Two transcription pause elements underlie aσ70-dependent pause cycleEric J. Strobel and Jeffrey W. Roberts1

Department of Molecular Biology and Genetics, Cornell University, Ithaca, NY 14853

Contributed by Jeffrey W. Roberts, July 2, 2015 (sent for review June 5, 2015; reviewed by Richard H. Ebright and Robert Landick)

The movement of RNA polymerase (RNAP) during transcriptionelongation is modulated by DNA-encoded elements that cause theelongation complex to pause. One of the best-characterized pausesequences is a binding site for the σ70 initiation factor that inducespausing at a site near lambdoid phage late-gene promoters. Anessential component of this σ70-dependent pause is the elementalpause site (EPS), a sequence that itself induces transcription paus-ing throughout the Escherichia coli genome and underlies othercomplex regulatory pause elements, such as the ops and his op-eron pauses. Here, we identify and provide a detailed kinetic anal-ysis of a transcription cycle analogous to abortive cycling thatunderlies the σ70-dependent pause. We show that, in σ70-depen-dent pausing, the elemental pause acts primarily to modulate therate at which complexes attempt to disengage the σ70:DNA inter-action. Our findings establish the σ70-dependent pause-encodingregion as a multipartite element in which several pause-inducingcomponents make distinct mechanistic contributions to the induc-tion and maintenance of a regulatory transcription pause.

transcription pausing | σ-dependent pausing | elemental pause |factor-dependent pausing | transcription

Transcription pausing is an important element of gene regu-lation at the level of RNA synthesis in both prokaryotes and

eukaryotes, serving to coordinate the appearance of RNA withits utilization in cellular functions, and to modulate the interactionof regulatory proteins with RNA polymerase (RNAP) (1, 2). Themechanisms by which nucleic acid sequence elements and tran-scription factors induce transcriptionpauses are varied (3–6)but areunified in that theymust retain the transcription elongation complex(TEC) at a defined position on the DNA template for a period. Anotable example of regulatory transcription pausing exists in theEscherichia coli lambdoidphage late-geneoperon,whereexpressionof the controlled genes depends upon a σ70-dependent pause thatallows engagement of the antiterminator Q to RNAP (7); the ter-minator-resistant TEC can then “read through” an intrinsic ter-minator that precedes the coding sequence of the operon.The transcription pause that facilitates assembly of Q into com-

plex with RNAP is dependent on binding of the σ70 initiation factorto a repeat of the −10 promoter element, named the “−10-likesequence” (Fig. 1) (3). The anchoring of RNAP to the DNA site isfollowed by further RNA synthesis, as downstream DNA is un-wound and drawn into RNAP to form a “scrunched” DNA struc-ture that is exactly analogous to the scrunched initial transcribingcomplex from the promoter (Fig. 1) (8–11). Consistent with itsstructural similarities to an initial transcribing complex, the σ70-dependent pause comprises a transcription cycle analogous toabortive cycling (10, 12). From the stable pause (at +16 for λpR′),continuation one position further results in either escape down-stream and release of σ70, or backtracking that is rescued by Gre-mediated cleavage into a second cycle of synthesis and pausing (Fig.1) (10, 13). In fact, backtracking is a central component of the pausebecause it retains RNAP in the pause region.A second pause-inducing sequence, located immediately

downstream of the −10-like sequence, is essential to the σ70-dependent pause (Fig. 1) (14). Although this segment was firstdescribed as a backtrack-inducing sequence, a property it does

display, it primarily acts as an elemental pause site (EPS) (15–19).The EPS is a pause-inducing sequence that was identified throughglobal analysis of RNAP distribution within the E. coli genome, andwhich functions by halting the nucleotide addition cycle so thatfurther elongation cannot occur (17, 18). The EPS may include, orlead to, further conformational rearrangements that contribute tothe pause, especially an opening of the clamp that encloses thenucleic acids (20–22).The most important elements of the EPS are a GC-rich segment

at the upstream edge of the RNA:DNA hybrid and a pyrimidine atthe pause site followed immediately by a G. These elements interactwith various structures of the enzyme, especially a “core recognitionelement” that is proposed to counteract pausing by favoring trans-location (18). The existence of two discrete pause-inducing se-quences within the σ70-dependent pause-encoding region raises thequestion of how each element contributes to σ70-dependentpause kinetics.We have characterized the function of the elemental pause in

σ70-dependent pausing by relating the effects of a −10-like se-quence consensus mutation and a series of EPS mutations to σ70-dependent pause cycle kinetics. We have analyzed three EPSdeterminants as follows: (i) the pause site nucleotide; (ii) theGC-rich upstream edge of the RNA:DNA hybrid; and (iii)the internal segment of the RNA:DNA hybrid. We find thatthe major function of the EPS in the context of σ70-dependentpausing is to slow the pause cycle by limiting the rate at whichTECs transition from +16 to +17 and thus attempt to disengagethe −10-like sequence. In our model, the −10-like sequence isresponsible for inducing the initial capture of RNAP, whereas theEPS determines the frequency at which an escape attempt can bemade. Both the strength of σ70 binding to the −10-like sequenceand the EPS influence the probability of pause escape. Our studyof the λpR′ elemental pause has led to novel mechanistic insightsinto the nature of σ70-dependent pausing as a transcription pause

Significance

During transcription, RNA polymerase encounters DNA-encodedsignals that interrupt RNA synthesis and cause the transcriptionelongation complex to pause. An important function of transcrip-tion pausing in gene regulation is the halting of transcription tofacilitate assembly of regulatory factors into complex with RNApolymerase. An example of this function is found in the bacterio-phage λ late-gene transcription, in which the initiation factor σ70

induces a pause that mediates recruitment of the antiterminationfactor Qλ, thereby allowing the expression of genes downstreamof an intrinsic terminator. In this study, we characterize the con-tributions of two pause signals to σ70-dependent pause kineticsand propose a comprehensive model for σ70-dependent pausing.

Author contributions: E.J.S. and J.W.R. designed research; E.J.S. performed research; E.J.S.and J.W.R. analyzed data; and E.J.S. and J.W.R. wrote the paper.

Reviewers: R.H.E., Rutgers University; and R.L., University of Wisconsin–Madison.

The authors declare no conflict of interest.1To whom correspondence should be addressed. Email: jwr7@cornell.edu.

This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.1073/pnas.1512986112/-/DCSupplemental.

www.pnas.org/cgi/doi/10.1073/pnas.1512986112 PNAS Early Edition | 1 of 7

BIOCH

EMISTR

YPN

ASPL

US

that uses, first, a cycling effect analogous to that of transcriptioninitiation and, second, a universal pause sequence, to establish along-lived regulatory pause.

ResultsAnalysis of the σ70-Dependent Pause Cycle. To analyze σ70-dependentpausing in vitro, we performed single-round transcription reactionsin the presence and absence of the transcript cleavage factor GreB.Inclusion of GreB approximates the in vivo pause cycle becauseTECs that have backtracked upon failing to disengage the −10-likesequence can reactivate transcription by transcript cleavage and makesubsequent escape attempts. Omission of GreB reveals the kineticproperties of individual steps in the pause cycle by showing thechange in distribution of paused, backtracked, and escaped com-plexes in a single cycle. The measurements we make across these twoconditions provide a detailed understanding of σ70-dependent pausecycle kinetics and lead to a comprehensive model of σ70-dependentpausing. We note that, as is usual in similar experiments, thekinetic rates we obtain are not purported to be identical tothose in vivo, primarily because substrate nucleoside triphosphateconcentrations are arbitrary; relative values are, however, mean-ingful for understanding the pause mechanism.In the text below, the term “σ70-dependent pause” refers to the

σ70-dependent pause cycle as a whole, from the initial binding of σ70

at the −10-like sequence to the release of RNAP into elongationdownstream; it is characterized by the terms PrInduction (pause in-duction, the percentage of complexes that have detectably engagedin the σ70-dependent pause cycle) and t1/2[σP] (overall pause half-life). The term “+16 pause” refers to the transit from +16 to +17within the λpR′ σ70-dependent pause cycle; it is characterized by thetermsCE[16P] (capture efficiency, the percentage of complexes thatenter an elemental paused state at +16) and t1/2[16P] (+16 pause

half-life). PrEscape is the percentage of complexes that escapethe σ70:DNA interaction in a single cycle.

The −10-Like Sequence Modulates σ70-Dependent Pause Kinetics byInfluencing the Probability of Pause Escape. We previously used aλpR′ consensus −10-like sequence mutant (TATAAT substitutedfor AACGAT) to study the dynamics of σ70-dependent pause es-cape in the presence of the Qλ antiterminator (Table S1) (11).Below, we use the same consensus variant to examine the kineticconsequences of strengthening the interaction between σ70 andthe −10-like sequence. However, the consensus substitution has asecond useful property: it lacks a putative elemental pause at +12that delays transcription from the open complex to +16. Evidencefor this pause is a prominent abortive product at +12 from the WTtemplate (Figs. S1, S2A, and S3A) (23). The +12 pause very likelycauses asynchrony in further elongation; thus, kinetic analysis withthe consensus template benefits from higher synchrony.The σ70-dependent pause cycle is evident in the pause kinetics at

λpR′ compared with pause kinetics at ConsλpR′. In the presence ofGreB, pause induction (PrInduction) is 100% for both templates,whereas overall pause half-life (t1/2[σP]) at λpR′ (53 s) is increasedat ConsλpR′ (181 s) (Fig. 2). In the absence of GreB, the probabilityof escape (PrEscape) is reduced by ∼3.5-fold at ConsλpR′, a changethat drives the increase in overall pause half-life by increasing thenumber of cycles necessary for −10-like sequence-bound TECs tobreak the σ70:DNA interaction (Fig. 2C). The +16 pause half-life(t1/2[16P]) at λpR′ (47 s) appears reduced at ConsλpR′ (28 s);however, we suggest that this is due to an artificial increase in +16pause half-life at λpR′ resulting from the asynchrony describedabove. We argue below that the actual λpR′ +16 pause half-lifeshould be close to that of ConsλpR′; in fact, it is implausible thatthe −10-like sequence should affect the rate of movement of theTEC from +16 to +17.The increase in overall pause half-life at ConsλpR′ relative to

λpR′ in the presence of GreB shows the importance of escapeprobability in σ70-dependent pause kinetics and exposes the

Fig. 1. Architecture of σ70-dependent paused TECs. As the TEC reaches +13, σ70-dependent pausing is induced through contacts between the nontemplate strand−10-like and discriminator sequences by σ70 regions 2 and 1.2, respectively. The+16 paused, scrunched complex is formed when the −10-like sequence-boundTEC continues to synthesize RNA by DNA scrunching. The TEC encounters anEPS at +16 and can enter into a stable, paused state. Upon synthesis to +17,if RNAP fails to break the σ70:DNA interaction, the TEC backtracks into anarrested state in which σ70 remains bound to the −10-like sequence and theRNA 3′-OH is displaced downstream from the active center.

A

C

B

Fig. 2. Strengthening the σ70:DNA interaction increases σ70-dependentpause cycling. (A and B) Single-round in vitro transcription experimentsassessing σ70-dependent pause kinetics at λpR′ (A) and ConsλpR′ (B) in thepresence and absence of GreB. Scatter plots show the decay of total σ70-dependent paused TECs (+GreB) and +16 paused TECs (−GreB) over time.Dashed trend lines are generated using the average of two replicates, thevalues of which are shown as points. ConsλpR′ plots are composed of dataalso shown in Figs. 3A and 4A (+GreB) and Figs. 3B and 4B (−GreB), and arederived from the experiments of Figs. S2A and S4A. (C) Measured values fromA and B. PrInduction and t1/2[σP] are total σ70-dependent pause induction andhalf-life, respectively. CE[16P] and t1/2[16P] are +16 elemental pause captureefficiency and half-life, respectively. PrEscape is the probability of −10-like se-quence disengagement following a single escape attempt.

2 of 7 | www.pnas.org/cgi/doi/10.1073/pnas.1512986112 Strobel and Roberts

pause cycle clearly. Furthermore, the increased synchrony ofinitiation at ConsλpR′ provides us with a system in which tostudy σ70-dependent pause cycle kinetics without removing thepause from its natural promoter context.

The Pause Site Nucleotide Modulates σ70-Dependent Pause Kinetics.Consistent with the proposal that downstream portions of theσ70-dependent pause constitute an EPS, both the elementalpause and the σ70-dependent pause are favored by U or C at thepause site and are disfavored by A or G (16, 19). The pause sitenucleotide (+16 at λpR′) should not influence the initial σ70-binding event because, at the time, the RNAP active center ispositioned several nucleotides upstream of the eventual pausesite (Fig. 1). However, the terminal nucleotide could affect thekinetic properties of the +16 elemental pause.In the presence of GreB, pause induction (PrInduction) is 100%

for all +16 nucleotide variants of ConsλpR′ (Table S1), whereasthe overall pause half-life (t1/2[σP]) is reduced when the native +16C(181 s) is mutated to G (43 s) or A (30 s), but is almost unchangedwhen mutated to T (171 s) (Fig. 3 A and E, and Fig. S2A). Thisindicates that the +16 nucleotide does not contribute to pauseinduction but does affect the rate of −10-like sequence disen-

gagement. In the absence of GreB, +16 pause capture efficiency(CE[16P]) at ConsλpR′ is reduced when the native +16C (82%) ismutated to G (57%) or A (39%), but is almost unchanged whenmutated to T (79%) (Fig. 3 B and E, and Fig. S2A). The +16 pausehalf-life (t1/2[16P]) at ConsλpR′ is longer for the native +16C (28 s)and the T (30 s) mutant than for the G (20 s) and A (22 s) mutants(Fig. 3 B and E). Furthermore, the probability of escape (PrEscape)is increased for G (20%) and A (17%) mutants relative to thenative +16C (12%) and the T (10%) mutant (Fig. 3E). The con-sistency of mutant effects in the presence and absence of GreBsuggests a direct link between the kinetic properties of the +16elemental pause and the rate at which TECs traverse the σ70-dependent pause cycle. We interpret the short overall pause half-life at Cons16G and Cons16A to result from an increase in thefrequency at which −10-like sequence-bound TECs attempt tobreak the σ70:DNA interaction and an increased probability ofsuccess on each attempt.

The Upstream RNA:DNA Hybrid Edge Modulates σ70-Dependent PauseKinetics. Early studies of the λpR′ σ70-dependent pause identifiedtwo positions immediately downstream of the −10-like sequencethat, when mutated from G to A, reduce the level of pausingthrough both the nontemplate and template DNA strands (24).The nontemplate strand effect of this sequence is the result of aninteraction with σ70 region 1.2 that supplements the binding of σ70to the −10-like sequence (Fig. 1) (25, 26). The template strandelement is part of an EPS at the 82pR′ and λpR′ σ70-dependentpauses that also determines the site of the pause (Fig. 1) (19).Here, we consider the mechanism through which the GC-richupstream RNA:DNA hybrid edge influences σ70-dependent pausecycle kinetics.To characterize the contribution of the upstream RNA:DNA

hybrid edge to σ70-dependent pausing, we made a ConsλpR′ het-eroduplex mutant, named ConsUE, that contains C-to-T mutationsat positions +7 and +8 in the template strand, but maintains theWT sequence in the nontemplate strand to preserve σ70 region 1.2binding (Table S1). To facilitate labeling of the RNA, the experi-ments of Fig. 3 C and D were performed using 4× lower [UTP] and4× higher [GTP] than those of Fig. 3 A and B; because G is the nextnucleotide incorporated following the +16 pause, the overall pausehalf-life at ConsλpR′ is reduced from 181 to 77 s, as has been ob-served for other transcription pauses (27). This does not affect thecomparisons made.In the presence of GreB, pause induction (PrInduction) and overall

pause half-life (t1/2[σP]) at ConsλpR′ (100% and 77 s) are reducedat ConsUE (54% and 40 s) (Fig. 3 C and E, and Fig. S2B). Becausethe ConsUE mutations do not interfere with σ70:DNA bind-ing, decreased pause induction likely reflects complexes thatbind the −10-like sequence but disengage rapidly so as not to bedetectably paused. In the absence of GreB, +16 pause capture ef-ficiency (CE[16P]) at ConsλpR′ (63%) was reduced at ConsUE(10%), whereas +16 pause half-life (t1/2[16P]) was comparable (16and 18 s, respectively) (Fig. 3 D and E, and Fig. S2B). The prob-ability of escape (PrEscape) was increased at ConsUE (41%) relativeto ConsλpR′ (9%) (Fig. 3E). Notably, in contrast to most complexeswe describe, ConsUE TECs at +17 appear to chase by 300 s, sug-gesting that these complexes exist in equilibrium between a GreB-sensitive backtracked state and transcription-competent post-translocated state (Fig. S2B). Because this effect is specific to +17complexes (TECs from +18 to +24 appear stably arrested), it isunlikely that the intrinsic cleavage activity of RNAP contributesmeaningfully to the decay at +17. Our findings are consistent withthe proposal that the elemental pause is stabilized by a GC-richupstream RNA:DNA hybrid edge (17, 18). An AT-rich upstreamhybrid edge increases the rate at which TECs traverse the σ70-dependent pause cycle by promoting unimpeded transit from+16 to +17. Furthermore, the favorability of restoring strongbase pairs at the upstream hybrid edge promotes backtracking

A C

B

E

D

Fig. 3. The pause site nucleotide and upstream RNA:DNA hybrid edge modulateσ70-dependent pause kinetics. (A and B) Single-round in vitro transcription exper-iments assessing the role of the pause site nucleotide (+16) in σ70-dependent pausekinetics (A) and +16 pause kinetics (B). Dashed trend lines are generated using theaverage of two replicates, the values of which are shown as points. ConsλpR′ plotsare composed of data also shown in Figs. 2B (±GreB), 4A (+GreB), and 4B (−GreB),and are derived from the experiments of Figs. S2A and S4A. (C andD) Single-roundin vitro transcription experiments assessing the role of the GC-rich upstream RNA:DNA hybrid edge (template strand positions +7 and +8) in σ70-dependent pausekinetics (C) and +16 pause kinetics (D). Dashed trend lines are generated using theaverage of two replicates, the values of which are shown as points. (E) Measuredvalues from A–D. PrInduction and t1/2[σP] are total σ70-dependent pause inductionand half-life, respectively. CE[16P] and t1/2[16P] are +16 elemental pause captureefficiency and half-life, respectively. PrEscape is the probability of −10-like se-quence disengagement following a single escape attempt.

Strobel and Roberts PNAS Early Edition | 3 of 7

BIOCH

EMISTR

YPN

ASPL

US

at +17, as in WT, whereas weak base pairs at these positionsincrease the probability of escape by reducing the propensityfor TECs to backtrack.

The λpR′ +10 to +15 Region Modulates σ70-Dependent Pause Kinetics.The composition of the RNA:DNA hybrid has been shown to in-fluence the rate of translocation by RNAP and the strength of fac-tor-independent transcription pausing (6). Previous work indicatedthat the internal sequence of the EPS downstream of the G/C richelement is an important determinant of σ70-dependent pausing at82pR′ (14), but did not distinguish the separate role of the terminalnucleotide. Here, we replaced the λpR′ native sequence from +10to +15, 5′-TTAATT-3′, with the sequence 5′-CCGGCC-3′ to createthe mutant λpR′IE[CG] (Table S1). Consistent with previousresults, σ70-dependent pausing at λpR′IE[CG] was dramaticallyreduced both in the presence and absence of GreB (Fig. S3A).To further characterize the RNA:DNA hybrid internal segment,

we made a −10-like sequence consensus variant of λpR′IE[CG](Table S1). In the presence of GreB, pause induction (PrInduction)is 100% for both ConsλpR′ and ConsIE[CG], whereas overall pausehalf-life (t1/2[σP]) at ConsλpR′ (181 s) is reduced at ConsIE[CG]

(39 s) (Fig. 4 A and E, and Fig. S4A). In the absence of GreB,+16 pause capture efficiency (CE[16P]) at ConsλpR′ (82%) isreduced at ConsIE[CG] (23%), whereas +16 pause half-life(t1/2[16P]) is modestly reduced (28 and 23 s, respectively) (Fig. 4B and E, and Fig. S4A). The probability of escape (PrEscape) isincreased at ConsIE[CG] (24%) relative to ConsλpR′ (12%)(Fig. 4E). As for ConsUE, TECs that have moved to +17 atConsIE[CG] are GreB sensitive while also chasing by 300 s,suggesting that these can isomerize between the backtrackedand posttranslocated states (Fig. S4A). This may result from theGC-rich RNA:DNA hybrid stabilizing the forward position of theactive center relative to the backtracked complex. Our results in-dicate that the internal RNA:DNA hybrid is a determinant of ele-mental pausing and further emphasize the influence of the +16pause on overall σ70-dependent pause kinetics.

A Detailed Mutant Scan of the λpR′ +10 to +15 Region. To identifyelements within the GC-rich segment of λpR′IE[CG] that interferewith σ70-dependent pausing, we performed a scan comprising 3-, 2-,and 1-bp mutations in which λpR′IE[CG] sequence was restored toWT (Table S1 and Fig. S3 B and C). These templates are named inthe format nxp, where n refers to the number of bases restored andp refers to the position of the first restored base. The mutants withthe greatest increase in pausing for each category (3x12, 2x13, and1x14) are clustered within a 3-bp stretch of the +10 to +15 segment,from +12 to +14 (Fig. S3B and Table S1). To determine whether the+12 to +14 segment is sufficient to produce the full effect of λpR′IE[CG], we mutated λpR′ such that positions +12 to +14 are replacedwith the corresponding λpR′IE[CG] sequence (Table S1; 12GGC).Consistent with the observation that 2- and 3-bp substitutions out-side the +12 to +14 segment can also significantly affect pausing atλpR′IE[CG], the 12GGC mutation reduced pausing but failed toreproduce the full effect of λpR′IE[CG] (Fig. S3B).We then tested consensus variants of 3x12, 2x13, and 1x14

to determine whether changes in the kinetic properties of the+16 pause account for improved overall σ70-dependent pausing(Table S1). In the presence of GreB, pause induction (PrInduction)is 100% for all mutants, and overall pause half-life (t1/2[σP]) atCons3x12 (347 s), Cons2x13 (136 s), and Cons1x14 (126 s), isincreased relative to ConsIE[CG] (39 s) (Fig. 4 C and E, and Fig.S4B). In fact, the Cons3x12 overall pause half-life of 347 s isnearly twice that of ConsλpR′. By measurements in the absenceof GreB, we find that the overall pause half-life for Cons3x12,Cons2x13, and Cons1x14 correlates with increased +16 pausecapture efficiency (CE[16P]; 85%, 68%, and 86%, respectively),decreased probability of escape (PrEscape; 8%, 9%, and 13%,respectively), and, for Cons3x12, increased +16 pause half-life(t1/2[16P]; 36 s) (Fig. 4 D and E, and Fig. S4B).The restoring scan yielded a second class of four mutants (1x11,

1x12, 2x11, and 3x10) that appear to backtrack at positions up-stream (+14, +15) of the native pause (+16) as shown by the per-sistence of arrested TECs to at least 300 s in the absence of GreB(Fig. S3C). These mutants are related in that they restore positions+11 and +12 to WT λpR′ sequence and may result from the for-mation of a sequence that promotes the rapid entry of −10-likesequence-bound complexes into a backtracked state.

Analysis of σ70-Dependent Pause Kinetics. In Fig. 5, we plot mea-sured σ70-dependent pause half-lives against predicted σ70-dependent pause half-lives that were calculated by applying the+16 pause measurements t1/2[16P], CE[16P], and PrEscape to thefollowing equation:

kσPause =ðk+16PauseÞ

�PrEscape

�

Pr+16Capture. [1]

In Eq. 1, we assume that backtracking, transcript cleavage, andresynthesis to +16 are instantaneous relative to the elemental

A C

B

E

D

Fig. 4. The RNA:DNA hybrid internal segment modulates σ70-dependentpause kinetics. (A and B) Single-round in vitro transcription experiments as-sessing the role of the RNA:DNA hybrid internal segment (+10 to +15) in σ70-dependent pause kinetics (A) and +16 pause kinetics (B). Dashed trend lines aregenerated using the average of two replicates, the values of which are shownas points. ConsλpR′ plots are composed of data also shown in Figs. 2B (±GreB),3A (+GreB), and 3B (−GreB), and are derived from the experiments of Figs. S2Aand S4A. (C and D) Single-round in vitro transcription experiments assessing σ70-dependent pause kinetics (C) and +16 pause kinetics (D) for restoring scanmutants. Dashed trend lines are generated using the average of two replicates,the values of which are shown as points. (E) Measured values from A–D.PrInduction and t1/2[σP] are total σ70-dependent pause induction and half-life,respectively. CE[16P] and t1/2[16P] are +16 elemental pause capture efficiencyand half-life, respectively. PrEscape is the probability of −10-like sequence dis-engagement following a single escape attempt.

4 of 7 | www.pnas.org/cgi/doi/10.1073/pnas.1512986112 Strobel and Roberts

pause half-life. Therefore, kσPause should be dependent on theproperties of the +16 elemental pause (k+16Pause and Pr+16Capture)and the probability that a TEC will disengage the −10-likesequence (PrEscape). Eq. 1 is a good predictor of the measuredσ70-dependent pause half-life for 9 of the 10 templates tested(Fig. 5; black circles). The one exception, ConsUE (Fig. 5; redX), displayed a decreased tendency to backtrack, indicating thatthe ConsUE pause cycle may also depend on the rate of back-tracking and transcript cleavage. (This might be true as well forConsIE[CG], although to a lesser extent.) Nonetheless, the strongcorrelation between measured and calculated overall σ70-dependentpause half-lives suggests that the +16 elemental pause and the prob-ability of −10-like sequence disengagement are the primary determ-inants of pause cycle kinetics.Consistent with our proposal that asynchrony of initiation ob-

scured the +16 pause half-life at λpR′, the application of λpR′ −

GreB measurements to Eq. 1 produces a calculated overallhalf-life of 100 s, nearly twice the measured value (53 s). Becausethe strength of the interaction between σ70 and the −10-like se-quence should not interfere with +16 pause kinetics, the +16elemental pause should behave identically at both λpR′ andConsλpR′ and only the probability of escape should differ.Therefore, we should be able to use the PrEscape measurementfrom λpR′ and the Pr+16Capture and k+16Pause measurementsfrom ConsλpR′ to approximate t1/2[σP] at λpR′. This calcula-tion produces a predicted λpR′ t1/2[σP] of 57 s, which ap-proximates the measured t1/2[σP] of 53 s and therefore furthervalidates Eq. 1 as a description of σ70-dependent pause ki-netics. Furthermore, if the correct t1/2[16P] for WT is actually∼28 s, whereas t1/2[σP] is 53 s, Gre-mediated cleavage andcycling contribute importantly to the strength of the WT pause.

DiscussionWe have characterized the kinetics of a transcription cycle thatunderlies the σ70-dependent pause of the lambdoid bacteriophagelate-gene regulatory region. The pause depends upon σ70:DNAbinding and a widespread transcription pause sequence, theEPS, which stabilizes the TEC in an off-pathway paused andscrunched state. The elemental pause prevents the TEC fromadvancing to the next nucleotide or beyond, which wouldcreate an unstable state leading to either disengagement of σ70and escape from the pause, or backtracking to relieve thescrunched conformation. In effect, the elemental pause limitsthe rate at which TECs attempt to disengage the σ70:DNAinteraction that anchors RNAP to the −10-like sequence; fur-thermore, the elements of the EPS favor backtracking and arrestover pause escape. We provide a detailed kinetic model of thetransit of RNAP through the pause site, in which the overallpause characteristics are explained by the rates and probabilitiesof individual steps.In Fig. 6, we show a refined model of the pause at λpR′.

Transcription through the σ70-dependent pause can be dividedinto three stages: induction, cycling, and escape. Induction andescape are defined by the engagement and disengagement, re-spectively, of σ70 with the −10-like sequence. The cyclic nature ofσ70-dependent pausing results from Gre-mediated transcriptcleavage allowing TECs that have failed to escape the σ70:DNA

Fig. 5. Plot of measured vs. calculated t1/2[σP] values. Measured t1/2[σP]values taken in the presence of GreB from Figs. 3 A and C, and 4 A and C, areplotted against calculated t1/2[σP] values that were obtained by applying themeasurements taken in the absence of GreB from Figs. 3 B, D, and E, and 4 B,D, and E, to Eq. 1. The trend line equation and R2 value are shown. Pointsplotted as black circles are included in the trend line. The ConsUE value isshown as a red X and is not included in the trend line. For reasons describedin the text, WT λpR′ is not plotted.

Fig. 6. The λpR′ σ70-dependent pause cycle. Pause induction occurs when, at +13, σ70 binds the −10-like sequence and anchors RNAP to the DNA template.The −10-like sequence-bound TEC continues RNA synthesis to +16, at which point RNAP either proceeds rapidly to a posttranslocated conformation or entersan elemental paused state induced by the EPS. Upon entering a posttranslocated conformation at +16, RNA synthesis continues to +17 and RNAP eitherdisengages the −10-like sequence and escapes the pause or fails to disengage the −10-like sequence and enters a backtracked conformation. The backtrackedcomplex is then rescued by Gre-mediated transcript cleavage and so that subsequent escape attempts can be made until σ70 releases the −10-like sequence.

Strobel and Roberts PNAS Early Edition | 5 of 7

BIOCH

EMISTR

YPN

ASPL

US

bond and consequently have backtracked, to repeat the pausecycle until successful escape. The overall pause time thus de-pends upon (i) the rate at which TECs traverse the cycle, and(ii) the probability of successful escape.The rate at which TECs traverse the λpR′ σ70-dependent pause

cycle is limited by the +16 elemental pause. Correspondingly,EPS mutations that reduce +16 pause capture and/or half-life,including those of Cons16G, Cons16A, ConsUE, and ConsIE[CG], allow TECs to proceed from +16 to +17 rapidly, resultingin a short overall σ70-dependent pause half-life when the presenceof GreB permits the entire pause cycle to occur. The probabilitythat σ70 will disengage the −10-like sequence on a given escapeattempt determines the number of cycles necessary for all com-plexes to escape the pause. Because the σ70:DNA interaction isthe barrier to escape, a stronger σ70:DNA interaction reduces theprobability of pause escape, as shown directly by substitution ofthe −10 element consensus for the natural −10-like sequence. Asecond element that should determine the probability of σ70:DNA disengagement is the energy provided by the competingreaction, backtracking, which relieves unwinding energy withoutloss of σ70. Therefore, an important function of the GC-richRNA:DNA hybrid upstream edge is to favor backtracking. Mu-tation of this sequence in ConsUE increases the probability ofescape, likely by disfavoring backtracking. Consistent with thismodel, translocation to +17 breaks two G:C base pairs at theupstream RNA:DNA hybrid edge that stabilize the elementalpause (Fig. 6). We suggest that, as the TEC advances from +16 to+17, it reaches an EPS-induced “tipping point” from whichRNAP either escapes the σ70:DNA bond or backtracks and ar-rests. This model is consistent with the observation that insertingforeign segments between the −10-like sequence and the EPSmoves the site of σ70-dependent pausing downstream, implyingthat the extent of scrunching does not define the tipping point;instead, it is the stabilization provided by the EPS (19).We have identified an internal segment of the λpR′ EPS that

contributes importantly to the pause, consisting mostly of thesequence AAU 3–5 nt from the pause end. Original reports ofthe EPS found a preference for U at −3 (15, 17), and an analysis ofbases that oppose translocation (6) also showed preference for U orA at this site. Changing the AAU sequence and surrounding nu-cleotides to a G/C-rich segment strongly inhibited capture of TECsin an elemental pause. It is noteworthy that a similar sequencesimilarly located promotes abortive initiation (11).We can conclude that the σ70-dependent pause depends entirely

on both the σ70:DNA interaction and the elongation effects of theEPS. Rendering the −10-like sequence inactive by mutating theessential +2A and +6T nucleotides completely removes pausing at+16 (28). Similarly, our kinetic model for the σ70-dependent pausecycle predicts that removal of the slow elemental pause step wouldcause even a consensus −10-like sequence σ70-dependent pause todecay on a millisecond timescale and be undetectable. Thus, thefunctions of the −10-like sequence and the EPS are completelyinterdependent in σ70-dependent pausing.The role of the EPS in σ70-dependent pausing offers some insight

into the antipausing activity of the antiterminator Q. Incorporationof Q into a σ70-dependent paused TEC effects rapid escape fromthe σ70-dependent pause (29). In the absence of GreB, Q causes the+16 pause to decay rapidly; however, Q-dependent arrested com-plexes appear both from +17 to +19 and further downstream in acluster focused at +25 (11). Notably, these complexes all backtrackto around +13 to +15, indicating that σ70 remains bound to the −10-like sequence. Therefore, because Q does not appear to disrupt theσ70:DNA interaction [and in fact, could supplement it throughcontacts with the DNA-encoded Q-binding element (23, 30)], theantipausing activity of Q likely functions by modifying the propertiesof the +16 elemental pause such that the overall rate of pause cy-cling, and thus the frequency of escape attempts, is increased.

Materials and MethodsPlasmids. The source of template for transcription and mutagenesis waspM650 (31). Plasmids for protein purification were pET-28a-σ70 (32), pES3(for GreB) (11), and pVS10 (for RNAP; a gift from I. Artsimovitch, The OhioState University, Columbus, OH) (33). All mutants were constructed usingQuikChange site-directed mutagenesis.

Proteins. RNAP (33), GreB (34), σ70 (32), and NusA (35) were purified asdescribed.

In Vitro Transcription. Reaction mixtures containing 5 nM template and 25 nMRNAP (25 nM core reconstituted with 50 nM σ70) were incubated in tran-scription buffer (20 mM Tris·HCl, pH 8.0, 0.1 mM EDTA, 1 mM DTT, and50 mM KCl), 0.1 mg/mL BSA, a nucleotide mix of either 200 μM ATP, UTP,CTP, and 50 μM GTP containing 0.5 μCi/μL [α-32P]GTP (Figs. 2, 3 A and B, and4, and Figs. S2A, S3, and S4) or 200 μM ATP, GTP, CTP, and 50 μM UTP con-taining 0.5 μCi/μL [α-32P]UTP (Fig. 3 C and D, and Fig. S2B), and 150 nM NusA for10 min at 37 °C to form open complexes. Single-round transcription reactionswere initiated by addition of magnesium chloride to 5 mM and rifam-picin to 10 μg/mL. Total reaction volume was 25 μL. When present, GreB wasadded to 200 nM before formation of open complex. Reactions were stopped byadding 125 μL of stop solution (0.6 M Tris, pH 8.0, 12 mM EDTA, 0.16 mg/mLtransfer RNA).

Heteroduplex Templates. Heteroduplex templates were constructed as de-scribed previously (24) and sequenced to confirm purity. Transcription wasperformed as described above.

Purification, Fractionation, and Analysis of Transcription Reactions. Stoppedtranscription reactions were phenol extracted by addition of 150 μL ofphenol/chloroform/isoamyl alcohol (25:24:1), vortexing, centrifugation, andcollection of the aqueous phase. Ethanol precipitation of RNA was per-formed by adding 450 μL of 100% ethanol to each reaction, followed bystorage at −20 °C overnight. Precipitated RNA was resuspended intranscription loading dye [1× transcription buffer, 80% (vol/vol) form-amide, 0.05% bromophenol blue, and xylene cyanol]. Reactions were frac-tionated by electrophoresis using 12% (wt/vol) denaturing polyacrylamide gelscontaining 7 M urea. Reactive bases were detected by an Amersham Bio-sciences Typhoon 9400 Variable Mode Imager. Quantitation was performedusing ImageQuant.

Data Analysis.Normalization. For all experiments, individual bands were normalized for in-corporation of [α-32P]GTP (or [α-32P]UTP in the case of Fig. 3 C and D) and nu-cleotide content to obtain relative numbers of RNA molecules.Quantification of +GreB experiments. For +GreB reactions, “%Paused” wascalculated for each time by dividing the sum of all paused complexes, asdetermined by decay over time in the presence of GreB, by the sum of allquantifiable bands (paused complexes, terminated, and readthrough bands)at 120 s, when accumulation of long products is complete. The 120-s timetotal was used as the divisor for all times to account for the presence ofnonquantifiable bands at early times. By 120 s, the complexes represented bythese bands have accumulated in the terminated and readthrough bands andare therefore quantifiable.

The %Paused values described above were used to generate an expo-nential trend line from the average of two replicates. For most templates, thepause half-life remained constant until the %Pause dropped below ∼30%,after which a slower decay occurred. For this reason, the number of datapoints to which each exponential trend was fit varied from 3 to 5 and wasdetermined by assessing the point at which measured values deviated fromthe initial decay of the pause. The trend line generated as described abovewas then used to determine pause half-life and induction (the y intercept ofthe trend line). For most templates, pause induction was several percentagepoints above 100%; we report these values as 100%.Quantification of −GreB experiments. For −GreB reactions, “%Paused at +16”was calculated for each time by dividing the +16 pause band by the sum of allquantifiable bands (arrested, terminated, and readthrough). In contrast tothe calculation described for +GreB samples, the low level of nonquantifiablebands allowed us to perform this calculation within each lane. The exceptionto this is ConsUE, which, because of the prevalence of nonquantifiable bandsat early time points, required use of the 300-s time point as the divisor for alltime points.

The %Paused at +16 values were used to determine +16 pause half-life andcapture efficiency by generating an exponential trend line from the average of

6 of 7 | www.pnas.org/cgi/doi/10.1073/pnas.1512986112 Strobel and Roberts

two replicates. A third value, the probability that σ70 will disengage the −10-likesequence on a given escape attempt, was determined for the 60-s time point bydividing the sum of the terminated and readthrough bands by the sum ofarrested, terminated, and readthrough bands. The 60-s time point was used forthis calculation because at 60 s most elongation complexes have proceededbeyond +16; however, there is not sufficient time for the endogenous transcript

cleavage activity of RNAP to rescue arrested TECs, which would allow additionalcleavage attempts and could obscure the measurement of escape probability.

ACKNOWLEDGMENTS. We thank David Shalloway for helpful discussionsregarding kinetic models, and the referees for their insightful suggestions.This work was funded by NIH Grant GM21941.

1. Kwak H, Lis JT (2013) Control of transcriptional elongation. Annu Rev Genet 47:483–508.

2. Santangelo TJ, Artsimovitch I (2011) Termination and antitermination: RNA poly-merase runs a stop sign. Nat Rev Microbiol 9(5):319–329.

3. Ring BZ, Yarnell WS, Roberts JW (1996) Function of E. coli RNA polymerase sigmafactor sigma 70 in promoter-proximal pausing. Cell 86(3):485–493.

4. Artsimovitch I, Landick R (2000) Pausing by bacterial RNA polymerase is mediated bymechanistically distinct classes of signals. Proc Natl Acad Sci USA 97(13):7090–7095.

5. Chan CL, Landick R (1993) Dissection of the his leader pause site by base substitutionreveals a multipartite signal that includes a pause RNA hairpin. J Mol Biol 233(1):25–42.

6. Bochkareva A, Yuzenkova Y, Tadigotla VR, Zenkin N (2012) Factor-independenttranscription pausing caused by recognition of the RNA-DNA hybrid sequence. EMBOJ 31(3):630–639.

7. Roberts JW, et al. (1998) Antitermination by bacteriophage lambda Q protein. ColdSpring Harb Symp Quant Biol 63:319–325.

8. Kapanidis AN, et al. (2006) Initial transcription by RNA polymerase proceeds througha DNA-scrunching mechanism. Science 314(5802):1144–1147.

9. Revyakin A, Liu C, Ebright RH, Strick TR (2006) Abortive initiation and productiveinitiation by RNA polymerase involve DNA scrunching. Science 314(5802):1139–1143.

10. Marr MT, Roberts JW (2000) Function of transcription cleavage factors GreA and GreBat a regulatory pause site. Mol Cell 6(6):1275–1285.

11. Strobel EJ, Roberts JW (2014) Regulation of promoter-proximal transcription elon-gation: Enhanced DNA scrunching drives λQ antiterminator-dependent escape from aσ70-dependent pause. Nucleic Acids Res 42(8):5097–5108.

12. Hsu LM (2002) Promoter clearance and escape in prokaryotes. Biochim Biophys Acta1577(2):191–207.

13. Borukhov S, Sagitov V, Goldfarb A (1993) Transcript cleavage factors from E. coli. Cell72(3):459–466.

14. Perdue SA, Roberts JW (2010) A backtrack-inducing sequence is an essential compo-nent of Escherichia coli σ70-dependent promoter-proximal pausing. Mol Microbiol78(3):636–650.

15. Herbert KM, et al. (2006) Sequence-resolved detection of pausing by single RNApolymerase molecules. Cell 125(6):1083–1094.

16. Hein PP, Palangat M, Landick R (2011) RNA transcript 3′-proximal sequence affectstranslocation bias of RNA polymerase. Biochemistry 50(32):7002–7014.

17. Larson MH, et al. (2014) A pause sequence enriched at translation start sites drivestranscription dynamics in vivo. Science 344(6187):1042–1047.

18. Vvedenskaya IO, et al. (2014) Interactions between RNA polymerase and the “corerecognition element” counteract pausing. Science 344(6189):1285–1289.

19. Bird J (2013) Characterization of novel DNA elements necessary for sigma-dependentpromoter proximal transcriptional pausing. PhD thesis (Cornell University, Ithaca, NY).

20. Toulokhonov I, Zhang J, Palangat M, Landick R (2007) A central role of the RNApolymerase trigger loop in active-site rearrangement during transcriptional pausing.Mol Cell 27(3):406–419.

21. Imashimizu M, et al. (2013) Intrinsic translocation barrier as an initial step in pausingby RNA polymerase II. J Mol Biol 425(4):697–712.

22. Weixlbaumer A, Leon K, Landick R, Darst SA (2013) Structural basis of transcriptionalpausing in bacteria. Cell 152(3):431–441.

23. Yarnell WS, Roberts JW (1992) The phage lambda gene Q transcription antiterminatorbinds DNA in the late gene promoter as it modifies RNA polymerase. Cell 69(7):1181–1189.

24. Ring BZ, Roberts JW (1994) Function of a nontranscribed DNA strand site in tran-scription elongation. Cell 78(2):317–324.

25. Haugen SP, et al. (2006) rRNA promoter regulation by nonoptimal binding of sigma re-gion 1.2: An additional recognition element for RNA polymerase. Cell 125(6):1069–1082.

26. Feklistov A, et al. (2006) A basal promoter element recognized by free RNA poly-merase sigma subunit determines promoter recognition by RNA polymerase holo-enzyme. Mol Cell 23(1):97–107.

27. Landick R, Wang D, Chan CL (1996) Quantitative analysis of transcriptional pausing byEscherichia coli RNA polymerase: His leader pause site as paradigm.Methods Enzymol274:334–353.

28. Deighan P, Pukhrambam C, Nickels BE, Hochschild A (2011) Initial transcribed regionsequences influence the composition and functional properties of the bacterialelongation complex. Genes Dev 25(1):77–88.

29. Grayhack EJ, Yang XJ, Lau LF, Roberts JW (1985) Phage lambda gene Q antiterminatorrecognizes RNA polymerase near the promoter and accelerates it through a pausesite. Cell 42(1):259–269.

30. Nickels BE, Roberts CW, Sun H, Roberts JW, Hochschild A (2002) The σ70 subunit ofRNA polymerase is contacted by the λQ antiterminator during early elongation. MolCell 10(3):611–622.

31. Ko DC, Marr MT, Guo J, Roberts JW (1998) A surface of Escherichia coli sigma 70required for promoter function and antitermination by phage lambda Q protein.Genes Dev 12(20):3276–3285.

32. Marr MT, Roberts JW (1997) Promoter recognition as measured by binding of poly-merase to nontemplate strand oligonucleotide. Science 276(5316):1258–1260.

33. Belogurov GA, et al. (2007) Structural basis for converting a general transcriptionfactor into an operon-specific virulence regulator. Mol Cell 26(1):117–129.

34. Perederina AA, et al. (2006) Cloning, expression, purification, crystallization and ini-tial crystallographic analysis of transcription elongation factors GreB from Escherichiacoli and Gfh1 from Thermus thermophilus. Acta Crystallogr Sect F Struct Biol CrystCommun 62(Pt 1):44–46.

35. Schmidt MC, Chamberlin MJ (1984) Amplification and isolation of Escherichia colinusA protein and studies of its effects on in vitro RNA chain elongation. Biochemistry23(2):197–203.

Strobel and Roberts PNAS Early Edition | 7 of 7

BIOCH

EMISTR

YPN

ASPL

US