Bacterial transcription elongation factors: new insights into molecular mechanism of action


E-mail; Tel. (+1) 718 270 3752; Fax (+1) 718 270 2656.


Like transcription initiation, the elongation and termination stages of transcription cycle serve as important targets for regulatory factors in prokaryotic cells. In this review, we discuss the recent progress in structural and biochemical studies of three evolutionarily conserved elongation factors, GreA, NusA and Mfd. These factors affect RNA polymerase (RNAP) processivity by modulating transcription pausing, arrest, termination or anti-termination. With structural information now available for RNAP and models of ternary elongation complexes, the interaction between these factors and RNAP can be modelled, and possible molecular mechanisms of their action can be inferred. The models suggest that these factors interact with RNAP at or near its three major, nucleic acid-binding channels: Mfd near the upstream opening of the primary (DNA-binding) channel, NusA in the vicinity of both the primary channel and the RNA exit channel, and GreA within the secondary (backtracked RNA-binding) channel, and support the view that these channels are involved in the maintenance of RNAP processivity.


RNA synthesis in all cellular organisms is mediated by DNA-dependent RNA polymerase (RNAP) whose structure and function are conserved from bacteria to human. Bacterial RNAP core enzyme has a molecular mass of ≈400 kDa, and a conserved subunit composition of α2ββ′ω. The core enzyme possesses the catalytic activity, which includes the synthesis of RNA complementary to the DNA template in the presence of nucleoside triphosphates, and the degradation of nascent RNA (nucleolytic activity) through pyrophosphorolysis or hydrolysis. The core enzyme by itself is incapable of recognizing specific promoter DNA sequences, or of melting the DNA and initiating transcription. To carry out these functions, it must bind one of several specificity factors, σ, to form a holoenzyme (Gross et al., 1998). Alternative σ factors bind competitively to core, generating multiple forms of holoenzyme that can utilize different classes of promoters under various growth conditions (Ishihama, 2000).

In the absence of additional regulatory input, the gene expression level depends only on promoter strength and concentration of specific holoenzyme species. To orchestrate transcription of ≈4000 genes in a temporally and spatially co-ordinated manner during cell growth and development, and in response to different environmental signals, bacterial cells require a much higher level of complexity in transcription regulation. Indeed, the RNAP enzymatic activity in the cell is tightly regulated by a host of transcription factors that act on RNA, DNA or RNAP during all stages of transcription cycle. These factors include proteins, small peptides, non-coding RNAs (Escherichia coli 6S RNA), polyphosphates, amino acids, vitamins and other molecules (see Storz, 2002; Grundy and Henkin, 2004).

Among the estimated 285 E. coli protein factors known to affect transcription (Encyclopedia of E. coli K12 Genes and Metabolism, the vast majority (265) is DNA-binding proteins (Ishihama, 2000) that act during transcription initiation by either increasing or decreasing the occupancy of promoter by holoenzyme. Many activators recruit RNAP to specific promoter sites by direct protein–protein interactions, whereas repressors occlude promoters from RNAP or render the RNAP–promoter complex functionally inactive. Some anti-σ factors can  regulate  transcription  initiation  by  sequestering σ  from association with core, causing selective inhibition/activation of different σ-specific promoters (see Dove et al., 2003).

A small but growing number of proteins are known to act during elongation and termination stages of transcription by direct modification of RNAP properties. In E. coli, they include: Nus factors [NusA, NusB, NusG and NusE (S10)], RfaH, ribosomal protein S4, Gre-factors (GreA, GreB), Mfd, RapA (HepA) and ρ (Rho) ( Bailey et al., 2000; Squires and Zaporojets, 2000; Sukhodolets et al., 2001; Fish and Kane, 2002; Nudler and Gottesman, 2002; Roberts and Park, 2004). Also in this group are bacteriophage factors such as Alc of T4, P7 of Xp10, Q and N of λ and Nun of HK022 (Nechaev and Severinov, 2003). These factors affect RNAP processivity by modulating transcription pausing (temporary interruption of transcription), arrest (permanent stalling of elongation complex without dissociation), termination (release of RNA transcript and complex dissociation) or anti-termination (prevention of termination). Pausing, which may play an integral role in transcription termination or anti-termination, can occur through at least two pathways (Artsimovitch and Landick, 2000): (I) by RNAP interaction with the nascent RNA secondary structures (e.g. his- and trp-pause RNA hairpins) and (ii) by physical barriers to RNAP translocation caused by DNA-binding proteins, misincorporated substrates, DNA lesions and special DNA sequences (Fish and Kane, 2002). In the latter case, RNAP may ‘backtrack’, resulting in disengagement of the RNA 3′-terminus from the catalytic centre, and concomitant extrusion of the 3′ tail of the RNA through the secondary channel of RNAP (see below). Both types of pausing, under certain conditions, could lead to transcription arrest or dissociation of ternary elongation complex (TC) (Richardson and Greenblatt, 1996).

To understand the mechanism of action of elongation/termination factors, we must define the structural basis for RNAP processivity. Biochemical studies have identified four structural determinants in RNAP responsible for the stability of TCs: the downstream DNA-binding clamp, the RNA/DNA hybrid binding site, the single-stranded RNA binding site and the upstream RNA binding site (Nudler, 1999; Korzheva et al., 2000). This is corroborated by the high-resolution structures of RNAP and its complexes with nucleic acids that show three major channels in RNAP: (i) the primary channel, which includes the downstream DNA-binding clamp and the hybrid binding site, (ii) the RNA exit channel, which includes the single-stranded RNA and upstream RNA binding sites and (iii) the secondary channel, which allows substrate diffusion, and binds the extruded RNA 3′-terminus in backtracked TCs (Zhang et al., 1999; Vassylyev et al., 2002;). Many mutations affecting elongation/termination properties of RNAP localize in and around these channels, suggesting that they are involved in the maintenance of RNAP processivity, and are likely to be the sites of regulation by elongation/termination factors.

This review focuses on the structure–function relationship of three evolutionarily conserved elongation factors, Gre (GreA and GreB), NusA and Mfd. All three proteins have been characterized biochemically and functionally, and mechanistic models for their action have been described. The structures of GreA and NusA are known, and a partial structure of Mfd can be inferred by homology with the known structures of DNA helicases. Thus, with the wealth of structural information now available for RNAP, it is possible to model the interaction between these factors and RNAP, and glean new insight into the mechanisms that have evolved to regulate transcription elongation.

GreA and GreB

Prokaryotic transcription factors GreA and GreB suppress RNAP pausing and arrest in vitro and in vivo by stimulating the intrinsic nucleolytic activity of RNAP, which is an evolutionarily conserved function of all multisubunit RNAPs (Borukhov et al., 1993; Marr and Roberts, 2000; Toulme et al., 2000; reviewed by Fish and Kane, 2002). When RNAP encounters a roadblock during elongation and backtracks, the 3′-end of the nascent RNA gets backpedaled into the secondary channel of RNAP. Gre-induced cleavage of this extruded portion of RNA generates a new 3′-terminus, giving RNAP a second chance to transcribe over the roadblock and resume elongation. GreA-induced hydrolysis generates mostly di- and trinucleotides, while GreB-induced hydrolysis generates fragments of up to 18 nt long, depending on the extent of RNAP backtracking. GreA can only prevent transcription arrest; GreB on the other hand can reactivate pre-arrested TCs as well (Borukhov et al., 1993).

Besides anti-pausing and anti-arresting function, the factor-induced endonucleolytic reaction may enhance transcription fidelity by inducing the excision of misincorporated nucleotides, and facilitate transition of RNAP from the initiation to the elongation stage of transcription by preventing product release during abortive synthesis (Fish and Kane, 2002). Gre factors are evolutionarily conserved, and their homologues have been found in more than 60 bacterial species. E. coli cells lacking both Gre factors are viable, but temperature sensitive (Orlova et al., 1995).

All members of Gre family are ≈160-amino-acid-long polypeptides with similar structural organization consisting of two domains: an N-terminal coiled-coil domain (Gre-NTD) and a C-terminal globular domain (Gre-CTD) (Fig. 1A). The NTD is responsible for the induction of type-specific nucleolytic and anti-arrest activities, whereas the CTD is responsible for the high-affinity binding of Gre to RNAP (Fish and Kane, 2002). A cluster of positively charged residues on the surface of the NTD is responsible for the interaction of Gre with nascent RNA in the TC, and is required for anti-arrest function.

Figure 1.

Structural organization of Gre proteins and the models of Gre–RNAP complexes.
A. Ribbon structure of E. coli GreA (Stebbins et al., 1995) (left) with the side-chains of essential D41 and E44 shown in yellow. Charge distribution on the solvent accessible surface of GreA (middle) and GreB (right). The surface is coloured by the electrostatic potential: white, uncharged; red, negative (Asp, Glu); blue, positive (Arg, Lys). Only one side of the proteins is shown.
B. Two views of the 3-D structural model of GreA/GreB–TC complex shown in ribbon diagram with colour coding as follows: white, α; light green, β′; orange, β; dark green, ω; blue, GreA; yellow, non-template DNA; green, template DNA; red, nascent RNA; and magenta, catalytic Mg2+ ions. Left panel is the secondary channel view of RNAP. Right panel showing the main channel view is obtained by rotating the left view 90° clockwise about the vertical axis. The structural model of Gre–TC was generated using the structure of T. thermophilus holoenzyme (Vassylyev et al., 2002) and the models of T. aquaticus TC (adapted from Korzheva et al., 2000) and Gre–RNAP complex (adapted from Laptenko et al., 2003; Opalka et al., 2003).
C. Two structural models of Gre insertion into the RNAP secondary channel. Left panel shows the blow-up view of the right panel of (B), and the right panel shows an alternative model of Sosunova et al. (2003). The two Mg2+ ions and the side-chains of three catalytic aspartates are shown in magenta, and two complementing GreA residues, D41 and E44, are shown in yellow.

The molecular mechanism of Gre action and the structural models of Gre–RNAP and Gre–TC complexes were recently proposed independently by three laboratories based on biochemical, mutational and structural analyses and modelling (Laptenko et al., 2003; Opalka et al., 2003; Sosunova et al., 2003). In all three structural models, Gre-NTD protrudes into the RNAP secondary channel, placing the two invariant acidic residues of the NTD tip, D41 and E44, in the immediate vicinity of the catalytic site of RNAP (Fig. 1B and C). According to the models, the occupation of the secondary channel by Gre-NTD still allows diffusion of the substrates and extrusion of backtracked RNA. The mechanistic models of Gre-induced hydrolysis proposed by all three groups are based on the two-metal-ion mechanism of phosphoryl transfer reactions (Steitz and Steitz, 1993), and assume that one of the two catalytic Mg2+ ions is chelated by three invariant Asp residues in the RNAP catalytic site, while the second Mg2+ is weakly bound by RNAP. In the models, D41 and E44 of Gre directly participate in the co-ordination of the second Mg2+ D41 also co-ordinates the attacking water molecule (or hydroxide ion) to stabilize the pentavalent transition state of the phosphate. In the eukaryotic functional analogue of Gre, TFIIS, the two essential acidic residues D290 and E291 (Saccharomyces cerevisiae nomenclature) located at the tip of its RNA-binding Zn-ribbon domain play a similar role according to the three dimensional (3-D) crystal structure of TFIIS–Pol II complex (Kettenberger et al., 2003). Thus, Gre and TFIIS represent the first known examples of transcription factors that modify RNAP activity by directly participating in the catalytic processes.

In addition to co-ordinating the metal-ion, Gre-NTD may stabilize the backtracked TC conformation through its interaction with the G-loop and F-helix regions of β′ (Laptenko et al., 2003), and with RNA (through its basic patch residues), to facilitate hydrolysis. Detailed footprint analyses also suggest that Gre allosterically affects the downstream DNA-binding clamp (β′ downstream jaw domain) and the DNA/RNA hybrid binding site (β rifampicin binding region). Thus, Gre binding may affect the stabilization and destabilization of the interactions between downstream DNA and the DNA-binding clamp, and between the RNA/DNA hybrid and its binding site, both of which could affect the stability of TC during initiation and elongation (Ederth et al., 2002).

The models place Gre differently in spatial relation to RNAP; Laptenko et al. (2003) and Opalka et al. (2003) place the Gre-CTD at the rim of the secondary channel, where it hooks over the β′ coiled-coil (E. coliβ′ 647–704), while Sosunova et al. (2003) place the Gre-CTD parallel to the β′ coiled-coil. The first model maximally conforms to the extensive footprint data and is consistent with the results of electron microscopy, while the second model specifically accommodates the RNA-basic patch cross-linking data and satisfies the steric constraints in modelled backtracked TC. The two orientations of Gre in TC may represent two stages of Gre action or reflect the conformational flexibility of Gre domains.

Several other factors also modulate the activity of RNAP by their action within the secondary channel, such as Thermus thermophilus Gre-like repressor, Gfh1 (Laptenko and Borukhov, 2003), ppGpp-binding protein, DksA (Paul et al., 2004; Perederina et al., 2004), antibiotics microcin MccJ25 and streptolidigin (Adelman et al., 2004; Darst, 2004; Mukhopadhyay et al., 2004). Unlike the primary channel and the RNA exit channel, which are occupied by nucleic acids in TC, the secondary channel remains unoccupied, and provides a direct and unobstructed passage from solution to the catalytic site for potential regulatory factors. As more factors are likely to be discovered that specifically interact with RNAP secondary channel, it is considered an attractive target for future drug design.


NusA is an essential multifunctional transcription elongation factor that is universally conserved among eubacteria and archaea (Nudler and Gottesman, 2002). Depending on the RNA/DNA sequence context, and the presence or absence of auxiliary factors, NusA may elicit opposite effects on transcription (Richardson and Greenblatt, 1996). By itself, NusA stimulates certain types of pausing, such as those caused by his- or trp-pause hairpins, and ρ-independent intrinsic transcription termination. It can also induce anti-termination at ρ-dependent terminators with RNA sequences containing nut or nut-like elements (BoxA, B, C). In complex with other Nus factors (NusG, NusB, NusE) or λ phage proteins N and Q, it stimulates anti-termination at both ρ-dependent and ρ-independent terminators (Richardson and Greenblatt, 1996; Nudler and Gottesman, 2002). NusA anti-termination function plays an important role in the expression of ribosomal rrn operons. During transcription of many other genes, NusA-induced RNAP pausing provides a mechanism for synchronizing transcription and translation (Squires and Zaporojets, 2000). The multiple functionality of NusA is reflected by its multidomain structural organization (Fig. 2A). According to the high-resolution crystal structures of NusA obtained from Thermotoga maritima and Mycobacterium tuberculosis (Gopal et al., 2001; Worbs et al., 2001; Shin et al., 2003), it is an elongated protein composed of several discrete domains. The N-terminal RNAP-binding domain (NTD) is connected through a flexible hinge helix to three globular domains, S1, KH1 and KH2, with homology to RNA-binding motifs found in the ribosomal protein S1 and pre-mRNA-binding protein K. The three RNA-binding domains are held by interdomain interactions as a rigid body. In the structures of NusA of T. maritima and M. tuberculosis, these domains lie at different angles with respect to NTD (Shin et al., 2003) (Fig. 2B). The orientation of RNA-binding domains might change upon NusA binding to RNAP, or in response to specific sequences of the nascent RNA transcript. Finally, E. coli NusA has an additional C-terminal extension consisting of two auto-inhibitory domains, AR1 and AR2, which act by preventing S1 and KH domains from binding RNA by free NusA. The inhibition is relieved by AR1 and AR2 binding to the C-terminal domain of RNAP α-subunit (α-CTD) (Mah et al., 2000).

Figure 2.

NusA structure and the model of NusA–RNAP complex.
A. NusA domain organization and surface charge distribution. Top, ribbon structure of T. maritima (T.m.) NusA (Worbs et al., 2001; Shin et al., 2003) (left) is shown with structural domains indicated. Bottom, solvent accessible surface of NusA shown in the same orientation as above and coloured by the electrostatic potential (with colour coding as shown in Fig. 1A).
B. Structural comparison of M. tuberculosis NusA (yellow) (Gopal et al., 2001) and T.m. NusA (cyan) shown in the same orientation as in (A). The two molecules were aligned based on structural and sequence homology of NusA-N-terminal domains (NTDs).
C. Structural alignment of the RNAP-binding triple α-helical bundle of T. thermophilus (T.t.) σ (residues L192-K226 and F233-R259) (Vassylyev et al., 2002) and the corresponding homologous structural element in T.m. NusA-NTD (residues I3-K37 and F102-K125). Left panel shows the same view orientation of NusA as in (A) and (B). Right panel is obtained by rotating the left view 90° counterclockwise about the vertical axis.
D. Two views of the 3-D structural model of NusA–TC complex shown in ribbon diagram using the same colour coding for RNAP subunits and NusA as indicated in Figs 1B and 2A respectively. The model was generated using the structural model of TC (as in Fig. 1B) and the structure of T.m. NusA with RNA-binding domains rotated as in M.t. structure (see Fig. 2B). Left panel shows the upstream DNA entrance into the primary channel of RNAP. Right panel showing the RNA exit channel view of RNAP is obtained by rotating the left view 90° clockwise about the horizontal axis.
E. Two molecular surface views of NusA–TC complex with the same colour-coding as in (D). Left panel shows the same view as in the left panel of (D). Right panel view is obtained by rotating the left panel view 90° about the vertical axis as indicated.

Several mechanistic models explaining NusA actions have been proposed based largely on the results of biochemical and genetic studies (Richardson and Greenblatt, 1996; Nudler and Gottesman, 2002). These models can now be re-examined in the light of new structural information. Early biochemical studies have shown that NusA competes with σ70 for core-binding, and forms a complex with RNAP in TC when σ is released (Greenblatt and Li, 1981; Richardson and Greenblatt, 1996). This is supported by Fe-BABE footprinting analyses and mapping data which show that NusA binds to essentially the same region on E. coli RNAP core as σ70 (Traviglia et al., 1999), the N-terminal coiled-coil element in β′ (E. coliβ′ 264–308). It is now apparent from the 3-D structures of NusA and σ70 that a portion of the NusA-NTD, which is the core-binding determinant, shares strong structural homology with the region 2 of σ70, also an RNAP-binding domain (Fig. 2C). As in σ70–core interaction, the binding of NusA to core may be initiated by anchoring of NusA-NTD to the β′ coiled-coil (β′ 264–308), near the upstream opening of its primary channel. The structure of NusA–core complex can be modelled by replacing σ70 in the structure of holoenzyme with NusA, aligning NusA-NTD with σ70 region 2 (Fig. 2). Depending on which structure of NusA is used, the three RNA-binding domains project either above or below the RNAP β-flap domain. The latter model better accommodates the results of RNA–RNAP and RNA–NusA cross-linking and other biochemical and genetic data (Gusarov and Nudler, 2001; Toulokhonov et al., 2001).

The proposed structural model has several implications. First, like σ70 region 2, the positively charged surface of NusA-NTD is likely to contact the non-template strand DNA as it is emerging from the RNA/DNA hybrid, and/or part of the upstream duplex DNA (Fig. 2). This could result in both electrostatic (non-specific) and sequence-specific NusA–DNA interactions, which may contribute to transcription pausing at sequences lacking any specific secondary structures of RNA. Next, the S1 and KH domains of NusA located at the RNAP β-flap can form a complex with λ N, NusG, NusB and NusE factors; at the same time, they can interact with the distal secondary structures of RNA such as nut BoxB. As part of this anti-termination complex, the three RNA-binding domains of NusA could be well positioned to interact with the 5′-proximal arm of the RNA termination hairpin, and sequester it from annealing to the 3′-proximal arm. Consistent with the proposed mechanism of NusA enhancement of ρ-independent termination (Gusarov and Nudler, 2001), the S1 and KH domains may interact with the RNAP β- and β′-regions that form the RNA exit channel opening, and indirectly facilitate the formation of termination hairpin by disrupting the RNA–β,β′ contacts. On the other hand, it appears that the stimulation of pausing by NusA at his- and trp-pausing hairpins requires an active involvement of S1 and KH domains, both in the formation of RNA stem–loop structure and in strengthening its specific interactions with the structural elements of RNAP surrounding the RNA exit channel, specifically, the β-flap tip helix (Toulokhonov and Landick, 2003). These interactions may induce transcription pausing allosterically (Toulokhonov et al., 2001), or directly by pressing down on the flap domain, resulting in the constriction of RNA exit channel and the ‘trapping’ of nascent RNA.


Transcription repair coupling factor Mfd (TRCF) is an evolutionarily conserved protein of 130 kDa (1148 aa) which functions in the cell to reactivate or recycle stalled or arrested RNAPs during elongation (Selby and Sancar, 1993; Park et al., 2002; Roberts and Park, 2004). It does so by ‘reverse backtracking’ RNAP, allowing its catalytic centre to re-engage the RNA 3′-end (Roberts and Park, 2004). Mfd also recruits DNA excision-repair machinery to damaged DNA sites in a transcription-coupled manner through its recognition of stalled RNAPs. E. coli Mfd is structurally related to superfamily II helicases, specifically to RecG, a protein involved in the DNA recombination/repair processes in bacteria (Mahdi et al., 2003) (Fig. 3A). The C-terminal half of Mfd comprises seven conserved helicase motifs (aa 597–906), a translocase motif TRG (aa 926–965) and the C-terminal region (aa 966–1148) (Roberts and Park, 2004). Helicases and translocases are proteins that use the chemical energy of NTP hydrolysis to effect nucleic acid strand separation and protein movement along DNA or RNA respectively. The helicase motifs provide the ATPase and, most probably, the double-stranded DNA (dsDNA)-binding functions, while TRG, in analogy with RecG, may provide the motor function for translocation (Chambers et al., 2003). The C-terminal region may contribute to both RNAP binding and its release. The N-terminal portion of Mfd contains a region of homology to UvrB that is necessary for the recruitment of excision repair complex UvrA2B to RNAP, but otherwise is not needed for Mfd activity (Roberts and Park, 2004). The mid-section of Mfd polypeptide (aa 379–571) is responsible for RNAP binding (Selby and Sancar, 1993).

Figure 3.

Structural organization of Mfd and the model of Mfd–TC complex.
A. Functional map of Mfd. RNAP-binding domain and conserved regions of sequence homology to UvrB and RecG are shown schematically on the linear diagram of the 1148-amino-acid-long E. coli Mfd protein and are colour coded.
B. Ribbon structure views of the modelled RecG-like translocase domain of Mfd protein (residues 553–966) (left) and its complex with dsDNA (right). Position of Mg2+ ion in the ATPase centre is shown in dark magenta. The other colour coding is the same as in (A). The location of functionally important elements is indicated. Residue R905 of the sensor α-helix forms a salt bridge with the γ-phosphate of ATP; D889 of the same helix forms hydrogen bonds with the conserved R929 and R953 of TRG helical hairpin; Q963 of the ratchet loop contacts DNA. The right panel view is obtained by rotating the left view 90° counterclockwise about the vertical axis. Models were generated using the structure of T. maritima RecG complex with replication fork DNA (Chambers et al., 2003; Mahdi et al., 2003).
C. Two views of the ribbon structure model of Mfd–TC complex are shown using the same colour coding for RNAP subunits as in Fig. 1B. For Mfd only the TRG motif and helicase motif VI are indicated by different colours. The Mfd RNAP-binding domain (for which the structural model is not available) is shown schematically as dark yellow oval. The segment of T. thermophilus RNAP β-subunit (corresponding to 1–142 of E. coli) which interacts with Mfd is coloured bright yellow. Right panel view shows the same orientation of TC as in the right panel of Fig. 1B. Left panel view is obtained by rotating the right view by 45° about the vertical axis and 45° about the horizontal axis.

The binding site for Mfd on RNAP resides within the N-terminal 142 residues of RNAP β-subunit (Park et al., 2002) which, in the structural model of TC (Korzheva et al., 2000), localize near the upstream junction of the transcription bubble (Fig. 3). Consistent with this, Mfd requires unobstructed access to ≈25 bp of the duplex DNA immediately upstream of the transcription bubble to form a complex with TC (Park et al., 2002). Moreover, the presence of σ70 interferes with Mfd binding. These data indicate that Mfd binds to TC in the vicinity of the upstream opening of RNAP primary channel.

Mfd, which is not a helicase (meaning, it is unable to unwind DNA), nevertheless possesses both ATPase and dsDNA-binding activities and, like RecG protein, appears to be an ATP-dependent translocase. A mechanistic model of Mfd action has been proposed (Fig. 3B) based on the sequence homology between Mfd and RecG, and the available structure of RecG–fork DNA complex (Chambers et al.,  2003; Mahdi et al.,  2003).  According  to the model, Mfd is recruited to TC through its RNAP-binding domain. In the presence of ATP, the dsDNA-binding domain binds the duplex DNA immediately upstream of the transcription bubble (Fig. 3C). The model proposes that an α-helix located in the helicase motif VI acts as a sensor, and changes its conformation upon ATP hydrolysis, disrupting an existing network of hydrogen bonds with a helical hairpin structure in the TRG motif. Distortion of the TRG helical hairpin may translate to a ‘swinging arm’ motion of an adjacent loop structure, which enters the dsDNA-binding cleft and establishes new contacts with DNA (Fig. 3B). The loop may thus act as a mechanical lever or a ratchet. Because of tethering of RNAP to Mfd, and specific orientation of the ratchet, its action pushes RNAP and DNA in opposite directions.

When Mfd encounters a backtracked TC, it pushes RNAP forward until the RNA 3′-end is in proper register with the catalytic centre. However, in the presence of intractable roadblocks (such as strongly bound proteins or non-pairing lesions in DNA), or low NTP concentrations, the force generated by the ratcheting motion of Mfd action may be used to dislodge RNAP from DNA, thereby allowing arrested enzyme to recycle. The rate of Mfd translocation depends on the turnover rate of ATP hydrolysis, which, at three molecules per minute, is relatively slow (Selby and Sancar, 1993). Moreover, Mfd may be released from TC upon ATP hydrolysis, which implies that Mfd may bind and dissociate from the target TC several times during its course of action. Thus, it may not act on TCs moving at the normal speed of elongation, and the target of Mfd action may be limited to slow moving or stalled TCs.

Another bacterial helicase-like transcription factor associated with RNAP is the 110 kDa RapA (HepA) protein, which belongs to SWI/SNF superfamily of helicase-like proteins (Muzzin et al., 1998; Sukhodolets et al., 2001). Like Mfd, RapA is not a helicase, but has seven putative helicase motifs and possesses an ATPase activity that is stimulated by RNAP. Unlike Mfd, it binds all types of nucleic acids with a weak preference for DNA/DNA and DNA/RNA duplexes. RapA and RNAP form a tight binary complex in which the RapA binding site appears to be at the interface between α- and β′-subunits (Sukhodolets et al., 2001). Unlike Mfd, which acts on any stalled TCs, RapA is functionally specialized and recognizes only TCs that are arrested at intrinsic termination sites under special conditions.

RapA stimulates multiround transcription of RNAP both in vivo and in vitro by remodelling the post-transcription/post-termination complex (PTC) on tightly supercoiled DNA induced at high salt concentrations (Sukhodolets et al., 2001). Similar to Mfd, RapA may also use the energy of ATP hydrolysis to facilitate the release of RNAP trapped in arrested PTCs, thereby promoting RNAP recycling. Consistent with this view, rapA deletion causes severe cell growth inhibition at high salt concentrations. Many others factors or conditions may exist that induce hypersupercoiling of chromosomal DNA, contributing to persistent arrest of PTC. Thus, RapA anti-arrest activity could be an important component of the stress–response system. It is still unclear how DNA supercoiling could lead to termination complex arrest, and what the precise molecular mechanism is that allows RapA to release RNAP trapped in such complexes.

Concluding remarks

Transcription elongation and termination complexes are important targets for regulatory factors in the cell. Ultimately, our understanding of molecular mechanisms of this regulation will rely on high-resolution structures of multicomponent elongation and termination complexes, including intermediate complexes, such as backtracked, paused, or pre-termination TCs. The 3-D structures of bacterial RNAP core and holoenzyme are now being used to build models of higher-order structures containing nucleic acid components and transcription factors such as CRP (Lawson et al., 2004), λcI (Jain et al., 2004) and others. As illustrated in this review of GreA/GreB, NusA and Mfd, computer-assisted modelling can provide preliminary views of possible structures of elongation/termination complexes. This approach can be applied to NusB and NusG, for which the structures are known, as the relevant biochemical data will become available that identify the interaction sites on RNAP. For other elongation factors such as N, Q, Nun and RfaH, we first need to obtain their high-resolution structures alone or in complex with RNA recognition elements.

The reconstructed models can then be tested directly by biochemical and genetic experiments. Furthermore, additional experiments should be designed to address more global and physiological aspects of transcription elongation and termination processes. For instance, NusG, Mfd, RapA, GreA and GreB appear to have overlapping or redundant functions in vivo. Individually, each of these factors affects the in vivo phenotypes of E. coli cells to different extents, but none of them (except NusG) is essential for viability. This raises the possibility of functional interplay among elongation/termination factors, such as synergistic effects between Gre, Mfd, NusG, RapA or competition between NusA, Mfd and RapA and other elongation factors. Like NusA, which can manifest opposite functions when it is in complex with NusG, NusB, NusE and N, other elongation/termination factors may also use complex formation to diversify their functions.


Research in Sergei Borukhov's laboratory is funded by a grant from NIH. We are grateful to reviewers of this manuscript for their comments and helpful suggestions. We apologize to those whose work was not cited because of space limitations.