Molecular Cell, Vol. 1, 1033–1042, June, 1998, Copyright 1998 by Cell Press

Transcriptional Pausing at 162of the HIV-1 Nascent RNA ModulatesFormation of the TAR RNA Structure

et al., 1989; Marciniak et al., 1990; Kato et al., 1992;Graeble et al., 1993 and references therein).

Tat appears to act by stimulating phosphorylation ofRNAPII by the kinase subunits of either TFIIH, P-TEFb,or both (Parada and Roeder, 1996; Cujec et al., 1997;

Murali Palangat,* Timothy I. Meier,†§

Richard G. Keene,†‖ and Robert Landick*‡

*Department of BacteriologyUniversity of Wisconsin—MadisonMadison, Wisconsin 53706†Division of Biology and Biomedical Sciences Zhu et al., 1997). The position in the HIV-1 transcriptionalWashington University unit at which the critical phosphorylation occurs andSt. Louis, Missouri 63130 how it alters RNA chain elongation are unknown. How-

ever, formation of TAR from at least the first 42 nt ofthe nascent RNA and its interaction with Tat are required

Summary prior to complete assembly of an elongation-proficientTC (Berkhout et al., 1989). Thus, Tat–TAR interaction

A strong transcriptional pause delays human RNA must occur after U42 exits RNAPII and becomes avail-polymerase II three nt after the last potentially paired able for base-pairing with A20. Since 16–20 nt of nascentbase in HIV-1 TAR, the RNA structure that binds the RNA are protected by RNAPII (Rice et al., 1991; Gu ettransactivator protein Tat. We report here that the al., 1996), synthesis of a 58–62 nt nascent RNA shouldHIV-1 pause depends in part on an alternative RNA demarcate the earliest point for Tat–TAR interaction.structure (the HIV-1 pause hairpin) that competes with What molecular mechanism ensures that Tat–TAR in-formation of TAR. By probing the nascent RNA struc- teraction is established before the unmodified TC termi-ture in halted transcription complexes, we found that nates transcription prematurely? In some other casesthe transcript folds as the pause hairpin before and of regulated transcriptional elongation, a delay in RNAat the pause, and rearranges to TAR concurrent with synthesis prior to the interaction of a regulatory mole-or just after escape from the pause. The pause signal cule is programmed by a pause signal that temporarilytriggers a 2 nt reverse translocation by RNA polymer- halts RNAP. For instance, promoter-proximal pausingase that may block the active site and be counteracted

in Drosophila heat shock genes halts RNAPII aroundby formation of TAR. Thus, the HIV-1 pause site modu-

120 until interaction with the heat shock transcriptionlates nascent RNA rearrangement from a structurefactor (O’Brien and Lis, 1991). Similar promoter-proximalthat favors pausing to one that both recruits Tat andpauses have been detected in many Drosophila genespromotes escape from the pause.(Rougvie and Lis, 1990), in mammalian genes like c-myc(Krumm et al., 1995), and in the late operons of severallambdoid phages where interaction of the Qantitermina-Introductiontion protein releases the paused RNAP (Ring and Rob-erts, 1994).The activation of HIV-1 transcription is the best-under-

A transcriptional pause that could help present TARstood example of a eukaryotic control mechanism thatfor timely interaction of Tat has been reported duringregulates the efficiency of RNA chain elongation rathertranscription of HIV DNA both in vivo (Kessler and Ma-than the rate of initiation (for review, see Jones andthews, 1992) and in vitro (Parada et al., 1995) aroundPeterlin, 1994; Bentley, 1995; Jones, 1997). In the ab-

senceof the transactivator protein Tat,which isencoded 160, very close to the position at which the target forby twodistal exons in thesingle 10 kb HIV-1 transcription Tat-binding emerges from RNAPII, but beyond theunit, RNA polymerase II (RNAPII) initiates mRNA synthe- position of known promoter-proximal pause sites. Thissis at the HIV-1 promoter in the upstream long terminal pause more resembles the class of promoter-distalrepeat, but terminates transcription prematurely, thus pause signals that are found in the leader regions of E.yielding primarily short untranslated RNAs rather than coli operons like his and trp and that depend in partprogramming viral replication. Tat can interact with nu- on nascent RNA secondarystructures. These pause sig-merous components of a transcription complex (TC), nals halt RNAP until a ribosome initiates translation ofincluding RNAPII (Mavankal et al., 1996), the cyclin- a leader peptide coding region, releases the pauseddependent kinase 7 subunit of TFIIH (Parada and Roeder, RNAP, and regulates transcriptional attenuation (Lan-1996; Cujec et al., 1997; Garcia-Martinez et al., 1997), dick et al., 1996a). Like the his and trp pause sites, theand the RNAPII CTD-kinase complex known as P-TEFb HIV-1 pause occurs just after the addition of the last(Zhu et al., 1997; Wei et al, 1998). However, to trigger paired base in an important RNA secondary structureefficient RNA chain elongation either in vivo or in vitro, (TAR for HIV-1, the A:B or 1:2 RNA structures for hisTat must bind to the TAR RNA structure, which forms and trp). Further, the his pause site appears to modulatefrom the initial portion of the HIV-1 transcript (Berkhout initial formation of this RNA structure through a direct

RNAP–RNA interaction (Wang and Landick, 1997). We‡To whom correspondence should be addressed. report here that the HIV-1 pause signal includes an alter-§Present address: Lilly Research Laboratories, Indianapolis, Indiana

native RNA secondary structure that inhibits TAR forma-46285.tion in the nascent transcript and thus may control the‖ Present address: Department of Molecular Biology, Cleveland

Clinic Foundation, Cleveland, Ohio 44195. availability of TAR for interaction with Tat.

Molecular Cell1034

Results

Experimental ApproachTo study intrinsic pausing by RNAPII, we needed toexamine synchronous RNA chain elongation by RNAPIITCs unperturbed by the excess of elongation factorsand DNA-binding proteins that are present in nuclearextracts. To accomplish this, we first formed elongationcomplexes halted after addition of U14 to the HIV-1transcript by incubating an immobilized, 774 bp DNAfragment encoding the HIV-1 promoter and early tran-scribed region fused to the human c-myc exon 1 in aHeLa nuclear extract with dATP, CTP, GTP, and a-32P-UTP (see Figure 1A, Experimental Procedures, and Katoet al., 1992). These halted U14 TCs were recovered fromthe extract and washed with buffer containing 0.15%sarkosyl to remove residual preinitiation complexes,transcription factors, and nonspecific DNA-binding pro-teins (Izban and Luse, 1991). Addition of all 4 NTPs to thesarkosyl-washed complexes restarted chain elongationefficiently, allowing synchronous transcription of theHIV-1 pause site (Figure 1B). To obtain homogenouscomplexes halted before, at, or after the HIV-1 pause,we elongated the RNA chains in the immobilized TCsby repeated incubation with different sets of 3 NTPs,followed by washing to remove each NTP set (stepwisetranscription; see Figure 4, Experimental Procedures,and Kashlev et al., 1993).

RNAPII Recognizes a Strong Intrinsic Pause SignalFigure 1. In Vitro Transcription from the HIV-1 Promoterbefore the Addition of G63 in the HIV-I(A) Experimental approach and sequence of the RNA transcribedEarly Transcribed Regionfrom a template that carries the HIV-I promoter andearly transcribed

Synchronous elongation of the halted U14 complexes region (2138 to 185 with respect to the HIV-I start site) followedrevealed a strong pause RNA band of z62 nt that ap- by c-myc sequences (see Experimental Procedures).peared z18 s after resumption of transcription at 1 mM (B) Time course of transcription through the HIV-1 pause site. Sam-

ples were removed at indicated times after addition of 1 mM eacheach NTP and graduallyelongated into run-off transcriptNTP to sarkosyl-washed U14 complexes, processed, and separatedover a period of several minutes (Figure 1B). Two addi-on a denaturing 11% polyacrylamide gel (see Experimental Proce-tional RNAs z65–67 nt in length appeared at the samedures). (M), 32P-labeled MspI digest of pBR322 with fragment sizes

time, only a fraction of which were further elongated indicated in nt.even after 5 min. We attribute these persistent RNAs to (C) Relative concentrations of the pause RNA (open circles) andtranscriptional arrest. The z62 nt pause and z65–67 nt total RNA at and above the pause (closed circles) in each lane of

Figure 1B. Back-extrapolation of the [pause RNA] to time 0 yieldsarrest RNAs presumably correspond to the HIV-1 pausea maximum pause efficiency of #0.7; nonlinear regression of theRNAs reported in previous studies (Kessler and Ma-[pause RNA] as a function of time gives a pause half-life of 22 sthews, 1992; Parada et al., 1995). Quantitation of the(see Landick et al., 1996b).

z62 nt RNA revealed that the paused complex resumed (D) RNA sequencing ladder compared to pause and arrest RNAs.transcription with a pseudo-first-order half-life of 22 s The RNA sequencing ladder was generated by elongating U14 TCsand that at most 70% of transcribing RNAPII molecules in the presence of all 4 NTPs and a 39-dNTP as a chain terminator

(see Experimental Procedures). Samples obtained at 40 s and 60 sentered the paused state (Figure 1C; see Experimentalduring a time course of elongation without 39dNTPs were electro-Procedures and Landick et al., 1996b). Approximatelyphoresed next to the RNA sequence ladder on a 6% denaturing10% of the transcribing RNAPII became arrested atpolyacrylamide gel.

z65–67. Since pol II requires z10 s to add the 48 nt (E) Location of HIV-1 pause and Tat-binding sites relative to a con-between U14 and the pause (Figure 1B), the 22 s pause ventional depiction of the TAR RNA structure. The shaded box indi-half-life represents a slowing of the rate of chain elonga- cates the segment of nascent RNA likely protected by RNAPII at

the pause site (see text).tion by a factor of .100.To determine the precise 39 ends of the pause and

arrest RNAs, we compared them to an RNA sequenceladder generated with chain-terminating 39dNTPs (Fig- HIV-1 pause exhibits two striking similarities to the well-

characterized his and trp pause signals: (1) pausing oc-ure 1D). The pause RNA terminated with U62 and thearrest RNAs with U65 and U66. These RNAs end 3, 6, curs immediately after an RNA structure that can pair

to within a few nt of the pause site in free RNA, andand 7 nt after the last paired nucleotide in the stem ofthe TAR RNA structure, only a portion of which could (2) pausing occurs where the active site must catalyze

reaction of a 39-terminal uridine with GTP.actually form in the paused TC (Figure 1E). Thus, the

Pausing Modulates HIV-1 Nascent RNA Folding into TAR1035

the upper portion of the TAR structure with the RNAhairpin that is bound by the MS2 coat protein (Figure2B; see Valegard et al., 1994, and references therein).Since overall stability but not sequence is important ina prokaryotic pause hairpin (Chan and Landick, 1993),we were surprised to find that the MS2 substitutionreduced the HIV-1 pause half-life by a factor of 3, eventhough it should be more stable than TAR in the portionof nascent RNA available for base-pairing at the HIV-1pause site (Figure 2B). The 39-proximal substitution es-sentially eliminated the pause, whereas the downstreamchange had little effect. Thus, the HIV-1 pause signalis multipartite since both RNA hairpin and 39-proximalsequences influenced pausing. Dissection of the role of39-proximal sequences and more rigorous testing foreffects of downstream DNA must await future studies;we chose to focus our effort on the interesting effecton pausing of base substitutions in TAR.

RNA Secondary Structure Influencesthe HIV-1 PauseWe next tested the effect on pausing of several pre-viously characterized deletions and substitutions in TAR(Dbulge, CA-loop, UA-loop, and tetraloop, Figure 2B; Wuet al., 1991; Churcher et al., 1993). All four substitutionsdecreased the pause half-life in sarkosyl-washed TCassays, and the most significant effects were causedby the substitutions that should increase the stability ofthe TAR RNA structure (Figure 2B). In particular, the

Figure 2. Pausing at U62 on Mutant Templates and with Purified UUCG tetraloop is known to stabilize RNA structures byRNAPII z2 kcal/mol (Tuerk et al., 1988); on TAR, it reduced the(A) Effects on pausing of base substitutions in the TAR hairpin, 39- pause half-life by a factor of 3 (Figure 2B). Since theproximal and downstream DNA mutants (using sarkosyl-washed tetraloop substitutions are .25 nt from the pause RNAU14 complexes as in Figure 1B), and of transcription with purified

39 end, whereas the RNAPII footprints on template DNAcalf-thymus RNAPIIa (see Experimental Procedures). TAR hairpinand nascent RNA extend at most z20 bp or nt upstreammutant, MS2 is shown in Figure 2B. 39-proximal mutant, replacementin halted complexes (Linn and Luse, 1991; Rice et al.,of GGGAACCCACT62 before pause with GTCTTAGGGAT62. Down-

stream mutant, replacement of G63CTTAAGCCTCAATAAAGCTT 1991; Gu et al., 1993, 1996), it is likely that these substitu-after pause with G63CCGAGACTAGGGAACCTGCA. Cross symbol tions affected pausing through their influence on RNAindicates new pause created in the downstream DNA mutant. secondary structure. However, the strong effects on(B) Pause half-lives with HeLa nuclear extract-generated TCs (black) pausing of substitutions that should stabilize rather thanor purified calf-thymus RNAPII (shaded). Each value is an average

destabilize the TAR structure seemed puzzling, unlessof at least three experiments. Base substititions are shown in bold.they were mediated by an auxiliary factor that was pres-The predicted free energies of formation for the portions of the RNA

structures that should be able to form in the paused TC (base pairs ent in the HeLa nuclear extract and survived the 0.15%C18·G44 and above) are: wild-type, 27.5 kcal/mol; MS2, 29.3 kcal/ sarkosyl wash. To test this idea, we performed transcrip-mol; Dbulge, 213.5 kcal/mol; CA loop, 27.5 kcal/mol; UA loop, 27.4 tion assays using highly purified calf-thymus RNAPII.kcal/mol; tetraloop, 210.8 kcal/mol (Zuker and Stiegler, 1981). Thepredicted DG for the tetraloop was decreased by 2 kcal/mol basedon measurements of its unusual thermal stability (Tuerk et al., 1988);

Recognition of the HIV-1 Pause Sitethe UA loop also is reported to be more stable than the predictedby RNAPII Required No Additionalvalue (Churcher et al., 1993), but its DG was not adjusted because

no measurement is available. Transcription FactorsSince purified RNAPII cannot initiate transcription at apromoter sequence, we synthesized an HIV-I templatewith a defined 39 extension (a “tailed” template) thatThe HIV-1 Pause Signal Is Multipartite

In addition to bases in the active site, the his and trp serves as an efficient initation site for RNAPII (Kadeschand Chamberlin, 1982; Lang et al., 1994). Highly purifiedpause signals depend on an RNA secondary structure,

the 39-proximal RNA or DNA, and the downstream DNA calf-thymus RNAPII initiated in the absence of ATPhalted at position U34 on this DNA template (equivalentduplex (Chan and Landick, 1993; Chan et al., 1997). To

ask expeditiously if the HIV-1 pause signal is similarly to position U14 of the HIV-1 transcript; see ExperimentalProcedures). After addition of 1 mM each NTP, thesemultipartite, we checked the effect on pausing of substi-

tutions that should dramatically alter the TAR RNA struc- complexes recognized the HIV-1 pause similarly to ex-tract-initiated human RNAPII (pause half-life ≈20 s; Fig-ture, the 39-proximal sequence, or the downstream DNA

sequence (Figure 2A). We first tested a replacement of ures 2A and 2B). Further, a template that specified the

Molecular Cell1036

UUCG tetraloop pause signal reduced the pause half-life by a factor of 2 (Figure 2B), consistent with thebehavior of human RNAPII initiated in a HeLa nuclearextract. Transcripts synthesized on these tailed tem-plates were sensitive to RNase T1 digestion, but not toRNase H digestion, establishing that they were dis-placed from the template strand (data not shown; forma-tion of persistent hybrids is a frequent complication oftranscription by purified RNAPII on tailed templates;Kadesch and Chamberlin, 1982).

The calf-thymus RNAPII preparation used in theseexperiments was not phosphorylated on its C-terminalrepeat (RNAPIIa). However, purified human RNAPIIo (thehyperphosphorylated form of RNAPII) gave similar re-

Figure 3. The Pause RNA Hairpin Is a Functional Component of thesults (data not shown), suggesting that the extent ofHIV-1 Pause Signal

phosphorylation of the C-terminal domain was not im-RNA structures in the first 44 nt of HIV-1 RNA predicted by the

portant for pause site recognition. We conclude that method of Zuker and Stiegler (1981): the TAR RNA hairpin (DG 5recognition of the HIV-1 pause site is an intrinsic prop- 27.5 kcal/mol) and the alternative pause RNA hairpin (DG 5 212.6erty of the mammalian RNAPII enzyme and requires only kcal/mol). The shaded box represents the z18 nt segment of RNA

that should be protected by RNAPII. Potentially compensatory basethe nucleic acid sequences and structures present whenchanges for the TAR RNA hairpin and the pause RNA hairpin arethe TC arrives at U62 on an HIV-1 template.shown in boxes.(Inset) Relative pause half-lives for these substitutions alone and incombination were measured as described in the legend to FigureThe HIV-I Pause Is Stimulated by a Pause RNA 1 and in the Experimental Procedures.

Hairpin that Competes with Formationof the TAR RNA HairpinSince the experiments with pure RNAPII appeared to

These results are most easily explained if the alternativerule out involvement of an auxiliary factor in pausing,RNA structure, which we term the HIV-1 pause hairpin,and since we could not explain the effects on pausingstimulates pausing by preventing formation of TAR,of substitutions in TAR by destabilization of the TARwhich itself stimulates escape from the pause site. ThisRNA structure, we next asked whether an RNA structurewould explain why substitutions that disrupt both struc-other than TAR might be formed in the HIV-1 pausedtures had little effect, whereas selective disruption ofTC. We used an RNA structure prediction algorithm tothe pause hairpin or stabilization of TAR reduced paus-examine the possible folding patterns in the first 44 nting and stabilization of the pause hairpin restored paus-of the pause RNA (Zuker and Stiegler, 1981), assuminging (Figures 2 and 3).that z18 nt of RNA upstream from the paused RNA 39

end should be unavailable for structure formation (Riceet al., 1991; Gu et al., 1996). Interestingly, 11 to 144 of Mapping RNA Structures in Active TCs

To test whether the HIV-1 pause hairpin actually forms inthe pause RNA was predicted to fold as an alternativeRNA structure that was both more stable than and mutu- the paused TC, we probed the RNA secondary structure

present in halted TCs by partial digestion with RNaseally exclusive of the corresponding portion of TAR (Fig-ure 3). T1, which cuts 39 to unpaired Gs. To prepare these TCs

and to simplify interpretation of the RNase T1 digestionTo test whether this alternative RNA structure couldbe part of the HIV-1 pause signal, we constructed two patterns, we needed to form the complexes by stepwise

transcription and to incorporate 32P-labeled nucleotidesets of substitutions that would replace base pairs ineither TAR or the alternative structure. The first set of only at or near the 39 end. Thus, we restarted transcrip-

tion in nonradioactive, sarkosyl-washed U14 complexessubstitutions (nt 26–29 [a] or 36–39 [b]; Figure 3) sepa-rately would disrupt both TAR and the alternative RNA by addition of only ATP, GTP, and CTP, so that RNAPII

next halted before addition of U23. After washing withstructure, but restore base pairing only in TAR when theywere combined. Separately, substitution of nt 26–29 or transcription buffer to remove NTPs, RNAPII could be

moved stepwise to the pause site by repeating these36–39 had little effect on the pause half-life (18 s and19 s, as compared to 22 s for wt). However, when they steps with different sets of 3 NTPs (Figure 4). A control

experiment with 32P-NMP incorporated during U14 com-were combined, the pause half-life decreased by a fac-tor of z3 (to 7 s). This result is inconsistent with stimula- plex formation, rather than during the last step, illus-

trates this procedure (Figure 4).tion of pausing by the TAR RNA structure.The second set of substitutions (nt 5–15 [c] or 25–36 To help interpret the RNase T1 digestion patterns,

we analyzed purified RNAs that were predicted to form[d]; Figure 3) would disrupt either the alternative RNAstructure alone (5–15) or both it and TAR (25–36), but either the pause RNA hairpin or the TAR RNA structure

(1–42 and 1–59, respectively; Figure 5), in addition torestore only the alternative RNA structure when the twogroups of substitutions were combined. Separately, TCs halted before, at, and after the pause. We synthe-

sized and purified the 42mer and 59mer RNAs by step-these substitutions reduced the pause half-life signifi-cantly (to 12 s for 5–15) or modestly (to 17 s for 25–36), wise elongation of transcripts in unlabeled U14 com-

plexes, incorporation of 32P-CMP in the last step tobut increased the pause half-life to 24 s when combined.

Pausing Modulates HIV-1 Nascent RNA Folding into TAR1037

Figure 4. Stepwise Transcription by RNAPII up to the HIV-1 PauseSite

Positions of RNA 39 end in TCs halted sequentially by NTP depriva-tion are indicated in large italics within the pause hairpin RNA struc-ture, with lines connecting them to the corresponding RNA bandsseparated in a denaturing 15% polyacrylamide gel. Asterisks indi-cate positions of 32P label. The 59-labeled TCs used in this experi-ment generated significantly more of the undesired weak bands atthe positions other than the indicated ladder than the 39-labeledcomplexes used for the RNA structure mapping experiment (Fig-ure 5).

Figure 5. The HIV-1 Nascent RNA Rearranges from the Pause Hair-pin to TAR at or after the Pause Sitegenerate a 39 label, and recovery from denaturing poly-Purified 42mer and 59mer RNA, and TCs halted at C59, U62, andacrylamide gels (see Experimental Procedures). TheC64 were partially digested with RNase T1 (20 U/ml) in transcriptionC59, U62(paused), and C64 TCs were synthesized andbuffer alone (lanes 4, 5, 9, 10, 12–14, 16–18, and 20–22) or with 4labeled similarly, but not disrupted prior to partial RNaseM urea (lanes 2, 3, 7, and 8). Reactions were carried out for 10 s

T1 digestion. After digestion, the 32P-labeled, 39-proxi- (lanes 2, 4, 7, 9, 12, 16 and 20), 20 s (lanes 3, 5, 8, 10, 13, 17, andmal fragments remained associated with RNAPII in ac- 21), or 30 s (lanes 14, 18, and 22). Samples in lanes 1, 6, 11, 15, andtive TCs since they could be extended further upon 19 were not digested with RNase T1. The 59 end of the digested

RNA is indicated. Cross symbol indicates an RNA band that couldaddition of NTPs (data not shown). Thus, the partialnot be assigned. RNA hairpin structures in purified model RNAsdigestion patterns of the nascent transcripts in these(U42 and C59) and in TCs (U62(P) and C64) with the RNase T1-halted TCs should reflect the actual structure that thesensitive (arrows) and hypersensitive (bold arrows) sites are shown

nascent RNA assumes when RNAPII is located at a given below the gel panels. Asterisks indicate positions of 32P label.template position; comparison of these patterns tothose obtained from the model RNAs was used to iden-tify these structures (Figure 5). and 16–18), whereas the nascent RNA in the C64 com-

plex appeared almost completely rearranged into theRNase T1 cut the model pause hairpin RNA stronglyafter G16, G21, and G36, and to a lesser extent after TAR structure, with strong cleavage in the loop regions

Gs (G32, G33, and G34) and near complete protectionG26, G28, G32, G33, and G34 (Figure 5, lanes 4 and 5).In contrast, RNase T1 cut the model TAR to significant elsewhere (Figure 5, lanes 20–22; note that the band

near G21 is a partially extended RNA present in theextent only in the loop regions bases G32, G33, and G34(Figure 5, lanes 9 and 10; see also Berkhout et al., 1989). undigested sample in lane 19). A comparison of cutting

after G36 and G44, versus cutting after G32–34, is espe-Thus, the first 42 nt of the HIV-1 transcript fold in astructure that is dramatically different from TAR and cially revealing. Although G36 and G44 are further from

the region that would be protected by RNAPII in the C64that is consistent with the predicted pause hairpin struc-ture. Interestingly, the nascent RNAs in both C59 and complex, they are cut much less or not at all relative to

in C59 and U62(paused) complexes. However, the ratioU62(paused) complexes gave RNase T1 digestion pat-terns similar to the model pause RNA pattern (strong of G32–34 to G34 cutting is greater in the U62(paused)

complex than in the purified 42mer RNA (compare lanescutting after G16, G21, and G36; Figure 5, lanes 12–14

Molecular Cell1038

Figure 6. Effect of TFIIS and Pyrophosphate on the HIV-1 PausedTC

U62(paused) and C64 TCs with 59-proximal 32P label were preparedby stepwise transcription (see Experimental Procedures and Figure4) and then incubated with either 2.5 nM TFIIS (U62 TCs, lanes 1–7;C64 TCs, lanes 15–21) or with 50 mM pyrophosphate (U62 TCs,lanes 8–14; C64 complexes, lanes 22–28) for indicated times. RNAsamples were prepared and electrophoresed on a denaturing 7.5%polyacrylamide gel as described in the Experimental Procedures.Bands were assigned by comparison to markers from TCs haltedat 1 nt intervals from C59 to C64 (not shown).

4 and 17, Figure 5), suggesting that TAR may form asa minor constituent of interconverting RNA structuresin the paused complex. We conclude that the nascentRNA in U62 TCs folds predominantly as the HIV-1 pause

Figure 7. Model for Effect of HIV-1 Pausing on Formation of TARRNA hairpin, which inhibits complete formation of TARThe pause hairpin allows backtracking of RNAP in response to theuntil after RNAPII escapes the pause site by no more39-proximal RNA and DNA sequence. The accompanying reversethan 2 nt.translocation of the RNA transcript through the active site blocksRNA chain elongation. Formation of TAR forces the RNA 39 end

The Nascent RNA in the HIV-1 Paused TC back into the active site to allow NTP binding and escape from theIs Reverse Translocated pause.The hatched segment of RNA transcript is paired in the pause

hairpin, but not in the portion of TAR that can form at the pauseTranscriptional pausing likely is caused by inability to(see Figure 3).maintain proper alignment of the RNA 39 end in the active

site, either because the RNAP backtracks so that the39-proximal RNA blocks the active site (reverse translo-

pause signal is an RNA secondary structure, the HIV-1cation) or because the RNA is pulled upstream out of thepause hairpin, that allows reverse translocation and pre-active site (hypertranslocation; see Chan et al., 1997). Invents formation of the TAR RNA structure. Third, re-the other well-characterized case of hairpin-dependentarrangement of the nascent RNA into TAR promotespausing (the his leader pause), formation of the pauseescape from the pause and occurs when RNAPII is athairpin appears to hypertranslocate the nascent RNAor has just escaped from the HIV-1 pause site.based on resistance of the paused TC to transcript hy-

drolysis and to pyrophosphorolysis (Feng et al., 1994;Chan et al., 1997). To distinguish these possibilities for Determinants of Pausing in the HIV-1the HIV-1 paused TC, we tested its sensitivity to TFIIS- Leader Regionstimulated transcript cleavage and to pyrophospho- Several typesof prokaryoticpause sites have been stud-rolysis. We found that the HIV-1 pause RNA is sensitive ied in sufficient detail to establish that the signals areto concentrations of TFIIS (2.5 nM) and pyrophosphate generally multipartite and that some but not all include(50 mM) that have lesser or no effect on other halted an RNA secondary structure (Chan and Landick, 1994).TCs, most notably the C64 complex in which the TAR Two types of RNAPII pause sites have been examinedRNA structure is formed (Figure 6). Further, both treat- todate: (1) promoter-proximal pause sites such as foundments shorten the pause RNA by 2 nt. We conclude that in the Drosophila hsp and mammalian c-myc genesRNA chain elongation most likely is inhibited in the HIV-1 (Rougvie and Lis, 1990; O’Brien and Lis, 1991; Krummpaused complex because RNAPII is backtracked and et al., 1995), which like the l late operon pause sitethe RNA 39 end is reverse translocated downstream by (Ring et al., 1996) may be factor-dependent and reflect2 nt, thus blocking the active site (Figure 7). incomplete loss of promoter contacts; and (2) the U-rich

sites such as adenovirus T1 site and the histone H3.3site, which trigger both intrinsic pausing and transcrip-Discussiontional arrest by causing RNAPII to reverse-translocatethe RNA and DNA chains, presumably because theOur results lead to three principal conclusions. First,

strong transcriptional pausing at 162 of the HIV-1 tran- U-rich RNA:DNA hybrid present in the elongation-com-petent conformation is less stable than thehybrid gener-script is stimulated by a multipartite signal that favors

backtracking of RNAPII and reverse translocation of the ated upon backtracking of RNAPII to an elongation-incompetent conformation (Gu et al., 1993; Reeder andnascent RNA by 2 nt. Second, one component of the

Pausing Modulates HIV-1 Nascent RNA Folding into TAR1039

Hawley, 1996; Komissarova and Kashlev, 1997; Nudler oligonucleotides or upstream RNA sequences to na-scent RNA near the TC that is described by Reeder andet al., 1997). Arrest, but not pausing, at the U-rich sites

is relieved by the TFIIS-dependent transcript cleavage, Hawley (1996) and Komissorova and Kashlev (1997).The HIV-1 pause hairpin could play both indirect andwhich reactivates the arrested TC to allow multiple at-

tempts to overcome the thermodynamic barrier to chain direct roles in pausing. By preventing TAR formation,the pause hairpin may indirectly promote pausing byelongation (reviewed in Reines et al., 1996).

The HIV-1 pause signal represents a third class of allowing RNAPII to reverse translocate and remove theRNA 39 end from the active site. It also is possible thatRNAPII pause signal that depends on neither extrinsic

factors nor U-rich sequences, but rather in part on an interaction of the pause hairpin with RNAPII in a mannersimilar to the putative interaction of the his pause hairpinRNA secondary structure. In this way, it is analogous to

the prokaryoticpause signals found in the leader regions with E. coli RNAP (Wang and Landick, 1997) could stabi-lize it relative to TAR or directly slow RNA chain elon-of amino acid biosynthetic operons regulated by attenu-

ation. Like these prokaryotic signals, the HIV-1 pause gation.Our results also suggest that HIV-1 pause site controlsalso is multipartite, since base changes in both the

pause hairpin and the 39-proximal sequence between the timing of TAR formation. Thus, the most importantimplication of our results is that TAR is unlikely to formthe hairpin and RNA 39 end can reduce pausing (Fig-

ure 2). in the HIV-1 nascent transcript until RNAPII reaches orescapes from the 162 pause site. Since Tat binding toHowever, unlike these prokaryotic pause signals,

which appear to pull the RNA 39 end upstream out of TAR is a central, and presumably early, step in the pro-cess of transactivation, strong transcriptional pausingthe active site and render it resistant to both transcript

cleavage and pyrophosphorolysis, the HIV-1 pause sig- in the HIV-1 leader where TAR can first form could playa role in transactivation. If Tat acts by stimulating phos-nal appears to cause reverse translocation by 2 nt,

based on sensitivity to TFIIS-stimulated transcriptcleav- phorylation of RNAPII (reviewed in Jones, 1997), thenrearrangement of the nascent transcript into TAR, eitherage and pyrophosphorolysis (Figure 6). The primary

cause of this reverse translocation likely is the relative upon Tat binding at the pause or spontaneously afterRNAPII escapes the pause, presumably precedes theinstability of the RNA:DNA hybrid when the 39 end of

the paused transcript is properly positioned in the active critical phosphorylation events. Further, if TAR formsfor only a small fraction of the time that RNAPII residessite (see Guajardo and Sousa, 1997; Komissarova and

Kashlev, 1997; Landick, 1997; Nudler et al., 1997). As- at the pause (i.e., the pause hairpin and TAR are in anequilibrium that favors the pause hairpin; Figure 7), thensuming that the RNA:DNA hybrid for RNAPII is z8 bp,

as it appears to be for E. coli RNAP and RNAPI (see Lee the interaction of Tat with TAR would stabilize TAR rela-tive to the pause hairpin and favor release from theand Landick, 1992; Jeong et al., 1996; Nudler et al.,

1997), a 2 bp reverse translocation would add two un- pause.usually stable rG·dC bp to the hybrid with loss of rC·dGand rU·dA bp, making it z1.2 kcal/mol more stable than

Does the HIV-1 Pause Play a Rolea 39-terminal hybrid (Sugimoto et al., 1995). In contrast,in Transcriptional Regulation?a shift of the hybrid 2 bp upstream in the 39-proximalAlthough the HIV-1 pause may help recruit Tat beforesubstitution that virtually eliminates pausing (Figure 2A)premature termination of RNAPII, there is currently nowould confer no additional predicted stability. Thus, ourcompelling evidence that it plays an essential role inresults are consistent with the idea that the relativeviral replication. In fact, the pause is only likely to bestrength of nascent RNA:template DNA hybrids controlsimportant for transactivation when Tat is present in lim-the lateral stability of the TC; systematic variation of theiting amounts. Transactivation as assayed by cotrans-39-proximal sequence at the HIV-1 pause should testfection of reporterand Tat-expression plasmids into cul-this idea rigorously.tured cells, conditions that overproduce Tat to highlevels, occurs independently of pausing. Under theseconditions, Tat can activate HIV-1 transcription whenFormation of the TAR RNA Structure Releases

RNAPII from the HIV-1 Pause tethered to a binding site upstream from the promoter,even when TAR and the pause signal are not presentReverse translocation of the HIV-1 pause RNA also of-

fers an attractive explanation for the decrease in pause (Kamine et al., 1991; Southgate and Green, 1991). Fur-ther, transactivation occurs in cotransfection experi-half-life that was observed when the TAR RNA structure

was stabilized by base substitution. Since the stem of ments when the 39-proximal region critical for pausingis replaced with several different sequences (Berkhoutthe TAR hairpin can extend closer to the RNA 39 end

than the pause hairpin stem, TAR formation may pull et al., 1989) and when TAR is replaced with different RNAstructures that can recruit Tat fusion proteins targetedtothe RNA back into reactive alignment by driving forward

translocation of RNAPII, thus favoring escape from the the new structure (Selby and Peterlin, 1990; Southgateet al., 1990).pause (Figure 7). In the view that multiple laterally trans-

located states of RNAP exist in equilibrium (positional Other arguments for or against a role of pausingare contradictory and currently unresolved. Jeang andequilibrium; Guajardo and Sousa, 1997; Landick, 1997;

Komissarova and Kashlev, 1997), this would trap the Berkhout (1992) argue against a role of pausing in Tatrecruitment because placement of a self-cleaving ham-conformation with a 39 OH properly positioned for NTP

addition. This antipausing effect of TAR is similar to merhead RNA structure just downstream from the 160region, but not 500 bp later, abolishes transactivation.the inhibition of arrest caused by pairing of antisense

Molecular Cell1040

Oslo, Norway) or on streptavidin-agarose (Sigma, St. Louis, MO)Because removing TAR from the TC soon after it formsthrough a streptavidin-biotin linkage according to the manufactur-blocks transactivation, they conclude that Tat must ordi-ers’ recommendations.narily bind TAR after RNAPII moves past the 160 to

1100 region. However, this interpretation contradicts Proteins and Substratesthe earlier findings of Berkhout et al. (1989), who con- HeLa cells were obtained from the Cell Culture Center (Connrapids,cluded that Tat must bind TAR before nascent RNA MN). RNAPIIa was purified from frozen calf thymus (ANTEC, Tyler,

TX) using immobilized 8WG16 antibody as described by Thompsonaround 160 region emerges from RNAPII because inser-and Burgess (1996). RNAPIIo was purified from HeLa nuclei as de-tion at 160 of a sequence known to destroy the TARscribed by Lu et al. (1992). Transcription factor TFIIS was overex-structure by base-pairing with the 59 side of its stempressed in E. coli and purified as described by Yoo et al. (1991).

does not block transactivation. It seems most likely that Purified NTPs and 29-dNTPs were obtained from Pharmacia (Piscat-Tat binds TAR when or soon after it forms (for instance away, NJ) and 39-dNTPs from Boehringer Mannheim Biochemicalsat the pause site), but that the subsequent events of (Indianapolis, IN).transactivation such as RNAPII phosphorylation occur

Promoter-Dependent Transcriptionas RNAPII moves further downstream, and that removalPreinitiation complexes were formed by incubating the immobilizedof TAR (and Tat) when a hammerhead nuclease is pres-DNA template (0.5 mg) with 2 ml of HeLa nuclear extract (Shapiro et

ent blocks completion of those steps. Keen et al. (1997) al., 1988) for 20 min at 308C in 16 ml of transcription buffer (10 mMhave shown that these events include transfer of Tat HEPES [pH 7.9], 33 mM KCl, 8 mM MgCl2, 0.1 mM Na2EDTA, 1 mMto stable interaction with RNAPII after the initial and DTT, 6 %, v/v glycerol) and 1 U of Inhibitace (59-39, Boulder, CO).

Transcription was allowed to initiate for 1 min at 308C after additionnecessary Tat–TAR interaction, but prior to templateof NTPs in 4 ml of transcription buffer to give final concentrationsposition z175 where TAR can be removed by nucleaseof 100 mM each 29dATP, CTP, and GTP and 0.5 mM [a-32P]UTP (10digestion without loss of Tat or transactivation.mCi). The reaction was diluted with 10 vol of 0.15% sarkosyl inIn contrast to the conditions used in relevant experi- BC100 buffer (20 mM Tris–HCl [pH 8.0], 100 mM KCl, 10 mM

ments to date, it seems most likely that transcriptional b-mercaptoethanol, 0.2 mM Na2EDTA, 20% v/v glycerol). The TCspausing would play a role in HIV-1 transactivation early were collected by centrifugation at 2000 rpm for 2 min at 48C,

washed with 5 vol of BC100 buffer without sarkosyl, then washedin HIV-1 replication, when premature termination ofwith transcription buffer containing acetylated BSA (1 mg/ml), andHIV-1 transcription keeps Tat levels low. In these condi-suspended in the same buffer (20 ml). For pause half-life measure-tions, recruitment of Tat to the complex prior to prema-ments, the U14 TCs were then elongated at 308C in the presence

ture termination should be inefficient. The pause signal, of all 4 NTPs at 1 mM each. Samples were removed at the indicatedwhich slows RNAPII by a factor of .100 relative to its times and mixed with 100 ml of stop buffer (125 mM Tris–HCl [pHaverage elongation rate, could increase the effective 8.0], 15 mM Na2EDTA, 333 mM NaCl, 1.25% SDS, 170 mg tRNA/ml).

The samples were extracted with phenol:chloroform (1:1), ethanolconcentration of TCs at the critical position for Tat re-precipitated, dissolved at 908C for 1 min in formamide loading dyecruitment by a corresponding amount. Our results es-(95% formamide, 15 mM Tris–HCl [pH 7.9], 5 mM EDTA, 0.1% brom-tablish that the HIV21 pause site defines the positionphenol blue, 0.1% xylene cyanol), and separated on an 11% poly-at which the nascent RNA can rearrange to TAR, interactacrylamide gel (19:1) containing 8 M urea and 44 mM Tris·borate

with Tat, and initiate the events leading to transacti- [pH 8.3]. Gels were analyzed with a Phosphorimager (Molecularvation. Dynamics, Sunnyvale, CA) and quantitated as described previously

(Landick et al., 1996b).

Experimental ProceduresTranscription with Purified Calf Thymus RNAPIIPurified RNAPII (250 ng) was preincubated with 1 mg of immobilizedTemplate DNAtailed template (see above) for 10 min at 308C in 15 ml of RNAPIIThe 774 bp DNA fragment used for transcription of the wild-typebuffer (70 mM HEPES–KOH [pH 7.9], 50 mM NH4Cl, 6 mM MgCl2,HIV-1 pause signal was amplified by PCR from plasmid pLL2830.15 mM DTT, 5 mM spermidine, 20 %; v/v glycerol, 1 U of Inhibitace)using universal M13 forward and biotinated reverse primers. pLL283containing 400 mM UpG dinucleotide (Kadesch and Chamberlin,was constructed in multiple steps from plasmid pBSIIKS- (Stra-1982; Lang et al., 1994). Transcription was allowed to initiate for 1tagene; La Jolla, CA) by insertion of the following DNA fragmentsmin at 308C by the addition of 5 ml of NTP mix in transcription bufferinto the HincII site (listed from the forward primer side toward theto give final concentrations of 100 mM each CTP, GTP, and 0.5 mMreverse primer side of pBSIIKS2): a 220 bp ScaI–HindIII fragment[a-32P]UTP (10 mCi). The TCs were washed three times with 10 volcontaining the promoter and early transcribed region from HIV-1; aof RNAPII buffer, suspended in 20 ml of the same buffer, and allowed36 bp HindIII–SmaI fragment from the polylinker region of pUC19;to transcribe in the presence of all 4 NTPs (1 mM each). At thea 292 bp RsrII(blunt)–BsmI(blunt) fragment from exon 1 and intronindicated times, aliquots were removed, processed, electropho-1 of the human c-myc gene.resed, and analyzed as described above.Nucleotide changes in pLL283 were constructed by PCR-ampli-

fying a DNA fragment from pLL283 using the M13 forward or reverseRNA Sequencingprimer and a second primer that specified the change and thatRNAsequence ladders were generatedby elongating the 32P-labeledoverlapped the unique NheI site inTAR. After digestionwith NheI andU14 TCs in the presence of all 4 NTPs (1 mM each) and a giveneither EcoRI (forward-primer-amplified fragments) or XhoI (reverse-39dNTP (0.8 mM) at 308C for 20 min (Figure 1). After processingprimer-amplified fragments), fragments were ligated between theof samples as described above, the RNA was separated in a 6%appropriate sites in pLL283.polyacrylamide, 8 M urea gel next to pause and arrest RNAs.The template for purified RNAPII (Figure 2) was constructed by

first introducing into pLL283 a unique BsmBI site that could producea 4 nt 59 staggered cut beginning at the HIV-I transcription start Stepwise Transcription and 39-Proximal Transcript Labeling

U14 TCs were formed on the HIV-1 template immobilized on strep-site. A 20 bp synthetic DNA fragment with a 10 nt 39 overhang and4 nt 59 overhang (59-GGGTCTCTTCTCGGTCGTCT top strand; 39- tavidin-agarose beads as described above, either with [a-32P]UTP

(Figures 4 and 6) or with all nonradioactive NTPs (Figure 5). The U14accaaaaaaaCCCAGAGAAGAGCCAGCAGAccca bottom strand; du-plex portion in uppercase) was then ligated to a BsmBI-digested complexes were moved to position A22 by incubation with ATP,

GTP, and CTP (50 mM each) for 5 min at 308C. The complexes werePCR fragment generated from the new plasmid as described above.DNA templates were immobilized either on Dynabeads (Dynal, then washed four times with 10 vol of transcription buffer to remove

Pausing Modulates HIV-1 Nascent RNA Folding into TAR1041

the unincorporated NTPs and moved stepwise along the DNA by translocation of the elongation complex on DNA. J. Biol. Chem. 268,25604–25616.repeated incubation with different sets of 3 NTPs at 50 mM each

separated by washing (shown in Figure 4 with 59-proximal 32P-UMP- Gu, W., Wind, M., and Reines, D. (1996). Increased accommodationlabeling). To label RNAs near their 39 ends (Figure 5), [a-32P]CTP of nascent RNA in a product site on RNA polymerase II during arrest.was included in the final step of the procedure or in the next to final Proc. Natl. Acad. Sci. USA 93, 6935–6940.step (C64 TCs). Guajardo, R., and Sousa, R. (1997). A model for the mechanism of

polymerase translocation. J. Mol. Biol. 265, 8–19.Structural Probing of the Pause RNA Structure Izban, M.G., and Luse, D.S. (1991). Transcription on nucleosomalwith Ribonuclease T1 templates by RNA polymerase II in vitro: inhibition of elongationGel-purified C42 or C59 RNAs (10 ml each) or C59, U62(paused), or with enhancement of sequence-specific pausing. Genes Dev. 5,C64 TCs (30 ml each) were incubated at 308C with RNase T1 (20 683–696.U/ml) in transcriptionbuffer with orwithout 4 M ureafor 10,20, or 30 s Jeang, K.T., and Berkhout, B. (1992). Kinetics of HIV-1 long terminal(see legend to Figure 5). The reaction was terminated by addition to repeat trans-activation. Use of intragenic ribozyme to assess rate-an equal volume of phenol:chloroform (1:1). Samples were pro- limiting steps. J. Biol. Chem. 267, 17891–17899.cessed, separated on a polyacrylamide (20%; 19:1)–urea (8M) gel,

Jeong, S.W., Lang, W.H., and Reeder, R.H. (1996). The yeast tran-and analyzed as described above.

scription terminator for RNA polymerase I is designed to preventpolymerase slippage. J. Biol. Chem. 271, 16104–16110.

Acknowledgments Jones, K.A. (1997). Taking a new TAK on Tat transactivation. GenesDev. 11, 2593–2599.

We thank Irina Artsimovitch, Kati Geszvain, Chad Metcalf, RachelJones, K.A., and Peterlin, B.M. (1994). Control of RNA initiation andMooney, Vladimir Svetlov, and Wilma Ross for helpful discussionselongation at the HIV-1 promoter. Annu. Rev. Biochem. 63, 717–743.and comments on the manuscript. This work was supported byKadesch, T.R., and Chamberlin, M.J. (1982). Studies of in vitro tran-grants from the National Institutes of Health (GM38660) and thescription by calf thymus RNA polymerase II using a novel duplexNational Science Foundation (MCB9696042).DNA template. J. Biol. Chem. 257, 5286–5295.

Kamine, J., Subramanian, T., and Chinnadurai, G. (1991). Sp1-Received February 10, 1998; revised April 13, 1998.dependent activation of a synthetic promoter by human immunode-ficiency virus type 1 Tat protein. Proc. Natl. Acad. Sci. USA 88,

References 8510–8514.

Kashlev, M., Martin, E., Polyakov, A., Severinov, K., Nikiforov, V., andBentley, D.L. (1995). Regulation of transcriptional elongation by RNA Goldfarb, A. (1993). Histidine-tagged RNA polymerase: dissection ofpolymerase II. Curr. Opin. Genet. Dev. 5, 210–216. the transcription cycle using immobilized enzyme. Gene 130, 9–14.Berkhout, B., Silverman, R.H., and Jeang, K.T. (1989). Tat trans- Kato, H., Sumimoto, H., Pognonec, P., Chen, C.H., Rosen, C.A., andactivates the human immunodeficiency virus through a nascentRNA Roeder, R.G. (1992). HIV-1 Tat acts as a processivity factor in vitro intarget. Cell 59, 273–282. conjunction with cellular elongation factors. Genes Dev. 6, 655–666.Chan, C., Wang, D., and Landick, R. (1997). Spacing from the tran- Keen, N., Churcher, M., and Karn, J. (1997). Transfer of Tat andscript 39 end determines whether a nascent RNA hairpin interacts release of TAR RNA during the activation of the human immunodefi-with RNA polymerase to prolong pausing or triggers termination. J. ciency virus type-1 transcription elongation complex. EMBO J. 16,Mol. Biol. 268, 54–68. 5260–5272.Chan, C.L., andLandick, R. (1993). Dissection of the his leader pause Kessler, M., and Mathews, M.B. (1992). Premature termination andsite by base substitution reveals a multipartite signal that includes processing of human immunodeficiency virus type 1-promoted tran-a pause RNA hairpin. J. Mol. Biol. 233, 25–42. scripts. J. Virol. 66, 4488–4496.Chan, C.L., and Landick, R. (1994). New perspectives on RNA chain Komissarova, N., and Kashlev, M. (1997). RNA polymerase switcheselongation and termination by E. coli RNA polymerase. In Transcrip- between inactivated and activated states by translocating back andtion:Mechanisms and Regulation, R.C. Conaway and J.W.Conaway, forth along the DNA and the RNA. J. Biol. Chem. 272, 15329–15338.eds. (New York: Raven Press), pp. 297–321. Krumm, A., Hickey, L.B., and Groudine, M. (1995). Promoter-proxi-Churcher, M.J., Lamont, C., Hamy, F., Dingwall, C., Green, S.M., mal pausing of RNA polymerase II defines a general rate-limitingLowe, A.D., Butler, JG, Gait, M.J., and Karn, J. (1993). High affinity step after transcription initiation. Genes Dev. 9, 559–572.binding of TAR RNA by the human immunodeficiency virus type-1 Landick, R. (1997). RNA polymerase slides home: pause and termi-tat protein requires base-pairs in the RNA stem and amino acid nation site recognition. Cell 88, 741–744.residues flanking the basic region. J. Mol. Biol. 230, 90–110.

Landick, R., Turnbough, C., Jr., and Yanofsky, C. (1996a). Transcrip-Cujec, T., Okamoto, H., Fujinaga, K., Meyer, J., Chamberlin, H., tion attenuation. In Escherichia coli and Salmonella: Cellular andMorgan, D., and Peterlin, B. (1997). The HIV transactivator TAT binds Molecular Biology, 2nd Edition, F. Neidhardt, R. Curtiss, III, J. Ingra-to the CDK-activating kinase and activates the phosphoyrlation of ham, E. Lin, K. Low, B. Magasanik, W. Reznikoff, M. Riley, M.the carboxy-terminal domain of RNA polmyerase II. Genes Dev. 11, Schaechter, and H. Umbarger, eds. (Washington, DC: ASM Press),2645–2657. pp. 1263–1286.Feng, G., Lee, D.N., Wang, D., Chan, C.L., and Landick, R. (1994). Landick, R., Wang, D., and Chan, C. (1996b). Quantitative analysisGreA-induced transcript cleavage in transcription complexes con- of transcriptional pausing by RNA polymerase: the his leader pausetaining Escherichia coli RNA polymerase is controlled by multiple site as a paradigm. Methods Enzymol. 274, 334–352.factors, including nascent transcript location and structure. J. Biol. Lang, W.H., Morrow, B.E., Ju, Q., Warner, J.R., and Reeder, R.H.Chem. 269, 22282–22294. (1994). A model for transcription termination by RNA polymerase I.

Cell 79, 527–534.Garcia-Martinez,L.F., Mavankal, G.,Neveu, J.M., Lane, W.S., Ivanov,D., and Gaynor, R.B. (1997). Purification of a Tat-associated kinase Lee, D.N., and Landick, R. (1992). Structure of RNA and DNA chainsreveals a TFIIH complex that modulates HIV-1 transcription. EMBO in paused transcription complexes containing Escherichia coli RNAJ. 16, 2836–2850. polymerase. J. Mol. Biol. 228, 759–777.Graeble, M.A., Churcher, M.J., Lowe, A.D., Gait, M.J., and Karn, J. Linn, S.C., and Luse, D.S. (1991). RNA polymerase II elongation(1993). Human immunodeficiency virus type 1 transactivator protein, complexes paused after the synthesis of 15- or 35-base transcriptstat, stimulates transcriptional read-through of distal terminator se- have different structures. Mol. Cell. Biol. 11, 1508–1522.quences in vitro. Proc. Natl. Acad. Sci. USA 90, 6184–6188. Lu, H., Zawel, L., Fisher, L., Egly, J.M., and Reinberg, D. (1992).Gu, W., Powell, W., Mote, J., Jr., and Reines, D. (1993). Nascent RNA Human general transcription factor IIH phosphorylates the C-termi-

nal domain of RNA polymerase II. Nature 358, 641–645.cleavage by arrested RNA polymerase II does not require upstream

Molecular Cell1042

Marciniak, R.A., Calnan, B.J., Frankel, A.D., and Sharp, P.A. (1990). Wu, F., Garcia, J., Sigman, D., and Gaynor, R. (1991). Tat regulatesHIV-1 Tat protein trans-activates transcription in vitro. Cell 63, binding of the human immunodeficiency virus trans-activating re-791–802. gion RNA loop-binding protein TRP-185. Genes Dev. 5, 2128–2140.

Mavankal, G., Ignatius Ou, S.H., Oliver, H., Sigman, D., and Gaynor, Yoo, O.J., Yoon, H.S., Baek, K.H., Jeon, C.J., Miyamoto, K., Ueno,R.B. (1996). Human immunodeficiency virus type1 and2 Tat proteins A., and Agarwal, K. (1991). Cloning, expression and characterizationspecifically interact with RNA polymerase II. Proc. Natl. Acad. Sci. of the human transcription elongation factor, TFIIS. Nucleic AcidsUSA 93, 2089–2094. Res. 19, 1073–1079.

Nudler, E., Mustaev, A., Lukhtanov, E., and Goldfarb, E. (1997). The Zhu, Y., Pe’ery, T., Peng, J., Ramanathan, Y., Marshall, N., Marshall,RNA:DNA hybrid maintains the register of transcription by pre- T., Amendt, B., Mathews, M., and Price, D. (1997). Transcriptionventing backtracking of RNA polymerase. Cell 89, 33–41. elongation factor P-TEFb is required for HIV-1 Tat transactivation

in vitro. Genes Dev. 11, 2622–2632.O’Brien, T., and Lis, J.T. (1991). RNA polymerase II pauses at the59 end of the transcriptionally induced Drosophila hsp70 gene. Mol. Zuker, M., and Stiegler, P. (1981). Optimal computer folding of largeCell. Biol. 11, 5285–5290. RNA sequences using thermodynamics and auxilliary information.

Nucleic Acids Res. 9, 133–148.Parada, C.A., and Roeder, R.G. (1996). Enhanced processivity ofRNA polymerase II triggered by Tat-induced phosphorylation of itscarboxy-terminal domain. Nature 384, 375–378.

Parada, C.A., Yoon, J.B., and Roeder, R.G. (1995). A novel LBP-1-mediated restriction of HIV-1 transcription at the level of elongationin vitro. J. Biol. Chem. 270, 2274–2283.

Reeder, T., and Hawley, D. (1996). Promoter proximal sequencesmodulate RNA polymerase II elongation by a novel mechanism. Cell87, 767–777.

Reines, D., Conaway, J.W., and Conaway, R.C. (1996). The RNApolymerase II general elongation factors. Trends Biochem. Sci. 21,351–355.

Rice, G.A., Kane, C.M., and Chamberlin, M.J. (1991). Footprintinganalysis of mammalian RNA polymerase II along its transcript as analternate view of transcription elongation. Proc. Natl. Acad. Sci. USA88, 1245–1249.

Ring, B., Yarnell, W., and Roberts, J. (1996). Function of E. coliRNA polymerase s factor s70 in promoter-proximal pausing. Cell 86,485–493.

Ring, B.Z., and Roberts, J.W. (1994). Function of a nontranscribedDNA strand site in transcription elongation. Cell 78, 317–324.

Rougvie, A., and Lis, J. (1990). Postinitiation transcriptional controlin Drosophila melanogaster. Mol. Cell. Biol. 10, 6041–6045.

Selby, M.J., and Peterlin, B.M. (1990). Trans-activation by HIV-1 Tatvia a heterologous RNA binding protein. Cell 62, 769–776.

Shapiro, D., Wahli, W., and Keller, M. (1988). A high-efficiency HeLacell nuclear transcription system. DNA 7, 47–55.

Southgate, C., Zapp, M.L., and Green, M.R. (1990). Activation oftranscription by HIV-1 Tat protein tethered to nascent RNA throughanother protein. Nature 345, 640–642.

Southgate, C.D., and Green, M.R. (1991). The HIV-1 Tat proteinactivates transcription from an upstream DNA-binding site: implica-tions for Tat function. Genes Dev. 5, 2496–2507.

Sugimoto, N., Nakano, S., Katoh, M., Matsumura, A., Nakamuta, H.,Ohmichi, T., Yoneyama, M., and Sasaki, M. (1995). Thermodynamicparameters to predict stability of RNA/DNA hybrid duplexes. Bio-chemistry 34, 11211–11216.

Thompson, N.E., and Burgess, R.R. (1996). Immunoaffinity puri-fication of RNA polymerase II and transcription factors usingpolyol-responsive monoclonal antibodies. Methods Enzymol. 274,513–526.

Tuerk, C., Gauss, P., Thermes, C., Groebe, D.R., Gayle, M., Guild,N., Stormo, G., d’Aubenton-Carafa, Y., Uhlenbeck, O.C., Tinoco, I.J.,Brody, E.N., and Gold, L. (1988). CUUCGG hairpins: extraordinarilystable RNA secondary structures associated with various biochemi-cal processes. Proc. Natl. Acad. Sci. USA 85, 1364–1368.

Valegard, K., Murray, J.B., Stockley, P.G., Stonehouse, N.J., andLiljas, L. (1994). Crystal structure of an RNA bacteriophage coatprotein-operator complex. Nature 371, 623–626.

Wang, D., and Landick, R. (1997). A nascent RNA hairpin that stimu-lates transcriptional pausing interacts with RNA polymerase nearamino acid 900 of the b subunit. Proc. Natl. Acad. Sci. USA 94,8433–8438.

Wei, P., Garber, M., Fang, S.-M., Fischer, W., and Jones, K. (1998).A novel CDK9-associated C-type cyclin interacts directly with HIV-1Tat and mediates in high-affinity, loop-specific binding to TAR RNA.Cell 92, 451–462.