The distribution of RNA polymerase in Escherichia coli is dynamic and sensitive to environmental cues


  • Julio E. Cabrera,

    1. Laboratory of Molecular Biology, National Cancer Institute, 9000 Rockville Pike, Bethesda, MD 20892, USA.
    Search for more papers by this author
  • Ding J. Jin

    Corresponding author
    1. Laboratory of Molecular Biology, National Cancer Institute, 9000 Rockville Pike, Bethesda, MD 20892, USA.
    Search for more papers by this author
    • Present address: Gene Regulation and Chromosome Biology Laboratory, NCL-Frederick, Frederick, MD 21702, USA.


Despite extensive genetic, biochemical and structural studies on Escherichia coli RNA polymerase (RNAP), little is known about its location and distribution in response to environmental changes. To visualize the RNAP by fluorescence microscopy in E. coli under different physiological conditions, we constructed a functional rpoC–gfp gene fusion on the chromosome. We show that, although RNAP is located in the nucleoid and at its periphery, the distribution of RNAP is dynamic and dramatically influenced by cell growth conditions, nutrient starvation and overall transcription activity inside the cell. Moreover, mutational analysis suggests that the stable RNA synthesis plays an important role in nucleoid condensation.


Escherichia coli RNA polymerase (RNAP) is a multisubunit enzyme that exists in two forms: core (α2ββ′) and holoenzyme (α2ββ’σ), with the latter having an additional sigma polypeptide involved in promoter discrimination (Burgess et al., 1987). Unlike eukaryotes, in which three different RNAPs (Pol I, Pol II and Pol III) (Sentenac, 1985) synthesize different RNA species (rRNA, mRNA and tRNA/5S rRNA respectively), a single RNAP synthesizes all RNA types in E. coli. The E. coli RNAP containing the most abundant sigma factor σ70 is responsible for the expression of the majority of the genes during exponential growth.

Escherichia coli RNAP synthesizes two classes of RNA during exponential growth: unstable mRNA and stable RNA consisting of rRNA and tRNA (Jinks-Robertson and Nomura, 1987). Approximately 85% of stable RNA is rRNA encoded by seven nearly identical operons, and transcription from these promoters is sensitive to growth and/or nutrient conditions of the medium (Nomura et al., 1984; Bremer and Dennis, 1996). For example, it is estimated that, under optimal growth conditions, about 80% of the active RNAP molecules synthesize rRNA and tRNA, which are encoded by genes that represent <1% of the genome, resulting in over 85% of total cellular RNA being the stable RNA (Bremer and Dennis, 1996). Consistent with this, direct electron microscopic visualization of chromatin spreads from rapidly growing cells has shown that elongating RNAP molecules are densely packed on the stable RNA operons, whereas transcription from the remainder of the genome is dramatically less active with only one RNAP molecule per 10–20 kb (French and Miller, 1989).

In contrast, when cells are growing in nutrient-poor (but not starvation) media, fewer RNAP molecules synthesize stable RNA, and cells have reduced growth rates. Moreover, when cells are shifted from nutrient-rich to amino acid starvation conditions leading to the stringent response (Cashel et al., 1996), the cellular transcription machinery is reprogrammed. Although the synthesis of stable RNA and other stringent genes is inhibited, the expression of some mRNA genes, such as amino acid biosynthetic operons, is activated during the stringent response. The stringent response depends on the relA gene, which encodes a guanosine 3′,5′ bis-diphosphate (ppGpp, also known as the ‘magic spot’) synthetase (Cashel and Gallant, 1969; Cochran and Byrne, 1974). In relA mutants, ppGpp does not accumulate, nor is stable RNA synthesis inhibited during amino acid starvation (Metzger et al., 1989). On the other hand, some rifampicin-resistant rpoB mutants exhibit the stringent response phenotype even when they are grown in rich media in the absence of amino acid starvation: they have a reduced transcription rate from stable RNA promoters and an increased transcription rate from other mRNA promoters that are positively regulated during the stringent response (Zhou and Jin, 1997; 1998). Thus, these mutant RNAPs have been named stringent RNAPs.

Despite extensive genetic and biochemical studies, very little is known about whether and how the distribution of RNAP inside the cell responds to environmental cues, such as nutrient richness (rich versus poor media), nutrient starvation or inhibition of transcription. This limitation is largely a result of the small size of an E.  coli cell (≈ 2–4 µm). Inside E. coli cells, the chromosomal DNA is confined to an irregularly shaped structure named the nucleoid (Hobot et al., 1985; Nanninga and Woldringh, 1985; Drlica, 1987; Robinow and Kellenberger, 1994; Pettijohn, 1996). The relative compactness of the nucleoid depends on many factors including transcription and translation (Woldringh et al., 1995; Nordstrom and Dasgupta, 2001; Zimmerman, 2003). However, it is not known whether changes in the ratio between stable RNA and mRNA synthesis affect the condensation of the nucleoid. Elegant work using chromatin-spreading techniques provided direct visualization of different transcription units by electron microscopy (French and Miller, 1989); however, E. coli subcellular structures are not preserved by this method. Recently, it was reported that E. coli RNAP is evenly distributed within the nucleoid using immunofluorescence with an antibody against the α subunit of RNAP (Azam et al., 2000).

The green fluorescent protein (GFP) from marine jellyfish provides a unique tool for the study of bacterial cell biology (Gordon et al., 1997; Lemon and Grossman, 2000; Margolin, 2000; Phillips, 2001; Li et al., 2002; Southward and Surette, 2002). In this work, we studied the subcellular localization of RNAP in E. coli by visualization of RNAP–GFP, expressed from a single chromosomal locus. In particular, we studied the effects of growth conditions (nutrient-rich versus nutrient-poor medium), nutrient starvation and transcription on the distribution of RNAP in wild-type and different mutant strains. Our results show that there are transcription foci in rapidly growing cells and that these foci are sensitive to environmental cues. Moreover, our study suggests that stable RNA synthesis is implicated in the condensation of the E. coli nucleoid.


Construction of a functional rpoC–gfp fusion on the E. coli chromosome

We sought to tag one of the endogenous RNAP subunits with GFP in order to follow the subcellular localization of the enzyme under various conditions. Because RNAP is essential for cell viability, it is critical to fuse GFP to an appropriate position in RNAP without affecting its function. We chose to fuse GFP to the C-terminus of the β′ subunit of RNAP encoded by rpoC because it has been shown that E. coli cells containing an rpoC–lacZ fusion can grow normally (Rowland and Glass, 1995). Indeed, we found that a plasmid containing the rpoC–gfp fusion was able to complement a cold-sensitive rpoC mutant strain (data not shown), demonstrating that the RNAP–GFP is functional. Thus, we constructed strain DJ2599 (MG1655 rpoC–gfp), in which the chromosomal rpoC gene was replaced by an rpoC–gfp fusion by linear DNA recombination in vivo (Yu et al., 2000), allowing the β′-GFP fusion protein to be expressed from the native promoter of the rpoBC operon (a detailed description is given in Experimental procedures). Western blot analysis with anti-GFP or anti-β′ antibodies showed a full-length fusion protein in total protein extracts from DJ2599 cells (data not shown).

The growth of DJ2599 was only slightly slower (10–15%) than the parental strain MG1655 below 37°C, although the plating efficiency of DJ2599 was reduced about 100-fold compared with MG1655 above 37°C. It is common for GFP-tagged proteins to be temperature sensitive (Siemering et al., 1996; Gordon et al., 1997; Levin et al., 1999), presumably because GFP misfolding at higher temperatures interferes with the overall structure of the fusion proteins. Indeed, purified RNAP–GFP from DJ2599 cells grown at 30°C was as active as wild-type RNAP in various in vitro transcription assays performed from 25°C to 42°C, and both enzymes behaved essentially the same in the transcription of stringent and non-stringent promoters (data not shown). In vivo, transcription from the stringent ribosomal promoter rrnB P1 and the non-stringent promoter lacUV5 was comparable in both MG1655 and DJ2599 strains at 30°C (Table 1). Thus, our results demonstrate that the RNAP–GFP-containing strain behaves similarly to wild type and, therefore, the fusion should faithfully reflect the distribution of wild-type RNAP inside the cell.

Table 1. Transcription from the rrnB P1 and lacUV5 promoters is comparable in the wild-type and rpoC–gfp strains.
  rrnB P1–lacZ lacUV5–lacZ
  1. The β-galactosidase activities expressed from the rrnB P1 and the lacUV5 promoters in the MG1655 strain and its rpoC–gfp derivative were determined from cultures grown to an OD of ≈ 0.4 in LB at 30°C as described in Experimental procedures. The values are the averages of three independent assays.

MG16551027 ± 4074 ± 2
MG1655 rpoC–gfp1048 ± 3569 ± 2

Challenges of visualizing RNAP in E. coli

We carried out fluorescence microscopy of the cells carrying the rpoC–gfp fusion gene. Unlike other reported bacterial proteins tagged with GFP, which are generally insensitive to growth conditions and form only one or two focal points per cell, visualization of RNAP was more problematic. Cells embedded in LB-agarose and mounted on microscope slides with coverslips showed dim GFP signals covering the whole nucleoid (data not shown), similar to the observations reported previously in Bacillus subtilis (Lewis et al., 2000). Fortuitously, when we examined cells adsorbed to a flow cell into which fresh LB broth was constantly pumped, we found that the GFP signal was stronger and that the distribution of RNAP was significantly different from that observed on microscope slides (data not shown). These observations suggested that RNAP distribution is extremely sensitive to growth conditions as cells under a coverslip are very likely in a different physiological state compared with those in a flow cell, presumably because of mainly reduced oxygen availability. However, adsorption of E. coli to flow cells was poor, and the cells remained motile, complicating the imaging process. Thus, we used a modified sampling procedure in which the cells were fixed rapidly with formaldehyde immediately after they were removed from batch cultures to ‘freeze’ RNAP and subcellular structures before microscopic examination. Importantly, formaldehyde does not affect the GFP chromophore (Chalfie et al., 1994; Ward, 1998). This procedure enabled us to obtain reproducible results in visualizing the RNAP enzyme inside E. coli cells under different physiological conditions.

RNAP is concentrated in transcription foci in rapidly growing cells

We examined the effect of cell growth conditions on the distribution of RNAP in E. coli(Fig. 1). The DJ2599 strain grown at 30°C on nutrient-rich medium (LB broth) and on nutrient-poor medium (minimal + glucose) had a doubling time of ≈ 45 min and 170 min respectively. The cells with different growth rates were examined for their shape (Fig. 1A and B, DIC), RNAP–GFP fluorescence (Fig. 1A and B, β′-GFP) and nucleoid position (Fig. 1A and B, DAPI) respectively. We found that almost all the GFP fluorescence signals were either coincident with the signals from the DNA stain DAPI (orange and yellow areas in Fig. 1A and B, β′-GFP + DAPI) or located at the edge of the DNA stain DAPI (green areas in Fig. 1A, β′-GFP + DAPI). This indicates that most, if not all, RNAP molecules are localized in the nucleoid and its periphery regardless of cell growth conditions. Note that there were distinctive cytoplasmic spaces in cells grown in either medium (Fig. 1A and B, DIC + DAPI). Also, slow-growing cells in nutrient-poor medium were smaller in size, as reported previously (Nanninga and Woldringh, 1985).

Figure 1.

RNA polymerase distribution in cells grown in different media. Images of rpoC–gfp cells grown in LB (A and C) and in glucose-minimal media (B and D). DIC, phase-contrast images; β′-GFP, fluorescence from the β′-GFP fusion protein; DAPI, DAPI images false coloured red; β′-GFP + DAPI, merge of rpoC–gfp and DAPI images; DIC + DAPI, composite images with the DAPI images overlapping the phase-contrast images.
C and D. Distribution of RNAP in more detail with higher amplification. The regions with increased accumulation of RNAP are indicated by arrows. Scale bar indicates 1 µm.

However, the distribution of RNAP in the nucleoid was sensitive to cell growth. In slow-growing cells, the GFP fluorescence signal was relatively homogeneous, indicating that the RNAP was evenly distributed throughout the nucleoid (compare β′-GFP in Fig. 1B and A and compare Fig. 1C and D). In fast-growing cells, RNAP was concentrated in spots of the nucleoid where the GFP fluorescence signals were much more intense compared with the overall signals (Fig. 1A, β′-GFP, and Fig. 1C). Some of these regions with higher GFP fluorescence signals were at the periphery of nucleoids (Fig. 1A, green areas in β′-GFP and DAPI).

To determine whether transcription is necessary to produce these regions with higher GFP fluorescence signals in fast-growing cells, we added rifampicin to growing cultures in LB broth. The antibiotic rifampicin inhibits overall transcription in the cell by preventing RNAP from initiation and/or reinitiation (Hartmann et al., 1967). We found that those intense GFP fluorescence signals disappeared in nucleoids shortly after the addition of rifampicin (Fig. 2B, β′-GFP). Parallel experiments with a rifampicin-resistant mutant showed that the target for the antibiotic is RNAP, and specifically transcription, because transcription foci were maintained in the mutant in the presence of rifampicin (Fig. 2D, β′-GFP). Thus, these data indicate that transcription is needed for the accumulation of RNAP molecules in certain regions of the nucleoid. We named these regions ‘transcription foci’ as they probably represent areas where active RNAP molecules accumulate in rapid-growing cells. Note that nucleoids of wild-type cells became less condensed after rifampicin treatment as the surrounding cytoplasmic space was diminished (compare Fig. 2B and A, DAPI). On the contrary, the nucleoids of the rifampicin-resistant strain remained compact after rifampicin treatment, and the surrounding cytoplasmic space was evident (compare Fig. 2C and D, DAPI).

Figure 2.

The effect of rifampicin on RNAP distribution. Cells were imaged before (A and C) and after 10 min of treatment with rifampicin (B and D).
A and B. Correspond to rpoC–gfp cells (DJ2599).
C and D. Correspond to a rifampicin-resistant mutant rpoC–gfp strain (DJ2796A).
DIC, phase-contrast images; β′-GFP, fluorescence from the β′-GFP fusion protein; DAPI, fluorescence from the DNA-binding stain DAPI. Scale bar indicates 1 µm.

The number of transcription foci in rapid-growing cells appears to be variable

We analysed the number of apparent transcription foci in 260 nucleoids (Fig. 3). Most nucleoids (61%) had only one or two distinctive transcription foci followed by three (21%) and four (9%) transcription foci, and the percentage of cells with more than four foci was about 3%.

Figure 3.

Number of apparent transcription foci per nucleoid in a population of rpoC–gfp cells. The number of transcription foci per nucleoid was measured in 260 nucleoids. The number of nucleoids containing 0–6 transcription foci is plotted. The numbers in parenthesis indicate the percentage of nucleoids with the specified amount of apparent transcription foci.

To measure the degree of homogeneity of the RNAP distribution in nucleoids from different cells, we quantified the intensity variations in the RNAP–GFP fluorescence signal within individual nucleoids. Texture analysis (Haralick et al., 1973) was used to measure these differences in fluorescence intensities within nucleoids. This approach has been used extensively to analyse nuclear structure in eukaryotic cells (Weyn et al., 2000; Murata et al., 2001a,b). Of the many existing texture parameters, we chose the so-called contrast textural feature because it reflects the variations in grey intensities within an image: the higher these variations are, the larger the contrast value will be (for details, see Experimental procedures). The results of this analysis from 100 cells are shown in Fig. 4. Clearly, the distribution of RNAP from fast-growing cells (LB) is relatively heterogeneous in nucleoids because of the transcription foci, whereas the distribution of RNAP in slow-growing cells (Glc-minimal) is relatively homogeneous as foci are not obvious. Thus, this analysis is indicative of the extent to which transcription foci are present in nucleoids.

Figure 4.

Normalized contrast parameter of the β′-GFP fluorescence signal in nucleoids of rpoC–gfp cells under different physiological conditions. The normalized contrast parameter was measured from 100 nucleoids as described in Experimental procedures. In this analysis, larger values of normalized contrast indicate a more heterogeneous distribution of RNAP within the nucleoids, which in turn is indicative of the presence of transcription foci. The labels across the top specify the relevant genetic markers of the strains used in this study. wt, DJ2599; other strains are DJ2599 derivatives. The labels at the bottom specify the different physiological conditions: LB, cells grown in LB media; Glc-minimal, cells grown in glucose-minimal media; AA starvation, cells after 20 min of treatment with serine hydroxamate.

RNAP redistribution during the stringent response

The two types of RNAP distribution described above were observed in cells under balanced growth conditions. We asked whether a physiological change such as the stringent response, which alters global gene expression, would affect the RNAP distribution in the cell. We induced the stringent response in wild-type cells (DJ2599) grown in LB by the addition of serine hydroxamate to the cultures. Serine hydroxamate is an amino acid analogue that causes starvation for the amino acid serine leading to the stringent response (Tosa and Pizer, 1971; Cashel et al., 1996). Cell samples were taken before and at different times after induction of the stringent response.

We found that RNAP distribution changes during the stringent response (Fig. 5A). Only 5 min after the addition of serine hydroxamate, the transcription foci became more dispersed compared with those before the stringent response. This trend continued and, 20 min after the addition of serine hydroxamate, RNAP molecules were almost evenly redistributed throughout the nucleoid. In addition, the nucleoids decondensed upon the stringent response in wild-type cells as the cytoplasmic space diminished (Fig. 5A, right). Texture analysis also showed that the RNAP distribution becomes homogeneous after stringent response induction (Fig. 4).

Figure 5.

A. RNAP distribution during the stringent response. Fluorescence micrographs of rpoC–gfp (labelled as wt) and rpoC–gfp relA251 (labelled as relA) cells before and after the induction of the stringent response with serine hydroxamate as described in Experimental procedures. The times after the addition of serine hydroxamate are indicated. The first four rows correspond to the fluorescence from the β′-GFP fusion protein. The last row corresponds to the DAPI fluorescence. Scale bar indicates 1 µm.
B. Changes in RNAP distribution before, during and coming out of the stringent response. The stringent response was induced by 3-amino-1,2,4-triazole (1,2 AT) as described in Experimental procedures. The β′-GFP fluorescence images of rpoC–gfp (labelled as wt) and rpoC–gfp relA251 (labelled as relA) cells before, 20 min after the induction of the stringent response and 20 min after the addition of histidine are presented. Scale bar indicates 1 µm.

To confirm that the apparent RNAP redistribution is the consequence of the stringent response, we performed parallel assays with an isogenic DJ2599 relA::kan mutant strain. The relA gene is essential for the stringent response, and cells lacking the gene are defective in ppGpp production and the stringent response (Metzger et al., 1989). In contrast to wild-type cells, transcription foci were clearly maintained during amino acid starvation in the relA mutant (Fig. 5A). Note that the nucleoids became more contracted in the relA mutant cells during amino acid starvation as the cytoplasmic space was enlarged (Fig. 5A, right). Texture analysis indicated that RNAP distribution was heterogeneous in the relA mutant during amino acid starvation (Fig. 4), consistent with the fact that transcription foci were retained.

To determine whether the changes in RNAP distribution are reversible, we followed the RNAP distribution in cells before, during and coming out of the stringent response (Fig. 5B). In this set of experiments, the stringent response was induced by 3-amino-1,2,4-triazole (3-AT), which inhibits the synthesis of histidine, leading to starvation for the amino acid (Tosa and Pizer, 1971). As with serine hydroxamate, 3-AT caused the disappearance of transcription foci in wild type but not in the relA mutant (Fig. 5B, centre). However, when histidine was added to these cultures, the transcription foci reappeared in wild-type cells (Fig. 5B, right). These results demonstrate that RNAP distribution is dynamic and responds rapidly to changes in nutrient conditions.

The distribution of a stringent RNAP mutant grown in rich medium mimics the distribution of the wild-type RNAP during the stringent response

Strains with a class of mutant RNAPs originally called ‘stringent RNAPs’ have reduced transcription from stable RNA genes, which probably contribute to their slow growth phenotype when grown in rich media (Jin and Gross, 1989; Zhou and Jin, 1997; 1998). We introduced one such mutant, rpoB3443 (encoding a L533P mutation in the β subunit), into DJ2599 and asked whether the distribution of the mutant RNAP during steady-state growth in a rich medium is aberrant. Indeed, the distribution of this mutant RNAP in cells growing in LB media was more homogeneous, with far fewer transcription foci compared with wild-type cells (Fig. 6B, β′-GFP, compare with Fig. 6A, β′-GFP). Texture analysis showed that the distribution of the mutant RNAP in DJ2599 rpoB3443 grown in LB resembles that of the isogenic wild-type strain grown in nutrient-poor medium (Fig. 4). Note that the rpoB3443 cells were smaller and that the nucleoid was expanded, as evidenced by the diminished cytoplasmic space. This result further suggests that high RNAP concentrations at stable RNA operons and/or highly active transcription at those loci are necessary for transcription foci establishment and maintenance.

Figure 6.

Comparison of RNAP distribution in wild type and an isogenic stringent RNAP mutant rpoB3443. Images of rpoC–gfp cells carrying a wild-type rpoB allele (A) or the rpoB3443 allele (B). DIC, phase-contrast images; β′-GFP, fluorescence from the β′-GFP fusion protein; DAPI, fluorescence from the DNA stain DAPI. Scale bar indicates 1 µm.


In this report, we visualized E. coli RNAP under different physiological conditions by fluorescence microscopy using a functional rpoC–gfp gene fusion. Formaldehyde fixation of cells enabled us to visualize RNAP distribution in more detail than previously possible and to identify new features of that distribution. Our results show that, although most, if not all, RNAP molecules are located within the nucleoid and its periphery under the conditions tested, the distribution of RNAP is dynamic and is dramatically influenced by environmental cues. Moreover, our mutational analyses suggest that stable RNA synthesis plays an important role in nucleoid compactness in the cell.

Overcoming intrinsic difficulties imaging RNAP in cells under different physiological conditions

In initial experiments with living cells embedded in LB-agarose mounted on microscope slides, we observed very little difference in the distribution of RNAP under different physiological conditions. Given the extreme sensitivity of RNAP distribution to growth conditions, this probably reflects the cellular response to a sudden shift to a limited oxygen environment and not the condition being investigated. Thus, fixation is essential to preserve an accurate snapshot of the RNAP distribution under different physiological conditions. This procedure should be useful for studies of other cellular molecules with distributions that are also sensitive to growth rate and overall metabolic activity.

RNAP is located within and in the surroundings of the nucleoid

Within the limits of resolution of the microscope used, the RNAP–GFP signals were coincident with the DAPI-stained DNA and its periphery, indicating that most, if not all, RNAP molecules are located either within or in the surroundings of the nucleoid, independent of growth and nutrient conditions. This is consistent with a previous report indicating that there is no free RNAP in the E. coli cytoplasm (Shepherd et al., 2001), and is probably attributable to the fact that RNAP is a DNA-binding protein.

The distribution of RNAP in E. coli is dynamic and sensitive to environmental cues

We found that growth rate (resulting from different growth conditions), nutrient starvation and overall transcription activities profoundly affect the distribution of RNAP. In fast-growing cells, RNAP distribution is relatively heterogeneous, being concentrated in regions of the nucleoid and its surroundings that we named transcription foci. In slow-growing cells, the distribution of RNAP was relatively homogeneous, and transcription foci were not evident (Fig. 1B and D). In the presence of rifampicin, which inhibits all transcription activity in the cell, RNAP becomes evenly distributed within nucleoids (Fig. 2B). Moreover, the transcription foci disappear when fast-growing cells are subject to the stringent response by amino acid starvation (Fig. 5A) and reappear if the missing amino acid is added back into the culture (Fig. 5B). Clearly, the distribution of RNAP is dynamic and correlates with the overall transcription activity inside the cell in response to environmental cues.

The fluidity of RNAP distribution in the cell suggests that the enzyme is mobile within nucleoids

This is in contrast to the behaviour of β-galactosidase, a cytoplasmic protein with a similar molecular weight to RNAP, which is found to be rather static inside E. coli cells (Elowitz et al., 1999). It is conceivable that a sliding mechanism (Park et al., 1982; von Hippel and Berg, 1989; Guthold et al., 1999) coupled with interDNA strand-crossing ability of RNAP within nucleoids, as well as DNA movement, facilitate the mobility of RNAP on the E. coli chromosome.

The transcription foci are probably RNAP molecules engaged in stable RNA synthesis

What is the nature of the transcription foci? Several lines of evidence suggest that the transcription foci are the sites of highly active stable RNA synthesis. First, the transcription foci were observed in fast-growing cells in which most RNAP molecules are known to be engaged in stable RNA synthesis (Nomura et al., 1984; Bremer and Dennis, 1996). Secondly, although the transcription foci disappear in wild-type cells during amino acid starvation, they are maintained in an isogenic relA mutant strain in which stable RNA is still actively synthesized (Metzger et al., 1989). Thirdly, electron microscopy has shown that, in rapidly growing cells, RNAP molecules are densely packed inside the rRNA and tRNA operons (French and Miller, 1989). Because the rRNA operons are longer than those for tRNAs, and therefore can load more polymerase molecules, we believe that, to a first approximation, the transcription foci represent RNAP molecules engaged in rRNA operon transcription. However, we also note that a large number of tRNA operons/genes are clustered with rRNA operons so that transcription from the foci probably involves the synthesis of both stable RNAs. Direct determination of the genomic regions involved in these foci is needed to resolve this issue.

Somewhat surprisingly, the number of transcription foci per nucleoid does not appear to correspond to the number of stable RNA operons per genome. In E. coli, there are seven rRNA operons, most of which are located near the origin of replication (oriC), and the direction of transcription of all rRNA operons is parallel to that of replication. However, the effective number of rRNA operons is actually more than seven in most fast-growing cells because of reinitiation of DNA replication at oriC before cytokinesis (Bremer and Dennis, 1996). Interestingly, among the 260 nucleoids examined, most (≈ 90%) contain four or fewer transcription foci per nucleoid (Fig. 3). It is possible that the disparity between the number of apparent transcription foci and the number of rRNA operons reflects the limitation of resolution of the microscope, the limitations that arise from counting transcription foci in two-dimensional images of three-dimensional structures and/or the subjectivity involved in counting transcription foci. However, an alternative explanation that we favour is that these foci are transcription ‘factories’ for the synthesis of stable RNA, and form a structure(s) analogous to the eukaryotic nucleolus (Iborra et al., 1996; Cook, 1999). In this model, the rRNA operons that are spatially distant on the chromosome could be brought together in these transcription foci, contributing to the condensation of the genome (see below and Fig. 7). Intriguingly, some transcription foci appear to be at the periphery of the nucleoid (Fig. 1A, β′-GFP + DAPI), presumably to facilitate the release of transcription products into the cytoplasm. It should be noted that, because our images are two-dimensional, foci that appear to be within the nucleoid could actually be in the periphery of the nucleoid. Three-dimensional imaging techniques, such as confocal microscopy or the deconvolution of Z-stacks of individual images, are required to address this issue.

Figure 7.

Model of stable RNA synthesis, RNAP distribution and the condensation of the E. coli chromosome. The E. coli chromosome is represented as black lines folded in loops. The oriC region, the rRNA/tRNA operons and the RNAP molecules are represented by a black square, red circles and green circles respectively. Each of the rrn operons is labelled with its corresponding letter. For simplicity, only two tRNA operons are drawn. The top part of the diagram represents an expanded nucleoid in cells where stable RNA synthesis or cell growth is inhibited during the stringent response. When transcription of the stable RNAs is resumed, the RNAP molecules transcribing the stable RNA operons assemble into two putative transcription factories (bottom part, big green circles labelled 1 and 2) pulling different stable RNA operons into proximity and making the nucleoid more compact (transcription foci 1 includes the ribosomal operons C, A, B and E; transcription foci 2 includes the ribosomal operons H, D, G and two tRNA operons, tRNA1 and tRNA2). The supercoilings introduced by active transcription in the transcription factories reinforce the condensation of the chromosome. The bottom part represents only one of many possible arrangements.

Is the stable RNA synthesis a driving force for the condensation of the E. coli chromosome?

Two lines of evidence argue for the involvement of stable RNA synthesis in nucleoid condensation in E. coli. First, the nucleoids became loosened in wild-type cells when stable RNA synthesis was preferentially inhibited during the stringent response (Fig. 5). In contrast, nucleoids remained condensed in relA mutant cells where stable RNA synthesis was maintained during amino acid starvation. Secondly, nucleoids are loosened in the rpoB3443 mutant containing a stringent RNAP defective in rRNA synthesis even when grown in nutrient-rich LB broth (Fig. 6). It has been reported that RNase treatment destroys isolated bacterial nucleoids, indicating that RNA is important for the integrity of nucleoids (Pettijohn, 1996; Murphy and Zimmerman, 2002); however, the nature of RNA species responsible for this is unknown. Our study suggests that stable RNA synthesis is a driving force in the condensation of the E. coli chromosome.

We propose a working model to explain how stable RNA synthesis affects RNAP distribution and the condensation of the chromosomal DNA in bacteria (Fig. 7). It is known that the E. coli chromosome is folded in loops within the cell (Pettijohn, 1996). It is conceivable that the chromosomal loops are dynamic with many transient interactions among them. In growing cells, high concentrations of elongating RNAP on stable RNA genes solidifies interactions between chromosomal loops containing those operons, by RNAP–RNAP interactions (Shaner et al., 1982) and/or transcription-induced supercoiling (Wu et al., 1988). This results in the formation of transcription foci leading to nucleoid condensation (Fig. 7, bottom). During the stringent response, inhibition of stable RNA synthesis eliminates the interaction or folding of chromosomal loops leading to decondensation of the nucleoid (Fig. 7, top). In rapidly dividing cells, the majority of RNAP is in transcription foci engaged in stable RNA synthesis, with the remainder involved in mRNA synthesis elsewhere in the genome. Conversely, during the stringent response, stable RNA synthesis is dramatically reduced, leading to the dissolution of the foci and, thus, an increase in available RNAP needed for mRNA synthesis. This type of redistribution of RNAP as a consequence of the stringent response has been proposed in our previous work (Zhou and Jin, 1998). Recently, RNAP has been proposed as a force that helps chromosome segregation (Dworkin and Losick, 2002). It is an intriguing possibility that stable RNA synthesis-mediated chromosome condensation, as outlined here, facilitates chromosome segregation.

Experimental procedures

Bacterial strains

The basic bacterial techniques used have been described elsewhere (Miller, 1972). The bacterial strains used in this study are described in Table 2.

Table 2. . Bacterial strains used in this work.
StrainRelevant genotypeReference or source
MG1655Wild-type E. coli K-12Laboratory collection, Carol Gross
CF1651MG1655 relA251::kan Metzger et al. (1989)
DY329W3110 ΔlacU169 nadA::Tn10 gal490 λcI857Δ(cro-bioA) Yu et al. (2000)
DJ2599MG1655 rpoC–gfp AmpRThis work
DJ2796AMG1655 rpoC–gfp AmpRRif RThis work
RLG1319MG1655 ΔlacX74/lacUV5–lacZRichard Gourse, Zhou and Jin (1998)
RLG1350MG1655 ΔlacX74/rrnB P1–lacZRichard Gourse, Zhou and Jin (1998)
DJ2608B1RLG1319 rpoC–gfp AmpRThis work
DJ2608C1RLG1350 rpoC–gfp AmpRThis work
DJ2650DJ2599 relA251::kanThis work
DJ2609DJ2599 rpoB3443This work

The replacement of the rpoC gene by a chimeric rpoC–gfp gene was carried out using an efficient recombination system (Yu et al., 2000) and is described elsewhere (Cabrera and Jin, 2003). Briefly, we amplified by polymerase chain reaction (PCR) a DNA fragment from the pGFPuv plasmid (Clontech) using the oligos JC99A and JC99C. The sequence of the oligo JC99A was: 5′-CCAGCCTGGCAGAACTGCTGAACG CAGGTCTGGGCGGTTCTGATAACGAG C T A GAAATAATGA GTAAAGGAGAAGAACTTTTCACTGG-3′. The sequence of the oligo JC99C was: 5′-CCCCCCATAAAAAAACCCGCC GAAGCGGGTTTTTACGTTATTTGCGGATTATGGTCTGAC AGTTACCAATGC-3′. The first 50 bp of the 5′ end of the DNA fragment obtained are identical to the last 50 bp before the stop codon of the rpoC gene (underlined in the sequence of oligo JC99A). Immediately downstream of this 50 bp, the DNA fragment has three codons encoding for a three-amino-acid linker (bold characters in the sequence of oligo JC99A, encoding for the amino acid sequence Leu-Glu-Ile) that replaces the stop codon of the rpoC gene. Downstream from the three-amino-acid linker, the DNA fragment consists of the gfpuv and the ampicillin resistance (Ampr) genes. For simplicity, we refer to the gfpuv gene as gfp. The last 51 bp of the 3′ end of the DNA fragment are identical to the bases located 4 bp downstream from the stop codon of the rpoC gene (underlined in the sequence of oligo JC99C). After purification of the linear DNA fragment, we used it to electroporate DY329 cells. The recombinant clones were selected on LB-ampicillin plates and analysed further by Western blotting and DNA sequencing. The rpoC–gfp allele was moved into MG1655 by phage P1 transduction. One of the resulting Ampr transductants was named strain DJ2599.

To measure the transcription from the rrnB P1 and lacUV5 promoters in an rpoC–gfp strain, we moved the rpoC–gfp allele into the RLG1319 and RLG1350 strains (obtained from W. Ross and R. Gourse, University of Wisconsin, Madison, WI, USA) by phage P1 transduction. RLG1350 is MG1655 lacX74/rrnB P1 (81-+1)-lacZ and RLG1319 is MG1655 lacX74/lacUV5-lacZ (Zhou and Jin, 1998). The rpoC–gfpuv Ampr derivative  of  RLG1350  was  named  DJ2608B1,  and the rpoC–gfp Ampr derivative of RLG1319 was named DJ2608C1. A spontaneous rifampicin-resistant rpoC–gfp strain (DJ2796A) was isolated by plating 0.1 ml of overnight cultures of DJ2599 on LB + rifampicin (50 µg ml−1) plates as described previously (Jin et al., 1988). The mutant DJ2796A had a similar growth rate to the parental strain DJ2599.

The relA251::kan allele was moved into strain DJ2599 by P1 transduction with a lysate made from strain CF1651(Metzger et al., 1989), and Kanr transductants were selected. The relA251 derivative of DJ2599 was named DJ2650.

The rpoC–gfpuv allele was moved into the MG1655 strains carrying the rpoB3443 mutation by P1 transduction, and the Ampr transductants were selected. The transductants were scored for rifampicin-resistant and slow growth phenotypes (Jin et al., 1988). The rpoC–gfp derivative of the rpoB3443 strain was named DJ2609.

Chemicals and reagents

All chemicals were obtained from Sigma. Antibodies against the GFPuv protein (Anti-Living Colors) were obtained from Clontech. Monoclonal antibodies against the β′ subunit were a gift from R. Burgess (University of Wisconsin-Madison).

Bacterial growth

All cultures were grown with vigorous agitation in a water bath at 30°C. Fresh overnight cultures were diluted 1:500 into fresh media. Cells were grown in M63 media supplemented with glucose (0.2% final concentration) or in Luria–Bertani (LB) medium. Samples for microscopic observation were taken at OD600 of ≈ 0.4. To induce the stringent response, serine hydroxamate was added to a final concentration of 100 µg ml−1. In experiments in which histidine starvation was induced, the cells were first grown in a defined media that contains glucose (0.2% final concentration) as the carbon source, nucleosides, bases and all amino acids except histidine. To induce histidine starvation, 3-amino-1,2,4-triazole was added to a final concentration of 40 mM. To interrupt histidine starvation, histidine was added to a final concentration of160 mM.

β-Galactosidase assays

β-Galactosidase assays were performed as described previously (Cabrera and Jin, 2001). Culture aliquots were lysed in a microtitre plate and exposed to the chromogenic substrate ONPG. Kinetic measurements were made using a SpectraMax 250 microtitre plate reader (Molecular Devices) to obtain the value of Vmax. Units were presented as specific activities, which were the products of the ratio (Vmax divided by A600 of the culture) times a factor of 25. The specific activities thus obtained have been determined empirically to correspond to the standard Miller units. For each culture, triplicate samples were taken at each time point, and the results were expressed as the mean of the three measurements. The standard deviations of the triplicates were < 5%.


All samples for microscopy were fixed with formaldehyde. To fix the cells, 5 ml aliquots of culture were removed from the culture flasks, and formaldehyde was added to a final concentration of 3.7%. Cells were fixed during 1 h at room temperature, centrifuged and resuspended in 1 ml of 1× PBS. Before mounting the cell mixture on the slides, 15 µl of a 10 µg ml−1 solution of DAPI was added. The final mixture of fixed cells was mounted on slides using 1% low-melting-point agarose. Microscopy was performed in a Zeiss Axiophot II microscope equipped with a Plan-Apo 100× objective, epifluorescence filters and a 2.5 optovar. Images were captured with a CCD camera (Micromax) working at 2 × 2 binning. The images were processed with Adobe photoshop.

Contrast analysis of nucleoids

At each position in the nucleoid, the β′-GFP fluorescence signal is proportional to the concentration of RNAP. The homogeneity in the RNAP distribution can be evaluated by measuring the differences in fluorescence signals between each position of the nucleoid and its neighbouring positions. In nucleoids in which RNAP is not distributed homogeneously (nucleoids with transcription foci), there will be on average more differences in the fluorescence signal between neighbouring positions than in nucleoids in which the RNAP is distributed homogeneously (nucleoids without transcription foci). Because the nucleoids are rod-shaped, a decrease in the grey levels is expected in the edge regions. This decrease in the grey levels in the edge regions is not interesting for our purposes. We then chose to analyse the central region of each nucleoid. To make this selection systematically, we chose the region of the nucleoid that has grey levels above the mean of the grey levels of the entire nucleoid. We called this the region of interest (ROI) of each nucleoid. In each ROI, we measured the intensity of each pixel and its eight neighbouring pixels. The differences in intensities were used to feed a grey level co-occurrence matrix as described by Haralick et al. (1973). This grey level co-occurrence matrix is a representation of the grey level transitions within the ROI, and the contrast textural feature can be calculated from it. The measurement of the contrast textural feature was done with nucleoids of similar grey levels (i.e. similar fluorescence intensities). To normalize the contrast parameter by area and grey level intensity, we divided it by the area and by the mean's square. We named the resulting number normalized contrast. To carry out the computations described here, a Java applet named Nucleoid Analyzer was written. This applet has a graphical user interface that allows one to visualize the pictures taken by the microscope's camera. Upon selection of the nucleoid by the user, the Nucleoid Analyzer applet performs automatic selection of the ROI and the calculations described before.


We thank Mr Albert Mao (currently at Duke University) for writing the Java Applet that we used in the analysis of the nucleoids during his summer internship in the laboratory. We also thank Drs James McNally and Tatiana Karpova (NCI Core Fluorescence Imaging Facility) for their advice during our initial studies. Finally, we are grateful for the productive interactions and comments from Dr Tim Durfee (University of Wisconsin, Madison) and from Drs Stuart Austin, Don Court, Mikhail Kashlev and Jeff Strathern (Gene Regulation and Chromosome Biology Laboratory, NCI, Frederick, MD, USA). We also thank Drs Sigal Ben-Yehuda (Harvard) and Michael Yarmolinsky (NCI) for their comments.