Transcription-induced barriers to supercoil diffusion in the Salmonella typhimurium chromosome

Abstract

Transcription and replication both influence and are influenced by superhelical changes in DNA. Explaining how supercoil movement is channeled in living chromosomes has been a major problem for 30 years. Transcription of membrane-associated proteins leads to localized hypersupercoiling of plasmid DNA, and this behavior indicates the presence of aberrant supercoil diffusion. Using the lambda Red recombination system, we constructed model domains in the Salmonella typhimurium chromosome to analyze supercoiling dynamics of regions encoding membrane proteins. Regulation of Tn10-derived tetracycline resistance involves a repressor, TetR, and a membrane-bound export pump, TetA. Strains deficient in TetR activity had 60-fold higher transcription levels (from P_A) than TetR-positive strains. High tetA transcription caused a 10- to 80-fold decrease in the γδ resolution efficiency for the domain that includes the Tet module. Replacing tetA with genes encoding cytosolic proteins LacZ and Kan also caused the appearance of supercoil diffusion barriers in a defined region of the chromosome. In strains containing a functional TetR located next to a regulated lacZ reporter (P_RtetR-P_AlacZ), induction of transcription with chlortetracycline caused a 5-fold drop in resolution efficiency in the test domain interval. A short half-life resolvase showed that barriers appeared and disappeared over a 10- to 20-min span. These studies demonstrate the importance of transcription in chromosome structure and the plasticity of supercoil domains in bacterial chromosomes.

Numerous studies demonstrate that bacterial chromosomes are partitioned into independent domains that limit supercoil diffusion (1, 2). Early analyses of bacterial chromosomes liberated from cells by treatment with lysozyme and detergent showed that “nucleoids” were negatively supercoiled objects and that multiple single-strand breaks were needed to relax all supercoils (3, 4). Sinden and Pettijohn subsequently confirmed that chromosomes were segregated into multiple independent domains in vivo by using tri-methylpsoralen crosslinking (5). In 1996, an assay was developed to measure supercoil diffusion inside living cells by using the site-specific recombination reaction of Tn3 or the closely related element γδ (6). In this assay, the ability to delete specific chromosome segments separated by two γδ res sites spaced along the chromosome measures supercoil diffusion (7, 8). What elements are responsible for domain structure is a long-standing question.

One advantage of the site-specific resolution assay is that it shows how supercoil structure at specified locations is modified when growth rate or biological activity changes. Most resolution barriers are random with respect to DNA sequence, but the appearance of a new barrier has been documented. For example, induction of the early promoter of the bacteriophage Mu caused a new barrier to supercoil diffusion that was restricted to the region near the virus center (9).

The impact of transcription on DNA structure is well documented in plasmids (10). For example, transcription of the pBR322 tetA gene causes dramatic hypersupercoiling when the cell also harbors a mutation in the topA gene, which encodes a type I topoisomerase (ω protein) (11–14). Studies by Lynch and Wang showed that hypersupercoiling from a moderate-strength promoter required both the presence of a topA mutation plus transcription of a protein that was cotranscriptionally inserted into the membrane or exported to the periplasm or outer membrane (15). Similar studies have been difficult to conduct in the bacterial chromosome for two reasons. First, it has been difficult to construct test intervals in the 4-Mb bacterial genome. Second, bacterial chromosomes have supercoiling domains larger than most multicopy plasmids.

A new supercoil diffusion barrier appears after induction of tetA transcription. Moreover, transcription of two cytosolic proteins, LacZ and Kan, showed similar effects. A short half-life resolvase revealed that barriers appear and disappear over a 10- to 20-min span. These studies demonstrate an extremely dynamic domain structure for bacterial chromosomes.

Materials and Methods

Media. LB and minimal media were prepared as described (6). Antibiotics were added to medium at concentrations of 50 μg/ml for kanamycin and ampicillin, 20 μg/ml for chloramphenicol, 12 μg/ml for tetracycline, and 10 μg/ml for gentamicin. Bochner plates were made as described (16).

Plasmids. Plasmid pJBREScI is a pACYC184-derived vector carrying the tnpR (WT Res resolvase protein) gene of transposon γδ cloned under control of the λP_L promoter with a nearby copy of λ cIts repressor gene (17). Plasmid pJBREScI-SSRA carries a short half-life Res resolvase in which the sequence tag encoded by ssrA RNA was added to the carboxyl terminus of Res. λ Red helper plasmid pKD20 is a derivative of pINT-ts that encodes Red recombinase under the control of the P_araB promoter (18).

Chromosomal Modification with Red Recombination. Insertion mutagenesis of the Salmonella typhimurium chromosome was done with the lambda Red recombination system of plasmid pKD20 (18). Synthetic oligos used for each construct is described in Table 2, which is published as supporting information on the PNAS web site. Oligonucleotides were purchased from IDT (Coralville, IA). Taq polymerase (Sigma) and Taq extender (Stratagene) were mixed 1:1 to generate DNA for sequence analysis and to make gene-replacement substrates. Recombinants with the correct insertions were selected for tetracycline resistance or kanamycin resistance or screened for blue-color phenotype and tested as described (6, 19). Details of strain constructions are given in Supporting Text, which is published as supporting information on the PNAS web site.

Assays. To measure resolution, log-phase cultures grown at 30°C to 50 Klett units (filter no. 54) were induced for Res expression by a 10-min incubation at 42°C followed by 10-fold dilution in LB containing chloramphenicol. Cells were incubated overnight at 30°C and then diluted 1:10⁶ and plated on minimal medium with chloramphenicol and 5-bromo-4-chloro-3-indolyl β-d-galactoside. Deletion frequencies were scored by loss of color or loss of antibiotic resistance (6). To measure lacZ expression, overnight bacterial cultures diluted 1:100 in 5 ml of LB were grown to 50 Klett units. β-Galactosidase (β-gal) assays were done in triplicate as described (20).

Results

Supercoil Diffusion in the 17-Min Region of S. typhimurium. Transcription can lead to anomalous DNA structure and a hypernegative supercoiling phenotype in plasmid DNA (14, 15, 21, 22). For efficient promoters like the hybrid tac promoter under control of the Lac repressor, transcriptional effects on plasmid DNA topology can be documented in WT cells (23). With weaker promoters like the one regulating the tetracycline resistance gene on pBR322, cotranscriptional translation of membrane proteins stimulate hypernegative supercoiling only in topA mutants (15, 24). Although several examples of translation-driven supercoiling are restricted to topA mutants, the behavior has led to theories of cotranscription membrane association, helping to organize chromosomal DNA after replication (25, 26). A serendipitous result suggested that the tetA gene of Tn10 might also cause topological anomalies in the chromosome. A genetic interval in S. typhimurium was flanked with a pair of res sites by using two combinations of drug markers. Strain NH3470 had a Tn5 kanamycin resistance gene (kan) upstream and a Tn10-derived tetR/tetA module downstream. Strain NH3471 had aTn1696-derived gentamicin resistance gene upstream (gen) and the Tn5 kan gene downstream (Table 1 and Fig. 1A). Recombination between res sites in strain NH3470 (Fig. 1 A Upper) results in sensitivity to tetracycline, and in NH3471 (Fig. 1 A Lower) resolution leads to kanamycin sensitivity.

Table 1. Strains used.

Strain	Genotype	Plasmid
NH3470	zbj8904::Tn10d(Gn)⋄(kan) (ylil)⋄(PRtetR435-PAtetA-res)	pJBREScl
NH3471	zbj8904::Tn10dGn (ylil)⋄(kan-res)	pJBREScl
NH3472	zbj8904::Tn10d(Gn)⋄(kan) (ylil)⋄(PRtetR-PAtetA-res)	pJBREScl
NH3475	zbj8904::Tn10d(Gn)⋄(kan) (ylil)⋄(PAtetA-res)	pJBREScl
NH3477	(ylil)⋄(PRtetR435-PAlacZ-re s)
NH3478	(ylil)⋄(PRtetR-PAlacZ-res)
NH3479	(ylil)⋄(PAlacZ-res)
NH3480	zbj8904::Tn10dkan (ylil)⋄(PAtetA-rrnB-res)	pJBREScl
NH3481	cobP712::Tn10dGn cobT714::MudJr2	pJBREScl
NH3482	cob-708::Tn10dGn cobT714::MudJr2	pJBREScl
NH3483	cobT712::Tn10dGn cobT714::MudJr2(kan)⋄(PAtetA)	pJBREScl
NH3484	cobP712::Tn10dGn cobT714::MudJr2(kan)⋄(PAtetAinv)	pJBREScl
NH3485	cob-708::Tn10dGn cobT714::MudJr2(kan)⋄(PAtetA)	pJBREScl
NH3486	cob-708::Tn10dGn cobT714::MudJr2(kan)⋄(PAtetAinv)	pJBREScl
NH3487	zbj8904::Tn10d(Gn)⋄(kan) (ylil)⋄(PRtetR-PAlacZ-rrnB- res)	pJBREScl
NH3488	zbj8904::Tn10d(Gn)⋄(kan) (ylil)⋄(PAlacZ-rrnB-res)	pJBREScl
NH3489	zbj8904::Tn10dGn (ylil)⋄(PRtetR-PAkan-rrnB-res)	pJBREScl
NH3490	zbj8904::Tn10dGn (ylil)⋄(PAkan-rrnB-res)	pJBREScl
NH3491	(recN)⋄(PRtetR-PAtetA) zbj8904::Tn10d(Gn)⋄(kan) (ylil)⋄(PAlacZ-rrnB-res)	pJBREScl
NH3492	zbj8904::Tn10d(Gn)⋄(kan) (ylil)⋄(PRtetR-PAlacZ-rrnB- res)	pJBREScl-SSRA

All strains were derived from S. typhimurium LT2 and were constructed during this work. The Tn10dGn and MudJr2 elements were described (6). ⋄ indicates a replacement generated by lambda Red recombineering (47).

Fig. 1.

Effect of WT tetR, a mutant tetR, and a tetR deletion on γδ resolution efficiency. (A) Physical and genetic map of a 12-kb interval between zbj8904 and yliI in the 17- to 20-min region of S. typhimurium includes eight ORFs flanked by drug markers and res sites. Two arrangements of selectable drug markers involve kanamycin and tetracycline resistance (Upper) and gentamicin and kanamycin resistance (Lower). (B) Resolution efficiency comparison among different module insertions in the gene yliI in the 12-kb interval in the log phase. Resolution efficiency for a 12-kb interval in the 43- to 45-min region (6) is designated as the expected resolution efficiency.

Based on studies of the genetic segment between the cob and his operons, we expected resolution efficiency for a 12-kb interval to be 80% (Fig. 1B, first column). However, NH3470 had a resolution efficiency of only 3%, which is 26-fold lower than expected (see Fig. 1B, second column). Strain NH3471, analyzed by using the same protocol except that kanamycin sensitivity was scored, had a resolution frequency of 57%, which is close to the expected level (Fig. 1B, third column). This unexpected low efficiency suggested that the tet cassette in strain NH3470 caused a low-resolution phenotype.

We have used Tn10-derived tet-res elements for resolution analyses in many different intervals where they did not show a low-resolution phenotype. One explanation is that a mutation (like an altered res site) generated during PCR amplification reduced resolution efficiency. To test this hypothesis, a new strain was made by using different PCR reactions to generate the tet-res module. A resulting strain (NH3472) had a resolution efficiency of 56%, which was similar to the efficiency of the strain NH3471 (57%). This experiment supports the notion that the tet module in strain NH3470 was unusual.

Sequence analysis of tet modules from NH3470 and NH3472 showed that NH3470 res site was unaltered, but it had a deletion of base pair 435 in the tetR gene (ΔA-435). Hereafter, we refer to this allele as tetR435. The deletion causes a frameshift and premature translation stop that eliminates 66 amino acids of the WT TetR protein (Fig. 6, which is published as supporting information on the PNAS web site). Although the DNA-binding domain of TetR435 repressor is unaltered (27), destabilization of the TetR dimer interface might lead to constitutive tetA gene expression (28).

Tet expression in strains with WT tetR and tetR435 alleles was measured in two ways. First, Bochner plates were used to test sensitivity to fusaric acid. These plates are commonly used to counterselect against strains carrying Tn10-derived tetracycline resistance (16). TetA expression is toxic at 42°C in the presence of fusaric acid. Standard Bochner plates contain chlortetracycline hydrochloride (CLT), which induces the Tet operon by binding the TetR, causing its dissociation from DNA and transcription of both tetA and tetR (28). Fusaric acid sensitivity was measured in normal Bochner plates and modified plates lacking CLT. Strain NH3472 with a WT tetR gene grew well on Bochner plates without CLT but was killed on a standard Bochner plate (Fig. 7, which is published as supporting information on the PNAS web site). Strain NH3470 died on Bochner plates with or without CLT, which indicates constitutive TetA expression.

Second, to quantify transcription from the tetA promoter, chromosomal lacZ was fused to a tetA promoter (P_A) by using lambda Red substitution. NH3478 (WT tetR) made 15 units of β-gal in the absence of CLT and 688 units of β-gal after addition of 5 μg/ml CLT. NH3477 (tetR435) made 959 units of β-gal with or without CLT addition (Table 3, which is published as supporting information on the PNAS web site). Thus, tetR435 had a higher level of tetA expression than a WT tetR construct induced with the optimal amount of CLT. The effect of tetA expression was local and not generally inhibitory to resolution, because a Tet element 1,800 kb away from the test interval had no impact when the region had no tetA promoter (data not shown).

Transcription Hinders Supercoil Diffusion. Although the expression data shown above reveal a correlation between tetA expression and reduced γδ resolution, transcription and translation levels of TetA are not the only differences between these strains. The TetR435 protein might cause DNA damage or alter DNA structure by unexpected mechanisms. To check this possibility, strains NH3475 and NH3479 were made, which lack TetR and constitutively express tetA or lacZ, respectively, from the P_A promoter. NH3475 (P_AtetA allele) was similar to NH3470, being inviable on Bochner plates with or without CLT (Fig. 7). Strain NH3479 was similar to NH3477 because it expressed 915 units of β-gal (Table 3). The resolution efficiency of a strain lacking TetR (NH3475) (yliI)<>(P_AtetA-res) was identical (3%) with the strain carrying the tetR435 allele (Fig. 1B, fifth column). Thus, tetR435 and a tetR deletion have the same phenotype.

The hypothesis at this point was that TetA protein expression induces a position-specific barrier to chromosome supercoil diffusion, perhaps by handcuffing the DNA to the cell membrane during translation. However, a trivial explanation is that RNA polymerase transcribes through the res site, causing Res protein to dissociate. This mechanism would block site-specific resolution by occluding Res binding to one res site. To address this possibility, a pair of efficient transcription terminators (rrnB T/T2) was placed at the 3′ end of tetA. After the TetA termination codon is a 105-bp sequence followed by the rrnB T1/T2 terminators (158 bp in length), and the γδ res site is 129 bp beyond the terminators. In this configuration the res site is shielded from occlusion by efficient (>95%) transcription terminators. Strain NH3480, which contains this construct (yliI)<>(P_AtetA-rrnB-res), retained a low-resolution phenotype (6% recombinants; see Fig. 1B, sixth column). A 2-fold occlusion effect is possible, because the resolution efficiency rose from 3% without the terminator to 6% when the terminator was present. Nonetheless, most of the inhibitory effect of tetA expression on resolution remained to be explained.

TetA Barriers Are Mobile. To see whether tetA expression alters supercoiling behavior in a different chromosomal setting, several tetA configurations were moved to the his-cob region. This genetic interval has been thoroughly characterized, and it has no sequence- or region-specific barriers (6, 19, 29). Two intervals that normally resolve at high frequency include a 14-kb segment defined by cobP712::Tn10dGn and cobT714::MudJr2 (Fig. 2A, NH3481) and a 28-kb interval defined by cob-708::Tn10dGn and cobT714::MudJr2 (Fig. 2 A, NH3482). In these experiments, a ΔtetR P_AtetA module was substituted for the kan gene in cobT714::MudJr2 (Fig. 2B). Replacements were made in two orientations to control for transcription occlusion. In NH3483 and NH3485, tetA transcription points toward the nearby res site [Fig. 2B, MudJr2 (kan)<>(P_AtetA)]. In NH3484 and NH3486 tetA transcription is directed away from the res site [Fig. 2B, MudJr2 (kan)<>(P_AtetA-inv)]. In this orientation RNA polymerase cannot occlude a res site.

Fig. 2.

γδ resolution in the 43- to 45-min of S. typhimurium chromosome. (A) Physical map of two intervals. The 14-kb interval from cobP712 to cobT714 is labeled a, and the 28-kb interval from cob-708 and cobT714 is labeled b. (B) Genetic map of res elements placed at the cobT714 location: MudJr2, MudJr2(kan)<>(P_AtetA), and MudJr2(kan)<>(P_AtetA-inv). (C) Resolution efficiencies for different 14-kb (a) and 28-kb (b) intervals.

Resolution efficiency over the 14-kb interval between cobP712::Tn10dGn and cobT714::MudJr2 (NH3481) was 98% (Fig. 2Ca, MudJr2). When the kan gene in MudJr2 was replaced by P_AtetA (NH3483), resolution efficiency dropped to 2% (Fig. 2Ca, P_AtetA). When the kan gene was replaced by inverted P_AtetA (NH3484), resolution efficiency was 5% (Fig. 2Ca, P_AtetA-inv). Thus, constitutive tetA expression caused a 20- to 50-fold decrease in resolution efficiency for this 14-kb interval. Resolution efficiency for the 28-kb interval between cob-708::Tn10dGn and cobT714::MudJr2 (NH3482) was 78% (Fig. 2Cb, MudJr2). Replacement of the kan gene in MudJr2 with P_AtetA (NH3485) caused resolution efficiency to decrease to 1% (Fig. 2Cb, P_AtetA). When kan gene was replaced by an inverted P_AtetA (NH3486), resolution efficiency was 2% (Fig. 2Cb, P_AtetA-inv). Constitutive tetA expression reduced resolution efficiency in the 28-kb interval by 40- to 80-fold. These results prove that the tetA gene is a mobile element that can disrupt supercoil diffusion. Occlusion here is unlikely in either case, so the major impact is due to transcription and/or translation of tetA.

Genes Encoding Cytoplasmic Proteins Also Create Supercoil Barriers. Is membrane association necessary to disrupt supercoil diffusion? To address this question, we measured γδ resolution in intervals with lacZ and kan genes fused to the tetA promoter. When lacZ expression was repressed by TetR (NH3487), γδ resolution was efficient (65% in Fig. 3 A and B, column 2), but when lacZ expression was unregulated (NH3488), γδ resolution was very inefficient (4% in Fig. 3 A and B, column 3). The magnitude of disruption to supercoil diffusion in NH3488 was similar to a strain constitutive for TetA expression (6% in Fig. 3 A and B, column 1). Similarly, when kan was placed at P_A under the control of TetR (NH3489), γδ resolution was efficient (70% in Fig. 3 A and B, column 5). But unregulated kan expression (NH3490) caused γδ resolution to fall to the level of cells expressing TetA (7% in Fig. 3 A and B, column 6). Clearly, a membrane anchor is not a prerequisite for generating barriers to supercoil diffusion; transcription of two genes encoding cytoplasmic proteins reduced γδ resolution as much as transcription of tetA.

Fig. 3.

Effect of genes encoding cytoplasmic proteins on the γδ resolution efficiency in the 12-kb interval between zbj8904 and yliI. (A) Genetic map of different protein coding sequences (tetA, lacZ, and kan) fused to tetA promoter P_A. (B) Resolution efficiency comparison among different insertions in the gene yliI in the 12-kb interval in the log phase.

To see whether resolution could be modulated by using a distant source of tetR repressor, tetR was cloned into recN (NH3491), which is 1,800 kb away from gene yliI. With no source of TetR, strain NH3488 made dark-blue colonies, and resolution was an inefficient 4% (Fig. 3 A and B, column 3). When TetR protein was provided in trans, NH3491 made light-blue colonies and γδ resolution was efficient (67%, Fig. 3 A and B, column 4). Thus, TetR regulation of the tetA promoter clearly determines the efficiency of resolution reactions.

Barrier Kinetics. Supercoil barriers in two regions of the genome appear when genes encoding tetA, lacZ, or kan are expressed from a tetA promoter P_A (Figs. 1, 2, 3). To explore the correlation between transcription level and γδ resolution efficiency, a dose–response experiment was carried out in strain NH3492, which has the lacZ fused to the tetA promoter next to a WT tetR repressor. Cultures grown overnight in LB were diluted 100-fold in medium containing 0, 0.1, 0.5, 1, 5, 10, or 20 μg/ml CLT. Cell growth at 30°C was monitored in Nephlo flasks by using a Klett meter, and when cultures reached the value of 50 Klett units, half of the cells were harvested by centrifugation and assayed for β-gal activity. The remaining half of each culture was incubated at 42°C for 10 min to induce expression of a tagged form of γδ Res protein (see below). Cells were put back in culture at 30°C to allow chromosome segregation and then diluted and spread on plates containing the chromogenic indicator 5-bromo-4-chloro-3-indolyl β-d-galactoside. Resolution efficiency was calculated as the percentage of white colonies among all colonies. WT Res protein is stable, and once induced, it can catalyze recombination for a time lasting longer than two cell divisions, which obscures the kinetic analysis of barrier disappearance. The Res-SsrA protein has a C-terminal extension of 11 amino acid residues that are identical with the Clp-XP proteolysis tag of the SsrA system (30). Research (R.S., unpublished work) shows that this protein has a half-life of 5 min in exponential cultures of S. typhimurium. The Res-SsrA protein is fully functional and has the same in vivo resolution efficiency as WT Res if either the expressing cells lack functional ClpXP protease or the 11-aa tag is modified so that it is no longer recognized as a ClpXP/P substrate (data not shown). For a 10-min induction followed by outgrowth at 30°C, 90% of all resolution in strains with the Res-SsrA protein occurs in a 5- to 12-min window after the shift to 42°C.

A reciprocal relationship between β-gal activity and γδ resolution efficiency was observed (Fig. 4). At 0.1 μg/ml CLT, a small induction of β-gal expression produced a small inhibitory effect on resolution (Fig. 4 Inset). With the increase of CLT concentration, β-gal expression increased and resolution decreased. At 5 μg/ml CLT, β-gal activity peaked and γδ resolution efficiency approached a minimum. At very high concentrations of CLT, cells grow poorly, which complicates both resolution and β-gal assays.

Fig. 4.

Effect of chlortetracycline concentration on β-gal activity and γδ resolution efficiency. NH3492 was subcultured in 0, 0.1, 0.5, 1, 5, 10, and 20 μg/ml chlortetracycline and was grown to a cell density of 50 Klett units. Both β-gal assays (•) and γδ assays (○) were performed on each sample.

To study the kinetics of barrier appearance, fresh overnight cultures of NH3492 were diluted 100-fold into LB at 30°C, and cell growth was followed in a Klett meter. At a density of 50 Klett units, 5 μg/ml CLT was added. One aliquot was immediately induced for the resolution assay, then at various later time points aliquots were removed and shifted to 42°C for resolution assays (Fig. 5). Each time point in Fig. 5 marks the elapsed time from addition of CLT to the end of the 42°C induction period. Resolution frequencies changed 5-fold from 50% to 10% during a 20-min window of growth and induction with CLT.

Fig. 5.

Barrier appearance and disappearance in response to chlortetracycline. □, overnight cultures of NH3492 subcultured in LB without chlortetracycline. γδ resolution assays were done when cell density reached 50 Klett units. ▪, 5 μg/ml chlortetracycline added at a cell density of 50 Klett units. Samples were taken at various points after addition of chlortetracycline and subjected to γδ resolution assay. ○, NH3492 grown in LB with 5 μg/ml chlortetracycline. γδ resolution assays were initiated when cell density reached 50 Klett units. •, chlortetracycline removed by quickly washing and then resuspending cells in LB. Samples taken at various times were tested for γδ resolution.

The kinetics of barrier disappearance was also examined. An overnight culture of NH3492 was diluted 100-fold into LB plus 5 μg/ml CLT. At a density of 50 Klett units, cells were harvested and resuspended in CLT-free medium, and 10-min shifts to 42°C were done in the same manner described above. Resolution efficiency increased from 10% to 50% after CLT removal. Barriers appeared and disappeared with similar kinetics (Fig. 5).

Discussion

Transcription changes the dynamic behavior of plasmid DNA in several ways. First, RNA polymerase unwinds the double helix as it transcribes RNA. Induction of a moderate-strength promoter can add six new RNA polymerase molecules to a small plasmid, which changes the average linking number by 10 negative supercoils (31). Second, transcription causes a temporary segregation of positive supercoils in front of a transcribed gene and negative supercoils in the transcription wake (32–34). In strains carrying a topA mutation, expression of a membrane-inserted protein can result in a dramatic increase in negative supercoil of plasmids (14, 15, 21). Both of these effects are visible because rapid plasmid extraction freezes the population in its current topological state, and agarose gel electrophoresis can detect subtle changes in plasmid topology (35). Two effects of transcription have been noted in the large bacterial chromosome. First, if a highly transcribed operon like rrnB, which encodes ribosomal RNAs, is positioned head to head with a replication fork, transcription markedly delays the time required for forks to cross the transcription unit (36). Second, induction of the strong phage Mu early promoter causes a new supercoil barrier to appear in the vicinity of the transcription unit (9).

In this work we found three effects of transcription on supercoil dynamics in the S. typhimurium genome. First, constitutive tetA transcription caused a persistent 20-fold reduction in γδ site-specific recombination over a 14-kb interval by (Fig. 2). The effect was larger for longer intervals, with an impact of a 40-fold reduction for a 28-kb domain. Second, transcription effects were not restricted membrane proteins. Similar effects on γδ resolution efficiency were found with aminoglycoside-3′-O-phosphotransferase, which is the product of the Tn5 kan gene and β-gal, the product of Escherichia coli lacZ. Barriers were detected by using a WT resolvase in combination with modules that either lacked a functional copy of the TetR repressor, or modules with WT TetR protein but in the presence of the inducer CLT. The critical determinant was transcription strength. Even though the eight genes separating the res sites in strain NH3470 include five membrane-associated proteins, the level of transcription of these genes, estimated by microarray analysis of total-cell RNA (data not shown), was too low to disrupt supercoil diffusion. At the atp operon, consisting of nine genes encoding membrane-associated proteins, and the nmpc operon, which regulates a single gene, the mRNA abundance measured from microarray analysis accurately predicted the impact on γδ resolution (Fig. 8, which is published as supporting information on the PNAS web site). Third, by using a modified form of γδ resolvase with a cellular half-life of 5 min, barriers to supercoil diffusion appeared and disappeared within 10–20 min of the point when CLT was either added or washed out of cell cultures. The impact of transcription over a 14-kb interval was correlated with the level of transcription over a 40-fold range of expression measured with β-gal assays.

Because transcription alters chromosome dynamics of both plasmid and the bacterial chromosomal DNA, the question arises as to how many transcription-related barriers exist in a bacterial cell. Microarray analyses provide important information on genome-wide transcription profiles (37, 38). Cells growing under the same conditions used in our experiments (exponential cultures in LB) give detectable RNA signals for >4,000 E. coli genes (38). However, steady-state mRNA abundance exhibits a strikingly skewed distribution. For 70% of the genes (3,034 ORFs) mRNA abundance is less than one molecule per cell. Genes in this category include the uninduced RNA for lacZ (which yields ≈15 Miller units of β-gal activity) and 69 of the 321 known essential genes. Genes in this low-transcription category probably have little impact on DNA dynamics. The second class is genes with a steady-state abundance of one to four copies of RNA/DNA, which results in 40–160 Miller units of β-gal activity. Genes in this class (≈1,000) include 86 essential genes, and they are also predicted to have a modest topological impact. Fig. 4 suggests that over a 14-kb interval these genes would decrease resolution by 30% or less. The third gene class includes genes with RNA/DNA ratios >4. Included in this category would be a lacZ gene induced with 1 mM isopropyl β-d-thiogalactoside, which causes a 40- to 60-fold derepression of transcription to ≈600 Miller units. Protein-coding genes in E. coli with RNA/DNA ratios of 10 or greater represent 85 genes. The predicted impact on chromosome domains of these top 85 genes is given in Table 4, which is published as supporting information on the PNAS web site. The abundance of lacZ mRNA in uninduced cultures yields an RNA/DNA ratio of 0.45 (38), so a culture with 600 Miller units would have an RNA/ DNA ratio of ≈20. This expression change is nearly identical with the change in levels of β-gal expression we observe in S. typhimurium going from a repressed P_RtetR-P_AlacZ culture to steady-state expression in the presence of 5 μg/ml CLT (Fig. 4). Extrapolating from the plot in Fig. 4, the number of chromosomal sites with diffusion barriers that would inhibit resolution by 5-fold or more would be <8. If we add the seven ribosomal operons, we predict that only ≈15 sites would show persistent 5-fold effects on resolution. This number of barriers is small compared with our estimate of 300 barriers that are stochastic and associated with DNA replication.

Several implications of these results are noteworthy. First, because transcription of many genes, including a significant fraction of the essential genes, is intermittent in many parts of the chromosome, domain structure is likely to be both dynamic and variable. Recent studies show that cell-to-cell expression variation is high for the lac operon at low inducer levels (39, 40). In specific instances, a burst of transcription can lead to a large amount of protein production in some cells and none at all in most cells. Considering the fact that operons like araBAD respond in similar fashion (41), no two organisms in a bacterial culture may ever have the same genome organization. This dissimilarity makes chromosome structure behave like a quasispecies, which is a collection of organisms that vary in structure around a common mean (42). Second, the genome behavior of the highly expressed operons may have special mechanisms to insulate their transcriptional effects from surrounding chromosome regions. The highly transcribed genes may be under selection for interspersion. Examples are known where chromosomal inversions involving specific inverting sites are disfavored, whereas other inversions using the same end points are permitted (43, 44). Finding the rules governing such behavior has been difficult. It is possible that apposition of two chromosomal segments with high-transcription activity may be unstable, just as placing a replication origin in front of a strong transcription unit leads to an impediment to replication (36). Third, recent reports in both bacteria and yeast show that high levels of transcription can cause formation of persistent RNA:DNA hybrids (R-loops) that interfere with both transcription and replication (22, 45, 46). Whether R-loop formation is responsible for some examples of chromosomal domain behavior is a difficult question, but it is one well worth pursuing.