Preferential interaction of the his pause RNA hairpin with RNA polymerase β subunit residues 904–950 correlates with strong transcriptional pausing

Abstract

RNA secondary structures (hairpins) that form as the nascent RNA emerges from RNA polymerase are important components of many signals that regulate transcription, including some pause sites, all ρ-independent terminators, and some antiterminators. At the his leader pause site, a 5-bp-stem, 8-nt-loop pause RNA hairpin forms 11 nt from the RNA 3′ end and stabilizes a transcription complex conformation slow to react with NTP substrate. This stabilization appears to depend at least in part on an interaction with RNA polymerase. We tested for RNA hairpin interaction with the paused polymerase by crosslinking 5-iodoUMP positioned specifically in the hairpin loop. In the paused conformation, strong and unusual crosslinking of the pause hairpin to β904–950 replaced crosslinking to β′ and to other parts of β that occurred in nonpaused complexes prior to hairpin formation. These changes in nascent RNA interactions may inhibit reactive alignment of the RNA 3′ end in the paused complex and be related to events at ρ-independent terminators.

The his leader pause hairpin is a well characterized nascent RNA structure that modulates RNA chain elongation (1–4). It forms concomitantly with a paused conformation of the transcription complex midway through the his biosynthetic operon leader region, where it increases the lifetime of the paused complex at least 6-fold. The pause allows time for a ribosome to initiate synthesis of the leader peptide and release the paused polymerase (probably by disrupting the pause hairpin), thereby synchronizing ribosome and polymerase movement during transcriptional attenuation (5). This and other pause sites are regulatory timing mechanisms that both prevent transcription beyond a region where regulatory input is effective and put polymerase in the proper conformation to interact with regulatory molecules.

Similar hairpins are involved in some, but not all, pause signals; are required at ρ-independent terminators (6–8), where they trigger dissociation of the transcription complex; and alone (e.g., HK022 put; ref. 9) or in association with proteins (e.g., λ N-nut complex; ref. 8) can function as antiterminators, which modify the transcription complex to block recognition of both pause sites and terminators. Nascent RNA hairpins also are components of some eukaryotic regulatory mechanisms (e.g., HIV TAR RNA; ref. 10) and can trigger pausing or termination by RNA polymerase II in vitro (11). Although RNA hairpin–RNA polymerase interactions have long been suggested to play roles in pausing (12), termination (13–15), and antitermination (16), no direct evidence for these interactions or their effects exists.

The his pause RNA hairpin offers an excellent paradigm to study these interactions. It is one of four components of the multipartite his pause signal that also depends on a 3′-proximal region of RNA or DNA, alignment of the 3′-terminal nucleotide and the incoming NTP, and duplex DNA that contacts polymerase downstream from the active site (1, 2). Both the his pause signal and ρ-independent terminators (which differ chiefly by shorter, U-rich 3′-proximal regions) destabilize DNA and RNA contacts in transcription complexes approaching the pause site, causing a propensity for backward sliding of the enzyme along the RNA and DNA chains (3, 17–20). The potential for backtracking may slow translocation and thus dictate the position of hairpin formation (20); conversely, the hairpin, once formed, blocks backtracking in the rearranged complex. At the pause site, this complex is stable although poorly reactive with NTP substrate; at terminators, it dissociates, presumably because the remaining contacts are insufficient to stabilize the ternary complex (4).

Interaction of the pause hairpin appears to be at least partially ionic and to involve a region of RNA polymerase that is easily disordered because certain anions and chaotropes (both ionic and nonionic) reduce pausing but only when the pause hairpin is intact (21). Likewise, several results suggest that terminator hairpins may interact with polymerase: (i) conserved sequences in the stem region are important for efficient termination (22, 23); (ii) termination efficiency is often salt-sensitive (24, 25); and (iii) both the size of most terminator hairpins (5- to 9-bp stem; 3- to 5-nt loop) and distance between the hairpin and release point (7–9 nt) are remarkably invariant (22, 26–28). Although several segments within both the β and β′ subunits of RNA polymerase are known to influence pausing and termination (29–31), neither the site of hairpin interaction nor the mechanism by which it might slow elongation are known. We have tested for the putative RNA secondary structure interaction by crosslinking the his pause RNA hairpin to RNA polymerase.

MATERIALS AND METHODS

Proteins and DNAs.

His₆-RNA polymerase was prepared as described previously (3). DNA templates were prepared from pCL102 (wild type) and pDW308 (MS-hairpin) (3, 32) by PCR amplification with primers that hybridized upstream and downstream from the promoter-leader segment: #947, 5′-GAGAGACAACTTAAAGAGAC (98 bp upstream of the transcription start site); #1454, 5′-GGCTTTCTCATGCGTTCATGC (132 bp downstream of transcription start site). The resulting 230-bp PCR fragment contained the T7 A1 promoter and a WT or MS-hairpin pause site at +71 (Fig. 1; see ref. 3).

Figure 1.

his leader pause RNA hairpin and strategy for crosslinking to RNA polymerase. (A) Wild-type pause hairpin (WT) and pause site. Positions at which 5-iodoUMP and [α³²P]CMP were incorporated during preparation of the paused complexes and at which RNase T₁ cuts to generate the crosslinked oligonucleotide are indicated. (B) The multisubstituted pause hairpin (MS), which decreases pause half-life by a factor of ≈6 (1). (C) Method for preparing and crosslinking transcription complexes.

In Vitro Transcription Reactions.

In vitro transcription reactions using His₆-tagged polymerase immobilized on Ni²⁺-agarose beads were performed as described previously (3). To incorporate 5-iodoUMP at positions U54–56 (Fig. 1A), U52 complexes were first prepared with nonradioactive NTPs by stepwise transcription and washing, then incubated with 1 μM [α-³²P]CTP (NEN, 3,000 Ci/mmol; 1 Ci = 37 GBq) and 10–20 μM 5-iodoUTP (Sigma) for 5 min at 25°C, and then incubated with 5 μM unlabeled CTP for an additional 3 min to ensure extension to position C57. The C57 complexes containing [³²P]CMP and 5-iodoUMP were elongated to desired positions by further stepwise transcription with nonradioactive NTPs. Incorporation of 5-iodoUMP at other positions was accomplished similarly.

Photocrosslinking Reactions.

Crosslinking reactions were performed essentially as described by Stump and Hall (33). After halting transcription complexes at the desired position, the beads containing transcription complexes were placed in a 1.5-ml microfuge tube cap in 100 μl transcription buffer (TB; 20 mM Tris⋅acetate (pH 8.0), 20 mM MgCl₂, 20 mM NaCl, 7 mM 2-mercaptoethanol, 2% glycerol, 25 μg acetylated BSA/ml) and exposed to UV light (312 nm maximum, 300 nm cutoff; ≈30 mW/cm²) for 40 min at 4°C. After crosslinking, the beads were recovered and resuspended in TB for further analysis.

Electrophoresis of Crosslinked Proteins and Fragments.

After crosslinking, samples were either directly mixed with an equal volume of loading buffer (0.125 M Tris⋅HCl, pH 6.8/4% SDS/20% glycerol/1.43 M 2-mercaptoethanol/0.2% bromphenol blue), or first treated with trypsin (see below) or RNase T₁ at 65°C for 10 min, or both. After heating to 100°C for 5 min, the samples were electrophoresed through a 11 × 14 × 0.75-cm 8% polyacrylamide SDS gel with 3% stacking gel (separating gel, 33.5:0.3 acrylamide:bisacrylamide; stacking gel, 30:0.44 acrylamide:bisacrylamide) in 1 × SDS running buffer (25 mM Tris/192 mM glycine/0.1% SDS) for ≈4 hr at 5 W and analyzed using a PhosphorImager. To recover crosslinked proteins for further analysis, gel slices were excised, washed twice with H₂O, and soaked in 1% SDS overnight at 37°C. The eluents were combined with 5 μg purified β subunit as carrier, extracted with 1-butanol to remove all aqueous phase, centrifuged to pellet the protein, dried in vacuo after removing the supernatant, and resuspended in 10 μl 1% SDS.

Trypsin Digestion.

After crosslinking, samples were washed once with 1.5 ml of cold TB, digested with 0.15 μM trypsin in 50 μl TB + 150 mM NaCl overnight at 24°C, and then stopped by addition of Trypsin Inhibitor (Boehringer Mannheim) to 4 μM.

CNBr Digestion.

CNBr digestion was performed in 30 μl of 70% formic acid/1% SDS with 0.3 M CNBr (final) at room temperature for 24 hr (34). The peptides were extracted to dryness with 1-butanol, washed with cold 70% ethanol, dried in vacuo, and resuspended in 10 μl 0.5× loading buffer. The samples were then electrophoresed through a 10–20% gradient polyacrylamide SDS gel (Novex) and compared with the CNBr fragments of purified β subunit. β subunit was purified by preparative SDS/PAGE of inclusion bodies from a 500-ml saturated culture of JM109 pRL385 (35) in a Bio-Rad Model 491 preparative cell (1 cm stacking gel, 4%, 30:0.44 acrylamide:bisacrylamide; 6 cm separating gel, 6%, 33.5:0.3 acrylamide:bisacrylamide). Fractions containing β subunit were combined and stored at 1 mg/ml and −80°C until use.

RESULTS

The Pause Hairpin Makes Strong and Unusual Contact to a Paused RNA Polymerase.

To characterize the interactions of the his pause hairpin with RNA polymerase, we used a stepwise, immobilized transcription method to incorporate the photocrosslinker 5-iodoUMP and [³²P]CMP specifically into the loop of the nascent pause RNA hairpin (see Materials and Methods). 5-IodoUMP can be efficiently activated at >308 nm, which reduces nonspecific crosslinking and protein damage dramatically (33, 36). Substitution of UMP with 5-iodoUMP throughout the his pause hairpin had no effect on the half-life of pausing at the his pause site (data not shown).

We incorporated 5-iodoUMP residues at U54–U56 in the loop of the pause hairpin, where they were flanked by two [³²P]CMP residues within an eight-base oligonucleotide that could be separated from the rest of the pause RNA by RNase T₁ digestion (Fig. 1 A and C; see Materials and Methods). In parallel, we used a template encoding three base substitutions that destabilize the pause hairpin (MS hairpin, ref. 1; Fig. 1B). RNA polymerase pauses strongly before addition of G72 (Pause, Fig. 1A) and weakly before U71 (32); the MS hairpin reduces the pause half-life by a factor of ≈6. When transcription complexes that contain just these three iodoUMP and two [³²P]CMP residues were UV irradiated at the pause site or 2 bp preceding it (−2 complex), crosslinking occurred exclusively to the β and β′ subunits of RNA polymerase (Fig. 2A; position of α not shown, but see Fig. 3B). We estimated the efficiency of crosslinking to be ≈5% at the pause site; no crosslinking was observed when 5-iodoUMP or UV irradiation was omitted (data not shown).

Figure 2.

Detection of pause hairpin crosslinking to RNA polymerase. Transcription complexes containing [³²P]CMP (at 53 and 57) and 5-iodoUMP (at 54–56) were formed at various positions on the WT or MS templates (Fig. 1 A and B) by stepwise transcription, UV-irradiated, and separated by PAGE (see Materials and Methods). (A) RNase T₁ digestion and nucleotide addition in crosslinked, paused complexes. Lanes 1, 3, and 5, −2, wild-type paused complexes (P), and MS paused complexes without RNase T₁ digestion. Lanes 2, 4, and 6, same complexes after RNase T₁ digestion. Lane 7, paused complex prepared with CMP in place of [³²P]CMP, crosslinked, and then incubated with 1 μm [α-³²P]ATP. Lane 8, as lane 7 but with addition of 10 μm GTP during incubation. (B) Complexes formed at −8, −2, −1, pause, +2, +4, and +6 on the wild-type template, UV irradiated, and treated with RNase T₁.

Figure 3.

Localization of the pause hairpin crosslink. (A) Split-β polymerases. Pause complexes were generated using either wild-type (lane 1) or split-β polymerases (S1, lane 2; S2, lanes 3 and 4) on the WT (lanes 1–3) or MS templates (lane 4). Samples were prepared, UV irradiated, treated with RNase T₁, and electrophoresed as described in Materials and Methods. The band labeled BSA resulted from nonspecific crosslinking to acetylated BSA, as evidenced by its disappearance when BSA was omitted (e.g., lane 4). (B) Trypsin digestion. After crosslinking, −2 and paused complexes were either digested with trypsin (lanes 2, 4–6) or left intact (lanes 1 and 3). The samples were then either treated (lanes 1–4) or not treated (lanes 5 and 6) with RNase T₁ and resolved on a 12% polyacrylamide SDS gel. (C) Complete CNBr cleavage. Crosslinked β subunit and β904–1342 fragment were generated as in B, gel purified, subjected to CNBr cleavage, and resolved on a 10–20% gradient gel (see Materials and Methods). Other weak bands in the lanes reflect partially digested fragments (lanes 2 and 3) or radiodegradation of isolated β subunit (lane 4). (D) Location of pause hairpin crosslink relative to other features of the β subunit. A–I (light grey), conserved regions; DR1 and 2, dispensable regions 1 and 2; S, streptolydigin-resistance substitutions (black); R, rifampicin-resistance substitutions and region likely contacting nascent RNA (dark grey); AS, near active site by crosslinking; DD, near downstream DNA. ∗, location of radioactivity from pause hairpin crosslink in β subunit fragments.

Relative to the −2 complex, the crosslinking pattern of the paused complex and, to a lesser extent, the MS paused complex suggest relocation of the nascent RNA from β′ to β as well as novel β contact. Prior to digestion with RNase T₁, much of the radioactivity in the paused samples migrated more slowly than the β or β′ subunits from the −2 complex (Fig. 2A, lanes 1, 3, and 5). Digestion with RNase T₁ converted the novel bands to crosslinked β that comigrated with β from the −2 sample, but of greater intensity (β:β′ = ≈1:1, 15:1, 4:1, for −2, WT, and MS, respectively; lanes 2, 4, and 6, Fig. 2A; see Table 1). The absence of the shifted β band in the −2 complex sample and its lesser amount and different position in the MS hairpin sample suggest that the aberrant mobility results from unusual contacts of the pause hairpin that are unique to the paused conformation and that, upon crosslinking, are retained in the presence of SDS. For example, the crosslinked hairpin could trap part or all of β (or even β and β′) in an SDS-resistant, folded conformation. Because the MS hairpin template directs weak pausing (1, 3), it may weaken but not abolish these unusual contacts in the paused complex.

Table 1.

Percentage of RNA crosslinking at different positions

Crosslinker	Complex	Site of crosslinking*
		β1–903		β904–1,342	β′
U54–56	−6	5		45	50
	−5	5		30	65
	−4	5		5	90
	−3	0		20	80
	−2	0		60	40
	WT	5		90	5
	MS	10		70	20
	+2		60†		40
	+4		40		50
	+6		50		50
U52	WT	5		85	10
	MS	20		60	20
U49	WT	5		80	15
	MS	25		50	25

^*The percentage of ³²P crosslinked to β and β′ was determined from PhosphorImager scans of gels containing RNase T₁-digested samples, assuming ±5% accuracy. The percentage localized on each β trypsin fragment was then determined from scans of trypsin-digested samples. 

^†Trypsin digestion was not performed on the +2, +4, and +6 samples. 

Similar experiments with complexes positioned throughout the pause region revealed that the change to predominant β contact occurred only at the −1 and pause positions (Fig. 2B; supershifting of β was similarly specific to −1 and pause; data not shown). Minus 1 and pause correspond exactly to the positions where weak and strong pausing, partial and complete hairpin formation, and stabilization against backtracking were observed previously (3, 32). Thus, formation of the pause RNA hairpin in the paused complex appears to relocate the RNA away from β′ into an unusual contact to β that alters its gel mobility upon crosslinking.

To test whether crosslinking inactivated the paused complexes, we attempted to extend and label the RNA in nonradioactive paused complexes after UV treatment. Incubation of an unlabeled sample with GTP and [α-³²P]ATP labeled both β and β′ (Fig. 2A, lane 7). No labeling occurred if GTP was omitted (Fig. 2A, lane 8), demonstrating that ³²P incorporation resulted from extension of the crosslinked pause RNA and not from a nonspecific reaction. Thus, the crosslinked complexes remained paused and did not become arrested. The decrease in efficiency of labeling the shifted β band may reflect the reduced reactivity of the paused complex. Further, because labeling of the shifted β species required extension to at least +3, even though +2 and beyond complexes did not produce the β supershift when crosslinked directly (not shown), supershifting of β must reflect the properties of the paused complex and not the structure of the RNA attached to β.

The Principal Pause Hairpin Contact Is to β Subunit Residues 904–950.

To determine where the pause hairpin contacts β, we first tested crosslinking in transcription complexes in which β was split into two functional pieces (37). One type contained fragments 1–235 and 235–1,342 (S1, Fig. 3D), whereas the other contained fragments 1–952 and 950–1,342 (S2, Fig. 3D). The pause hairpin crosslinked to the larger of the two fragments in each sample, mapping its contact to between residues 235 and 952 (Fig. 3A). We next narrowed the crosslink to between the 952 split site and a trypsin cut at R903. In free RNA polymerase, trypsin cleaves β at R903 and K909 in a 3:1 ratio (38). Trypsin digestion of crosslinked −2 and paused complexes localized the β crosslinks in both C-terminal to R903 (Fig. 3B, lanes 2 and 4). Interestingly, cleavage at K909 appeared to be inhibited in the paused complex and activated in the −2 complex (the ratio of 903:909 cuts is 6:1 for pause and 1:2 for −2; Fig. 3B lanes 5 and 6, fragment separation is better without RNase T₁ treatment). Because we typically observed 903:909 cutting ratios between 2:1 and 1:2 in other complexes (not shown), inhibition of 909 cutting in the paused complex may reflect steric occlusion of K909 by the pause hairpin but also could result from structural rearrangement caused by contact anywhere in the 904–952 segment.

To confirm mapping of the β crosslink, we took advantage of the location of this segment within the largest CNBr fragment of β: 806 to 951 (Fig. 3D). CNBr digestion of crosslinked samples either alone or following trypsin digestion yielded the expected fragments (Fig. 3C). Generation of a smaller CNBr fragment from the trypsin-digested sample allowed assignment of the hairpin crosslink to β904–950, because crosslinking to M951 probably would block CNBr digestion.

Even in nonpaused complexes (−6 through −2 and +2 through +6) and the weakly paused MS hairpin complex, the principal β crosslink of loop bases U54–56 occurred to β904–950 (data not shown). However, the overall level of crosslinking was lower (see Fig. 2), trypsin could cut at βK909, and the distribution of label between the β and β′ subunits varied significantly (Table 1). We conclude that a strong and unusual interaction of the pause hairpin to β904–950, perhaps near K909, is unique to the paused conformation.

Base Substitutions That Lessen the Contribution of the Pause Hairpin to Pausing Also Alter Its Contact to RNA Polymerase.

A key question was whether the strong pause hairpin crosslinking near K909 of β reflects a mechanistically important interaction. To address this issue, we compared crosslinking by the WT pause hairpin and the MS hairpin more extensively. In addition to weaker crosslinking and altered electrophoretic retardation, two lines of evidence argue that weaker pausing by the MS hairpin is associated with altered contacts. First, when the MS hairpin was crosslinked in paused complexes containing split β subunit S2 (Fig. 3D), the principal crosslinks were to β950–1342 and to the β′ subunit rather than to the 1–952 fragment that was contacted by the WT pause hairpin (compare lanes 3 and 4, Fig. 3A). However, the split at 952 also increased β′ crosslinking at the wild-type pause site. Thus, disrupting the peptide backbone near the normal site of pause hairpin contact appears to weaken the important β contact, especially when combined with the already weakened MS hairpin, and allow increased interaction with β′. Second, when we placed 5-iodoUMP at positions 49 or 52 in the MS and wt pause hairpin loops, different distributions of crosslinks by the WT and MS hairpins with wild-type polymerase were observed (Table 1). These results also suggest that the MS substitutions change, but do not eliminate, hairpin contacts to RNA polymerase.

Hairpin Interaction Probably Occurs to β904–936, Which Exhibits Weak Sequence Similarity to RNA Polymerase II but No Net Positive Charge.

The β subunits of prokaryotic, archaeal, and eukaryotic RNA polymerases contain significant and colinear sequence similarities, previously designated A-I for β and A-H for β′ (ref. 39 and references therein). However, the region around K909 is not among these highly conserved regions (Fig. 3D). Conserved region G of β extends to P889, whereas G938 begins an ≈100-aa insertion present only in some eubacteria (Fig. 4; refs. 38, 40). We found that RNA polymerase containing a viable deletion of β937–983 paused normally (data not shown), making it likely that hairpin interaction occurs to β904–936. This region is significantly conserved among eubacteria (Fig. 4). Further, the corresponding segments of chloroplast, archaeal, and eukaryotic type II RNA polymerases, but not RNA polymerases I or III, exhibit recognizable similarities (Fig. 4).

Figure 4.

Alignment of sequences from representative RNA polymerases with the pause hairpin interaction site in E. coli (Ec) RNA polymerase. Regions shown as white on black are the most notably conserved regions. Sequences were obtained from GenBank. Hi, Haemophilus influenzae; Ba, Buchnera aphidicola; Pp, Pseudomonas putida; Nm, Neisseria meningitidis; Ss, Staphylococcus aureus; Po, Porphyra purpurea; Bs, Bacillus subtilis; Sp, Spiroplasma citri; Mt, Mycobacterium tuberculosis; Tm, Thermotoga maritima; Hc, Heterosigma carterae; Cp, Cyanophora paradoxa; Cj, Campylobacter jejuni; Mg, Mycoplasma genitalium; Os, Odontella sinesis; Bb, Borrelia burgdorferi; Cr, Chlamydomonas reinhardtii; Sa, Sulfolobus acidocaldarius; Sc, Saccharomyces cerevisiae pol II; Dm, Drosophila melanogaster pol II; Hs, Homo sapiens pol II.

Although one R and two K residues are present in E. coli β904–936, its net charge is 0 at neutral pH; the only overrepresented amino acid (relative to full-length β) is V. It seems likely that the anticipated ionic interaction of hairpins with polymerase involves some other segment of the protein. This is not surprising because 5-iodoU crosslinking should detect side chains near bases in the loop of the pause hairpin and not to the phosphodiester backbone. β subunit residues 904–936 may provide a nonionic component of the interaction between nascent RNA hairpins and RNA polymerase.

DISCUSSION

Our major finding was that a nascent RNA structure, the his pause RNA hairpin, that is known to stabilize a paused conformation of the E. coli transcription complex alters contacts of the nascent RNA with RNA polymerase. The contacts of the hairpin loop to the paused polymerase differed from contacts made by the same RNA bases prior to hairpin formation at the pause or after RNA polymerase escapes the pause in three ways: (i) they relocate the nascent RNA away from the β′ subunit and toward the 904–950 segment of the β subunit; (ii) they inhibit trypsin cleavage of K909; and (iii) after crosslinking to β904–950, they retard the electrophoretic mobility of β in the presence of SDS. Base substitutions in the stem of the pause hairpin that reduce both its stability and the pause half-life also reduce all three of these effects.

Hairpin Contact to β904–950 May Weaken RNA Polymerase’s Grip on DNA but Stabilize the Paused Transcription Complex.

Our results suggest both direct and indirect roles of RNA hairpin interactions in the related mechanisms of pausing and termination. The paused complex forms when the pause hairpin pairs to within 11 nt of the 3′ end and somehow inhibits nucleotide addition, whereas 7–9 U-rich nt 3′ to a terminator hairpin leads to termination (see Introduction). Direct interaction of the hairpin loop with β904–950 may both contribute to the active site–RNA 3′ end geometry that slows NTP addition and help prevent release of the nascent RNA after the paused complex forms.

However, the indirect effect of losing β′ contact when strong β904–950 interaction is established also could contribute to slow nucleotide addition and, at a terminator, to dissociation of the transcription complex. We recently found that the β′ contact lost in the paused complex (Table 1) maps N-terminal to a trypsin cut at K81 of β′ (D.W., unpublished data), which is the only trypsin cut of β′ when DNA is bound to RNA polymerase (38). Loss of RNA contact to this N-terminal segment of β′ is interesting for several reasons. Together with the β C-terminal region, β′1–81 forms a “clamp” that contacts DNA in front of the active site and confers stability to the elongating transcription complex (41, 42). β′1–81 contains a C₄ Zn-finger-like motif (38), and both genetic studies (43) and the viability of a β::β′ fusion (K.S., S. A. Darst, R. A. Mooney, and R.L., unpublished work) suggest it lies near the C-terminal region of β (which also contains a C₄ Zn-finger-like motif in eukaryotes but not in bacteria). Finally, amino acid substitutions that abrogate HK022 put-dependent antitermination occur in the β′ Zn-finger-like motif (9).

Nudler et al. suggest that stabilization of the initiating transcription complex upon growth of the nascent RNA to 8–10 nt reflects strengthening of the DNA clamp when RNA enters the exit channel, and that termination could involve loss of this stabilizing influence (41). Our finding of a shift in RNA interaction away from β′ and to β904–950 is consistent with this idea and suggests that interaction of RNA with the N-terminal region of β′ (which occurs at least for bases 12–17 nt upstream from the RNA 3′ end in the absence of a hairpin; D.W., unpublished data) may stabilize the transcription elongation complex (Fig. 5A). Disruption of this interaction upon hairpin formation may favor pausing by altering the protein conformation optimal for reactive alignment of the RNA 3′ end and NTP in the active site (Fig. 5B). Further disruption of RNA contacts by ρ-independent terminator hairpins may lead to RNA release because the remaining rU-dA-rich RNA:DNA heteroduplex and weakened downstream contacts cannot hold the transcription complex together long enough for the reactive alignment to be established. Interaction of put RNA (and conceivably other antiterminators) with this region of β′ may suppress pausing and termination by stabilizing optimal DNA clamp and active-site geometries. Binding of the λnut-N complex to β904–950 could localize N’s antiterminator domain for β′1–81 contact.

Figure 5.

Model for the effect of nascent RNA hairpin interaction. (A) Transcription complex prior to hairpin formation. The downstream DNA duplex passes through a large channel in the RNA polymerase structure (41, 44), whereas the RNA transcript exits from an ≈8-bp RNA:DNA hybrid through a small tunnel (1, 19, 20, 45). (B) Paused transcription complex. Dotted line indicates uncertainty in position of the paused RNA 3′ end, which loses the reactive alignment depicted in A. Preferential interaction of the pause hairpin with β904–950 removes RNA from a β′ interaction that may help configure the active site and downstream DNA contacts. Closer approach of a ρ-independent terminator hairpin to the RNA 3′ end may remove RNA from remaining critical contacts in the exit tunnel and further destabilize the downstream DNA interaction, possibly leading to opening of the DNA clamp (see refs. 4, 19, and 44). Partial clamp opening may also contribute to pausing but is not depicted here because there is no direct evidence for such a change.

Does RNA Polymerase Possess RNA Chaperone Activity?

Our findings also raise the speculative but intriguing possibility that RNA interaction with β904–950 assists in formation of nascent RNA hairpins. Although the MS hairpin reduces the dwell time of polymerase at the pause site significantly, it still allows pausing, contacts the hairpin-interaction site, and retards the gel mobility of β upon crosslinking. Interestingly, when β was split near the interaction site, the MS hairpin lost contact to the 904–950 region (Fig. 3A), suggesting that combining changes in the interacting protein and RNA moieties destroyed the interaction. A possible explanation is that protein–RNA interaction stabilizes both the MS hairpin and the interaction site. Perhaps facilitating RNA folding is a general or even the major function of the hairpin-interaction site. For instance, transient interactions with the loop region of terminator hairpins might ensure that the hairpins form in the short time window before polymerase elongates past the possible termination site (20).

Interactions of RNA polymerase that stabilize nascent RNA structures also might assist in their folding into biologically active forms. T7 RNA polymerase cannot replace E. coli RNA polymerase for synthesis of functional 23S rRNA in vivo at 37°C (46). This may reflect synthesis at too rapid a rate for proper folding (T7 polymerase is five times faster than the E. coli enzyme at 37°C but yields functional 23S when slowed at 25°C). However, it could reflect an active role of E. coli RNA polymerase in the folding process through interactions that also control the rate of transcription. If such a function were essential, it could explain why β904–950 is not among the regions in which termination-altering substitutions were found previously, because dominant lethal mutations could not be recovered (30, 31). Examination of directed amino acid substitutions and deletions in the hairpin-interaction site will be crucial to test these ideas.