key: cord-0824850-woe6knpv
authors: Joyeux, Marc
title: Requirements for DNA-bridging proteins to act as topological barriers of the bacterial genome
date: 2020-08-12
journal: Biophys J
DOI: 10.1016/j.bpj.2020.08.004
sha: 2b64380ac3f927f80b9bfcf798ca419f5a6b2b55
doc_id: 824850
cord_uid: woe6knpv

Abstract Bacterial genomes have been shown to be partitioned into several kilobases long chromosomal domains that are topologically independent from each other, meaning that change of DNA superhelicity in one domain does not propagate to neighbors. Both in vivo and in vitro experiments have been performed to question the nature of the topological barriers at play, leading to several predictions on possible molecular actors. Here, we address the question of topological barriers using polymer models of supercoiled DNA chains that are constrained such as to mimic the action of predicted molecular actors. More specifically, we determine under which conditions DNA-bridging proteins may act as topological barriers. To this end, we developed a coarse-grained bead-and-spring model and investigated its properties through Brownian dynamics simulations. As a result, we find that DNA-bridging proteins must exert rather strong constraints on their binding sites: they must block the diffusion of the excess of twist through the two binding sites on the DNA molecule and, simultaneously, prevent the rotation of one DNA segment relative to the other one. Importantly, not all DNA-bridging proteins satisfy this second condition. For example, single bridges formed by proteins that bind DNA non-specifically, like H-NS dimers, are expected to fail with this respect. Our findings might also explain, in the case of specific DNA-bridging proteins like LacI, why multiple bridges are required to create stable independent topological domains. Strikingly, when the relative rotation of the DNA segments is not prevented, relaxation results in complex intrication of the two domains. Moreover, while the value of the torsional stress in each domain may vary, their differential is preserved. Our work also predicts that nucleoid associated proteins known to wrap DNA must form higher protein-DNA complexes to efficiently work as topological barriers.

The genetic information of most bacteria is encoded in a circular DNA molecule comprising up to several millions of base pairs (bp). Such closed molecules are topologically constrained, because they lack free ends capable of rotating and releasing the torsional stress (1). In vivo alterations of the torsional state of circular DNA is mediated by the recruitment of enzymes called topoisomerases, which can either increase or decrease the twist of the double helix by opening transiently one or two strands (1). The net result of the action of all types of topoisomerases is that the genomic DNA of most bacteria is significantly undertwisted (negatively supercoiled) and winds about itself to transfer part of the torsional stress to the bending degrees of freedom, thereby forming plectonemes (1). If the DNA were not subject to any additional constraint beyond closure, then one single nick of one strand or one single break of the two strands would suffice to release the torsional stress of the full molecule. It has however long be known that at least several tens of nicks are required to achieve this goal in Escherichia coli (2, 3) . This indicates that there exist barriers which block the diffusion of the torsional stress along the DNA molecule and organize the chromosome of E. coli into many independent topological domains, whose torsional state is not affected by the relaxation of other domains. It is currently estimated that the genomic DNA of bacteria like E. coli is composed of several hundreds of different topological domains with variable size (average size ≈10 kbp) and position (4) . More generally, the partitioning of the bacterial chromosome into ≈10 kbp independent topological domains is believed to hold in most bacteria (5) and to be related, in part, to the insulation of fundamental co-expression units that are neighbors along the genome (6) .

The work reported here deals with the topological barriers which make this partitioning possible, that is, the barriers which block the diffusion of torsional stress along the DNA molecule and are responsible for its division into independent topological domains.

Although this question has now been addressed for nearly two decades (7) (8) (9) , the mechanisms underlying the formation of topological barriers remain mostly elusive. Most plausible models for the formation of topological barriers are (i) actively transcribing RNAP, which generate both positive and negative supercoils (10) and may consequently block the displacement and dissipation of plectonemes (11, 12) , and (ii) formation of DNA loops by certain nucleoid proteins, which may serve as topological barriers (13-16). In the present paper, we focus on this latter mechanism and establish under which conditions DNA-bridging proteins may serve as topological barriers.

Several DNA-bridging proteins have been studied in some detail, including transcription regulators like the LacI repressor (17) , H-NS (18) and H-NS-like proteins (19) , Lsr2 (20) , Fis (21) and Lrp (22) . In their functional form, all of these proteins have at least two independent DNA-binding domains, so that they can interact with two DNA duplexes simultaneously and form a bridge between two sites that are widely separated from the genomic point of view. Most of these proteins also bind non-specifically but with high affinity to the DNA molecule. However, among all these DNA-bridging proteins, LacI is to date the only one that has been proved to work as a topological barrier (15, 23, 24) .

The guiding line of the present paper is consequently the experimental demonstration in (15) that the binding of a DNA-binding protein to its recognition sites in two different locations on a supercoiled DNA molecule can confine free supercoils to a defined region and divide the DNA molecule into two distinct topological domains (15) . In order to gain more detailed information on this mechanism, we determined the minimal physical properties that must be added to a standard model of circular DNA to reproduce the results described in (15) .

The starting model for this study was similar to those we proposed recently to investigate facilitated diffusion (25) (26) (27) , the interactions of DNA and H-NS nucleoid proteins (28) (29) (30) , the formation of the bacterial nucleoid (31) (32) (33) (34) (35) , and the interplay of DNA demixing and supercoiling in compacting the nucleoid (36) .

The results presented in this article reveal that the formation of DNA loops is by no means sufficient to create topologically independent domains and that DNA-bridging proteins must exert rather strong constraints on their binding sites in order to act as topological barriers: They must block the diffusion of the excess of twist through both binding sites and must additionally block the rotation of one DNA segment relative to the other one.

The coarse-grained bead-and-spring model developed for the present study is described in detail in Model and Simulations in the Supporting Material. In brief, bacterial DNA is modeled, as in (30, 36) , as circular chains of n beads with radius 0 . As in (31, 36) , the torsional energy term was borrowed from (37) and requires the introduction of a body-fixed frame ( , , ) k k k u f v , where k u denotes the unit vector pointing from bead k to bead 1 k + . The torsional rigidity opposing rotation of ( , , )

) was adjusted so that at equilibrium the writhe contribution accounts for approximately 70% of the linking number difference (38) , as is illustrated in Fig. S1 . The values of the bending and torsional rigidities are close together, in agreement with experimental results (38) . It may be worth emphasizing that introduction of the body-fixed frame ( , , )

k k k u f v is crucial for modeling correctly the torsion of double-stranded DNA with a "single-stranded" bead-and-spring model. This procedure is quite realistic and has already proved very useful for studying, for example, the interplay of DNA demixing and supercoiling in nucleoid compaction (36) , the buckling transition in double-stranded DNA and RNA (39) , the influence of nucleoid-associated proteins on DNA supercoiling (40) 

As a starting point, we recall the experimental results in (15) , where the torsional relaxation properties of 4100 bp plasmids with initial superhelical density 0.06

Of special interest to us were the experiments performed with plasmids containing tandem copies of the binding site of the sequencespecific DNA-binding protein LacI at two different locations, such that the plasmid was divided into two stable loops of respective length 2900 bp and 1200 bp upon binding of LacI.

The 1200 bp region moreover contained the recognition site for one nicking enzyme (Nt.BbvC1) and the 2900 bp region the recognition site for a second nicking enzyme (Nb.BtsI). In the absence of LacI, addition of Nt.BbvC1 alone or Nb.BtsI alone was sufficient to release the full torsional stress of the plasmid. In contrast, in the presence of LacI, addition of Nt.BbvC1 alone removed only the 7 negative supercoils of the 1200 bp region, while addition of Nb.BtsI alone removed only the 17 negative supercoils of the 2900 bp region (15) .

Simultaneous addition of both enzymes was required to relax the full plasmid (15) . This experiment demonstrates very clearly that pairs of LacI bridges block the diffusion of torsional stress and divide the plasmids into two independent topological domains.

We report below on our efforts to determine the minimal set of mandatory properties that allow molecular bridges to act as topological barriers and block the diffusion of torsional stress in out-of-equilibrium supercoiled chains.

Preparation of torsionally out-of-equilibrium circular chains. , which were chosen so that the distance between their centers is smaller than 10

nm. This step is illustrated in Fig. 1 (a), which shows an equilibrated chain with 

where 12 r W quantifies the winding of one moiety of the chain around the other one, that is, loosely speaking, their intrication. The out-of-equilibrium chains prepared as described above

The main purpose of this work is to understand how DNA-bridging proteins like LacI (15, 23, 24) can maintain such an unbalance when the whole chain is relaxed. Since DNAbridging proteins form DNA loops by dynamically cross-linking widely separated DNA sites, the question amounts here to determine the constraints that must be imposed to beads α and β J o u r n a l P r e -p r o o f to divide the circular DNA chain into two topologically independent loops. Note that DNAbridging proteins are not introduced explicitly in the model: Rather, constraints on beads α and β are meant to model the mechanisms that allow the proteins to act as topological barriers.

Twist equilibrates much more rapidly than writhe.

In order to understand in more detail how the diffusion of torsional stress proceeds, the out-of-equilibrium circular chains prepared as described above were allowed to relax without constraint and all relevant quantities were monitored. From a practical point of view, four different initial out-of-equilibrium conformations were prepared for each value of 0 k L ∆ and 100 different relaxation trajectories were integrated for 1 ms for each initial conformation.

, and r j W ( 1 j = , 2) was then averaged over these 100 trajectories. In the absence of any constraint on beads α and β, all conformations obtained at the end of the 1 ms integration time satisfy 

where w 

Based on recent experimental results (12) (13) (14) (15) (16) , one may wonder whether DNA loops are ipso facto independent topological domains, or whether some additional, more subtle properties of DNA-bridging proteins are required to achieve this goal. We emphasize that the word "loop" is used here according to its biological meaning and not the geometrical one.

More precisely, it does not refer to a closed curve, but rather to a conformation of the DNA chain, where two DNA sites that are widely separated from the genomic point of view are maintained close to each other by some other macromolecule, so that the segment of DNA located between these two sites looks like a loop when it is observed from a long distance or projected on a plane. However, the chain is not closed and torsional stress cannot diffuse directly from one extremity of the loop to the other one through the bridging macromolecule.

In order to ascertain this point, we launched a series of simulations with the crudest model for DNA-bridging proteins, namely a simple bond between beads α and β. The , where pot E is the energy of the naked circular DNA chain described in Eqs. (S1)-(S5) and BP E an additional term, which models the action of DNA-bridging proteins. For the simplest model, we used

where d αβ is the distance between the centers of beads α and β and 0 4 d αβ = nm. For the sake of simplicity, we used the same value of the stretching rigidity as for the DNA chain, that is topological barriers is consequently related to some property, which is subtler than the mere connection between two genetically widely separated DNA sites.

Still, it is worth noting that the αβ bond has a significant effect on the time evolution of 12 r W , as can be checked in Fig. S5 . The reason is the following. In order to compute the partial writhes r j W (see the Supporting Material), one has to introduce two "virtual" closed chains 1 C and 2 C . 1 C is composed of beads α to 1 β − , plus the "unphysical" segment between beads β and α, while 2 C is composed of beads β to 1 α − , plus the same "unphysical" segment between beads α and β. When beads α and β are bonded, they remain close to each other for 0 t > , as indicated by the positions of the two arrows in Fig. 3(a) , so that the unphysical segment between beads α and β remains small and 1 C is a good representation of the initially torsionally relaxed DNA loop (and 1 r W is a measure of its writhe), 2 C is a good representation of the initially torsionally stressed DNA loop (and 2 r W is a measure of its writhe), and 12 r W is a measure of the intrication of the two loops. In this case, 12 r W remains close to zero at all times, as can be checked in Fig. S5(b) . This indicates that, for bridged DNA, equilibration of the torsional stress is not accompanied by any significant intrication of the two circular moieties. In contrast, for the system without the BP E term, beads α and β do not remain close to each other for 0 t > , as indicated by the positions of the two arrows in Fig. 1(c) , so that the unphysical segment between beads α and β becomes arbitrarily large and 1 C , 2 C , the r j W and 12 r W lose their physical meaning. In this case 12 r W varies quite widely with changing DNA conformations, as is illustrated in Fig. S5 

The reason why forming a loop is not sufficient to create independent topological domains is of course that the bond between beads α and β described in the previous subsection is not able to block the rapid diffusion of the excess of twist through beads α and β and the subsequent equilibration of the writhe. In order to form independent topological domains, the diffusion of the excess of twist must absolutely be blocked. Stated in other words, rotation of the internal basis ( , , )

k k k u f v around k u must be forbidden at ,

From 

for ,

We note that the k δ remain small, and the corrections in Eq. (5) 

where α ξ denotes the angle formed by beads β, α and 1 α + , and β ξ the angle formed by beads α, β and 1 β + . Note that the bending rigidity for these two angles was assumed to be five time larger than the bending rigidity of the DNA chain, in order for α ξ and β ξ to deviate only moderately from / 2 π . In this second model, the action of DNA-bridging proteins is consequently modeled by the potential energy terms in Eq. Fig. 4(b) . It is stressed that this behavior has no precise physical meaning but arises instead from the mathematical definition of the total writhe r W in Eq. (S11) and its somewhat arbitrary partitioning into the sum of the two partial quantities r j W and 12 r W in Eq.

(S17). In particular, r W does not display any sharp jump along single trajectories.

As can be checked in Fig. 4 Fig. 4(b) ). According to Eq. (1), this limit is equivalent to 12 r r W W =

. ) superimpose at short times, which indicates that the relaxation speed is proportional to the total amount of torsional stress. However, for longer times and/or larger values of the linking number difference, a slower regime sets in, which is probably due to the resistance opposed by plectonemes against the reorganization of the chain and the intrication of the two moieties. Finally, we note that after full relaxation the writhe contribution accounts for approximately 90% of the linking number difference, which is significantly larger than the 70% contribution in the absence of domains (see Fig. S1 ).

In conclusion, this second model of DNA-bridging proteins indicates that blocking the diffusion of the excess of twist at the two DNA loci bound by the proteins is sufficient to maintain a constant difference between the linking number difference, excess of twist, and writhe of the two moieties, but that it is not sufficient to insure true topological independence.

Indeed, the DNA chain evolves towards very compact and intricated final conformations,

where the topology of each moiety is different from the initial one. The relaxation limit 12 r r W W = (or equivalently 1 2 r r 0 W W + = ) appears non trivial and intriguing to us. We suspect that the explanation involves subtle considerations that should deserve further work.

The question that remains to be answered is consequently: What additional constraint must the DNA-bridging protein exert on its binding sites in order to preserve not only the difference in torsional stress between the two moieties of the DNA chain, but rather the precise initial torsional stress in each moiety ? The feeling that emerges upon examination of conformations like the one shown in Fig. 3(b) , is that the winding of the initially relaxed moiety around the initially stressed one probably requires the relative rotation of the two DNA segments around the αβ bond, and that no winding would be possible if this rotation were blocked. In order to check this conjecture, a third bending term was added to BP E , namely 0 2 2 2 0 2 BP , , 16) . In this context, we surmise that the slow relaxation observed in (15) for single DNA-bridges may correspond to the slow intrication obtained with our second model for DNA-bridging proteins, while the more stable topological separation observed in (15) for bridges in tandem corresponds to true barriers obtained with the third model for DNA-bridging proteins. We finally note that the intrication properties discussed here are also expected to contribute to the formation of topological domains that are, in effect, larger than the domain delineated by the binding sites of LacI as recently observed in (16) .

DNA-bridging proteins form DNA loops by dynamically cross-linking DNA sites that are widely separated from the genomic point of view (17) (18) (19) (20) (21) (22) . They usually bind DNA nonspecifically but with high affinity. DNA-bridging proteins have now been investigated for more than two decades, but it is only very recently that it has been demonstrated that a DNAbridging protein like the LacI repressor is able to separate circular plasmids into two topologically independent loops (15, 23, 24) . To the best of our knowledge, this is at present the only demonstrated case of a DNA-bridging protein acting as a topological barrier. In the present work, we aimed at going one step further and determining the conditions that DNAbridging proteins must fulfill to act as topological barriers. We showed that these proteins This work also brings some light on another mechanism that has been proposed to explain the formation of independent topological domains, namely two proteins wrapping the DNA in two different locations (Ref. (15) , Fig. 6 ). There is experimental evidence that certain wrapping proteins, like GalR (15), the λ O protein (15) , and Fis (62) Another model which is frequently proposed for the formation of topological barriers, namely actively transcribing RNAP (10) (11) (12) , is more difficult to discuss in terms of the results described in this paper, because several time scales need be considered. Indeed, it is likely that the two parts of the DNA molecule located respectively upstream and downstream from the transcribing RNAP will tend to increase their intricacy, in order to reduce the torsional stress induced by the blocking of the diffusion of twist at the RNAP. However, the RNAP is itself moving forward along the DNA, so that the ability of transcribing RNAP to act as topological barriers probably depends on the relative speeds of the two mechanisms. More work is needed to clarify this point. Similarly, the case of DNA-bending nucleoid proteins, like HU (63), certainly deserves further investigations.

Finally, we note with interest that the intricate structure obtained with DNA-bridging proteins that block the diffusion of twist along the DNA but do not prevent the relative rotation of the DNA segments are highly compact. It turns out that the mechanism leading to the formation of the bacterial nucleoid is a longstanding but still lively debated question (31, 32, (64) (65) (66) (67) . The point that has puzzled scientists for decades is that the volume of the Fig. 1(b) under the constraint that beads α and β are bonded (Eq. (3)).

(b) Representative snapshot obtained upon relaxation of the out-of equilibrium conformation shown in Fig. 1(b) for the second model of DNA-bridging proteins (Eqs. (5) and (6)). (c)

Representative snapshot obtained upon relaxation of the out-of equilibrium conformation shown in Fig. 1(b) for the third model of DNA-bridging proteins (Eqs. (5) and (7)). The 

imagined the research project and performed preliminary simulations. M.J. developed the models, ran the simulations and analyzed the results. M.J. and I.J. wrote the manuscript

Université Lumière Lyon 2) for her hospitality during the Covid-19 epidemic confinement

DNA topology

On the structure of the folded chromosome of Escherichia coli

Chromosomes in living Escherichia coli cells are segregated into domains of supercoiling

Topological domain structure of the Escherichia coli chromosome

Universal and idiosyncratic characteristic lengths in bacterial genomes

Conserved units of co-expression in bacterial genomes: An evolutionary insight into transcriptional regulation

The structure and function of the bacterial chromosome

Topological insulators inhibit diffusion of transcription-induced positive supercoils in the chromosome of Escherichia coli

Supercoiling of the DNA template during transcription

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

Organization of supercoil domains and their reorganization by transcription

Recombination by resolvase is inhibited by lac repressor simultaneously binding operators between res sites

DNA looping alters local DNA conformation during transcription

Dividing a supercoiled DNA molecule into two independent topological domains

Protein-mediated loops in supercoiled DNA create large topological domains

Some repressors of bacterial transcription

H-NS mediated compaction of DNA visualised by atomic force microscopy

The structural and functional organization of H-NS-like proteins is evolutionarily conserved in Gram-negative bacteria

Lsr2 of Mycobacterium tuberculosis is a DNA-bridging protein

Mechanism of chromosome compaction and looping by the Escherichia coli nucleoid protein Fis

The role of high affinity nonspecific DNA binding by Lrp in transcriptional regulation and DNA organization

DNA supercoiling, a critical signal regulating the basal expression of the lac operon in Escherichia coli

bp LacI-mediated, negatively supercoiled DNA loop by atomic force microscope

Description of non-specific DNA-protein interaction and facilitated diffusion with a dynamical model

Dynamical model of DNA protein interaction: effect of protein charge distribution and mechanical properties

Comparison of kinetic and dynamical models of DNA-protein interaction and facilitated diffusion

A model of H-NS mediated compaction of bacterial DNA

Equilibration of complexes of DNA and H-NS proteins on charged surfaces: A coarse-grained model point of view

Role of salt valency in the switch of H-NS proteins between DNAbridging and DNA-stiffening modes

Compaction of bacterial genomic DNA: Clarifying the concepts

In vivo compaction dynamics of bacterial DNA: A fingerprint of DNA/RNA demixing ?

Coarse-grained model of the demixing of DNA and non-binding globular macromolecules

A segregative phase separation scenario of the formation of the bacterial nucleoid

Preferential localization of the bacterial nucleoid

Bacterial nucleoid: Interplay of DNA demixing and supercoiling

Kinetics of DNA supercoiling studied by Brownian dynamics simulation

Structure of plectonemically supercoiled DNA

Dynamics of the buckling transition in double-stranded DNA and RNA

Influence of nucleoid-associated proteins on DNA supercoiling

Simulation of DNA supercoil relaxation

Competition between B-Z and B-L transitions in a single DNA molecule: Computational studies

A multiscale dynamic model of DNA supercoil relaxation by topoisomerase IB

Brownian dynamics simulations of sequence-dependent duplex denaturation in dynamically superhelical DNA

Transcription-driven twin supercoiling of a DNA loop: A Brownian dynamics study

Dynamics of site juxtaposition in supercoiled DNA

Salt effects on the structure and internal dynamics of superhelical DNAs studied by light scattering and Brownian dynamics

Limiting laws and counterion condensation in polyelectrolyte solutions. I. Colligative properties

Polyelectrolytes. Marcel Dekker

Use of site-specific recombination as a probe of DNA structure and metabolism in vivo

Structure and reactions of closed duplex DNA

DNA gyrase: structure and function

Transcription-coupled hypernegative supercoiling of plasmid DNA by T7 RNA polymerase in Escherichia coli topoisomerase I-deficient strains

Potent stimulation of transcription-coupled DNA supercoiling by sequence-specific DNA-binding proteins

Large-scale conformational transitions in supercoiled DNA revealed by coarse-grained simulation

Hyperplectonemes: A higher order compact and dynamic DNA self-organization

Viscous nonlinear dynamics of twist and writhe

Twist and writhe dynamics of stiff polymers

Structural basis for recognition of AT-rich DNA by unrelated xenogeneic silencing proteins

Domain organization and oligomerization among H-NS like nucleoid-associated proteins in bacteria

Protein occupancy landscape of a bacterial genome

Fis protein forms DNA topological barriers to confine transcription-coupled DNA supercoiling in Escherichia coli

Genes on a wire: The nucleoid-associated protein HU insulates transcription units in Escherichia coli

Shape and compaction of Escherichia coli nucleoids

DNA condensation in bacteria: Interplay between macromolecular crowding and nucleoid proteins

Physical descriptions of the bacterial nucleoid at large scales, and their biological implications

From structure to function of bacterial chromosomes: Evolutionary perspectives and ideas for new experiments

Polymer solutions: An introduction to physical properties

A combined wormlike-chain and bead model for dynamic simulations of long linear DNA

Conformational and thermodynamic properties of supercoiled DNA

References (70,71)