Department of Chemistry and Biochemistry, Florida International University, FL, USA
Address correspondence to: Fenfei Leng, Department of Chemistry and Biochemistry, Florida International University, 11200 SW 8th Street, FL 33199, USA. Tel: +305-348-3277. Fax: +305-348-3772. E-mail: email@example.com
Introduction: the Twin-Supercoiled-Domain Model of Transcription
DNA supercoiling is a fundamental property of chromosomal DNA in living cells [1, 2]. Not only does it affect DNA structure but also has great influence on essential DNA transactions, such as transcription, DNA replication, and recombination [1, 2]. Now it is clear that DNA supercoiling is under a tight homeostatic control by different DNA topoisomerases in vivo . In Escherichia coli, there are four DNA topoisomerases, that is, topoisomerase I, gyrase, topoisomerase III, and topoisomerase IV . Although these four topoisomerases have different functions during a cell cycle, the DNA supercoiling level is primarily set by opposing actions of DNA topoisomerase I and gyrase . Inactivation of either enzyme results in the production of (+) or (−−) supercoiled plasmid DNA in vivo [5, 6].
A number of early studies have shown that transcription is related to DNA supercoiling in E. coli cells. For instance, Lockshon and Morris  showed that the treatment of E. coli cells with DNA gyrase inhibitors, such as novobiocin, generated (+) supercoiled DNA topoisomers after transcription. Using Salmonella typhimurium and E. coli topA strains, Pruss  showed that plasmid pBR322 became (−−) supercoiled, which required transcription and translation of the tetracycline resistance gene, a membrane-insertion protein. In 1987, Liu and Wang  proposed a “twin-supercoiled-domain” model to explain these experimental results in which transcription plays a critical role in the formation of (−−) supercoiled DNA in topA mutants or (+) supercoiled DNA after the treatment of DNA gyrase inhibitors. This elegant model proposed that during transcription elongation, as the nascent RNA transcript becomes longer and longer, the transcription complex cannot continuously rotate along the DNA double helix, which forces the DNA strand to rotate around its own helix axis. As a result, two supercoiled domains are created, a (+) supercoiled domain ahead of the advancing transcriptional complex, and a (−) supercoiled domain behind it. After two decades of studies from different laboratories, both in vitro and in vivo experimental results clearly support the twin-domain mechanism for transcription-coupled DNA supercoiling (TCDS) [8-14]. Liu and Wang  also pointed out that in a diluted aqueous solution the friction force against the transcribing RNA polymerase was too small to cause significant supercoiling of the DNA template containing just one transcription unit. We now know that inside of a living cell is far away from the scenario of a dilute aqueous solution . In fact, cellular cytoplasm and nucleus are viscous and also extremely crowded . TCDS should be different from the situation in the dilute aqueous solution. Indeed, increasing viscosity significantly enhances the efficiency of TCDS in a defined protein system , indicating that the in vivo situation is far more complex than that of the dilute aqueous solution. More recently, a few groups reconsidered the induced torsional stress by a transcribing RNA polymerase and demonstrated that the torsional force of RNA polymerase was sufficient to generate the twin-domains even in a dilute aqueous solution. For example, Nelson  showed that small natural bends in the DNA helix backbone could increase a few thousand-fold of torsional stress even in linear unanchored DNA. This torsional force is sufficient to induce the formation of a (+) supercoiled domain in front of the transcribing RNA polymerase and a (−) supercoiled domain behind it . Another case is a Brownian dynamic study performed by Mielke et al. . These authors clearly demonstrated that a transcribing RNA polymerase alone could drive the formation of a (+) and a (−) supercoiled domains in a naked plasmid DNA template (for details, see Fig. 1 of ref. 18).
For experimental studies of TCDS in vitro and in vivo, we almost solely rely on the utilization of small circular plasmid DNA templates to determine the topology of the DNA topoisomers after transcription [9, 13]. As demonstrated previously [9, 11, 12], TCDS on plasmid DNA templates requires two barriers: a friction barrier stemming from preventing or retarding transcriptional machinery from rotating around the DNA double helix and a topological barrier to prevent the cancellation of the (+) and (−) supercoiled domains (Fig. 1A). We have briefly discussed the formation of the friction barrier by a transcribing RNA polymerase. In the following sections, we will continue discussing the function of the friction barrier and also review our understanding of the topological barrier and its role in TCDS.
TCDS in Defined Protein Systems
The early in vitro studies used simple defined protein systems to test the twin-supercoiled-domain model . A few parameters were examined in these defined protein systems, including the promoter strength, the length of the RNA transcripts, and the effects of certain nucleoprotein complexes on DNA templates. For instance, Tsao et al.  used a defined protein system containing a relaxed plasmid DNA template, T7 or T3 RNA polymerase, and a prokaryotic topoisomerase I in the presence of the four ribonucleoside triphosphates for transcription. As prokaryotic topoisomerase I specifically removed (−) supercoils generated during transcription elongation, transcription by T7 or T3 RNA polymerase resulted in a rapid accumulation of (+) supercoils on DNA templates . The production of (+) supercoils was also dependent on the length of RNA transcripts. These results clearly demonstrated that the twin-supercoiled-domain model was working.
In 1994, Drolet et al.  used another defined protein system to study transcription-driven (−−) supercoiling of the DNA templates by prokaryotic DNA gyrase. As DNA gyrase selectively converted (+) supercoils into (−) supercoils, (−−) supercoiled DNA topoisomers were generated during transcription elongation in the presence of DNA gyrase. Interestingly, the production of (−−) supercoiled DNA was linked to the R-loop formation. However, it was not clear whether a twin-supercoiled-domain mechanism played a role in the R-loop-linked supercoiling. If so, the R-loop-linked DNA supercoiling should depend on the length of RNA transcripts, a key parameter to prevent RNA polymerase from rotating around the DNA double helix.
More recently, we set up a similar in vitro-defined protein system to study TCDS in the presence of DNA gyrase [9, 10]. Additionally, HU protein and RNase HI were added to the transcription-supercoiling assays to suppress the R-loop formation. As the DNA template contained a T7 promoter, it allows specific initiation of transcription from the T7 promoter by T7 RNA polymerase. We also added one or multiple Rho-independent transcription terminators to each plasmid which enabled us to restrict transcription to selected regions and modulate the length of RNA transcripts. Finally, we inserted into most plasmids one or more recognition sites for a sequence-specific DNA-binding protein. Using this defined protein system, we found that a variety of sequence-specific DNA-binding proteins, such as the bacteriophage λ O replication initiator (λ O) or the E. coli lac repressor (LacI) or gal repressor (GalR), strongly stimulated TCDS. We demonstrated that this stimulation required the presence in the DNA template of a recognition sequence for the relevant DNA-binding protein and depended on the production of long RNA chains by an RNA polymerase. Our data are most consistent with a model in which specific DNA-binding proteins facilitate a twin-domain mechanism to enhance DNA supercoiling during transcription. More precisely, we suggest that some nucleoprotein complexes, perhaps those that contain topological constrains such as DNA-loops or wraps (Figs.1B and 1C), can form topological barriers that impede the diffusion and merger of independent chromosomal supercoil domains. We also demonstrated that TCDS in the defined protein systems takes place by two separate, and apparently independent, mechanistic pathways in vitro . The first supercoiling pathway, which is not likely to be significant in vivo, was found to be dependent on R-loop formation but not on the length of the RNA transcripts, and could be suppressed by the presence of RNase HI or HU protein. The second pathway for TCDS was much more potent, but became predominant in vitro only when sequence-specific DNA-binding proteins were present during transcription. This major supercoiling route was shown to be resistant to RNase HI and had functional properties consistent with those predicted for the twin-supercoiled-domain model. Under optimal conditions, the twin-domain pathway of TCDS rapidly and efficiently generated superhelicity levels more than twice that typically found in vivo .
DNA Topological Barriers
Recently, our laboratory successfully constructed a series of plasmid DNA templates that carry several tandem copies of one or two DNA-recognition sites in two different locations (Fig. 1 of ref. 21). These DNA-recognition sites divide the plasmid DNA templates into two regions of different size, 1.2 and 2.9 kb. We also placed nicking restriction endonucleases Nt.BbvCI and Nt.BtsI into these plasmids such that Nt.BbvCI resides in the 1.2-kb region and Nt.BtsI in the 2.9-kb region. These novel DNA templates are powerful tools to study DNA topological barriers in defined protein systems. Indeed, using two newly developed biochemical methods, that is, DNA-nicking and -gyrase methods and atomic force microscopy (AFM), we discovered that several sequence-specific DNA-binding proteins, such as LacI, GalR, and λ O, function as DNA topological barriers to block supercoil diffusion and divide these DNA molecules into two independent topological domains. For example, LacI was able to bind to lac O1 operators of the plasmid DNA templates pCB115 and pCB109 and divide these two molecules into two DNA-loops (Fig. 2). Interestingly and importantly, LacI was able to divide these two plasmids into two topologically independent domains, one relaxed domain and one supercoiled domain (Fig. 2). Under our experimental conditions, these two domains are highly stable (t1/2 = 2 h; ref. 21). Our unpublished results showed that one molecule of LacI is sufficient to block supercoil diffusion upon binding to O1 and O2 or O3 operators, suggesting that this DNA-looping protein functions as a topological barrier in lac operon under physiological conditions. Intriguingly, our data also showed that two DNA-wrapping proteins, GalR and λ O, were able to block DNA supercoil diffusion. For instance, a combination of these two unrelated proteins also confined free supercoils to define regions and divide a DNA molecule into different topological domains .
Two models are proposed to explain our experimental results (Fig. 3). Model (I) is for DNA-looping proteins, such as LacI, which are able to bring two DNA-recognition sites together to fold into a topologically constrained nucleoprotein complexes. These nucleoprotein complexes serve as DNA topological barriers to block supercoil diffusion. We have to point out here that the topology inside the nucleoprotein complexes may be critical for the formation of the topological barriers . Indeed, our AFM images showed that LacI upon binding to lac O1 operators result in the wrapping of DNA around the LacI molecules (Fig. 2). We would also like to emphasize that DNA topological barriers can be formed from a multiple protein complex binding to two different kinds of DNA-binding sites. For example, protein complexes of certain enhancers are able to bring together DNA elements thousands of base pairs away . It is possible that the DNA topology inside the enhancers is constrained and the enhancers by themselves are topological barriers. Model (II) is for DNA-wrapping proteins, such as λ O and GalR. Upon binding to their recognition sites, these sequence-specific DNA-wrapping proteins induce the wrapping of specific DNA sequence around them to form certain unique nucleoprotein complexes, such as O-some . As pointed out previously , these unique nucleoprotein structures are able to form topological barriers that slow or prevent diffusion of supercoils past the nucleoprotein complexes. For circular DNA molecules, two such topological barriers are required to divide them into two topologically independent domains (Fig. 3).
The discovery of DNA topological barriers has great biological significance. For instance, the DNA topological barrier models provide a framework to explain the reason why and how chromosomes are divided into different topological domains. Previous studies showed that E. coli chromosome is segregated into 400–500 dynamic, topological domains  by topological barriers . Although so far the structural nature of topological barriers in vivo has not been revealed yet, our results suggest that DNA-binding proteins are important components of topological barriers. E. coli cells contain a few nucleoid-associated proteins with relatively high concentrations, such as HU, H-NS, IHF, and FIS. These proteins are able to constrain (−) supercoils upon binding to DNA [27, 28] and induce the formation of higher order structures of chromosomes . It is possible that the nucleoprotein complexes generated from these nucleoid-associated proteins serve as general topological barriers to modulate localized DNA supercoiling in E. coli cells. Indeed, recent results showed that H-NS plays a critical role in global chromosome organization in bacteria  and our unpublished data demonstrated that HU and H-NS are able to confine supercoils to a define region. E. coli cells also contain many sequence-specific DNA-binding proteins, such as LacI and GalR. These sequence-specific DNA-binding proteins may be part of the dynamic topological barriers in vivo. For example, as discussed above, LacI may function as a topological barrier in the lac operon under physiological conditions. As (−) supercoils also assist LacI binding to lac operators , LacI-mediated topological barriers may be part of the mechanism of transcription inhibition by LacI in vivo. Additionally, a transcribing RNA polymerase may also serve as a topological barrier . First, RNA polymerases cause DNA wrapping around themselves . Second, as discussed above, a transcribing RNA polymerase is capable of generating a (+) supercoil domain ahead of the RNA polymerase and a (−) supercoil domain behind it . These two topological domains should be able to block supercoil diffusion. For eukaryotes, the basic structure of chromosome is the nucleosome  where a 146 bp of DNA wraps itself around a histone octamer in ∼1.8 turns. According to our model (II) (Fig. 3), nucleosomes should be able to function as topological barriers, block supercoil diffusion, and modulate DNA topology in defined regions. Additionally, it was shown that in the interphase nuclear DNA is attached to nuclear matrix via binding of matrix attachment regions (MAR) to MAR-binding proteins to form topologically independent supercoiled loops or domains . Our results strongly support these hypotheses. Nevertheless, experimental evidence is required to determine whether nucleosomes and MAR-binding proteins are able to block supercoil diffusion and function as topological barriers. Furthermore, our models present a reasonable explanation for transcriptional activation of DNA replication initiation for bacterial phage λ where TCDS plays a critical role for the activation of λ DNA replication. Specifically, the O-some  assembled from wrapping DNA around O protein in the replication origin blocks, confines, and captures TCDS, which causes structural changes in λ DNA replication origin . In this case, the DNA replication origin is unwound and DNA replication is initiated.
TCDS in E. coli topA Strains
In 1985, Pruss  reported that (−−) supercoiled plasmid pBR322 was isolated from S. typhimurium and E. coli topA mutant strains lacking DNA topoisomerase I. In addition, the isolated pBR322 DNA was extremely heterogeneous in linking number . A following study showed that transcription of tet gene was responsible for the effects on DNA supercoiling in the topA mutant strains . Furthermore, a series of studies showed that (−−) supercoiling of plasmids in topA strains required anchoring the transcribing RNA polymerase to cell membrane through a nascent hydrophobic membrane-bound peptide or protein [8, 11, 12]. For example, after analyzing TCDS of pBR322 and its derivatives, Lodge et al.  suggested that transcription-driven twin-supercoiled-domains were produced only if the DNA templates were anchored to a large cellular structure via coupled transcription, translation, and membrane insertion of a nascent protein. Indeed, Lynch and Wang  clearly demonstrated that (−−) supercoiling of plasmid DNA templates in topA mutants required anchoring the transcribing RNA polymerase to the membrane through a nascent hydrophobic membrane-bound peptide or protein. In addition, Cook et al.  demonstrated that transcription of oppositely oriented membrane-associated gene products rapidly supercoiled the plasmid DNA templates in topA strains. In these cases, the membrane-bound transcriptional machinery cannot freely rotate around the DNA double helix and therefore provides a friction barrier to produce the twin-supercoiled-domains on the DNA templates. We have noticed that all these studies regarding TCDS in E. coli topA strains utilized a combination of E. coli RNA polymerase and its promoters. As demonstrated previously, pBR322 and its derivatives contain several E. coli RNA polymerase promoters  where the length and location of RNA transcripts cannot be precisely controlled. TCDS may result from simultaneously transcribing several transcriptional units on these plasmid DNA templates. Therefore, it was difficult to determine the factors influencing TCDS in topA strains. Apparently, a more specific model system was needed to identify the parameters that regulate TCDS in E. coli.
We also developed a similar system to study TCDS in E. coli . This system consists of a E. coli topA strain, such as VS111(DE3), in which a λ DE3 prophage containing a T7 RNA polymerase gene under the control of the lacUV5 promoter has been integrated into the cell chromosome, along with a set of plasmids producing RNA transcripts of various lengths by T7 RNA polymerase. As each plasmid carries a T7 promoter, transcription can be induced upon adding IPTG into the cell culture containing VS111(DE3) harboring one of the plasmids. Using this unique system, we were able to examine the effects of the length of the RNA transcripts on TCDS and found that transcription by T7 RNA polymerase strikingly induced the formation of (−−) supercoiled plasmid DNA. We also discovered that TCDS was dependent on the length of RNA transcripts, precisely predicted by the twin-supercoiled-domain model. One surprising finding is that (−−) supercoiling of plasmid DNA by T7 RNA polymerase did not require anchoring of DNA to the bacterial cytoplasmic membrane, which is in contrast to those described previously [8, 11, 12], indicating that a transcribing RNA polymerase alone is sufficient to cause change of local DNA superhelicity.
Recently, in a separate study, to examine whether promoter strength affects TCDS, we developed a two-plasmid system in which a linear, nonsupercoiled plasmid was used to express lac repressor constitutively, whereas a circular plasmid was used to gauge TCDS in E. coli cells . Using this two-plasmid system, we found that TCDS in topA strains is dependent on promoter strength. We also demonstrated that transcription-coupled (−−) supercoiling of plasmid DNA did not need the expression of a membrane-insertion protein for strong promoters, which is consistent with our results by T7 RNA polymerase . Regardless, it might require cotranscriptional synthesis of a polypeptide. Furthermore, we found that for weak promoters the expression of a membrane-insertion tet gene was not sufficient for the production of (−−) supercoiled DNA. Our results can be explained by the “twin-supercoiled-domain” model where the friction force applied to E. coli RNA polymerase plays a critical role in the generation of (−−) supercoiled DNA.
Effects of TCDS on the S. typhimurium Leu-500 Promoter
In bacterial genomes, many promoters are divergently coupled . For example, in the ilvYC operon of E. coli, the ilvY promoter is divergently coupled to the ilvC promoter . Results from Hatfield laboratory clearly demonstrated that the transcriptional activities of the ilvY and ilvC promoters are dependent on the localized superhelical density around the promoter region  where transcription was able to modulate the localized superhelicity. Another well-characterized example is the activation of the S. typhimurium leu-500 promoter by divergently coupled transcription. The leu-500 mutation is a single A-to-G point mutation in the −10 region of the promoter controlling the leu operon, which results in leucine auxotrophy . The leucine prototrophy can be restored by mutations in the supX locus (initially named su leu 500) whose location was identified in S. typhimurium between the tryptophan operon and the cysB locus (trp-cysB). Several studies demonstrated that supX was topA, the gene coding for DNA topoisomerase I [41, 42]. The AT to GC mutation is expected to increase the energy barrier for the formation of a functional transcription open complex. Not surprisingly, this phenotype can be suppressed by a mutation in the topoisomerase I gene which results in a loss of topoisomerase I's activities. Intriguingly, Lilley and Higgins  demonstrated that the activation of the leu-500 promoter was dependent only on the topA background but did not correlate with the level of global supercoiling, as measured for extracted plasmid DNA templates. In addition, when the leu-500 promoter was cloned into a plasmid, its activation no longer required topA background . These results suggested that a different regulating factor rather than topA background was responsible for the activation of the leu-500 promoter. More recent results clearly showed that transcription-driven localized supercoiling rather than the global superhelical density was responsible for the activation of the leu-500 promoter and that the “mystery” regulating factor is TCDS . Furthermore, after the analysis of leu operon and the surrounding elements in S. typhimurium chromosome, Wu et al.  revealed that the ilvH promoter and the leuO gene were located upstream of, and were transcribed divergently to the leu-500 promoter. Based on these analyses, they proposed a promoter relay mechanism to explain the expression of genes in the ilvIH-leuO-leuABCD gene cluster which is coordinated in a sequential manner . The key component in this model is TCDS, which causes transient localized structural changes on DNA templates. Furthermore, in a recent study, El and Bossi  showed that the orientation of the DNA supercoiling had opposite effects on neighboring promoters where transcription-driven (−) supercoiling activated the leu-500 promoter and transcription-induced (+) supercoiling suppressed the promoter.
As Liu and Wang  published the twin-supercoiled-domain model of transcription in 1987, most studies focused on confirming the formation of a friction barrier by a transcribing RNA polymerase, which creates two supercoiled domains around the RNA polymerase. It is evident that a transcribing RNA polymerase alone is sufficient to provide the friction force to supercoil the DNA template in defined protein systems and in E. coli cells [9, 13]. A question arises from these studies whether TCDS especially the transient supercoiling is physiologically significant. Indeed, a few cases showed that TCDS is able to activate transcription, DNA replication, and recombination in vivo [39, 49]. However, as mentioned above, these studies almost solely relied on circular plasmids for determining the topological change of the DNA templates. Nevertheless, it is technically difficult to study transcription-induced transient DNA supercoiling and the effects on other DNA transactions in vitro and in vivo. Now, it is time to think about new strategies to study transient DNA supercoiling in defined protein systems and in vivo. Another outstanding question is how DNA supercoils move along the DNA helix axis in vivo. Recently, at single-molecule level, van Loenhout et al.  observed the movement of DNA supercoils along DNA double helix in vitro. Their results showed that supercoils move along DNA quite rapidly using two different mechanisms, diffusion and long-range hopping . It will be interesting to examine whether certain DNA-binding proteins or nucleoprotein complexes, such as HU, H-NS, and nucleosomes, affect this dynamic process. Of course, the next challenge is to devise methods to study DNA dynamics in vivo. A third unanswered question is which components contribute to the formation of topological barriers in vivo. As discussed above, our recent results suggest that topological barriers can be formed through binding some sequence-specific DNA-binding proteins to their binding sites, which forms topological structures, such as DNA loops and wraps (Fig. 3; ref. 21). Are these topological barriers biologically significant? Can histones or histone-like protein form these types of topological barriers? Can R-loops or certain DNA structures form topological barriers? The nature of the topological barriers in vivo still needs to be determined.
This study was supported by grant 5SC1HD063059-04 from the National Institutes of Health (to F.L.).