© 2008 Nature Publishing Group http://www.nature.com/nsmb

NEWS AND VIEWS in the Wzc tetramer (~55 Å) interact with the extensive periplasmic domains in the Wza octamer9,20, resulting in a periplasmic complex spanning a distance of ~145 Å between the cytoplasmic and outer membranes. Notably, the larger periplasmic domain of Wzc interacts with an extensive α-helical domain at the base of the Wza octamer, and cryo-EM data suggest significant conformational changes in both partners near the junction20. Thus, if Wzz does interact with an outer-membrane partner, the extended periplasmic domain in the resulting structure may be significantly altered from the view provided by the isolated protein. A crucial question for all of these systems is how they span the stress-bearing peptidoglycan layer without compromising viability. Genes encoding chain length–regulating PCP proteins have been known for more

than a decade, but despite their importance in surface assembly and virulence the mechanisms underlying their action have remained unknown. The availability of structural data for representative PCP-1 proteins now sets the stage for biochemical investigations to unravel the activities of these fascinating proteins and poses important questions concerning the structure and function of other PCP proteins. 1. Raetz, C.R. & Whitfield, C. Annu. Rev. Biochem. 71, 635–700 (2002). 2. Whitfield, C. Annu. Rev. Biochem. 75, 39–68 (2006). 3. Morona, R. et al. Microbiology 146, 1–4 (2000). 4. Daniels, C. et al. Environ. Microbiol. 4, 883–897 (2002). 5. Tocilj, A. et al. Nat. Struct. Mol. Biol. 15, 130–138 (2008). 6. Batchelor, R.A. et al. J. Bacteriol. 173, 5699–5704 (1991). 7. Grangeasse, C. et al. Trends Biochem. Sci. 32, 86–94 (2007).

8. Dong, C. et al. Nature 444, 226–229 (2006). 9. Collins, R.F. et al. Proc. Natl. Acad. Sci. USA 104, 2390–2395 (2007). 10. Tang, K.H. et al. Biochemistry 46, 11744–11752 (2007). 11. Marolda, C.L. et al. J. Bacteriol. (in the press) 190, (2008). 12. Marolda, C.L. et al. J. Bacteriol. 188, 5124–5135 (2006). 13. Bastin, D.A. et al. Mol. Microbiol. 7, 725–734 (1993). 14. Morona, R. et al. J. Bacteriol. 177, 1059–1068 (1995). 15. Daniels, C. et al. Mol. Microbiol. 28, 1211–1222 (1998). 16. Saldiás, M.S. et al. Microbiology 154, 440–453 (2008). 17. Goldman, R.C. & Hunt, F. J. Bacteriol. 172, 5352–5359 (1990). 18. Wu, T. et al. Proc. Natl. Acad. Sci. USA 103, 11754–11759 (2006). 19. Sperandeo, P. et al. J. Bacteriol. 189, 244–253 (2007). 20. Collins, R.F. et al. J. Biol. Chem. 281, 2144–2150 (2006).

DNA torsional stress propagates through chromatin fiber and participates in transcriptional regulation Christophe Lavelle Transient mechanical stresses induced by molecular motors that move DNA can propagate through the chromatin fiber and trigger local DNA alterations that in turn allow for specific DNA-protein interactions. Hence, DNA supercoiling modulations and chromatin conformational dynamics concur in gene regulation. DNA transaction events such as replication and transcription are always linked to changes in the topological state of the double helix. As DNA is rotating inside the polymerase, positive and negative supercoilings are induced downstream and upstream, respectively1. Transcriptionally generated torsion may be regarded as a waste product to be dissipated through DNA distortion and chromatin conformational dynamics, and/or disposed of by topoisomerases, but it may instead be part of a regulatory signal detected by molecular partners. Free energy of negative supercoiling can be used to locally unpair the DNA helix, facilitating the opening of promoter regions, or to drive the formation of alternative (non-B) DNA structures such as Z-DNA, cruciforms, intramolecular triplexes and quadruplexes, which may influence protein binding to regulatory regions. One could indeed hypothesize that these odd DNA structures

Christophe Lavelle is at the Laboratoire Physico-Chimie Curie, Unité Mixte de Recherche 168, Centre National de la Recherche Scientifique 168, Institut Curie, 11 rue P. et M. Curie, 75231 Paris Cedex 05, France.

e-mail: [email protected]

are more than by-products of ongoing events, because they have escaped negative selection despite the challenge they represent for DNA replication. Supporting this hypothesis, sequence analysis showed that stress-induced DNA destabilization occurs preferentially at regulatory loci2. However, despite considerable speculation and accumulation of in vitro data pointing toward transcriptional regulation mechanisms based on non-B DNA, the occurrence of such alternative structures and their effective participation in cellular processes, either as passive (as topological buffers) or active (as signal providers) transcription partners, had never been directly shown in eukaryotic cells. In this issue, Kouzine et al.3 managed to directly measure transcription-generated supercoiling in vivo. Using an activatable site-specific Cre recombinase to excise a chromatin fragment positioned between two divergent promoters, they show that torque generated by elongating RNA polymerase created sufficient supercoiling density in surrounding chromatin to enable transition to non-B-DNA of specific supercoilingsensing sequences positioned between the promoters. For this, cells were transfected with an autonomously replicating plasmid in which two loxP cassettes were inserted between the two

NATURE STRUCTURAL & MOLECULAR BIOLOGY VOLUME 15 NUMBER 2 FEBRUARY 2008

identical promoters; Cre recombinase–mediated excision and circularization of the 1-kb region between loxP sites in vivo enabled the trapping of supercoils that were structurally restrained or dynamically transiting through the segment. The topological difference between transcribed and nontranscribed episome (~–1.4 supercoils) progressively disappeared when relaxation by endogenous topoisomerases was allowed to proceed. Complete relaxation took more than 30 minutes, confirming that topoisomerases do not instantaneously relax supercoils, as has previously been observed in yeast4. Indeed, the preservation of some superhelical tension in active chromatin was suggested long ago to maintain a transcription-poised chromatin state5. In the present system, it could help to stabilize odd DNA structures. The role of topoisomerases was further investigated by treating the cells with both camptothecin (a selective topo-I inhibitor) and S1 nuclease, followed by indirect end labeling. Topo-I cleavage sites were found to be concentrated in internucleosomal linker DNA, where unconstrained torsional stress is supposed to accumulate. To test whether the dynamic unconstrained supercoiling was sufficient to promote transient non-B-DNA

123

structures in susceptible sequences, Kouzine et al. inserted the far upstream element (FUSE) sequence from the human MYC gene between the two promoters and monitored its conformation using potassium permanganate and UV–cross-linking psoralen assays3. Both approaches confirmed the melting of FUSE upon transcription activation in vivo. Finally, the authors used a chromatin immunoprecipitation assay with antibodies against two proteins, FBP and FIR, that are known to regulate MYC transcription upon binding to melted FUSE6,7, to confirm that transcription-induced FUSE melting initiated and sustained their binding. The cooperation of a mechanosensor of ongoing transcription (FUSE) with two competing effectors (FBP and FIR) provides a real-time mechanism for controlling MYC expression7. It remains to be determined how frequently a similar mechanism is used throughout the genome, and what factors influence its efficiency—for example, the arrangement of promoters, transcription rate, available topoisomerase activity, local chromatin structure and potential competition between different non-B-DNA structures8. Concretely, how do we fit all of that into the eukaryotic nucleus? Let’s put together a few ideas and build a speculative model (Fig. 1). First, chromatin is not only a reservoir of constrained negative supercoiling (with a mean of –1 supercoil stored per nucleosome); it may also act as a dynamic buffer, able to transiently absorb torsional stress, either negative or positive, because of the inherent conformational flexibility of nucleosomes, that is, their ability to change the crossing status of entry and exit DNAs9. Alternatively, part of the torsional stress can be dispersed as torsion in linker DNA, and the question of what balances these two possibilities in vivo remains to be answered. Second, by destabilizing nucleosomes, positive torsional stress may trigger their modification into hexasomes or tetrasomes (nucleosomes lacking one or two H2A-H2B dimers, respectively), lexosomes (distorted nucleosomes) or reversomes (chirally reversed nucleosomes)10. Thus, along with nonB-DNA structures, noncanonical nucleosomal structures might appear upon mechanical stress and participate in transcription regulation10. Third, supercoiling occurs when the free rotation of the DNA is restrained, which clearly happens in circular plasmids and bacterial chromosomes, but also in linear eukaryotic chromosomes as a result of the viscous drag of the huge chromatin fiber. Indeed, the same group showed that, even in topologically open linear DNA, transcriptionally generated torque can be intense enough to melt susceptible sequences upstream of active promoters in vitro6.

124

a

b

Direction of polymerase (relative movement) Transcribing polymerase

Transcription factory (RNA PoI II foci)

Chromatin pumping through polymerase

FUSE MYC promoter Topological domain

Chromosome territory

Insulator

Transient torsional stress –

+

Promoter firing

FUSE

MYC promoter

FBP Promoter extinction FIR Melted FUSE

Kim Caesar

© 2008 Nature Publishing Group http://www.nature.com/nsmb

NEWS AND VIEWS

Figure 1 Schematic overview of chromatin dynamics during MYC transcription. Left: when MYC is switched off, the domain is condensed into the chromosome territory and the FUSE is wrapped around a nucleosome7. Right: when MYC is switched on, the chromatin loop—delimited by two insulators—expands at the surface of the chromosome territory, and the locus rapidly relocates to a transcription factory18. FUSE is freed from the nucleosome through the action of a chromatin remodeler7 and, upon melting— triggered either by negative supercoiling produced by the same remodeler or by previous transcription initiation—recruits FBP and FIR to set up the feedback loop that will regulate transcription by modifying the rate of promoter firing6,7. While chromatin is rotated through the polymerase, supercoiling is generated in front and behind the enzyme as a result of the viscous drag1,6,16. Downstream positive supercoiling may (i) help to destabilize nucleosomes to clear DNA before it enters the polymerase16; (ii) be partially absorbed by chromatin structural dynamics (positively crossed nucleosomes, tetrasomes, hexasomes and reversomes)9,10; (iii) be released by topo-II switching positively crossed nucleosomes into negatively crossed ones4. Upstream negative supercoiling may (i) help to reincorporate nucleosomes on DNA that exits from the polymerase16; (ii) be partially absorbed by nucleosomes adopting the negatively crossed conformation9; (iii) be partially absorbed by melting linker DNAs, with potential induction of alternative DNA secondary structures3,6–8; (iv) be released by topo-I sensing constraints in these linkers3,4,16.

Independent topological domains have been suggested to form in the nucleus through the interaction of specific DNA regions with protein clusters or attachment to the nuclear matrix or other cellular structures, but limited information is available on the exact nature of these boundaries to DNA rotation11. Large transcription factories have been proposed by Cook and co-workers to participate in the partitioning of the genome12. These local concentrations of polymerases support a model in which DNA moves through immobile polymerases, instead of polymerases tracking like locomotives along their DNA template12. Hence, polymerases may be more appropriately viewed as DNA movers (or pumpers) rather than strict DNA-tracking enzymes13. Fourth, as shown by Kouzine et al., transcription-induced supercoiling escapes relaxation in vivo and could serve as a signal for other topology-dependent DNA transactions3. Generally, chromatin is thought to delay the relief of dynamic supercoiling by hindering the operation of topoisomerases14. More specifically, it has recently been shown in yeast that topo-II was five-fold more efficient than

topo-I in relaxing chromatin, whereas under the same conditions naked DNA was relaxed by topo-I twice as fast as it was relaxed by topo-II4. The poor ability of topo-I, a torque-sensitive topoisomerase15, to relax chromatin may reflect the high torsional resilience of chromatin9; chromatin structure might instead favor topo-II activity, by increasing the juxtaposition probability of DNA4,16. One can then speculate that topo-I should mainly relax negative supercoiling that is produced in the wake of the polymerase and is present as negative torsional constraint in linker DNA, as observed in the Kouzine et al. experiment3, whereas topo-II would be more efficient in relaxing positive supercoiling that is produced in front of the polymerase and is present as positive writhe, because no sink for torsional stress such as melting exists in this case . In conclusion, the work of Kouzine et al.3 enriches the palette of DNA roles. Whereas in its early days this long polyelectrolyte was seen as a mere passive carrier of genetic information, DNA has been found to contain regulatory signals that actively participate in the regulation of its metabolism through its

VOLUME 15 NUMBER 2 FEBRUARY 2008 NATURE STRUCTURAL & MOLECULAR BIOLOGY

NEWS AND VIEWS

© 2008 Nature Publishing Group http://www.nature.com/nsmb

own packaging, as shown by sequence-based rules for nucleosome positioning17. Now, DNA also seems to function as a driving belt for the transmission of mechanical stress in the dynamic control of gene expression. This definitely makes DNA supercoiling an essential participant in the regulation of transcription rather than just a bystander product of transcriptional elongation. ACKNOWLEDGMENTS I thank D. Levens for discussions and A. Prunell for comments on the manuscript.

1. Liu, L.F. & Wang, J.C. Proc. Natl. Acad. Sci. USA 84, 7024–7027 (1987). 2. Benham, C.J. Proc. Natl. Acad. Sci. USA 90, 2999–3003 (1993). 3. Kouzine, F., Sanford, S., Elisha-Feil, Z. & Levens, D. Nat. Struct. Mol. Biol. 15, 146–154 (2008). 4. Salceda, J., Fernandez, X. & Roca, J. EMBO J. 25, 2575–2583 (2006). 5. Villeponteau, B., Lundell, M. & Martinson, H. Cell 39, 469–478 (1984). 6. Kouzine, F., Liu, J., Sanford, S., Chung, H.J. & Levens, D. Nat. Struct. Mol. Biol. 11, 1092–1100 (2004). 7. Liu, J. et al. EMBO J. 25, 2119–2130 (2006). 8. Kouzine, F. & Levens, D. Front. Biosci. 12, 4409–4423 (2007). 9. Bancaud, A. et al. Nat. Struct. Mol. Biol. 13, 444–450

(2006). 10. Lavelle, C. & Prunell, A. Cell Cycle 6, 2113–2119 (2007). 11. Marenduzzo, D., Faro-Trindade, I. & Cook, P.R. Trends Genet. 23, 126–133 (2007). 12. Cook, P.R. Science 284, 1790–1795 (1999). 13. Cozzarelli, N.R., Cost, G.J., Nollmann, M., Viard, T. & Stray, J.E. Nat. Rev. Mol. Cell Biol. 7, 580–588 (2006). 14. Udvardy, A. & Schedl, P. Mol. Cell. Biol. 11, 4973–4984 (1991). 15. Koster, D.A., Croquette, V., Dekker, C., Shuman, S. & Dekker, N.H. Nature 434, 671–674 (2005). 16. Lavelle, C. Biochimie 89, 516–527 (2006). 17. Segal, E. et al. Nature 442, 772–778 (2006). 18. Osborne, C.S. et al. PLoS Biol. 5, e192 (2007).

Understanding how the replisome works Kenneth J Marians Recent papers shed insight into the architecture and dynamics of the components of the bacterial replisome. Three enzymatic activities are required to replicate chromosomal DNA: synthesis of the nascent DNA by DNA polymerases, unwinding of the parental duplex DNA by a DNA helicase and priming of the Okazaki fragments on the lagging strand by a DNA primase. In bacteria, these enzymes associate physically to form a replisome, a protein machine that faithfully replicates millions of nucleotides of DNA incredibly quickly. As the replication fork progresses, the replisome must accommodate the fact that synthesis of the nascent leading and lagging strands is inherently asymmetric. The leading strand is synthesized in a continuous fashion, thus demanding a highly processive DNA polymerase. In contrast, the lagging strand is synthesized in discontinuous pieces of about 2 kb, requiring only a moderately processive DNA polymerase but demanding a rapidly repeated cycle of primer synthesis, dissociation of the lagging-strand DNA polymerase from the penultimate Okazaki fragment, binding of the polymerase to the new primer and extension of the new nascent Okazaki fragment. Protein-protein interactions between replisome components serve to orchestrate these tasks. Chief among these interactions in directing the cycle of lagging-strand synthesis is the one between the primase, DnaG, and the helicase, DnaB. Several recent papers1–4, two of which appear in this issue, provide insight into how these proteins coordinate their efforts.

Kenneth J. Marians is the chair of the Molecular Biology Program, Memorial Sloan-Kettering Cancer Center, 1275 York Avenue, New York, New York 10021, USA. e-mail: [email protected]

The primase-DnaB interaction is distributive with respect to the Okazaki fragment cycle5— Okazaki fragment size is inversely proportional to the primase concentration and increases upon dilution of primase in the reaction. Mutational analysis of the region of primase that interacts with DnaB (the C-terminal helicase binding domain, HBD)6 demonstrated that primase variants with reduced affinity for DnaB had to be present in the reaction at higher concentrations than the wild type to effect the synthesis of Okazaki fragments of similar size, thus leading to the proposal that the primase-DnaB interaction controlled the lagging-strand cycle7. Similar observations have led to the same conclusion with respect to the bacteriophage T4 replication system8. The reports by Bailey et al.1 and Wang et al.2 reveal the architecture of the replication fork helicase and its interaction with primase. DnaB from Bacillus stearothermophilus and G40P, a DnaB homolog from Bacillus subtilus phage SPP1, both crystallized as double-tiered hexamers (Fig. 1a). The C-terminal domain (CTD), which contains the helicase motor and ATPase active site, shows the roughly six-fold symmetry observed previously for hexameric helicases and forms one tier of the structures. The second tier is composed of the N-terminal domain (NTD) formed into a trimer of dimers, with the monomer-monomer interface of one dimer mediated by interactions between α-helix hairpins and the dimer-dimer interfaces mediated by the globular portions of the NTD. Thus, the helicase will be oriented on the lagging-strand template with the CTD leading at the apex of the replication fork and the NTD behind. Mutational analysis by Wang et al.2 indicates that, although the monomer-monomer interface of the NTD

NATURE STRUCTURAL & MOLECULAR BIOLOGY VOLUME 15 NUMBER 2 FEBRUARY 2008

is required for helicase activity, the globular domain of the NTD is not. Once primase binds to DnaB, it must somehow scan the single-stranded, laggingstrand template emerging from the central cavity of the helicase motor for a recognition site at which it can initiate primer synthesis. The orientation of primase on DnaB and the question of how scanning might occur have now been clarified by the current reports. Bailey et al.1 also cocrystallized the HBD of primase bound to DnaB. It binds by capping the DnaB NTD dimerdimer interface. This observation immediately underscores several important features of the primase-DnaB interaction: capping of the dimer-dimer interfaces of the DnaB NTD will lock the three-fold symmetry of that tier in place, indicating that the different symmetries inherent to the two tiers of the enzyme represent the active form and not two different forms of the enzyme necessary for function, as had been suggested for G40P on the basis of EM reconstructions9. The capping function of the primase HBD also provides a simple explanation as to how primase stimulates DnaB helicase activity10: stabilizing the conformation of the DnaB hexamer on DNA will result in an increase in the processivity of unwinding. This structure also confirms that the stoichiometry of the primase– DnaB complex is three molecules of primase bound to one DnaB hexamer. Furthermore, communication from primase to the helicase motor is evident in a variant G40P engineered by Wang et al.2, where interfering with the interaction between the globular domain of the G40P NTD and the CTD yields a protein that still binds primase but is not stimulated to the same extent as the wild type with respect to its ATPase and helicase activities.

125