Translocation of Escherichia coli RNA polymerase against a protein roadblock in vivo highlights a passive sliding mechanism for transcript elongation


E-mail; Tel. (+33) 2 38 25 76 08; Fax (+33) 2 38 63 15 17.


Current models for transcription elongation infer that RNA polymerase (RNAP) moves along the template by a passive sliding mechanism that takes advantage of random lateral oscillations in which single basepair sliding movements interconvert the elongation complex between pre- and post-translocated states. Such passive translocational equilibrium was tested in vivo by a systematic change in the templated NTP that is to be incorporated by RNAP, which is temporarily roadblocked by the lac repressor. Our results show that, under these conditions that hinder the forward movement of the polymerase, the elongation complex is able to extend its RNA chain one nucleotide further when the incoming NTP is a kinetically favoured substrate (i.e. low Km). The addition of an extra nucleotide destabilizes the repressor–operator roadblock leading to an increase in transcriptional readthrough. Similar results are obtained when the incoming NTPs are less kinetically favoured substrates (i.e. high Kms) by specifically increasing their intracellular concentrations. Altogether, these in vivo data are consistent with a passive sliding model in which RNAP forward translocation is favoured by NTP binding. They also suggest that fluctuations in the intracellular NTP pools may play a key role in gene regulation at the transcript elongation level.


DNA transcription by RNA polymerase (RNAP) is the first reaction in a cascade of events that allows the cells to produce RNAs and proteins necessary for metabolism, growth and multiplication (Richardson and Greenblatt, 1996; Uptain et al., 1997). The polymerase achieves RNA synthesis through three steps, initiation, elongation and termination, all of which are subject to control by regulatory factors that modulate the expression of genes and operons according to cellular needs. In the elongation phase, the ternary complex (TC) made by RNAP and its accessory factors, the DNA template and the nascent RNA transcript, is characterized by two apparently contradictory features. The complex must be stable enough to avoid any random dissociation before reaching the termination point, but at the same time, flexible enough to allow the lateral movement of the polymerase along the template at every cycle of nucleotide addition.

In vitro, multisubunit RNAPs carry out transcript elongation in a discontinuous manner (Landick, 1997; von Hippel, 1998; Nudler, 1999). Frequently, DNA sequences through which the polymerase must pass to complete an RNA chain lead the enzyme to suspend RNA synthesis for varying lengths of time before resuming elongation (Artsimovitch and Landick, 2000; Forde et al., 2002). In some instances, a fraction of the paused polymerases eventually transforms into an arrested state in which the TC neither elongates nor dissociates (Davenport et al., 2000). Extensive footprinting and cross-linking studies of RNAP halted at specific template positions by NTP deprivation have provided evidence that the loss of catalytic activity results from a backward translocation of RNAP along the DNA and RNA chains. The upstream movement shifts the transcription bubble, the RNA–DNA hybrid and the catalytic centre away from the 3′ end of the transcript. Thus, unless the process is reversed, resumption of RNA synthesis requires trimming of the extruded RNA 3′ tip to generate a new 3′ hydroxyl group in register with the catalytic centre of the enzyme (Reeder and Hawley, 1996; Komissarova and Kashlev, 1997a,b; Nudler et al., 1997; Samkurashvili and Luse, 1998; Sidorenkov et al., 1998).

These biochemical studies as well as recent crystal structure determinations of TCs (Zhang et al., 1999; Korzheva et al., 2000) suggest that the transcribing complex moves along the template as a passive sliding clamp in which the protein–nucleic acid interactions confer stability to dissociation while the 8–9 bp RNA–DNA hybrid maintains the register. In this view (Guajardo and Sousa, 1997; von Hippel and Pasman, 2003), random lateral diffusion drives RNAP back and forth between two adjacent template positions (n and n + 1) with the growing 3′ end of the nascent RNA alternating between a substrate and a product binding subsites (pretranslocated and post-translocated states of the complex respectively). At each cycle of nucleotide addition, the binding of the incoming NTP at the substrate subsite within the post-translocated complex shifts the translocational equilibrium forward by blocking the re-entry of the RNA 3′ end. After bond formation, the substrate subsite is vacated by the new RNA 3′-terminus in a rapid equilibrium between the pre- and the post-translocated states (Fig. 1A). Thus, for each subsequent round of nucleotide addition, an efficient binding of a new incoming NTP will decrease the propensity of RNAP to slide back and will allow the nucleotide addition cycle to continue effectively and rapidly. Accordingly, the binding of the incoming NTP at any template position will determine the outcome of the transcription complex at this position by modulating the partition between the different reaction pathways occurring during the nucleotide addition cycle (pausing, elongation, termination). Consistent with this, the pausing at the end of a T-stretch of an intrinsic terminator has been partially alleviated with C substitutions as expected if CTP, which is the kinetically favoured NTP substrate, facilitates the forward movement of RNAP (Gusarov and Nudler, 1999). Thus, the passive sliding model of translocation raises the attractive possibility that the regulation of gene expression during the elongation phase could occur in response to variations in NTP pools or through interactions of the TC with factors that modulate its apparent Kms for the NTPs. Indeed, the N protein of bacteriophage lambda has been shown to increase three- to fivefold the overall rate of transcript elongation in vitro, presumably by decreasing the apparent Kms for the NTPs (Rees et al., 1997).

Figure 1.

A. Model for the passive sliding mechanism of RNAP translocation. The elongating RNAP alternates in vitro between two adjacent positions on the template (n and n + 1). In the upper part (pretranslocated state of the complex), the 3′ end of the nascent RNA occupies the substrate binding subsite of the RNAP active site. Random lateral diffusion drives RNAP in the forward direction to the next template base, thus leaving the substrate binding subsite vacant and prone to bind the next template-directed NTP (lower part). In the post-translocated state of the complex, the product binding subsite of the RNAP is filled by the 3′ end of the nascent RNA.
B. Schematic representation of the lateral oscillations experienced by the TC roadblocked in vivo by the lac repressor (only the relevant part of the plasmid is shown). RNA polymerase initiates transcription from the constitutive hisR promoter derivative (which is not under stringent control) and goes through the trinucleotide (ATC/TAG)n repeating sequence until it encounters the lac repressor bound to the operator sequence (Op/Rep). The roadblocked TC is in a dynamic equilibrium between downstream and upstream translocated positions on the template. The catalytic site of the RNAP located in the downstream position is in register with position −6 relative to the upstream border of the lac operator sequence. Upon collision with the lac repressor, the RNAP slides backward with a concomitant upstream shift of the transcription bubble and the RNA–DNA hybrid such that the 3′ end of the nascent RNA becomes extruded out of the polymerase. The upstream movement of RNAP relocates the catalytic site in register with position −9 where GreA/B-mediated cleavage of the RNA 3′ tip occurs. Upon RNA cleavage, RNAP re-elongates the transcript down to position −6. The top of the figure indicates the base changes at position −5 within the plasmids pATC10-A/C/G/T used in this study.

Although the aforementioned in vitro studies with starved TCs lend some support to the passive sliding model, a direct demonstration of the process is still lacking. Furthermore, the backward translocation of RNAP over long distances that leads to catalytically incompetent complexes may not be relevant to the mechanism that moves the polymerase in single base steps during the dynamic conditions of transcript elongation. An alternative mechanism for RNAP movement, ‘powerstroke’, in which the translocation of the enzyme is powered by the hydrolysis of NTP upon its incorporation, has been also considered (see Gelles and Landick, 1998). The most valuable information supporting the passive sliding model comes from kinetic studies of monomeric RNAP of the bacteriophage T7, in which the effects of NTP concentration on individual transcript extension steps were evaluated (Guajardo et al., 1998). However, these in vitro experiments have been performed with complexes that are at the late initiation phase, in which the polymerase carries out the first rounds of RNA synthesis while maintaining strong interactions with the promoter and thus does not translocate as a whole on the template. Therefore, the results may have only limited relevance to the mechanism of translocation during the elongation phase.

We have shown previously that a transcription elongation complex halted within a trinucleotide repeating sequence (ATC/TAG)n as the result of a roadblock imposed by the lac repressor in Escherichia coli is able to elongate its RNA chain down to position −6 relative to the upstream edge of the operator motif (Guérin et al., 1996; Toulméet al., 1999). In the model in which RNAP translocation occurs by a passive sliding mechanism, such a roadblocked complex is expected to oscillate between pre- and post-translocated states that should interconvert on a time scale similar to that of nucleotide addition (bond formation). Therefore, we asked whether this translocational equilibrium could be shifted forward by a kinetically preferred NTP substrate or by increasing the concentration of a less preferred NTP substrate. The results reported below show that, under optimal NTP-binding conditions in vivo, the roadblocked TC readily extends its RNA chain one nucleotide further (down to position −5 relative to the upstream edge of the operator motif). The addition of an extra nucleotide destabilizes the repressor–operator roadblock leading to an increase in transcriptional readthrough. Thus, these data support the passive sliding model in which NTP binding drives RNAP forward translocation.


Dynamic features of an E. coli transcription elongation complex roadblocked by the lac repressor in vivo

We have reported previously the design of an E. coli system in which RNAP that initiates transcription from a constitutive promoter within a plasmid is roadblocked at a downstream position by the lac repressor bound to its operator motif (Guérin et al., 1996; Toulméet al., 1999). Excess lac repressor is provided inside the cell by a second plasmid harbouring the lacIq gene. The elongation complex was halted within a trinucleotide repeating sequence (ATC/TAG)n in which the structural status of each base residue could be analysed in situ by a variety of chemical and physical probes. By combining DNA footprinting and RNA 3′ end mapping experiments, we have shown that the TC halted within the (ATC/TAG)n sequence is in dynamic equilibrium between downstream and upstream translocated positions (Fig. 1B). The roadblocked polymerase is repeatedly switching between the two positions on the template with accompanying transcript cleavage and resynthesis. In this view, the polymerase first elongates the transcript until the operator-bound lac repressor becomes a physical barrier. Within such a roadblocked complex, the catalytic centre of the enzyme is in register with position −6 (the location of the 3′-terminal RNA nucleotide with respect to the upstream border of the operator motif). At this point, the polymerase becomes prone to slide backward. The upstream movement relocates the catalytic centre in register with position −9, where transcript cleavage mediated by GreA and GreB occurs (Toulméet al., 2000). Re-elongation of the transcript from the new 3′-terminus brings RNAP back to the downstream roadblocked location. Thus, in agreement with several in vitro and in vivo studies by others (reviewed by King et al., 2003), the E. coli RNAP halted in the elongation phase by the lac repressor does not terminate transcription and dissociate from the template. Rather, our previous results reveal the dynamic features of the roadblocked TC that mimic the conditions of active transcription.

A kinetically favoured incoming NTP allows a one-nucleotide extension of the nascent RNA

Once the transcribing RNAP arrives at position −6 relative to the repressor binding motif, it is expected to occupy the opposite side of the DNA helix with the 5–6 bp spacing being equal to the half-turn angular separation between the individual footprints of the two proteins (Guérin et al., 1996). At this point, the polymerase will come into a steric clash with the repressor, which will impede its forward movement. The hindrance may constitute an ultimate block to translocation with the TC being trapped in the pretranslocated state as the RNA 3′-terminus cannot clear the substrate subsite. Alternatively, on account of structural flexibility within the leading edge of the polymerase and/or variations in DNA twist induced by thermal fluctuations, the TCs roadblocked at position −6 may still be able to populate the post-translocated state. However, because the forward movement of the polymerase is disfavoured, this will increase the apparent Km of the NTP to be incorporated after translocation (Guajardo et al., 1998; King et al., 2003).

Our previous experiments described above have been performed with plasmid derivatives such as pATC10-A (Fig. 1B), which have a perfect (ATC/TAG)n repeat, and thus the next nucleotide to be incorporated at position −5 by the roadblocked TC is an A (Toulméet al., 2000). The fact that the four NTPs are known to have different apparent Kms for elongating E. coli RNAP in vitro (KC < < < KA < KG < KU; see Kingston et al., 1981) led us to test whether another base residue at position −5 could be incorporated by the roadblocked polymerase. To this end, we constructed plasmid template derivatives (pATC10-C, pATC10-G and pATC10-T) in which the basepair at position −5 was changed in order to have a C, a G or a T, respectively, on the non-template strand (Fig. 1B). We next performed S1 nuclease protection experiments to map the 3′ ends of the truncated transcripts resulting from the roadblocked TCs within the different plasmids. In agreement with previous results (Toulméet al., 2000), the RNAs extracted from cells transformed with the plasmid pATC10-A and grown in the absence of the inducer IPTG contained short transcripts with 3′ ends distributed primarily between positions −6 and −9 with respect to the lac operator (Fig. 2A, lane A). This signature of the apparent ‘cleavage-and-restart’ process of the oscillating roadblocked TCs was also observed for the RNAs extracted from cells harbouring the plasmids pATC10-G and pATC10-T (Fig. 2A, lanes G and U respectively). In contrast, the mapping of the RNAs obtained from cells transformed with the plasmid pATC10-C shows an additional band corresponding to a truncated transcript with a 3′ terminus that is located at position −5 (Fig. 2A, lane C). Conceivably, the detection by S1 nuclease of an extra nucleotide at the 3′ end of part of the transcripts generated by pATC10-C could result from a different sensitivity of the enzyme during the mapping experiment. However, this possibility was excluded by running in vitro transcription assays in which transcripts with an additional nucleotide at position −5 were readily observed for the pATC10-C template (results not shown).

Figure 2.

Nuclease S1 mapping of the 3′ ends of the truncated RNAs produced in wild-type (A) and greAgreB mutant strains (B) by the roadblocked TC transcribing any of the four plasmid templates (pATC10-A/C/G/T). Above these autoradiograms, and in subsequent figures, the letters specify the incoming NTP to be incorporated at position −5 within the corresponding plasmid template. The arrows indicate the −9, −6, and −5 positions of the 3′ ends of the transcripts relative to the upstream border of the lac operator motif.

The RNA 3′ ends mapping experiments were extended with E. coli cells lacking functional GreA and GreB (Fig. 2B). As reported previously (Toulméet al., 2000), the results obtained with pATC10-A show a single major band at position −6, indicating that the TC roadblocked within the (ATC/TAG)n repeat backslides and that it resumes elongation upon transcript shortening promoted by GreA and GreB. In the absence of functional Gre factors, the roadblocked TC slides back and forth between the −6 and −9 positions without shortening the 3′ part of the transcript. Similar results were obtained with pATC10-G and pATC10-T (data not shown). For the pATC10-C template, however, two major RNA 3′ ends located at positions −6 and −5 were detected in the absence of functional Gre factors, which indicates that part of the roadblocked TCs (40–50%) have extended their RNA transcript one nucleotide further. These results show that the TC roadblocked at position −6 can interconvert between the pre- and the post-translocated states and that the efficient binding of the kinetically favoured CTP (i.e. low Km) within the post-translocated state leads to nucleotide addition at position −5.

The addition of an extra nucleotide increases the efficiency of transcriptional readthrough

Conflicting results about the efficiency of repression of downstream genes by the lac repressor have been reported in the literature (reviewed by King et al., 2003). Most of these studies use DNAs with different sequences flanking the operator motif. In theory, the operator-bound lac repressor is expected to impede RNAP progression only for a certain period of time determined by repressor concentration and operator affinity. The collision between the two complexes may temporarily halt the transcribing TC, which should resume elongation upon repressor dissociation. Accordingly, the duration of the block should be a function of the occupancy of the repressor on its binding site (Guajardo and Sousa, 1999). However, the rate of repressor dissociation is likely to depend on the configuration of the close apposition of the two complexes on the DNA and thus on the sequence context surrounding the operator motif. To test this hypothesis, we investigated the effects of the base substitutions at position −5 on transcriptional readthrough by measuring the amount of the downstream cat mRNA transcripts produced by the different plasmid templates (pATC10-A/C/G/T shown in Fig. 1B). A 32P end-labelled oligonucleotide complementary to the early part of the cat message was used to measure cat mRNA transcript levels by primer extension with reverse transcriptase. A second primer was included to measure plasmid-encoded β-lactamase transcript (bla) levels as an internal control. Figure 3A shows that, for all the plasmids, a significant fraction of RNAPs did transcribe beyond the operator motif under repressing conditions (–IPTG). The efficiency of transcriptional readthrough for each plasmid construct was evaluated by normalizing the level of cat mRNA produced under repressing conditions to that obtained under inducing conditions. This analysis revealed a clear difference between pATC10-C and the three other plasmid templates (Fig. 3B). The presence of a C at position −5 increases the readthrough efficiency by roughly 30%. These data together with the results presented above show that there is a direct correlation between the incorporation of a nucleotide at position −5 by the roadblocked TC and the efficiency with which the TC reads through the protein impediment. Apparently, by moving closer to the protein roadblock, the polymerase increases the rate of dissociation of the repressor from its operator motif.

Figure 3.

Determination of the efficiencies of transcriptional readthrough as a function of the NTP to be incorporated at position −5. Total mRNA was extracted from W3110 co-transformants containing one of the four plasmid templates (pATC10-A/C/G/T) and either induced (+IPTG) or not induced (–IPTG) 15 min before cell harvest. The cat and bla transcript levels were measured by quantitative primer extension analysis. The primer extension products were separated on an 8% denaturating polyacrylamide gel (A) and quantified by phosphorimager scanning (B). The readthrough efficiency was measured using the formula: [(cat/bla transcripts)–IPTG/(cat/bla transcripts)+IPTG] × 100. The error bars reflect the standard deviation from the mean of three experiments performed with two independent co-transformants. The asterisks in (A) indicate non-specific arrests of reverse transcriptase within the cat mRNA.

Modulating readthrough efficiency by varying the in vivo pool size of the NTP substrate

The above experiments indicate that the affinity of RNAP for the nucleotide to be incorporated at position −5 determines the frequency of reading through the repressor roadblock. A corollary of this interpretation is that it should be possible to modulate the readthrough efficiency of an RNAP transcribing the pATC10 template derivatives by varying the in vivo concentration of the NTP substrate to be incorporated at position −5. In this case, the readthrough level should remain insensitive to variations in the concentrations of the three non-substrate NTPs. This assumption was tested by measuring the readthrough efficiencies in strains in which the free NTP pools were varied.

Strains deficient in the first enzyme of the pyrimidine biosynthesis pathway (encoded by the carAB operon) have been used to modulate the pyrimidine and purine levels inside the cell (Wilson et al., 1992; Qi and Turnbough, 1995; Gaal et al., 1997; Pokholok et al., 1999; Cheng et al., 2001). The car-94 allele confers pyrimidine auxotrophy so that growth in glucose-minimal salts medium requires the addition of an exogenous pyrimidine source. Uracil (URA) is a readily metabolized pyrimidine source, which supports rapid cell growth and high intracellular levels of UTP and CTP. Conversely, UMP can be used as a poor pyrimidine source (in media such as NC, see Experimental procedures), which results in slower cell growth and low levels of UTP and CTP. Therefore, strain CLT42 (car-94) was transformed with the repressor-producing plasmid and either of the four test plasmids (pATC10-A/C/G/T). Co-transformants were grown under limiting and excess pyrimidine conditions (UMP and URA respectively), and the levels of downstream cat mRNA were measured as described above. The data summarized in Fig. 4A show that the efficiencies of transcriptional readthrough vary depending on both the incoming NTP substrate and the pyrimidine source available in the medium. Significant increases in the readthrough levels were observed for pATC10-C and pATC10-T when the cells were grown with uracil compared with growth with UMP, indicating a correlation between the efficiency of readthrough and the concentration of the NTP to be incorporated at position −5. However, the ratio between cat mRNA produced in uracil-grown cells to that obtained with UMP-grown cells is more pronounced for pATC10-T than for pATC10-C. This difference is presumably due to the more favourable Km for CTP under pyrimidine-limiting conditions and the relatively modest variation in the CTP pool in carAB mutant strains grown under the two conditions compared with the 20-fold variation reported for the UTP levels (Sadler and Switzer, 1977; Cheng et al., 2001). The efficiency of readthrough was unaffected by the pyrimidine availability when measured in cells bearing the pATC10-A and pATC10-G plasmid derivatives, which is consistent with the fact that changes in the concentrations of the NTPs not required to extend the transcript to position −5 do not affect the transcriptional readthrough.

Figure 4.

Determination of the efficiencies of transcriptional readthrough in strains limited for pyrimidine nucleotides.
A. CLT42 co-transformants harbouring one of the four plasmids (pATC10-A/C/G/T) were grown to mid-log in NC supplemented medium containing either uracil (excess pyrimidines, black histograms) or UMP (limiting pyrimidines, grey histograms), and RNA was extracted after a 15 min induction with IPTG when needed. The cat and bla transcripts were quantified, and the readthrough efficiencies were calculated by normalization between the –IPTG and +IPTG values using the formula described in the legend to Fig. 3. The car-94 co-transformants have a growth rate of 1.37 and 1 doublings per hour in uracil and UMP media respectively. The error bars reflect the standard deviation from the mean of three experiments performed with two independent co-transformants.
B. The CLT246 (car-403::Tn10) strain harbouring the pAC177IQ plasmid and either pATC10-A, pATC10-G or pATC10-T was grown in NC supplemented medium containing UMP as a pyrimidine source and MgSO4 concentrations varying from 0.2 to 1.2 mM to modulate the utilization of UMP and thereby vary the intracellular purine levels. RNA extractions were performed from cultures induced or not induced with IPTG for 15 min before harvesting the cells. The cat and bla transcripts were quantified, and the data were processed as described above. The large triangles above the graphs indicate the variations in purine (left) and pyrimidine (right) levels in CLT246 according to the growth rate variations. pATC10-A, triangles; pATC10-G, diamonds; pATC10-T, circles. The data shown represent the average of three independent experiments.

car-403::Tn10 is another allele of the carAB operon (Gaal et al., 1997; Pokholok et al., 1999). In strain CLT246 (car-403::Tn10), a poor pyrimidine source causes a decrease in the intracellular concentrations of UTP and CTP. The low level of pyrimidine triphosphates limits overall RNA synthesis and growth rate and, as a result, increases the intracellular concentrations of ATP and GTP. The growth rate and the purine pool size can be varied by growing the cells in UMP-supplemented minimal medium containing different concentrations of MgSO4. The MgSO4 concentration modulates UMP utilization as a pyrimidine source, with higher MgSO4 concentrations permitting more efficient use of UMP as a pyrimidine source. Strain CLT246 harbouring the lac repressor-overproducing plasmid was co-transformed with either of the two plasmid templates pATC10-A and pATC10-G. The amount of transcriptional readthrough in each co-transformant was then determined in cells grown in UMP-supplemented medium containing various MgSO4 concentrations. A gradual decrease in the efficiency of transcriptional readthrough was readily observed for the two plasmids by increasing pyrimidine availability (i.e. reducing ATP and GTP pool sizes), as reflected by faster growth rates (Fig. 4B, left). A sharp opposite response was observed in the control experiment conducted with the plasmid pATC10-T, in which the incoming NTP substrate at position −5 is UTP (Fig. 4B, right). These results establish a direct correlation between the intracellular concentration of the NTP to be incorporated at position −5 and the efficiency with which RNAP reads through the lac repressor impediment.


The complete cycle of nucleotide addition is a highly regulated process that requires three steps in order to move the elongation complex in the forward direction: binding of the incoming NTP into the substrate binding site, formation of a phosphodiester bond, and translocation of RNAP from the n to the n + 1 position to align the NTP binding site with the next template base. Current models for transcript elongation infer that the rate-limiting step in the process is biochemical (NTP binding and catalysis) rather than translocation of RNAP (Guajardo et al., 1998; von Hippel, 1998; von Hippel and Pasman, 2003; Holmes and Erie, 2003). The polymerase is believed to progress passively along the template by taking advantage of random lateral oscillations in which single basepair sliding movements reversibly interconvert the pre- and post-translocated states of the complex. The in vivo results presented in this study support this passive translocational equilibrium. They show that, under conditions that hinder the forward movement of the polymerase, the elongation complex will readily incorporate the most kinetically favoured NTP (low Km), whereas a less kinetically favoured NTP substrate will be less efficiently incorporated unless its concentration inside the cell is increased. The incorporation of an extra nucleotide mitigates the physical obstruction imposed by a protein roadblock leading to an increase in transcriptional readthrough.

The single basepair lateral movements of RNAP are difficult to observe directly. Nevertheless, because the partition between the pre- and the post-translocated states is shifted forward at each efficient binding of the incoming NTP, indirect evidence can be obtained if the effects that modulate the translocational equilibrium at a given individual extension step lead to a specific reaction outcome. By changing systematically the templated NTP that is to be incorporated by RNAP temporarily roadblocked in front of the lac repressor, we show that the nature and the concentration of that nucleotide determine the outcome of the stalled RNAP. The impediment is overcome with the best efficiency when the NTP to be incorporated is CTP, which is known to have the lowest apparent Km (Kingston et al., 1981). Lower readthrough levels are observed when the required NTP is UTP, ATP or GTP, substrates for which RNAP exhibits higher apparent Kms. In the two latter cases, however, a better readthrough efficiency is obtained by raising the concentration of the intracellular purine pools, thus compensating for the low affinity. Similarly, the transcriptional readthrough is increased when the required NTP is UTP or CTP by elevating the intracellular pyrimidine pools. Yet, the levels of readthrough remained unaffected by intracellular variations in the NTPs not required for that specific extension step, as expected if the efficiency with which the polymerase overcomes the protein roadblock depended strictly on the NTP required at position −5. Thus, if we consider that the translocation from position −6 to position −5 is the readthrough-determining event, our observations are inconsistent with an alternative mechanism for RNAP translocation in which the forward movement of the TC is actively driven by NTP hydrolysis (see Gelles and Landick, 1998). In effect, an active translocation ‘powerstroke’ mechanism would suppose an increase in the readthrough efficiency when the NTP required at position −5 is ATP or GTP by raising the CTP pool, hydrolysis of which (during the incorporation of a CMP at the preceding position −6) would be expected to fuel the movement.

Our in vivo results may have important implications for gene regulation in response to fluctuations in the nucleoside triphosphate pools. Several examples relating the influence of NTP pools on transcription have been reported in the literature. Expression of the operons of the pyrimidine biosynthetic and salvage pathways (pyrBI, pyrC, codBA, upp) is regulated by pyrimidine availability (Wilson et al., 1992; Liu et al., 1994; Qi and Turnbough, 1995; Cheng et al., 2001). During the highly regulated synthesis of rRNA, the activity of the rrnB P1 promoters increases in response to elevated levels of ATP or GTP inside the cell, thus providing sufficient amounts of rRNA for ribosome assembly and subsequent protein production (Gaal et al., 1997; Schneider et al., 2002). In all these examples, the rate of incorporation of a critical nucleotide at a defined position determines the outcome of the initially transcribing complex leading to promoter clearance and subsequent elongation. In another example, the expression of two enzymes of the pyrimidine biosynthetic pathway, the ATCase and OPRTase (aspartate transcarbamylase and orotate phosphoribosyltransferase) encoded, respectively, by the pyrBI and pyrE operons, is directly regulated by the UTP level at the elongation phase through a transcription attenuation control mechanism (Turnbough et al., 1983; Bonekamp et al., 1984; Poulsen et al., 1984; Donahue and Turnbough, 1994). All these mechanisms allow the bacteria to sense fluctuations in the NTP pools and thus to adjust gene expression with cellular needs.

The millimolar concentration range of NTPs in vivo is above the apparent Kms determined in the elongation phase for E. coli RNAP in vitro, which implies that active elongation may not be modulated by intracellular NTP pools under optimal laboratory growth conditions. However, nutritional limitations under natural growth conditions may cause large fluctuations in the NTP pools. Also, the in vitro determinations have shown significant variations in the apparent Kms for the incorporation of NTPs as a function of the transcribed DNA (Levin and Chamberlin, 1987). Thus, it appears likely that transcription elongation in vivo could be modulated by NTP pools at certain transcriptional impediments. These can be template sequences such as pause sites, in which the apparent Km for the incorporation of a specific nucleotide is anomalously high (see examples in Donahue and Turnbough, 1994; Artsimovitch and Landick, 2000). Chromosomal proteins as well as site-specific binding proteins constitute another class of hindrance to active transcription in the cellular context, where collisions between elongating RNAPs and DNA-bound proteins are presumably very common. Hence, as clearly shown by our artificial experimental system, fluctuations in the NTP pools in vivo may play a crucial role in gene expression by modulating the elongation rate at specific loci within the genome. Interestingly, some E. coli transcription elongation factors such as NusA and NusG as well as the lambda antitermination factor N have been shown to affect the rate of transcript elongation in vitro (Das, 1992; Burova et al., 1995; Burns et al., 1998). A possible in vivo role for these proteins could be to modify the apparent Kms and, thus, the NTP concentration requirements of the elongating RNAP to bypass some specific impediments. We speculate that such modifications of the apparent Kms could be exerted through conformational switches in the vicinity of the RNAP active centre. A putative target is the F bridge helix of the β′ subunit, which was recently proposed to alternate between bent and relaxed conformations concomitantly with RNAP lateral movements and the entry of NTP substrates (Epshtein et al., 2002).

Finally, besides regulation of gene expression, the mechanism of RNAP translocation driven by NTP binding could play a critical role in the regulation of other important metabolic processes in response to variations in NTP pools. By powering the movement of the replicating chromosome, the RNAP is a molecular motor that presumably regulates bacterial chromosome segregation by using the mechanism described here (Dworkin and Losick, 2002). Another example is genomic instability, as transcription-associated mutations are higher with RNAP mutants that have slower elongation rates (Beletskii et al., 2000).

Experimental procedures

Bacterial strains and plasmids

All experiments have been performed with the W3110 strain (E. coli Genetic Stock Center) unless otherwise stated. The strains AD8775 (W3110 greAgreB), CLT246 (car-403::Tn10ΔlacIZ), and CLT42 ([Fcar-94Δ(argF-lac)U169 rpsL150 thiA1 relA1 deoC1 ptsF25 flbB5301 rbsR]) have been described previously (Roland et al., 1985; Pokholok et al., 1999; Toulméet al., 2000). The pATC10-A/C/G/T plasmids have been constructed with two sets of complementary oligonucleotides. One set contains the repetitive sequence (CAT/GTA)8, and the other is made of the following sequences: 5′-CNTCATGAATTGTGAGCGCTCACAATTCand 5′-GAATTGTGAGCGCTCACAATTCATGAN*G, where N is either a A, a C, a G or a T, N* is the base complementary to N, and the underlined sequence represents the lac operator motif. The double-stranded oligonucleotide (CAT/GTA)8 was phosphorylated and ligated to any of the four duplexes of the second set. Ligation products were separated on an 8% polyacrylamide gel, and the relevant fragments with the appropriate size were excised from the gel, phosphorylated and inserted into the filled-in SalI site of the pKP14 vector (Figueroa-Bossi et al., 1998). The ligation reaction was transformed into electrocompetent DH5α bacteria, and the plasmids with the desired insert were selected upon digestion followed by dideoxy sequencing to control for the orientation of the insert. The resulting plasmids with either a A, a C, a G or a T at position −5 (on the non-template strand) relative to the first base of the lac operator motif were named pATC10-A, pATC10-C, pATC10-G and pATC10-T respectively. The hisR hpa promoter being used in this work, which drives the transcription through the repeat, the operator motif and the downstream cat gene, is not under stringent control (Figueroa-Bossi et al., 1998). The repressor-overproducing plasmids pAC177IQ and pAC184IQ have been described previously (Toulméet al., 1999; 2000).

Media and growth conditions

Bacterial cultures were grown at 37°C with the appropriate antibiotics at the following final concentrations: ampicillin 100 µg ml−1, chloramphenicol 15 µg ml−1 (greAgreB mutant strain), kanamycin 18 µg ml−1 (greAgreB mutant strain) or 30 µg ml−1 (pAC177IQ), tetracycline 5 µg ml−1 (car-403::Tn10 mutant strain) or 16 µg ml−1 (pAC184IQ). Standard LB medium was used to grow the cells unless otherwise stated. Intracellular pyrimidine fluctuations were obtained by growing the pyrimidine auxotrophic strain CLT42 (car-94) in NC salt medium (Alper and Ames, 1978) supplemented with 10 mM NH4Cl, 0.4% glucose, 0.015 mM thiamine hydrochloride, 1 mM arginine and either 1 mM uracil (pyrimidine excess) or 0.25 mM UMP (pyrimidine limitation). Variations in the purine nucleotide pools in strain CLT246 (car-403::Tn10) were obtained by growing cells in NC medium supplemented with 10 mM NH4Cl, 0.4% glucose, 0.015 mM thiamine hydrochloride, 1 mM arginine, 0.25 mM UMP and with the concentration of MgSO4 varied from 0.2 to 1.2 mM to modulate the uptake of UMP and thus control pyrimidine availability. Growth of the cultures in NC supplemented media was followed by monitoring the absorbance at 650 nm, and growth rates were calculated by measuring the slopes of the growth curves.

RNA extractions

The bacteria were grown in LB or NC supplemented medium at 37°C to an OD at 600 or 650 nm, respectively, of ≈ 0.5. When needed, the induction with 1 mM IPTG (to remove the repressor from the operator) was performed 15 min before cell harvest. The samples were processed essentially as described previously (Toulméet al., 2000) except that NaCl (5 g l−1) instead of LB medium was used for resuspension of NC-grown cells.

Nuclease S1 mapping and cat mRNA measurements

Cellular RNAs extracted from W3110 and AD8775 (W3110 greAgreB) co-transformants grown in LB medium were used to map by nuclease S1 protection experiments the 3′ ends of the short transcripts as described previously (Toulméet al., 2000). The protected fragments were separated on an 8% denaturing polyacrylamide gel and visualized by phosphorimaging.

Cellular transcripts produced under various growth conditions were analysed quantitatively with 32P-labelled primers that anneal between positions 248 and 265 for the cat gene and positions 4850 and 4868 for the bla gene (the positions are those of the original vector pKK232-8; Brosius, 1984) and with M-MLV reverse transcriptase exactly as described before (Toulméet al., 2000). Briefly, 20 µg of RNAs were mixed in water with 0.1 pmol of 32P-labelled cat and bla primers in a volume of 10 µl, heat denatured and hybridized for 20 min at 45°C after the addition of 4 µl of 5× reverse transcription buffer (250 mM Tris-HCl, pH 8.3, 375 mM KCl, 15 mM MgCl2), 2 µl of 100 mM dithiothreitol and 5 U of RNase inhibitor (RNAguardTM from Amersham). Synthesis of the cDNA was initiated upon addition of 4 µl of dNTPs (5 mM each) and 100 U of M-MLV reverse transcriptase. After a 1 h incubation at 45°C, the primer extension products were precipitated with 0.1 volumes of 3 M sodium acetate and 2 volumes of absolute ethanol. Products were analysed by urea PAGE (8 M urea, 8% polyacrylamide). The dried gels were quantified by scanning on a PhosphorImager (Molecular Dynamics) with imagequant software, version 5.1.


We are very grateful to F. Toulmé for expert technical assistance, and M. Boudvillain for reading the manuscript. This work was supported in part by the Association pour la Recherche sur le Cancer (♯5560), the ANRS (♯02003), the Ligue Contre le Cancer (comité du Loiret) and by an NIH grant (GM29466) to C. L. Turnbough.