Subcellular Partitioning of Transcription Factors in Bacillus subtilis

Abstract

RNA polymerase (RNAP) requires the interaction of various transcription elongation factors to efficiently transcribe RNA. During transcription of rRNA operons, RNAP forms highly processive antitermination complexes by interacting with NusA, NusB, NusG, NusE, and possibly several unidentified factors to increase elongation rates to around twice those observed for mRNA. In previous work we used cytological assays with Bacillus subtilis to identify the major sites of rRNA synthesis within the cell, which are called transcription foci. Using this cytological assay, in conjunction with both quantitative native polyacrylamide gel electrophoresis and Western blotting, we investigated the total protein levels and the ratios of NusB and NusG to RNAP in both antitermination and mRNA transcription complexes. We determined that the ratio of RNAP to NusG was 1:1 in both antitermination and mRNA transcription complexes, suggesting that NusG plays important regulatory roles in both complexes. A ratio of NusB to RNAP of 1:1 was calculated for antitermination complexes with just a 0.3:1 ratio in mRNA complexes, suggesting that NusB is restricted to antitermination complexes. We also investigated the cellular abundance and subcellular localization of transcription restart factor GreA. We found no evidence which suggests that GreA is involved in antitermination complex formation and that it has a cellular abundance which is around twice that of RNAP. Surprisingly, we found that the vast majority of GreA is associated with RNAP, suggesting that there is more than one binding site for GreA on RNAP. These results indicate that transcription elongation complexes are highly dynamic and are differentially segregated within the nucleoid according to their functions.

Unlike the situation in eukaryotes, a single RNA polymerase (RNAP) is responsible for carrying out transcription of all classes of genes in prokaryotes. As rRNA synthesis is the rate-determining step in ribosome assembly, the efficiency of this process determines the rate at which organisms can produce proteins and ultimately divide (23, 37). The rate of rRNA transcription has also been found to be around twice the rate of mRNA transcription (52), and this is due to the types of transcription elongation complexes (EC) formed during these two types of transcription. The vast majority of stable RNA (rRNA and tRNA) is produced by antitermination ECs that are resistant to transcription pause signals and ensure that full-length transcripts are efficiently produced (1, 11, 52), although transcription of tRNA genes that are located outside rRNA operons does not appear to depend on the same regulatory sequences that are responsible for the antitermination complex formation that occurs in rRNA operons (Doherty, unpublished data). mRNA is produced by ECs that tend to be highly susceptible to pausing, and this appears, among other things, to help ensure the close coupling of transcription and translation, transcription fidelity, to enhance responsiveness to transcription attenuators (5, 27, 29, 54).

The ability of RNAP to react so differently during the two types of transcription is dependent on the activity of transcription elongation factors that become associated with RNAP once it has cleared the promoter. The Nus and Gre factors have been studied intensively for many years, and much is known about how these factors regulate transcription elongation (4, 23, 37, 40, 49, 52). The Nus (N utilization substance) factors were first discovered due to their role in formation of terminator-resistant complexes during phage λ transcription, but they have since been found to be essential for regulation of host cell (Escherichia coli) transcription and are highly conserved proteins in both gram-positive and gram-negative organisms (3, 28, 37, 39).

NusA is an essential protein that plays a role in both transcription antitermination and pausing (37, 40). Although the precise mechanism by which NusA is able to exert these contradictory activities has not been fully elucidated, it is dependent, in part, on the activities of other elongation factors (13, 19, 40). There is also some evidence that the ratio of NusA to RNAP is important in λ antitermination complexes and in rRNA complexes from Bacillus subtilis that have been reported to contain two molecules of NusA (13, 22). Other elongation complexes are thought to contain only a single molecule of NusA, based on sedimentation equilibrium, molecular modeling, and transcription pull-down experiments (6, 13, 18). In E. coli, NusG increases elongation rates but also interacts with Rho to control both efficient elongation and termination (7, 8). The interaction of NusG with Rho appears to be less important in B. subtilis, possibly due to a more prevalent Rho-independent termination system (23). NusB and NusE (ribosomal protein S10) appear to have roles that are largely, if not exclusively, restricted to rRNA synthesis (19, 33, 37).

The transcription restart Gre factors have also been studied in great detail. These factors appear to bind within the nucleotide entry channel of RNAP, where they induce nucleolytic cleavage of the 3′ ends of transcripts that have been displaced from the active site through RNAP backtracking (30, 36, 46). Gre factors can also increase elongation efficiency during mRNA synthesis by suppressing elongation arrest, by enhancing transcription fidelity by excising misincorporated nucleotides, and by facilitating the transition from initiation to elongation (4, 5, 14). In E. coli there are two Gre factors, GreA and GreB, which have complementary and overlapping activities, but in other organisms, such as B. subtilis, there is only a single gene encoding a Gre factor (GreA). Despite their apparently important role, it appears that Gre factors are not essential, and synthetic lethal phenotypes are observed only when gre mutants are combined with other mutants, such as mfd (50).

Although much is known about the biochemical functions of transcription elongation factors, relatively little is known about how transcription is spatially organized within the cell. Electron microscopy of E. coli chromosomal spreads has shown that rRNA operons are heavily transcribed compared to neighboring mRNA operons at high growth rates with loads of around one RNAP molecule every 90 nucleotides (17). In previous work using green fluorescent protein (GFP) fusions in B. subtilis, we showed that RNAP was localized at high concentrations in nucleoids in regions termed transcription foci (TF). TF appeared more frequently as the growth rate increased, which coincided with the increased cellular demand for ribosomes under these conditions, and they colocalized with the seven rRNA operons located within about 200 kb of oriC, suggesting that the accumulation of signal within TF largely represented increased transcription of rRNA operons (12, 32). Therefore, we have a crude, but effective, visual assay that allows us to distinguish between mRNA synthesis sites and rRNA synthesis sites within the cell. Using our cytological approaches combined with other assays, we also determined that the ratio of NusA to RNAP was greater in rRNA complexes (2:1) than in mRNA complexes (1:1) (13). In this work, we examined the localization and cellular levels of the additional transcription elongation factors NusG, NusB, and GreA under standardized growth conditions. Our results provide a dynamic illustration of differential elongation complex composition within the cell.

MATERIALS AND METHODS

Bacterial strains and media.

All cloning was carried out in E. coli DH5α (Gibco-BRL) using vectors that have been described previously (15, 31) (Table 1). Transformation of B. subtilis was carried out by the method of Anagnostopoulos and Spizizen (2), as modified by Jenkinson (24). All strains were selected for with appropriate antibiotics; ampicillin (100 μg/ml) was used for E. coli, and chloramphenicol (5 μg/ml) and kanamycin (5 μg/ml) were used for B. subtilis. When appropriate, xylose was used at a concentration of 0.5% (wt/vol). B. subtilis strains were cultured in S and CH media, which have been described previously (44). The stringent response was induced as described by Davies and Lewis (12). To collapse nucleoids, strains were grown to an A₆₀₀ of 0.4 at 37°C in CH medium before chloramphenicol (50 μg/ml) was added. Then the culture was grown with shaking for an additional 30 min at 37°C.

TABLE 1.

Plasmids, strains, and primers used in this study

Strain, plasmid, or primer	Genotype or sequence	Reference or source/construction
E. coli strains
BL21(DE3)(pLysS)	λ DE3 pLysS; F−ompT(lon)hsdSB (rB− mB−)	48
DH5α	F−endA1 hsdR17 supE44 thi-1 λ−recA1 gyrA96 relA1 Δ(lacZYA-argF)U169 φ80dlacZΔ15	Gibco-BRL
B. subtilis strains
168	trpC2	C. Anagnostopoulous
BS61	trpC2 chr::pNG71 (nusB-gfp cat Pxyl-′nusB)	This study; 168 transformed with pNG71
BS67	trpC2 chr::pNG76 (greA-gfp cat Pxyl-greA)	This study; 168 transformed with pNG76
BS126	trpC2 chr::pNG126 (rpoC-yfp kan Pxyl-′rpoC)	13
BS68	trpC2 chr::pNG75 (nusB-cfp cat Pxyl-′nusB), pNG126 (rpoC-yfp kan Pxyl-′rpoC)	This study; BS126 transformed with pNG75
BS128	trpC2 chr::pNG128 (nusG-gfp cat Pxyl-nusG)	This study; 168 transformed with pNG128
PK9C8	trpC thy chrf::hbs-gfp lacI cat Pxyl-His6-hbs	25
Plasmids
pSG1164	bla cat Pxyl-gfp	31
pSG1186	bla cat Pxyl-cfp	15
pNG71	bla cat Pxyl-′nusB-gfp	This study; Acc65I- and XhoI- digested 371-bp 5′ end of ′nusB in Acc65I- and XhoI-cut pSG1164
pNG75	bla cat Pxyl-′nusB-cfp	This study; pNG71 digested with PstI and SpeI to replace gfp with cfp isolated from similarly digested pSG1186
pNG76	bla cat Pxyl-greA-gfp	This study; EcoRI- and XhoI-digested greA in EcoRI- and XhoI-cut pSG1164
pNG126	bla kan Pxyl-′rpoC-yfp	13
pNG128	bla cat Pxyl-nusG-gfp	This study; EcoRI- and XhoI-digested nusG in EcoRI- and XhoI-cut pSG1164
pETMCSIII	bla Pφ10-6HisTϕ	35
pNG127	bla Pφ10-His6-greA Tϕ	This study; EcoRI- and NdeI-digested greA inserted into EcoRI- and NdeI-cut pETMCSIII
Primers
′nusB Forward	5′-GAA AAA GCT GGT ACC GCA CTA TTT-3′
nusB Reverse	5′-GGG TTC CTC GAG TGA TTG TTC AAT-3′
nusG Forward	5′-GTC CTG CTC GAG ATG GAA AAG AAT-3′
nusG Reverse	5′-AGT TTT TTC GAA TTC CAA TTT ATC-3′
greA Forward	5′-GAG TGA CTC GAG ATG GCA CAA GAG-3′
greA Reverse	5′-GCT AAA CAG GAA TTC TGA AAT TTT-3′
nusB Forward check	5′-TAA CGA CAA GCG GAT AAA GAG GAT-3′
nusG Forward check	5′-CAA ATG CGG GGA GGG AAG GAC TGG-3′
greA Forward check	5′-TCG AAG ATG TTG AGG GAC AAC CAG-3′
gfp Reverse check	5′-AAG TCT GGT ACC TTA TTT GTA TAG-3′
Pxyl Forward check	5′-GTA CTC TAG AAA GGA GAT TCC TAG GAT GG-3′

DNA manipulations.

Oligonucleotide primers (Table 1) were used for PCR amplification of B. subtilis chromosomal sequences. DNA manipulations were performed as described by Sambrook et al. (41).

Image acquisition and analysis.

All imaging conditions, techniques, and software used in this study have been described previously (13). Briefly, cells were imaged at the onset of exponential growth (A₆₀₀, approximately 0.3) using a Ziess Axioskop 2 epifluorescence microscope equipped with a Photometrics Quantix cooled charge-coupled camera, and images were analyzed using MetaMorph v 6.0 (Universal Imaging Corp., United States).

GreA overproduction and purification.

An overnight culture of E. coli BL21(DE3)(pLysS) cells transformed with pNG128 (Table 1) was used to inoculate 500 ml of LB containing ampicillin. The resulting culture was then shaken at 37°C until the A₆₀₀ was 0.7. Isopropyl-β-d-thiogalactopyranoside (IPTG) was added to a final concentration of 0.5 mM, and the culture was shaken at 37°C for an additional 3 h. Then the culture was centrifuged at 6,500 rpm for 7 min at 4°C using a GS3 rotor (Sorvall, Australia). The pellet was resuspended in 30 ml lysis buffer (50 mM KH₂PO₄, 300 mM NaCl; pH 8.0), and lysozyme was added to a final concentration of 2 mg/ml. Samples were then incubated at 37°C for 15 min before 30 μl of DNase (10 mg/ml), 30 μl of RNase (10 mg/ml), and 30 μl of Triton X-100 were added, and the preparation was incubated at 37°C for 5 min. Samples were then sonicated, and following centrifugation each supernatant was applied to a Poly-Prep chromatography column (Bio-Rad, Sydney, Australia) containing Ni-CAM HC resin (Sigma). The column was washed with 4 column volumes of lysis buffer containing 20 mM imidazole, and samples were eluted with lysis buffer containing 250 mM imidazole. The imidazole was removed by dialysis with 50 mM KH₂PO₄-300 mM NaCl-50% glycerol (pH 8.0), and aliquots were stored at −80°C. The yield was determined using the Bio-Rad protein assay (Bio-Rad).

Protein quantification.

All of the culture and harvest conditions used for quantitative native polyacrylamide gel electrophoresis in this study were described previously by Davies et al. (13). Briefly, strains were grown to an A₆₀₀ of approximately 0.3, taking care to obtain the exact optical density, and cells from 10-ml samples were pelleted by centrifugation, snap frozen in liquid nitrogen, and stored at −80°C. Cells were lysed using the osmolysate method (42), and samples were electrophoresed on a 10% (wt/vol) native polyacrylamide gel alongside purified GFP, which was used to create a standard curve. The gels were electrophoresed at 40 V for 4 h at 4°C and were scanned using a Typhoon 9410 variable-mode imager (excitation wavelength, 488 nm; Molecular Dynamics). Quantitation was performed using the Image Quant 5.2 software. For quantitative Western blotting, 5-ml overnight cultures grown in CH medium were used to inoculate 50 ml of medium to obtain a final A₆₀₀ of 0.05. The resulting culture was grown at 37°C with shaking until the A₆₀₀ was approximately 0.3, and the cell density was determined carefully. One-milliliter aliquots were pelleted by centrifugation and stored at −80°C. Pellets were resuspended to an A₆₀₀ of 10 in 200 mM Tris-Hcl (pH 7.5)-5 mM EDTA-100 mM NaCl (TES) supplemented with 2 mg/ml lysozyme and incubated for 15 min at 37°C. To fragment the DNA, samples were sonicated, and carefully measured aliquots of cell extract were loaded onto a 15% sodium dodecyl sulfate-polyacrylamide gel. Purified GreA samples were also loaded in the remaining lanes on the same gel to create a standard curve. Transfer onto nitrocellulose membranes (Amersham) was performed overnight at 30 V. To detect GreA, rabbit polyclonal anti-GreA antibody and goat anti-rabbit antibody (Bio-Rad) were used, and they were isualized with the Opti4CN system (Bio-Rad). Quantitation of bands was performed using the Image Quant 5.2 software.

RESULTS

Construction of NusB-GFP-, NusG-GFP-, and GreA-GFP-labeled strains.

Fusions of gfp to the 3′ termini of nusB, nusG, and greA were created (Table 1) and transformed into B. subtilis, resulting in strains BS61, BS128, and BS67 (Table 1). In each strain the fusion was driven by its wild-type promoter, as shown in Fig. 1, which allowed the subcellular localization of each of the transcription factors to be visualized in live cells at physiological expression levels. Importantly, a xylose-inducible promoter was present in these constructs to drive downstream genes when the fusion integrated in an operon (31). As Fig. 1 shows, while the wild-type copy of nusB was functionally inactivated by insertion of the gfp fusion in BS61, a xylose-inducible promoter was able to drive expression of full-length nusG and greA in BS128 and BS67, respectively. Correct insertion of these constructs into the chromosome was confirmed by PCR analysis (not shown). In E. coli, loss of NusB results in transcription polarity and in an inability to form antitermination complexes, which results in a prolonged doubling time and cold sensitivity (37, 45, 49), while in B. subtilis, loss of NusG results in a reduced growth rate (23). No phenotypic effects of greA disruption have been observed in wild-type E. coli, but greA is required, along with other factors, such as Mfd, to resolve stalled RNAP complexes that can block the progress of DNA replication forks (50). As shown in Table 2, each of the labeled strains exhibited growth kinetics similar to those of wild-type strain 168 in both poor medium (S medium) and defined rich medium (CH medium). Thus, the fact that all the gfp fusion strains grew at the same rate as the wild-type strain suggests that functional fusions were created.

FIG. 1.

Chromosomal organization of BS61, BS128 and BS67. In each strain the full-length gfp gene fusion is under the control of the wild-type promoter. In BS61 the xylose-inducible promoter drives the expression of a truncated nusB gene (371 bp of the 3′ end) and folD. In BS128 the xylose-inducible promoter drives the expression of full-length nusG along with ribosomal genes rplK and rplA. As greA is not in an operon, the xylose-inducible promoter drives only the expression of the full-length wild-type greA gene. cat encodes chloramphenicol acetyltransferase.

TABLE 2.

Doubling times of strains 168, BS61, BS128, and BS67 under both slow- and medium-growth conditions in the presence and absence of xylose

Strain	Doubling time (min)
	Slow growth		Medium growth
	Without xylose	With xylose	Without xylose	With xylose
168a	107	NDc	47	ND
B61b	105	105	50	42
BS128b	105	110	48	47
BS67b	110	105	42	39

^aThe values are averages for triplicate determinations.

^bThe values are averages for duplicate determinations.

^cND, not determined.

Although these strains grew with kinetics identical to those of the wild-type strain in both media, it was surprising that the growth of BS61 (NusB-GFP) and BS128 (NusG-GFP) was xylose independent (Table 2), as both genes are located in operons (28) (Fig. 1). The xylose promoter is tightly controlled (23, 34) (see Fig. 6I), and it is unlikely that leaky expression was sufficient to explain the xylose independence. It has been reported previously that the highly expressed rplK and rplA genes encoding ribosomal proteins L11 and L1, respectively, are expressed independent of the nusG promoter (23), suggesting that there are unmapped promoters within or downstream of nusG. Loss of L11 in B. subtilis results in susceptibility to environmental stress and reduces the growth rate threefold (56). In BS128 grown in the absence of xylose no phenotype was associated with loss of L11. In BS61 the little-studied folD gene, encoding a predicted bifunctional protein that acts as both a methylenetetrahydrofolate dehydrogenase and a methenyltetrahydrofolate cyclohydrolase, which is involved in the metabolism of coenzymes and prosthetic groups, is downstream of nusB, as shown in Fig. 1. According to the functional database JAFAN (http://bacillus.genome.jp), folD has been found to be essential for growth of B. subtilis, and therefore we expected this strain to be xylose dependent if the nusB promoter is the only promoter in the operon. To investigate the possibility that previously unidentified promoters are present in the operons, 200 bp immediately upstream of the rplK and folD translation start sites was analyzed using a bacterial promoter prediction program (NNPP2.2, available at http://www.fruitfly.org/seq_tools/promoter.html), which identifies potential promoters based on conserved sequences. Several sequences satisfying the parameters of this program were identified in both of these regions, possibly accounting for the xylose independence of BS67 and BS128, although further biochemical analyses are needed to identify bona fide promoters.

FIG. 6.

Effect of a high concentration of chloramphenicol on nucleoid morphology and GFP fusion localization. (A to C) DAPI signals of HBsu-labeled (PK9C8), RNAP-labeled (BS1048), and GreA-labeled (BS67) strains, respectively, after they were exposed to a high concentration of chloramphenicol (50 μg/ml). In each case the nucleoid has clearly collapsed into a compact structure at the center of the cell. (D to F) GFP signals showing HBsu (D), RNAP (E), and GreA (F) localization following the collapse. (G) DAPI signal for BS67 cultured with 0.5% (wt/vol) xylose and exposed to chloramphenicol. (H) GFP image corresponding to panel G. (I) Western blot obtained using an anti-GreA polyclonal antibody for BS67 cultured in the presence and absence of 0.5% (wt/vol) xylose. In both lanes there is a 45-kDa band corresponding to the GreA-GFP fusion, while the 17-kDa band corresponding to wild-type GreA was present only when cells were cultured in the presence of xylose.

Subcellular localization of NusB, NusG, and GreA.

The subcellular localization of GFP-tagged RNAP in B. subtilis has been characterized in detail, and RNAP has been found to localize primarily in the nucleoid region (13, 32). Subnucleoid regions with high GFP intensity, termed transcription foci (Fig. 2H), were observed during mid-exponential growth of this organism at medium and high growth rates (32), and these TF were subsequently shown to represent sites of rRNA synthesis (12). In a small subset of cells, TF also formed in S medium (Table 3 and Fig. 2G), reflecting down-regulation of rRNA synthesis under these conditions.

FIG. 2.

Subcellular localization of NusG, NusB, GreA, RNAP, and HBsu. (A to C) Localization patterns for NusG, NusB, and GreA in S medium, respectively. (D to F) Localization patterns for NusG, NusB, and GreA in CH medium, respectively. For comparison the localization patterns for RNAP (32) in S medium (G) and in CH medium (H) and the localization pattern for HBsu (25) in CH medium (I) are also shown. The arrows indicate transcription foci. Scale bar = 5 μm.

TABLE 3.

Summary of TF quantitation data for RNAP, NusA, NusB, NusG, and GreA

Molecule	S medium		CH medium
	% of cells with foci	No. of foci per cell	% of cells with foci	No. of foci per cell
RNAPa	7.5	NDc	88.5	3.7
NusAa	66.9	ND	95.6	3.7
NusBb	91.6	1.91 ± 0.37	100.0	3.79 ± 0.61
NusGb	17.0	1.62 ± 0.56	91.7	3.59 ± 0.76
GreAb	0	0	0	0

^aData obtained from reference 13.

^bThe values are averages determined by using no fewer than 350 cells.

^cND, not determined.

The subcellular localization patterns of the transcription factors NusB, NusG, and GreA were determined in both S and CH media using fluorescence microscopy and are shown in Fig. 2. NusG localized in a pattern similar to that seen for RNAP cultured in CH medium (Fig. 2D and H), and 91.7% of the cells under these conditions had TF (Table 3). NusG also formed TF in S medium in 17% of the cells (Fig. 2A), which was more than the percentage observed for RNAP (Table 3), but the vast majority of the cells still lacked TF. These results suggest that NusG is closely associated with RNAP and complement biochemical data which indicate that NusG is involved in the regulation of both mRNA and antitermination transcription complexes (20, 37, 49).

Although NusB was recruited to TF, the localization patterns of the NusB fusion differed markedly from the localization patterns of RNAP and NusG. The majority of NusB was recruited to TF at low and medium growth rates, as shown in Fig. 2B and E, and relatively little signal was dispersed through the nucleoid. Furthermore, this pattern of recruitment to TF did not appear to be growth rate dependent, as observed with RNAP, with NusG, and, to a lesser extent, with NusA (13), and 91.6% of the cells contained TF under slow-growth conditions, as shown in Table 3. A dual-labeled NusB-cyan fluorescent protein (CFP)-RNAP-yellow fluorescent protein (YFP) strain (BS68) (Table 1) was also created and cultured in CH medium to determine if TF formed for NusB coincided with RNAP. Figure 3A shows the RNAP-YFP localization pattern, Fig. 3B shows the NusB-CFP localization pattern, Fig. 3C shows an overlay of the two signals, and Fig. 3D is a line scan taken through Fig. 3C. From the line scan it is clear that although the peaks for NusB are more pronounced than those for RNAP TF, the signal peaks colocalize, suggesting that NusB localizes principally at sites of rRNA synthesis within the cell, which is consistent with previously published biochemical data (3, 37, 49).

FIG. 3.

TF from RNAP (green) and NusB (red) colocalize. (A) Localization pattern for RNAP-YFP. (B) Localization pattern for NusB-CFP. (C) Overlay. The arrows indicate the same TF in panels A to C. (D) Line scan through the length of the cells, showing that the TF formed in RNAP- and NusB-labeled strains colocalize (green line, RNAP; red line, NusB). Scale bar = 5 μm. AU, arbitrary units.

In contrast to RNAP and the Nus factors, GreA appeared to localize uniformly throughout the nucleoid and did not appear to be recruited to TF, as shown in Fig. 2C and F; of all the images analyzed, none showed visible TF (Table 3). The lack of TF in the GreA-GFP strain suggests that this elongation factor plays little, if any, role in rRNA transcription and is principally involved in the regulation of mRNA synthesis through its role in restarting paused elongation complexes (5, 16). The nucleoid association of the labeled factors was confirmed by costaining cells with DAPI (4′,6′-diamidino-2-phenylindole), which revealed GFP-DAPI signal colocalization (data not shown). In addition, the histone-like protein HBsu shown in Fig. 2I has previously been shown to colocalize with the nucleoid (25), and although the GreA-GFP fusion was much brighter, the localization pattern was very similar. From the cytological data it is evident that the NusG, NusB, and GreA fusions are closely associated with the nucleoid and presumably with RNAP despite their strikingly different localization patterns.

TF formed in BS61 and BS128 represent sites of rRNA synthesis.

To confirm that the TF formed during exponential growth of BS61 and BS128 were representative of rRNA synthesis, the stringent response was induced (see Materials and Methods). The stringent response is a starvation response, which occurs as a result of high levels of uncharged tRNA molecules within the cell, resulting in RelA-induced production of the effector molecule ppGpp, which destabilizes the RNAP-promoter complex on ribosomal operons (10). TF have previously been shown to include rRNA operons within 200 kb of the replication origin (12), and inducing the stringent response has previously been shown to cause RNAP and NusA TF to disappear in B. subtilis (13, 32). In addition, in E. coli, RNAP TF have been shown to remain after induction of the stringent response in a relA mutant, suggesting that the disappearance of TF is a direct result of ppGpp destabilization of the RNAP-promoter complex on rRNA operons (9). Figure 4A and C show BS61 and BS128 imaged before induction of the stringent response, and the positions of TF are indicated. Figure 4B and D show BS61 and BS128, respectively, imaged 30 min after induction of the stringent response. Consistent with results obtained for GFP-tagged RNAP and NusA (13, 32), a dramatic reduction in visible TF was observed, which coincided with the down-regulation of rRNA operon promoters (12).

FIG. 4.

Effect of the stringent response on TF formed by NusB and NusG. (A and C) Localization patterns for NusB and NusG before induction of the stringent response. The arrows indicate TF. (B) Localization pattern for NusB after induction of the stringent response, showing that TF largely disappeared. (D) Localization pattern for NusG after the stringent response, showing that there were no visible TF.

Quantification of NusB, NusG, and GreA.

Using image analysis approaches that were used in other studies (13), the relative amounts of NusB and NusG in mRNA and antitermination transcription complexes were determined by quantifying determining the percentages of fluorescence in TF in exponentially growing BS61 and BS128 cells cultured in CH medium. By analyzing at least 200 cells, it was found that 53.1% ± 6% of cellular NusB and 27.4% ± 6% of NusG localized to TF, as shown in Table 4. This compares with 27.4% ± 4% for RNAP and 36.8% ± 5% for NusA when the values were calculated using the same method (13) (Table 4).

TABLE 4.

Summary of RNAP, NusA, NusB, NusG, and GreA quantitation data

Parameter	RNAPa	NusAa	NusBb	NusGb	GreAb
No. of molecules/cell	7,500 ± 1,000	8,900 ± 1,500	3,400 ± 700	6,050 ± 850	13,800 ± 1,300
Ratio	1.0	1.2	0.4	0.8	1.8
No. of molecules/genome	2,590	3,070	1,170	2,080	4,790
No. of molecules/nucleoid	4,060	4,810	1,820	3,260	7,490
% of signal in TF	27.4 ± 4	36.8 ± 5	53.1 ± 6	27.4 ± 6	NDc
No. of molecules in TF	1,110	1,770	1,060	890	ND
Ratio	1.0	1.6	1.0	0.8	ND
No. of molecules outside TF	2,950	3,040	760	2,370	ND
Ratio	1.0	1.0	0.3	0.8	ND

^aData obtained from reference 13.

^bThe values are averages from at least four experiments.

^cND, not determined.

The total numbers of NusB, NusG, and GreA molecules per cell were also determined using nondenaturing gel electrophoresis (native polyacrylamide gel electrophoresis) (13). As each of the gene fusions was under the control of its wild-type promoter, the GFP levels should have represented the physiological levels of the transcription factors. Our results indicated that there were 3,400 ± 700 molecules of NusB per cell, 6,050 ± 850 molecules of NusG per cell, and 13,800 ± 1,300 molecules of GreA per cell (Table 4).

From the image analysis and quantitation data, the ratio of transcription factors to RNAP within TF was determined (Table 4). Table 4 shows that the RNAP/NusA/NusB/NusG ratio in TF was 1:1.6:1:0.8. An RNAP/NusA/NusG/NusB ratio of 1:1:0.8:0.3 was obtained for mRNA transcription elongation complexes.

Surprisingly, the GreA levels were found to be nearly twice the RNAP levels. In E. coli the analogous GreB factor has been shown to bind to the nucleotide entry channel of RNAP at a 1:1 ratio (36). One artifact of a partially functional fusion can be overproduction of the fusion protein to compensate for any reduced activity. For this reason, quantitative Western blotting using an anti-GreA antibody was used to determine the levels of GreA in exponentially growing wild-type strain 168 (see Materials and Methods). Different levels of purified GreA were used to create a standard curve when samples were run alongside known amounts of cell extract (Fig. 5). The results of these experiments showed that the level of GreA was about 15,800 ± 2,550 molecules per cell, which correlated well with the fluorescence data, suggesting that the GreA-GFP fusion is not overproduced and that GreA is present at substantially higher levels than RNAP within the cell.

FIG. 5.

Quantitative Western blotting of GreA from wild-type strain 168. (A) Western blot obtained using anti-GreA polyclonal antibody. Lanes 1 to 5 contained 0.075, 0.1875, 0.375, 0.5626, and 0.75 pmol of purified GreA, respectively, while lanes 6 to 9 contained known volumes of strain 168 cell lysates. (B) Standard curve generated from the Western blot used to calculate GreA levels. AU, arbitrary units.

GreA-GFP is largely associated with RNAP.

Further cytological analysis was performed to investigate whether the apparent excess GreA is due to involvement of this molecule in other cellular processes that are not related to transcription. In addition to inhibiting growth at a translational level, exposing cultures to a high level of chloramphenicol (50 μg/ml) collapses nucleoids into compact structures by reducing the contact between the nucleoid and the cell envelope (53). As GreA is not a known DNA-binding protein, only GreA associated with RNAP would be expected to collapse with the nucleoid. The DNA-binding proteins HBsu and RNAP were used as controls.

Histone-like protein HBsu plays a role in DNA packaging and has been shown to colocalize with the nucleoid of B. subtilis in vivo (25) (Fig. 2I). As HBsu is a DNA-binding protein, it seems reasonable to hypothesize that if the nucleoid collapsed, HBsu would collapse with it. As shown in Fig. 6A to C, the DAPI signals in strains PK9C8 (HBsu-GFP), BS1048 (RNAP-GFP), and BS67 (GreA-GFP), respectively, clearly showed that there was nucleoid compaction. The corresponding GFP images in Fig. 6D to F also show signals that largely coincide with the region occupied by the DAPI stain. In the case of HBsu and RNAP, this is indicative of their close association with the DNA. As GreA has no known DNA-binding properties, the redistribution of this molecule is thought to be the result of the close association of GreA with RNAP.

We also wanted to investigate whether increasing the GreA levels within the cell affected the GreA-GFP localization pattern. As shown in Fig. 1, in BS67 the wild-type greA gene is downstream of the greA-gfp fusion and is under tight control of the xylose-inducible promoter. Thus, strain BS67 was grown in the presence of 0.5% (wt/vol) xylose to determine whether GreA localization could be due to a nonspecific interaction with RNAP or DNA. The DAPI image of the cells shows that there was a compact signal (Fig. 6G), similar to that observed for BS67 grown without xylose (Fig. 6C). However, the corresponding GFP image in Fig. 6H indicates that a significant portion of the GreA was not closely associated with either RNAP or DNA. This suggests that wild-type GreA and the GreA-GFP fusion compete for a limited number of binding sites on RNAP and that excess GreA (and GreA-GFP) is free to diffuse throughout the cell. Figure 6I shows a Western blot for BS67 in the absence and presence of xylose. Under both conditions a 45-kDa band corresponding to the GreA-GFP fusion was present, but only when cells were grown in the presence of xylose was the 17-kDa band corresponding to wild-type GreA observed, confirming that there is indeed tight control over gene expression in the absence of xylose from the P_xyl promoter. Therefore, it appears that the large amount of GreA naturally present in cells has a specific role that presumably involves binding to RNAP since when excess GreA is provided, the GFP fusion can be displaced, resulting in nonspecific whole-cell fluorescence.

DISCUSSION

Subcellular localization and quantitation of NusB and NusG.

In this study we analyzed in detail the subcellular localization of NusB and NusG, and we also determined the cellular levels of these molecules and related the data obtained to the distribution of RNAP grown under identical conditions. NusG was found to be distributed in a manner very similar to that of RNAP, reflecting its global role in regulating the activity of all classes of transcription ECs. The cellular level of NusG was also similar to the cellular level of RNAP, suggesting that these proteins interact at a 1:1 ratio. We did notice that slightly more TF were formed at low growth rates for NusG than for RNAP (17% and 7.5%, respectively) (Table 3), and the significance of the small difference is not clear. However, since there is slightly less NusG than RNAP in each cell, the difference may simply reflect a small preference in recruitment of NusG to RNAP ECs involved in rRNA synthesis.

The localization pattern for NusB was strikingly different from the localization pattern observed for RNAP, NusA, or NusG (13, 32) (Fig. 2). NusB appeared to be highly concentrated in foci, which colocalized both with the DNA (data not shown) and with RNAP TF (Fig. 3), suggesting that this factor is involved only in the regulation of rRNA synthesis. This is in good agreement with previous biochemical data for the role of NusB in antitermination complexes (19, 37, 49).

Interestingly, careful examination of micrographs of NusB-GFP fluorescence suggested that this fluorescence provides very high spatial resolution data for the location of rRNA synthesis in the cell. We have shown in previous work that TF represent the sites of the majority, but not all, of the rRNA synthesis in the cell and that TF represent transcription from the seven most origin-proximal rRNA operons in the B. subtilis genome (12). Three additional rRNA operons are located up to almost 1 Mb from the origin. Figure 4A shows a bright focus of NusB-GFP fluorescence that represents a TF comprising the seven most origin-proximal rRNA operons, and there are two smaller, fainter foci adjacent to this area. We suggest that these additional smaller foci could represent the locations of rRNA synthesis of some of the other more origin-distal rRNA operons. Therefore, localization of NusB-GFP fluorescence provides very high resolution information on the spatial location of rRNA synthesis within the cell. Since NusB is known to heterodimerize with NusE (ribosomal protein S10), we assume that, with respect to transcription, NusE localizes in exactly the same pattern as NusB.

Although TF for NusB are extremely bright, only 53% of cellular NusB was found to localize within TF, and clear, but low-level, background fluorescence was seen in the cells (Fig. 2E). There are a number of reasons why such a relatively low level of fluorescence was obtained for NusB in TF, even though this molecule is thought to be principally involved in regulation of rRNA synthesis. First, the background fluorescence in the cell could have been due to autofluorescence and not a signal from the NusB-GFP fusion. Since care was taken to subtract backgrounds from images, we believe that autofluorescence did not contribute much, if any, to the background signal. Second, there could have been fluorescence quenching in the TF that reduced the apparent level of NusB in TF. Such an effect has been observed in the prespore chromosomes of sporulating bacilli stained with the DNA-specific dye DAPI (21, 43). However, we believe that quenching is an unlikely explanation, as it requires large numbers of fluorescent molecules to be stacked close to each other, as would be the case for a dye such as DAPI but is unlikely for the 1,200 or so molecules of NusB-GFP present in transcription elongation complexes in TF. Third, our quantitation is correct, and only 53% of NusB was recruited to TF. The remainder may have been involved in translation (see below) or transcription of the three rRNA operons not present in TF or represented NusB that was being recycled for a new round of rRNA transcription. Indeed, recent work using an E. coli system showed that NusB binding to boxA is the first step in antitermination complex formation, and dissociation of NusB from the released, mature transcript is probably one of the last steps (19). Therefore, excess NusB would be necessary for efficient formation of antitermination complexes and hence the overall growth rate.

It has been proposed that both NusB and NusG also have roles in regulating the efficiency of translation, at least in E. coli (for a review see reference 47). Certain NusB mutants are able to suppress temperature-sensitive secretion mutations, and this effect has been linked to a reduction in translation efficiency (38). In addition, depletion of NusG was found to reduce the rate of peptide elongation but not formation of lacZ mRNA, linking NusG directly to translation efficiency (55). However, our data suggest that, at least in B. subtilis, it is unlikely that either NusB or NusG has a significant role in translation. First, ribosomes have been shown to localize in nucleoid-free regions, which in dividing cells correspond mainly to the cell poles in B. subtilis (32), placing these factors in domains that are spatially separate from the ribosomes. Second, the cellular quantities of NusB and NusG do not appear to allow much flexibility for roles outside transcription, certainly not in a process involving molecules that are as abundant as ribosomes. Both the cytological and quantitative data presented here suggest that both NusB and NusG are restricted to transcription.

GreA is very abundant and is closely associated with RNAP.

Our results for the subcellular localization of GreA suggested that while this molecule is closely associated with RNAP, there is little evidence which suggests that it plays a significant role in the regulation of rRNA ECs and is restricted to mRNA ECs. Since one of the principal activities of Gre proteins is to rescue stalled RNAP, it is perhaps not surprising that we did not observe GreA in TF that contain ECs specifically modified to resist stalling. We also determined that the cellular level of GreA was approximately twice the cellular level of RNAP. Similar results have been reported for Gre proteins in E. coli, in which levels of 2 to 6, 0.5 to 2, and 2 to 3 μM have been determined for GreA, GreB, and RNAP, respectively (27). However, while there are now data for both E. coli and B. subtilis which show that the molar ratios of the Gre protein to RNAP are approximately 2:1, structural models of GreB and RNAP suggest that only one molecule of GreB is bound in the nucleotide entry channel (36). Furthermore, our data on nucleoid collapse and GreA overproduction suggest that the vast majority of GreA is closely associated with RNAP. How can these conflicting data be resolved?

There have been at least two studies which have suggested alternative binding sites for Gre proteins on RNAP in E. coli, although the proposed sites are different (26, 51). Using targeted protein footprinting, Traviglia et al. (51) identified three unique binding sites for GreA on RNAP. The first site was in a region that placed GreA near the nucleotide entry channel, which is consistent with existing structural data (36), while the other two sites placed GreA on the underside of the catalytic core on the β subunit. It seems unlikely that binding sites in this region would allow GreA to perform its antiarrest and restart functions, and these sites may have represented nonspecific binding of GreA to RNAP in this assay. Conversely, using competition binding assays, Koulich et al. (26) found that the C-terminal globular domain of GreA occupies a single binding site on RNAP, while the N-terminal coiled-coil domain (NTD) occupies one of two sites on RNAP, a proximal site and a distal site with respect to the RNA in the active site. When it was bound to the proximal site, the NTD was within 2 Å of the RNA backbone, while when it was bound to the distal site, it was 4 to 12 Å from the RNA. Both of these sites place Gre proteins around the nucleotide entry channel of RNAP. Interestingly, in the same study it was found that full-length GreA was essential for antiarrest, while the NTD or an alkaline pH was found to be sufficient for inducing RNA cleavage. Furthermore, the GreA affinity for the 3′ terminus of RNA was increased when the NTD was added separately. This suggests that there are two GreA molecules, one of which is bound by the C-terminal globular domain with the NTD at the proximal site close to, and possibly interacting with, the 3′ terminus of RNA, while the other, bound by the NTD at the distal site, stabilizes the GreA-RNAP-RNA complex, giving it antiarrest and RNA cleavage properties. It is also interesting that the targeted protein footprinting method used by Traviglia et al. (51) identified only regions within 20 Å and was not able to identify the second binding site proposed by Koulich et al. (26), since it was within 12 Å of the first site.

Could both sets of results be correct? Let us speculate. Table 3 shows that approximately 3,000 molecules of RNAP per cell are involved in mRNA transcription. Assuming that none, or very little, GreA is involved in rRNA synthesis, the GreA/RNAP ratio is approximately 4:1 in mRNA ECs, which is consistent with the presence of two binding sites on the underside of the β subunit (51) and the presence of two closely juxtaposed sites around the nucleotide entry channel (26). This ratio may even be consistent with the helical crystal GreB-RNAP data of Opalka et al. (36), as the mass of a second GreB molecule near the nucleotide entry channel may not have been resolved and the method of assembling complexes may have precluded GreB interaction with the underside of the β subunit. The physiological significance of this is unclear, as one Gre molecule that is bound within the nucleotide entry channel accounts well for most, if not all, of the known activities of Gre proteins. Nevertheless, it is clear that further studies are required to clearly determine whether there is more than one specific binding site for GreA on RNAP.

In summary, using cytological and quantitative analyses, we can begin to imagine the subcellular spatial and molecular composition of transcription elongation complexes. Transcription foci representing the sites of rRNA synthesis contain transcription elongation complexes comprising RNAP, NusB, NusG, two molecules of NusA, and, we assume, NusE. mRNA transcription elongation complexes located throughout the bulk of the nucleoid contain RNAP, NusG, one molecule of NusA, and up to four molecules of GreA. Clearly, in addition to being tightly regulated at the molecular level, transcription is highly organized spatially within the bacterial cell.