligation products were separated by agarose gel

electrophoresis, and high-molecular-weight products were isolated from the gel.

These gel-puriﬁed products were then used to transform E. coli. The multimer

plasmid was linearized with a restriction enzyme (XhoI) that cuts in pBluescript

only. (B) Primer extension analysis of TK transcription of multimerized tem-

plates.

5828 KREBS AND DUNAWAY MOL.CELL.BIOL.

Hippel and Berg as indicative of a process in which facilitated

diffusion is important: the rate of the process is increased when

template length is increased (32). Although our experiments

do not address which step or steps in the transcription process

are rate limiting in these different templates, it is possible that

the increased rate at which sequence-speciﬁc transcriptional

activators ﬁnd their targets facilitates assembly of the tran-

scription complex, perhaps by providing a surface for nucle-

ation of the complex. There are other possible roles for a long

template, such as attachment to the nuclear architecture or

targeting to a nuclear compartment rich in transcription fac-

tors. Attachment, for example, could create DNA loops that

mimic the topological domains of the chromosome. In either

case, our results show that such attachment or targeting must

be triggered solely by DNA length, not speciﬁc sequences.

There are other examples of the importance of DNA length

in cellular processes. DNA length is important for chromo-

some segregation and nuclear assembly (9, 21, 22), though the

role of DNA length is not understood for either process. It is

clear that the dynamics of long DNAs play a critical role in

many nuclear activities, and studies of these dynamics will lead

to greater understanding not only of these processes but of the

structure and function of the chromosome itself.

ACKNOWLEDGMENTS

We thank N. Cozzarelli, M. Botchan, R. Harland, L. Zechiedrich, S.

Uptain, and C. Robinett for critical reading of the manuscript.

This work was supported by a grant to M.D. from the National

Institute of General Medical Sciences.

REFERENCES

1. Benyajati, C., and A. Worcel. 1976. Isolation, characterization, and structure of

the folder interphase genome of Drosophila melanogaster. Cell 9:393–407.

2. Berg, O. G., and M. Ehrenberg. 1982. Association kinetics with coupled

three- and one-dimensional diffusion. Chain length dependence of the asso-

ciation rate of speciﬁc DNA sites. Biophys. Chem. 15:41–51.

3. Berg, O. G., R. B. Winter, and P. H. von Hippel. 1981. Diffusion-driven

mechanisms of protein translocation on nucleic acids. Models and theory.

Biochemistry 20:6929–6948.

4. Brill, S. J., and R. Sternglanz. 1988. Transcription-dependent DNA super-

coiling in yeast DNA topoisomerase mutants. Cell 54:403–411.

5. Cook, P. R., and I. A. Brazell. 1975. Supercoils in human DNA. J. Cell Sci.

19:261–279.

6. Courey, A. J. 1986. Structural and functional aspects of DNA supercoiling.

Ph.D. thesis. Harvard University, Cambridge, Mass.

6a.Dunaway, M., and E. Ostrander. Unpublished observations.

7. Dunaway, M., and E. A. Ostrander. 1993. Local domains of supercoiling

activate a eukaryotic promoter in vivo. Nature (London) 361:746–748.

8. Ehbrecht, H.-J., A. Pingoud, C. Urbanke, G. Maas, and C. Gualerzi. 1985.

Linear diffusion of restriction endonucleases on DNA. J. Biol. Chem. 260:

6160–6166.

9. Forbes, D. J., M. W. Kirschner, and J. W. Newport. 1983. Spontaneous

formation of nucleus-like structures around bacteriophage DNA microin-

jected into Xenopus eggs. Cell 34:13–23.

10. Giaever, G. N., and J. C. Wang. 1988. Supercoiling of intracellular DNA can

occur in eukaryotic cells. Cell 55:849–856.

11. Goodrich, J. A., and R. Tjian. 1994. Transcription factors IIE and IIH and

ATP hydrolysis direct promoter clearance by RNA polymerase II. Cell 77:

145–156.

12. Harland, R. M., H. Weintraub, and S. L. McKnight. 1983. Transcription of

DNA injected into Xenopus oocytes is inﬂuenced by template topology.

Nature (London) 301:38–43.

13. Hirose, S., and Y. Suzuki. 1988. In vitro transcription of eukaryotic genes is

affected differently by the degree of DNA supercoiling. Proc. Natl. Acad. Sci.

USA 85:718–722.

14. Kanaar, R., and N. R. Cozzarelli. 1992. Roles of supercoiled DNA structure

in DNA transactions. Curr. Opin. Struct. Biol. 2:369–379.

15. Khoury, A. M., H. J. Lee, M. Lillis, and P. Lu. 1990. Lac repressor-operator

interaction: DNA length dependence. Biochim. Biophys. Acta 1087:55–60.

16. Labhart, P. L., and R. H. Reeder. 1984. Enhancer-like properties of the

60-81 bp elements in the ribosomal gene spacer of Xenopus laevis. Cell

37:285–289.

17. Lin, S.-Y., and A. D. Riggs. 1975. The general afﬁnity of lac repressor for E.

coli DNA: implications for gene regulation in procaryotes and eucaryotes.

Cell 4:107–111.

18. Liu, L. F., and J. C. Wang. 1987. Supercoiling of the DNA template during

transcription. Proc. Natl. Acad. Sci. USA 84:7024–7027.

19. McKnight, S. L., E. R. Gavis, and R. Kingsbury. 1981. Analysis of transcrip-

tional regulatory signals of the HSV thymidine kinase gene: identiﬁcation of

an upstream control region. Cell 25:385–398.

20. Mizushima-Sugano, J., and R. G. Roeder. 1986. Cell-type-speciﬁc transcrip-

tion of an immunoglobulin kappa light chain gene in vitro. Proc. Natl. Acad.

Sci. USA 83:8511–8515.

21. Murray, A. W., N. P. Schultes, and J. W. Szostak. 1986. Chromosome length

controls mitotic segregation in yeast. Cell 45:529–536.

22. Murray, A. W., and J. W. Szostak. 1983. Construction of artiﬁcial chromo-

somes in yeast. Nature (London) 305:189–193.

23. Parvin, J. D., and P. A. Sharp. 1993. DNA topology and a minimal set of

basal factors for transcription by RNA polymerase II. Cell 73:533–540.

24. Paulson, J. R., and U. K. Laemmli. 1977. The structure of histone-depleted

metaphase chromosomes. Cell 12:817–828.

25. Perucho, M., D. Hanahan, and M. Wigler. 1980. Genetic and physical link-

age of exogenous sequences in transformed cells. Cell 22:309–317.

26. Pruitt, S. C., and R. H. Reeder. 1984. Effect of topological constraint on

transcription of ribosomal DNA in Xenopus oocytes: comparison of plasmid

and endogenous genes. J. Mol. Biol. 174:121–139.

27. Riggs, A. D., S. Bourgeois, and M. Cohn. 1970. Lac repressor-operator

interaction. Kinetic studies. J. Mol. Biol. 53:401–417.

28. Riggs, A. D., H. Suzuki, and S. Bourgeois. 1970. Lac repressor-operator

interaction. Equilibrium kinetics. J. Mol. Biol. 48:67–83.

29. Schultz, M. C., S. J. Brill, Q. Ju, R. Sternglanz, and R. H. Reeder. 1992.

Topoisomerases and yeast rRNA transcription: negative supercoiling stim-

ulates initiation and topoisomerase activity is required for elongation. Genes

Dev. 6:1332–1341.

30. Sive, H. L., N. Heintz, and R. G. Roeder. 1986. Multiple sequence elements

are required for maximal in vitro transcription of a human histone H2B gene.

Mol. Cell. Biol. 6:3329–3340.

31. Tsao, Y. P., H. Y. Wu, and L. F. Liu. 1989. Transcription-driven supercoiling

of DNA: direct biochemical evidence from in vitro studies. Cell 56:111–118.

32. von Hippel, P. H., and O. G. Berg. 1989. Facilitated target location in

biological systems. J. Biol. Chem. 264:675–678.

33. Weber, F., and W. Schaffner. 1985. SV40 enhancer increases RNA polymer-

ase density within the linked gene. Nature (London) 315:75–77.

34. Weintraub, H., P. F. Cheng, and K. Conrad. 1986. Expression of transfected

DNA depends on DNA topology. Cell 46:115–122.

35. Winter, R., and P. H. von Hippel. 1981. Diffusion-driven mechanisms of

protein translocation on nucleic acids. The Escherichia coli repressor-oper-

ator interaction: equilibrium measurements. Biochemistry 20:6948–6960.

36. Winter, R. B., O. G. Berg, and P. H. von Hippel. 1981. Diffusion-driven

mechanisms of protein translocation on nucleic acids. The Escherichia coli

lac repressor-operator interaction: kinetic measurements and conclusions.

Biochemistry 20:6961–6977.

37. Wu, H.-Y., S. Shyy, J. C. Wang, and L. F. Liu. 1988. Transcription generates

positively and negatively supercoiled domains in the template. Cell 53:433–

440.

VOL. 16, 1996 DNA TOPOLOGY AND LENGTH EFFECTS ON TRANSCRIPTION 5829