Antiterminator-dependent modulation of transcription elongation rates by NusB and NusG


  • Malcolm Zellars,

    1. Department of Molecular Biology and Microbiology, Tufts University School of Medicine, 136 Harrison Avenue, Boston, MA 02111, USA.
    Search for more papers by this author
  • Catherine L. Squires

    1. Department of Molecular Biology and Microbiology, Tufts University School of Medicine, 136 Harrison Avenue, Boston, MA 02111, USA.
    Search for more papers by this author

Catherine L. Squires. E-mail; Tel. (+1) 617 636 6947; Fax (+1) 617 636 0337.


Ribosomal RNA is transcribed about twice as fast as messenger RNA in vivo, and this increased transcription rate requires the rrn boxA antitermination system. Because several Nus factors have been implicated in rrn antitermination, we have examined the role of NusB, NusE and NusG in controlling the rate of rrn boxA-mediated transcript elongation. In vivo RNA polymerase transcription rates were determined by measuring the rate of appearance of lacZ transcript using a plasmid that contained an inducible T7 promoter fused to the rrn boxA sequence followed by the lacZ gene. This plasmid was introduced into Escherichia coli mutant strains that can be conditionally depleted of NusG, or that carry a deficient nusB gene or a nusE mutation. We found that, in addition to the rrn boxA antiterminator sequence, both NusG and NusB were required to maintain the high transcription rate. The nusE mutation used in this study may be specific for lambda antitermination, as it did not influence the boxA-mediated increase in transcription rate.


Escherichia coli RNA polymerase (RNAP) transcription rates vary depending on the sequences being transcribed. For example, the transcription rate of rrn operons is ≈ 90 nucleotides per second and, in contrast, the rate of mRNA transcription is 40–50 nucleotides per second (Mosteller and Yanofsky, 1970; Vogel and Jensen, 1994).

The difference in rRNA and mRNA transcription rates has been attributed to the rrn antitermination system, but the mechanism by which the rate increase occurs remains unclear (Vogel and Jensen, 1995). Ribosomal RNA antitermination requires a boxA sequence (Li et al., 1984; Berg et al., 1989), which bears strong sequence similarity to the boxA element found in the bacteriophage λ N/nut antitermination system (Olson et al., 1982). Each of the seven rrn operons in E. coli possesses two boxA sequences, one within its leader and one within its spacer region (Berg et al., 1989; Condon et al., 1995). Mutations in these rrn boxA sequences reduce the expression of the rRNA operons, presumably by causing premature termination of some of the transcripts (Heinrich et al., 1995; Pfeiffer and Hartmann, 1997). An additional characteristic of the boxA element is that it apparently confers resistance to the elongation rate-lowering effects of guanosine tetra- and pentaphosphates [(p)ppGpp]. The (p)ppGpp-induced transcription rate reduction is dependent on the NusA protein (Vogel and Jensen, 1997).

NusA, NusB, NusG and NusE (ribosomal protein S10) form an integral part of the bacteriophage λ N/nut antitermination system and, because of the similarities of the two systems, are also likely to play a role in rrn antitermination (Richardson and Greenblatt, 1996). The essential protein NusA has two opposing effects on transcription. On the one hand, it decreases RNAP transcription rates by increasing the pause half-life of the RNAP elongation complex (Lau et al., 1983; Landick and Yanofsky, 1984). However, in another situation, the NusA protein and the rrn antiterminator boxA element increase the transcription rate from 45 nucleotides per second to 65 nucleotides per second on the lacZ gene (Vogel and Jensen, 1997). This paradoxical action of NusA may depend upon the other cellular factors that associate with the RNAP elongation complex. Vogel and Jensen (1995) also showed that the (p)ppGpp molecule exerts its effects through NusA, because the normal RNAP transcription elongation rate on lacZ is resistant to (p)ppGpp action in a nusA cold-sensitive mutant.

The role of NusB in rrn antitermination is demonstrated by an E. coli mutant strain, nusB5, which exhibits transcriptional polarity within rrn operons (Sharrock et al., 1985), and direct biochemical evidence that shows a requirement for NusB in rrn antitermination in vitro (Squires et al., 1993). Whether NusB influences the rate of transcription has not been investigated before this study.

The role of NusE in rrn antitermination is unclear. The NusE protein can form a heterodimer with NusB, and gel shift experiments have shown that this heterodimer can bind the rrn boxA sequence. These data suggest a possible role for NusE in the rrn antitermination system (Mason et al., 1992; Nodwell and Greenblatt, 1993). The effect of NusE on RNA transcription rates and whether this effect is boxA dependent awaits further investigation.

NusG is an essential protein that increases both Rho-dependent termination and RNAP transcription elongation rates. NusG interacts directly with Rho and RNAP (Sullivan and Gottesman, 1992; Li et al., 1992; 1993; Burova et al., 1995), and recent data indicate that NusG reduces the half-life of an RNAP paused complex, thereby increasing the rate of elongation (Burova et al., 1995; Burns et al., 1998). Because NusG increases the efficiency of Rho-dependent terminators and because rrn operons contain these termination sequences, NusG may adversely affect the transcription of rrn operons (Aksoy et al., 1984; Burns et al., 1998). In rRNA transcription involving antitermination, however, one might expect NusG to increase the RNAP transcription rate rather than increase RNAP termination efficiency, and this NusG-dependent modulation of rate during rrn antitermination might be boxA dependent.

Several researchers have suggested possible mechanisms by which antitermination may relate to RNAP transcription elongation rates (Jin et al., 1992; von Hippel, 1998). A relationship between elongation rates and antitermination is highly plausible, because several proteins involved in antitermination, such as lambda Q and NusG, appear to reduce pausing in an elongating transcription complex. This reduction in pausing correlates with increases in RNAP transcription rates (Yang and Roberts, 1989; Burns et al., 1998). Conversely, when RNAP transcription rates are slowed, Rho-dependent termination efficiency is increased (Jin et al., 1992).

To define the relationship between the components of the rrn antitermination system and RNAP transcription elongation kinetics further, we examined the influence of NusB, NusG and NusE on boxA-associated RNA transcription rates. We used the plasmid system developed by Vogel and Jensen (1994) to measure RNAP transcription rates. The plasmid, containing an IPTG-inducible promoter fused to boxA followed by lacZ, was introduced into strains that carried a deficient nusB, had been depleted of NusG or carried the mutant NusE71 protein. Our data showed that NusB and NusG were both required for the rrn boxA-mediated transcription rate increase. The nusE71 mutation did not influence RNA polymerase elongation rates in the presence of the rrn boxA.

Results and discussion

In this study, we have determined the influence of NusB, NusG and NusE71 on transcript elongation rates by measuring the appearance of a full-length lacZ mRNA transcribed from a construct of the T7A1 promoter that is IPTG inducible (vogel and Jensen, 1994). The rate data were collected for transcripts occurring in the presence and in the absence of the rrn boxA sequence. The test plasmids used are shown in Fig. 1. Measurements were made in both wild-type cells and in cells that were mutant for NusB, depleted of NusG or carried a nusE71 mutation. For NusB studies, we used an E. coli strain carrying an IS10 insertion within the nusB gene. An E. coli strain containing a kanamycin insertion into the chromosomal nusG gene was used for experiments involving the depletion of NusG (Sullivan and Gottesman, 1992). Unlike NusB, NusG is essential for survival of the cells. To provide the NusG protein in trans, the nusG gene was cloned under the control of the araBAD promoter, which is dependent on the presence of arabinose for expression (Guzman et al., 1995). Measurements involving NusE were carried out in a strain that contained the nusE71 missense mutation. The nusE gene specifies the ribosomal protein S10 and is an essential gene. The strain carrying the nusE71 mutation has an altered bacteriophage λ plating efficiency at 42°C (Friedman et al., 1984), but otherwise grows normally and makes a functional S10 protein. In the case of NusB and NusG, we also used a probe at the 5′ end of the lacZ transcript to ensure that depletion of these gene products did not affect induction times for the T7A1 promoter. Measurement of lacZ 5′ end mRNA appearance was carried out in wild-type and mutant nus strains to determine whether the nusB::IS10 mutation or NusG depletion influenced transcript initiation rates. Under four different conditions (wild type, nusB::IS10, NusG abundant and NusG depleted), induction was evident ≈ 15–20 s after the addition of IPTG to the cells and appeared as a linear function of time (data not shown). These values are in agreement with previously reported wild-type 5′ end rates (vogel and Jensen, 1994). Therefore, the transcription rates reported below are the result of transcript elongation and not IPTG uptake or other initiation-related events.

Figure 1.

. Diagrammatic representation of the plasmid assay system used to measure RNAP transcription rates in vivo. Open bars represent genes carried by the plasmids; p/o represents the T7A1 promoter–lac operator; cat′ indicates a truncated chloramphenicol acetyltransferase gene; boxA represents the rrn antiterminator sequence (Vogel and Jensen, 1995); solid lines under the genes indicate the positions of the complementary 5′ and 3′ oligonucleotide probes centred at 233 and 3056 bases respectively. The numbers above the genes indicate the total bases transcribed. These numbers were used for all rate calculations.

Effect of nusB::IS10 on transcription rates

The influence of the NusB protein on boxA-dependent transcription was determined using a nusB::IS10 mutant strain. This strain (Table 1) contains an IS10 insertion between nucleotide positions 248 and 249 of nusB (Taura et al., 1992). This insertion inactivates the NusB protein and confers a cold-sensitive phenotype on the cells (Shiba et al., 1986). To determine the effect of NusB on transcript elongation, we compared rates obtained in wild type and two different strain backgrounds carrying the nusB::IS10 allele and the test plasmids pUV12 (−boxA) or pUV16 (+ boxA) (Fig. 2A and B, Table 2[link]). In the wild type, the transcript elongation rates with pUV12 and pUV16 were 36 and 62 nucleotides per second respectively. The boxA feature conferred the expected almost twofold increase in transcription elongation rate (vogel and Jensen, 1994). In the nusB::IS10 background, however, the rates were the same plus or minus boxA, 25 and 26 nucleotides per second (Table 2). These results suggest that NusB is essential for the boxA-mediated transcription rate increase.

Table 2. . RNA polymerase transcription rates. a. Nucleotides per second. The number of nucleotides used for rate calculation was measured from the transcription start site to the middle of the 3′ probe, which is centred at position 3056 of lacZ.b. Rate data are the means ± SD. The experiments were repeated two to five times.c. ND, not determined.Thumbnail image of
Figure 2.

. Induction of lacZ synthesis in strains containing either pUV12 or pUV16. A. Induction of pUV12 and pUV16 in wild-type cells. B. Induction of pUV12 and pUV16 in the nusB::IS10 strain. C. Induction of pUV12 and pUV16 in the NusG-depleted background.

The wild-type strain carrying pUV12 (−boxA) had a significantly more rapid transcription rate than that obtained with the nusB::IS10 strain (36 versus 25 nucleotides per second; Table 2). We attribute the lower transcription rates in the nusB::IS10 background to the deleterious influence of the nusB gene inactivation, but we have not investigated the mechanism behind the nusB mutation's phenotypes. These phenotypes include cold sensitivity and reduced growth rate, as well as our observation of diminished transcription rates.

NusG depletion

The strain RARNGT had its chromosomal nusG gene inactivated by the insertion of a kanamycin resistance cassette, but carries a second nusG gene under the control of the arabinose-inducible araBAD promoter on a plasmid (see Table 1). When RARNGT cells were grown in the absence of arabinose and in the presence of glucose and fucose, transcription from the araBAD promoter was strongly repressed. Exponential growth of the culture continued over several generations, but eventually the growth of the culture slowed to a stop as the NusG concentration was depleted (Fig. 3). In this and the following sections, ‘NusG-abundant’ cells were cells grown in the presence of arabinose, and ‘NusG-depleted’ cells were cells that had been grown for 8 h after transfer to arabinose-free medium. Immunoblot analysis performed on NusG-depleted cells had levels of NusG less than 17 ag per sample (quantification data not shown), whereas NusG-abundant cells grown in the presence of arabinose contained ≈ 40 ng per sample (Fig. 4).

Figure 3.

. Growth curves of an E. coli nusG mutant strain dependent upon the presence of arabinose in the medium for the production of NusG protein. The solid squares represent cells grown with arabinose, and the open squares represent cells grown in the absence of arabinose.

Figure 4.

. Western blot analysis of the depletion of NusG. Samples were taken 8 h after inoculation from cultures that either contained or did not contain arabinose. The labels above the lanes indicate the growth condition in the strain RARNGT. Purified NusG of the indicated amounts was added to produce a standard curve for quantification of the plus arabinose sample.

NusG depletion and transcription rates

It has been reported that NusG increases the transcription rate of mRNA both in vivo (Burova et al., 1995) and in vitro (Burns et al., 1998). We wanted to determine whether the presence of NusG also influences the elevated transcription rate conferred by the rrn boxA feature. We assayed lacZ mRNA transcription rates in NusG-depleted cells in the presence (pUV16) and in the absence (pUV12) of the boxA sequence. The elongation rate of full-length lacZ mRNA, measured by the 3′ end probe (probe lacZ3; see Experimental procedures), was compared in the presence and absence of boxA for both NusG-abundant cells (Fig. 2A) and NusG-depleted cells (Fig. 2C, Table 2[link]). In the NusG-abundant cells, the rates for pUV12 (−boxA) and pUV16 (+ boxA) were 36 and 62 nucleotides per second, respectively, similar to rates observed previously for wild-type cells (vogel and Jensen, 1994; Vogel and Jensen, 1995). However, in the NusG-depleted cells, the rates were essentially the same for both pUV12 and pUV16, 34 and 37 nucleotides per second (Table 2). These results suggest that NusG was essential for the boxA-mediated transcription rate increase.

This finding is consistent with other observations showing that NusG protein in the presence of the λ N protein ensures processivity of λ N/nut antitermination (Li et al., 1992) and may also overcome the NusA-induced pausing of RNAP (Burns et al., 1998). An influence of transcription rate on antitermination has been noted previously, but how closely the two events are related is unclear. Jin et al. (1992) observed that a mutation in rho is suppressed by a rpoB mutation that slows the transcription rate of the mutated RNAP. These researchers, as well as von Hippel (1998), suggest that transcription rates may directly influence the antitermination mechanism.

It should be noted that we observed no decreased rate of gene transcription for pUV12 (−boxA) when the cells were depleted of NusG. This result differs from previous reports indicating that NusG increases mRNA transcription rates both in vivo (Burova et al., 1995) and in vitro (Burns et al., 1998). Burova et al. (1995) measured induced β-galactosidase expression, thus estimating the transcript elongation rate from the time of appearance of β-galactosidase protein activity. Burns et al. (1998) reported that the addition of NusG increases in vitro transcription rates and that this increase depends on NTP concentration. We repeated the rate measurement of Burova et al. (1995) in vivo on the single chromosomal copy lacZ gene. We found, as they found, a slower time of appearance of β-galactosidase protein activity under conditions of NusG depletion. However, lacZ mRNA rate measurements in the same strain showed no diminution in the absence of NusG (M Zellars, unpublished), consistent with the results reported here in Table 2. These results suggest an unexpected connection between translation rates and the cellular content of NusG. This connection is currently being investigated.

Effect of nusE71 on transcription rate

The nusE71 mutation is an A→C transversion at position 257 of the ribosomal S10 gene, rpsJ, that does not permit efficient bacteriophage λ growth at 42°C. It has been suggested that this effect on λ growth is caused by blockage of N action by NusE71 at 42°C (Friedman et al., 1984; Court et al., 1995). Because NusE can form a heterodimer with NusB, and because NusB plays a role in the rrn antitermination system, we wanted to explore the possibility that NusE may help to modulate transcript elongation rates in the presence of the rrn boxA sequence. We transformed the nusE71 mutant strain with pUV16 (+ boxA), and measured the boxA-mediated transcript elongation rate at 42°C. The polymerase transcription elongation rate was determined to be 76 and 75 nucleotides per second for the wild-type and nusE71 mutant strains respectively (Table 2). Thus, the presence of the nusE71 mutation had no effect on rrn boxA-mediated transcription rate.

To define a relationship between the NusE protein and the rrn antitermination system further, a nusE conditional lethal mutant should be used for transcription rate measurements. The fact that the nusE71 mutation did not influence boxA-mediated transcription rates is probably related to differences between rrn antitermination and the λ N/nut antitermination system that was used to select the nusE71 mutation.


The results presented here showing that both NusB and NusG were necessary for the rrn boxA-mediated transcription rate increase provide additional support for the notion that antitermination and transcription rate increases are closely related mechanisms for controlling gene expression. Understanding the role played by Nus factors in these two processes will require the construction of additional test plasmids carrying a terminator within the boxA–lacZ system. Such test plasmids will allow simultaneous monitoring of terminator readthrough efficiency and transcript elongation rates in mutant strains such as those used here. Such experiments should provide additional insights into the link between antitermination and transcription rate.

Experimental procedures

Bacterial strains, plasmids and phage

E. coli strains used in this study are listed in Table 1.

Plasmids and episome construction

Plasmids designated pUV12 and pUV16 have been described elsewhere (Vogel and Jensen, 1994) and are shown in Fig. 1. Construction of an F′ episome with the genotype lacI q lacZ::Tn10 proAB+ was accomplished by the introduction of the Tn10 marker from λ phage NK1323 into F′[lacI q lacZ proAB+] in CSHF109, according to the procedure of Miller (1992). After construction, the F′ episome was mated into NRF1000 to serve as a donor for subsequent matings. Construction of pNG33 carrying the nusG gene was achieved by ligation of a 625 bp pfu DNA polymerase (Stratagene)-derived primer extension reaction (PER) product. The oligonucleotide primers were engineered to contain a consensus Shine–Dalgarno site upstream of nusG and an XbaI site at the 3′ end of nusG. The sequences of the oligonucleotides used were as follows: CGAATTCAGGAGGATCTGAGATGTCTGAAGCTCC and GCGAAATTGTTCTAGAATCTCACGCCTTGTGCAA. The PER mixture was set up essentially as described by the manufacturer. All reactions were performed in a DNA engine thermocycler (MJ Research). The thermocycling profile consisted of an initial denaturation at 94°C for 2 min, annealing at 37°C for 15 s, extension at 72°C for 1 min, followed by 30 cycles of 45 s at 94°C, 45 s at 60°C and 1 min at 72°C. The PER product was resolved by 0.8% agarose gel electrophoresis, excised and purified using the Qiaquick gel extraction kit (Qiagen). The purified PER fragment containing nusG was restricted with XbaI and ligated into pBAD33 (Guzman et al., 1995), which had been restricted with SmaI and XbaI. The ligation products were transformed into DH5αSE-competent cells (BRL) and analysed for inserts.

Strain construction

All strains are isogenic to MC4100 or N99. To construct RAR4100 or RARK527, overnight cultures of MC4100 or IQ527 were plated onto M9 minimal media containing 0.2% arabinose to select for ara+ derivatives. Overnight cultures of the ara+ derivatives were plated onto media containing 40 mg each of serine, methionine and glycine (SMG) (Uzan and Danchin, 1976; 1978) to select for relA+ derivatives. Construction of RARNGT was accomplished by transformation of pNG33 (Table 1), followed by P1 transduction of the nusG::Kn allele from SS287 and mating to transfer the F′ from NRF1000. RARNBT was constructed by transducing the nusB::IS10 allele from IQKN into RAR4100 and mating of the F′ from NRF1000. Verification of the transduction of the nusB::IS10 allele was performed by primer extension using Taq DNA polymerase (Promega) and primers designed to amplify the region of the E. coli chromosome containing the nusB gene. RARNBT was also verified for tetracycline sensitivity to ensure that the zba525::Tn10 did not co-transduce with the nusB::IS10 allele. Construction of RARNEK was carried out by P1 transduction of the nusE71 allele from N99nusE71 into RAR4100, followed by mating of the F′ from RAR4100K. The transduction of the nusE71 allele was verified by the inability of the strain to support the growth of bacteriophage λc17 (Packman and Sly, 1968).

RNA polymerase rate measurements and RNA isolation

In all cases, rate measurements were determined using pUV12 or pUV16 transformed into RARNGT, RARNBT, RARK527 or RARNEK, and each rate experiment was repeated at least twice. All cultures, except for those used in the NusG depletion experiments, were grown in M9 minimal media containing 0.2% d-glucose and 0.2% casamino acids. The cells were grown to an OD436 of 0.4–0.5 before IPTG induction. After induction with 1 mM IPTG, 5 ml samples were removed at 10, 20 or 25 s intervals into stop solution (Vogel et al., 1992) that had been cooled to −10°C. The samples were vortexed and pelleted by centrifugation at 12 000 × g. Total mRNA was isolated by a modification of the procedure of Summers (1970). The isolated RNA was resuspended in 35 μl of 1× DNase I digestion buffer. Then, 1–5 units of RNase-free DNase was added and incubated for 30 min at 37°C. After DNase I digestion, the RNA was precipitated by bringing the volume to 100 μl with H2O, adding NaOAC (pH 5.2) to 0.3 M and a volume of 2-propanol equal to the final volume. The RNA was pelleted at 12 000 × g at room temperature to minimize salt precipitation, washed with 70% ethanol, dried and resuspended in 30 μl of H2O. The concentration of the RNA was determined at OD260.

NusG depletion

To deplete cells of NusG, RARNGT was grown in the absence of l-arabinose just until the end of exponential growth. A culture of RARNGT, grown overnight in the presence of arabinose, was inoculated at an OD436 of 0.01 into M9 media containing 0.2% d-glucose, 0.2% d-fucose and 0.2% casamino acids. The culture was allowed to grow until reaching an OD436 of 0.1 and diluted by 1:4 into prewarmed media. The culture was allowed to grow until reaching an OD436 of 0.4–0.5. At this OD (about 8 h after dilution), the culture was sampled for rate measurements and for immunoblot detection of NusG.


Samples containing 1.5 μg of total protein were resolved on a 12.5% SDS–PAGE gel. The resolved proteins were transferred, according to the manufacturer's instructions, onto polyvinylidene difluoride (PVDF) membrane using a mini trans-blot apparatus (Bio-Rad). The membrane was prepared for immunoblot analysis as follows: the filters were blocked at room temperature for 30 min with 10% non-fat dry milk in TBST (20 mM Tris base, pH 7.6, 137 mM NaCl, 0.5% Tween). The membrane was washed three times for 10 min in TBST. The polyclonal NusG antibody was diluted 1:2000 in TBST containing 10% non-fat dry milk and incubated with the membrane for 1 h to overnight. After incubation, the membrane was washed three times successively for 15 min in TBST. The secondary antibody, conjugated to horseradish peroxidase (HRP), was diluted 1:10 000 in TBST containing 10% non-fat dry milk and incubated with the membrane at room temperature for 30 min. The membrane was washed three times successively for 10 min in TBST. The immunoblot was visualized using a LumiGlo HRP system (Kirkegaard and Perry Laboratories) and exposure to autoradiographic film (Kodak). Quantification of the immunoblot was carried out using imagequant software (Molecular Dynamics).

Northern slot-blot analysis

Total RNA (5 mg) was denatured in 100 ml of a solution containing 6 × SSC and 7% formalin. The RNA was incubated at 65°C for 15 min. After denaturation, the RNA samples were immediately blotted onto Bio-Rad's Zeta-Probe membrane using a slot-blot apparatus (Hoefer Scientific). The blotted RNA was fixed to the filter by UV cross-linking in a Stratalinker UV oven (Stratagene).

Two oligonucleotide probes were used. The first oligo, located at the 5′ end (lacZ5), contained the sequence CGGCCTCAGGAAGATCGCACTCCAG. The 5′ probe was used to assay the induction times. The second oligo, located at the 3′ end (lacZ3), contained the sequence GACACCAGACCAACTGGTAATGGTA. This probe was used to measure the appearance of full-length lacZ transcript, which was later used to calculate the transcription rate. The probes were labelled using [γ-32P]-ATP (7000 ci mmol−1; ICN Radiochemicals) and T4 polynucleotide kinase (Promega).

The filter was prehybridized, hybridized and washed according to the procedure of Angelini et al. (1986). After washing, the filter was sealed in a plastic bag and exposed to a Molecular Dynamics phosphorimaging screen. The resulting image was scanned and quantified using the Storm phosphorimager and the imagequant software (Molecular Dynamics).

Calculation of RNA polymerase transcription rates

RNA polymerase elongation rates were determined as described previously (Vogel and Jensen, 1994). In our case, the lacZ3′ probe was centred at 3274 and 3294 for pUV12 and pUV16 respectively. To obtain the intersecting lines between early transcripts and the appearance of late transcripts, a linear regression was performed on the early transcript levels (from t0 to the point at which the late transcripts appear) and the late transcript levels (from the point at which the c.p.m. increased over background levels to tfinal). The intersecting point derived by this method is the transcription time. Thus, dividing the number of bases transcribed by the transcription time yields the transcription elongation rate in nucleotides per second.


We would like to thank Ulla Vogel for providing pUV12 and pUV16, Koreaki Ito for providing IQ527, Max Gottesman for providing SS287 and N99, Barbara Stitt for NusG protein and antibodies to NusG, and D. Raychaudhuri for providing pBAD33. We are grateful to Craig Squires for providing helpful comments on the manuscript. This work was supported by National Institutes of Health grant GM24751.