Galactose repressor mediated intersegmental chromosomal connections in Escherichia coli

Abstract

By microscopic analysis of fluorescent-labeled GalR, a regulon-specific transcription factor in Escherichia coli, we observed that GalR is present in the cell as aggregates (one to three fluorescent foci per cell) in nongrowing cells. To investigate whether these foci represent GalR-mediated association of some of the GalR specific DNA binding sites (gal operators), we used the chromosome conformation capture (3C) method in vivo. Our 3C data demonstrate that, in stationary phase cells, many of the operators distributed around the chromosome are interacted. By the use of atomic force microscopy, we showed that the observed remote chromosomal interconnections occur by direct interactions between DNA-bound GalR not involving any other factors. Mini plasmid DNA circles with three or five operators positioned at defined loci showed GalR-dependent loops of expected sizes of the intervening DNA segments. Our findings provide unique evidence that a transcription factor participates in organizing the chromosome in a three-dimensional structure. We believe that these chromosomal connections increase local concentration of GalR for coordinating the regulation of widely separated target genes, and organize the chromosome structure in space, thereby likely contributing to chromosome compaction.

The genes involved in d-galactose metabolism and regulation with their cognate promoters constitute a regulon, which is regulated by Gal repressor (GalR) in Escherichia coli. Purified GalR is a homodimer of a 37-kDa subunit (1). GalR is also known to form oligomers and higher-order structures, which give rise to paracrystals, at less than 0.2 M salt concentrations both in the absence and presence of DNA (1). GalR represses gene transcription by binding to specific operator DNA sequences that are associated with the gene promoters (2). There are at least five known promoters each associated with one or two GalR binding operators, located at 17.0 (galE operon), 48.2 (mgl operon), 48.2 (galS), 64.1 (galR), and 66.5 (galP) min on the chromosome. So far only three of the cognate promoters—of galE, galS, and galP—contain two operators (2). We have previously shown that GalR bound to the two operators, which encompass two promoters, P₁ and P₂, in the galE operon, associates to form a DNA loop (3). Whereas simple DNA binding represses the P₁ promoter, only DNA looping represses the P₂ promoter. We proposed that the GalR (dimers) bound to the regulon operators located around the chromosome associate in some order giving rise to a specific 3D network of d-galactose metabolism-related genes to better coordinate the regulation of the functionally related promoters and to maintain higher local concentrations of GalR around the operators,. The interactions may also help compaction of the chromosome. The most likely way to bring distant regulatory loci together is through interactions between DNA-bound proteins. Here we demonstrate, by both in vivo and in vitro methods, that operator-bound GalR located around the chromosome interact with each other. We observed that there are greater interactions in nongrowing cells than in growing cells. We used fluorescent Venus labeled GalR to trace location of GalR in cell (4), Chromosome Conformation Capture (3C) analysis to determine the aggregation of distally located DNA-bound GalR in vivo (5), and atomic force microscopy (AFM) to visualize DNA-bound GalR–GalR interactions in vitro (6, 7). The implication of the results is discussed.

Results

Fluorescence Microscopy Analysis of GalR-Venus.

Elf et al. demonstrated location of the LacI repressor protein bound to its DNA target in the lac operon (4). This was accomplished by using a fusion of the LacI protein to a modified rapidly maturing YFP fluorescent protein (Venus), and observing the cells under a fluorescent microscope. We used an anologous approach to find out the location of DNA-bound GalR around the chromosome. We genetically fused the Venus gene sequence to a single copy chromosomal GalR gene at the carboxy-end to generate galR-venus (Fig. 1A). Any potential effect of the fusion of Venus to GalR on the GalR expression level was examined by Western blot. As shown in Fig. S1, the expression level of the fusion protein under various growth conditions was almost identical. We also tested whether the fusion would affect the binding of GalR to its target DNA. The results showed that the gene-regulatory DNA-binding activity of the GalR-Venus fusion on the regulation of two promoters of the gal operon was the same as that of WT GalR (8) (Fig. S2). We concluded that the fusion of Venus did not fundamentally alter the normal GalR property. Moreover, a single dimer of GalR-Venus fusion bound to DNA in the cell cannot be observed under the conditions used, as opposed to findings by Xie et al, who observed single molecules of LacI-Venus fusion protein (5) using an EMCCD camera with their microscope set-up that is reportedly capable of detecting single photons, at the expense of resolution (5). They greatly increased the exposure time and calculated the midpoint of the fluorescence signal to assign “enhanced localization” for each focus. In our setup, exposure times were ∼1 s.

Fig. 1.

Live cell fluorescence microscopic analysis of Venus labeled GalR. (A) Schematic presentation of the galR locus and the genetic layout of the construction of the chromosomal galR-venus containing strain in which the venus is fused to the C-terminal of GalR. Kanamycin cassette was used as the selection marker. (B) Typical DIC and fluorescence images of GalR, GalR-Venus, and GalR^T322R-Venus strains grown to stationary phase (Left) and log phase (Right). All cells used were grown in M63 minimal media with fructose and casamino acids. Five images for each strain are shown. (C) Statistical analysis of fluorescent foci per cell in GalR-Venus strain cultured to stationary phase. A total of 284 and 292 cells were counted for cells growing in minimal medium without and with galactose, respectively.

Without a single-photon sensitive microscope, we hoped to detect clusters comprising multiple molecules of fluorescent GalR in our microscope if GalR molecules aggregate. As shown in Fig. 1B, when the stationary phase GalR-Venus cells were observed under a fluorescence microscope, 277 of 284 counted cells grown in minimal medium displayed 1–3 distinguishable fluorescent foci. The distribution of the number of foci per cell is shown in Fig. 1C. We rarely observed cells with four or more foci. At the resolution limits of our setup, we conclude that the foci represent diffraction-limited localization events. Of course, independent localizations that are closer than 200 nm apart will appear as a single focus in our diffraction-limited setup. As mentioned above, all of our fluorescence experiments were done with GalR-Venus fusion located at the chromosome in normal position that generates ∼60–70 dimers in the absence of d-galactose under the conditions used (37 °C) as determined by Western blot. Under the conditions used we cannot resolve whether each spot has many subnucleoid entities. There are ∼100 GalR dimers present in E. coli cells grown in minimal medium (9). It is likely that DNA-bound GalR associate to generate few GalR aggregates.

We have previously characterized a GalR mutant (GalR^T322R), which normally binds to the operators but inefficient in forming tetra- and higher-order oligomers (3). We generated a GalR^T322R-Venus mutant and first tested its gene-regulatory activity the same way as the WT (Fig. S2). The results showed that although GalR^T322R-Venus repressed the P₁ promoter of the gal operon by binding to a single cognate operator locus O_E, it is defective in tetramerization-mediated repression of the P₂ promoter of the gal operon by DNA looping. Therefore, if GalR-Venus foci are caused by GalR multimerization, then we do not expect to observe fluorescent foci in cells carrying the GalR^T322R-Venus fusion protein and grown under the same conditions. Consistently, we observed fluorescent foci in only four of 185 cells that contained the mutant GalR fusion. These results support the hypothesis that the foci observed with GalR-Venus were generated by association of two or more GalR dimers presumably bound to DNA. Furthermore, when log phase cells carrying the WT GalR-Venus fusion protein were inspected, the fluorescent foci were significantly reduced, caused by fast DNA replication fork of the chromosome presumably interfering with the intrachromosomal contacts (Fig. 1B). Moreover, we also investigated the effect of d-galactose (the inducer of gal operon) on the observed foci. As shown in Fig. S3, both in stationary phase and log phase, no significant difference could be found between the cells treated with and without d-galactose. The distribution of the number of foci per cell cultured to stationary phase with d-galactose is shown in Fig. 1C. Because d-galactose does not effectively induce the gal operon located at 17.0 min of the chromosome in stationary phase cells (10), d-galactose may not break GalR-mediated bridges detected here under similar conditions. Alternatively, the GalR-mediated intrachromosomal links in stationary phase cells may have more complex structure/compositions to be sensitive to d-galactose. The fluorescence data strongly suggest that the GalR binding sites along the chromosome are brought together through GalR polymerization. In addition, Xie et al reported that the signal intensities for single foci that they measured occurred in discrete quanta, suggesting that these investigators were measuring single molecules of fluorophores (5). As the intensities of our foci do not appear to occur in such discrete quanta, we are not concluding that we are detecting single molecules at each focus.

Intersegmental Chromosomal Associations.

The intersegmental chromosomal connections by DNA-bound GalR as suggested by the above fluorescent labeling of GalR experiment can be tested in vivo by the 3C method or its refinements (5). Such techniques have been used in chromosomal 3D structure analysis in yeast and human (11–13). Fig. S4 shows the principle of the 3C method as adapted from Dekker et al. (5). First, we performed the 3C analysis to test whether the operators of GalR are physically connected to each other in stationary phase cells. We designed appropriate primers listed in Table S1 for studying any connections between four of the operator loci located at 17.0, 48.2, 64.1, and 66.5 min around the E. coli chromosome (Fig. 2A). We also designed 22 other primers (also listed in Table S1) each proximate to binding sites of FruR, MalT, PurR, TyrR, and H-NS transcription factors as likely negative controls for interactions with GalR sites (shown in Fig. 2A). We used EcoRI for 3C analysis, as its digestion sites are appropriately located in the vicinity of all chosen DNA targets. The efficiency of digestion was estimated by the amount of PCR products obtained with primers around individual EcoRI sites. After digestion, we detected very little DNA amplification of these restriction sites reflecting successful digestions, although the PCR amplifications were abundant without EcoRI treatments. Typical results with the P_DF/P_DR primer pair are shown in Fig. 3A (lanes 2, 3, 5, and 6). There were very few amplification products when treated with DNA ligase after digestion, showing that the digested two ends could not be ligated and restore the original DNA sequence in noticeable amounts (Fig. 3A, lanes 1 and 4). In addition, we found that the digestion efficiencies in cross-linked and non–cross-linked samples were somewhat different (Fig. 3B). In the non–cross-linked samples, DNA was more or less completely digested in 1 h, whereas in the cross-linked samples, the maximum digestion efficiencies reached to ∼80% in 4 h. Thus, we set the digestion time as 4 h. The P_CF/P_CR primer pair was used as the internal control of template for a DNA segment containing no EcoRI site.

Fig. 2.

Map of E. coli circular chromosome. (A) Positions of DNA binding sites are marked for various DNA binding proteins used for potential 3C interactions (as discussed in the text). The experimentally confirmed GalR binding sites are shown in red. (B) Intersegmental DNA networks by GalR, as determined by 3C assays, are shown by red lines. The potential GalR binding sites near the binding sites of many other DNA binding proteins identified here, are now marked as GalR′ with an arbitrary number, and are also shown in red.

Fig. 3.

3C PCR analysis of E. coli chromosome using pairs of primers. (A) Comparison of digestion efficiencies in cross-linked (lanes 4–6) or non–cross-linked (lanes 1–3) samples. Details of cell growth and template preparations are described in SI Materials and Methods. After 3C template preparations, the PCR products using the indicated primer pairs were run in 2% agarose gel and visualized by EtBr. The digestion efficiency was detected by primers, P_DF, and P_DR. Both P_CF and P_CR are primers for the PCR control corresponding to a DNA region that has no EcoRI digestion site. (B) Kinetics of restriction digestion of chromosome DNA by EcoRI. The incubation time of restriction digestion is given in hours. (C) Samples of chromosomal interactions detected by pairs of primers, including P₁₄, P₂₄, and P₂₆ cognate to three of the known GalR targets (Table S1) in stationary phase cells of WT E. coli (lanes 1 and 2). Left margin shows primer combinations. Experiments performed in ΔgalR and ΔgalS strains are in columns 3–6. Lanes N and C represent non–cross-linked and cross-linked samples, respectively. (D) PCR results using two non-GalR primers in WT and ΔgalR cells of stationary phase. Primer combinations are shown at the top.

We performed 3C analysis between GalR and GalR or GalR and non-GalR DNA targets for potential contacts between them in stationary phase cells. Interaction efficiency between two sites was defined by normalizing the ratios of the amount of PCR products between the two in the cross-linked and non–cross-linked samples compared with the internal controls (13). We arbitrarily set a threefold change or higher, to assign a positive interaction between a given pair of targets. Among 94 combinations tested, we found contacts among 30 of them. Sample PCR results are shown in Fig. 4A and summarized in Fig. 4B. There were no visible signals among the remaining 64 combinations. Surprisingly, three of the four GalR targets showed positive signals when matched not only against each other but also with six of the non-GalR targets that were not expected to contact GalR targets. For example, primers P₁₄, P₂₄, and P₂₆ close to GalR targets gave interaction signals when paired against P₉ and P₁₂ primers designed to test contacts with PurR and TyrR targets, respectively.

Fig. 4.

Quantification of 3C PCR data after normalization to the signals of the P_CF-P_CR amplification product. (A) Agarose gel images of PCR products with primers indicated at the top. (B) Quantification analysis of PCR data by gel imaging system software. Fold change is the ratio between noncross linked (N) and cross linked (C) after normalization by P_CF-P_CR. Signal ratio of cross-linked to non–cross-linked samples indicates interaction frequency in WT strain in the absence of d-galactose. Threefold change is set as threshold.

GalR-Mediated Intrachromosomal Contacts.

To confirm the participation of GalR in intrachromosomal contacts mentioned above, we performed the 3C assays in a ΔgalR mutant strain (14). Compared with the interaction frequency in the WT, the deletion of the GalR encoding gene removed most of the observed interactions, both between any GalR-GalR targets and most but not all of GalR and non-GalR targets (Fig. 3C, lanes 1–4). The results show that GalR indeed mediates most of the interactions observed here (Fig. 2B). A GalR homolog, GalS, binds specifically to all GalR targets tested but with different affinities (15, 16). We investigated the effect of GalS on the GalR-mediated intrachromosomal contacts in a ΔgalS strain (17). The results showed that, in the absence of GalS, the interaction frequency of the connections did not diminish but rather frequently enhanced (Fig. 3C, lanes 5 and 6), which is reminiscent of two previous observations: DNA-bound GalS, unlike DNA-bound GalR, does not associate (15), and more GalR are made in ΔgalS cells (16). We believe that the enhancement of GalR-mediated interactions observed in many cases in ΔgalS strain occurs because GalS competes with GalR for binding to DNA sites in WT, thus reducing the potency of GalR-mediated bridges in DNA; in the absence of GalS and presence of an extra amount of GalR, the GalR-mediated connections are higher.

Numerous GalR Binding Sites in the Chromosome.

As shown above, three loci, galP (66.52 min), galR (64.11 min), and mgl-galS (48.22 min) interacted not only with each other but presumably also with or near targets of FruR, MalT, PurR, TyrR, and HNS. Two models may explain the latter results. (i) The latter binding proteins collaborate with GalR to form bridging complexes. (ii) There are GalR binding sites not identified previously near the non-GalR targets. These ideas were tested by 3C analysis in strains with deleted binding proteins, ΔfruR, ΔmalT, ΔpurR, ΔtyrR, or Δhns. We found no significant change in the observed PCR products in the deletion strains compared with WT (Fig. S5). On the other hand, in experiments in which one non-GalR site was tested against another non-GalR site, many linkages observed in the WT strain were not found in the ΔgalR strain (Fig. 3D). The latter observations suggested that the second hypothesis is a more likely reason for interactions involving the so-called non-GalR sites. DNA sequence search indeed revealed the existence of 91 potential GalR binding sites around the chromosome (with zero or one mismatch with the consensus bases in the gal operator sequence: TGNAANCGNTTNCA (2). In fact, there is at least one potential GalR-binding site between each EcoRI restriction site at a non-GalR locus tested and the cognate primer sequence used in the 3C assays. The potential GalR-binding sites identified by 3C approach are shown in Table 1. We do not know whether these newly identified potential GalR binding sites actually bind GalR and are involved in regulation of specific genes. Nonetheless, our results show that a transcription factor for the gal regulon connects distal segments in the bacterial chromosome. Curiously enough, we did not observe any contacts between the primer (P₁) designed for the GalR-regulated gal operon (17 min) and other primers tested by the 3C assays. A priori, this was an unexpected observation. As mentioned earlier, the gal operon contains two operators, and GalR binds to both and associates, generating locally a DNA loop of 113 bp (18, 19). Given that GalR can form polymers, why GalR bound to these sites does not participate in the GalR-mediated chromosomal interconnections identified above remains to be investigated.

Table 1.

Examples of known and potential GalR binding sequences

Primer	Binding site	Localization	No. of mismatches	Sequence*
P1	galEE	791357–791372	0	GTGGAATCGTTTACAC
P1	galEI	791244–791259	1	ATG TAACCGCTACCAC
P26	galP1	3086026–3086041	0	CTGAAACCGATTACAC
P26	galP2	3086208–3086223	0	GTGTAATCGCTTACAC
P24	galR	2974591–2974606	0	ATGTAAGCGTTTACCC
P14	galS	2239786–2239801	0	CTGTAACCGTTTCCAT
P14	mgl	2238640–2238655	0	ATGTAACCGCTTTCAA
P9	F9	1735896–1735911	1	GTGGAAACGTTTGCTC
P21	F21	2736478–2736493	1	ATGGAAAAGGTTGCAC

Sequences shaded in gray are used in the pMini-3 circle for AFM analysis.

*The GalR consensus sequence is TGNAANCGNTTNCA.

Having established that the transcription factor GalR forms a chromosomal network in stationary phase cells, we tested whether these chromosomal interconnections exist in growing cells. It appears that more than 80% of the observed contacts disappear when cells are growing exponentially (Fig. S6A). These results are consistent with the hypothesis that moving DNA replication forks may disrupt the connections and that reassociation may be a slow process. Of the five signals in the WT that survived in growing cells, perhaps because these associations may have faster kinetics, only one was sensitive to the presence of d-galactose during logarithmic growth (Fig. S6B).

GalR-Mediated Looping in Vitro by Atomic Force Microscopy.

We have previously used AFM to observe GalR- and LacI-mediated DNA loops (7). In this research, we engineered five different operators for GalR binding in two mini DNA circles (pMini-1 and pMini-2) with discrete distances between binding sites. The map of pMini-1 and pMini-2 are shown in Fig. 5 A and C. Purified mini DNA circles were mixed with or without GalR protein as described in SI Materials and Methods. Samples were then scanned by an atomic force microscope. The AFM images of DNA without the protein show plectonomic structures with tight superhelical stretches of DNA and occasional loops because of crossing over of two double helical chains (Fig. 5 B and D, Upper). In the presence of GalR, we observed 71 molecules of GalR-mediated looped-out DNA of 246 total pMini-1 molecules and 33 of 154 total pMini-2 molecules inspected. We also counted the numbers of observed loops per GalR-DNA complexes. For pMini-1, 30% of DNA molecules formed loops (71 of 246), in which 12.6% contain two loops, 9.7% three loops, 5.7% four loops, and 0.8% five loops. In pMini-2, almost 20% formed loops (33 of 154), in which 9.1% contain two loops, 3.2% three loops, 7.8% four loops, and 1.3% five loops (Fig. 5E). GalR mediation in the loop formation was inferred by measuring the height and width of the overlapping DNA chains (6). DNA loops frequently emanate from one or more such taller and broader globular particles at the DNA crossover points as shown by black arrows in Fig. 5 B and D, Lower Left. We assume that these particles are oligomers of DNA-bound GalR. To establish that oligomerization of DNA-bound GalR generates looping of the DNA intervals, the contour lengths of large numbers of loops in both DNA alone and DNA/GalR samples were traced and the DNA lengths measured (samples are shown in Fig. 5 B and D.). Without protein, the loops sizes were, as expected, random because of DNA crossovers. We note that the observed DNA crossover points in the DNA-only samples are of different volumes and are not large enough to account for at least a GalR dimer. Volume estimates of the “cores” in the GalR plus samples suggest that oligomers (up to octamers) connect two or more gal operators sites. We also found that each DNA molecule may contain more than one GalR core particle (Fig. 5D, Lower Left). This may be the result of group of DNA binding sites—in the current case, two and three—get together with GalR independent of each other. To confirm the GalR mediated formation of core particle, we used the GalR^T322R, which is incapable of tetramerization, to see segmental interactions by AFM analysis. We observed the dimeric GalR bound to DNA circles when incubated with GalR^T322R, as indicated by red arrows in the lower right panel of Fig. 5 B and D. However, we also observed that only two loops per molecule at most (49 of 526 for pMini-1 and 12 of 205 for pMini-2), whereas for the WT GalR, even five loops per molecule could be found (shown in lower panel of Fig. 5 B, D, and E). The formation of two loops per molecule, which means that GalR is tetramerized, may be because of the residual tetramerization activity in the mutant GalR. Moreover, the GalR^T322R protein did not show any DNA-bound GalR that contains higher than tetrameric structure, unlike the DNA-bound WT GalR.

Fig. 5.

AFM analysis of pMini-1 and pMini-2 circles. (A) Map of pMini-1 circle. Five reported GalR binding sites were constructed into the circle with defined distances by overlapping PCR approach. Details are given in SI Materials and Methods. Numbers indicate the relative location of GalR specific operators in the chromosome: 1 = operator of galR, 2 = gal operon O_I, 3 = gal operon O_E, 4 = galS, and 5 = mgl operon. Distances between two sites are given as numbers of DNA base pairs. (B) AFM imagines of pMini-1 in the absence (Upper) and presence (Lower) of GalR or GalR^T322R. Size of protein present in the AFM images is calculated. Black arrow indicates polymerized GalR particle core; red arrow indicates dimeric GalR. Length of each loop after measuring is shown in number of base pairs. (C) Map of pMini-2 circle. Five reported GalR binding sites are indicated as follows: 1 = operator O₁ of the galP, 2 = galR, 3 = galS, 4 = mgl operon, and 5 = operator O₂ of the galP. (D) AFM imagines of pMini-2 in the absence (Upper) or presence (Lower) of GalR or GalR^T322R. (E) Distribution of percentages of DNA molecules vs. loop numbers in each molecule among pMini-1 and pMini-2 incubated with GalR or GalR^T322R that were counted for analysis.

To further examine the bridging of GalR binding sites, we constructed a smaller 648 bp DNA circle (pMini-3), which contained only three operators—one from previously known sites and two from the newly proposed operators from 3C studies described above. It was important to include the latter two because direct GalR binding to these presumed operators has not been directly demonstrated. The operators were separated by 100, 200, and 300 bp in pMini-3 (Fig. 6A). AFM analysis of pMini-3 alone showed, as expected, mostly plectonomic DNA (Fig. 6B, Upper). The plasmid in the presence of GalR clearly showed 1–3 loops of DNA per molecule. We found that each loop-containing DNA molecule, unlike the previous two plasmids, contained only one GalR core per molecule. This is because three operators per DNA would not allow formation of more than one GalR core. We measured contour lengths of GalR-bound and unbound molecules of pMini-3. Based on the measured contour lengths, the inferred binding patterns of GalR to pMini-3 are shown in Fig. 6C. The contour lengths of the plasmids carrying protein are highly consistent with the expected values from the binding site loci.

Fig. 6.

AFM analysis of pMini-3 DNA circle. (A) Map of pMini-3 plasmid. The pMini-3 DNA contains three GalR binding sites, two of which are potential GalR binding sites predicted by the 3C analysis. Construction and preparation of the mini circle are described in SI Materials and Methods. (B) Typical AFM imagines of pMini-3 in the absence and presence of GalR. Arrows show GalR protein particles under microscopy. (C) Inferred looping patterns corresponding to measured contour lengths from lower panel of B. Blue lines with arrows at both ends signify interacting sites.

Discussion

We used two independent in vivo methods, microscopic visualization of fluorescent-labeled GalR, and conformation capture of chromosome by cross-linking, to locate DNA bound GalR protein in stationary phase cell. Results from these two experiments presented here suggest that in stationary phase cells, the E. coli chromosome is partially condensed by the DNA sequence-specific GalR transcription factor that connects remote segments of DNA in 3D space. Incidentally, we identified many more previously unknown GalR binding sites that participate in such interconnections. We confirmed the connections between remote DNA sites by GalR in the absence of other factors in vitro by direct visualization of the interactions by AFM. Although the mechanism, the interface, and the energetics of a small (∼100-bp) DNA loop formation by association of two DNA-bound GalR dimers are known in detail (20, 21), the frequency or the stability of the remote intersegmental chromosomal multiconnections presumably by multimeric GalR dimers remains unknown at this stage, although the connections are neither rare nor transient. Given the current findings with GalR, we believe that other proteins in the cell may make similar intrachromosomal connections and may also contribute to 3D folding. The implication of our finding of remote intersegmental chromosomal connections is of importance. GalR-mediated intrachromosomal connections may serve at least two functions: The “togetherness” of gal regulon members should increase the local concentration of GalR around the distant gal regulon promoters and thus coordinate regulation of the functionally regulated gene products, as argued by Dröge and Müller-Hill (22). One way to increase the local concentration of DNA binding proteins is to have their genes located next to their DNA targets. Another putative role of GalR-mediated chromosomal connections may be architectural; it may incidentally help chromosomal compaction. We note that specific biological roles of many putative GalR binding sites discovered here remain unknown; either they are part of yet-to-be discovered GalR regulated genes and members of the gal regulon, or they serve purely an architectural role. We note that a ΔgalR strain does not have any effect on cell growth except a change in intermediary metabolites (23). We are currently investigating the significance of associations of DNA-GalR complexes in stationary phase cells. Our modified 3C method developed based upon the principle of Dekker et al. (6) has been used by Wang et al. (24) to show some intersegmental chromosomal contacts based on the nucleoid protein HNS.

The E. coli chromosome has been shown to contain several kinds of topographical arrangements. (i) It contains six “macro-domains” with defined boundaries, four of which are structured and two of which are nonstructured (25–28). Attempts to have site-specific recombination between att sites engineered into the macrodomains showed that chromosomal inversions within and between domains frequently have physiological consequences, and sometimes they are not permissible, putting some limits to chromosomal rearrangements (plasticity) (29). (ii) The chromosome may be organized into supercoiled topological domains (30–34). (iii) Between replication cycles, the chromosome is condensed into a filament extending from one cellular pole to the other (35). The two ends of the filaments are connected by a stretched-out DNA segment that includes the “terminus.” How the topography of the “macro-domains,” “topological loops,” the “filament,” and the currently demonstrated remote “segmental connections” in stationary phase cells reconcile with each other in the bacterial nucleoid remains a challenging question, and we do not know whether the different chromosome conformations discussed above are present in both log and stationary phase cells.

Materials and Methods

Protocols for fluorescent microscopy, AFM, and 3C analysis used in this study are described in SI Materials and Methods. Cells used for FM and 3C analysis were cultured in M63 minimal medium. Constructions of GalR-Venus and GalRT322R-Venus strains and mini circles used for AFM assay are described in detail also in the SI Materials and Methods. All the primers used for constructions are listed in Table S2.