RNA polymerase II trigger loop residues stabilizeand position the incoming nucleotidetriphosphate in transcriptionXuhui Huanga,b,1, Dong Wangb,c, Dahlia R. Weissb,d, David A. Bushnellb, Roger D. Kornbergb, and Michael Levittb,1

aDepartment of Chemistry, Hong Kong University of Science and Technology, Clear Water Bay, Kowloon, Hong Kong; cSkaggs School of Pharmacy andPharmaceutical Sciences, University of California at San Diego, La Jolla, CA 92093; dDepartment of Pharmaceutical Chemistry, University of California,San Francisco, CA, 94158; and bDepartment of Structural Biology, Stanford University School of Medicine, Stanford, CA, 94305

Contributed by Michael Levitt, July 19, 2010 (sent for review May 16, 2010)

A structurally conserved element, the trigger loop, has been sug-gested to play a key role in substrate selection and catalysis of RNApolymerase II (pol II) transcription elongation. Recently resolvedX-ray structures showed that the trigger loop forms direct interac-tions with the β-phosphate and base of the matched nucleotidetriphosphate (NTP) through residues His1085 and Leu1081, respec-tively. In order to understand the role of these two critical residuesin stabilizing active site conformation in the dynamic complex, weperformed all-atom molecular dynamics simulations of the wild-type pol II elongation complex and its mutants in explicit solvent.In the wild-type complex, we found that the trigger loop is stabi-lized in the “closed” conformation, and His1085 forms a stable in-teraction with the NTP. Simulations of point mutations of His1085are shown to affect this interaction; simulations of alternativeprotonation states, which are inaccessible through experiment,indicate that only the protonated form is able to stabilize theHis1085-NTP interaction. Another trigger loop residue, Leu1081,stabilizes the incoming nucleotide position through interactionwith the nucleotide base. Our simulations of this Leu mutant sug-gest a three-component mechanism for correctly positioning theincoming NTP in which (i) hydrophobic contact through Leu1081,(ii) base stacking, and (iii) base pairing work together to minimizethe motion of the incoming NTP base. These results complementexperimental observations and provide insight into the role ofthe trigger loop on transcription fidelity.

nucleotide selection ∣ transcription fidelity

The process by which gene-encoding DNA is transcribed into acomplimentary messenger RNA sequence, termed transcrip-

tion, is carried out in the eukaryotic cell by complex molecularmachinery in which RNA polymerase II (pol II) is the key player(1–3). Nucleotide addition and translocation proceeds through aproposed four-step cycle (4–6). A nucleotide triphosphate (NTP)enters the enzyme by the entry site (“E” site). The NTP thenrotates into the nucleotide addition site (“A” site) where it is re-tained if complementary with the template DNA base. Ligationto the 3′ end of the nascent RNA chain results in the pretranslo-cation state (7). Finally, the enzyme translocates to the iþ 1position on the template DNA, leaving the A site empty inthe posttranslocation complex and ready for the next round.However, the mechanism of selection and catalysis is far fromunderstood (4).

A structurally conserved element, the trigger loop, has beensuggested to play a key role in both selection of the correctlymatched NTP and catalysis of transcription elongation in eukar-yotes and prokaryotes (8–10). Structures of the transcribingcomplex reported by Wang et al., (8) reveal the trigger loop ina previously unseen “closed” conformation, making a networkof interactions with the correctly matched NTP in the A site.Mismatched NTPs are proposed to be destabilizing, so thatthe trigger loop does not achieve the closed conformation thatallows it to trigger catalysis. Mutation of the trigger loop in both

eukaryotic pol II and the closely related bacterial enzyme RNAPalters the elongation behavior of the enzyme (11–15). However,mutation does not always render the enzyme inactive, even upondeletion of the entire trigger loop (12, 16). Thus, two modes oftranscription elongation emerge: a fast, high fidelity, trigger loop-dependent mode and a slow, low fidelity trigger loop-indepen-dent mode.

In the closed conformation, the trigger loop directly contactsthe β-phosphate and base of the matched NTP through residuesHis1085 and Leu1081, respectively (8). Specifically, the sidechain of His1085 is precisely positioned to directly interact withthe β-phosphate oxygen of the incoming NTP, which helps to sealoff the active site conformation through electrostatic interactions.Moreover, it is also suggested that the protonated form ofHis1085 is able to act as an electron-withdrawing group andthereby facilitate the formation of the phosphodiester bond uponnucleophillic attack (8, 17). Thus, the His1085-NTP interactionis hypothesized to couple selection with catalysis, greatly enhan-cing the rate at which correct NTPs are incorporated over the rateof 2′-deoxy-NTP or incorrectly matched NTP, as the energy ofcorrect base pairing alone cannot account for the enzyme’sremarkable selectivity. Substitution of tyrosine for His1085 byKaplan et al. (16) (which preserves the ability to hydrogen bond)exhibited severely deficient elongation at saturating concentra-tions of NTP; the impairment was smaller for the addition of2′-deoxy-NTP, which indicates that His1085 plays little, if any,role in selection of an incorrect NTP. Substitution of phenylala-nine for His1085 was lethal (16).

Leu1081 forms a hydrophobic contact with the incoming NTP(8), which is also expected to play important roles in stabilizingthe active site conformation and positioning the base. In thebacterial RNAP, mutation of the nonpolar residue at the sameposition of the trigger loop (M932A) increases the intrinsic paus-ing duration by an order of magnitude (18). In addition, Leu1081is also suggested to be important in translocation. In a recentcocrystal structure of the pol II elongation complex with theantibiotic α-amanitin in the active site, Leu1081 forms a wedgedconfiguration into the bridge helix, which may eventually triggerthe forward translocation of the hybrid (19).

Although experimental studies provide a wealth of informa-tion, it is difficult for experiments to provide dynamic informationat atomic resolution. Through use of a physics-based force field,all-atom molecular dynamics (MD) simulations can model

Author contributions: X.H., D.W., D.R.W., D.A.B., R.D.K., and M.L. designed research;X.H., D.W., and D.R.W. performed research; X.H., D.W., D.R.W., and M.L. analyzed data;and X.H., D.W., D.R.W., D.A.B., R.D.K., and M.L. wrote the paper.

The authors declare no conflict of interest.

Freely available online through the PNAS open access option.1To whom correspondence may be addressed. E-mail: michael.levitt@stanford.edu orxuhuihuang@ust.hk.

This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.1073/pnas.1009898107/-/DCSupplemental.

www.pnas.org/cgi/doi/10.1073/pnas.1009898107 PNAS ∣ September 7, 2010 ∣ vol. 107 ∣ no. 36 ∣ 15745–15750

BIOPH

YSICSAND

COMPU

TATIONALBIOLO

GY

Dow

nloa

ded

by g

uest

on

Feb

ruar

y 2,

202

2

detailed atomistic interactions and provide dynamic information.Thus, computer simulations may complement structural andbiochemical experiments by providing information inaccessibleto experiments. In the current study, computer simulation mayhelp address many remaining important questions. For example,the effect of the protonation state change of His1085 inducedby the environment is difficult to investigate experimentally. It ishypothesized that His1085 must be protonated in order to act asan electron-withdrawing group, facilitating the SN2 chemical re-action (8). Simulation is also able to provide information on dy-namics and flexibility, which experiments do not. Finally, in silicowe can design computer experiments to separately investigate dif-ferent components that contribute to the stability of the pol IIactive site conformation, for example, separating the interactionscontributing to the stability of the incoming NTP base into hydro-phobic contact, base-pairing, and base stacking.

In this study we perform a total of 500-ns molecular dynamicssimulations of the pol II elongation complex in explicit water(∼370;000 atoms) to study two trigger loop residues, His1085and Leu1081. Our goal is to gain a detailed understanding ofthe roles of these two critical residues in positioning and stabiliz-ing the incoming NTP in a catalytically competent conformation.

ResultsWe first investigate if RNA pol II transcription elongationcomplexes containing a variety of protein and nucleic acids ele-ments (see Fig. S1) are stable in molecular dynamics simulations.As shown in Fig. S2, the rmsd for different elements during the2.2 ns of the simulation and averaged over 20 runs is alwayssmaller than 4 Å. With more than 3,000 residues and a relativelyflexible protein and nucleic acid structure, we consider this to besmall. We extend one simulation to 46 ns, and even over this longtime period the rmsd for all the atoms is still reasonable (<6.5 Å).

1. Role of His1085 in Transcription Elongation. His1085 stabilizes theactive site conformation compared to point mutations. Trigger loopresidue His1085 has been shown to play an important role in polII function. In vivo, point mutation to Phe results in total invia-bility, whereas point mutation to Tyr induces a severe growthdefect (16). Further in vitro studies indicated that the defectivegrowth seen in the Tyr mutant is actually caused by defects in polII elongation (16). However, the molecular mechanisms by whichthese point mutations alter the transcription elongation rateremain unclear. Therefore, we first perform computer simula-tions to investigate these two single point His1085 mutations(see Fig. 1).

When His1085 is mutated to Phe, the interactions betweenPhe and NTP are totally lost in our simulations. Fig. 2 shows thatthe distance between the Pβ atom of NTP and the center of massof the aromatic ring is generally larger than 10 Å; the percentageof the stable conformations where the minimum distance be-tween any pair of atoms from Phe and GTP is less than 3.5 Åis also negligible (0.2%, see Table S1). The mutation of His1085to Phe not only causes the loss of hydrogen bonding interactions,but also causes steric repulsion as the six-membered aromaticring in Phe is bigger than the five-membered imidazole ring inHis. Our simulation results explain why this single point mutationcan render the enzyme totally inactive (16).

Simulation shows substitution of His1085 to Tyr apparently hashigher stability than the mutation to Phe. The distance distribu-tion in Fig. 2 has two peaks, one at ∼6 Å and another at ∼8 Å.The hydrogen bond existence map in Fig. 3 also indicates that afraction of Tyr conformations is able to hydrogen bond with theNTP. Interestingly, the majority of these hydrogen bonds areformed between Tyr and oxygen (OB1) of the β-phosphate (seeFig. 3). The larger six-membered aromatic ring of Tyr is moreconfined in the active site and prefers particular orientations thatfacilitate a hydrogen bond with the oxygen closer to the base of

the NTP. Therefore, the Tyr mutant may negatively affect, but nottotally shut down, the pol II elongation. This is consistent withexperimental observations that this mutant is still active, butdecreases the elongation rate by an order of magnitude comparedto the wild-type enzyme (16)

The protonation state of His1085 plays an important role in active sitestability. Wang et al. (8) suggested that the protonated imidazolegroup of His1085 plays a crucial role in the phosphodiesterbond formation of RNA synthesis. The protonated histidinecan withdraw electron density from the phosphate and facilitateSN2 attack of the RNA 3′-terminal OH group. Furthermore, theinteractions between His1085 and NTP can also stabilize the tran-sition state for the SN2 reaction. Similar mechanisms have beenobserved in many nucleic acid polymerases (17) as well as otherbiological systems such as serine/threonine metallo-phosphotases(20, 21).

Understanding the nature of the interactions between His1085and NTP is thus important for understanding the role of His1085in catalysis as well as in stabilizing the closed conformation.His1085 interacts with the phosphate of NTP through two typesof interactions: hydrogen bonds and salt bridges. Hydrogen bond-ing interactions exist when His1085 takes either the protonatedform or one of the unprotonated forms with hydrogen connectedtoNε (see Fig. 1); salt bridge interactions exist only when the his-tidine side chain is protonated.

The effect of the histidine protonation state on the His1085-NTP interactions is important, but difficult to investigate experi-mentally. By designing artificial systems, computer simulationscan study the independent contributions from hydrogen bondsand salt bridges to the interaction between His1085 and NTP.Specially, we simulate three protonation states of His1085 asshown in Fig. 1: HIP (protonated Nδ and Nε), HIE (protonated

Fig. 1. A schematic view of the locations of His1085 and Leu1081 in theRNA pol II active site. RNA is shown in red, template DNA in cyan, and thetrigger loop in magenta. Three different protonation states for His1085 areinvestigated in our simulations: HIP (the protonated form); HID (one of theunprotonated forms with a hydrogen bonded to Nδ); and HIE (the otherunprotonated form with a hydrogen bonded to Nε). An artificial state withan extra positive charge placed on Nε (HIDþ) is also studied in order to in-vestigate the role of the charge–charge interactions for the stabilizationof His1085 in the active site. Two single mutants (PHE and TYR) forHis1085 and two single mutants (GLY and ALA) for Leu1081 are also studied.The nucleotide on the DNA template forming the base paring with GTP iseliminated (Abasic) in order to investigate the role of base-pairing interac-tions in stabilizing the active site configuration. Finally, the 3′ terminal baseof the RNA is eliminated (Abasic 3′ RNA) to investigate the role of base stack-ing in stabilizing the active site configuration.

15746 ∣ www.pnas.org/cgi/doi/10.1073/pnas.1009898107 Huang et al.

Dow

nloa

ded

by g

uest

on

Feb

ruar

y 2,

202

2

Nε), and HID (protonated Nδ). In addition, we construct anartificially charged state by assigning a positive charge to the Nε

atom of HID (HIDþ).The protonated form (HIP) is indeed the most stable system in

our simulations because it can interact with the phosphate groupof NTP through both hydrogen bonding and salt bridge interac-tions. HIP is stable in most of the conformations (88.8%,Table S1), with the Nε-Hε group generally hydrogen bonding witheither OB1 or OG1 atoms on the phosphate group of NTP (seeFig. 3). Significantly less stable interaction between histidineand the phosphate group of NTP is found in the unprotonatedform, HIE (39.9%), which is able to form only hydrogen bondsbut not salt bridges. The distance distribution for this system asdisplayed in Fig. 2 is bimodal. Although the major peak is still at∼5 Å, corresponding to stable interactions, the existence of peaksat larger distances indicates that His1085 is less stable without thesalt bridge interactions. These distance distributions, generatedfrom 20 trajectories each, have been shown to reach convergence(see Fig. S3). The comparison between HIE and Tyr is interesting,because both of them form only hydrogen bonds. HIE (39.9%stable conformations) is slightly more stable than Tyr (25.1%stable conformations). This difference may be due to the stericeffects caused by the larger size of the aromatic ring in the Tyrside chain. The other His protonation form, HID, behaves like

the His1085 to Phe mutation that can form neither hydrogenbonding nor salt bridge interactions with NTP. For this system,only a very small fraction (4.0%) is observed to be stable andthe distance distribution peaks around 10 Å (see Fig. 2). Amongall the protonation forms, the protonated histidine (HIP) is theonly form that can stabilize the closed conformations (more than50% of the conformations are stable).

In order to compare the contribution of hydrogen bonds andsalt bridges to the stability, we designed the artificially protonatedform, HIDþ, which can form salt bridges but not hydrogen bonds(see Fig. 1). As shown in Fig. 2, there is only a single peak at ∼5 Åin the distance probability distribution for HIDþ. In addition, alarger fraction of conformations (59.8%) is stable for HIDþcompared to HIE where only hydrogen bonding interaction ispresent. These results suggest that the salt bridge contributesmore than hydrogen bonding to His1085-NTP interactions.

2. Leu1081 Positions the Incoming Nucleotide Base. The trigger loopresidue, Leu1081, makes hydrophobic contact with the incomingNTP base in the closed conformation (8). Our simulation resultsshow that Leu1081 is actually the least flexible residue of the trig-ger loop, as measured by root mean square fluctuation of Cα

atoms (see Fig. S4). This indicates a strong and stable interactionbetween Leu1081 and the NTP base. In order to further study this

Fig. 2. (A) Definition of the distance, d, between the Pβ atom in GTP and the center of mass of the imidazole ring in His1085. Histograms of the distancedistribution for systems with different protonation states of His1085 and single mutations of His1085 are shown in B and C, respectively. Only data collectedafter the first 1 ns of each simulation are included in the analysis, and the bin size is set to 0.5 Å.

Fig. 3. Mapping of the existence of hydrogen bonds. For each system, the existence of a certain hydrogen bond is plotted for multiple conformations. A totalnumber of 500 conformations were taken for each system; conformations were sampled every 50 ps from the last 1.2 ns of each of the 2.2-ns trajectory.Hydrogen bonds are shown for 25 snapshots from each one of the 20 trajectories, and trajectories are marked in alternating shades of gray. The standarddeviations are computed by bootstrapping the 20 independent trajectories 20 times with replacement.

Huang et al. PNAS ∣ September 7, 2010 ∣ vol. 107 ∣ no. 36 ∣ 15747

BIOPH

YSICSAND

COMPU

TATIONALBIOLO

GY

Dow

nloa

ded

by g

uest

on

Feb

ruar

y 2,

202

2

interaction, we investigate the dynamic profile of the incomingNTP using two vertical (up or down the nascent nucleic aciddouble helix) and one lateral distance (perpendicular to this helixaxis in the base-pairing direction) as shown in Fig. 4. Further-more, we compare the results of the wild-type with two singlepoint mutations: Leu1081Ala and Leu1081Gly, where the hydro-phobic interactions between residue 1081 and the incomingNTP base are eliminated. We measure the distance betweenthe Cα atom of the Leu1081 and the center of mass (c.o.m.)of the GTP base as an indicator of the vertical motion betweenthe trigger loop and GTP. Mutation of Leu1081 to Ala alters thisdistance distribution and induces a larger fluctuation (Fig. 4B).The distance between the c.o.m. of the GTP base and that ofthe nucleotide base in the −2 position of the nascent RNA chain(the second vertical motion; see Fig. 4C) is unchanged. Lateralmotion between the DNA template and the incoming GTP is alsounaffected (see Fig. 4A). Mutation to Gly has an even largereffect on the trigger loop to NTP vertical motion, but might alsodisrupt the trigger loop backbone stability. These results indicatethat Leu1081 maintains the correct position of the base by mini-mizing the vertical motion between the trigger loop and theincoming NTP.

To fully understand the nucleotide base positioning, we alsoinvestigate the effect of two additional components contributingto the base stability: base pairing and base stacking. In order toseparately study these two components, we designed the systemwith an abasic mutation of either the 3′ terminal RNA nucleotideor the DNA nucleotide at þ1 position, which eliminates the basestacking and base pairing, respectively (see Fig. 1). In the wildtype, when base pairing is present, there is no lateral motion.Without base pairing, as in the abasic DNA mutant, there isincreased lateral motion, greater distance distribution, and higherfluctuations. This also explains the slight increase in the fluctua-tion of the vertical motion to the trigger loop. A second system, inwhich the RNA 3′ nucleotide is mutated to abasic, has no basestacking with the incoming NTP. In this system, the lateral dis-tance is unaffected. So, too, is the vertical distance to the trigger

loop. However, the other vertical distance between the GTP baseand the nucleotide of the nascent RNA chain is greatly increased.

DiscussionIn this study, we use a total of 500-ns all-atom moleculardynamics simulations to investigate the role of two importanttrigger loop residues in the RNA pol II transcription elongationcomplex. Although this system including as it does explicit solventis large (∼370;000 atoms), it proved to be stable on the nanose-cond time scale. This allowed molecular dynamics to provide datathat are inaccessible through experimental observation.

We found that the wild-type complex, when simulated in themost likely His1085 protonation state (HIP), forms the moststable interaction with the GTP phosphate group and the triggerloop is stabilized in the closed conformation. Simulations of pointmutations of this residue are shown to adversely affect this inter-action, which is consistent with experimental data. We found bothhydrogen bond interactions and charge–charge interactionscontribute to the stabilization of the His1085-NTP interaction.The protonated form of His (HIP) may be the major form inthe presence of NTP due to the pKa shift of the imidazole toa higher value due to a negative charge group nearby (in this case,the triphosphate group). This information is very difficult to ob-tain experimentally, but important for understanding the catalyticreactions. The stabilization of the active site conformation issuggested to have catalytic significance: The closed conformationformed by the trigger loop upon entry of a correctly matched NTPis necessary for the proposed SN2 reaction leading to phospho-diester bond formation (8).

We propose a three-component mechanism for correctly posi-tioning the incoming NTP for selection and catalysis, as shown inFig. 5. While His1085 stabilizes a catalytically competent triggerloop conformation through interactions with the NTP phosphategroup, Leu1081 also stabilizes the incoming nucleotide positionthrough interaction with the base. Leu1081 positions the base byminimizing the vertical motion between the trigger loop and theincoming NTP. This hydrophobic interaction is crucial and thusconserved across species, as in bacterial RNAP, where a methio-

Fig. 4. (A) Distance distributions as a function of time and distance probability distributions in the wild-type, ALA (Leu1081Ala), GLY (Leu1081Gly), Abasic, andAbasic (3′ RNA) systems are shown in black, red, green, blue, and brown, respectively. Distance between center of mass of the GTP base and center of mass ofthe sugar ring of the DNA nucleotide at þ1 position is used (A Left) for the distance definition). Error bars in distance distributions as a function of time arecomputed from 20 simulations. For each system, only data after the first 1 ns of each of the twenty 2.2-ns simulations are included in the calculation. The binsize in distance probability distributions is set to be 0.5 Å. (B) The same as A except that distributions for distance between center of the mass of the GTP baseand the Cα atom of the trigger loop residue Leu1081 are shown. (C) The same asA except that distributions for distance between center of mass of the GTP baseand center of mass of the nucleotide base at the −2 position of the RNA.

15748 ∣ www.pnas.org/cgi/doi/10.1073/pnas.1009898107 Huang et al.

Dow

nloa

ded

by g

uest

on

Feb

ruar

y 2,

202

2

nine residue plays a similar role as leucine in pol II. Eliminatingthis interaction greatly increases the intrinsic pausing duration(18). Furthermore, we predict that mutation of Leu1081 to Ileor Val, which preserves the hydrophobic contact, may still func-tion properly in transcription elongation. In addition, the basestacking from the 3′ RNA terminal nucleotide, sandwiching withLeu1081, controls the positioning of incoming NTP in the verticaldirection. Finally, the template base was found to minimize thelateral motion between the incoming NTP and the templateDNA. Therefore, the template base may play important dualroles in transcription: It serves not only as a template for correctbase selection, but also contributes to correct nucleotide position-ing. The role of base pairing and stacking in positioning theincoming NTP has long been recognized (22). It is found experi-mentally that a single abasic site to eliminate base pairing servesas a strong block for pol II elongation (23). However, the role ofLeu in positioning the incoming NTP has only been recognizedrecently (8). Our mechanism explains how these three compo-nents work together to stabilize and correctly position the incom-ing NTPs.

Although this is a set of atomic resolved simulations of thewhole transcription complex in explicit water, it has some limita-tions. Because of the size of the system, we cannot sample toconvergence, making it impossible to calculate an accurate freeenergy difference between the different systems studied. We cir-cumvent this problem by focusing on the distribution of distancesthat are related to experimental observation, by running manyshort simulations and by estimating the expected errors in all cal-culated results by bootstrapping with replacement. Our results doqualitatively reflect the relative stabilities of the different systemsand do indicate the role of His1085 and Leu1081 in the active sitestability. We suggest that there could be an important link be-tween the active site stability and the catalysis and NTP selectionrate. The simulation system we set up may be used to investigateother key residues in pol II transcription. The same methodologycould be used in targeted studies of other large biologicallyimportant protein complexes.

System and MethodMolecular dynamics simulations for the RNA pol II transcriptionelongation complex were carried out with an all-atom model. The

whole transcription elongation complex consists of RNA Pol II, aDNA template, the RNA nascent chain, and GTP. The crystalstructure (PDB ID code 2E2H) reported by Wang et al. (8) isused as our initial configuration. We follow a similar procedureto set up the simulation as in our previous study of the RNA pol IIbacktracked complex (24). Missing residues in the middle of eachchain of pol II are filled in by Segmod (25). The pol II elongationcomplex is solvated in a water box, with nonperiodic water layersat least 7 Å from the pol II surface. The solvated configuration isshown in Fig. S1. Two Mg2þ ions in the catalytic site from thecrystal structure (8) are included and 100 Naþ counterions areadded to make the HIP system electrically neutral. The solvatedsystem for HIP has 368,799 atoms (other systems have a slightvariation in the number of atoms). The ENCAD program (26)is first used for minimization to eliminate bad contacts, and sub-sequently the GROMACS simulation package (27), renownedfor its speed, is used to perform molecular dynamics simulationsfor this large system. The AMBER03 force field (28) is used forthe protein, RNA, DNA, and ions; AMBER03 force field para-meters (28) for the phosphate and sugar are also used for theabasic residue with minor modifications; polyphosphate para-meters developed by Meagher et al. (29) are used for GTP;and the TIP3P (30) water model is used for the explicit solvent.For the long-range electrostatic interactions, the particle-meshEwald method (31) is used with the short-range cutoff of10 Å. For the van der Waals interactions, a typical 9-Å cutoffis used, and a switch function is applied to make the functionssmoothly go to zero at 8 Å. A 10-Å neighbor list is updated every10 time steps. A time step of 2.0 fs is used, and the lengths of thebonds connecting to hydrogen atoms are constrained. The Nose–Hoover thermostat (32) is used for temperature coupling.

We follow a standard equilibration procedure that includesconjugate gradient minimization and 200-ps constant numberof atoms, volume, and temperature MD simulation with positionrestraints on the solute. The final configurations from equilibra-tion are then used for production simulations. Twenty 2.2-nssimulations with different initial velocities are performed for eachof the 10 systems: three protonation states of His1085 (HIP, HIE,and HID), an artificial system with an extra position chargeplaced on Nδ (HIDþ), two single point mutations (Phe andTyr) for His1085, two single point mutations (Gly and Ala) forLeu1081, and two other changes making RNA terminal nucleo-tide and DNA nucleotide at þ1 position abasic. The systems aredetailed in Fig. 1. During the first 200 ps of the simulations, wegradually increase the temperature from 50 K to 300 K, and theremaining 2-ns simulations are run at 300 K. We extended one ofthe molecular dynamics simulations for HIP to 46 ns. Thus, thetotal simulation time is about 500 ns, and each simulation is run inparallel with 56 cores (Dell PowerEdge 1950 compute nodes) onStanford’s Bio-X2 supercomputing cluster.

ACKNOWLEDGMENTS. X.H. acknowledges Hong Kong Research Grants CouncilDAG09/10.SC03 and University Grants Committee RPC10SC03. M.L, X.H.,and D.R.W. acknowledge National Institutes of Health (NIH) Roadmapfor Medical Research Grant U54 GM072970 and NIH Grant GM041455(to M.L). D.W. acknowledges NIH Pathway to Independence AwardGM085136 and start-up fund from Skaggs School of Pharmacy and Pharma-ceutical Sciences, University of California at San Diego. R.D.K. acknowledgesNIH Grant GM49985. Computing resources were provided by NationalScience Foundation award CNS-0619926.

1. Stryer L (1975) Biochemistry (WH Freeman and Company, New York), pp 100–101.

2. Lee TI, Young RA (2000) Transcription of eukaryotic protein-coding genes. Annu Rev

Genet 34:77–137.

3. Shilatifard A, Conaway RC, Conaway JW (2003) The RNA polymerase II elongation

complex. Annu Rev Biochem 72:693–715.

4. Kornberg RD (2007) The molecular basis of eukaryotic transcription. Proc Natl Acad Sci

USA 104:12955–12961.

5. Gnatt AL, Cramer P, Fu JH, Bushnell DA, Kornberg RD (2001) Structural basis of

transcription: An RNA polymerase II elongation complex at 3.3 angstrom resolution.

Science 292:1876–1882.

Fig. 5. A proposed three-component mechanism for the base recognition ofPol II. The base stacking from the 3′ RNA terminal nucleotide, sandwichingwith Leu1081, maintains the correct positioning of incoming NTP base in thevertical direction. The template base positions the incoming NTP base byminimizing the lateral motion between the incoming NTP and the templateDNA.

Huang et al. PNAS ∣ September 7, 2010 ∣ vol. 107 ∣ no. 36 ∣ 15749

BIOPH

YSICSAND

COMPU

TATIONALBIOLO

GY

Dow

nloa

ded

by g

uest

on

Feb

ruar

y 2,

202

2

6. Brueckner F, Ortiz J, Cramer P (2009) A movie of the RNA polymerase nucleotideaddition cycle. Curr Opin Struct Biol 19:294–299.

7. Westover KD, Bushnell DA, Kornberg RD (2004) Structural basis of transcription: Nu-cleotide selection by rotation in the RNA polymerase II active center. Cell 119:481–489.

8. Wang D, Bushnell DA, Westover KD, Kaplan CD, Kornberg RD (2006) Structuralbasis of transcription: Role of the trigger loop in substrate specificity and catalysis.Cell 127:941–954.

9. Vassylyev DG, Vassylyeva MN, Perederina A, Tahirov TH, Artsimovitch I (2007)Structural basis for transcription elongation by bacterial RNA polymerase. Nature448:157–162.

10. Vassylyev DG, et al. (2007) Structural basis for substrate loading in bacterial RNApolymerase. Nature 448:163–168.

11. Archambault J, et al. (1998) Stimulation of transcription by mutations affectingconserved regions of RNA polymerase II. J Bacteriol 180:2590–2598.

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

13. Trinh V, Langelier MF, Archambault J, Coulombe B (2006) Structural perspective onmutations affecting the function of multisubunit RNA polymerases.Microbiol Mol BiolRev 70:12–36.

14. Bar-Nahum G, et al. (2005) A ratchet mechanism of transcription elongation and itscontrol. Cell 120:183–193.

15. Zhang J, Palangat M, Landick R (2010) Role of the RNA polymerase trigger loop incatalysis and pausing. Nat Struct Mol Biol 17:99–104.

16. Kaplan CD, Larsson KM, Kornberg RD (2008) The RNA polymerase II trigger loopfunctions in substrate selection and is directly targeted by alpha-amanitin. Mol Cell30:547–556.

17. Castro C, et al. (2009) Nucleic acid polymerases use a general acid for nucleotidyltransfer. Nat Struct Mol Biol 16:212–218.

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

19. Brueckner F, Cramer P (2008) Structural basis of transcription inhibition by alpha-amanitin and implications for RNA polymerase II translocation. Nat Struct Mol Biol15:811–818.

20. Mumby MC, Walter G (1993) Protein-serine threonine phosphatases—structure,regulation, and functions in cell-growth. Physiol Rev 73:673–699.

21. Silverman R (2000) The Organic Chemistry of Enzyme-Catalyzed Reactions (Academic,San Diego, CA), pp 85–88.

22. Chamberlin MJ (1976) RNA Polymerase (Cold Spring Harbor Lab Press, Cold SpringHarbor, NY), pp 17–67.

23. Tornaletti S, Maeda LS, Hanawalt PC (2006) Transcription arrest at an abasic site in thetranscribed strand of template DNA. Chem Res Toxicol 19:1215–1220.

24. Wang D, et al. (2009) Structural basis of transcription: Backtracked RNA polymerase IIat 3.4 angstrom resolution. Science 324:1203–1206.

25. Levitt M (1992) Accurate modeling of protein conformation by automatic segmentmatching. J Mol Biol 226:507–533.

26. Levitt M, Hirshberg M, Sharon R, Daggett V (1995) Potential energy function andparameters for simulations of the molecular dynamics of proteins and nucleic acidsin solution. Comput Phys Commun 91:215–231.

27. Hess B, Kutzner C, van der Spoel D, Lindahl E (2008) GROMACS 4: Algorithms for highlyefficient, load-balanced, and scalable molecular simulation. J Chem Theory Comput4:435–447.

28. Duan Y, et al. (2003) A point-charge force field for molecular mechanics simulations ofproteins based on condensed-phase quantum mechanical calculations. J ComputChem 24:1999–2012.

29. Meagher KL, Redman LT, Carlson HA (2003) Development of polyphosphate para-meters for use with the AMBER force field. J Comput Chem 24:1016–1025.

30. Jorgensen WL, Chandrasekhar J, Madura JD, Impey RW, Klein ML (1983) Comparisonof simple potential functions for simulating liquid water. J Chem Phys 79:926–935.

31. Darden T, York D, Pedersen L (1995) A smooth particle mesh Ewald potential. J ChemPhys 103:3014–3021.

32. Hoover W (1985) Canonical dynamics: Equilibrium phase-space distributions. Phys RevA 31:1695–1697.

15750 ∣ www.pnas.org/cgi/doi/10.1073/pnas.1009898107 Huang et al.

Dow

nloa

ded

by g

uest

on

Feb

ruar

y 2,

202

2