Single-Molecule Studies of RNA Polymerase: Motoring Along Kristina M. Herbert, 1 William J. Greenleaf, 2 and Steven M. Block 2,3 1 Biophysics Program, 2 Department of Applied Physics, and 3 Department of Biological Sciences, Stanford University, Stanford, California 94305; email: [email protected]Annu. Rev. Biochem. 2008. 77:149–76 First published online as a Review in Advance on April 14, 2008 The Annual Review of Biochemistry is online at biochem.annualreviews.org This article’s doi: 10.1146/annurev.biochem.77.073106.100741 Copyright c 2008 by Annual Reviews. All rights reserved 0066-4154/08/0707-0149$20.00 Key Words elongation, fluorescence, initiation, optical traps, termination, transcription Abstract Single-molecule techniques have advanced our understanding of transcription by RNA polymerase (RNAP). A new arsenal of ap- proaches, including single-molecule fluorescence, atomic-force mi- croscopy, magnetic tweezers, and optical traps (OTs) have been em- ployed to probe the many facets of the transcription cycle. These approaches supply fresh insights into the means by which RNAP identifies a promoter, initiates transcription, translocates and pauses along the DNA template, proofreads errors, and ultimately termi- nates transcription. Results from single-molecule experiments com- plement the knowledge gained from biochemical and genetic assays by facilitating the observation of states that are otherwise obscured by ensemble averaging, such as those resulting from heterogeneity in molecular structure, elongation rate, or pause propensity. Most studies to date have been performed with bacterial RNAP, but work is also being carried out with eukaryotic polymerase (Pol II) and single-subunit polymerases from bacteriophages. We discuss recent progress achieved by single-molecule studies, highlighting some of the unresolved questions and ongoing debates. 149 AR Further Click here for quick links to Annual Reviews content online, including: • Other articles in this volume • Top cited articles • Top downloaded articles • AR’s comprehensive search Annu. Rev. Biochem. 2008.77:149-176. Downloaded from arjournals.annualreviews.org by Yale University STERLING CHEMISTRY LIBRARY on 06/30/08. For personal use only.
Transcript
ANRV345-BI77-08 ARI 28 April 2008 11:12
Single-Molecule Studiesof RNA Polymerase:Motoring AlongKristina M. Herbert,1 William J. Greenleaf,2
and Steven M. Block2,3
1Biophysics Program, 2Department of Applied Physics, and 3Department of BiologicalSciences, Stanford University, Stanford, California 94305; email: [email protected]
Annu. Rev. Biochem. 2008. 77:149–76
First published online as a Review in Advance onApril 14, 2008
The Annual Review of Biochemistry is online atbiochem.annualreviews.org
This article’s doi:10.1146/annurev.biochem.77.073106.100741
AbstractSingle-molecule techniques have advanced our understanding oftranscription by RNA polymerase (RNAP). A new arsenal of ap-proaches, including single-molecule fluorescence, atomic-force mi-croscopy, magnetic tweezers, and optical traps (OTs) have been em-ployed to probe the many facets of the transcription cycle. Theseapproaches supply fresh insights into the means by which RNAPidentifies a promoter, initiates transcription, translocates and pausesalong the DNA template, proofreads errors, and ultimately termi-nates transcription. Results from single-molecule experiments com-plement the knowledge gained from biochemical and genetic assaysby facilitating the observation of states that are otherwise obscuredby ensemble averaging, such as those resulting from heterogeneityin molecular structure, elongation rate, or pause propensity. Moststudies to date have been performed with bacterial RNAP, but workis also being carried out with eukaryotic polymerase (Pol II) andsingle-subunit polymerases from bacteriophages. We discuss recentprogress achieved by single-molecule studies, highlighting some ofthe unresolved questions and ongoing debates.
149
AR FurtherClick here for quick links to Annual Reviews content online, including:
• Other articles in this volume• Top cited articles• Top downloaded articles• AR’s comprehensive search
Ann
u. R
ev. B
ioch
em. 2
008.
77:1
49-1
76. D
ownl
oade
d fr
om a
rjou
rnal
s.an
nual
revi
ews.
org
by Y
ale
Uni
vers
ity S
TE
RL
ING
CH
EM
IST
RY
LIB
RA
RY
on
06/3
0/08
. For
per
sona
l use
onl
y.
ANRV345-BI77-08 ARI 28 April 2008 11:12
RNAP: RNApolymerase
Promoter: aregulatory region ofDNA locatedupstream of a genebound by the RNAPholoenzyme
The information needed to create and sus-tain life is encoded within the DNA of ev-ery cell. The nanoscale machine that serves asthe molecular gatekeeper to this repository ofinformation is the enzyme RNA polymerase(RNAP). RNAP moves along the DNA tem-plate while transcribing selected portions intomessenger RNA, thereby initiating the pro-cess of gene expression. From a biophysi-cal perspective, the motion of RNAP alongDNA is reminiscent of the action of motorproteins, such as kinesin and myosin, whichtranslocate along microtubule or actin fila-ment substrates, respectively. The activitiesof RNAP are vastly more complex, however,befitting an enzyme that sits at the nexus ofpathways controlling cellular fate. The pro-cess of transcription can be divided broadlyinto three phases—initiation, elongation, andtermination—each characterized by distinctchemomechanical activities and levels of reg-ulation.
To initiate transcription, RNAP must firstrecognize and bind to an appropriate pro-moter sequence. A variety of initiation fac-tors influence RNAP’s specificity for differ-ent promoters. Some of these factors alsoaid the polymerase in forming an open pro-
moter complex (OPC), in which the DNA islocally melted to form a transcription bub-ble, exposing the bases of the template-strandDNA. From here, RNAP typically under-goes a process termed abortive initiation, whichinvolves the synthesis of a series of shortRNA transcripts, followed by their release andthe return of RNAP to the initial promotersite. Eventually, after a number of such fitsand starts, RNAP escapes the promoter re-gion, forming a stable, processive transcrip-tion elongation complex (TEC) capable oftranscribing the entire gene (1).
Elongation, during which individual nu-cleotides are added to the 3′ end of the grow-ing RNA chain, involves the coordinationof translocation along DNA with nucleosidetriphosphate (NTP) binding, nucleotide con-densation, and the release of inorganic py-rophosphate (PPi). As RNAP carries out elon-gation, the fundamental nucleotide additioncycle competes with a variety of off-pathwaystates, many of which have regulatory impor-tance. For example, upon encountering sitesof DNA damage, RNAP is thought to stop,and the subsequent backtracking of RNAPalong the DNA template triggers the pro-cess of transcription-coupled repair. More-over, the misincorporation of an incorrect nu-cleotide into the nascent RNA may activatenucleolytic activities inside the polymerase orrecruit additional cofactors that help excisethe base and correct the error. Finally, tran-scriptional pausing and arrest, i.e., the tran-sient or permanent entry into catalytically in-active states, also interrupt elongation; suchstates are targets of regulation by certain tran-scription factors (2).
Despite all these possibilities for interrup-tion, RNAP can successfully generate tran-scripts up to 106 nucleotides long (3). How-ever, this prodigious processivity must haltefficiently and precisely at the end of a gene.Transcriptional termination is induced by spe-cific structural elements that form in thenascent RNA or by active termination factors,which can act directly upon the TEC. Theeffect of these mechanisms is to release the
150 Herbert · Greenleaf · Block
Ann
u. R
ev. B
ioch
em. 2
008.
77:1
49-1
76. D
ownl
oade
d fr
om a
rjou
rnal
s.an
nual
revi
ews.
org
by Y
ale
Uni
vers
ity S
TE
RL
ING
CH
EM
IST
RY
LIB
RA
RY
on
06/3
0/08
. For
per
sona
l use
onl
y.
ANRV345-BI77-08 ARI 28 April 2008 11:12
newly minted RNA and dissociate the other-wise stable elongation complex, allowing thetranscription cycle to begin anew (4).
The repertoire of biomechanical processesdisplayed during transcription makes RNAPparticularly intriguing for study. Bulk bio-chemical studies have previously identifiedand characterized many of the essential ac-tivities of RNAP, but certain details are ob-scured by ensemble averaging or by the com-paratively limited time resolution available.Single-molecule techniques offer a meansto pick apart some of the catalytic statesand complex behaviors of individual macro-molecules with improved spatial and temporalresolution.
The physical mechanism of abortive ini-tiation was recently characterized using in-novative single-molecule fluorescence andmagnetic tweezers-based techniques. Thoseexperiments suggest that the template DNAbecomes “scrunched” within the footprint ofRNAP during the initiation phase of tran-scription (5, 6). Results from high-resolutionoptical trapping assays indicate that duringelongation, RNAP likely moves as a rigidbody along DNA in single-base increments,maintaining a tight coupling between move-ments along the DNA and the lengths ofRNA transcripts (7). The ability to apply con-trolled loads during active transcription hassupplied new information about chemome-chanical coupling, suggesting that RNAP mo-tion may be a consequence of the rectifi-cation of random thermal motion broughtabout by the binding and hydrolysis of nu-cleoside triphosphates (NTPs) (7–9). Detailedmeasurements of single-molecule elongationrates have identified a class of short-lifetimepauses that frequently interrupt transcription,even in genes previously thought to be devoidof strong regulatory pauses (10–13), addinganother layer of complexity to the kineticsof elongation. Finally, single-molecule stud-ies have indicated that the release of tem-plate DNA during the transcriptional termi-nation process is preceded by entry into anelongation-incompetent state (14).
Backtracking: thereverse translocationof RNAP (in theupstream directionalong the DNAtemplate) whilekeeping theRNA:DNA hybrid inregister
This review discusses these and other re-cent findings from single-molecule work, withan eye toward future applications.
SINGLE-MOLECULETECHNIQUES
Single-molecule techniques for investigatingthe transcription cycle fall into three classes:Atomic force microscopy (AFM), single-molecule fluorescence, and methods thattrack the motions of tiny particles to whichmolecules of interest are attached, such asmagnetic tweezers, optical traps (OTs) and thetethered particle motion (TPM) assay (15).
Scanning-mode AFM has been used suc-cessfully to image ultrastructural alterations inthe TEC, such as changes in the bend angles ofthe template DNA induced by RNAP (16). Tovisualize transcription, active TECs are gen-erally deposited onto an atomically flat sur-face, such as mica, then scanned with the tipof an AFM cantilever (Figure 1) as minute de-flections are detected by a laser that reflects offthe cantilever surface. Scanning-mode AFM
Laser
Positiondetector
Scanning stage
Sample
Image
AFM
Cantilever
Figure 1Atomic force microscopy (AFM). Transcription elongation complexes aredeposited onto an atomically flat surface (lower right panel ). Amicrofabricated cantilever with a sharp tip is scanned over the sample.Deflections of the cantilever are registered by means of laser light reflectedonto a position-sensitive detector. Detector signals are used to reconstruct atwo-dimensional image (simulated image shown in the upper right panel ).
www.annualreviews.org • Single-Molecule Studies of RNA Polymerase 151
Figure 2Single-molecule fluorescence methods. Fluorescence may be used to trackbinding and residence times of accessory factors (upper panel ) or theposition of the RNAP holoenzyme ( green) by covalently attaching afluorescent dye (star) and exciting it with an appropriate wavelength(wavy arrows). Forster (fluorescence) resonance energy transfer (FRET)(lower panel ) allows the determination of intramolecular distances throughfluorescent coupling between a donor ( yellow star) and an acceptor(red star) dye. In the lower left diagram, the donor ( yellow star) is excited(blue arrows) and emits light. When the donor fluorophore movessufficiently close to the acceptor (lower right), resonance energy transferresults in emission of a longer wavelength by the acceptor. The degree ofacceptor emission relative to donor excitation is sensitive to the distancebetween the attached dyes.
techniques have allowed the reconstructionof two-dimensional images of transcriptionalcomplexes to ∼2 nm accuracy; however, arti-facts produced by the process of sample de-position onto the surface, as well as the pro-jection of the three-dimensional molecularstructure into two dimensions, can complicatethe interpretation of images (17).
Single-molecule fluorescence tracking hasbeen used to monitor the binding and res-idence time of fluorescently tagged tran-scription factors that influence the catalyticproperties of RNAP (18) (Figure 2). By track-ing a tagged RNAP itself, or by monitoringthe incorporation of fluorescent nucleotidesinto an RNA chain, the processes of promotersearch or elongation can be studied with min-
imal perturbation (19–23). Structural rear-rangements of the TEC that occur during thetranscription cycle can also be monitored us-ing Forster (fluorescence) resonance energytransfer (FRET) (Figure 2) (6, 24, 25). FRETcan follow the distance between two appro-priately selected fluorophores by means ofthe nonradiative coupling of fluorescence en-ergy from one to another, which leads to achange in the emission properties. This tech-nique requires both a donor and an accep-tor dye, which are each covalently attached tothe molecule(s) of interest in close proximity,typically within 2–10 nm. When the donorfluorophore is exposed to excitation light, itcan transfer some of its excited-state energyto the acceptor fluorophore in a process thatdepends on the inverse sixth power of thedistance between fluorophores. By measuringthe intensity change in acceptor fluorescence,distances on the order of nanometers can cur-rently be measured in single molecules withmillisecond time resolution (26).
The final class of single-molecule tech-niques typically employs micrometer-sizedbeads attached to single RNAP molecules orto associated nucleic acids. Records of thepositions of these beads report on the lo-cations or rotational states of the enzyme(27–29). Such beads can also be used as “han-dles” through which forces may be applied tomolecules (28, 30). The position of a bead-tagged RNAP can be determined sensitivelyby measuring the light scattered from thebead, either by centroid tracking in video im-ages (31) or, more precisely, through laser-based light scattering (28).
By tethering the polymerase and the endof the DNA template between a polystyrenebead and the cover glass surface, the chang-ing length of the DNA, and therefore theprogress of the enzyme, can be determinedfrom the averaged amplitude of the Brown-ian motion of the tethered bead. This TPMassay is shown in Figure 3a (32). Improvedresolution in the length of the DNA tethermay be obtained by applying external force tothe bead, thereby straightening the tether and
152 Herbert · Greenleaf · Block
Ann
u. R
ev. B
ioch
em. 2
008.
77:1
49-1
76. D
ownl
oade
d fr
om a
rjou
rnal
s.an
nual
revi
ews.
org
by Y
ale
Uni
vers
ity S
TE
RL
ING
CH
EM
IST
RY
LIB
RA
RY
on
06/3
0/08
. For
per
sona
l use
onl
y.
ANRV345-BI77-08 ARI 28 April 2008 11:12
allowing for a more direct measurement ofdisplacement. Apart from improving the po-sitional resolution of the measurement, appli-cation of such a force can be used to ener-getically bias steps in the transcription cyclethat involve motion along the tether. Just aschanges in substrate concentration bias chem-ical reactions through mass action, variationsin applied force bias translocation, providingfurther insights into mechanisms of motormotion (Figure 4).
One way to apply force to the bead is withan OT. OTs consist of tightly focused beamsof infrared laser light that exert controlledforces on small dielectric particles, such aspolystyrene beads, by means of radiation pres-sure (28). Depending on the geometry of theassay, up to ∼30 pN of tension can be ap-plied either to the upstream DNA (Figure 3b)(an assisting load), to the downstream DNA(a hindering load) (12, 33), or to the nascentRNA (Figure 3c) (11). This surface-basedassay offers nanometer-scale positional reso-lution.
Alternately, force can be applied to thebead by means of laminar fluid flow. The dis-tal end of the template DNA is attached to asecond bead (rather than the cover glass sur-face), which is held by a micropipette, so thatfluid flow exerts force on the free bead, plac-ing tension on the DNA template (Figure 3d )(30).
In both OT-based and fluid-flow assays,the DNA tether is attached to a nominallyfixed reference point, i.e., to the cover glasssurface or to a micropipette. However, anyresidual motions of these points can result insignificant drift and signal noise. To circum-vent such sources of noise, the assay compo-nents may be levitated by using an additionalOT (Figure 3e) (34), thereby decoupling mo-tions of the surface. Such a “dumbbell” assaycan achieve single-base pair positional reso-lution, allowing transcription to be followedat the level of individual catalytic turnovers(7, 35).
Rotational motions are produced as RNAPtracks along the helical pitch of DNA. Such
Plectoneme:crossings of DNAhelices. Topologyconstrains the twist(number of right-handed turns of theDNA double helix)plus plectonemes
Holoenzyme: amacromolecularstructure comprisingthe core enzyme (α2,β, β′) plus σ-factor,required for RNAPbinding to apromoter sequence
rotations can be observed directly by tetheringa large bead decorated with smaller fluores-cent beads to the TEC (Figure 3f ) (27, 29).
The local melting of DNA that occurs dur-ing initiation can be detected using magnetictweezers, which can exert both tension andtorque on a DNA template attached to super-paramagnetic beads (36). When such a DNAmolecule is placed under small amounts oftension and then twisted, large loops (plec-tonemes) are formed (Figure 3g). The num-ber of plectonemes formed is related to thetotal amount of excess twist in the DNA. Pro-cesses that change the degree of twist, suchas melting of the promoter region that oc-curs during abortive initiation cycles, changethe number of plectonemes. This produces alarge change in the height of the magneticbead over the cover glass. A mere 1–2 basesof melting induces 5–10-nm changes in theaxial bead position, which can be detected byoptical techniques.
Together, these diverse single-moleculeapproaches supply an ample toolbox of tech-niques, each suited to measuring different as-pects of transcription.
INITIATION
Transcription initiation is a multistep process,requiring RNAP to locate and bind to a pro-moter, unwind dsDNA to form an open com-plex, begin RNA transcription, then escapefrom the promoter region, perhaps releas-ing initiation factors in the process. Single-molecule experiments have supplied someunique observations addressing the molecularmechanisms of several aspects of the intiationphase.
Promoter Search
The initiation of transcription requires spe-cific binding of the holoenzyme to DNApromoter sequences scattered throughout avast excess of genomic DNA, a search prob-lem that is common to all sequence-specificDNA-binding proteins. In 1970, LacI was
www.annualreviews.org • Single-Molecule Studies of RNA Polymerase 153
Ann
u. R
ev. B
ioch
em. 2
008.
77:1
49-1
76. D
ownl
oade
d fr
om a
rjou
rnal
s.an
nual
revi
ews.
org
by Y
ale
Uni
vers
ity S
TE
RL
ING
CH
EM
IST
RY
LIB
RA
RY
on
06/3
0/08
. For
per
sona
l use
onl
y.
ANRV345-BI77-08 ARI 28 April 2008 11:12
reported to bind to the lac operator site atrates 1000-fold faster than could be explainedby random, diffusional encounters with theDNA in three dimensions (37), and an analo-gous phenomenon has also been reported forRNAP (38). Two independent mechanisms,sliding and intersegment transfer, have beenproposed to account for the enhanced bindingrates. These mechanisms both serve to reducethe effective dimensionality of the search pro-cess, increasing its efficiency by orders of mag-nitude (39, 40). Sliding results when RNAPassociates weakly with nontarget DNA, al-lowing it to diffuse in a one-dimensionalrandom walk until it reaches the target site. In-tersegment transfer involves the polymerasecrossing from one position on the templateto another more distant position by meansof an intermediate state where the protein
is bound simultaneously to both proximaland distal DNA segments. Multiple transferevents occur until the promoter site is even-tually reached (Figure 5).
A number of biochemical assays have de-veloped indirect evidence that lends sup-port to the sliding mechanism (41–43).However, only single-molecule experimentspermit direct observation of the trajectoriesof individual RNAP molecules during thepromoter-search process, providing a uniquewindow into this phase of initiation. Usinga fluorescently labeled antibody for RNAP,Shimamoto and colleagues (20) imaged themotions of holoenzymes along nonspecific se-quences of DNA molecules oriented in thepresence of laminar fluid flow and observedstable binding only to specific promoter sites.The fluid flow used in the assay converts what
a b c
Bead-based single-molecule transcription assays
S N
F
F
F F
S N
SN
d
F
e
f
g
F
F
154 Herbert · Greenleaf · Block
Ann
u. R
ev. B
ioch
em. 2
008.
77:1
49-1
76. D
ownl
oade
d fr
om a
rjou
rnal
s.an
nual
revi
ews.
org
by Y
ale
Uni
vers
ity S
TE
RL
ING
CH
EM
IST
RY
LIB
RA
RY
on
06/3
0/08
. For
per
sona
l use
onl
y.
ANRV345-BI77-08 ARI 28 April 2008 11:12
would otherwise be bidirectional Brownianmotion of polymerase into largely unidirec-tional motion that can be observed via flu-orescence tracking (Figure 2). In follow-upexperiments designed to determine if groovetracking along the DNA helix occurred con-comitant with polymerase sliding, small, flu-orescently decorated beads were attached tothe DNA, which was then dragged overRNAP holoenzymes immobilized on the sur-face (Figure 3f ). The authors observed smallrotational motions of the beads that were con-sistent with polymerase groove tracking dur-ing linear diffusion along DNA (29).
Using total internal reflection fluorescencemicroscopy (in this case, in the absence ofany fluid flow or stage movements that mightperturb diffusion), Harada et al. (19) ob-served the binding and dissociation of Cy3-labeled RNAP from λ DNA molecules thatwere stretched between twin OTs. In rare in-
stances, individual RNAP molecules movedrandomly along the DNA template over dis-tances greater than 200 nm. These events sup-plied evidence for RNAP sliding along non-specific DNA. However, based on such events,Harada et al. estimated the linear diffusion co-efficient of RNAP to be 104 nm2/s, which is1–3 orders of magnitude smaller than (model-dependent) values implied by rates of pro-moter binding measured in solution studies(44). This discrepancy may be attributable tothe rarity of measurable events, to the per-turbing effects of tension on the system, orto the validity of bulk estimates of diffusionrates. Other estimates of diffusion rates havebeen obtained using time-resolved AFM toacquire sequential images every 100 secondsfor RNAP diffusing along λ DNA. The posi-tion of RNAP on the DNA varied from oneimage to the next, consistent with a randomwalk in one dimension with a diffusion rate of
←−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−Figure 3Bead-based, single-molecule transcription assays showing DNA (large blue strands), RNA (red strands),beads (blue spheres), optical traps (OTs, pink), fluorescent beads ( yellow), magnets and magnetic beads(orange/blue gradients). Directions of applied forces (F) are shown by straight black arrows. The coverglass surface is indicated by a blue horizontal line. (a) Tethered particle motion assay. This method trackstranscriptional progress by averaging the Brownian excursions of a bead tethered by a changing length ofDNA to a molecule of RNAP ( green) immobilized on the cover glass surface. (b) Surface-basedDNA-pulling OT assay. RNA polymerase is bound to a bead, and the distal end the DNA template isanchored to the cover glass. Force is exerted on the bead by an OT. Here, the force is shown assistingpolymerase motion; reversal of the template direction allows the application of hindering loads.(c) Surface-based RNA-pulling OT assay. A molecular handle consisting of dsDNA with acomplementary ssDNA overhang is annealed to the 5′ end of the nascent RNA. As in panel b, the RNAPis bound to a bead, and the DNA is anchored to the cover glass. Forces applied to the bead producetension on the transcript. (d ) Pipette-based DNA-pulling assay. An RNAP molecule is bound to a beadheld by a suction micropipette, and the distal end of the DNA template is attached to a second, free bead.Fluid flow exerts viscous forces on the free bead (right), placing tension on the tether. (e) Dumbbell OTassay. Two beads, one attached to an RNAP molecule and the other to the distal end of the DNAtemplate, are levitated above the surface by twin OTs. Transcriptional progress of RNAP can bemeasured free of the drift caused by motion of the cover glass surface. ( f ) Fluorescent particle rotationassay. A larger bead is decorated with smaller fluorescent reporter beads, which can be used to determineits angle about a vertical axis. Similar to panel a, the larger bead is tethered to a molecule of RNAP on thecover glass surface through the template DNA. Rotations of RNAP around the DNA template axisduring elongation or promoter search lead to rotations of the larger bead that can be directly visualized.( g) Magnetic tweezer assay. A superparamagnetic bead is tethered to one end of a DNA molecule whosedistal end is attached to the cover glass surface. External magnets are used to impart both twist and smallamounts of tension to the DNA. Rotations of these magnets underwind the DNA and induce theformation of plectonemes. Melting of the transcription bubble during initiation adds a positive twist tothe template, removing plectonemes and causing a large change in the height of the tethered bead thatcan be measured directly (36).
www.annualreviews.org • Single-Molecule Studies of RNA Polymerase 155
Ann
u. R
ev. B
ioch
em. 2
008.
77:1
49-1
76. D
ownl
oade
d fr
om a
rjou
rnal
s.an
nual
revi
ews.
org
by Y
ale
Uni
vers
ity S
TE
RL
ING
CH
EM
IST
RY
LIB
RA
RY
on
06/3
0/08
. For
per
sona
l use
onl
y.
ANRV345-BI77-08 ARI 28 April 2008 11:12
δtrans
δeq
F*δtrans
F*δeq
Natural energy landscapeEnergy landscape under force
Energy
X A B
Figure 4Force affects translocation. Force biases rates of chemical reactionsinvolving translocation. The graph depicts a notional energy landscape(solid red line) connecting states A (initial ) and B ( final ) with the reactioncoordinate, X, corresponding to displacement along the direction oftranslocation. In this example, a retarding force, F, tilts the energylandscape by an amount equal to the work performed against the appliedload, F ∗X, producing a changed landscape (dashed green line). Thetransition state for this reaction is raised by an amount F ∗δtrans, whereδtrans is the distance to the transition state located between A and B. Forcealso affects the thermodynamic equilibrium between states A and B,raising the relative energy of state B by an amount F ∗δeq, where δeq is theequilibrium distance between A and B. The capacity to bias bothtransition-state barriers as well as equilibrium constants of chemicalreactions makes force a powerful tool in probing the nature ofchemomechanical coupling. This general reaction picture is readilyinterpreted in terms of RNAP during its elongation phase. If state A andB are taken to represent pre- and posttranslocated states of thetranscription elongation complex, then the reaction diagram abovecorresponds to a force-dependent translocation mechanism (69). Thissame picture can be adapted equally well to represent the forcedependence of pausing or backtracking events, if state B is taken torepresent a configuration of RNAP competent for active elongation, andstate A is taken to represent a paused or backtracked state.
101 nm2/s (45, 46), which is again too low toexplain bulk search data.
Despite the troubling differences in the es-timated diffusion coefficients between single-molecule and bulk experiments, both provideconvincing evidence for sliding as a possiblemechanism. In addition, time-resolved AFMimaging has provided some support for theintersegment transfer mechanism. A single
RNAP molecule was observed contacting twopositions along DNA and then transferringfrom one position to the other. In anotherimage series, RNAP dissociated and reboundat a different template position (45, 46). Ona cautionary note, AFM measures molecularinteractions in only two dimensions that oc-cur near the surface, leading to the possibilitythat surface effects may restrict diffusion orthat the reduced dimensionality of the mea-surement may promote intersegment transferor rebinding.
Open-Complex Formation
Upon locating a specific promoter site,RNAP undergoes a structural transition fromthe closed promoter complex to the OPC(Figure 5). During this transition, RNAPbends and unwinds a local segment of DNAwith the help of initiation factors such asσ, forming the transcription bubble. The“housekeeping” factor, σ70, requires no addi-tional factors to unwind a promoter sequenceand directs RNAP to recognize the vast ma-jority of promoters in enteric bacteria. AFMimages of Escherichia coli RNAP-σ70 OPCsformed at the λPR or λPL promoters show thatthe DNA is bent between 55◦ and 88◦ (by con-vention, this angle refers to the deviation froma straight line, not to the included angle ofbend) (47, 48). These measurements are con-sistent with bend angles inferred from gel mo-bility assays that compared bent A-tract DNAto OPCs (47). RNAP with σ54, a factor thatrequires the transcriptional activator NtrC tounwind DNA, was imaged on the glnA pro-moter DNA in both the presence and absenceof NtrC, allowing comparison of closed andopen complexes. The closed promoter com-plex yielded a DNA bend angle of 49◦ ± 24◦,whereas the open complex bent the templateDNA 114◦ ± 18◦ (49). Experimental differ-ences in the DNA bend angles produced byOPCs carrying either σ70 and σ54 might be at-tributable, in principle, to the different sigmafactors or to different promoter sequences.A more detailed study of DNA bend angles
156 Herbert · Greenleaf · Block
Ann
u. R
ev. B
ioch
em. 2
008.
77:1
49-1
76. D
ownl
oade
d fr
om a
rjou
rnal
s.an
nual
revi
ews.
org
by Y
ale
Uni
vers
ity S
TE
RL
ING
CH
EM
IST
RY
LIB
RA
RY
on
06/3
0/08
. For
per
sona
l use
onl
y.
ANRV345-BI77-08 ARI 28 April 2008 11:12
Closed complex Open complex Abortive initiation Promoter clearance
σ-Factor release
Promoter search and initiation
Sliding
Intersegment transfer
Figure 5Promoter search and initiation. The RNAP holoenzyme (core polymerase in green; σ-factor in purple) ispostulated to find promoter target sequences through two mechanisms. During intersegment transfer,the polymerase binds loosely to a position on the DNA (blue) and then makes bridging contacts to asecond position on the DNA before transferring from one position to the other. The sliding mechanisminvolves the diffusion of weakly bound RNAP along the DNA. Once a promoter is found, theholoenzyme binds tightly to the DNA and bends it, forming the closed promoter complex. With the helpof σ-factor, a portion of the DNA melts, exposing bases of the template strand and forming the openpromoter complex. Subsequently, during abortive initiation, RNAP repeatedly synthesizes short RNAfragments before eventually clearing the promoter to form a stable transcription elongation complex, andσ-factor is likely released (1).
induced by static OPCs on various promot-ers might therefore be revealing and couldbe complemented by bulk studies comparingthe gel mobility of these complexes to A-tractDNA. To observe the dynamics of OPC for-mation, however, additional single-moleculetechniques must be employed.
Single-molecule magnetic tweezer experi-ments allow the observation of OPC forma-tion in real time (36), a process that producesconformational changes but no RNA tran-script and therefore is difficult to measure inbulk. Strick and coworkers (50) created tor-sionally constrained tethers of DNA, contain-ing either a strong lacCONS promoter or aweaker rrnB P1 promoter, by attaching oneend of the DNA template to the cover glassand the other to a magnetic bead. By rotat-ing external magnets, torque was applied tothe bead, introducing either positive or neg-ative supercoils into the DNA (Figure 3g).Topology constrains the linking number ofthe DNA (which, in this case, is the sum of the
positive plectonemes formed and the numberof right-handed turns of the DNA helix) sothat it remains fixed. For every helical turn ofDNA unwound by RNAP during OPC for-mation, one full positive plectoneme is cre-ated (or negative plectoneme destroyed). Thischange in the number of plectonemes resultsin an axial movement (up or down) of the mag-netic bead of roughly 50 nm for every 10 bpunwound, depending on the initial tension ap-plied. However, any compaction of the DNAresults in an axial reduction in the height ofthe bead, regardless of the sign of the super-coils, permitting unwinding signals to be dis-tinguished from compaction. The unwindingof as little as 1 bp and compaction as small as5 nm can be observed.
Using this sensitive technique, differencesin the amplitudes of transitions between pos-itively and negatively supercoiled DNA wereused to infer that the opening of the promoterregion unwinds 1.2 ± 0.1 turns and compacts(because of wrapping and/or bending of the
www.annualreviews.org • Single-Molecule Studies of RNA Polymerase 157
DNA) 15 ± 5 nm of DNA (50). This unwind-ing amount is consistent with that inferred byprevious biochemical footprinting assays (51),and the level of DNA compaction is consis-tent with AFM imaging experiments of theλPR promoter (47).
Negative supercoiling of DNA energeti-cally favors melting of the transcription bub-ble, whereas positive supercoiling makes itless favorable. Consistent with these effects oftorque, OPCs formed irreversibly on under-wound DNA carrying the strong lacCONSpromoter, whereas overwound DNA dis-played transitions from the closed to OPCsand vice versa. In contrast, on DNA carryingthe weak rrnB P1 promoter, reversible tran-
Scrunching
Inchworming
Transient excursions
Figure 6Abortive initiation models. The initially transcribing complex is shownwith RNA polymerase, DNA, RNA, and the enzyme active site. Threemechanisms have been proposed to explain abortive initiation. In thetransient excursions model (top), RNAP briefly breaks its contacts with thepromoter region ( green horizontal arrows) and transcribes a short segmentof RNA. Upon release of the aborted product, RNAP diffuses back torestart the cycle. In the inchworming model (middle), flexible elementswithin the enzyme allow the footprint of RNAP to grow as RNA issynthesized, and promoter contacts at the upstream face are maintained(blue vertical arrow). Upon release of the abortive RNA, the polymeraserelaxes to its normal footprint. In the scrunching model (bottom), RNAPmaintains its shape while increasing its effective footprint by pulling insome of the downstream DNA. Abortive loss of the RNA transcript thenresults in the release of this scrunched DNA, resetting the enzyme.
sitions were found on negatively supercoiledDNA only, suggesting that strong promotersmay be easier to melt than weak promoters.In the presence of initiating nucleotides orthe transcription effecter ppGpp, the stabil-ity of the OPC was dramatically increased ordecreased, respectively (50).
Plectonemes can form in the DNA tem-plate only under specific conditions of torqueand tension. Because the change in plec-toneme number constitutes the signal used forthe detection of initiation, torque and tensionare interdependent and cannot be varied atwill in magnetic tweezer assays. Newer formsof sensitive optical instrumentation are beingdeveloped that may permit a wider range oftorques and tensions to be explored at highbandwidth (52).
Abortive Initiation
After forming the OPC, RNAP begins syn-thesis of an RNA oligonucleotide comple-mentary to the DNA template strand. Al-though RNAP forms a highly stable, pro-cessive complex during the elongation phase,the initially transcribing complex (ITC) iscomparatively unstable, spontaneously releas-ing short RNA chains and restarting synthe-sis, a process known as “abortive initiation”(Figure 5). The ability to synthesize manyshort transcripts coupled with the capacity toreinitiate quickly once a transcript has beenaborted implies that the active site of RNAPis able to move forward along DNA while si-multaneously maintaining promoter contact.How can this occur?
One model postulates that the RNAPmolecule makes transient downstream excur-sions on the template, briefly breaking itsbonds with the promoter, until the short RNAis released, and then the enzyme diffuses backto the promoter (53) (Figure 6). Such a modelis not easily reconciled with bulk footprintingdata, which suggest that the abortive initiationprocess results from an inability of RNAP tobreak its promoter contacts (54–56). Theseobservations led Straney & Crothers (55) to
158 Herbert · Greenleaf · Block
Ann
u. R
ev. B
ioch
em. 2
008.
77:1
49-1
76. D
ownl
oade
d fr
om a
rjou
rnal
s.an
nual
revi
ews.
org
by Y
ale
Uni
vers
ity S
TE
RL
ING
CH
EM
IST
RY
LIB
RA
RY
on
06/3
0/08
. For
per
sona
l use
onl
y.
ANRV345-BI77-08 ARI 28 April 2008 11:12
propose that the energy required to break freeof the promoter might be somehow storedin a “stressed intermediate” and that abortiveinitiation was a consequence of this energynot being used productively. One particularinstance of this concept, the “inchworming”model, postulates that flexible elements in-side RNAP might allow the active center tomove forward transiently with respect to theupstream face during synthesis, storing up en-ergy like a stretched spring that retracts uponaborted synthesis (Figure 6).
In a third model, the flexible element thatstores the energy ultimately used for promoterescape lies not in RNAP but in the single-stranded DNA of the transcription bubbleand its interactions with the enzyme. In thisscrunching model, RNAP functions more orless as a rigid body. The downstream DNAis pulled progressively into the enzyme witheach nucleotide addition cycle, producing ascrunched form within the enzyme footprint(Figure 6). Abortive RNA transcripts lead tothe release of the scrunched DNA, which isthen extruded out the downstream face ofRNAP (1, 56–58), only to be reeled in againupon further RNA synthesis.
To distinguish among these three possi-bilities, single-molecule FRET was used tomonitor the relative motions of componentsof the transcription complex during the iso-merization from the OPC to the ITC. Thefollowing quantities were measured: (a) thedistance between the RNAP leading edge anda point on the downstream DNA; (b) the dis-tance between the RNAP trailing edge anda point on the upstream DNA; (c) any ex-pansion or contraction within RNAP itself;and (d ) any expansion or contraction betweenpoints on the upstream and downstream DNA(6). Freely diffusing complexes were observedby confocal microscopy, using the techniqueof alternating-laser excitation (ALEX). Thisdual-laser method facilitates measurements ofFRET efficiency in select molecules carryingboth an active donor and active acceptor dye,eliminating the background of singly labeledmolecules (59, 60). For the lacCONS pro-
moter, distance changes were only observedbetween FRET pairs located on the RNAPleading edge and the downstream DNA(∼7-A contraction), as well as pairs locatedon the upstream DNA and downstream DNA(∼4-A contraction), consistent with scrunch-ing of the DNA during abortive initiation (6,61).
In parallel work, magnetic tweezers wereemployed to monitor the winding and un-winding of the DNA bubble during initialtranscription (Figure 3g). These single-molecule experiments supplied complemen-tary data in support of the scrunching model(5). The scrunching model uniquely predictsthat the extent of DNA unwinding should in-crease proportionally for longer RNA tran-scripts. Because the formation of plectonemesin pretwisted DNA makes the axial position ofa tethered bead sensitive to small amounts ofadditional twist, even the unwinding of a sin-gle base can be observed. Abortive initiationwas halted after varying amounts of transcriptwere synthesized by supplying the polymerasewith a subset of the four NTPs. For transcriptlengths beyond 2 bp, unwinding was observedequivalent to slightly less than the number ofbases in the nascent RNA. Furthermore, com-plexes spent the majority of time in an un-wound state, suggesting that abortive productsynthesis was fast compared to transcript re-lease, consistent with an independent single-molecule FRET experiment (62). With allfour NTPs present, transcription cycles wereobserved with four distinct transitions: (a) un-winding of the promoter DNA correspond-ing to the closed-to-open promoter transi-tion; (b) further unwinding, corresponding tothe ITC with scrunched DNA; (c) rewind-ing to a state consistent with a transcriptionbubble (identical to the size expected for theTEC, with no scrunched DNA); and (d ) fur-ther rewinding, back to the initial state upontranscription termination (5). These results,coupled with the FRET observations, providestrong evidence that promoter escape involvesDNA scrunching during the initial phase oftranscription.
www.annualreviews.org • Single-Molecule Studies of RNA Polymerase 159
Ann
u. R
ev. B
ioch
em. 2
008.
77:1
49-1
76. D
ownl
oade
d fr
om a
rjou
rnal
s.an
nual
revi
ews.
org
by Y
ale
Uni
vers
ity S
TE
RL
ING
CH
EM
IST
RY
LIB
RA
RY
on
06/3
0/08
. For
per
sona
l use
onl
y.
ANRV345-BI77-08 ARI 28 April 2008 11:12
Sigma Release
To initiate transcription, core RNAP mustbind to a dissociable σ-factor, forming theholoenzyme. Previously, it was widely be-lieved that σ-factor was released upon thetransition into the TEC (Figure 5), permit-ting individual RNAP molecules to bind dif-ferent σ factors during successive rounds oftranscription and thereby to respond quicklyto changes in cell cycle or growth conditions(63).
Alternative pictures explain the timing andmechanism of σ release. In the obligate re-lease model, σ dissociation is mechanisticallycoupled to promoter escape and occurs as thegrowing RNA transcript reaches eight or ninenucleotides (nt) in length, although the exactvalue may vary with different promoter se-quences (55, 64, 65). The stochastic releasemodel proposes instead that the affinity ofRNAP for σ decreases as the TEC is formed,so that σ gets released stochastically from thecomplex after the transition to the elongationphase (66).
Recent work has questioned the paradigmthat σ is released concomitant with promoterescape and suggested that a subpopulation ofelongation complexes may not release σ70 atall. Such a possibility might facilitate morerapid transcription of genes whose promot-ers require σ70 (67). Bulk (solution) FRETmeasurements with dye labels incorporatedinto σ and DNA showed persistent signals inTECs that had synthesized RNA transcriptsup to 50 nt long, although these signals diddecay with increasing transcript length (68).However, bulk experiments cannot differen-tiate between a homogeneous population ofcomplexes with a single lifetime and a hetero-geneous population of both long- and short-lived lifetimes, nor can they score possible re-association events.
Single-molecule FRET measurements,similar to those performed in bulk, con-firmed that σ70 is not released obligatorilyupon promoter escape. In experiments withfreely diffusing complexes, a significant frac-
tion of early TECs (transcript lengths of 11or 14 nt) were bound to σ-factor. Further-more, the introduction of a promoter-like se-quence in the initially transcribed region sig-nificantly increased the half-life of bound σ70
(18). The authors argued that the σ-factorremained bound as a consequence of ineffi-cient release upon transitioning to the elon-gation phase. However, the experiments withfreely diffusing complexes allow for the possi-bility that σ might be released upon promoterescape but subsequently rebind, for exam-ple, at certain promoter-like sequences foundin the downstream DNA. Subsequent single-molecule FRET experiments using surface-immobilized TECs permitted measurementswith improved time resolution and ruled outthe release or exchange of σ (62). In addi-tion to showing that early TECs retain σ, theexperiments with freely diffusing elongationcomplexes also showed that mature TECs(those with transcript lengths of 50 nt) stillretained ∼50% of σ, with a retention half-life of 50 min (18). This long lifetime hintsat the possible existence of a subpopulationof TECs that remain bound to σ70 through-out multiple rounds of transcription. It seemsworthwhile to conduct follow-up experimentsin the presence of other factors that may mod-ulate rates of σ release, e.g., NusA, competingσ factors, and core RNAP, to determine if sucha long half-life is consistent with the cellularmilieu.
ELONGATION
The elongation phase of transcription isknown to be a highly complex, multistate pro-cess. Conceptually, one can separate the elon-gation phase into a series of “on-pathway”states associated with DNA-templated RNAsynthesis via the nucleotide condensationreaction with PPi release (along with anyassociated translocations) and various “off-pathway” states that are incompetent for elon-gation, such as paused or arrested states(Figure 7).
160 Herbert · Greenleaf · Block
Ann
u. R
ev. B
ioch
em. 2
008.
77:1
49-1
76. D
ownl
oade
d fr
om a
rjou
rnal
s.an
nual
revi
ews.
org
by Y
ale
Uni
vers
ity S
TE
RL
ING
CH
EM
IST
RY
LIB
RA
RY
on
06/3
0/08
. For
per
sona
l use
onl
y.
ANRV345-BI77-08 ARI 28 April 2008 11:12
On-Pathway Elongation
During the transcription of a typical gene,RNAP processively translocates along DNA,generating an mRNA that may reach thou-sands of nucleotides in length. Single-molecule experiments have probed the sta-bility of the actively transcribing complex, aswell as the chemomechanics and kinetics ofthe elongation process, providing a windowinto the inner workings of the enzyme as itcarries out its primary biological function.
Structure and stability. Once the nascenttranscript reaches ∼9–11 nt in length, RNAPbreaks free of the promoter region and en-ters the elongation phase. At this point, theTEC becomes highly stable and processive,remaining tightly associated to both the tem-plate DNA and nascent RNA throughout (po-tentially) thousands of cycles of nucleotideaddition. The robustness of the TEC is dra-matically displayed in single-molecule opticaltrapping assays that are able to exert ex-traordinarily large loads (up to 30 pN ten-sion, applied to either the DNA or RNA)without disrupting transcription (11, 12, 69)(Figure 3b-e). The stability of the TEC isunderlined by the fact that transcriptionallystalled TECs can be prepared in advance andstored for weeks or longer, then restarted dur-ing an experiment by the addition of NTPs(70, 71). The primary stabilizing factor ofthe TEC has been presumed to be base pair-ing within the RNA:DNA hybrid (72). How-ever, the forces that can be applied to thenascent RNA without impairing transcrip-tion vastly exceed the forces required to un-zip or shear apart an 8–9-bp RNA:DNA du-plex. A “sliding clamp” model, where exten-sive protein-nucleic acid contacts within thepolymerase contribute significantly to RNAretention, has been proposed to explain over-all TEC stability (73). Such a clamp, consist-ing of a narrow protein channel surroundingthe hybrid, would also prevent any significantshearing motions between the RNA and DNAstrands under load because confinement in-
side a channel would lead to significant stericclashes between bases (11).
Prior to determination of the crystal struc-ture of RNAP (74, 75) and to the elucidationof the paths taken by nucleic acids throughthe elongating enzyme (73), AFM images ofTECs correctly measured the large bend an-gle between the upstream and downstreamDNA, which is close to 90◦ (48). Longer RNAtranscripts could occasionally be visualized inTEC images as well (76). However, the anglesmeasured between the RNA and DNA armsproved to be inconsistent with the currentlymodeled location of nucleic acids in the crys-tal structures, perhaps because of confound-ing surface interactions with the RNA, whichis substantially less rigid than DNA, or be-cause of the difficulties inherent in imaginga three-dimensional structure in two dimen-sions (77).
Step size. During elongation, translocationof the nucleic acid scaffold with respect tothe enzyme active site must be coordinatedwith the nucleotide condensation reaction.Initially, RNAP was postulated to behave asa rigid body, maintaining an invariant foot-print as it advanced by one base pair for ev-ery nucleotide added to the growing RNAchain (78). However, an inchworming modelof elongation was subsequently proposed in anattempt to rationalize the apparent differencesin the size of the enzyme footprint when com-plexes were halted at successive template po-sitions (79). During the proposed inchworm-ing motion, a flexible element, hypothesizedto exist within RNAP, allowed upstream anddownstream portions of the enzyme to moveout of phase, quasi independently, while si-multaneously producing its transcript one nu-cleotide at a time. The discrepancies in foot-print size were eventually reinterpreted asresulting from the backtracking behavior ofTECs, which were found to slide upstreamalong the template DNA under certain condi-tions (80–82). The inchworming model con-sequently fell into disfavor. However, otherbulk biochemical experiments have supplied
www.annualreviews.org • Single-Molecule Studies of RNA Polymerase 161
Ann
u. R
ev. B
ioch
em. 2
008.
77:1
49-1
76. D
ownl
oade
d fr
om a
rjou
rnal
s.an
nual
revi
ews.
org
by Y
ale
Uni
vers
ity S
TE
RL
ING
CH
EM
IST
RY
LIB
RA
RY
on
06/3
0/08
. For
per
sona
l use
onl
y.
ANRV345-BI77-08 ARI 28 April 2008 11:12
some evidence for the existence of flexible el-ements within RNAP itself (83–85), and re-cent single-molecule data obtained during theinitiation phase support the notion that thefootprint of RNAP may have the capacity tochange (6, 61). All this leaves open the ques-
tion of the actual DNA step size during elon-gation.
Single-molecule techniques have success-fully measured the nanometer-scale step sizesof many motor proteins (86–90), but theAngstrom-scale step sizes expected for nucleic
Backtracking
ρ-Binding
MisincorporationElemental
pause
Termination hairpin formation
Termination
Hairpinpause
Pyrophosphorolysis
NTP addition NTP addition
Pyrophosphorolysis
PausingElongationTermination
162 Herbert · Greenleaf · Block
Ann
u. R
ev. B
ioch
em. 2
008.
77:1
49-1
76. D
ownl
oade
d fr
om a
rjou
rnal
s.an
nual
revi
ews.
org
by Y
ale
Uni
vers
ity S
TE
RL
ING
CH
EM
IST
RY
LIB
RA
RY
on
06/3
0/08
. For
per
sona
l use
onl
y.
ANRV345-BI77-08 ARI 28 April 2008 11:12
acid-based motors were experimentally inac-cessible until recently. The construction of anultrastable OT with Angstrom resolution al-lowed Block and coworkers (7, 91) to followindividual, actively elongating TECs with un-precedented precision. Records of transcrip-tional elongation obtained under conditionsof low nucleotide concentration and moder-ate loads (2.5–10 μM NTPs; 18 pN assistingforce) displayed clear steps of varying dura-tion averaging 3.7 ± 0.6 A in length, which isnearly the rise per base of double-stranded, B-form DNA (3.4 ± 0.5 A) (92). Larger steps,consisting of small integral multiples of thefundamental spacing, were also observed. Thelarger steps were statistically attributable tothe temporal resolution of the assay and anal-ysis, which fails to detect brief events belowthe integration time used for measurement.The observation of single-base stepping is in-consistent with either an obligate scrunch-ing or an inchworming model and supportsthe original concept of a rigid-body, sliding-clamp mechanism, where RNAP translocatesin single base pair increments that are tightlycoupled to nucleotide addition (7).
Kinetics: heterogeneity and state switch-ing. The rates at which genes are tran-scribed play an important regulatory role byfreeing the polymerase molecule to under-take additional rounds of transcription or,
conversely, by causing it to remain occu-pied. Single-molecule measurements of elon-gation dynamics offer insights into suchrate-based regulatory mechanisms (93). Thevery first single-molecule transcription exper-iments employed the TPM assay (Figure 3a)to measure rates of elongation by E. coli RNAPmolecules (32, 70). Measured speeds weregenerally consistent with the transcriptionrates reported in bulk studies. Interestingly,however, although the average speeds of in-dividual transcribing molecules did not varysubstantially with time, these speeds did varysignificantly from molecule to molecule (70).The time resolution of the TPM assay made itdifficult to determine if the heterogeneity wasdue to intrinsic differences in the on-pathwayelongation rates of molecules or to differentpropensities to enter into off-pathway, pausedstates. High-resolution optical trapping stud-ies subsequently permitted a more accurateseparation of active elongation from pausingin E. coli RNAP (10, 12). These studies as wellas measurements of T7 RNAP (9, 94), wherepausing is rarely observed, corroborated theoriginal observation of molecular heterogene-ity in the overall rates of transcription and de-termined that the variance in molecular rateswas largely attributable to on-pathway differ-ences in speeds.
Arguably, results from bulk steady-stateand presteady-state kinetic studies may also
←−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−Figure 7Transcription elongation pathway and a subset of off-pathway states showing RNA polymerase ( green),the template strand (light blue), the nontemplate strand (dark blue), RNA strand (red ), and ρ-factor( purple). Elongation (central panel ) corresponds to the template-directed condensation of nucleosidetriphosphates (NTPs, yellow) onto the 3′ end of the growing RNA chain, along with the release ofinorganic pyrophosphate. Individual nucleotides may occasionally be excised via pyrophosphorolysis. Anumber of paused states branch off the central elongation pathway (upper panel ). An elemental pause statehas been proposed as a common intermediate state preceding hairpin-stabilized and backtracking pauses(solid arrows), although both these states might be reached directly from the main elongation pathway(dashed arrows) (97). Misincorporation-induced pauses are triggered when a mismatched NTP ( yellow) isadded to the RNA chain; backtracking often results in such cases (34, 95). Two paths lead totranscriptional termination (lower panel ). Intrinsic termination occurs when RNAP transcribes specificsequence elements that code for a termination hairpin in the RNA followed by a U-rich tract, triggeringdissociation of the TEC. Another pathway to termination involves the binding of ρ, which is thought tomove along the transcript until reaching RNAP and ultimately dislodging the RNA from the enzyme(114).
www.annualreviews.org • Single-Molecule Studies of RNA Polymerase 163
Ann
u. R
ev. B
ioch
em. 2
008.
77:1
49-1
76. D
ownl
oade
d fr
om a
rjou
rnal
s.an
nual
revi
ews.
org
by Y
ale
Uni
vers
ity S
TE
RL
ING
CH
EM
IST
RY
LIB
RA
RY
on
06/3
0/08
. For
per
sona
l use
onl
y.
ANRV345-BI77-08 ARI 28 April 2008 11:12
be taken as evidence of heterogeneous kinet-ics (95, 96). However, rather than assuminga molecular population with a distribution ofintrinsic elongation rates, the data were in-terpreted in terms of a model in which mem-bers of an otherwise homogeneous populationswitched between two distinct elongationrates on a slow timescale compared to nu-cleotide addition. Whereas regulator bind-ing is known to switch RNAP into different,persistent states (97, 98), the phenomenon ofspontaneous switching behavior is disputed(99). One early single-molecule study, whichwas conducted at comparatively low spatialand temporal resolution, reported that wild-type RNAP can spontaneously switch veloc-ity states (100). However, subsequent stud-ies have failed to confirm any such switching(12, 13, 101), casting doubt on the finding. Todate, only one other observation of velocity-state switching has been reported, in this casefor rpoB8, a mutant RNAP bearing a pointmutation in the β-subunit (but not for wildtype) (13). The mechanistic basis of this par-ticular mutation is unclear, however, it poten-tially affects a specific contact with the nascentmRNA that is located more then 20 A fromthe enzyme’s active site.
Careful control experiments performed byGelles and coworkers (101) have ruled outa number of trivial explanations for the ap-parent intermolecular heterogeneity, such asthe effect of temperature, the solute concen-tration, and the immobilization technique.This suggests that the source of heterogene-ity may be structural in origin, perhaps owingto minor defects in RNAP caused by trans-lation errors, posttranslational modificationsof RNAP, or different conformers in RNAPfolding (101). Supporting the structural ori-gin of heterogeneity, the variance in FRETdistances between the RNA 5′ end and a la-beled base located on the template DNA ina TEC was significantly larger than the cor-responding variance in a control sample withthe identical dyes placed on a DNA molecule.In individual TECs, these FRET levels per-sisted for more than 10 min (24). Analo-
gous FRET measurements with active com-plexes initiated from natural promoters mayhelp establish whether any such heterogene-ity persists during elongation. Simultane-ous FRET and elongation-rate measurements(performed with TPM or OTing assays)may provide evidence for a correlation be-tween structural and velocity heterogeneity.
Chemomechanics: translocation mecha-nism and stall. Recent single-molecule workhas helped characterize the molecular mech-anism of RNAP translocation. Two differentmodels have been proposed. In the “powerstroke” model, a conformational change inthe enzyme generates translocation throughthe direct coupling of displacement to NTPhydrolysis and subsequent PPi release. In the“Brownian ratchet” model, random thermalfluctuations between the pre- and posttranslo-cated states of the enzyme are mechanicallyrectified by NTP binding and hydrolysis,leading to unidirectional motion. Experimen-tal (80, 102) and theoretical (103, 104) evi-dence has often been interpreted in terms ofa ratchet-like mechanism, although other in-terpretations certainly cannot be ruled out.The power stroke model was inspired bycrystal structures of T7 RNAP obtained innominally pre- and posttranslocated states. Inthat model, translocation is tightly coupledto PPi release through a structural changethat constitutes the power stroke (105). Op-tical trapping techniques, wherein force canbe used as a control parameter to modulatestepping rates, are well suited for differentiat-ing between these particular models becausethey supply quantitatively different predic-tions about the relationship between translo-cation rates and applied loads.
The elongation velocity, v(F), is expectedto fit a Boltzmann-type model, which returnsa distance parameter, δ, representing the ef-fective distance over which an applied force,F, acts to slow translocation (where �G = F·δsupplies the associated mechanical energy).For the general case of a Brownian ratchetmechanism, this parameter corresponds to the
164 Herbert · Greenleaf · Block
Ann
u. R
ev. B
ioch
em. 2
008.
77:1
49-1
76. D
ownl
oade
d fr
om a
rjou
rnal
s.an
nual
revi
ews.
org
by Y
ale
Uni
vers
ity S
TE
RL
ING
CH
EM
IST
RY
LIB
RA
RY
on
06/3
0/08
. For
per
sona
l use
onl
y.
ANRV345-BI77-08 ARI 28 April 2008 11:12
distance over which the enzyme fluctuatesduring the stepping cycle between transloca-tion states. For the specific case of an RNAPratchet, the distance between pre- and posttranslocated states subtends exactly one basepair. For a power stroke mechanism, however,the distance parameter corresponds to the dis-placement of the enzyme in moving from thestart of its cycle to a transition state located in-termediate between the pre- and posttranslo-cated positions, which is necessarily less thanone base pair. In addition to predicting dif-ferent values for the distance parameter, theBrownian ratchet and power stroke modelsalso predict significant differences in the forcedependence of transcriptional velocity over arange of nucleotide concentrations. Becausethe NTP-binding event is coupled to translo-cation in the ratchet model, velocity becomesmost sensitive to applied loads when NTPconcentrations are very low, which causesNTP binding to become rate limiting. In thisregime, force acts like a competitive inhibitorto NTP binding. Conversely, in the powerstroke model, NTP binding tends to be de-coupled from translocation, because it is sep-arated by one or more biochemical transitionsthat are very nearly irreversible, such as nu-cleotide condensation followed by PPi release(where the latter is presumed to be responsiblefor the power stroke itself ). The presence ofthese intervening biochemical steps thereforetends to make velocities largely independentof load at the lowest NTP concentrations andmost sensitive when concentrations are high.Thus, the Brownian ratchet and power strokemechanisms have diametrically opposite ef-fects on force-velocity curves as the NTP con-centrations are varied.
Accurate measurements of force-velocityrelationships for RNAP are made significantlymore challenging by the presence of off-pathway events, such as the entry into back-tracking and arrested states, which also ex-hibit a load dependence (34, 100, 106). Thiscomplication is evident in the variety of stallforces reported for E. coli RNAP, which rangefrom 14 to 25 pN, depending on how fast
load was applied (33, 69). Heterogeneity in thestall force from molecule to molecule has alsobeen observed, suggesting that polymerasesmay stall at different locations on the DNAtemplate, corresponding to different under-lying sequences. Some DNA sequences areprone to enzyme backtracking and arrest, andtherefore, this variability may be responsiblefor the different apparent stall forces (12, 69).In addition, a eukaryotic RNAP (yeast Pol II)reportedly stalled at a comparatively low forceof 7.5 pN. However, Pol II elongated success-fully against forces significantly beyond thisstall force for brief periods of time. Further-more, the apparent stall force was doubledwith the addition of elongation factor TFIIS(107). All together, the results from prokary-otic and eukaryotic RNAP suggest that stalloccurs not when translocation forces are ex-actly balanced by the application of an exter-nal load but rather when the probability ofencountering a backtracking-prone sequencebecomes significant, leading to enzyme inac-tivation. This makes measuring the true stallforce of RNAP considerably more challeng-ing than for the conventional motor proteins,such as kinesin and myosin, and highly depen-dent on the method of data collection and itsinterpretation.
The first measurements of the force-velocity relationships for E. coli and yeastRNAP were obtained by rapidly increasing ahindering load in such a way that RNAP tran-scribed only a short distance before stalling.Force-velocity characteristics obtained in thisfashion were conducted under conditions ofsaturating NTP concentrations, and the is-sue of RNAP heterogeneity was addressed bynormalizing each record in both force andvelocity before ensemble averaging, so thatindividual records could be combined (12,69, 107). However, these normalization pro-cedures may have obscured the underlyingforce-velocity relationships somewhat, mak-ing the velocity appear rather insensitive toload until stall was reached, and at this pointit dropped precipitously with additional force(8). This insensitivity to external load under
www.annualreviews.org • Single-Molecule Studies of RNA Polymerase 165
Ann
u. R
ev. B
ioch
em. 2
008.
77:1
49-1
76. D
ownl
oade
d fr
om a
rjou
rnal
s.an
nual
revi
ews.
org
by Y
ale
Uni
vers
ity S
TE
RL
ING
CH
EM
IST
RY
LIB
RA
RY
on
06/3
0/08
. For
per
sona
l use
onl
y.
ANRV345-BI77-08 ARI 28 April 2008 11:12
saturating NTP concentrations implies thatforce does not significantly affect overalltranslocation rates until other (off-pathway)processes intervene, such as backtracking oran irreversible stall. For saturating nucleotideconcentrations, a force-insensitive velocity isbroadly consistent with a Brownian ratchet-type model where translocation is coupled toNTP binding, as discussed above.
Recent experiments, however, on T7RNAP and E. coli RNAP, provide evidencefor a greater load sensitivity in the elongationvelocity. For T7 RNAP, under conditionsof limiting NTP concentrations, hinderingloads reduced transcription rates, consistentwith the load acting as a competitive in-hibitor of NTP binding (9). Velocity wasnot load dependent, however, under satu-rating NTP concentrations. These findingswere interpreted in terms of a Brownianratchet model similar to one first proposedby Guajardo & Sousa (104), wherein RNAPmolecules fluctuate between pre- and post-translocated states until NTP binding locksthese into the posttranslocated state. In ex-periments on E. coli RNAP, Block and cowork-ers (7) used a high-resolution, passive opticalforce clamp to measure an entire ensembleof force-velocity curves, varying the NTPconcentration over more than two orders ofmagnitude. The improved spatial resolutionachieved using this approach allowed poten-tially confounding backtracking events (whichare off-pathway) to be identified in individualrecords and removed from further analysis,thereby isolating the force dependence of the(on-pathway) translocation events. The en-semble of experimental force-velocity curveswere globally fit to a Boltzmann-type rela-tion, which returned a distance parameter of1 bp. Furthermore, elongation velocity wasmore sensitive to force at low NTP concen-trations (7). A separate experiment conductedby Wang and coworkers (8) also returned adistance parameter of 1 bp. Taken all together,these findings lend support to the notion thatRNAP moves by means of a Brownian ratchetmechanism.
The angular motion of RNAP was alsoprobed using an externally applied torque(27). Magnetic beads, decorated with smallfluorescent beads used to track the rela-tive angular motion of RNAP and DNA(Figure 3f ), displayed rotations of ∼8.7 ±3.7 bp/revolution during elongation. The ob-served rate was within the error of the ro-tational speed expected if RNAP tracked theDNA helix, which has 10.4 bp per turn. In ad-dition, RNAP stalled under external torquesgreater then 5 pN nm. Newly developed tech-nology, such as the optical torque wrench (52),which is capable of exerting both forces andtorques in single-molecule experiments, mayeventually allow the simultaneous acquisitionof torque-velocity and force-velocity curves.
Off-Pathway Events
The process of active, on-pathway elonga-tion is frequently interupted by entry intooff-pathway states that can be important forthe regulation of RNA synthesis. By avoid-ing the ensemble averaging inherent in tradi-tional biochemistry, single-molecule methodsallow for the direct observation of these asyn-chronous states and have thereby led to animproved understanding of their origin.
Transcriptional pausing. Transcriptionalinitiation has long been identified as a criticalpoint of regulation, but mechanisms forcontrolling expression levels during theelongation phase have received compara-tively little attention until recently (108).Transcriptional pausing can not only reducerates of mRNA production, but also recruitregulatory factors to the TEC that modifysubsequent transcription (109–112), functionas a precursor to transcriptional arrest andtermination (113, 114), help synchronizetranscription and translation in prokary-otes (115), or lead to messenger splicingor polyadenylation in eukaryotes (116,117). The long-lived, “stabilized pauses”that are known to play a regulatory roleare often associated with the formation of
166 Herbert · Greenleaf · Block
Ann
u. R
ev. B
ioch
em. 2
008.
77:1
49-1
76. D
ownl
oade
d fr
om a
rjou
rnal
s.an
nual
revi
ews.
org
by Y
ale
Uni
vers
ity S
TE
RL
ING
CH
EM
IST
RY
LIB
RA
RY
on
06/3
0/08
. For
per
sona
l use
onl
y.
ANRV345-BI77-08 ARI 28 April 2008 11:12
RNA hairpins in the transcript (which arethought to allosterically inactivate RNAP)or with the formation of energetically weakRNA:DNA hybrids (which are thought toinduce backtracking) (118).
High-resolution, single-molecule studiesof transcriptional pausing complement tradi-tional biochemical studies, helping to over-come some of their intrinsic limitations. Bio-chemical measurements of pausing typicallyemploy gel-based assays that score the pres-ence or absence of bands generated by RNAsproduced by an ensemble of transcribingcomplexes. To form a sharp band, an ini-tially synchronized population of transcribingmolecules must simultaneously encounter asignal on the DNA template and pause therefor long enough to be detected. To produceadequate signal levels, the NTP concentra-tions used in gel-based assays are often re-duced to nonphysiological levels to slow tran-scription. Furthermore, once a synchronizedpopulation encounters the initial pause loca-tion, the stochastic duration of the pause life-time results in molecules leaving this posi-tion over a distribution of times, thereafterdesynchronizing the population and reducingthe sharpness of all succeeding bands. Thisdesynchronization makes it difficult to studypausing over significant distances along theDNA template. Single-molecule methods, bycontrast, are not subject to desynchroniza-tion (which is an ensemble property), and theycurrently allow for the detection of pauses asshort as ∼1 s and separated by as little as ∼2 bp(10) at physiological concentrations of NTPs.In addition, OT-based methods allow pausingto be probed as a function of the force appliedto the DNA or the RNA. Pausing that involvesany longitudinal motion of polymerase alongthe template or transcript is generally a strongfunction of the applied load, and therefore,such measurements can supply additional in-sights into the mechanism.
Early on, numerous brief transcriptionalpauses were noted in records of RNAPtranscription obtained using OTs (69). Suchevents, which would eventually come to be
Stabilized pause:long-lived pausesresulting fromformation of ahairpin in thenascent RNA,backtracking ofpolymerase, orinteractions oftranscriptionalregulators
Elemental pause: aclass of short-livedpause that is believedto be a precursor tolonger-livedstabilized pausestates
known as “ubiquitous” pauses, occur evenin regions of the template DNA previouslythought devoid of regulatory pauses (basedon biochemical assays), such as the E. colirpoB gene. These ubiquitous pauses, mak-ing up approximately 95% of all detectedpauses, have lifetimes <25 s and occur ev-ery ∼100 bp, on average, along the tem-plate (12, 13). Limitations in the spatial res-olution of earlier work made it impossibleto determine whether ubiquitous pauses oc-curred stochastically, independent of the un-derlying sequence, or were instead triggeredin a sequence-dependent fashion by codingelements located at frequent, apparently ran-dom intervals. Two recent single-moleculestudies succeeded in improving the preci-sion of assays to a point where displacementrecords could be correlated with the underly-ing DNA sequences with base pair (or near-base pair) resolution. In brief, these studiesrelied upon imbedded “fiducial marks” (reg-istration points), which were used to align in-dividual records and supply improved accu-racy in the absolute position on the basis ofeither the release points of RNAP at the endsof the template (119) or on the transcrip-tional behavior of RNAP moving on repet-itive (concatamer) templates carrying char-acterized pause sequences (10). Using thelatter method, Herbert et. al. (10) determinedpause positions within a base pair over nearly2000 bp of overall transcription and con-cluded that ubiquitous pauses were inducedby specific commonly occurring sequences.
How, then, do the ubiquitous pauses ofsingle-molecule assays, which are brief andsequence specific, relate to the longer-livedpauses identified in biochemical assays, manyof which are regulatory? Bulk studies foundthat short-lifetime pauses still persist afterRNA hairpin formation or DNA backtrackingare suppressed, suggesting that such eventsmay stabilize and thereby prolong preexist-ing, but weaker, pauses. This observation ledto the proposal that long-lived pauses are pre-ceded by a common, elemental pause that canbe further stabilized (97). The existence of
www.annualreviews.org • Single-Molecule Studies of RNA Polymerase 167
Ann
u. R
ev. B
ioch
em. 2
008.
77:1
49-1
76. D
ownl
oade
d fr
om a
rjou
rnal
s.an
nual
revi
ews.
org
by Y
ale
Uni
vers
ity S
TE
RL
ING
CH
EM
IST
RY
LIB
RA
RY
on
06/3
0/08
. For
per
sona
l use
onl
y.
ANRV345-BI77-08 ARI 28 April 2008 11:12
an elemental pause (inactivated) state gainedadditional support from cross-linking studies(120), and it has been invoked to explain thekinetics of misincorporation and nucleotideaddition (95, 96, 121). In contrast to stabi-lized pauses, which involve either large-scalebacktracking motions or postulated allostericmodulation of the enzyme by RNA hairpins,an elemental pause likely requires only a verysmall structural isomerization affecting theactive site (97, 118, 120).
Because rates of entry into the presump-tive elemental paused state, determined bio-chemically, are similar to corresponding ratesof entry into the ubiquitous pause state, iden-tified in single-molecule experiments, it wasconjectured that ubiquitous pauses representthe elemental pause state (12). Consistent withthis hypothesis, ubiquitous pause states lie offthe main elongation pathway and are inducedby sequences generally similar to those foundin biochemically characterized hairpin andbacktracking pauses (10). Furthermore, theduration and frequency of ubiquitous pausesare largely independent of the applied load,implying that such pauses did not involvetranslocations of polymerase along the RNAor DNA (11, 12). Finally, ubiquitous pausesoccurred in single-molecule records where ex-ternal loads were applied to the transcript suf-ficient to remove any secondary structure, in-dicating that RNA hairpin formation was notresponsible for such events (11).
In contrast to studies of E. coli RNAP, a re-cent single-molecule study of transcription inyeast Pol II concluded that ubiquitous paus-ing in eukaryotic polymerase resulted mainlyfrom backtracking (107). No backtracking atubiquitous pause sites was actually observedin the single-molecule records from these ex-periments, however, owing to limitations inspatial resolution. Instead, backtracking wasinferred by modeling of the pause lifetime dis-tribution, which was fit to a t−3/2 power law. Apower-law relationship of this type is expected(at long times) for the first passage time to abarrier by a continuum random-walk process,
such as diffusion. Pausing was therefore mod-eled as inactivation of the enzyme induced bybacktracking, followed by diffusional returnto an active state, where the 3′ end of the RNAis realigned with the active site (107). A back-tracking model for pausing is tantalizing, butthe pause lifetime distributions acquired forboth bacterial and yeast RNAP appear to becomplex and composed of multiple compo-nents. For E. coli RNAP, the lifetime distribu-tions of short pause events were previously fitby a sum of two exponentials rather than bya power law (10–12, 34). Longer pause life-times, such as those arising from misincorpo-ration events, constitute a third componentin the tail of the overall distribution that issensitive to load (34) (see below). Because in-dividual pauses may have different character-istic lifetimes, pooling all pauses observed ona given template in a global distribution re-sults in a superposition of multiple exponen-tials, generating a composite curve in whichthe shortest and longest lifetimes predomi-nate (10). Such a relationship might also givethe appearance of a power law. It may there-fore be difficult to draw definitive conclusionsbased solely on models of global lifetime dis-tributions. In any event, considerable atten-tion needs to be paid to details of the analysisprocedure and any estimates of error, as wellas to alternative models. Assuming that back-tracking is responsible for ubiquitous pausingin Pol II, pause lifetimes should be signifi-cantly affected by external loads.
One previous study of E. coli RNAP re-ported load dependence for a particular pausesite (�tR2), which displayed a significantincrease in transcriptional dwell time withincreasing (hindering) load, implying back-tracking (119). However, because these exper-iments were conducted at low uridine triphos-phate (UTP) levels, the possibility remainsthat the observed load dependence may reflecta force-dependent decrease in elongation rateat this site produced by its sequence, whichrequires the addition of a number of uri-dine bases (119). It seems clear that our
168 Herbert · Greenleaf · Block
Ann
u. R
ev. B
ioch
em. 2
008.
77:1
49-1
76. D
ownl
oade
d fr
om a
rjou
rnal
s.an
nual
revi
ews.
org
by Y
ale
Uni
vers
ity S
TE
RL
ING
CH
EM
IST
RY
LIB
RA
RY
on
06/3
0/08
. For
per
sona
l use
onl
y.
ANRV345-BI77-08 ARI 28 April 2008 11:12
understanding of backtracking for both thebacterial and eukaryotic forms of polymerasescould be improved by the characterization ofpausing behavior at specific sites for moleculessubjected to various loads. Such studies mayplace models of pausing mechanisms onfirmer ground.
The ability of single-molecule experi-ments to discern subtle changes in pausingand elongation kinetics has led to a vari-ety of other interesting results. Experimentson rpoB8, a point mutant of E. coli RNAP,showed that it elongates more slowly, andbecause pausing is an off-pathway state thatcompetes with elongation, it should thereforepause more frequently per unit distance trav-eled (13). Another study that modulated tem-perature found that the elongation rate forE. coli RNAP increased by ∼1.5 bp/s per ◦Cfor small changes in temperature. However,neither the pause frequency nor the pauselifetime varied with temperature, suggestingthat elongation has a large enthalpic contribu-tion, whereas pausing is dominated by entropy(122). In another study, the effect of microcinJ25, an antibiotic known to bind RNAP anddecrease overall transcription, was shown notto affect the active elongation rate. Instead,the frequency of pausing was significantly en-hanced. These data, taken together with re-sults from cross-linking studies showing thatmicrocin J25 occupies the NTP entry chan-nel, suggest that microcin J25 acts to inhibittranscription by blocking access of NTPs tothe active site (123). Finally, using an ultra-stable assay with base pair resolution and scor-ing the positions of pauses induced by limitinga single NTP species, the DNA sequence ofthe template was reconstructed from the mo-tions of as few as four RNAP molecules (35).
Proofreading. Two single-molecule studiesconducted at comparatively low temporal res-olution studied the effects of load on tran-scription (100, 106). The first such study, re-stricted to pause events lasting ∼15 s or longerunder hindering loads, concluded that the
propensity to arrest was force dependent, butnot the propensity to pause (100). The sec-ond study, which employed both hinderingand assisting loads, reported that both pausingand arrest were force dependent (106). Sub-sequent high-resolution measurements sup-plied direct evidence for load-induced paus-ing. In averaged records of long (but notshort-lifetime) pauses, enzymes subjected tomoderate hindering loads were found to back-track (34). The density of pauses lasting 20 s ormore (∼1 per kilobase) corresponds closely tomeasured rates of base misincorporation dur-ing RNA synthesis in vitro (124), suggestingproofreading as the likely explanation for suchevents.
The prevailing model for proofreading byRNAP, based on both structural and biochem-ical data, invokes the backtracking of RNAPalong the DNA template in response to a mis-incorporation event. This backtracking is fol-lowed by endonucleolytic cleavage of the 3′
RNA fragment carrying the error, which canbe promoted by transcription factors GreAand GreB in prokaryotes or by TFIIS in eu-karyotes (96, 125, 126). Single-molecule ex-periments provide compelling support for thismodel, showing that the frequency of longpauses increases in the presence of the nu-cleotide analog inosine triphosphate (ITP),which mimics misincorporation, and that longpauses lead to enzyme backtracking by an av-erage of ∼5 bp. Addition of the transcriptionfactors GreA or GreB can relieve long pausesinduced by ITP in single-molecule assays (34).Single-molecule studies complement the bio-chemical picture developed for proofread-ing, providing a real-time window into thisprocess during active elongation at saturat-ing nucleotide levels, rather than in stalledcomplexes under subsaturating conditionsthat are forced to incorporate an incorrectnucleotide.
Enzyme “memory” and heterogeneity.Single-molecule experiments have shown thattranscriptional elongation rates do not tend
www.annualreviews.org • Single-Molecule Studies of RNA Polymerase 169
Ann
u. R
ev. B
ioch
em. 2
008.
77:1
49-1
76. D
ownl
oade
d fr
om a
rjou
rnal
s.an
nual
revi
ews.
org
by Y
ale
Uni
vers
ity S
TE
RL
ING
CH
EM
IST
RY
LIB
RA
RY
on
06/3
0/08
. For
per
sona
l use
onl
y.
ANRV345-BI77-08 ARI 28 April 2008 11:12
to change in any systematic way upon re-covery from a pause (12, 13). However, thepossibility that individual RNAP moleculesmight retain a kind of “intramolecular mem-ory,” whereby pausing at an upstream siteaffects the propensity of RNAP to pausethereafter downstream, remains an openquestion (99, 127). The binding of transcrip-tion regulators, such as RfaH, causes RNAPto respond differently to downstream pausesignals (109). Could an analogous effect ex-ist in the absence of bound transcription fac-tors, triggered instead by DNA sequence in-formation? Optical trapping experiments onDNA templates containing eight repeats ofthe identical sequence allow this and sim-ilar questions about molecular memory tobe examined quantitatively. On repetitivetemplates, the pause probability at a given se-quence site was correlated to pause probabil-ities at subsequent sites, implying that indi-vidual molecules exist in heterogeneous stateswith greater or lesser propensities to pause.However, the degree of correlation did notdecay with the distance between pause sites,suggesting that molecular pause propensitieswere stable on the timescale of the experimentand were not due to a transient memory effect.The pause correlation analysis was restrictedto pause sites that did not display any correla-tion with elongation velocity, suggesting thatthe molecular heterogeneity in pause propen-sity may be caused by a different mechanismthan the previously observed heterogeneity invelocity states (10). Further experiments arerequired to identify the source of this hetero-geneity.
TERMINATION
The TEC is extremely stable, but ultimatelyRNAP must dissociate accurately in responseto termination signals, releasing the tran-script and the DNA template. In prokary-otes, intrinsic termination occurs at specificsequence elements that code for a stablehairpin in the nascent RNA followed by a
U-rich tract, which together are thought togenerate an unstable RNA:DNA hybrid in theenzyme. Termination in prokaryotes also oc-curs via a different mechanism involving thetermination factor ρ, which can translocatealong RNA until it encounters the polymerase(114), leading to release of the nascent chain(Figure 7). In general, termination might beproduced indirectly through allosteric inter-actions between the RNA hairpin (or ρ-factor)and RNAP that signal the TEC to release itssubstrates (128, 129). Alternatively, termina-tion might be produced directly by forces ex-erted during folding of the terminator hair-pin (or by ρ-factor), which push the enzymeforward in the absence of RNA synthesis, sothat the hybrid is shortened and the TECbecomes mechanically destabilized (98, 130,131). Because up to 30 pN of tension can beapplied to the RNA without causing the re-lease of a transcript, the ρ-factor must exert atleast this much force if the latter mechanism isresponsible for ρ-based termination. For thecase of intrinsic termination, the force gen-erated by a terminator hairpin during fold-ing is not thought sufficient to release RNAat most sites along the DNA (11). There-fore, the hybrid-destabilizing effect of theU-rich tract, possibly aided by other mech-anisms, such as hairpin stem invasion (73), al-lostery (128), or forward translocation (131),must play some role in the energetics of in-trinsic termination. Future single-moleculeexperiments should be able to probe theseenergetics.
Previous biochemical experiments havesuggested that termination is an off-pathwaystate that competes with on-pathway elonga-tion because the termination efficiency can beincreased by slowing the rate of elongation(132). However, controversy arises concern-ing the pathway involved in intrinsic termina-tion. Some studies conclude that terminationis preceded by an intermediate elongation-incompetent state (133, 134), whereas oth-ers find that termination occurs quickly, withno stable intermediate (on a timescale of
170 Herbert · Greenleaf · Block
Ann
u. R
ev. B
ioch
em. 2
008.
77:1
49-1
76. D
ownl
oade
d fr
om a
rjou
rnal
s.an
nual
revi
ews.
org
by Y
ale
Uni
vers
ity S
TE
RL
ING
CH
EM
IST
RY
LIB
RA
RY
on
06/3
0/08
. For
per
sona
l use
onl
y.
ANRV345-BI77-08 ARI 28 April 2008 11:12
seconds) (135). Using TPM (Figure 3a),Gelles and coworkers (14) found that RNAPpauses with a ∼1 min characteristic lifetimebefore releasing DNA at the his terminator.A corresponding pause was not observed forpolymerases that read through the terminatoror transcribed a template lacking the termina-tion site, suggesting that termination may beirreversibly preceded by an inactivated inter-mediate state that is committed to termina-tion. In these experiments, the surface immo-bilization of RNAP is thought to significantlyreduce the rate at which DNA can diffuseaway from the enzyme, thereby slowing dis-sociation of the TEC and facilitating the ob-servation of a paused intermediate state. Thissame reduced diffusion rate might, however,also promote nonspecific rebinding to a re-leased DNA template, thereby generating afalse signal (135).
In principle, single-molecule fluorescenceexperiments could be designed to pinpoint ifRNA transcripts are released from the TECbefore, after, or synchronously with releaseof the DNA and if the template or the tran-script are both present during the proposedcommitted intermediate state. Optical trap-ping experiments where controlled loads areapplied to the DNA or RNA during termina-tion should be able to distinguish whether theforward translocation of RNAP is requiredand whether forces applied to the RNA areable to dissociate TECs with weak RNA:DNAhybrids.
CONCLUSION
Seventeen years have elapsed since the pio-neering single-molecule assay for RNAP tran-scription was developed by Schafer et al. (70),and enormous progress has been achievedduring this period. Many of the questions thatwere previously identified as uniquely wellsuited to the use of single-molecule meth-ods have already been addressed (3). Single-molecule experiments have shown that RNAPadvances by one base at a time along DNA,
with translocation tightly coupled to RNAsynthesis, and that it likely operates by a Brow-nian ratchet-type mechanism (7–9). Single-molecule experiments have identified andcharacterized long-lived heterogeneities inRNAP conformations (24), elongation rates(12, 13, 101), and pause propensities (10).Other single-molecule studies have found thattranscriptional initiation involves scrunchingof the DNA template until contacts withthe promoter are released (5, 6) and sup-plied evidence for an elongation-incompetentintermediate state preceding transcriptionaltermination (14). Steady improvements insingle-molecule assays have allowed the ef-fects of transcriptional cofactors and effectorsto be studied as well, including σ-factor (18),Gre A and GreB (34), ppGpp (50), and mi-crocin J25 (123).
Progress in single-molecule work on tran-scription has also raised deeper questions andcreated new avenues of potential research.Results from new assays for eukaryotic poly-merases have revealed large differences inmechanical stability between prokaryotic andeukaryotic polymerases, raising intriguingquestions with regard to functional differ-ences between these enzymes (107). The ad-vent of base pair resolution in single-moleculeassays (7) opens the possibility of probing en-zyme reaction rates at individual sequencesites. Advanced techniques, such as multi-color, single-molecule FRET (61), should alsomake it possible to observe directly the as-sembly of large macromolecular complexes,such as Mediator, which serves as a coacti-vator of Pol II transcription, and may ulti-mately provide insights into the coordinationof RNA splicing and transcription. The re-cent development of assays that exert con-trolled forces on the nascent RNA (11) willlikely be important in eventually establishingthe mechanism of transcriptional terminationand also in determining how elongation ki-netics can affect the structures of cotranscrip-tionally folded RNAs. Single-molecule workon RNAP is really motoring along!
www.annualreviews.org • Single-Molecule Studies of RNA Polymerase 171
Ann
u. R
ev. B
ioch
em. 2
008.
77:1
49-1
76. D
ownl
oade
d fr
om a
rjou
rnal
s.an
nual
revi
ews.
org
by Y
ale
Uni
vers
ity S
TE
RL
ING
CH
EM
IST
RY
LIB
RA
RY
on
06/3
0/08
. For
per
sona
l use
onl
y.
ANRV345-BI77-08 ARI 28 April 2008 11:12
SUMMARY POINTS
1. RNAP can slide along the DNA template to search for promoter sites.
2. Some fraction of RNAP molecules retain σ-factor upon the transition from the OPCto TEC in vitro.
3. Transcription initiation involves scrunching of the template DNA within the enzyme.
4. Translocation occurs in single-base increments, consistent with a tight coupling be-tween the length of the RNA transcript and the position of the RNAP on the DNAtemplate.
5. The force and nucleotide concentration dependence of transcription velocity is mostconsistent with a Brownian ratchet model for translocation.
6. TFIIS, a eukaryotic transcription accessory factor, modulates the stall force.
7. Ubiquitous, short-lifetime pauses interrupt transcription by E. coli RNAP, even undersaturating nucleotide concentrations. Ubiquitous pauses are sequence dependent andindependent of the applied load, and they may represent an elemental pause statefrom which stabilized, regulatory pauses are derived.
8. Heterogeneity with respect to elongation rates and the propensity to enter the pausedstate has been observed in populations of molecules.
FUTURE ISSUES
1. Does σ-factor remain bound to the TEC, even in vivo?
2. During abortive initiation, where does the scrunched DNA reside within RNAP?
3. Are short-lifetime pauses caused by backtracking, small conformational rearrange-ments, or something else? Is pausing caused by the same mechanism at all sites? Arethese mechanisms the same in both prokaryotes and eukaryotes?
4. What is responsible for intermolecular heterogeneity, and are changes in the elonga-tion rate correlated to structural changes?
5. What are the mechanisms by which accessory factors, such as GreA, GreB, NusA,NusG, λQ, and N, affect transcription?
6. How do specific sequence elements in the transcribed DNA sequence affect the ratesof next nucleotide addition to RNA?
7. How does torque affect transcriptional elongation and termination processes?
8. How is transcriptional termination modulated by force applied to either the DNAtemplate or the RNA transcript?
DISCLOSURE STATEMENT
The authors are not aware of any biases that might be perceived as affecting the objectivity ofthis review.
172 Herbert · Greenleaf · Block
Ann
u. R
ev. B
ioch
em. 2
008.
77:1
49-1
76. D
ownl
oade
d fr
om a
rjou
rnal
s.an
nual
revi
ews.
org
by Y
ale
Uni
vers
ity S
TE
RL
ING
CH
EM
IST
RY
LIB
RA
RY
on
06/3
0/08
. For
per
sona
l use
onl
y.
ANRV345-BI77-08 ARI 28 April 2008 11:12
ACKNOWLEDGMENTS
The authors thank Megan T. Valentine and Kirsten L. Frieda for comments on the manuscript.Funding for our work was provided by the N.I.G.M.S.
LITERATURE CITED
1. Hsu LM. 2002. Biochim. Biophys. Acta 1577:191–2072. Greive SJ, von Hippel PH. 2005. Nat. Rev. Mol. Cell Biol. 6:221–323. Gelles J, Landick R. 1998. Cell 93:13–164. Nudler E, Gottesman ME. 2002. Genes Cells 7:755–685. Revyakin A, Liu C, Ebright RH, Strick TR. 2006. Science 314:1139–436. Kapanidis AN, Margeat E, Ho SO, Kortkhonjia E, Weiss S, Ebright RH. 2006. Science
12. Neuman KC, Abbondanzieri EA, Landick R, Gelles J, Block SM. 2003. Cell 115:437–4713. Adelman K, La Porta A, Santangelo TJ, Lis JT, Roberts JW, Wang MD. 2002. Proc. Natl.
Acad. Sci. USA 99:13538–4314. Yin H, Artsimovitch I, Landick R, Gelles J. 1999. Proc. Natl. Acad. Sci. USA 96:13124–2915. Greenleaf WJ, Woodside MT, Block SM. 2007. Annu. Rev. Biophys. Biomol. Struct. 36:171–
9016. Dame RT, Wyman C, Goosen N. 2003. J. Microsc. 212:244–5317. Alessandrini A, Facci P. 2005. Meas. Sci. Technol. 16:R65–9218. Kapanidis AN, Margeat E, Laurence TA, Doose S, Ho SO, et al. 2005. Mol. Cell 20:347–5619. Harada Y, Funatsu T, Murakami K, Nonoyama Y, Ishihama A, Yanagida T. 1999. Biophys.
J. 76:709–1520. Kabata H, Kurosawa O, Arai I, Washizu M, Margarson SA, et al. 1993. Science 262:1561–
6321. Wu T, Schwartz DC. 2007. Anal. Biochem. 361:31–4622. Kim JH, Larson RG. 2007. Nucleic. Acids Res. 35:3848–5823. Gueroui Z, Place C, Freyssingeas E, Berge B. 2002. Proc. Natl. Acad. Sci. USA 99:6005–1024. Coban O, Lamb DC, Zaychikov E, Heumann H, Nienhaus GU. 2006. Biophys. J.
90:4605–1725. Mukhopadhyay J, Mekler V, Kortkhonjia E, Kapanidis AN, Ebright YW, Ebright RH.
2003. Methods Enzymol. 371:144–5926. Ha T. 2001. Methods 25:78–8627. Harada Y, Ohara O, Takatsuki A, Itoh H, Shimamoto N, Kinosita K Jr. 2001. Nature
409:113–1528. Neuman KC, Block SM. 2004. Rev. Sci. Instrum. 75:2787–80929. Sakata-Sogawa K, Shimamoto N. 2004. Proc. Natl. Acad. Sci. USA 101:14731–3530. Wuite GJ, Davenport RJ, Rappaport A, Bustamante C. 2000. Biophys. J. 79:1155–6731. Gelles J, Schnapp BJ, Sheetz MP. 1988. Nature 331:450–53
www.annualreviews.org • Single-Molecule Studies of RNA Polymerase 173
Ann
u. R
ev. B
ioch
em. 2
008.
77:1
49-1
76. D
ownl
oade
d fr
om a
rjou
rnal
s.an
nual
revi
ews.
org
by Y
ale
Uni
vers
ity S
TE
RL
ING
CH
EM
IST
RY
LIB
RA
RY
on
06/3
0/08
. For
per
sona
l use
onl
y.
ANRV345-BI77-08 ARI 28 April 2008 11:12
32. Yin H, Landick R, Gelles J. 1994. Biophys. J. 67:2468–7833. Yin H, Wang MD, Svoboda K, Landick R, Block SM, Gelles J. 1995. Science 270:1653–5734. Shaevitz JW, Abbondanzieri EA, Landick R, Block SM. 2003. Nature 426:684–8735. Greenleaf WJ, Block SM. 2006. Science 313:80136. Revyakin A, Allemand JF, Croquette V, Ebright RH, Strick TR. 2003. Methods Enzymol.
370:577–9837. Riggs AD, Bourgeois S, Cohn M. 1970. J. Mol. Biol. 53:401–1738. Belintsev BN, Zavriev SK, Shemyakin MF. 1980. Nucleic Acids Res. 8:1391–40439. von Hippel PH, Berg OG. 1989. J. Biol. Chem. 264:675–7840. Berg OG, Winter RB, von Hippel PH. 1981. Biochemistry 20:6929–4841. Khoury AM, Lee HJ, Lillis M, Lu P. 1990. Biochim. Biophys. Acta 1087:55–6042. Singer PT, Wu CW. 1988. J. Biol. Chem. 263:4208–1443. Winter RB, Berg OG, von Hippel PH. 1981. Biochemistry 20:6961–7744. Singer P, Wu CW. 1987. J. Biol. Chem. 262:14178–8945. Guthold M, Zhu X, Rivetti C, Yang G, Thomson NH, et al. 1999. Biophys. J. 77:2284–9446. Bustamante C, Guthold M, Zhu X, Yang G. 1999. J. Biol. Chem. 274:16665–6847. Rivetti C, Guthold M, Bustamante C. 1999. EMBO J. 18:4464–7548. Rees WA, Keller RW, Vesenka JP, Yang G, Bustamante C. 1993. Science 260:1646–4949. Rippe K, Guthold M, von Hippel PH, Bustamante C. 1997. J. Mol. Biol. 270:125–3850. Revyakin A, Ebright RH, Strick TR. 2004. Proc. Natl. Acad. Sci. USA 101:4776–8051. deHaseth PL, Zupancic ML, Record MT Jr. 1998. J. Bacteriol. 180:3019–2552. La Porta A, Wang MD. 2004. Phys. Rev. Lett. 92:19080153. Gralla JD, Carpousis AJ, Stefano JE. 1980. Biochemistry 19:5864–6954. Krummel B, Chamberlin MJ. 1989. Biochemistry 28:7829–4255. Straney DC, Crothers DM. 1987. J. Mol. Biol. 193:267–7856. Carpousis AJ, Gralla JD. 1985. J. Mol. Biol. 183:165–7757. Carpousis AJ, Stefano JE, Gralla JD. 1982. J. Mol. Biol. 157:619–3358. Carpousis AJ, Gralla JD. 1980. Biochemistry 19:3245–5359. Kapanidis AN, Laurence TA, Lee NK, Margeat E, Kong XX, Weiss S. 2005. Acc. Chem.
Res. 38:523–3360. Kapanidis AN, Lee NK, Laurence TA, Doose S, Margeat E, Weiss S. 2004. Proc. Natl.
Acad. Sci. USA 101:8936–4161. Lee NK, Kapanidis AN, Koh HR, Korlann Y, Ho SO, et al. 2007. Biophys J. 92:303–1262. Margeat E, Kapanidis AN, Tinnefeld P, Wang Y, Mukhopadhyay J, et al. 2006. Biophys.
J. 90:1419–3163. Travers AA, Burgess RR. 1969. Nature 222:537–4064. Korzheva N, Mustaev A, Nudler E, Nikiforov V, Goldfarb A. 1998. Cold Spring Harb.
Symp. Quant. Biol. 63:337–4565. Straney DC, Crothers DM. 1985. Cell 43:449–5966. Shimamoto N, Kamigochi T, Utiyama H. 1986. J. Biol. Chem. 261:11859–6567. Bar-Nahum G, Nudler E. 2001. Cell 106:443–5168. Mukhopadhyay J, Kapanidis AN, Mekler V, Kortkhonjia E, Ebright YW, Ebright RH.
2001. Cell 106:453–6369. Wang MD, Schnitzer MJ, Yin H, Landick R, Gelles J, Block SM. 1998. Science 282:902–770. Schafer DA, Gelles J, Sheetz MP, Landick R. 1991. Nature 352:444–4871. Levin JR, Krummel B, Chamberlin MJ. 1987. J. Mol. Biol. 196:85–10072. Wilson KS, Conant CR, von Hippel PH. 1999. J. Mol. Biol. 289:1179–9473. Korzheva N, Mustaev A, Kozlov M, Malhotra A, Nikiforov V, et al. 2000. Science 289:619–
25
174 Herbert · Greenleaf · Block
Ann
u. R
ev. B
ioch
em. 2
008.
77:1
49-1
76. D
ownl
oade
d fr
om a
rjou
rnal
s.an
nual
revi
ews.
org
by Y
ale
Uni
vers
ity S
TE
RL
ING
CH
EM
IST
RY
LIB
RA
RY
on
06/3
0/08
. For
per
sona
l use
onl
y.
ANRV345-BI77-08 ARI 28 April 2008 11:12
74. Cramer P, Bushnell DA, Kornberg RD. 2001. Science 292:1863–7675. Gnatt AL, Cramer P, Fu J, Bushnell DA, Kornberg RD. 2001. Science 292:1876–8276. Kasas S, Thomson NH, Smith BL, Hansma HG, Zhu X, et al. 1997. Biochemistry 36:461–
6877. Rivetti C, Codeluppi S, Dieci G, Bustamante C. 2003. J. Mol. Biol. 326:1413–2678. Yager TD, von Hippel PH. 1991. Biochemistry 30:1097–11879. Chamberlin MJ. 1994. In The Harvey Lectures, Series 88, 1992–1993, pp. 1–21. New York:
Wiley80. Komissarova N, Kashlev M. 1997. J. Biol. Chem. 272:15329–3881. Komissarova N, Kashlev M. 1997. Proc. Natl. Acad. Sci. USA 94:1755–6082. Reeder TC, Hawley DK. 1996. Cell 87:767–7783. Landick R. 1997. Cell 88:741–4484. Guerin M, Leng M, Rahmouni AR. 1996. EMBO J. 15:5397–40785. Mustaev A, Kashlev M, Zaychikov E, Grachev M, Goldfarb A. 1993. J. Biol. Chem.
268:19185–8786. Mallik R, Carter BC, Lex SA, King SJ, Gross SP. 2004. Nature 427:649–5287. Finer JT, Simmons RM, Spudich JA. 1994. Nature 368:113–1988. Sowa Y, Rowe AD, Leake MC, Yakushi T, Homma M, et al. 2005. Nature 437:916–1989. Noji H, Yasuda R, Yoshida M, Kinosita K. 1997. Nature 386:299–30290. Svoboda K, Schmidt CF, Schnapp BJ, Block SM. 1993. Nature 365:721–2791. Greenleaf WJ, Woodside MT, Abbondanzieri EA, Block SM. 2005. Phys. Rev. Lett.
95:20810292. Watson JD, Crick FH. 1953. Nature 171:737–3893. Kerppola TK, Kane CM. 1991. FASEB J. 5:2833–4294. Skinner GM, Baumann CG, Quinn DM, Molloy JE, Hoggett JG. 2004. J. Biol. Chem.
279:3239–4495. Foster JE, Holmes SF, Erie DA. 2001. Cell 106:243–5296. Erie DA, Hajiseyedjavadi O, Young MC, von Hippel PH. 1993. Science 262:867–7397. Artsimovitch I, Landick R. 2000. Proc. Natl. Acad. Sci. USA 97:7090–9598. Yarnell WS, Roberts JW. 1999. Science 284:611–1599. Pasman Z, von Hippel PH. 2002. J. Mol. Biol. 322:505–19
100. Davenport RJ, Wuite GJ, Landick R, Bustamante C. 2000. Science 287:2497–500101. Tolic-Norrelykke SF, Engh AM, Landick R, Gelles J. 2004. J. Biol. Chem. 279:3292–99102. Bar-Nahum G, Epshtein V, Ruckenstein AE, Rafikov R, Mustaev A, Nudler E. 2005. Cell
120:183–93103. Bai L, Shundrovsky A, Wang MD. 2004. J. Mol. Biol. 344:335–49104. Guajardo R, Sousa R. 1997. J. Mol. Biol. 265:8–19105. Yin YW, Steitz TA. 2004. Cell 116:393–404106. Forde NR, Izhaky D, Woodcock GR, Wuite GJ, Bustamante C. 2002. Proc. Natl. Acad.
Sci. USA 99:11682–87107. Galburt EA, Grill SW, Wiedmann A, Lubkowska L, Choy J, et al. 2007. Nature 446:
820–23108. Guenther MG, Levine SS, Boyer LA, Jaenisch R, Young RA. 2007. Cell 130:77–88109. Artsimovitch I, Landick R. 2002. Cell 109:193–203110. Tang H, Liu Y, Madabusi L, Gilmour DS. 2000. Mol. Cell Biol. 20:2569–80111. Palangat M, Meier TI, Keene RG, Landick R. 1998. Mol. Cell 1:1033–42112. Ring BZ, Yarnell WS, Roberts JW. 1996. Cell 86:485–93113. Kireeva ML, Hancock B, Cremona GH, Walter W, Studitsky VM, Kashlev M. 2005.
Mol. Cell 18:97–108
www.annualreviews.org • Single-Molecule Studies of RNA Polymerase 175
Ann
u. R
ev. B
ioch
em. 2
008.
77:1
49-1
76. D
ownl
oade
d fr
om a
rjou
rnal
s.an
nual
revi
ews.
org
by Y
ale
Uni
vers
ity S
TE
RL
ING
CH
EM
IST
RY
LIB
RA
RY
on
06/3
0/08
. For
per
sona
l use
onl
y.
ANRV345-BI77-08 ARI 28 April 2008 11:12
114. Richardson JP, Greenblatt J. 1996. See Ref. 136, pp. 822–48115. Landick R, Turnbough CJ, Yanofky C. 1996. See Ref. 136, pp. 1263–86116. de la Mata M, Alonso CR, Kadener S, Fededa JP, Blaustein M, et al. 2003. Mol. Cell
12:525–32117. Yonaha M, Proudfoot NJ. 1999. Mol. Cell 3:593–600118. Landick R. 2006. Biochem. Soc. Trans. 34:1062–66119. Shundrovsky A, Santangelo TJ, Roberts JW, Wang MD. 2004. Biophys. J. 87:3945–53120. Markovtsov V, Mustaev A, Goldfarb A. 1996. Proc. Natl. Acad. Sci. USA 93:3221–26121. Holmes SF, Erie DA. 2003. J. Biol. Chem. 278:35597–608122. Abbondanzieri EA, Shaevitz JW, Block SM. 2005. Biophys. J. 89:L61–3123. Adelman K, Yuzenkova J, La Porta A, Zenkin N, Lee J, et al. 2004. Mol. Cell 14:753–62124. Erie DA, Yager TD, von Hippel PH. 1992. Annu. Rev. Biophys. Biomol. Struct. 21:379–415125. Thomas MJ, Platas AA, Hawley DK. 1998. Cell 93:627–37126. Jeon C, Agarwal K. 1996. Proc. Natl. Acad. Sci. USA 93:13677–82127. Harrington KJ, Laughlin RB, Liang S. 2001. Proc. Natl. Acad. Sci. USA 98:5019–24128. Toulokhonov I, Artsimovitch I, Landick R. 2001. Science 292:730–33129. Arndt KM, Chamberlin MJ. 1990. J. Mol. Biol. 213:79–108130. Park JS, Roberts JW. 2006. Proc. Natl. Acad. Sci. USA 103:4870–75131. Santangelo TJ, Roberts JW. 2004. Mol. Cell 14:117–26132. Reynolds R, Bermudez-Cruz RM, Chamberlin MJ. 1992. J. Mol. Biol. 224:31–51133. Gusarov I, Nudler E. 1999. Mol. Cell 3:495–504134. Chan CL, Wang D, Landick R. 1997. J. Mol. Biol. 268:54–68135. Kashlev M, Komissarova N. 2002. J. Biol. Chem. 277:14501–8136. Neidhardt F, Curtiss IR, Ingraham JL, Lin ECC, Low KB, et al., eds. 1996. Escherichia
coli and Salmonella: Cellular and Molecular Biology. Washington, DC: ASM Press
Integrating Diverse Data for Structure Determinationof Macromolecular AssembliesFrank Alber, Friedrich Förster, Dmitry Korkin, Maya Topf, and Andrej Sali � � � � � � � �443
