Gene order and chromosome dynamics coordinate spatiotemporal gene expression during the bacterial growth cycle

Patrick Sobetzkoa, Andrew Traversb,c, and Georgi Muskhelishvilia,1

a School of Engineering and Science, Jacobs University Bremen, D-28759 Bremen, Germany; bMedical Research Council Laboratory of Molecular Biology, Cambridge CB2 0QH, United Kingdom; and cFondation Pierre-Gilles de Gennes pour la Recherche, Laboratoire de Biologie et Pharmacologie Applique, Ecole Normale Suprieure de Cachan, 94235 Cachan, France

Edited by Sankar Adhya, National Cancer Institute, National Institutes of Health, Bethesda, MD, and approved November 23, 2011 (received for review May 23, 2011)

In Escherichia coli crosstalk between DNA supercoiling, nucleoid-associated proteins and major RNA polymerase initiation factors regulates growth phase-dependent gene transcription. We show that the highly conserved spatial ordering of relevant genes along the chromosomal replichores largely corresponds both to their temporal expression patterns during growth and to an inferred gradient of DNA superhelical density from the origin to the terminus. Genes implicated in similar functions are related mainly in trans across the chromosomal replichores, whereas DNA-binding transcriptional regulators interact predominantly with targets in cis along the replichores. We also demonstrate that macrodomains (the individual structural partitions of the chromosome) are regulated differently. We infer that spatial and temporal variation of DNA superhelicity during the growth cycle coordinates oxygen and nutrient availability with global chromosome structure, thus providing a mechanistic insight into how the organization of a complete bacterial chromosome encodes a spatiotemporal program integrating DNA replication and global gene expression.
gene order conservation protein gradients

selection of mutations in s and tRNA dihydrouridine synthase (dusB) (essential for s expression) and also in topA (31), as well as in rpoC (the subunit of RNAP) under conditions of adaptive evolution (32). Although there is substantial evidence for integrated regulation of NAPs, DNA superhelicity, and RNAP selectivity during the growth cycle, the mechanism by which this regulation is accomplished remains obscure. We report here that the conserved ordering of the stage-specic regulatory genes and their targets along the replichores corresponds with their temporal expression patterns during the growth cycle. We propose that this ordering reects a gradient of DNA gyrase-binding sites and hence negative superhelicity from chromosomal origin (OriC) to terminus (Ter) of replication and that the generation of this superhelicity gradient is coupled to energy availability. During the growth cycle changes in local superhelicity drive morphological changes in chromosome structure that facilitate the integration of DNA replication and gene expression. Results
Gene Order. We observed that the ordering, relative to OriC, of

| transcriptional regulatory network |

n Escherichia coli cells the physiological transitions induced by the changing growth environment are accompanied by changes in DNA superhelical density (13), nucleoid structure (46), and the promoter selectivity of the RNA polymerase (RNAP) holoenzyme (3, 7). During the growth cycle both the relative and absolute concentrations of the abundant nucleoid-associated proteins (NAPs; Table S1) change substantially and correspondingly generate bacterial chromatin of variable composition (8, 9). The NAPs stabilize distinct supercoil structures (1012) selectively favoring particular RNAP holoenzymes (1315). These variable nucleoprotein complexes modulate DNA topology during the growth cycle (Fig. 1A), optimizing the channeling of supercoil energy into appropriate metabolic pathways (16, 17). The expression of the genes determining superhelical density, polymerase selectivity, and nucleoid structure is coordinated by cross-regulation. Thus, factor for inversion stimulation (FIS), a NAP abundant during the early exponential phase (18), regulates expression not only of the superhelicity determinants DNA gyrase subunits A and B (gyrA and gyrB) and topoisomerase I (topA) (19 21) but also other NAP-encoding genes including hns, subunit of histone-like protein from E. coli strain U93 (hupA), and DNA binding protein from starved cells (dps) (2224) and components of the transcription machinery such as 38 subunit of RNA polymerase rpoS (25). Similarly mutations affecting the selectivity of RNAP inuence NAP production (2628). Again, mutations in the genes controlling DNA superhelicity affect the production of both NAPs and the basal transcription machinery (27, 29). This pattern of integrated control constitutes a heterarchical network coordinating chromosome structure with cellular metabolism (27, 28, 30). A further pointer to this integrated network is the observed
E42E50 | PNAS | January 10, 2012 | vol. 109 | no. 2

the genes encoding the NAPs specic to particular stages of the growth cycle [i.e., s, hupA, subunit of histone-like protein from E. coli strain U93 (hupB), suppression of td phenotype (stpA), lrp, dps, cbpA, and subunit of integration host factor (ihfB)] approximately reects their relative abundance during the growth cycle (8, 9). With one exception, hns, the NAPs associated with the higher overall superhelicity characteristic of exponential growth are closer to the origin. Conversely, those associated with the lower superhelical density characteristic of the stationary phase are closer to the replication terminus (Fig. 1B). The ordering of the genes for the transcriptional-machinery components exhibits a pattern similar to that of the NAPs, with rpoD, encoding the 70 factor for vegetative growth, located closer to OriC than rpoS, encoding the stationary phase S factor. Similarly, gyrB (but not gyrA), encoding subunit B of DNA gyrase, is located in close proximity to OriC. Gyrase increases negative superhelicity, especially with the higher ATP/ADP ratios prevailing on nutritional shift-up (33). In contrast, both topA and topoisomerase III (topB), encoding the DNA-relaxing topoisomerases, are closer to the

Author contributions: A.T. and G.M. designed research; P.S. performed research; P.S., A.T., and G.M. analyzed data; and P.S., A.T., and G.M. wrote the paper. The authors declare no conict of interest. This article is a PNAS Direct Submission. Freely available online through the PNAS open access option.
1

To whom correspondence should be addressed. E-mail: g.muskhelishvili@jacobs-university. de.

See Author Summary on page 355. This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10. 1073/pnas.1108229109/-/DCSupplemental.

www.pnas.org/cgi/doi/10.1073/pnas.1108229109

A
NAP abundance

RNA polymerase sigma factors Plasmid superhelical density ( ) -0.068 Growth stage Shift up Early log Mid/late log

-0.043 Transition (ppGpp spike) Stationary

B Chromosomal
macrodomains Aerobic/anaerobic metabolism, DNA replication, rrn genes, etc DNA topology

Right OriC Left

Ori Ori C A B E NS

NS

Right Left

Ter Ter Ter

arcA arcB D yacG parC parE

H G

seqA

rmf

fnr Ter topA Ter

OriC atp dnaA OriC gyrB hupA

gyrA hupB stpA

sbmC

topB hns Ter ihfA

hfq crp fis

dps ihfB lrp cbpA

NAPs

OriC rsd nusG rpoBC rho OriC rpoZ

RNA polymerase modulators

fecI

Ter greB rpoD ssrS rpoH nusA rpoA rpoN greA rpoS rpoE fliA

Fig. 1. Spatiotemporal organization of chromosomal expression. (A) Temporal changes of NAPs, RNAP composition, and average plasmid DNA superhelicity () during bacterial growth. The phases of growth cycle are correlated with preferred expression of particular NAPs, RNAP holoenzymes and the supercoiling temporal gradient are indicated below. A transient increase in guanosine tetraphosphate (ppGpp) levels occurs at the transition between the exponential and stationary phases. (B) Spatial ordering of regulatory genes on the E. coli chromosome along the OriCTer axis. (Top line) Correspondence of macrodomains dened by Valens et al. (40) to linear map. (First bar) Selected genes involved in aerobic/anaerobic metabolism (dark blue), DNA replication (orange), rrn genes (red), and transition phase (brown). Genes on the clockwise (right) replichore are above the bar, and genes on anti-clockwise (left) replichore are below the bar. The atp operon encodes ATP synthase. arcA/arcB encode a two-component system active under microaerobic conditions (61, 62). ArcA also represses rpoS (63). fnr has a dominant role under more strictly anaerobic conditions (61). dnaA, encoding the principal initiator of DNA replication, maps close to OriC, whereas seqA, an inhibitor of replication initiation at OriC, maps closer to Ter. rmf decreases the availability of ribosomes and maps to a macrodomain immediately adjacent to the Ter macrodomain. (Second bar) Selected genes involved in control of DNA topology. gyrB, a component of DNA gyrase responsible for increasing negative superhelicity, maps close to OriC, whereas the gyrase inhibitor susceptibility to B17 microcin, locus C (sbmC), and topA and topB, both responsible for relaxing DNA, map either close to or within the Ter macrodomain. DNA gyrase inhibitor (yacG), encoding an inhibitor of GyrB, maps close to the center. Chromosomal partition genes C and E (parC and parE) encode the subunits of topoisomerase IV, responsible for decatenation of newly replicated DNA in the terminal region (64) and relaxation of negative supercoils (65). (Third bar) Selected genes encoding NAPs. The NAP-encoding gene closest to OriC is hupA, encoding histone-like protein from E. coli strain U93 (HU). Its early expression relative to hupB, encoding Hu (9), could buffer high negative superhelicity generated by DNA gyrase (36). HU2 and HU, but not HU2, constrain high superhelical densities in vitro (9). A mutation in hupA both increases growth rate and antagonizes histone-like nucleoid-structuring protein (H-NS) regulation of certain transcription units (6). High frequency of recombination (Hfq) is a nucleic acid-binding protein whose major role is that of an RNA chaperone, but it also may act as a DNA-binding NAP (8). lrp is activated by ppGpp (38). (Fourth bar) Selected genes involved in modulating RNAP activity, including factor-utilization regulators, secondary channel-binding proteins, termination/elongation factors, and RNAP subunits. factor-utilization regulators (light green): subunit of RNA polymerase (rpoZ), mapping close to the origin, encodes the subunit of RNAP, which confers a preference for utilization of 70 (28). Regulator of sigma D (rsd) encodes an anti-70 (66), whereas crl confers a strong preference for S utilization (67). Note that both rpoZ and crl map closer to OriC than do the respective factors whose activity they affect. The encoded regulatory pattern thus reects a shift from predominantly 70 use close to OriC to S availability in the central region of the chromosome. Secondary channel-binding proteins (plum): growth regulator A and B, transcription elongation factors (greA and greB) both map in the region containing many genes expressed during rapid growth. GreA has been shown to stimulate initiation and transcription of genes involved in aerobic metabolism, including the atp operon (68, 69) as well as the rrnB P1 promoter in vitro (70, but also see ref. 71). DksA, like the plasmid-encoded quorum sensing regulator (TraR) protein (72), inhibits ribosomal protein promoters and rrn initiation (73, 74) and is more distant from OriC than is greA. In vivo it would act to reduce the rate of rrn initiation and hence antagonize transcription foci formation. Termination/elongation factors (red): The termination factor Rho is encoded by a gene located very close to OriC. This location may compensate for the antagonistic effect of high negative superhelicity on transcription termination, which involves the rewinding of DNA. RNAP subunits: The map positions relative to OriC of rpoD and rpoS, respectively encoding 70 and S, correspond to their relative order of temporal expression.

Ter. Spatial organization of genes modifying the transcription machinery [transcriptional terminator Rho (rho), the N utilization substance gene (nus) factors, the gre factors, dnaK suppressor
Sobetzko et al.

(dksA), curly (crl)] and genes sustaining catabolism and energy production under aerobic [ATP synthase (atp) operon], microaerobic [two component signal transduction system (arcA/B)], and
PNAS | January 10, 2012 | vol. 109 | no. 2 | E43

GENETICS

crl dksA

PNAS PLUS

such genes from the origin and terminus was highly conserved (Fig. 2A). Similar analysis of all orthologous genes demonstrated that, although the relative distance was conserved, the specic replichore was not conserved to the same extent (Fig. 2B). Simulations supported this observed bias (Fig. S1). Furthermore, essentially the same picture emerged when we analyzed the more distantly related Gram-positive bacteria (Fig. S2 A and B). We conclude that conservation of relative distance along the OriCTer axis overrides the replichore coherence.
Targets. Not only are the genes coordinating the major regulatory pathways ordered; their targets are ordered as well. Analyses of the distribution density of binding sites for E70 and ES holoenzymes compiled in RegulonDB (34) show opposite spatial biases. For the vegetative 70 factor the highest percentage of targets is found around the origin, whereas for the stationary phase S factor the highest target density is close to the terminus (Fig. 3 A and B), consistent with both the closer location of rpoD to OriC and the temporal division of labor between 70 and S during the bacterial growth cycle (7). Similarly, the average density of binding sites for DNA gyrase (35) diminishes by ve- to 10-fold (Fig. 3C) from OriC to Ter (36). This organization could generate a gradient of superhelical density (Fig. 3D) correlating with that of E70 targets and anti-correlating with ES targets, as expected from the opposite supercoiling preferences of these holoenzymes (3, 28) and in keeping with the requirement of high negative superhelicity for initiation of OriC replication (37). For the major NAPs, despite distinct chromosomal locations and abundances during the growth cycle, a high percentage of the binding sites compiled in RegulonDB occurs around the origin (Fig. 3 EH). Among the NAPs encoded in the Terproximal region, only integration host factor (IHF) targets activating binding sites in the vicinity of Ter (Fig. 3F), whereas the stationary-phase regulator leucine responsive protein (LRP) (38) and the global repressor H-NS (39) both preferentially target the OriC-proximal region (Fig. 3 G and H). Additionally HU, the major supercoil-constraining NAP for which no binding site information is available, has distinct and opposite functional effects at the OriC and Ter ends of the chromosome, respectively reducing and increasing transcription (Fig. S3). To explore the relevance of this apparent chromosomal ordering of regulators for chromosomal expression, we used the Gene Ontology (GO) database describing the gene products in terms of associated metabolic processes and investigated the spatial organization of functional groups of genes. Many of these genes would be the ultimate targets of the regulation. Using scanning windows of variable sizes (0.10.5 Mb), we mapped the genes in the GO database, identifying the signicant matches between functionally complementing windows on the chromosome (Figs. S4 and S5). We determined the ratio of all signicant combinations for cis (matching windows located on the same replichore) and trans (matching windows on distinct replichores) arrangements. Although the GO tree organization comprises 13 complexity levels from the most broad down to specic functions, most signicant matches were observed at levels three and four (Figs. S6 and S7). Fig. 4A shows that functionally matching groups are localized predominantly on opposite replichores within the rrn macrodomain (36) comprising OriC and the nonstructured left (LNS) and right (RNS) macrodomains (40). Importantly, the predominantly trans organization of these groups was signicant only when the chromosomal gene order was aligned along the OriCTer axis. We infer that a majority of the functionally related groups are organized at comparable distances from the center of symmetry at OriC and are related in trans across the replichores. We denote these spatially coordinated matching GO groups of genes as maGOGs. We next analyzed the interactions between the E. coli DNAbinding transcriptional regulators and their targets compiled in the RegulonDB in the form of a static transcriptional regulatory
Sobetzko et al.

Fig. 2. Arrangement of important regulatory elements according to their position relative to OriC in the -Proteobacteria. (A) Genes located on the right and left replichores are indicated above and below the chart, respectively, as in Fig.1B. Horizontal bars show spatial distributions of orthologs; black color indicates the highest density for each individual distribution. The plot shows a strong conservation of chromosomal positions suggesting that origin-focused gene positioning is a major selection criterion in -Proteobacteria. (B) Relationship among the phylogenetic distance, the correlation of distance to origin, and the replichore coherence (the conserved replichore identity of orthologs) for all -Proteobacteria. The points represent the data on pair-wise species comparisons computed using the Pearson correlation coefcient of either distances to origin or replichore identity (right/left) of all orthologous pairs. The points are color-coded in the 3D plot. Red indicates a higher correlation of distance to origin than replichore coherence; blue indicates higher replichore coherence. The predominance of red points indicates the stronger conservation of distance to origin. The phylogenetic distances in B were derived from the tree of -Proteobacteria (http://www.cbrg.ethz.ch/research/orthologous/speciestrees).

anaerobic (fnr) conditions, as well as those involved in activation and negative modulation of replication [DNA replication initiator protein DnaA (dnaA) and sequestration of origin (seqA)] and translation [ribosomal RNA (rrn) operons and ribosome maturation factor (rmf)], exhibit a similar pattern of chromosomal ordering (legend of Fig. 1B and Table S2). We asked whether the gene order of regulatory elements associated with the temporal pattern of transcription in Escherichia coli was conserved in other -Proteobacteria. Analysis of 131 -proteobacterial genomes showed that the relative distance of
E44 | www.pnas.org/cgi/doi/10.1073/pnas.1108229109

H
GENETICS

Fig. 3. Organization of binding sites for the major initiation factors, DNA, gyrase, and NAPs in the chromosome (RegulonDB). Distributions were calculated by using a sliding window of 400 kb and normalizing over the total gene number for each window. The replichores are organized from OriC to Ter (left to right). The frequency distributions (ordinate) are plotted above the zero in the ordinate for the right replichore and below the zero for the left replichore. (A) Genomic distributions for RNAP70-regulated promoters. (B) Genomic distributions for S-regulated promoters. (C) Genomic distributions of gyrase-binding sites. (D) Inferred gradient of negative superhelical density along the OriCTer axis in exponentially growing cells. (E) Genomic distribution of FIS-binding sites. (F) Genomic distribution of IHF-binding sites. (G) Genomic distribution of H-NSbinding sites. (H) Genomic distribution of LRP-binding sites. The chromosomal position for each regulator gene is indicated, and the direction of the effect is rainbow color-coded with blue for repression and red for activation. The Ori, Ter, left, and right macrodomains (green, cyan, blue, and red lines, respectively) are indicated on the chromosomal replichores above and below the distributions.

network (TRN). These analyses again revealed macrodomains in which spatially coordinated interactions between the regulators and targets were signicantly enriched. However, in contrast to the spatial organization of maGOGs, signicant TRN interactions occur mainly in cis along the replichores, largely corresponding to intermacrodomain communication (Fig. 4B). We also analyzed the spatial organization of couplons, entities corresponding to intersections of regulons of distinct factors and NAPs and containing functionally related genes (27). Couplon analyses again demonstrated preferential communications in trans across the replichores (Fig. 4C). We infer that chromosomal arrangement of functionally related genes and the spatial communications between transcriptional regulators and their targets are essentially orthogonal with respect to each other.
Expression Proles. Any valid regulatory network must describe meaningfully the changes in gene expression both during normal growth and when expression is perturbed by mutation. The
Sobetzko et al.

analyses shown in Figs. 3 and 4 use the static E. coli Regulon database, which contains neither temporal nor spatial ordering information. To analyze the dynamic organization of communications in the effective transcript proles (41) obtained under dened conditions, we integrated the maGOGs, TRN, and couplon networks in a single combined heterarchical network (HEN). In such a network, as opposed to a hierarchical network, there is no subordination to one single dominant component. The HEN (Fig. 4D) served as a template for mapping the effective transcript proles. The observed gene ordering predicted that relative transcription during exponential and stationary growth phases would be ordered in a similar spatial manner relative to OriC and Ter. The results fully conrmed this prediction, clearly demonstrating a division of labor between the OriC and Ter chromosomal ends in temporal expression proles. In exponentially growing wild-type cells the OriC-proximal region was transcriptionally more active than the Ter-proximal region with the active regions corresponding largely to the rrn domain
PNAS | January 10, 2012 | vol. 109 | no. 2 | E45

PNAS PLUS

Fig. 4. Organization of functional groups and regulatory communications in the E. coli chromosome. (A) maGOGs. (B) TRN. (C) Couplons. (D) HEN. The circular genome is represented as a pair of multicolored parallel lines corresponding to the right (Upper line) and left (Lower line) replichores. On the replichores the macrodomains (colored as in Fig. 3) and the rrn functional domain (orange dashed line) are indicated. All trans communications occur between the upper and lower lines, whereas cis communications occur along the lines. The order of regulatory genes on the right and left replichores is indicated above and below each network, respectively, organized from OriC to Ter.

comprising the Ori and LNS macrodomains (Fig. 5 A and B). Similar spatiotemporal division of labor was observed in an rpoS mutant lacking S, but the pattern relative to wild-type cells was reversed, with the Ter-proximal region being more favored during the exponential phase and the OriC-proximal region during the stationary phase (Fig. 5 C and D). Also, in an rpoZ mutant favoring the ES holoenzyme (28), the Ter-proximal region was favored during the exponential phase, an effect that could be reversed by overproduction of 70 activating the OriC end (Fig. 5 E
E46 | www.pnas.org/cgi/doi/10.1073/pnas.1108229109

and F). Importantly, although the effects of factors mainly involved trans communications, consistent with the orthogonal organization of regulatory communications in HEN, the mutations of s and hns genes also substantially affected the cis communications, which were largely buffered in wild-type cells (Fig. 5 G L). Furthermore, some of these spatiotemporal patterns could be closely imitated by manipulating the composition of NAPs and DNA superhelicity (Fig. S8), validating the usefulness of HEN for gene-expression analyses.
Sobetzko et al.

Fig. 5. Spatiotemporal HEN patterns. The circular genome is represented as a pair of thin parallel lines corresponding to the right (Upper line) and left (Lower line) replichores as in Fig. 4. Macrodomains are indicated in color on the right (Upper) and left (Lower) replichores organized from OriC to Ter. (A) Effective HEN (eHEN) pattern of exponentially growing wild-type cells. (B) eHEN pattern of transcripts from stationary wild-type cells. Red indicates coherently enhanced and blue indicates coherently reduced numbers of expressed genes in matching windows; black indicates an absence of coherence. Note the switch in the activation of the OriC and Ter ends of the chromosome at the transition from exponential to stationary growth. (C) eHEN pattern of rpoS cells lacking S obtained during exponential phase. (D) eHEN pattern of rpoS cells in stationary phase. Note the spatiotemporal inversion of the activation of the OriC and Ter ends of the chromosome with respect to wild-type cells. (E) eHEN pattern of exponentially growing rpoZ cells favoring ES. (F) As in E, but the cells were overproducing 70 from an episome (28). Note the 70-dependent coherent activation and repression of communications in the OriC and Ter ends of the chromosome, respectively. (G and H) eHEN patterns of wild-type cells grown under conditions of high and low superhelicity, respectively. (I and J) Patterns of hns-mutant cells grown under conditions of high and low superhelicity, respectively. Note the supercoiling-dependent coherent activation and repression of communications in the left and right replichores, respectively. (K and L) eHEN patterns of s-mutant cells grown under conditions of high and low superhelicity, respectively. Note the coherent activation of distinct supercoiling-dependent cis and trans communications. In AD the cells were grown in minimal medium. In EL the cells were harvested during exponential growth in rich double-YT medium. (M and N) Model of chromosomal morphology changing with DNA superhelicity and NAP gradients on transition from exponential (M) to stationary (N) growth. For clarity, only the interactions between the gradients of FIS (pink) and H-NS (blue) are shown. Although FIS levels decline dramatically on transition to stationary phase, the compaction of the nucleoid along the OriCTer axis enables H-NS to establish repression. The chromosome is depicted as a plectoneme, but the model is equally consistent with a toroidal scaffold, which would maintain the separation of the replichores (49, 50). The macrodomains are indicated by colors, and approximate chromosomal positions of the s and hns genes are shown. The expression data for mapping onto the HEN connectivity patterns were taken from Dong and Schellhorn (59) in AD, from Geertz et al. (28) in E and F, and from Blot et al. (16) in GL.

Discussion We have demonstrated that during the E. coli growth cycle in batch culture the temporal pattern of stage-specic gene expression corresponds largely with the gene order along the two replichores. This correspondence is apparent not only for the principal regulatory genes but also for their targets. Importantly, this property is a highly conserved characteristic of the replichores in both Gram-negative -Proteobacteria and Gram-positive bacteria, implying that the ordering is related to DNA replication. Moreover, the most conserved property for a particular gene appears to be not the precise location but the relative distance of the gene from OriC and Ter, independent of replichore (Fig. 2 and Figs. S1 and S2 A and B). This nding suggests that the selective determinant is itself a variable property that also is dependent on the relative OriC/Ter distance and is consistent with macrodomains sharing many functions (Fig. S9). In fast-growing E. coli the bidirectional replication starting from OriC creates a gradient of gene dosage from OriC to Ter, such that the resulting origin-proximal relative increase in gene dosage potentially could increase gene expression in that region. Indeed, it is precisely this region that is preferentially active during exponential growth (Fig. 5A). However, variation in gene dosage is likely to be only one of a number of regulatory inuences. Thus, the originproximal region contains a ve- to 10-fold higher average density of gyrase-binding sites than the Ter-proximal region (Fig. 3C) and also is the preferred target region for E70 holoenzyme and for FIS protein (Fig. 3 A and E), implicated in evolutionary modulation of DNA superhelicity (25). We infer that the gradient of gyrase binding sites is indicative of potential gradients of negative superhelical density from OriC to Ter along the replichores (Fig. 3D). The existence of such gradients corresponds well with both the pattern of holoenzyme and NAP targets and the greater density of supercoiling-sensitive genes (e.g., rpoZ, s, hupA, and rrn) in the OriC end (16, 36, 42), as well as with the activation of the rrn macrodomain under conditions of high negative superhelicity during exponential growth (Fig. S10 A and B). The gradient of gyrase-binding sites would provide a simple mechanism for coupling energy availability to superhelical density (2, 33) and, because initiation of DNA replication is enhanced by high negative superhelicity (37), also to replication itself. Although the spatial order of selected stage-specic genes corresponds well to the temporal order of expression during the growth cycle, the position of certain other genes, although highly conserved, does not. Genes in this category in E. coli include hns and gyrA. Although in Gram-positive bacteria gyrA and gyrB map in close proximity to OriC (Fig. S2C), a rationale for the separate conserved locations of gyrA and gyrB in -Proteobacteria is not obvious. On shift-up expression of gyrB increases to a proportionally much greater extent than expression of gyrA (19). GyrA and GyrB together form a heterotetramer, and if GyrB were limiting the distinct locations would favor the use of the gyrasebinding sites in the origin-proximal region. We hope to clarify any requirement for separate placement of the gyrA and gyrB genes in E. coli by switching their chromosomal positions. Both FIS and H-NS principally target the origin-proximal region, potentially delimiting short topological domains (43). However, s is expressed at maximal levels during the exponential phase and maps in the origin-proximal region, in contrast to hns, which maps in the Ter macrodomain. In this context Montero Llopis et al. (44) recently reported that both Escherichia coli and Caulobacter crescentus spatially organize translation so that the mRNA product of a gene is translated in close proximity to its position in the nucleoid. This observation has profound implications for gene regulation and chromosome structure. If NAP production is localized, then diffusion of the resultant DNAbinding proteins within the nucleoid will be anomalous, generating concentration gradients. The outcome for nucleoid organization at
PNAS | January 10, 2012 | vol. 109 | no. 2 | E47

Sobetzko et al.

GENETICS

PNAS PLUS

a given locus then depends on the relative local concentrations of different NAPs, which in turn depend on the distance from the site of production and the total number of molecules available. In this model FIS and H-NS would be expected to be more dominant on nucleoid structure and function in the origin-proximal region during the exponential and stationary phases, respectively. The relation of NAP function to negative superhelicity is consistent with the concept that NAPs can act to buffer superhelical density not only at the local level, as previously reported for the rrnA P1 promoter (45), but also at the more global level of the chromosome and macrodomains (Fig. 5 GL). In this respect a correlation between the gradient of gyrase-binding sites and the functional effect of HU (36) is particularly striking (Fig. S3). In the context of the transcriptional effect of HU, the apparently anomalous positions of the rrnG and rrnH genes approximately halfway between OriC and Ter delimit the rrn functional macrodomain (36). Importantly, although all seven rRNA operons have been shown to be necessary for rapid adaptation, only ve are necessary to support near-optimal growth (46), and it is not known how the rebalancing of rRNA transcription after shift-up is distributed among the rrn operons. Also, in a strain lacking ppGpp (the negative regulator of rRNA transcription), the regions containing up-regulated genes whose expression is stimulated by high negative superhelicity are in close proximity to rrnG and rrnH (Fig. S10 C and D). This nding together with the propensity for forming variable transcription foci (47) raises the possibility that, in growing cells with a full complement of rrn operons, the regulation of individual rrn operons depends on their position in the chromosome. A related issue is how the scattered distribution of tRNA genes around the chromosome can be reconciled with chromosomal positioning (Fig. S11). Unlike rRNA genes, the expression of tRNA genes must be consistent both with growth and with differences in the relative amounts of iso-accepting tRNAs and codon use at different growth rates (ref. 48 and legend of Fig. S11). For both rRNA and iso-accepting tRNA genes, as for other genes described here, distance from the origin is more conserved than replichore (Fig. S11B).
Topological Model for Temporal Gene Expression. The deduced logic of spatial communications in the E. coli chromosome suggests a simple topological model of the circular chromosome folded as a negative supercoil in which close spatial proximity of replichores either within the rrn macrodomain comprising the Ori and anking exible macrodomains or between the rrn and Ter macrodomains supports alternative communication during exponential and stationary growth. These morphological transitions could occur as a consequence of changes in superhelical density, which would affect the conguration of DNA directly and also would alter the relative afnities of the different NAPs for DNA. The alternative congurations, with varying OriC and Ter separation, are consistent with the formation of transcription foci (47), the reorganization and physical extension of nucleoids during exponential growth (4951), and the compaction of nucleoids in the stationary phase (51, 52). This last effect is counteracted by the activator of rRNA operons FIS, which is abundant in exponential phase (4, 18), suggesting that the conformational transition in the nucleoid is associated with changing gradients of competing NAPs (Fig. 5 M and N). Although the extent to which gene expression has a linear relationship with gene copy number is unknown, it is conceivable that the relative increase in gene dosage resulting from the initiation of replication would increase the competitive advantage of regulators encoded in the origin-proximal region, whereas a balanced OriC-to-Ter stoichiometry on cessation of growth would abolish such an advantage (Fig. S12). How could such morphological transitions be coupled to energy availability? A primary response to nutritional shift-up under aerobic conditions is an increase in the ATP/ADP ratio (2, 33) that activates DNA gyrase, favoring the DnaAATP complex
E48 | www.pnas.org/cgi/doi/10.1073/pnas.1108229109

active in replication initiation (53) and coordinating RNAP activity with energy availability (54). We propose that gyrase activity creates a gradient of superhelical density corresponding to the gradient of gyrase-binding sites (Fig. 3 C and D and Fig. S10 A and B) such that the selectively increased superhelicity of the origin-proximal region and the resultant change in chromosome morphology precede the initiation of replication. Such selectivity suggests a lower superhelical density in the Ter region, consistent with the nding that in the prereplicative state in slow-growing cells the Ter region is extended (55). Concomitantly the increased spatial separation between origin-proximal H-NS target genes and the Ter-proximal location of the hns gene (Fig. 5M) would attenuate the silencing effect H-NS (39, 56, 57) on the rrn macrodomain. On transition to stationary phase. a dramatic decline of FIS levels (18) in conjunction with DNA relaxation (Fig. S10 A and B) and compaction of the nucleoid would enable H-NS, as well as the stationary-phase regulator LRP (38), to reestablish efcient repression (Fig. 5N) in the vicinity of replication origin. This model provides a dynamic mechanism for the concerted impact of the NAPs and DNA superhelicity on chromosome structure and cellular physiology (46, 9, 10, 16, 17, 19, 36, 39, 58). Conclusion The conservation of gene order and spatial organization of regulatortarget interactions in chromosomal macrodomains reported in this paper argue that the gene order species the temporal pattern of gene expression during the bacterial growth cycle. Our work thus illuminates why the timing of the expression of a gene is linked to its position on the chromosome, providing a view of the bacterial nucleoid as a highly organized dynamic entity optimized for coordinating oxygen and nutrient availability with spatiotemporal gene expression during rapid growth and its cessation. Materials and Methods
Identication of Functional Groups in the GO Tree and Modeling of the TFGs. To address the similarity of chromosomal regions, we applied a sliding-window approach using a 500-kb window size and a 4-kb window shift creating 1,160 windows, each containing about 500 genes. All genes within a window were assigned to their specic locations on the GO tree (http://www.geneontology. org/). For a xed level of the GO tree we summed up all genes assigned to the specic subtree, determining a unique one-dimensional pattern of GOsubtree (cluster) sizes for each window. Subsequently the similarity score of two windows (i and j) was determined by the equation: ni;k nj;k si;j ni;k nj;k
k1 k1 k1 c c c

where c is the number of clusters on the current GO-tree level, and n is the number of genes in a certain branch of this level, reecting their covariance with a mean that equals zero. Hence, high similarity scores (s values) indicate a high correlation of the functional composition. All s values were compared with scores of 10,000 random genomes, generated by a random remapping of gene (operon) IDs to gene (operon) positions. For subsequent analyses window pairs with a P value < 0.05 were considered signicant, and all others, including scores of overlapping windows [distance(i,j) <500 kb], were excluded. However, higher signicance values up to a false-discovery rate of 0.05 indicated no qualitative change. In practice, the exclusion of overlapping windows introduces a bias toward trans matches. To address this issue and to compile a reference data set with the same bias, we shifted the origin position along the chromosome and compared cis/trans ratios with the native origin location. Subsequently, the signicance of the cis/trans ratio peaks for a shifted origin position was compared with 10,000 random gene sets with the same number of best-scoring matches to rule out random noise peaks. GO levels three and four turned out to be statistically signicant; lower GO levels lacked a reasonable functional resolution, whereas higher levels represented an extremely detailed functional resolution that

Sobetzko et al.

suffered from varying resolutions among distinct functional clusters because of a nonuniform depth of the GO tree. Modeling of the TRN. In correspondence with the GO-tree function analysis, we parsed the TRN communication along the chromosome by counting the number of connections between the subnetworks of any two 500-kb windows normalized by the total number of connections spreading from both windows into the genome. Using procedures analogous to GO similarity analysis, we determined highly connected regions on the chromosome and determined the peaks for the cis/trans ratio by shifting the origin. Couplon Similarity. As a measure of couplon similarity, we determined for each couplon (27) whether the number of contained genes for the current window was greater or less than the expected number of genes. Hence, the similarity of two windows equals the number of couplons coherently overor underrepresented in both windows. The determination of the signicance of matches and cis/trans ratio peaks was carried out as for the GO-tree similarity with random sets of both genes and operons. Regulator Targets. To determine target site frequency distributions, the relevant data derived from the RegulonDB database were analyzed using a sliding 400-kb window with the map position at the window midpoint.

Expression Patterns. The expression data for mapping onto the HEN connectivity patterns were taken from Dong and Schellhorn (59) in Fig. 5 AD, from Geertz et al. (28) in Fig. 5 E and F, and from Blot et al. (16) in Fig. 5 GL. Phylogenetic Analysis. In the phylum of -Proteobacteria the majority of regulators is conserved in a plethora of species including human and plant pathogens such as Vibrio cholerae and Pseudomonas syringae. In total we investigated 131 species (using a reciprocal best blast hit approach at the protein sequence level followed by the determination of orthologous clusters by a modied GirvanNewman algorithm, resulting in densely connected clusters (node degree n/2), where n denotes the cluster size. Species information containing protein sequences was derived from the RefSeq sequence database and origin positions using the DoriC database (60). The phylogenetic distances in Fig. 2B were derived from the tree of -Proteobacteria (http://www.cbrg. ethz.ch/research/orthologous/speciestrees). The same methodology was applied for analysis of all the annotated genomes of Gram-positive bacteria. ACKNOWLEDGMENTS. The authors thank Malcolm Buckle and Sylvie Rimsky for fruitful discussions. This work was supported by grants from the Ecole Normale Suprieure de Cachan and Deutsche Forschungsgemeinschaft (to G.M.). A.T. received a Chaire dExcellence award from lAgence Nationale de la Recherche.

1. Balke VL, Gralla JD (1987) Changes in the linking number of supercoiled DNA accompany growth transitions in Escherichia coli. J Bacteriol 169:44994506. 2. Hsieh LS, Burger RM, Drlica K (1991) Bacterial DNA supercoiling and [ATP]/[ADP]. Changes associated with a transition to anaerobic growth. J Mol Biol 219:443450. 3. Bordes P, et al. (2003) DNA supercoiling contributes to disconnect sigmaS accumulation from sigmaS-dependent transcription in Escherichia coli. Mol Microbiol 48:561571. 4. Ohniwa RL, et al. (2006) Dynamic state of DNA topology is essential for genome condensation in bacteria. EMBO J 25:55915602. 5. Rimsky S, Travers A (2011) Pervasive regulation of nucleoid structure and function by nucleoid-associated proteins. Curr Opin Microbiol 14:136141. 6. Kar S, Edgar R, Adhya S (2005) Nucleoid remodeling by an altered HU protein: Reorganization of the transcription program. Proc Natl Acad Sci USA 102: 1639716402. 7. Ishihama A (2000) Functional modulation of Escherichia coli RNA polymerase. Annu Rev Microbiol 54:499518. 8. Ali Azam T, Iwata A, Nishimura A, Ueda S, Ishihama A (1999) Growth phasedependent variation in protein composition of the Escherichia coli nucleoid. J Bacteriol 181:63616370. 9. Claret L, Rouvire-Yaniv J (1997) Variation in HU composition during growth of Escherichia coli: The heterodimer is required for long term survival. J Mol Biol 273: 93104. 10. Dame RT (2005) The role of nucleoid-associated proteins in the organization and compaction of bacterial chromatin. Mol Microbiol 56:858870. 11. Guo F, Adhya S (2007) Spiral structure of Escherichia coli HUalphabeta provides foundation for DNA supercoiling. Proc Natl Acad Sci USA 104:43094314. 12. Maurer S, Fritz J, Muskhelishvili G (2009) A systematic in vitro study of nucleoprotein complexes formed by bacterial nucleoid-associated proteins revealing novel types of DNA organization. J Mol Biol 387:12611276. 13. Colland F, Barth M, Hengge-Aronis R, Kolb A (2000) Sigma factor selectivity of Escherichia coli RNA polymerase: Role for CRP, IHF and lrp transcription factors. EMBO J 19:30283037. 14. Shin M, et al. (2005) DNA looping-mediated repression by histone-like protein H-NS: Specic requirement of Esigma70 as a cofactor for looping. Genes Dev 19:23882398. 15. Maurer S, Fritz J, Muskhelishvili G, Travers A (2006) RNA polymerase and an activator form discrete subcomplexes in a transcription initiation complex. EMBO J 25: 37843790. 16. Blot N, Mavathur R, Geertz M, Travers A, Muskhelishvili G (2006) Homeostatic regulation of supercoiling sensitivity coordinates transcription of the bacterial genome. EMBO Rep 7:710715. 17. Sonnenschein N, Geertz M, Muskhelishvili G, Htt MT (2011) Analog regulation of metabolic demand. BMC Syst Biol 5:40. 18. Ball CA, Osuna R, Ferguson KC, Johnson RC (1992) Dramatic changes in Fis levels upon nutrient upshift in Escherichia coli. J Bacteriol 174:80438056. 19. Schneider R, Travers A, Kutateladze T, Muskhelishvili G (1999) A DNA architectural protein couples cellular physiology and DNA topology in Escherichia coli. Mol Microbiol 34:953964. 20. Keane OM, Dorman CJ (2003) The gyr genes of Salmonella enterica serovar Typhimurium are repressed by the factor for inversion stimulation, Fis. Mol Genet Genomics 270:5665. 21. Weinstein-Fischer D, Altuvia S (2007) Differential regulation of Escherichia coli topoisomerase I by Fis. Mol Microbiol 63:11311144. 22. Claret L, Rouvire-Yaniv J (1996) Regulation of HU and HU by CRP and FIS in Escherichia coli. J Mol Biol 263:126139. 23. Falconi M, Brandi A, La Teana A, Gualerzi CO, Pon CL (1996) Antagonistic involvement of FIS and H-NS proteins in the transcriptional control of hns expression. Mol Microbiol 19:965975.

24. Grainger DC, Goldberg MD, Lee DJ, Busby SJ (2008) Selective repression by Fis and HNS at the Escherichia coli dps promoter. Mol Microbiol 68:13661377. 25. Hirsch M, Elliott T (2005) Fis regulates transcriptional induction of RpoS in Salmonella enterica. J Bacteriol 187:15681580. 26. Zhou YN, Jin DJ (1998) The rpoB mutants destabilizing initiation complexes at stringently controlled promoters behave like stringent RNA polymerases in Escherichia coli. Proc Natl Acad Sci USA 95:29082913. 27. Muskhelishvili G, Sobetzko P, Geertz M, Berger M (2010) General organisational principles of the transcriptional regulation system: A tree or a circle? Mol Biosyst 6: 662676. 28. Geertz M, et al. (2011) Structural coupling between RNA polymerase composition and DNA supercoiling in coordinating transcription: A global role for the omega subunit? mBio 2, 4; e00034-11. 29. Peter BJ, et al. (2004) Genomic transcriptional response to loss of chromosomal supercoiling in Escherichia coli. Genome Biol 5:R87. 30. Travers A, Muskhelishvili G (2005) DNA supercoiling - a global transcriptional regulator for enterobacterial growth? Nat Rev Microbiol 3:157169. 31. Crozat E, et al. (2010) Parallel genetic and phenotypic evolution of DNA superhelicity in experimental populations of Escherichia coli. Mol Biol Evol 27:21132128. 32. Conrad TM, et al. (2010) RNA polymerase mutants found through adaptive evolution reprogram Escherichia coli for optimal growth in minimal media. Proc Natl Acad Sci USA 107:2050020505. 33. van Workum M, et al. (1996) DNA supercoiling depends on the phosphorylation potential in Escherichia coli. Mol Microbiol 20:351360. 34. Gama-Castro S, et al. (2008) RegulonDB (version 6.0): Gene regulation model of Escherichia coli K-12 beyond transcription, active (experimental) annotated promoters and Textpresso navigation. Nucleic Acids Res 36(Database issue):D120D124. 35. Jeong KS, Ahn J, Khodursky AB (2004) Spatial patterns of transcriptional activity in the chromosome of Escherichia coli. Genome Biol 5:R86. 36. Berger M, et al. (2010) Coordination of genomic structure and transcription by the main bacterial nucleoid-associated protein HU. EMBO Rep 11:5964. 37. Fuller RS, Kornberg A (1983) Puried dnaA protein in initiation of replication at the Escherichia coli chromosomal origin of replication. Proc Natl Acad Sci USA 80: 58175821. 38. Tani TH, Khodursky A, Blumenthal RM, Brown PO, Matthews RG (2002) Adaptation to famine: A family of stationary-phase genes revealed by microarray analysis. Proc Natl Acad Sci USA 99:1347113476. 39. Dorman CJ (2004) H-NS: A universal regulator for a dynamic genome. Nat Rev Microbiol 2:391400. 40. Valens M, Penaud S, Rossignol M, Cornet F, Boccard F (2004) Macrodomain organization of the Escherichia coli chromosome. EMBO J 23:43304341. 41. Marr C, Geertz M, Htt MT, Muskhelishvili G (2008) Dissecting the logical types of network control in gene expression proles. BMC Syst Biol 2:18. 42. Ferrndiz MJ, Martn-Galiano AJ, Schvartzman JB, de la Campa AG (2010) The genome of Streptococcus pneumoniae is organized in topology-reacting gene clusters. Nucleic Acids Res 38:35703581. 43. Hardy CD, Cozzarelli NR (2005) A genetic selection for supercoiling mutants of Escherichia coli reveals proteins implicated in chromosome structure. Mol Microbiol 57:16361652. 44. Montero Llopis P, et al. (2010) Spatial organization of the ow of genetic information in bacteria. Nature 466:7781. 45. Rochman M, Aviv M, Glaser G, Muskhelishvili G (2002) Promoter protection by a transcription factor acting as a local topological homeostat. EMBO Rep 3:355360. 46. Condon C, Liveris D, Squires C, Schwartz I, Squires CL (1995) rRNA operon multiplicity in Escherichia coli and the physiological implications of rrn inactivation. J Bacteriol 177:41524156.

Sobetzko et al.

PNAS | January 10, 2012 | vol. 109 | no. 2 | E49

GENETICS

PNAS PLUS

47. Cabrera JE, Jin DJ (2003) The distribution of RNA polymerase in Escherichia coli is dynamic and sensitive to environmental cues. Mol Microbiol 50:14931505. 48. Dong H, Nilsson L, Kurland CG (1996) Co-variation of tRNA abundance and codon usage in Escherichia coli at different growth rates. J Mol Biol 260:649663. 49. Lau IF, et al. (2003) Spatial and temporal organization of replicating Escherichia coli chromosomes. Mol Microbiol 49:731743. 50. Adachi S, Fukushima T, Hiraga S (2008) Dynamic events of sister chromosomes in the cell cycle of Escherichia coli. Genes Cells 13:181197. 51. Kim J, et al. (2004) Fundamental structural units of the Escherichia coli nucleoid revealed by atomic force microscopy. Nucleic Acids Res 32:19821992. 52. Frenkiel-Krispin D, et al. (2004) Nucleoid restructuring in stationary-state bacteria. Mol Microbiol 51:395405. 53. Olliver A, Saggioro C, Herrick J, Sclavi B (2010) DnaA-ATP acts as a molecular switch to control levels of ribonucleotide reductase expression in Escherichia coli. Mol Microbiol 76:15551571. 54. Travers AA, Kari C, Mace HAF (1981) Transcriptional regulation by bacterial RNA polymerase. Genetics as a Tool in Microbiology, SGM Symposia, eds Glover S, Hopwood D (Cambridge Univ Press, Cambridge, UK), Vol XXXI, pp 111130. 55. Wiggins PA, Cheveralls KC, Martin JS, Lintner R, Kondev J (2010) Strong intranucleoid interactions organize the Escherichia coli chromosome into a nucleoid lament. Proc Natl Acad Sci USA 107:49914995. 56. Aferbach H, Schrder O, Wagner R (1998) Effects of the Escherichia coli DNAbinding protein H-NS on rRNA synthesis in vivo. Mol Microbiol 28:641653. 57. Spurio R, et al. (1992) Lethal overproduction of the Escherichia coli nucleoid protein H-NS: Ultramicroscopic and molecular autopsy. Mol Gen Genet 231:201211. 58. Scolari VF, Bassetti B, Sclavi B, Lagomarsino MC (2011) Gene clusters reecting macrodomain structure respond to nucleoid perturbations. Mol Biosyst 7:878888. 59. Dong T, Schellhorn HE (2009) Control of RpoS in global gene expression of Escherichia coli in minimal media. Mol Genet Genomics 281:1933. 60. Gao F, Zhang CT (2007) DoriC: A database of oriC regions in bacterial genomes. Bioinformatics 23:18661867. 61. Alexeeva S, Hellingwerf KJ, Teixeira de Mattos MJ (2003) Requirement of ArcA for redox regulation in Escherichia coli under microaerobic but not anaerobic or aerobic conditions. J Bacteriol 185:204209.

62. Levanon SS, San KY, Bennett GN (2005) Effect of oxygen on the Escherichia coli ArcA and FNR regulation systems and metabolic responses. Biotechnol Bioeng 89:556564. 63. Mika F, Hengge R (2005) A two-component phosphotransfer network involving ArcB, ArcA, and RssB coordinates synthesis and proteolysis of sigmaS (RpoS) in E. coli. Genes Dev 19:27702781. 64. Zechiedrich EL, Khodursky AB, Cozzarelli NR (1997) Topoisomerase IV, not gyrase, decatenates products of site-specic recombination in Escherichia coli. Genes Dev 11: 25802592. 65. Zechiedrich EL, et al. (2000) Roles of topoisomerases in maintaining steady-state DNA supercoiling in Escherichia coli. J Biol Chem 275:81038113. 66. Jishage M, Ishihama A (1998) A stationary phase protein in Escherichia coli with binding activity to the major sigma subunit of RNA polymerase. Proc Natl Acad Sci USA 95:49534958. 67. Robbe-Saule V, et al. (2006) Physiological effects of crl in Salmonella are modulated by SigmaS and promoter activity. J Bacteriol 189:29762987. 68. Stepanova E, et al. (2007) Analysis of promoter targets for Escherichia coli transcription elongation factor GreA in vivo and in vitro. J Bacteriol 189:87728785. 69. Susa M, Kubori T, Shimamoto N (2006) A pathway branching in transcription initiation in Escherichia coli. Mol Microbiol 59:18071817. 70. Potrykus K, et al. (2006) Antagonistic regulation of Escherichia coli ribosomal RNA rrnB P1 promoter activity by GreA and DksA. J Biol Chem 281:1523815248. 71. Rutherford ST, et al. (2007) Effects of DksA, GreA, and GreB on transcription initiation: Insights into the mechanisms of factors that bind in the secondary channel of RNA polymerase. J Mol Biol 366:12431257. 72. Blankschien MD, et al. (2009) TraR, a homolog of a RNAP secondary channel interactor, modulates transcription. PLoS Genet 5:e1000345. 73. Paul BJ, et al. (2004) DksA: A critical component of the transcription initiation machinery that potentiates the regulation of rRNA promoters by ppGpp and the initiating NTP. Cell 118:311322. 74. Lemke JJ, et al. (2011) Direct regulation of Escherichia coli ribosomal protein promoters by the transcription factors ppGpp and DksA. Proc Natl Acad Sci USA 108: 57125717.

E50 | www.pnas.org/cgi/doi/10.1073/pnas.1108229109

Sobetzko et al.