Biochimica et Biophysica Acta 1577 (2002) 191 – 207

www.bba-direct.com

Review
Promoter clearance and escape in prokaryotes
Lilian M. Hsu *
Program in Biochemistry, Mount Holyoke College, South Hadley, MA 01075, USA
Received 21 June 2002; accepted 21 June 2002

Abstract

Promoter escape is the last stage of transcription initiation when RNA polymerase, having initiated de novo phosphodiester bond
synthesis, must begin to relinquish its hold on promoter DNA and advance to downstream regions (DSRs) of the template. In vitro, this
process is marked by the release of high levels of abortive transcripts at most promoters, reflecting the high instability of initial transcribing
complexes (ITCs) and indicative of the existence of barriers to the escape process. The high abortive initiation level is the result of the
existence of unproductive ITCs that carry out repeated initiation and abortive release without escaping the promoter. The formation of
unproductive ITCs is a widespread phenomenon, but it occurs to different extent on different promoters. Quantitative analysis of promoter
mutations suggests that the extent and pattern of abortive initiation and promoter escape is determined by the sequence of promoter elements,
both in the promoter recognition region (PRR) and the initial transcribed sequence (ITS). A general correlation has been found that the
stronger the promoter DNA-polymerase interaction, the poorer the ability of RNA polymerase to escape the promoter. In gene regulation,
promoter escape can be the rate-limiting step for transcription initiation. An increasing number of regulatory proteins are known to exert their
control at this step. Examples are discussed with an emphasis on the diverse mechanisms involved. At the molecular level, the X-ray crystal
structures of RNA polymerase and its various transcription complexes provide the framework for understanding the functional data on
abortive initiation and promoter escape. Based on structural and biochemical evidence, a mechanism for abortive initiation and promoter
escape is described.
D 2002 Elsevier Science B.V. All rights reserved.

Keywords: Promoter escape; Abortive initiation; Sigma release/displacement; Scrunching/translocation; Stressed intermediate

1. Overview first step towards elongation. Promoter escape is linked

1.1. Scope and substance of this review
Promoter escape is the stage of transcription straddling
the initiation and elongation phases. Depending on one’s
research emphasis, promoter escape can be viewed as the
last step of transcription initiation, or the ensuing tentative
Abbreviations: ITC, initial transcribing complex; ITS or ITR, initial
transcribed sequence or region; PRR, promoter recognition region; RPc,
RNA polymerase – promoter ‘‘closed’’ complex; RPo, RNA polymerase –
promoter ‘‘open’’ complex; TEC, ternary elongation complex; nt,
nucleotides; bp, base pair; NTP, nucleoside triphosphates; a-CTD, the C-
terminal domain of the alpha subunits of bacterial RNA polymerase; UP,
the upstream sequence element; FRET, fluorescence resonance energy
transfer; DSR, downstream region; DBS, downstream binding site; j2, j3,
j4, sigma domains 2, 3, 4; jR1.1, sigma region 1.1; FeBABE, iron-(S)-1-
[ p-(bromoacetamido)benzyl] ethylenediaminetetraacetate; yeast Pol II,
yeast RNA polymerase II
*
Tel.: +1-413-538-2609.
E-mail address: lhsu@mtholyoke.edu (L.M. Hsu).
intimately to the phenomenon of abortive initiation which,
when first discovered, was considered enigmatic and puz-
zling. Yet its mechanism stands poised to be solved, due to
substantial recent advances—quantitative and qualitative—
in our knowledge of transcription. The diverse experimental
approaches applied to studying the functional and structural
interactions between RNA polymerase and nucleic acids are
leading to the development of a mechanism of promoter
escape. Thus, this is an exciting time to be in the area of
promoter escape.
Two novel properties of RNA polymerase reported at the
beginning of the 1990s turned the tide of the transcription
field. These were the observations that RNA polymerase
contains a transcript cleavage activity [1] and that at specific
sequences, can undergo transcriptional arrest forming
‘‘dead-end’’ complexes [2,3]. These observations spurred
the discovery of protein factors GreA, GreB, and TFIIS [4–
6; also see Fish and Kane, this volume], which were shown
subsequently to stimulate the intrinsic cleavage activity of

0167-4781/02/$ - see front matter D 2002 Elsevier Science B.V. All rights reserved.
PII: S 0 1 6 7 - 4 7 8 1 ( 0 2 ) 0 0 4 5 2 - 9
192 L.M. Hsu / Biochimica et Biophysica Acta 1577 (2002) 191–207
RNA polymerase [7,8]. At the same time, transcriptional
arrest was shown to result from a relative movement of the
RNA polymerase active site on the template DNA. This
movement repositions the RNA polymerase active center to
an upstream site, thus putting it out of alignment with the 3V-
OH growing point of the nascent transcript [9 – 13]. To
revive the arrested complexes, the relocated RNA polymer-
ase active center performs hydrolytic or pyrophosphorolytic
cleavage, providing itself with a new RNA 3V-OH, and
elongation synthesis can resume. The phenomena of abor-
tive initiation and transcriptional arrest were joined when
transcript cleavage stimulatory factors were shown to facil-
itate productive RNA synthesis at several promoters where
RNA polymerase is stalled at the escape stage, stimulating
the cleavage of abortive RNAs by the polymerase, and
allowing re-synthesis [14 – 17]. Thus, some abortive tran-
scripts may have been transiently associated with an arrested
initial transcribing complex (ITC) prior to release.
Cleavage and arrest activities of RNA polymerase raised
the issue of how, in general, transcriptional elongation
complexes are stabilized. Two models for the elongation
complexes triggered a flood of experiments to test their
central postulates [18,19]. Enzymatic and chemical foot-
printing analysis and cross-linking – mapping studies on
halted ternary elongation complexes (TECs) played a major
role in revealing the molecular properties and configuration
of elongation complexes (reviewed in Refs. [20,21]). Recent
unveiling of X-ray crystal structures further provided a
three-dimensional view of RNA polymerase and its com-
plexes [22 – 30], making it possible to reconcile the varied
biochemical activities with the dynamic nature of this
enzyme machine [31].
The topic of promoter escape was reviewed in 1996
when there was a renewed effort to subject abortive initia-
tion and promoter escape to systematic quantitative analysis
[32]. This effort has begun to yield answers. In this review, I
will focus on updating the biochemical reactivity of abortive
initiation and promoter escape by E. coli RNA polymerase
holoenzyme Ej70. The abortive initiation properties of T7
RNA polymerase are fundamentally similar to that of Ej70
[33 –35] and the details will not be examined. Investigation
of promoter escape in eukaryotic RNA polymerases has
proceeded with an emphasis somewhat different than that of
the prokaryotes, largely due to the complexity of RNA
polymerase molecules as well as the difficulty of obtaining
promoter-specific transcription. This topic is reviewed sep-
arately [see Dvir, this volume].
2. Promoter escape in the context of transcription
initiation
2.1. Definition and distinction of terms
Promoter clearance and promoter escape were both
coined [36,37] to describe the phenomenon of abundant
transcription and release of short template-specified RNAs
(i.e. abortive initiation), delaying the movement of initiated
RNA polymerase into the elongation phase [37 – 40].
‘‘Escape’’ deals directly with issues affecting the down-
stream movement of a polymerase molecule. ‘‘Clearance,’’
on the other hand, implies a sufficient movement of the
enzyme downstream to avail the core promoter elements for
binding a second polymerase. Of the two terms, I will use
promoter escape preferentially in this review.
2.2. Promoter escape: the process
Promoter escape refers to the last stage of transcription
initiation spanning the de novo synthesis of the first phos-
phodiester bond to the formation of an elongation complex.
On many promoters, this distance of 10 –15 bp on the DNA
template prescribes an arduous journey for the polymerase
molecule. During this process, RNA polymerase at each
promoter appears to synthesize and release a characteristic
set of abortive RNAs. These short transcripts usually range
2 – 12 nucleotides (nt) in length, although abortive tran-
scripts as long as 15 –17 nt have been reported [16,41].
At most promoters where transcription initiation has been
quantitatively examined, abortive initiation occurs at high
molar excess to template concentration under single cycle
conditions [37 – 40]. Abortive transcription is not affected
by heparin (which binds free RNA polymerase) but is
greatly diminished by rifampicin (which inhibits the for-
mation of the second phosphodiester bond). These observa-
tions led to the conclusion that, during abortive initiation,
RNA polymerase cycles repeatedly at the promoter without
dissociating from the template but rather remains bound as
open and initial transcribing complexes [37]. The picture of
an ITC inferred from these observations is one of high
instability: when the short transcripts are released, these
complexes resume the open complex structure and can
begin another round of chain initiation. Footprinting anal-
yses confirm that ITCs are essentially indistinguishable
from the open complexes [3,42].
2.3. Kinetic diagram of transcription initiation
To understand the promoter escape process and its
contribution to promoter activity, it is necessary to place
this process in the context of the multi-step program of
transcription initiation (Fig. 1A). On most E. coli Ej70
promoters, RNA polymerase holoenzyme binds the pro-
moter DNA to set up an active catalytic complex through
two major conformational transitions. Upon first binding to
the double-stranded promoter DNA, a complex lacking
catalytic activity called the closed complex (RPc) is formed.
Subsequently, the enzyme melts apart 13– 14 base pairs (bp)
of template DNA from 11 to + 2/ + 3 around the + 1 start
site of transcription, forming a catalytically competent open
complex (RPo). Kinetic investigations show that the closed
complex formation reaction is a rapid equilibrium governed
 L.M. Hsu / Biochimica et Biophysica Acta 1577 (2002) 191–207 193

Fig. 1. Kinetic scheme of transcription initiation. The process of transcription initiation is described in detail in the text. (A) The sequential model of abortive
and productive initiation. In this scheme, R represents RNA polymerase; P, promoter DNA; RPc, the closed complex; RPo, the open complex; and TEC,
ternary elongation complex. Numbers 2, 3, 4, 5, and 6 represent ITCs with a nascent RNA of the same length. Rate constants k1 and k 1 govern the forward
and reverse steps of closed complex formation; k2 and k 2, forward and reverse rate constants for open complex formation; kE, a composite rate constant of
promoter escape, measures the rate of productive initiation from an open complex. An arrow underneath each ITC is of different length to represent the
different rates of abortive release. Whether sigma (j) is released or not at the initiation – elongation transition is discussed in the text. (B) The branched pathway
model of abortive and productive initiation. Here, RPoVrepresents the unproductive ITCs that abortively initiate without undergoing escape.

by a binding constant KB (i.e. k1/k 1); the isomerization to
form the open complex can be described by forward and
reverse rate constants k2 and k 2. On several promoters,
kinetically significant intermediates within each major tran-
sition have been demonstrated (reviewed in Refs. [43,44]).
Here, we present a simplified view of open complex
formation, assuming these steps are not rate-limiting for a
promoter blocked at the escape stage.
On many promoters, the open complex is highly stable
and its formation is essentially irreversible (i.e. k 2 is
negligible). For these promoters, the value KB  k2 corre-
lates with the frequency of chain initiation, as measured by
formation of the first (few) phosphodiester bond(s) using
limited nucleoside triphosphates (NTP) [45,46]. For a subset
of promoters, however, this in vitro indicator of promoter
strength is inadequate for predicting the level of promoter-
specified gene product in vivo [47]. The discrepancy has
been attributed to differences in promoter escape at different
promoters [48].
In Fig. 1A, promoter escape is illustrated (for simplicity’s
sake) as a six-step process spanning chain initiation by the
open complex, to sigma release yielding an elongation
complex. Each ITC has three rate constants governing its
formation and disappearance, making the detailed kinetic
analysis of this phase complicated and difficult. To study the
kinetics of promoter escape, the rate of synthesis of pro-
ductive RNA from pre-formed open complexes is measured
to obtain a composite rate constant designated here as kE
[49,50]. Relatively short run-off templates are used such that
productive RNA synthesis is not complicated by down-
stream kinetic bottlenecks from the elongation phase.
In the kE measurement, even when the RNA chain
initiation reaction is synchronized by adding NTP to pre-
formed open complexes, the reaction soon yields a hetero-
geneous mixture of ITCs—due to their varied abortive
release and reinitiation rates. The product of this mixed
population of ITCs is a heterogeneous collection of abortive
RNAs which, when fractionated on high percentage poly-
acrylamide gels according to their size and nucleotide
composition, appear as a ladder of RNA bands (see Fig. 2).
2.4. Promoter escape, the event, may or may not involve
sigma release
To undergo promoter escape, an RNA polymerase mol-
ecule must relinquish its contacts to the core promoter
elements. This event is best documented by changes in the
DNase I footprinting analysis where the footprint of an open
complex (from 50 to + 20) shrinks to that of an elonga-
tion complex (from 5 to + 25); this change was thought
to coincide with the release of the sigma subunit [42,51,53].
On several promoter templates where sigma release has
been specifically examined, this transition occurs when the
nascent transcript reaches f 9 – 10 nt in length [51 – 54].
With the release of sigma factor, abortive initiation ceases,
the initiation phase is deemed complete, and the elongation
phase begins. Thus, achieving the elongation complex
structure defines the endpoint of promoter escape.
194 L.M. Hsu / Biochimica et Biophysica Acta 1577 (2002) 191–207

Fig. 2. Gel profile of abortive and productive initiation. Shown are the products of transcription from a set of T5 N25 promoter-random ITS mutants on a PCR-
generated linear template. These templates differ only in + 3 to + 19 of the transcribed region. Among the samples are the wild-type T5 N25 (lane 4) and the
N25antiDSR templates (lane 13) referred to in the text. Transcription was performed as described [67] except NTP is present at 100 AM, with [g-32P] ATP label.
The identity of abortive products for lanes 1 and 21 are indicated. The convention of designating the abortive RNA is by a letter and a number: the letter
corresponds to the 3Vterminal nucleotide of the RNA, the number indicates its size. Abbreviations: C, for minus enzyme control; RO, run-off RNA ( f 60 nt).

Recently, the criterion of sigma release accompanying
the initiation – elongation transition has become an issue
[55,56]. One study found that under conditions of solid-
state transcription, a fraction of RNA polymerase retains its
j70 subunit throughout elongation on several promoters
tested [55]. A second study, applying fluorescence reso-
nance energy transfer (FRET) analysis on ‘‘in gel’’ tran-
scribed complexes of lacUV5, found that 100% of the j70
subunit is retained on an enzyme positioned at + 15 [56].
(Earlier studies showed that, on lacUV5, sigma was released
when transcription had proceeded to + 11 [51].) In earlier
experiments, although sigma release was inferred rather than
directly measured, the results were consistent with alteration
and weakening in the association of sigma to core polymer-
ase, such that additional physical manipulation resulted in
release [51 – 53,57,58]. An experiment explicitly examining
the status of sigma using sucrose gradient centrifugation of
35
S-labeled P. putida RNA polymerase showed that, while
sigma was not released from open complexes, it was
released from complexes engaged in RNA synthesis [59].
Furthermore, binding of RNA polymerase to polyribonu-
cleotides or tRNA led to almost quantitative release of
sigma [59].
Given the ambiguous status of sigma release, I shall, in
subsequent sections of this review, refer to promoter escape
as an event culminating in sigma displacement—that is, the
displacement of sigma from its binding to the 10 and
35 promoter elements. After displacement, sigma may be
repositioned over another stretch of DNA or released,
depending on template sequence [60]. Sigma displacement
 L.M. Hsu / Biochimica et Biophysica Acta 1577 (2002) 191–207
can account for two observations affiliated with the initia-
tion – elongation transition: the shrunken DNase I footprints
of elongation complexes, and the cessation of repetitive
abortive initiation.
T7 RNA polymerase sets the precedence of utilizing the
displacement of a crucial protein domain contact to mark the
initiation – elongation transition. This single polypeptide
enzyme carries out all steps of the transcription cycle with-
out additional specificity subunits. Instead, the enzyme
utilizes a protein feature called the ‘‘specificity loop’’ to
contact promoter DNA during open complex formation.
Later, this loop contacts the nascent RNA during the
initiation –elongation transition [35].
3. Promoter escape can be a rate-limiting step in
transcription initiation
The goal of research on promoter-specified transcription
is twofold: one, to understand how a promoter achieves its
strength; and two, to understand its regulation. Both require
knowledge of the rate-limiting step for that promoter in the
context of the sequential scheme of transcription initiation
(Fig. 1A). Past research on E. coli Ej70 promoters has
identified the rate-limiting step for the vast majority of them
to be at either the KB or the k2 step [36,43]. Only recently
have a group of stringently controlled promoters been
shown to be rate-limited at the k 2 step; the open com-
plexes formed from these promoters exhibit extremely short
half-lives [61 – 64]. A few promoters were deemed to be
rate-limited at the promoter escape step; they are lacUV5
[49], T5 N25 [48,65], and malT [50,66]. These promoters
earned such a designation essentially by default; that is,
determination of association constants as well as open
complex half-life led to the conclusion that the slowest step
in transcription initiation occurs after open complex for-
mation.
Two features characterize promoters that are rate-limited
at the promoter escape stage: one, kE governs the slowest step
for these promoters; and two, they synthesize high levels of
abortive RNA relative to the productive RNA. The compo-
site constant kE measures the rate of productive initiation
under typical in vitro conditions of single-cycle transcription
from pre-formed open complexes. Accordingly, RNA poly-
merase was found to escape the lacUV5 promoter with a t1/2
of productive synthesis of f 1 min [49], T5 N25, f 5 – 10 s
[65; Hsu, unpublished results], and malT, f 10 min [50].
The knowledge of kE, however, does not explain the pattern
of abortive initiation displayed by these promoters. To gain
insight into the pattern, a quantitative abortive initiation
assay was devised [67]. This assay is performed under
multi-cycle, fixed-time conditions, for the reason that abor-
tive initiation is a reiterative process regardless of whether
the reaction condition imposes single- or multi-cycle tran-
scription. Under single-cycle conditions, the relative extent
of productive initiation would actually be underestimated.
195
3.1. Quantitative assessment of abortive initiation
One of the first questions raised regarding abortive
initiation is whether it reflects a substrate-binding limitation,
thus displaying an apparent high KS value for a certain NTP.
As illustrated in Fig. 1A, at each of the early template
positions, RNA polymerase can either catalyze the incorpo-
ration of the next template-specified nucleotide, or release
the nascent short RNA if it is held unstably. Presumably, the
relative rate of these two reactions—incorporation versus
release—determines the amount of abortive RNA released
at each template position.
To address this issue, titration studies of NTP concen-
tration were performed with T7 A1, T5 N25, and T5
N25antiDSR promoters. Interestingly, NTP concentrations as
high as 800 AM did not abolish abortive initiation [68,69,
and Hsu, unpublished results]. Instead, the level of abortive
and productive initiation both rise with NTP concentration,
due to increased frequency of chain initiation at higher
substrate concentrations. The relative yield of productive
to abortive RNA increases with NTP concentration to a
point, but then reaches a plateau; different promoters appear
to plateau at different NTP concentrations [68]. High KS
barriers are seldom found on natural promoters (i.e. T7 A1
and T5 N25), but may arise in a mutant promoter with an
altered initial transcribed sequence (ITS) (i.e. T5 N25antiDSR).
Regardless of the KS barriers, there is an intrinsic level of
abortive initiation that is not overcome even at high substrate
concentration.
The high degree of abortive initiation in vitro prompted
an investigation of the abortive and productive initiation
kinetics by pulse-labeling. This experimental approach
uncovered the existence of a ‘‘moribund’’ fraction of initial
complexes that are trapped in abortive transcription contin-
uously and cannot undergo promoter escape [70]. There also
exists a productive fraction of ITCs that can escape the
promoter to yield full-length RNA, but only after going
through the obligate sequence of abortive initiation steps
[68]. Thus, the abortive RNAs derived from a promoter are
made by both the productive and unproductive ITCs. How-
ever, the unproductive complexes contribute disproportion-
ately to the abortive yield and their existence can greatly
complicate the quantitative analysis of abortive initiation.
The partitioning of open complexes into productive and
unproductive ITCs during in vitro transcription is wide-
spread [68,71]. The fraction of unproductive ITCs varies
depending on the promoter; however, it can be reduced to a
minimum at higher NTP concentrations (i.e. 100 AM or
above) [68]. Depending on the promoter, the unproductive
ITCs may or may not become ‘‘moribund’’—that is, become
inactivated and undergo transcriptional arrest. The different
classes of unproductive ITC can be distinguished by their
differential susceptibility to Gre-factor mediated reactivation
of the unproductive complexes [68,71,72].
The above observations suggest the importance of opti-
mizing the concentration of NTP in transcription assays
196 L.M. Hsu / Biochimica et Biophysica Acta 1577 (2002) 191–207
designed to measure the abortive and productive activity of
a promoter. The existence of multiple conformations of
ITCs demands a revision of the sequential transcription
initiation scheme to include branched pathways (see Fig.
1B), where only the productive branch contributes to
productive RNA synthesis, while both branches contribute
to abortive RNA synthesis [68].
An interesting feature of abortive initiation is the
different abortive pattern displayed at each promoter [68;
also see Fig. 2]. Abortive pattern refers to the unique
collec-tion of abundant and scarce abortive products
obtained at each promoter. Although abortive release from
each ITC occurs stochastically, the abortive RNA pro-
file—qualitative and quantitative—obtained at each pro-
moter is reproducible. This led to the definition of a
quantitative para-meter called abortive probability—a cal-
culation of the probability an ITC positioned at a given
template nucleotide would release its RNA [67]. The
prominent abortive RNAs give rise to high abortive prob-
ability values, suggesting the ITCs at these template posi-
tions are highly unstable. The overall abortive pattern,
therefore, conveys a telltale description of the promoter
escape process at a given promoter, identifying the multi-
ple high-barrier steps, and suggesting where promoter es-
cape might be complete.
The maximum length of the abortive transcripts provides
another gauge of promoter escape. The abortive RNAs
forming a ladder are detected because of their high abun-
dance from reiterative cycles of initiation and release.
Where the abortive ladder ceases, a fundamental change in
the transcription complex has occurred: either it can no
longer reinitiate or the complex has become highly stabi-
lized. The maximum length of the abortive RNA was shown
to correlate with the position of sigma release on several
promoters [51,53,54]. On most Ej70 promoters with wild-
type ITS (roughly defined as positions + 1 to + 20), the
abortive ladders span 2– 12 nt [Hsu, unpublished observa-
tions]. However, abortive ladders longer than 12 nt have
been reported, usually from promoters with altered ITS, 14
nt for E PR AL [70] and 15 nt for T5 N25antiDSR [16],
suggesting that the ITS plays a role in the abortive initia-
tion – promoter escape program. The wild-type ITS of E PRV
prescribes a long abortive or paused RNA ladder of + 16/
+ 17. In this case, the ITS is crucial in mediating the loading
of the antiterminator protein, Q [41].
3.2. Factors that influence abortive initiation and promoter
escape
In the above, I have alluded to various factors that can
affect abortive initiation and promoter escape. These factors
can be grouped into intrinsic and extrinsic elements. Among
the intrinsic elements are the core promoter recognition
region (PRR), ITS, and the conformational state of the
template DNA. The extrinsic elements include activator
and repressor proteins, transcript cleavage stimulatory fac-
tors, as well as reaction solution conditions, such as ionic
strength, pH, etc.
Of the intrinsic elements, the PRR functions primarily to
recruit RNA polymerase through specific binding interac-
tions. RNA polymerase binds first at the 35 and 10
hexamers via sigma contacts, and secondarily via a-CTD
contacts either to an upstream (UP) DNA element [73] or
transcription factors [74] to set up an open complex with a
distinct level of RNA chain initiation capability. E. coli
promoters can differ in chain initiation frequency over three
to four orders of magnitude [46]. The ITS, first shown to
play a role in promoter clearance from the T5 N25 and T5
N25antiDSR promoters [47,48], does so by prescribing dis-
tinctly different escape programs for each [16]. Although
the PRR and the ITS appear to exert control over separate
aspects of the transcription initiation process, they do not
function independently [53,68]. For example, the down-
stream ITS sequence can affect the open complex stability,
start site selection and initiation frequency of a promoter
[75]. In turn, a promoter element can alter the abortive
probability at different downstream positions, lengthen or
shorten the maximum length of abortive transcripts [68].
A comprehensive analysis of the contribution of each
PRR element from 60 to 1, including the UP element,
the 35 hexamer, the spacer DNA, the 10 hexamer, and
the discriminator (sequence between the 10 hexamer and
the + 1 site) on promoter escape has been performed [68].
The results overwhelmingly support an inverse correlation:
starting with a consensus promoter, any PRR sequence
change that weakens the RNA polymerase – promoter inter-
action invariably facilitates promoter escape. These results
extend the previous observation with an UP element: its
presence on a consensus promoter increased promoter-
proximal stalling in vivo [76] and delayed promoter clear-
ance in vitro [77]. Thus, PRR interactions that enhance
RNA polymerase binding and subsequent melting isomer-
ization raise the barrier for promoter escape, presumably
because later these same interactions must be broken for the
enzyme to move away from the promoter region. Remark-
ably, Carpousis et al. [78] noted this dichotomy regarding
the dual functional roles of sigma during transcription
initiation as early as 1982.
That the RNA polymerase –promoter DNA interaction
can affect promoter escape predicates the existence of RNA
polymerase mutations with altered properties of abortive
initiation. A number of such mutations have been identified.
Some of the rif R mutations mapped to the h subunit were
found to be defective in promoter escape due to damaged
catalytic (site) activity [79 – 81]. Others were defective in
abortive initiation versus slippage synthesis because a high
KS requirement for UTP was created [82,83]. Several
mutations in j70 revealed more interesting phenotypes in
abortive initiation and promoter escape. RNA polymerase
holoenzyme with mutations in j70 region 3.1 gave rise to
reduced abortive transcription on T7 A1, T5 N25, T5
N25antiDSR, and E PR promoters [84,85]. On E PR, the
 L.M. Hsu / Biochimica et Biophysica Acta 1577 (2002) 191–207
reduction resulted from the lowered stability of open com-
plexes formed with the mutant holoenzyme [85]. In another
example, jR2.2 mutations gave rise to mutant holoenzymes
that transcribe T5 N25antiDSR promoter to yield a shorter
abortive RNA ladder [86]. Both jR2.2 and jR3.1 have been
implicated in the binding of sigma to core RNA polymerase
[85,86]. Thus, the abortive initiation properties, the abortive
yield and the maximum length of abortive RNA, can be
varied by changes in the sigma subunit, suggesting that how
sigma interacts with the rest of the polymerase can impact
the promoter escape process.
Vo [68] also examined the effect of mutating the
N25antiDSR ITS sequence placed downstream of a consensus
PRM promoter by replacing different sequence blocks within
+ 1 to + 20. A variety of sequence-dependent results were
obtained. Some changes affect the total rate of initiation by
the promoter, others alter the ratio of abortive to productive
initiation, and still others bring about a reduction in the
maximum length of abortive transcripts. Although the
results are complex, they show that the ITS plays an
important role in promoter escape.
An intriguing example of the effect of the ITS on
abortive initiation and promoter escape is shown in Fig. 2
[Hsu, unpublished results]. When the ITS of T5 N25
promoter was randomly mutagenized from + 3 to + 19,
the ‘‘mutant’’ promoters invariably underwent abortive
initiation to a greater distance on the template DNA—
generally to + 15/ + 16, compared to + 11 for the ‘‘wild-
type’’ T5 N25 sequence, although abortive initiation to
+ 19/ + 20 has also been observed—while giving a 30-fold
fluctuation in the productive yield (i.e. 150% to 5% of T5
N25). One can draw the following conclusions from this
analysis. One, the ITS can greatly influence the extent of
productive synthesis. Two, there is no correlation between
promoter escape efficiency and the length of the abortive
initiation program. Three, the natural ITS sequence of T5
N25 prescribes the shortest abortive initiation program
( + 11, see lane 4 in Fig. 2), suggesting this promoter has
evolved an ‘‘optimal’’ ITS to bring about the most efficient
promoter escape in a promoter that is rate-limited at this
stage. Ongoing investigations of this type will provide
further insight towards understanding the sequence-pre-
scribed idiosyncrasies of transcription initiation.
An intrinsic element not yet systematically examined for
its effect on promoter escape is the state of the DNA
template. To the extent that DNA in vivo is negatively
supercoiled, and supercoiling can bend or melt DNA,
position or strengthen polymerase contacts, DNA topology
is potentially an important determinant of promoter escape
[87]. For a promoter optimized at recruiting the polymerase,
rate limitation at the promoter escape stage manifests itself
in vivo as stalling of the transcription bubble at + 6 to + 12
[88]. Whether stalling in vivo is accompanied by the
production of abortive products is unclear.
Several extrinsic factors are now known to affect pro-
moter escape. Transcript cleavage stimulatory factors GreA
197
and GreB can prevent or rescue arrested complexes [4,5] or
correct misincorporation [89], by inducing polymerase
active center-mediated transcript cleavage [see Fish and
Kane, this volume]. That GreA and GreB act on short
RNA-bearing ITCs to stimulate promoter escape in vitro
and in vivo suggests these complexes may become arrested
prior to releasing the transcripts [15 – 17]. Another means by
which Gre factors can stimulate promoter escape is to
facilitate the conversion of the moribund complex into an
active escape-competent complex [72].
A number of transcriptional activator and repressor
proteins have been shown to exert their regulatory effect
at the promoter escape step. These include cAMP receptor
protein (CRP), Lac repressor (LacI), bacteriophage P22 Arc
protein, and protein p4 of B. subtilis phage A29. These
examples are discussed in detail in the next section.
3.3. Promoters regulated at the promoter escape step
Extensive investigation of bacterial promoter activity has
led to the view that promoters are individually optimized to
achieve in vivo strength consonant with their physiological
requirement [65,90]. Regulation of a promoter, whether by
activation or repression, is most effectively exerted at the
slowest step [91]. Several promoters known to be regulated
at the promoter escape step also fit this scheme.
Regulation during promoter escape is implemented in
various ways. For example, lacUV5 promoter is repressed
by the LacI protein. Unlike other repressors that regulate
transcription by occluding the binding of polymerase or
interfering with isomerization, LacI and RNA polymerase
can bind simultaneously at the promoter. When both pro-
teins are bound, only abortive RNA is synthesized [92]. The
lacUV5 promoter has multiple lac repressor binding sites;
of these, LacI binding at the operator sequence which
overlaps the initial transcribed region (ITR) created a high
KS block at the + 6/ + 7 junction, limiting abortive initiation
to + 6 and preventing escape which requires transcription
past + 9 [93].
The case of the malT promoter presents an interesting
paradox. This weak promoter is rate-limited at the kE step
and requires activation through the upstream binding of
CRP. In the absence of CRP, the malT promoter produces
mainly abortive RNA, but the addition of CRP greatly
stimulates productive RNA synthesis [50]. CRP binding,
however, was found not to affect the rate of isomerization in
open complex formation, nor change the rate of promoter
escape (i.e. kE). Rather, CRP was shown to facilitate
promoter escape by two means: one, by raising the affinity
for the nucleotide substrate UTP [50]; and two, by forming a
more stable open complex [66].
These conclusions seem incongruent with the expectation
that a regulatory molecule would act at the rate-limiting
step. However, with the knowledge of the existence of
productive and unproductive ITCs—the former facile at
promoter escape and the latter doomed to abortive cycling
198 L.M. Hsu / Biochimica et Biophysica Acta 1577 (2002) 191–207
[68,70]—an alternative explanation for CRP action can be
proposed. That is, there are two open complex conforma-
tions in the absence of CRP: the active one, a precursor of
the productive ITC; and the inactive one, a precursor to an
unproductive ITC [68,70]. CRP binding favors the equili-
brium distribution towards the active precursor open com-
plex. This interpretation is consistent with the recent finding
that CRP acts transiently at a diverse set of promoters to
stimulate the formation of productive open complexes [94].
Viewed in this light, CRP activates promoter escape by
favoring the ‘‘productive’’ branch of ITC formation. The
increased affinity for UTP may be an indirect consequence
of promoter escape from this productive ITC.
Another example of a regulatory protein acting at the
promoter escape stage was reported for the bacteriophage
P22 Arc protein which represses the Ej70 Pant promoter
during late lytic growth [95]. Normally, Arc binds as a
dimer at tandem operator sites (designated as 35 proximal
or 10 proximal) in the spacer region of the Pant promoter.
It was found that the binding of a cooperativity-defective
Arc-SL35 dimer at the 35 proximal site was sufficient to
bring about regulation. However, the regulatory conse-
quence was different depending on the nature of the variant
Pant promoter used; the non-consensus (NC) variant differs
from the consensus (C) variant in harboring a single AT
transversion at 8. Arc-SL35 binding decreased the rate of
open complex formation at both promoters, and caused
repression of the NC promoter but activation of the C
promoter. What is the basis of this paradox?
Interestingly, the answer resides with the rate-limiting
step involved. The non-consensus promoter is rate-limited at
the open complex formation step. Since Arc-SL35 binding
further depresses this rate, repression results. For the con-
sensus variant, the AT transversion created a consensus
promoter, greatly elevating its intrinsic rate of open complex
formation such that it was no longer the slowest step. As a
result, the C variant is now rate-limited at promoter escape.
Arc-SL35 binding to C decreases the rate of open complex
formation, but to a degree insufficient to cause a reversion in
the rate-limiting step; instead, it accelerates the rate of
promoter escape to yield the productive RNA. It is not clear
how Arc binding at the spacer region facilitates promoter
escape, but its activity serves to confirm the principle of
regulation at the rate-limiting step.
The regulatory scenario presented by the B. subtilis
phage f29 p4 protein is not unlike that of Arc [96]. Protein
p4 binds upstream of two promoters—late A3 and early
A2c, activating the former but repressing the latter. Activa-
tion or repression is not a consequence of the different
positioning of p4, resulting in different interactions with
RNA polymerase via a-CTD contacts. Rather, the mode of
regulation depends on the presence or absence of a 35
element in the target promoter. Protein p4 represses the
activity of the A2c promoter, which contains a 35
element, presumably by raising the overall polymerase –
DNA binding affinity above a threshold level such that the
added affinity keeping the polymerase at the promoter could
not be overcome to achieve escape. Repression in this case
was detected as increased abortive initiation [96]. Thus,
analogous to Arc, p4 binding may have altered the rate-
limiting step for the A2c promoter. The activity of p4 on the
A2c promoter confirms the dichotomy between tight pro-
moter binding and poor promoter escape; interactions favor-
ing one may be detrimental to the other.
A well-known example of in vivo regulation of tran-
scription exerted at the promoter escape stage involves
slippage synthesis. In vitro, slippage synthesis occurs in
response to template sequence with repeated nucleotide
triplets or longer. RNA polymerase ‘‘stutters’’ at the repeat
sequence to make complementary homopolymers, some of
extraordinary length. This reaction is particularly prominent
in the initial region and can affect gene expression by
competing with normal template-specified transcription
[97]. A group of pyrimidine biosynthetic operon promoters
in E. coli has evolved a mechanism to sense the intracellular
UTP level as a means of regulating promoter escape in vitro
and in vivo [98 – 101]. High UTP concentration favors
nonproductive slippage synthesis, while low UTP concen-
tration stimulates the ‘‘normal’’ process of abortive initiation
and promoter escape.
A final example of regulating gene expression at the
promoter escape stage can be found at E PRV promoter.
PRVensures the transcriptional expression of E late genes by
Q-mediated antitermination. Q protein is brought onto the
transcription complex paused at + 16, subsequently facili-
tating promoter escape and the conversion of the elongating
RNA polymerase into the anti-terminating conformation.
How Q facilitates promoter escape is not known [102]. All
of the examples presented here indicate that the cell has
taken full advantage of promoter escape for regulatory
options.
4. Abortive initiation and promoter escape in the context
of RNA polymerase structure
In recent years, the study of transcription has received a
huge boost from the resolution of X-ray crystal structures of
RNA polymerase. The breakthrough started with T7 RNA
polymerase and its various complexes [22 – 25], followed by
the multi-subunit RNA polymerases from Thermus aqua-
ticus and yeast [26,27]. With the bacterial polymerases, it
was possible to model the vast amount of functional data
obtained with E. coli Ej70 onto the Taq core enzyme
structure [103 – 105]. These modeling efforts are valid
because the multi-subunit polymerases all show a high
degree of sequence homology and structural and functional
conservation of the catalytic core of the proteins, suggesting
that they share a common catalytic mechanism [106 – 109].
It is pertinent to point out that structural information
obtained by modeling offers great advantages, but with a
caveat. This is illustrated by efforts to obtain the structure of
 L.M. Hsu / Biochimica et Biophysica Acta 1577 (2002) 191–207
elongation complexes. Having solved the Taq core enzyme
structure, Korzheva et al. [103] modeled the nucleic acid
component of an elongation intermediate based on cross-
linking data (containing 40 bp of DNA and 14 nt of nascent
RNA) onto the core structure and obtained the first picture
of an elongation complex. Later, a crystal structure of a
yeast RNA polymerase II (yeast Pol II) elongation complex
containing a 9-nt RNA made on a tailed template was solved
[28]. While the two structures show remarkable agreement,
the comparison highlights the omission by modeling of
protein conformational changes between the core enzyme
and the elongation complex. However, by modeling, the
path of the nucleic acid component was augmented, not only
confirming the existence of an 8-bp RNA –DNA hetero-
duplex in the elongation complex, but also identifying an
RNA exit channel for the nascent transcript.
Another advantage of the modeled bacterial elongation
complex structure came recently. Its coordinates were
merged with those of a Taq holoenzyme/fork-junction
DNA structure to achieve a first open complex structure
of Taq RNA polymerase [30].
T7 RNA polymerase is structurally distinct from that of
the multi-subunit enzymes reflecting RNA polymerase’s
divergent evolutionary history. However, by all evidence,
T7 RNA polymerase seems to have evolved a functional
mechanism highly similar to that of the multi-subunit
enzymes [110 – 112]. Thus, mechanistic insights gained
from the investigation of T7 RNA polymerase activity can
be applied towards solving a common mechanism of tran-
scription.
The present cast of RNA polymerase and transcription
complex structures shows us that RNA polymerase under-
goes a well-orchestrated sequence of conformational
changes at each step of transcription, brought about by
protein – protein and protein – nucleic acid interactions.
Abortive initiation – promoter escape is an activity of the
ITCs, the structures of which have been difficult to probe
due to their high degree of heterogeneity and instability
[3,42]. In the following section, I shall describe the struc-
tures of RNA polymerase molecules in conformations
bracketing the initiation – elongation transition to identify
the molecular features and strategies important for achieving
promoter escape. I will also consider the structures (crystal
and biochemical) of initiation complexes to elucidate the
relevant features of the intermediate states. The goal is to
achieve a descriptive mechanism of abortive initiation and
promoter escape.
4.1. Structures of open and elongation complexes
4.1.1. Core enzyme [26,27]
Both Taq and yeast Pol II core enzymes fold into a
‘‘crab-claw’’ structure, with h (Rpb2) and hV(Rpb1) sub-
units each contributing a pincer, forming a central channel
with a diameter of 27 Å that can accommodate double-
stranded nucleic acids. The channel on both sides is lined
199
by h and hVprotein mass, and the active site Mg2 + is
located centrally on the back wall. Behind the active site
Mg2 +, a secondary channel opens out to the rear surface of
the enzyme. Through this channel, NTP substrates can
reach the active site, and released short transcripts can
diffuse out. While RNA polymerase is overall a highly
acidic molecule, there is an uneven distribution of the
charged residues, with the basic residues congregating
along the active site channel.
4.1.2. Holoenzyme [29,104,105]
Holoenzyme is formed from the joining of sigma and
core enzyme. Taq jA is a member of the j70 family proteins
that show a high degree of sequence and functional homol-
ogy in four regions and their sub-regions [29,113 –115]. A
Taq jA molecule lacking sub-region 1.1 folds into three
independent domains (designated j2, j3, and j4) joined by a
flexible loop (eight residues between j2 and j3) and a linker
(33 residues between j3 and j4) (see Fig. 3). In promoter
DNA recognition, j2 mediates 10 element binding and
melting, j3, binding to the extended 10 nucleotides, and
j4, binding to the 35 element. Sigma binding to the core
enzyme leads to large conformational changes in domains of
h and hV(including the h flap and the hVclamp) that result in
closing of the active site channel to 15 Å, a distance too
narrow to accommodate the double-stranded DNA. To form
an open complex, it is proposed that the holoenzyme must
‘‘open’’ first to accept the DNA before closing again to yield
the open complex structure, with the opening and closing
mediated by protein conformational changes.
On the holoenzyme, the three sigma domains are splayed
out on the surface across the upstream opening of the active
site channel, j2 binding to the hVclamp and j3 binding to
the h2 domain across the channel (see Fig. 3). The j4
domain is anchored at the tip helix of h flap, separated by a
surface distance of 45 Å from j3. The connection between
these two domains is made through the interior of the
polymerase active site crevice by a long polypeptide linker
(jR3.2) that first descends from j3 towards the active site,
then turns and wanders underneath the h flap out toward j4.
Two surprising features of this linker are noted. One, the
path of the jR3.2 linker occupies the entire length of the
RNA exit channel through which a nascent transcript longer
than 8 nt must also traverse. Two, this polypeptide segment
is highly acidic, and therefore, binds to neutralize the basic
nature of the upstream portion of the active site channel. The
persistent path of the jR3.2 polypeptide led to the sugges-
tion that nascent RNA transcripts would encounter steric
hindrance during the initial steps of nucleotide incorpora-
tion, and therefore, undergo abortive release [29].
In the Taq holoenzyme structure, jR1.1 is absent, but its
orientation can be inferred from the direction of the poly-
peptide backbone, pointing it towards the downstream bind-
ing site (DBS) for double-stranded DNA. This disposition
of jR1.1 is confirmed by FRET analysis of the Ej70
holoenzyme and open complex—jR1.1 occupies the DBS
200 L.M. Hsu / Biochimica et Biophysica Acta 1577 (2002) 191–207

Fig. 3. Schematic representation of bacterial RNA polymerase and transcription complex structures. Color designations are as follows: beige, hVsubunit; blue,
h subunit; gold star, the catalytic center; red, j domains and connecting polypeptide segments; dark blue, DNA template strand; green, DNA non-template
strand; and purple, nascent RNA (with its 5Vend indicated by a dot). Horizontal hatches highlight the prominent protein features, the hVclamp and h flap.
Vertical hatches indicate the channels (or tunnels) located in the enzyme interior. Abbreviations: T, template; NT, non-template; and DBS, downstream (double-
stranded DNA) binding site. These drawings aim to convey the overall configuration of the protein and nucleic acid components in the transcription complexes.
Shown are the top-down views, with the top h subunit made partially transparent to reveal the components in the interior bound between h and hV. For clarity,
the a2N subunits are not shown. Top: The holoenzyme structure is drawn to represent the sigma subunit as a ‘‘molecular plug’’ (see text). The conformation of
the linker between jR1.1 and j2 domain is unknown and shown as dashed red line. Middle: As bound in the open complex, DNA is bent sharply twice: once at
the junction of downstream double-stranded DNA and the melted bubble, and again at the junction of the bubble and upstream double-stranded DNA. DNA
single strands of the transcription bubble are bound in separate tunnels (indicated by vertical hatches). The binding of downstream double-stranded DNA in the
DBS displaces jR1.1, which is removed from view. The jR3.2 linker occupies the RNA exit channel as in the holoenzyme. Bottom: The elongation complex is
shown without the j subunit, as in the structures solved. The nascent RNA is held as an RNA – DNA heteroduplex at the 3Vend, while the 5Vend has traversed
the RNA exit channel and emerged from the enzyme.

in the holoenzyme, but is displaced about 51 Å out of the
channel to interact with the h pincer domain. The large
conformational switch of jR1.1 is consistent with the
finding that it plays a role in open complex formation
[116,117]. Interestingly, jR1.1 is also highly acidic and,
when bound to the active site channel, can complement its
positive charges. Thus, in the holoenzyme, sigma can be
viewed as a ‘‘molecular plug,’’ filling the channel (with
jR1.1 and jR3.2) and ‘‘clasping’’ it closed at the top
through the j2 – j3 binding (see Fig. 3).
4.1.3. Open complex [30,104,105]
Holoenzyme, as described above, is perfectly configured
to bind to the upstream double-stranded promoter DNA
 L.M. Hsu / Biochimica et Biophysica Acta 1577 (2002) 191–207
through the surface exposed domains of sigma by major
groove contacts, forming a closed complex (RPC1; see Ref.
[44]). The downstream DNA is then ‘‘brought’’ into the
DBS channel, displacing jR1.1 and forming RPC2 [44]. As
bound, j2 is perfectly positioned to nucleate the opening of
the transcription bubble, from upstream at 12/ 11 to
downstream at + 3/ + 4 [30].
The formation of the open complex again involves
substantial conformational changes leading to a complete
closure of the active site channel. Within this enclosed
structure, the two strands of DNA in the transcription bubble
are buried in separate tunnels formed by a hVprotein mass
that pokes through the two strands to interact with the h
protein mass. Of the protein mass that separates the two
tunnels is the highly conserved structural feature called the
hV rudder; its position implicates a role in open complex
formation [118]. One can view the open complex as a
‘‘solid’’ protein unit with five interior tunnels—the DBS
tunnel, the RNA exit tunnel, the template strand (T) tunnel,
the non-template strand (NT) tunnel, and the secondary
tunnel through which NTP and small RNA diffuse to and
from the active site (see Fig. 3). The separately buried nature
of the template and non-template strands suggests DNA
must be threaded through the enzyme during the progression
of transcription.
4.1.4. Elongation complex [28,103]
The structures of the elongation complex reveal most
clearly the path of the DNA and RNA (strands) inside the
enzyme. DNA inside the enzyme protein is kinked sharply
(about 90j) near the catalytic site, where a chelated Mg2 +
ion is positioned between the i and i + 1 nucleotide substrate
binding sites. (The i and i + 1 sites correspond to transcribed
positions + 1 and + 2 in the open complex. In steps after the
first phosphodiester bond, the i site is occupied by the 3V-
OH nucleotide of the nascent transcript, while the i + 1 site
accommodates the successive incoming NTP.) Downstream
of the active site, the bubble is open for 2– 3 nt before
annealing to form double-stranded DNA securely held by
the DBS. Upstream of the catalytic center, the template and
non-template strands of the bubble rise up towards the edge
of the active site crevice. The RNA transcript is bound to the
template strand as an 8-bp RNA –DNA heteroduplex. In the
modeled elongation complex, RNA longer than 8 nt would
be peeled away from the template strand and directed into
the RNA exit channel (see Fig. 3).
4.2. Structural insights of ITCs
4.2.1. Initiated complex [24,25,119]
The only available structure of an initiated complex is
that of T7 RNA polymerase containing a 3-nt RNA base-
paired to the template DNA in the form of a perfect A-helix.
Comparison of this structure to that of a T7 RNA polymer-
ase open complex reveals that during translocation, template
DNA is ‘‘scrunched’’ into the enzyme interior volume.
201
Scrunching not only involves threading of the DNA strands,
but also the generation of a strain that manifests as distortion
in the DNA backbone. In the structure of the T7 initiated
complex, this strain is localized to the immediate upstream
(5V) edge of the RNA – DNA heteroduplex, and two nucleo-
tides are scrunched, corresponding to the two steps of
translocation performed to synthesize the 3-nt RNA.
Using fluorescence quenching to monitor the collapse of
the open complex bubble, Liu and Martin [112] found that
for T7 RNA polymerase, the upstream edge of the initial
open complex bubble does not rewind until the nascent
RNA has reached + 9, giving a maximum bubble size of 13
melted base pairs. This evidence also indicates that T7 RNA
polymerase can scrunch in as many as 8 bp of DNA during
the early steps of transcription without causing upstream
rewinding. Furthermore, the 5Vend of the nascent RNA does
not peel off the template strand until past + 10, suggesting
an RNA – DNA heteroduplex of at least 10 bp.
It is not clear how many basepairs of DNA can be
scrunched into the interior volume of E. coli RNA polymer-
ase, but if its catalytic mechanism is identical to that of T7
RNA polymerase, then at least 8 bp. There is a caveat in
extrapolating these measurements for the E. coli enzyme,
however. Owing to the different promoter architecture of E.
coli and T7 RNA polymerase and the much smaller size of
the T7 enzyme, the above dimensions should be regarded as
minimum estimates.
Structurally, a complex perched on the brink of the
initiation – elongation transition was obtained by FeBABE
footprinting of the E PRV + 16 complex [60]. In the open
complex, probes in j2 and j4 mapped to the 10 and 35
elements of the promoter DNA, respectively, as expected.
When the polymerase paused at + 16, j2 and j4 have been
repositioned to the + 1 to + 6 region on the non-template
DNA and the 25 region of the spacer DNA, respectively.
That the contacts of sigma domains to the promoter DNA
have been displaced classifies the + 16 complex to be a
paused elongation structure, perhaps one that has just
undergone the initiation – elongation transition.
A more revealing look was obtained with a probe in
jR3.2 located near the N terminus of the polypeptide
linker that occupies the RNA exit channel. In the open
complex, it mapped in the vicinity of the catalytic center.
Upon transcription initiation, this region moved in a
‘‘retrograde’’ fashion, eventually contacting the upstream
surface DNA in the + 16 complex. This ‘‘disjunctive’’
movement of the separate sigma domains was puzzling at
first, but now can be reconciled with the positions of the
sigma domains on the holoenzyme [29,30,60]. The data
suggest that between the open complex and the paused
elongation conformations, j2 and j4 stay anchored on the
surface of the enzyme, making contacts with the threaded
DNA that has translocated towards the upstream, while
jR3.2 has ‘‘moved’’ outward (from the enzyme interior)—
probably because of displacement by the advancing RNA,
converging onto the vicinity of j4. In the E PRV + 16
202 L.M. Hsu / Biochimica et Biophysica Acta 1577 (2002) 191–207
structure, the interaction of the sigma subunit with the core
enzyme has been altered.
4.3. Key issues affecting abortive initiation and promoter
escape
The discussion above has identified several factors crit-
ical for understanding abortive initiation and promoter
escape. In the open complex, the location of the jR3.2
linker appears to pose a formidable physical block for the
advancing RNA. Steric clash would be encountered by a
transcript 6 nt or longer, but maybe even as short as 2 nt,
leading to abortive initiation [29]. Given this barrier, can
transcription ever proceed past the abortive initiation stage?
In the elongation complex structure, an RNA – DNA hetero-
duplex of 8 bp is observed [28]. Mechanistically, what role
does this heteroduplex play during the early steps of tran-
scription to bring about the initiation – elongation transition?
Regarding the jR3.2 block, this barrier appears formi-
dable but is not insurmountable. On most E. coli promoters
under routine transcription conditions, some fraction of the
RNA polymerase can escape the promoter to form produc-
tive RNA, accompanied by varying levels of abortive
initiation. Thus, this linker can be displaced during tran-
scription, and the FeBABE footprinting experiment
described above offers supporting evidence [60]. It is a
challenge, therefore, to understand how the efficiency of
displacement of this linker might vary from promoter to
promoter.
The role of the RNA – DNA heteroduplex in a tran-
scription complex has been more controversial (reviewed
in Refs. [20,120]). Many groups have demonstrated the
existence of an 8– 9 bp RNA –DNA heteroduplex in elon-
gation complexes [12,121,122]. However, claims that the
RNA – DNA hybrid is the central stability determinant for an
elongation complex were soon faced with the observation
that, depending on the sequence context, certain 8-bp
hybrids actually stimulate the back-and-forth sliding of the
polymerase, resulting in transcriptional arrest [18,122,123].
The RNA – DNA heteroduplex observed in the yeast Pol
II crystal structure is in fact caught in a poorly stabilized
conformation. This heteroduplex adopts a nonstandard
underwound helix where only the 3V-most 2– 3 bp show a
high degree of order due to protein binding. The upstream 5
nt of the nascent RNA are held mainly through hydrogen
bonding interaction with the template DNA, devoid of any
nearby protein contacts. Gnatt et al. [28] conclude that the
yeast Pol II elongation complex is stabilized mainly by
protein – protein interactions, but reason that stabilization
must not interfere with the required functional mobility
necessary for repeated translocation during transcription.
The loosely held nature of the RNA – DNA heteroduplex
makes possible this mobility. In summary, the RNA – DNA
heteroduplex may contribute to the stability of an elongation
complex as one of many determinants, and then, in a
sequence-dependent manner.
The existence of an RNA – DNA heteroduplex in an early
elongation complex indicates it was formed during the
initiation reactions. For T7 RNA polymerase, promoter
escape occurs only when the heteroduplex has grown to
8 – 9 bp [110,112]. To account for the high level of abortive
initiation, it has been suggested that initiation reactions
maybe concerned with the synthesis of an RNA primer; as
the RNA chain length increases, the initial complexes
should gain stability and the abortive RNA yield should
decrease [42]. This proposal was soon questioned with the
observation that the length and/or GC content of an RNA
shows no correlation with its abortive potential [51,68,86,
and Hsu, unpublished results]. The question remains: if not
stability, then what role does the heteroduplex play in
bringing about promoter escape?
In the following section, I shall describe a mechanism to
account for abortive initiation and promoter escape by
making two assumptions. One, E. coli RNA polymerase
performs translocation by scrunching DNA into the enzyme
interior [119]. Two, the RNA – DNA heteroduplex serves a
role in guiding the development of strain in the ITC from
steps of scrunching – translocation. The growing length of
the RNA – DNA heteroduplex increasingly places the strain
locus closer to the upstream edge of the transcription
bubble, eventually causing the collapse of the upstream
edge of the transcription bubble to achieve promoter escape.
4.4. A descriptive mechanism of promoter escape
The following descriptive mechanism is aimed at
accounting for the distinct pattern of abortive initiation seen
at different promoters; an example of the widely different
patterns is shown in Fig. 2. Starting with an open complex,
the transcription bubble is buried inside the protein tunnels
and the DNA single strands emanating to the surface become
reannealed upstream and are held in place by j2, j3, and j4
domains that, in turn, are anchored to the polymerase. In
addition, the jR3.2 linker occupies the tunnel for the nascent
RNA (see Fig. 3). The transcription bubble spans 13 – 14 bp
(from 11 to + 2/ + 3) and represents the most stable (i.e.
relaxed) conformation at the initiation stage.
At de novo initiation, NTP is brought in from the
secondary channel, positioned in the i and i + 1 sites, and
phosphodiester bond formation occurs. A translocation
event must follow rapidly to reposition the next template
base before the catalysis of the next phosphodiester bond
can occur. Translocation involves flipping out the next
template base, scrunching 1 bp of template and non-tem-
plate DNA into the enzyme interior volume to position the
next template base at the new i + 1 site.
As scrunching takes place, the size of the transcription
bubble increases, and strain develops. This strain is man-
ifested as a distortion in the DNA backbone and is at first
localized to the immediate upstream border of the RNA –
DNA heteroduplex along the template strand [119]. The
strain elicits stress and the ITC now becomes a ‘‘stressed
 L.M. Hsu / Biochimica et Biophysica Acta 1577 (2002) 191–207
intermediate’’ [51]. (Here, I wish to draw an analogy
between a ‘‘stressed intermediate’’ with a transition-state
intermediate in an enzyme-catalyzed reaction. As a transi-
tion-state intermediate, the stressed intermediate would con-
tain more internal energy—scrunching-induced—and is
more unstable and more reactive. Thus, for each step of
nucleotide incorporation, the transcription complex goes
through a higher-energy transition state.) To achieve tempo-
rary stabilization, the localized strain and the energy asso-
ciated with it must be distributed and dissipated. This can
occur by propagating the distortion along the DNA backbone
towards either the upstream or downstream regions (DSRs)
of the template strand. Movement of the distortion upstream
should be favorable, as backbone deformation can be more
readily accommodated by single-stranded region of DNA.
Moving the strain locus upstream away from the RNA –
DNA heteroduplex should also stabilize the structural con-
formation at the catalytic site, enabling the incorporation of
the next nucleotide. By contrast, propagation of the strain
energy downstream would disrupt the RNA – DNA hetero-
duplex, leading to release of the short RNA. (In the above
scheme, the non-template strand must also accommodate the
scrunching-induced strain. However, without the constraint
of an RNA –DNA heteroduplex, the subsequent steps to
relieve the stress can take a random course.)
After the incorporation of the second nucleotide,
scrunching – translocation follows, and the strain genera-
tion-stress relief cycle repeats. Strain is developed and
localized to the upstream border of the heteroduplex, which
is now 3-bp long. Propagation of the strain energy along the
downstream DNA backbone disrupts the RNA – DNA het-
eroduplex, releasing the RNA. Propagation of the strain
energy towards the upstream stabilizes the new intermedi-
ate, leading to the incorporation of the next nucleotide (see
illustration in Fig. 4, panel A).
This scheme satisfactorily explains two features of abor-
tive initiation. One, abortive release at a given position is a
‘‘population’’ phenomenon—that is, some complexes
release their transcripts, while others can proceed to incor-
porate the next nucleotide. In this scheme, the fraction that
propagates the strain downstream gives rise to abortive
release; the fraction that distributes its strain energy
upstream facilitates incorporation. Two, at the next template
position, a new ‘‘equilibrium’’ of abortive and incorporating
complexes is set up, leading to a different level of abortive
release from that position. This hypothesis can explain how
the abortive probability at any position is distinct, shows no
correlation with the length of the RNA, but is dependent on
the sequence context of the position.
As the nascent transcript grows (for example, to + 6),
the longer RNA – DNA heteroduplex increasingly places
the scrunching strain locus closer to the mid section of the
transcription bubble. The longer length of the heteroduplex
segment should discourage the downstream propagation of
strain energy, since more of the heteroduplex interaction
must be disrupted. Consequently, incorporation of the next
203
nucleotide should be favored. However, an additional factor
comes into play after a few incorporation/translocation
cycles—that is, the scrunching strain from the earlier steps
has been accumulating, because the collapse of the
upstream edge of the transcription bubble does not occur
until at least + 9 [112]. The cumulative strain located
upstream of the heteroduplex segment can ‘‘backfire’’ by
favoring downstream propagation and now, with more
force, can disrupt the 6-bp heteroduplex. Cumulative strain,
therefore, can affect the equilibrium distribution of abortive
versus incorporating complexes at the early steps in an
unpredictable manner, favoring abortive release unexpect-
edly at position + 6 and beyond (see illustration in Fig. 4,
panel B).
Presumably, when RNA –DNA heteroduplex reaches 8–
9 bp in length, the cumulative strain energy becomes
sufficiently large to force the collapse of the transcription
bubble upstream. This leads to rewinding of the 10
promoter DNA, dislodging it from the original (open com-
plex) interactions with j2 and j3, allowing the DNA to
‘‘translocate’’ in the upstream direction and simultaneously
dislodging the 35 promoter DNA contacts to j4. (Here, I
wish to propose the term ‘‘outward translocation’’ to refer to
the upstream movement of the DNA away from the enzyme,
to distinguish it from the step-wise ‘‘inward translocation’’
that occurs after each nucleotide incorporation step.) This
first major outward translocation of the promoter DNA
would encompass a distance of 8 –9 bp, equal to the number
of nucleotides scrunched inside the enzyme interior, and is
reminiscent of an enzymatic step in the ‘‘discontinuous’’
movement model of transcription elongation [19]. Coinci-
dentally, the protein contacts to the 10 and 35 pro-
moter sequences have been displaced. This loss of
interaction is sufficient to give the reduced DNase I foot-
prints observed in complexes that have just undergone the
promoter escape transition [3,42]. Thus, upon the first major
outward translocation, the sigma – promoter DNA contacts
are lost, reiterative initiation should cease, and promoter
escape is effectively accomplished.
In a complex that has just come through the promoter
escape transition, the transcription bubble would be restored
to a stable size, an 8– 9 bp RNA –DNA heteroduplex would
be in place at the active site, and the RNA 5Vend would be
directed into the RNA exit channel. During the subsequent
steps of nucleotide incorporation, the scrunching strain
would be placed immediately near the mid to upper segment
of the transcription bubble, facilitating the distribution of
strain energy in the upstream direction to effect the rewind-
ing of the upstream edge of the bubble. If the strain
development – stress relief cycle becomes monotonic, that
is, one basepair of inward translocation followed by one
basepair of outward translocation, the transcription bubble
would be maintained at a constant size and the progress of
elongation would proceed at an even pace—one template
step at a time [124]. However, monotonic translocation
appears not to be the default mode of movement for the
204 L.M. Hsu / Biochimica et Biophysica Acta 1577 (2002) 191–207

Fig. 4. Schematic representation of the strain generation/stress relief cycle during translocation involving scrunching. (For further detail, see text.) Only the
nucleic acid component of an ITC is shown: blue, template strand; green, non-template strand; and purple, nascent RNA with a dot marking its 5Vend. The gold
star denotes the catalytic center which performs catalysis from below the template strand. During the abortive initiation-promoter escape phase, inward
translocation occurs without outward translocation; this movement of the DNA is gauged by the black and red rectangles, each marking a particular upstream
and downstream base pair, respectively. Scrunching generates strain shown as distortion in the DNA backbone. The strain is at first localized at the 5Vedge of
the RNA – DNA heteroduplex (structures a and d) and subsequently delocalized by propagating the backbone distortion upstream (indicated by ‘‘ f up’’) or
downstream (‘‘ f dn’’). Downstream propagation disrupts the RNA – DNA heteroduplex, leading to abortive release of the short RNA and rewinding of the
bubble downstream, re-establishing the open complex structure (see a – b and d – b transitions). Upstream propagation temporarily stabilizes the conformation
around the active site (structures c and e), enabling the incorporation of the next nucleotide (see c – d transition). In this scheme, upstream propagation of the
strain is reversible, while downstream propagation is irreversible. Panel A shows a strain generation/stress relief cycle undergone by an ITC bearing a 3-nt RNA
(ITC3; structure a). This ITC has scrunched in 2 bp of DNA. Panel B shows an analogous stress relief cycle in ITC6 that harbors cumulative strain from
scrunching in 5 bp of DNA (see structure f). When propagated upstream, the cumulative stress can eventually lead to the collapse of the bubble, leading to the
first major outward translocation (here shown as occurring at + 10; see black rectangle in structure h), yielding a ternary elongation complex (TEC10).
 L.M. Hsu / Biochimica et Biophysica Acta 1577 (2002) 191–207
polymerase during elongation. Many studies have observed
the asymmetric expansion and contraction in the size of the
transcription bubble and/or in the elongation complex foot-
print for complexes elongating over a contiguous stretch of
10 –20 bp, suggesting that RNA polymerase translocates in
a discontinuous fashion [121,125,126]. Specifically, the
expansion of the bubble and footprint is continuous in the
downstream boundary in sync with the growth of the RNA,
while the upstream boundary remains fixed until a ‘‘jump’’
occurs [125,126]. These observations support a transloca-
tion mechanism that involves many steps of inward trans-
location for each step of outward translocation.
Thus, the scrunching –translocation mechanism appears
to function throughout transcription. During transcription,
each complex must become a stressed intermediate for a
time, and individual complexes can differ in the amount
of internal stress depending on the cumulative strain each
has harbored. Stress relief, especially by distributing the
stress in the downstream direction, would bring about
different consequences for the initiation versus elongation
complexes. In the ITCs, downstream distribution of the
stress disrupts the short RNA – DNA heteroduplex, caus-
ing the RNA to be abortively released. During elongation,
an occasional complex might have accumulated a level of
strain that is insufficient to cause bubble rewinding up-
stream (to lead to outward translocation) but, when
propagated downstream, is sufficient to disrupt the 8-bp
RNA – DNA heteroduplex. In such a complex, the RNA
has grown too long to be released. Instead, the RNA is
caught in a partly released configuration, with its 3V tail
portion exuded into the secondary channel and an up-
stream stretch—recently retracted from the RNA exit
channel—now tied up in an RNA –DNA heteroduplex,
resulting in the formation of an arrested complex [10 –
13].
Note that up to the point of the first major outward
translocation, sigma release has not been accounted for,
although sigma displacement has occurred. At the first
major outward translocation step, it is unclear whether the
dislodging of promoter DNA – j2 interactions might also
loosen the j2 –core enzyme interaction. However, as the
nascent transcript becomes longer than 8 nt, it gains addi-
tional stabilization from entry and binding in the RNA exit
channel, displacing the jR3.2 linker. By 14 nt, the jR3.2
linker should be completely displaced. This would loosen
the binding of j4 to the flap tip helix located at the exit of
the RNA channel. With multiple sigma domain – core inter-
actions broken, sigma release might eventually occur [127].
The above description can account for promoter escape
occurring around the + 10 position. However, many pro-
moters undergo abortive initiation till a greater distance on
the template, for example, to + 16. Since reiterative initia-
tion requires the maintenance of the promoter DNA –sigma
domain contacts, by the above scheme, longer abortive
RNA can result only if scrunching continues past + 10 to
accumulate even more strain energy, leading to the first
205
major outward translocation at + 16. This begs the question
about how many bases of the melted DNA can be scrunched
into the E. coli RNA polymerase.
To summarize, a model for abortive initiation and pro-
moter escape suggests the formation of ‘‘stressed intermedi-
ates’’ during initial transcription [51]. Stress is induced by
the scrunching of DNA during inward translocation [25].
The length of the RNA –DNA heteroduplex serves to place
the strain locus to different segments of the transcription
bubble and can influence the subsequent stress relief steps.
Cumulative strain from many steps of scrunching is required
to cause the collapse of the open complex bubble and lead to
promoter escape.
5. Concluding remarks and future prospects
In this review, I have described a stepwise mechanism of
abortive initiation and promoter escape to account for the
unique pattern of abortive products seen at different pro-
moters. The scheme proposes that abortive release results
from conformational rearrangements occurring along the
template DNA strand of the transcription bubble. However,
the RNA polymerase protein clearly plays the central role in
catalysis; thus, the next major challenge is to elucidate the
polymerase – DNA interactions and the stepwise conforma-
tional readjustments in both the DNA and the enzyme
protein during transcription. The proposed scheme makes
several assumptions and borrows heavily the mechanistic
insights from T7 RNA polymerase. For example, a major
assumption is that translocation proceeds by scrunching.
This hypothesis has been tested for the T7 RNA polymerase
[128,129] but has yet to be examined for E. coli RNA
polymerase. Equally important for the E. coli enzyme is
information regarding the dimension of the transcription
bubble and the extent of cumulative scrunching during
initiation. The above scheme cannot account for sigma
release at the first major outward translocation step. The
issue of sigma release, whether it occurs at the promoter
escape transition, or if it occurs at all during the tran-
scription cycle, awaits clarification. Biochemical character-
ization has shown the widespread existence of the
unproductive ITCs during abortive initiation. These com-
plexes carry out abortive initiation continuously and are
prevented from undergoing promoter escape. With a struc-
ture-based mechanism of abortive initiation and promoter
escape, one can begin to uncover the structural basis for the
unproductive conformation. Finally, a major question is
concerned with how the promoter DNA sequence can set
up open complexes with different abortive versus productive
capabilities. In this context, the role of the ITS on promoter
escape is especially intriguing. For example, how is the
sequence information in this region transmitted to the
enzyme to yield ITCs of widely different stability? Is the
information relayed through the template or the non-tem-
plate strand? How might this sequence influence the posi-
206 L.M. Hsu / Biochimica et Biophysica Acta 1577 (2002) 191–207
tion of sigma displacement as reflected in the different
points where abortive initiation ceases? The wealth of
biochemical and biophysical knowledge now available
places us in a good position to solve the mechanism of
transcription.
Acknowledgements
LMH gratefully acknowledges the generous hospitality
and intellectual support of members of the joint C.M. Kane
and M.J. Chamberlin labs during her third sabbatical visit to
the University of California at Berkeley, and the funding
support by National Science Foundation (MCB-0077941)
under the Research in Undergraduate Institutions program.
She wishes to thank Seth Darst and Richard Ebright for
permissions to cite data prior to publication, Caroline Kane,
Nam Vo, and Rachel Fish for critical reading of the
manuscript, and Kenneth Howe for being a patient and
astute art teacher.
References
[1] C.K. Surratt, S.C. Milan, M.J. Chamberlin, Proc. Natl. Acad. Sci.
U. S. A. 88 (1991) 7983 – 7987.
[2] K.M. Arndt, M.J. Chamberlin, J. Mol. Biol. 213 (1990) 79 – 108.
[3] B. Krummel, M.J. Chamberlin, J. Mol. Biol. 225 (1992) 239 – 250.
[4] S. Borukhov, A. Polyakov, V. Nikiforov, A. Goldfarb, Proc. Natl.
Acad. Sci. U.S. A. 89 (1992) 8899 – 8902.
[5] S. Borukhov, V. Sagitov, A. Goldfarb, Cell 72 (1993) 459 – 466.
[6] M.G. Izban, D.S. Luse, Genes Dev. 6 (1992) 1342 – 1356.
[7] M.D. Rudd, M.G. Izban, D.S. Luse, Proc. Natl. Acad. Sci. U. S. A.
91 (1994) 8057 – 8061.
[8] M. Orlova, J. Newlands, A. Das, A. Goldfarb, S. Borukhov, Proc.
Natl. Acad. Sci. U. S. A. 92 (1995) 4596 – 4600.
[9] G.A. Kassavetis, E.P. Geiduschek, Science 259 (1993) 944 – 945.
[10] A. Mustaev, E. Zaychikov, M. Grachev, A. Goldfarb, J. Biol. Chem.
268 (1993) 19185 – 19187.
[11] T.C. Reeder, D.K. Hawley, Cell 87 (1996) 767 – 777.
[12] E. Nudler, A. Mustaev, E. Lukhtanov, A. Goldfarb, Cell 89 (1997)
33 – 41.
[13] N. Komissarova, M. Kashlev, Proc. Natl. Acad. Sci. U. S. A. 94
(1997) 1755 – 1760.
[14] M.G. Izban, D.S. Luse, J. Biol. Chem. 268 (1993) 12864 – 12873.
[15] G.H. Feng, D.N. Lee, D. Wang, C.L. Chan, R. Landick, J. Biol.
Chem. 269 (1994) 22282 – 22294.
[16] L.M. Hsu, N.V. Vo, M.J. Chamberlin, Proc. Natl. Acad. Sci. U. S. A.
92 (1995) 11588 – 11592.
[17] M.T. Marr, J.W. Roberts, Mol. Cell 6 (2000) 1275 – 1285.
[18] P.H. von Hippel, D.G. Bear, W.D. Morgan, J.A. McSwiggen, Ann.
Rev. Biochem. 53 (1984) 389 – 446.
[19] M.J. Chamberlin, Harvey Lect. 88 (1995) 1 – 21.
[20] S.M. Uptain, C.M. Kane, M.J. Chamberlin, Ann. Rev. Biochem. 66
(1997) 117 – 172.
[21] E. Nudler, J. Mol. Biol. 288 (1999) 1 – 12.
[22] R. Sousa, Y.J. Chung, J.P. Rose, B.C. Wang, Nature 364 (1993)
593 – 599.
[23] D. Jeruzalmi, T.A. Steitz, EMBO J. 17 (1998) 4101 – 4113.
[24] G.M. Cheetham, D. Jeruzalmi, T.A. Steitz, Nature 399 (1999) 80 – 83.
[25] G.M. Cheetham, T.A. Steitz, Science 286 (1999) 2305 – 2309.
[26] G. Zhang, E.A. Campbell, L. Minakin, C. Richter, K. Severinov,
S.A. Darst, Cell 98 (1999) 811 – 824.
[27] P. Cramer, D.A. Bushnell, R.D. Kornberg, Science 292 (2001)
1863 – 1876.
[28] A.L. Gnatt, P. Cramer, J. Fu, D.A. Bushnell, R.D. Kornberg, Science
292 (2001) 1876 – 1882.
[29] K.S. Murakami, S. Masuda, S.A. Darst, Science 296 (2002) 1280 –
1284.
[30] K.S. Murakami, S. Masuda, E.A. Campbell, O. Muzzin, S.A. Darst,
Science 296 (2002) 1285 – 1290.
[31] J. Gelles, R. Landick, Cell 93 (1998) 13 – 16.
[32] M.J. Chamberlin, L.M. Hsu, in: E.C.C. Lin, A.S. Lynch (Eds.),
Regulation of Gene Expression in Escherichia coli, R.G. Landes
Company, Austin, 1996, pp. 7 – 25.
[33] C.T. Martin, D.K. Muller, J.E. Coleman, Biochemistry 27 (1988)
3966 – 3974.
[34] W.T. McAllister, Nucleic Acids Mol. Biol. 11 (1997) 15 – 25.
[35] J. Huang, R. Sousa, J. Mol. Biol. 303 (2000) 347 – 358.
[36] W.R. McClure, Ann. Rev. Biochem. 54 (1985) 171 – 204.
[37] A.J. Carpousis, J.D. Gralla, Biochemistry 19 (1980) 3245 – 3253.
[38] D.E. Johnston, W.R. McClure, in: R. Losick, M. Chamberlin (Eds.),
RNA Polymerase, Cold Spring Harbor Laboratory, Cold Spring Har-
bor, 1976, pp. 413 – 428.
[39] W.R. McClure, C.L. Cech, D.E. Johnston, J. Biol. Chem. 253 (1978)
8941 – 8948.
[40] L.M. Munson, W.S. Reznikoff, Biochemistry 20 (1981) 2081 – 2085.
[41] B.Z. Ring, W.S. Yarnell, J.W. Roberts, Cell 86 (1996) 485 – 493.
[42] A.J. Carpousis, J.D. Gralla, J. Mol. Biol. 183 (1985) 165 – 177.
[43] M.T. Record Jr., W.S. Reznikoff, M.L. Craig, K.L. McQuade, P.J.
Schlax, in: F.C. Neidhardt, R. Curtiss III, J.L. Ingraham, E.C.C. Lin,
H.E. Umbarger (Eds.), Escherichia coli and Salmonella typhimu-
rium: Cellular and Molecular Biology, 2nd ed., ASM Press, Wash-
ington, DC, 1996, pp. 792 – 821.
[44] P.L. DeHaseth, M.L. Zupancic, M.T. Record Jr., J. Bacteriol. 180
(1998) 3019 – 3025.
[45] W.R. McClure, Proc. Natl. Acad. Sci. U. S. A. 77 (1979) 5634 – 5638.
[46] M.E. Mulligan, D.K. Hawley, R. Entriken, W.R. McClure, Nucleic
Acids Res. 12 (1984) 789 – 800.
[47] U. Deuschle, W. Kammerer, R. Gentz, H. Bujard, EMBO J. 5 (1986)
2987 – 2994.
[48] W. Kammerer, U. Deuschle, R. Gentz, H. Bujard, EMBO J. 5 (1986)
2995 – 3000.
[49] J.E. Stefano, J. Gralla, Biochemistry 18 (1979) 1063 – 1067.
[50] M. Menendez, A. Kolb, H. Buc, EMBO J. 6 (1987) 4227 – 4234.
[51] D.C. Straney, D.M. Crothers, J. Mol. Biol. 193 (1987) 267 – 278.
[52] U.M. Hansen, W.R. McClure, J. Biol. Chem. 255 (1980)
9564 – 9570.
[53] B. Krummel, M.J. Chamberlin, Biochemistry 28 (1989) 7829 – 7842.
[54] W. Metzger, P. Schickor, T. Meier, W. Werel, H. Heumann, J. Mol.
Biol. 232 (1993) 789 – 800.
[55] G. Bar-Nahum, E. Nudler, Cell 106 (2001) 443 – 451.
[56] J. Mukhopadhyay, A.N. Kapanidis, V. Mekler, E. Kortkhonjia, Y.W.
Ebright, R.H. Ebright, Cell 106 (2001) 453 – 463.
[57] A.A. Travers, R.R. Burgess, Nature 222 (1969) 537 – 540.
[58] C.W. Wu, L.R. Yarbrough, Z. Hillel, F.Y. Wu, Proc. Natl. Acad. Sci.
U. S. A. 72 (1975) 3019 – 3023.
[59] G.F. Gerard, J.C. Johnson, J.A. Boezi, Biochemistry 11 (1972)
989 – 997.
[60] M.T. Marr, S.A. Datwyler, C.F. Meares, J.W. Roberts, Proc. Natl.
Acad. Sci. U. S. A. 98 (2001) 8972 – 8978.
[61] Y.N. Zhou, D.J. Jin, Proc. Natl. Acad. Sci. U. S. A. 95 (1998)
2908 – 2913.
[62] I.K. Pemberton, G. Muskhelishvili, A.A. Travers, M. Buckle, J. Mol.
Biol. 299 (2000) 859 – 864.
[63] M.M. Barker, T. Gaal, C.A. Josaitis, R.L. Gourse, J. Mol. Biol. 305
(2001) 673 – 688.
[64] J.E. Cabrera, D.J. Jin, J. Bacteriol. 183 (2001) 6126 – 6134.
 L.M. Hsu / Biochimica et Biophysica Acta 1577 (2002) 191–207
[65] R. Knaus, H. Bujard, in: F. Eckstein, D.M. Lilley (Eds.), Nucleic
Acids Mol. Biol., vol. 4, Springer-Verlag, Heidelberg, Germany,
1990, pp. 100 – 122.
[66] P. Eichenberger, S. Dethiollaz, H. Buc, J. Geiselmann, Proc. Natl.
Acad. Sci. U. S. A. 94 (1997) 9022 – 9027.
[67] L.M. Hsu, Methods Enzymol. 273 (1996) 59 – 71.
[68] N.V. Vo, PhD thesis, University of California, Berkeley (1998).
[69] J.D. Gralla, A.J. Carpousis, J.E. Stefano, Biochemistry 19 (1980)
5864 – 5869.
[70] T. Kubori, N. Shimamoto, J. Mol. Biol. 256 (1996) 449 – 457.
[71] M. Susa, R. Sen, N. Shimamoto, J. Biol. Chem. (2002), Online
manuscript #M112481200.
[72] R. Sen, H. Nagai, N. Shimamoto, Genes Cells 6 (2001) 389 – 401.
[73] S.E. Aiyar, R.L. Gourse, W. Ross, Proc. Natl. Acad. Sci. U. S. A. 95
(1998) 14652 – 14657.
[74] A. Ishihama, J. Bacteriol. 175 (1993) 2483 – 2489.
[75] M.L. Craig, O.V. Tsodikov, K.L. McQuade, P.E.J. Schlax, M.W.
Capp, R.M. Saecker, M.T. Record Jr., J. Mol. Biol. 283 (1998)
741 – 756.
[76] T. Ellinger, D. Behnke, R. Knaus, H. Bujard, J.D. Gralla, J. Mol.
Biol. 239 (1994) 466 – 475.
[77] M.G. Strainic Jr., J.J. Sullivan, A. Velevis, P.L. deHaseth, Biochem-
istry 37 (1998) 18074 – 18080.
[78] A.J. Carpousis, J.E. Stefano, J.D. Gralla, J. Mol. Biol. 157 (1982)
619 – 633.
[79] M. Kashlev, J. Lee, K. Zalenskaya, V. Nikiforov, A. Goldfarb, Sci-
ence 248 (1990) 1006 – 1009.
[80] J. Lee, M. Kashlev, S. Borukhov, A. Goldfarb, Proc. Natl. Acad. Sci.
U. S. A. 88 (1991) 6018 – 6022.
[81] V. Sagitov, V. Nikiforov, A. Goldfarb, J. Biol. Chem. 268 (1993)
2195 – 2202.
[82] D.J. Jin, J. Biol. Chem. 271 (1996) 11659 – 11667.
[83] D.J. Jin, C.L. Turnbough Jr., J. Mol. Biol. 236 (1994) 72 – 80.
[84] V.J. Hernandez, L.M. Hsu, M. Cashel, J. Biol. Chem. 271 (1996)
18775 – 18779.
[85] R. Sen, H. Nagai, V.J. Hernandez, N. Shimamoto, J. Biol. Chem. 273
(1998) 9872 – 9877.
[86] C.L. Chan, C.A. Gross, J. Biol. Chem. 276 (2001) 38201 – 38209.
[87] X. Dai, L.B. Rothman-Denes, Curr. Opin. Microbiol. 2 (1999)
126 – 130.
[88] T. Ellinger, D. Behnke, H. Bujard, J.D. Gralla, J. Mol. Biol. 239
(1994) 455 – 465.
[89] D.A. Erie, O. Hajiseyedjavadi, M.C. Young, P.H. von Hippel, Sci-
ence 262 (1993) 867 – 873.
[90] F. Rojo, J. Bacteriol. 181 (1999) 2987 – 2991.
[91] M. Lanzer, H. Bujard, Proc. Natl. Acad. Sci. U. S. A. 85 (1988)
8973 – 8977.
[92] S.B. Straney, D.M. Crothers, Cell 51 (1987) 699 – 707.
[93] J. Lee, A. Goldfarb, Cell 66 (1991) 793 – 798.
[94] H. Tagami, H. Aiba, EMBO J. 17 (1998) 1759 – 1767.
[95] T.L. Smith, R.T. Sauer, Proc. Natl. Acad. Sci. U. S. A. 93 (1996)
8868 – 8872.
[96] M. Monsalve, B. Calles, M. Mencia, M. Salas, F. Rojo, Mol. Cell 1
(1997) 99 – 107.
[97] J.-P. Jacques, D. Kolakofsky, Genes Dev. 5 (1991) 707 – 713.
[98] C. Liu, L.S. Heath, C.L. Turnbough Jr., Genes Dev. 8 (1994)
2904 – 2912.
207
[99] A.H. Tu, C.L. Turnbough Jr., J. Bacteriol. 179 (1997) 6665 – 6673.
[100] X. Han, C.L. Turnbough Jr., J. Bacteriol. 180 (1998) 705 – 713.
[101] Y. Cheng, S.M. Dylla, C.L. Turnbough Jr., J. Bacteriol. 183 (2001)
221 – 228.
[102] W.S. Yarnell, J.W. Roberts, Cell 69 (1992) 1181 – 1189.
[103] N. Korzheva, A. Mustaev, M. Kozlov, A. Malhotra, V. Nikiforov, A.
Goldfarb, S.A. Darst, Science 289 (2000) 619 – 625.
[104] N. Naryshkin, A. Revyakin, Y. Kim, V. Mekler, R.H. Ebright, Cell
101 (2000) 601 – 611.
[105] V. Mekler, E. Kortkhonjia, J. Mukhopadhyay, J. Knight, A. Revya-
kin, A.N. Kapanidis, W. Niu, Y.W. Ebright, R. Levy, R.H. Ebright,
Cell 108 (2002) 599 – 614.
[106] L.A. Allison, M. Moyle, M. Shales, C.J. Ingles, Cell 42 (1985)
599 – 610.
[107] D. Sweetser, M. Nonet, R.A. Young, Proc. Natl. Acad. Sci. U. S. A.
84 (1987) 1192 – 1196.
[108] L. Minakhin, S. Bhagat, A. Brunning, E.A. Campbell, S.A. Darst,
R.H. Ebright, K. Severinov, Proc. Natl. Acad. Sci. U. S. A. 98 (2001)
892 – 897.
[109] S.A. Darst, Curr. Opin. Struck. Biol. 11 (2001) 155 – 162.
[110] D. Temiakov, P.E. Mentesana, K. Ma, A. Mustaev, S. Borukhov, W.T.
McAllister, Proc. Natl. Acad. Sci. U. S. A. 97 (2000) 14109 – 14114.
[111] C. Place, J. Oddos, H. Buc, W.T. McAllister, M. Buckle, Biochem-
istry 38 (1999) 4948 – 4957.
[112] C. Liu, C.T. Martin, J. Biol. Chem. 277 (2002) 2725 – 2731.
[113] J.D. Helmann, M.J. Chamberlin, Ann. Rev. Biochem. 57 (1988)
839 – 872.
[114] M. Lonetto, M. Gribskov, C.A. Gross, J. Bacteriol. 174 (1992)
3843 – 3849.
[115] J.A. Bown, J.T. Owens, C.F. Meares, N. Fugita, A. Ishihama, S.J.
Busby, S.D. Minchin, J. Biol. Chem. 274 (1999) 2263 – 2270.
[116] C. Wilson, A.J. Dombroski, J. Mol. Biol. 267 (1997) 60 – 74.
[117] S. Vuthoori, C.W. Bowers, A. McCracken, A.J. Dombroski, D.M.
Hinton, J. Mol. Biol. 309 (2001) 561 – 572.
[118] K. Kuznedelov, N. Korzheva, A. Mustaev, K. Severinov, EMBO J.
21 (2002) 1369 – 1378.
[119] G.M. Cheetham, T.A. Steitz, Curr. Opin. Struck. Biol. 10 (2000)
117 – 123.
[120] C.L. Chan, R. Landick, in: R.C. Conaway, J.W. Conaway (Eds.),
Transcription: Mechanism and Regulation, Raven Press, New York,
1994, pp. 297 – 321.
[121] D.N. Lee, R. Landick, J. Mol. Biol. 228 (1992) 759 – 777.
[122] M.L. Kireeva, N. Komissarova, D.S. Waugh, M. Kashlev, J. Biol.
Chem. 275 (2000) 6530 – 6536.
[123] N. Komissarova, M. Kashlev, J. Biol. Chem. 272 (1997) 15329 –
15338.
[124] E. Nudler, K. Kashlev, V. Nikiforov, A. Goldfarb, Cell 81 (1995)
351 – 357.
[125] E. Zaychikov, L. Denissova, H. Heumann, Proc. Natl. Acad. Sci.
U. S. A. 92 (1995) 1739 – 1743.
[126] S.M. Uptain, PhD thesis, University of California, Berkeley (1997).
[127] M.M. Sharp, C.L. Chan, C.Z. Lu, M.T. Marr, S. Nechaev, E.W.
Merritt, K. Severinov, J.W. Roberts, C.A. Gross, Genes Dev. 13
(1999) 3015 – 3026.
[128] M. Jiang, M. Rong, C. Martin, W.T. McAllister, J. Mol. Biol. 310
(2001) 509 – 522.
[129] L.G. Brieba, R. Sousa, EMBO J. 20 (2001) 6826 – 6835.