Transcription factor dynamics

Prokaryotic RNA polymerase (RNAP) is a large (∼400 kDa) multi-subunit enzyme comprising α₂ββ'ω subunits which form a crab-claw-like structure. Although little sequence homology exists between eubacterial RNAP, archaeal RNAP and eukaryotic RNAPII, the crab-claw structure is remarkably conserved (Zhang et al., 1999; Cramer et al., 2001; Hirata et al., 2008). The two α subunits act as a scaffold to hold the catalytic β and β' subunits together, forming the crab-claw-like structure (Zhang et al., 1999). The exact role of the ω subunit is unclear but it is related in both structure and sequence to the eukaryotic polymerase subunit Rpb6 (Minakhin et al., 2001). It appears to be responsible for controlling transcription in response to nutrient shifts, correct folding of the β' subunit and its assembly into the core multi-subunit enzyme (Mukherjee et al., 1999; Vrentas et al., 2005; Chatterji et al., 2007). The channel formed by β and β' is referred to as the primary channel, which contains a deep positively charged cleft housing the enzyme's active site. During transcription, downstream double-stranded DNA separates into a single-stranded DNA template, which enters the primary channel and contacts the active site to allow polymerization of RNA (Borukhov & Nudler, 2008). Due to the crowding of the primary channel by the DNA : RNA hybrid, nucleotide triphosphates (NTPs) must access the active site through an alternative route. They do this through a pore on RNAP called the secondary channel (sometimes referred to as the nucleotide entry channel), which allows unobscured access to the active site not only for NTPs but also for other regulatory proteins and molecules. The elongating RNA molecule is separated from the DNA template by a wedge-like domain on RNAP to redirect the nascent RNA molecule through a third channel called the RNA exit channel, which then allows upstream DNA to reanneal as it exits RNAP (for a review see Borukhov & Nudler, 2008). Many of the regulatory roles of transcription factors are exerted through interaction with these structural elements (Borukhov et al., 2005).

For initiation of transcription to occur, RNAP must first associate with a σ factor to form what is termed the holoenzyme (α₂ββ'ωσ), which allows RNAP to recognize and bind promoter DNA sequences. The housekeeping σ factor is σ⁷⁰ in Escherichia coli and σ^A in Bacillus subtilis and is responsible for initiating transcription from most promoters. Other σ factors also exist and are usually stress induced to allow the organism to become virulent or adapt to any number of environmental cues such as hyperosmolarity, heat shock, oxidative stress, nutrient deprivation and variations in pH (Gruber & Gross, 2003; Kazmierczak et al., 2005).

Following σ factor dissociation and promoter escape, RNAP enters the elongation phase of transcription, where a raft of other co-factors combine to influence the sensitivity of RNAP to regulatory pause and termination signals. There are two basic classes of elongation complexes (ECs): the antitermination ECs involved in rRNA synthesis, and mRNA ECs, which respond in dramatically different ways to intrinsic and extrinsic pause and termination signals (Condon et al., 1995; Landick et al., 1996; Torres et al., 2004). As the name suggests, mRNA ECs are responsible for transcribing protein-encoding genes. These complexes are highly sensitive to both intrinsic and extrinsic pause and termination signals to ensure efficient regulation of gene expression, particularly in response to cellular metabolic needs, while maintaining the fidelity of the transcript. Antitermination ECs are highly processive, capable of synthesizing transcripts several kilobases in length, and are usually associated with the transcription of rRNA operons (Condon et al., 1995). They are largely resistant to pause and termination signals and have an overall elongation rate double that of mRNA ECs (Vogel & Jensen, 1994), and this helps satisfy the cellular thirst for rRNA at high growth rates. Although representing only about 1 % of chromosomally encoded genes, rRNA accounts for up to 80 % of cellular RNA at high growth rates to ensure the demand for ribosome production is met, making the formation of antitermination complexes vitally important.

Some of the better-studied transcription factors involved in regulating ECs are the Nus and Gre factors. NusA and NusG are global transcription elongation factors that regulate both classes of ECs. During mRNA transcription NusA acts to decrease the rate of transcription elongation, which has been shown to be important for ensuring the efficient coupling of transcription and translation (Landick et al., 1996; Burns et al., 1998; Ingham et al., 1999; Gusarov & Nudler, 2001). NusA has a somewhat contradictory role when it comes to regulating elongation of antitermination ECs. Instead of increasing the half-life of paused complexes as it does during mRNA transcription, NusA increases the elongation rate of antitermination complexes (Richardson & Greenblatt, 1996; Burns et al., 1998). The mechanism by which NusA achieves such opposing roles in transcription remains unclear, but it may have to do with the stoichiometry of NusA to RNAP. It has been found that during mRNA transcription there is one NusA molecule involved, while antitermination complexes have two, the second of which probably binds a conserved box element in the nascent rRNA transcript (Davies et al., 2005). NusG acts to increase elongation rates of both mRNA and antitermination ECs by decreasing pausing (Burns et al., 1998). NusG has been found to be essential in Gram-negative bacteria, most likely due to interactions with the termination factor Rho during mRNA transcription, whilst a drastically reduced growth rate is observed when NusG is knocked out in Gram-positives (Li et al., 1993; Ingham et al., 1999; Zellars & Squires, 1999; Pasman & von Hippel, 2000). This may be due in part to the much lower levels of Rho present in Gram-positive organisms such as B. subtilis, compared to the Gram-negative E. coli (Ingham et al., 1999). NusB and NusE (also known as ribosomal protein S10) are only known to play roles in regulating antitermination ECs. NusB is the first transcription factor to be recruited to antitermination ECs and binds the nascent rRNA transcript during the initial stages of elongation. NusE then forms a heterodimer with NusB to stabilize this complex, which allows the other factors such as NusA and NusG to bind (Greive et al., 2005). Each of the Nus factors are essential for antitermination, and (at the very least) cause increased doubling times in organisms when their genes are mutated or deleted.

Potential obstacles to transcription elongation that ECs encounter in vivo include DNA lesions and DNA-binding proteins, which can act as road blocks. In some instances these cause the EC to stall, which can lead to a further build up of trailing ECs. These stalled arrays of transcription complexes are not only highly detrimental to transcription, but can also block DNA replication forks (Borukhov et al., 2005). The Gre factors are involved in restarting these paused complexes to relieve the genome of these obstacles (Borukhov et al., 1993; Erie et al., 1993; Orlova et al., 1995). Although the Gre factors have only been thought to regulate mRNA ECs, there is some evidence that they may also be involved in regulating initiation from both mRNA and rRNA promoters (Hsu et al., 1995; Potrykus et al., 2006; Stepanova et al., 2007).

Bacterial DNA often consists of a single circular chromosome containing genes that are often grouped into operons. These operons are usually transcribed from a single promoter and contain genes of related function, allowing the cell to streamline the control of transcription. The order and orientation of genes and operons around the chromosome is highly strategic and driven by biophysical forces to ensure the cell can rapidly adapt to environmental cues by regulating gene expression. This is thought to be the reason why genes encoding transcription factors such as lacI are often in close proximity to the operons they regulate, in this case the lac operon (Kolesov et al., 2007). These forces have also driven the duplication of rRNA operons to maximize rRNA expression, which is directly linked to growth rate. Rapidly growing organisms such as Vibrio natriegens contain up to 13 rRNA operons and have a doubling time of 10 min, whilst a slower-growing organism such as Mycobacterium tuberculosis contains a single rRNA operon and has a doubling time of over 24 h. B. subtilis and E. coli contain 10 and 7 rRNA operons, respectively, and both have a doubling time under 20 min in rich media (see Lewis, 2007, for references on rRNA operon levels in bacterial genomes).

Chromosome replication occurs bi-directionally from the origin until the replication forks meet at the termination site. Therefore, during the replication cycle, there is more origin-proximal template DNA available for transcription than sequences closer to the termination site. Due to the ability of rapidly growing bacteria to undergo multi-fork DNA replication, up to three rounds of replication can be ongoing in a cell, exponentially increasing the transcription potential of sequences located close to the origin of replication (Couturier & Rocha, 2006). Rapidly growing organisms have taken advantage of this gene dosage effect by localizing highly transcribed genes around the origin. It is, therefore, no coincidence that rRNA operons are clustered in groups in the origin-proximal half of the chromosome in both B. subtilis and E. coli (Fig. 1A).

(24K): Fig. 1. Subcellular organization of RNAP in B. subtilis and E. coli. The circular maps of the 4.2 Mb B. subtilis (left) and 4.6 Mb E. coli (right) chromosomes are shown in (A), with the position of rRNA (rrn) operons shown relative to the origin of chromosome replication (oriC) and the terminus (terC). The images in (B) present the subcellular localization of RNAP through GFP-tagging of the C terminus of the β' subunit in B. subtilis (left) and E. coli (right), showing the accumulation of RNAP into transcription foci (TF; red arrows) in both species. Cartoons are shown below the cells to illustrate cell dimensions. There are two B. subtilis cells each containing two nucleoids and one E. coli cell containing two nucleoids. In both images cells are growing rapidly and contain two nucleoids on which RNAP is assembling, as is depicted in the cartoon in (C). RNAP localized to TF represents enzyme heavily loaded onto rRNA operons as well as mRNA encoding structural genes, whereas less intensely fluorescent regions represent loading principally onto structural genes (C). The cartoon cell depicted has two nucleoids (ovals) each with two TF and is a representation of the cells shown on the left in (B). Thus, the two nucleoids in this cartoon cell represent chromosomes that are undergoing replication that contain two segregated origin regions which also correspond to the TF (see text for details). The E. coli cells shown in (B) were taken from Fig. 1 of Cabrera & Jin (2006) with permission from Elsevier.

Nucleoli are nuclear organelles dedicated to the transcription of rRNA and have generally been assumed to be unique to eukaryotes. The lack of a nuclear membrane, let alone subnucleoid organelles, would strongly suggest that prokaryotes do not contain structures akin to the eukaryotic nucleolus. Furthermore, rRNA operons are generally separated by other highly expressed mRNA genes such as those encoding the transcriptional and ribosomal machinery. It would, therefore, seem reasonable to assume on the face of this that no nucleoli-like structures exist in prokaryotes. However, examination of GFP-labelled RNAP from E. coli and B. subtilis indicated that such a subcellular structure could exist (Lewis et al., 2000; Cabrera & Jin, 2003). During periods of rapid growth, RNAP-GFP localizes in a region coincident with the nucleoid, in a manner expected of a DNA-binding protein. Closer examination has revealed that RNAP-GFP localizes in discrete subnucleoid regions of higher intensity, called transcription foci (TF; arrows in Fig. 1B). Several lines of evidence have shown that these foci predominantly represent sites of rRNA synthesis (Lewis et al., 2000; Lewis, 2007). One of these lines of evidence showed that by synthetically inducing the stringent response, which directs transcription away from rRNA operons towards biosynthetic operons, these TF were found to disappear in GFP-labelled RNAP strains of both B. subtilis and E. coli (Lewis et al., 2000; Cabrera & Jin, 2003; Davies et al., 2005). Furthermore, relA mutant strains in E. coli that are unable to synthesize the small molecule alarmone ppGpp, and do not undergo the stringent response, retain TF under these conditions (arrows in Fig. 2; Cabrera & Jin, 2003). High transcriptional activity is known to cause nucleoid compaction due to overwinding of downstream DNA, and this can also be observed in Fig. 2, where the nucleoids in wild-type cells become more diffuse on induction of the stringent response, whereas those of the relA mutant remain highly compacted (Cabrera & Jin, 2006). These data strongly suggest that TF represent the sites of rRNA synthesis. However, with rRNA operons scattered around half of the chromosome in both B. subtilis and E. coli (Fig. 1A) an important question to address was whether these nucleoli-like structures contained only rRNA operons, and if so, how could these structures form if the rRNA operons were widely distributed around the chromosome? One model of how this may occur is through complex folding of the nucleoid. The more origin-distal operons could be brought to a nucleolus-like structure by reorganizing the nucleoid to co-localize all of the rRNA operons. However, the nucleoid is not static and this model fails to take into account the cell cycle of rapidly growing bacteria, in which multi-fork replication and simultaneous chromosome segregation is under way, and in B. subtilis TF have been shown to duplicate and segregate in a manner very similar to that of origin regions (Lewis, 2007). It is difficult to resolve how the newly replicated sequences could be brought to a nucleolus-like structure and at what stage in the replication/segregation cycle this would occur. Indeed, work in B. subtilis suggests that the formation of nucleoli does not occur, and that TF represent the heavy loading of RNAP onto the seven rRNA operons clustered close to the origin of replication (Davies & Lewis, 2003). In this work several of the rRNA operons were labelled through insertion of lacO repeats to which fluorescently labelled LacI could bind, and the origin regions were fluorescently co-labelled (using the origin-organizer Spo0J); the results indicated little co-localization of the more origin-distal rRNA operons with oriC regions. Although there is little doubt that the signal from TF is predominantly attributed to the transcription of rRNA operons, a small proportion must also be attributed to highly transcribed mRNA genes that are located around the origin of replication. Although still highly transcribed, we propose that the signals from the more origin-distal rRNA operons are probably drowned out by the noise from highly transcribed mRNA genes adjacent to them. Thus, TF represent the heavy loading of RNAP onto origin-proximal rRNA genes, plus RNAP involved in mRNA synthesis in that region, whereas less intensely fluorescent regions represent RNAP loading onto origin-distal structural genes (Fig. 1C).

(45K): Fig. 2. TF disappear on induction of the stringent response in E. coli. The figure shows a time-course of cells of the wild-type (wt, top) and a relA mutant (bottom) following exposure to the amino acid analogue serine hydroxamate, used to induce the stringent response. Representative TF are indicated with arrows. The bright TF in the wild-type cells rapidly disappear on exposure to serine hydroxamate so that within 10 min, no TF are visible and fluorescence is due to GFP-labelled RNAP. Conversely, in the relA mutant, which is unable to synthesis ppGpp, the stringent response is not induced by serine hydroxamate and TF remain clearly visible throughout the time-course. Figure adapted from Fig. 5 of Cabrera & Jin (2003) with permission.

Since we now know the sites of different classes of transcription, a great deal of information can be gained from determining the subcellular location of transcription factors. The transcription factors NusA, NusB, NusG and GreA are highly conserved throughout prokaryotes and are even present in the minimal genome of Mycoplasma genitalium. Using the same system as that used for tagging B. subtilis RNAP in vivo (Lewis & Marston, 1999; Lewis et al., 2000), functional GFP fusions were made to NusA, NusB, NusG and GreA and their subcellular localization patterns observed (Davies et al., 2005; Doherty et al., 2006). The localization of these fusions is shown in Fig. 3, and all have been shown to be nucleoid associated (Davies et al., 2005; Doherty et al., 2006). However, they display vastly different patterns which reflect the biochemical roles they have been assigned. NusA and NusG, as mentioned earlier, are global transcription factors in E. coli and are known to participate in both mRNA and antitermination EC formation (Richardson & Greenblatt, 1996; Ingham et al., 1999). Similar to RNAP, both of these are localized throughout the nucleoid as well as being recruited to TF (Fig. 3A, D). In the case of NusA, these foci have been shown to co-localize with TF from RNAP in a dual-labelled strain (Davies et al., 2005) and disappear (as do NusG and B foci) on induction of the stringent response (Fig. 3B). As for RNAP, this strongly suggests that the foci formed by Nus factors predominantly represent sites of rRNA synthesis. NusB shows a striking pattern of highly defined TF, with very little signal dispersed throughout the rest of the nucleoid (Fig. 3C), consistent with biochemical studies from E. coli and Mycobacterium tuberculosis suggesting that it is restricted to antitermination complex formation (Torres et al., 2004; Greive et al., 2005). Furthermore, a dual-labelled strain showed that these foci co-localize with RNAP (Doherty et al., 2006), and are rapidly reduced upon induction of the stringent response. One striking difference between NusB (and NusA to a lesser extent) and RNAP localization to TF is observed in slow-growing cells when NusB (and NusA) TF are still clearly visible, but they are almost totally absent with the RNAP-GFP fusion (Lewis et al., 2000; Davies et al., 2005; Doherty et al., 2006). This is due to the relatively low level of recruitment of RNAP to rRNA operons at low growth rates, when the demand for ribosomes is much less than at higher growth rates (Lewis et al., 2000). However, NusB is still required for antitermination complex formation at low growth rates, and so TF can still be observed under these conditions, making this a useful fusion for monitoring sites of rRNA synthesis at all growth rates (Doherty et al., 2006). Interestingly, we think the well-defined foci observed with the NusB-GFP fusion represent antitermination complexes loaded on individual rRNA operons or clusters. In the boxed cell in Fig. 3C, the white asterisk indicates several overlapping foci, likely to represent ECs loaded onto origin-proximal rRNA operons, whereas the yellow asterisk most likely represents loading onto an origin-distal operon, such as rrnB (see Fig. 1A). GreA localizes homogeneously throughout the nucleoid and shows no evidence of being recruited to TF, which is consistent with its biochemical role of predominately regulating mRNA ECs (Fig. 3E).

(43K): Fig. 3. Localization of GFP-labelled transcription elongation factors in B. subtilis. The distribution pattern of NusA is shown in (A); it is very similar to that of RNAP. Panel (B) shows the distribution of NusA following induction of the stringent response. TF have disappeared and fluorescence is homogeneously distributed throughout the nucleoids, showing that NusA foci also correspond to sites of rRNA transcription. The distribution of NusB is shown in (C) and is largely restricted to sites of rRNA synthesis. The cell marked with an arrow is magnified in the inset box; the foci marked with the white asterisk represent NusB thought to be loaded onto rRNA transcription elongation complexes on origin-proximal rrn genes, and the yellow asterisk possibly represents NusB loaded onto complexes transcribing an origin-distal rrn (see text for further details). The localization of NusG is shown in (D), and is very similar to that of RNAP and NusA. Panel (E) shows GreA localization. Fluorescence is homogeneously distributed throughout the nucleoids, suggesting little or no role for GreA in transcription of rRNA genes. Panel (F) shows GreA distribution following treatment of the culture with high concentrations of chloramphenicol to collapse the nucleoids. GreA signal remains coincident with nucleoids, showing that despite the molar excess of GreA over RNAP, it is specifically bound to RNAP, suggesting that there is more than one binding site for Gre factors. Cells in all panels are shown at the same magnification. Scale bar, 5 µm.

As these fusions were created by single crossover into the locus of the respective wild-type gene, they are all driven by the wild-type promoter. For this reason, all expression levels should be equal to that of the wild-type protein, allowing quantitative analysis to be performed. Using a combination of native PAGE, quantitative Western blots and image analysis, the cellular levels, along with the percentage of each respective protein that was recruited to TF was determined for RNAP, NusA, NusB, NusG and GreA (Davies et al., 2005; Doherty et al., 2006). This allowed a model for the composition of mRNA and rRNA antitermination complexes to be determined; it is summarized in Table 1. For each RNAP there are two NusA molecules (confirmed by quantitative pull-down assays; Davies et al., 2005) and one molecule each of NusB and NusG in rRNA antitermination complexes, whilst during mRNA transcription there is just one NusA, one NusG and possibly as many as three GreA molecules needed to form mRNA transcription complexes. The high levels of GreA are real and similarly large amounts of GreA and GreB have been reported in E. coli (Koulich et al., 1998). However, structural models suggest that only a single Gre factor is required to bind RNAP (via the secondary channel) to exert its biochemical role (Opalka et al., 2003). Nevertheless, the high amounts of GreA are most likely RNAP-associated. Gre proteins bind RNAP, not nucleic acids, and on compaction of nucleoids with high levels of chloramphenicol, GreA remains coincident with them (Doherty et al., 2006; Fig. 3E, F). Also, on overexpression of untagged GreA in the GreA-GFP labelled strain, fluorescence becomes delocalized, indicating that the high natural levels of Gre factors represent binding to specific sites on RNAP that are usually saturated (Doherty et al., 2006). The reason why there are such high levels of Gre factors in the cell, and where they bind on RNAP, remains to be determined.

Table 1. Summary of the localization and quantification data for B. subtilis RNA polymerase and transcription elongation factors

Although not as defined as in eukaryotes, it is clear that significant partitioning of transcription exists in prokaryotes. Through selective tagging of the transcription machinery we can identify global events as well as specific classes of transcription, such as rRNA synthesis using NusB. Using standardized growth media and imaging approaches we can determine the relative levels of different components of the transcription apparatus at an individual cell level, and combined with quantitative in vitro approaches, determine the composition of complexes involved in different classes of transcription (Davies et al., 2005; Doherty et al., 2006). As imaging techniques continue to improve it will also be possible to monitor the dynamics of transcription complexes in more detail using approaches similar to those that have been adopted for analysis of signal transduction in Rhodobacter spheroides and DNA replication in E. coli (Leake et al., 2006; Reyes-Lamothe et al., 2008). We can also monitor dynamic responses to stimuli, such as induction of the stringent response (Figs 2 and 3; Cabrera & Jin, 2003), showing that live cell monitoring could represent a rapid, cheap and sensitive approach to assessing the effects of novel antimicrobial compounds. It is interesting to note that the levels of transcription factors are very high within the cell (close to equimolar with RNAP), suggesting that the bulk of RNAP within the cell is present in the form of ECs. Genomic CHIP-ChiP experiments that are under way in several laboratories will provide information on global RNAP and transcription factor distribution and enable a more detailed analysis of ECs involved in transcription of individual genes/operons. Finally, through investigation of the subcellular localization pattern it will be possible to assign (at least partially) function to unknown proteins identified through isolation of protein complexes by various affinity approaches currently being used to investigate transcription complex composition. Work in the Lewis laboratory is supported by funding from the ARC, NHMRC, DEST and the University of Newcastle. The authors wish to thank Din Jin for permission to use data in this review. Due to the limitations of minireviews, it has not been possible to use primary references for all sources of information, and the authors wish to acknowledge the work of those whom we have not been able to directly include in this article.

Footnotes

†Present address: Flinders Microscopy and Image Analysis Facility, Department of Anatomy and Histology School of Medicine, Flinders University, GPO Box 2100, Adelaide 5001, Australia.