Regulation of cellular growth implies spatiotemporally coordinated programmes of gene transcription. A central question, therefore, is how global transcription is coordinated in the genome. The growth of the unicellular organism Escherichia coli is associated with changes in both the global superhelicity modulated by cellular topoisomerase activity and the relative proportions of the abundant DNA‐architectural chromatin proteins. Using a DNA‐microarray‐based approach that combines mutations in the genes of two important chromatin proteins with induced changes of DNA superhelicity, we demonstrate that genomic transcription is tightly associated with the spatial distribution of supercoiling sensitivity, which in turn depends on chromatin proteins. We further demonstrate that essential metabolic pathways involved in the maintenance of growth respond distinctly to changes of superhelicity. We infer that a homeostatic mechanism organizing the supercoiling sensitivity is coordinating the growth‐phase‐dependent transcription of the genome.
Understanding the mechanisms of concerted rearrangements of gene activities during growth and development is a fundamental problem. How is genomic expression organized in a cell and is there a common coordinating mechanism? Almost five decades ago, John von Neumann proposed that the flow of genetic information is coordinated by a specific relationship between the ‘digital’ (discontinuous) properties of unique genes and ‘analog’ (continuous) properties of the gene products (von Neumann, 1958). Such a relationship evidently develops during pattern formation in Drosophila embryogenesis, in which the essentially ‘analog’ information provided by concentration gradients of transcription factors is converted into ‘digital’ patterns of transcripts by means of differential protein interactions occurring on spatially separated transcriptional enhancers (Sauer et al, 1996). Most clearly, this fundamental principle can be demonstrated in Escherichia coli, a classical model organism that shows strict correlations between the global gene expression patterns and different growth states (Tao et al, 1999; Wei et al, 2001; Weber et al, 2005) and also enables the modulation of genomic transcription by alterations of global DNA superhelicity (Jeong et al, 2004; Peter et al, 2004; Willenbrock & Ussery, 2004; Travers & Muskhelishvili, 2005a). Such alterations are associated with both growth transitions and stress responses to environmental challenge, supporting the idea that DNA supercoiling itself might act as a principal coordinator of global gene expression (Balke & Gralla, 1987; Dorman, 1996; Tse‐Dinh et al, 1997; Cheung et al, 2003; Travers & Muskhelishvili, 2005b). Nevertheless, as less than 8% of specific genes are found to respond to supercoiling in E. coli (Peter et al, 2004), the role of superhelicity in organizing the global growth‐phase‐dependent transcription remains obscure.
Previous studies proposed that binding of abundant chromatin proteins could selectively direct the supercoiling energy to gene promoters (Travers & Muskhelishvili, 1998; Hatfield & Benham, 2002; Muskhelishvili & Travers, 2003). In this study, we investigated the transcriptional effects of two such proteins, factor for inversion stimulation (FIS) and histone‐like nucleoid structuring protein (H‐NS), which act as global pleiotropic regulators in E. coli. H‐NS can activate transcription but is predominantly a universal repressor for the bacterial genome, whereas FIS modulates the transcription of many genes implicated in regulation of metabolism and growth (Gonzalez‐Gil et al, 1996; Dorman, 2004; Kelly et al, 2004; Rimsky, 2004). Previously, it was shown that the loss of H‐NS and FIS affects the superhelical density of plasmid DNA (Owen‐Hughes et al, 1992; Schneider et al, 1997). By combining the effects of mutations in the fis and hns genes with experimentally induced changes of global superhelicity, we demonstrate here that the organization of global transcription is tightly coupled to distribution of supercoiling sensitivity in the genome.
Results And Discussion
FIS and H‐NS are present in several tens of thousands of copies per cell and can both activate and repress many genes by direct binding (Dorman & Deighan, 2003). However, FIS forms a concentration gradient decreasing by two orders of magnitude on transition to the stationary phase, whereas the concentration of H‐NS remains relatively constant (Ball et al, 1992; Azam & Ishihama, 1999). As both these regulators modulate supercoiling, we expected to observe both direct and indirect effects on global transcription during growth. We used a novel DNA microarray‐based strategy to link the growth‐phase‐dependent transcription of genes with supercoiling sensitivity. In one set of experiments, we compared the growth‐phase‐dependent transcript profiles of the wild type with both the fis and hns mutants to distinguish between the genes expressed in the presence or absence of each regulator (Fig 1A, data set A). In the second set of experiments (data set B), we used genetically engineered E. coli LZ41 and LZ54 strains containing drug‐resistant topoisomerase gene alleles to inhibit DNA gyrase or topoisomerase IV activity selectively and thereby induce relaxation or high negative supercoiling, respectively (Zechiedrich et al, 1997). On introduction of the fis and hns mutations in the LZ41 and LZ54 strains, the global supercoiling response to drug addition was not substantially altered (σ=−0.082 for LZ54 wild type and −0.086 for both LZ54fis and LZ54hns mutants; σ=<−0.031 for both LZ41 wild type and mutants; supplementary Fig 2 online). Thus, using distinct combinations of the LZ41 and LZ54 strains and their mutant derivatives, we could vary superhelical density to distinguish gene transcripts associated with either relaxation (Rel) or high negative supercoiling (Hyp) in each genetic background (Fig 1A, Exp.1; the term ‘associated’ emphasizes the difficulty of identifying whether these transcripts are activated by a given supercoiling regime or inactivated by the opposite one) and also those either activated (Act) or repressed (Rep) by each regulator at various superhelicities (Exp.2; we note that the regulator effects can be either direct or indirect). These ‘supercoiling‐associated’ transcripts were then mapped on the growth‐phase‐dependent transcript profiles of data set A to evaluate the input of supercoiling sensitivity in the organization of global transcription.
Supercoiling sensitivity of global transcription
The results of mapping are presented on genomic wheels in Fig 1B,C. About 50% of all transcripts could be mapped (supplementary Table III online), indicating that these variable patterns are largely organized by supercoiling‐associated genes. Notably, out of a total of 1,596 Hyp and Rel transcripts identified, only 145 were common to the wild‐type, fis and hns backgrounds, indicating that these supercoiling response depends on the genetic context. By using fis and hns mutants, we thus greatly expanded the repertoire of supercoiling‐associated transcripts identified. The observed interspersed pattern of the domains of relaxation and high negative superhelicity agrees with both the formation of fluid short‐range domains of distinct topologies in the E. coli chromosome (Postow et al, 2004; Stein et al, 2005) and the complex dependence of long‐range transcriptional patterns on supercoiling (Jeong et al, 2004). This supports the idea that transcription is governed by ‘effective’ superhelicity—a parameter reflecting the dynamic nature of competition between topoisomerases, transcription machinery and chromatin proteins for available supercoils (Travers & Muskhelishvili, 2005b). Variable distributions of supercoiling sensitivity reflect a compound effect of several cooperating factors in addition to FIS and H‐NS, including all chromatin architectural proteins, more dedicated transcriptional regulators and perhaps the chromosome partitioning proteins (Malik et al, 1996; Sawitzke & Austin, 2000; Beloin et al, 2003; Frenkiel‐Krispin et al, 2004). However, our data show that the lack of a single key component of bacterial chromatin—either FIS or H‐NS—can reorganize the supercoiling sensitivity of genomic transcription, consistent with the previously proposed global homeostatic feedback (Schneider et al, 2000). As expected from such a mechanism, all strains grew with comparable rates, despite remarkable variations of transcript patterns (supplementary Fig 1 online). Nevertheless, during prolonged growth, mutations in fis can confer a competitive advantage or disadvantage depending on the precise environment (Schneider et al, 1997; Crozat et al, 2005).
Variations of supercoiling‐associated transcripts
Distributions of supercoiling‐associated transcripts in the wild‐type, fis and hns cells varied with growth. The exponentially growing fis mutant showed an increase of both the Hyp and Rel transcripts compared with those in wild type, whereas in the hns mutant the relative proportion of Hyp transcripts increased, in keeping with in the observed increase of global negative superhelicity (Fig 1D; supplementary Fig 2B,C online). Furthermore, on transition to the stationary phase, a lower proportion of Rel genes were observed in the fis mutant than in the wild type, but not in the stationary phase itself. As expected, the effect of hns mutation was most pronounced for the Rep genes. In the fis mutant, the Rep genes increased only during the exponential phase, whereas in the hns mutant, the Rep genes predominated during the entire growth cycle. Also the ‘double‐coded’ transcripts responding to specific combinations of supercoiling and genetic background were enriched for Rep/Hyp and Rep/Rel genes, especially in hns cells. Taken together, these observations are in excellent agreement with differential abundance of FIS and H‐NS during the growth cycle and the general repressor role of H‐NS (Ball et al, 1992; Dorman, 2004).
Supercoiling sensitivity is coupled to metabolic function
Interestingly, the analysis of distribution of Hyp and Rel genes among the functional groups involved in essential cellular metabolism demonstrated a higher proportion of Hyp genes in the anabolic than in the catabolic pathways, which was especially remarkable in the wild‐type cells (Table 1). Analysis of a pathway of exceptional importance for vitality—the citric acid cycle—demonstrated that the crucial steps producing combustible fuel in the form of reducing equivalents NADH and FADH2, which are required for generation of ATP by oxidative phosphorylation, are associated with DNA relaxation. In contrast, when we examined the glyoxalate bypass, topping up the cycle and increasing the net synthesis of carbohydrate, we found this pathway to be associated with high negative superhelicity (Fig 2A; supplementary Table II online). This suggests that high ATP/ADP ratios favouring supercoiling of DNA by gyrase would facilitate glyoxalate bypass and plastic substrate synthesis, whereas DNA relaxation would favour the production of ATP required to maintain gyrase activity. Furthermore, we found that the de novo pathway of nucleotide biosynthesis explicitly involves the Hyp genes (Fig 2B). Most of these genes have promoters with a GC‐rich ‘discriminator’ sequence, which confers sensitivity to both high negative supercoiling and the regulatory nucleotide ppGpp (Cashel, 1970; Zalkin & Nygaard, 1996; Figueroa‐Bossi et al, 1998). Notably, both the superhelicity and ppGpp concentration are elevated in fis cells (Travers & Muskhelishvili, 2005b), indicating a homeostatic regulation mechanism. The observed coupling of distinct supercoiling sensitivities to essential metabolic pathways provides new insights into the mechanisms that coordinate central metabolism, and also sheds light on the puzzling observation that mutations in metabolic genes can affect DNA topology (Hardy & Cozzarelli, 2005). An important future task is the clarification of mechanistic effects of changes in superhelicity on transcription, which could modulate not only transcription initiation but also elongation and termination.
Our mapped transcript profiles describe the distributions of supercoiling sensitivity during growth, rather than the strength of transcriptional response to superhelicity and are consistent with the proposed role of FIS and H‐NS in forming topological barriers on the chromosome (Hardy & Cozzarelli, 2005; Dame, 2005). However, Hardy & Cozzarelli failed to detect any alterations of plasmid supercoiling in fis and hns mutants, most probably because they did not analyse the entire growth cycle, which is necessary to show the dynamic changes of superhelicity (supplementary Fig 2B,C online; Schneider et al, 1997). Despite substantial differences in the experimental design and assay conditions, we identified the supercoiling sensitivity of previously described genes (Peter et al, 2004). Also, the functional gene groups discovered in this study are consistent with those reported for a fis mutant of Salmonella typhimurium and hns mutant of E. coli (Hommais et al, 2001; Kelly et al, 2004).
We thus infer that the coordination of growth‐phase‐dependent transcription involves a homeostatic mechanism that organizes the supercoiling sensitivity in the genome. As proposed in our previous work (Schneider et al, 1999, 2000), this feedback implicates chromatin proteins that constrain DNA supercoils and act as ‘optimizers’ of cellular metabolism. Whereas the metabolic status determines the overall supercoiling level (van Workum et al, 1996), we show here that the genomic distributions of superhelicity can specify the patterns of transcripts during growth. Global regulation thus seems to be a genuine device converting the ‘analog’ information of torsional energy distributions into ‘digital’ patterns of responding genes. Our observations provide a holistic conceptual framework for analysis of global transcription and reinforce the notion of John von Neumann of coordination of information flow in the genetic system. We note that DNA supercoiling is implicated in the regulation of eukaryotic transcription (Mizutani et al, 1991; Dunaway & Ostrander, 1993; Caserta & Di Mauro, 1996; Kouzine et al, 2004). Thus, studies of regulated alterations of DNA superhelicity might be essential for understanding the coordinated gene functions in eukaryotes as well.
DNA microarray analysis. The E. coli K12 strains used in this study for RNA isolation are described elsewhere (Zechiedrich et al, 1997). The fis and hns mutants of LZ54 and LZ41 strains were obtained by P1 phage transduction. The transcript profiling for the LZ strains was carried out after brief (5 min) treatment of cells growing exponentially in 2 × YT medium at 30°C with moderate concentrations of norfloxacin (20 μg/ml). All other strains were grown in 2 × YT medium at 37°C. DNA microarray experiments were performed according to OciChipTM‐Application Guide (http://www.ocimumbio.com) as two biological replicates with two technical replicates each (except for the transition state of CSH50fis and the stationary phase of CSH50hns strains; supplementary information online). Scanned array images were quantified and normalized using the TM4 software package. A one‐class t‐test was applied to replicated experiments to obtain genes with significant P‐values (P<0.05). Further details are provided in the supplementary information online.
Real‐time PCR. QuantiTect® SYBR® Green one‐step Real‐Time PCR reactions (Qiagen GmbH, Hilden, Germany) were performed in triplicate, following the manual of the manufacturer and using an Mx3000P™ Real‐Time cycler (Stratagene®, La Jolla, CA, USA).
Supplementary information is available at EMBO reports online (http://www.emboreports.org).
We thank L. Steil, U. Völker and S. Wegener for help with preliminary experiments, F.‐O. Glöckner and C. Würdemann for providing facilities for microarray analyses, M. Keiser for useful comments and C. Burau for excellent technical assistance. This work was supported by the Deutsche Forschungsgemeinschaft through grant DFG‐MU‐2FIS to G.M.Array Express accession number: I‐TABM‐86.
- Copyright © 2006 European Molecular Biology Organization