Understanding the spatial and functional organization of bacterial genomes throughout the cell cycle is important to understanding their physiology. It is fascinating to learn what mechanisms have evolved to organize the genome and reduce its effective volume below that of a cell (Dame, 2005; Luijsterburg et al., 2008; Dillon and Dorman, 2010; Dame et al., 2011). The organization of the bacterial genome is hierarchical, occurring at different length scales. On the micrometer scale it is organized in macrodomains (Niki et al., 2000; Boccard et al., 2005). In Escherichia coli four such domains, each of them on the order of 1 Mbp in size, have been described: the Ori and Ter domains and the Left and Right domains flanking Ter. Although, it is poorly understood how these domains are organized and what demarcates their borders, recent studies have demonstrated that certain architectural proteins are exclusively associated to certain domains (Mercier et al., 2008; Sanchez-Romero et al., 2010; Cho et al., 2011; Tonthat et al., 2011). Within macrodomains at intermediate scales the genome is organized in topologically independent domains (loops) of an average size of 10 kbp (Postow et al., 2004; Deng et al., 2005; Noom et al., 2007). Also in this case it is not well understood how these loops are formed, but it has been suggested that they arise by clustering of DNA-bound proteins, such as RNA polymerase [in transcription factories (Cook, 2010)] or architectural proteins that bind and bridge sites or regions that are distant along the genome, such as H-NS and SMC/MukBEF (Dame et al., 2000; 2006; Petrushenko et al., 2010; Schwartz and Shapiro, 2011). Finally, there are numerous architectural proteins that shape the path of the genome on the nanometre scale by their ability to bridge or bend the DNA at their binding site. In E. coli this includes proteins such as H-NS, FIS, HU and IHF (Luijsterburg et al., 2008; Dillon and Dorman, 2010). Although not all of the relevant architectural proteins identified in E. coli are conserved in Caulobacter crescentus, it is reasonable to assume that similar general principles of genome organization apply also to this organism. Organization and remodelling of the genome throughout the cell cycle and in response to environmental cues are key determinants in tuning the physiological state of bacteria.
Upon replication of the C. crescentus genome active transport of the ori (followed by the rest of the newly replicated genome) towards the distal pole occurs. This transport is facilitated by interaction of the parS site adjacent to the ori with the ParB protein, which in turn associates with ParA. ParA is present as polymerized structure extending from the distal pole and it is believed that depolymerization of this structure is the driving force to pull the newly replicated chromosome via its parS site to the distal pole (Ptacin et al., 2010; Schofield et al., 2010; Shebelut et al., 2010; Banigan et al., 2011). Previous studies have demonstrated that the speed by which the two parS sites separate is not constant throughout the segregation phase (Viollier et al., 2004; Shebelut et al., 2010). In the current study, Hong and McAdams have measured the separation velocity of the parS sites and sites up to ∼150 kb away from parS. The authors used time-lapse fluorescent microscopy to track the subcellular positions of these sites during chromosome segregation, which permits calculation of translocation speeds. Their experiments show that a site proximal to parS and sites that are further away exhibit different behaviour. Loci ∼100 and 150 kb away separate at a much higher speed than the parS sites. The site at ∼50 kb exhibits a mixed segregation pattern behaving like the parS site in one-third of the observed cells and separating at the same velocity as the more distant loci in the remaining two-thirds of the cells. Loci at ∼10 kb distance generally exhibit only slow segregation, identical to that of parS. For comparison, it takes the parS site 5 min to reach the distal pole, whereas transport of distant loci is completed in about half a minute. The authors suggest that these differences are due to the existence of two segregation mechanisms. They propose that in addition to the slow segregation mediated by the Par system, a second – currently undefined – mechanism is operating that facilitates fast segregation. An alternative explanation is that these differences can be attributed to the build-up of tension during transport that might relate to differences in structural organization of the regions nearby parS (here defined as the region up to the first locus at 12 kb) and those further away. In fact, the locus at 12 kb in part of the cells follows the parS site at the same slow speed, whereas in other cells it follows after a time delay at high speed. Loci further away generally segregate at high speed (up to 2.5 times the speed of the parS site). It is unclear where this heterogeneity in behaviour of the parS proximal region derives from, but at least in a fraction of cells this observation can be interpreted as this region being less sensitive to stretching by force than distal regions. This could imply the existence of a stable compact region (characterized by a high stiffness) close to parS and distal regions that are more easily extended by the application of force. When reaching a critical tension this ‘extended spring’ relaxes and the DNA can be transported to the distant pole. The authors give qualitative insight into the response to force of the region up to 12 kb away from parS. The distance from the 12 kb locus to parS increases as a function of parS speed (a proxy for the applied force). As similar data are lacking for longer distances our predictions relating to the different stiffness of the two regions can unfortunately not be verified.
An interesting question is: what delays the transport of these distant loci? An obvious candidate to delay transport is the replisome and the rate of replication. A back of the envelope calculation shows that this is indeed a possibility. Replication proceeds at a rate of on average about 21 kbp min−1 (Dingwall and Shapiro, 1989), while the speed of parS movement measured by Hong and McAdams is in the order of 1–3 µm min−1, i.e. approximately 3–10 kbp min−1 assuming that the DNA is fully extended during segregation. However, this is not the case as the genome is compacted (see next paragraph). As the authors show, for instance a region of 20 kb of DNA in the parS proximal region has an average end-to-end distance of 0.3 µm.
In addition to segregation speed, Hong and McAdams investigated the folding of an approximately 30 kb genomic region directly adjacent to parS by measuring the mean physical distance R between parS and ten different loci up to 33 kbp away. They found a scaling law of the form R∝g0.34 with g denoting the genomic distance between measured loci. The relation between the physical and the genomic distance can be compared with polymer models to estimate the folding state of the region of interest. This approach has not yet been applied to studies in bacteria but has been widely used in the field of eukaryotic chromatin during the last one and a half decade (Emanuel et al., 2009; Tark-Dame et al., 2011). A hallmark of polymer models is that the mean distance between pairs of monomers, R, exhibits scaling laws of the form R∝gν where g denotes the distance between the monomers along the chain. ν is a scaling exponent that takes different values depending on the chain conformation. For a self-avoiding chain one finds ν ≈ 3/5 whereas for a situation where the excluded volume can be neglected, one finds the ideal chain exponent ν = 1/2. Both exponents are larger than 0.34, the value observed by Hong and McAdams. That value seems to suggest a compact object for which the overall size scales like the number of building blocks to the power 1/3. In fact, the size of a polymer globule in a poor solvent scales like N1/3 with N denoting the total number of monomers. Remarkably, however, this law does not apply to intermonomer distances inside the globule. In fact, screening of the excluded volume due to the presence of other monomers leads to an exponent 1/2 (Lua et al., 2004). All the standard polymer models thus show larger exponents than 0.34.
How can the small value of the exponent be explained? To start with, let us stress the remarkable fact that a similar value of that exponent, approximately 1/3, has been derived for eukaryotic DNA organization from the contact probability between pairs of loci between ∼ 500 kb and ∼ 7 Mb (Lieberman-Aiden et al., 2009; Mirny, 2011). The authors of that paper suggested that this might reflect the organization of eukaryotic chromatin in the form of a fractal globule, an object as compact as the ordinary globule discussed above but in addition self-similar on all length scales as each of its subchains is condensed in itself. A remarkable feature of such structures is that they have very few topological entanglements. This might be advantageous from a biological point of view since it allows large-scale opening and closing of sections of the genome. Based on that idea the fractal (or crumpled) globule had already been suggested long ago on purely theoretical grounds as a possible candidate for eukaryotic DNA organization (Grosberg et al., 1988; 1993; Halverson et al., 2011; Mirny, 2011). It should be mentioned, however, that unlike the highly entangled standard globules, this is a non-equilibrium structure. Such structures might be possible due to the huge lengths of eukaryotic DNA that would need hundreds of years to reach equilibrium in the absence of enzymes involved in disentanglement (e.g. topoisomerases) (Rosa and Everaers, 2008). It should be noted that in addition to the fractal globule model there are several other non-standard polymer models explaining chromatin folding (Emanuel et al., 2009; Tark-Dame et al., 2011).
For the much shorter bacterial genomes where relaxation times are on the order of a few minutes, it seems unlikely that the DNA shows a non-equilibrium conformation. How can then the small value of the scaling exponent be explained in this case? Instead the small value might reflect genomic conformations resembling the ones of branched or looped polymers. Both types of polymers are much more compact than linear polymers (Cunha et al., 2001; Bohn et al., 2007). Indeed, as mentioned by Hong and McAdams, supercoiled DNA resembles branched polymers. This might thus explain the small exponent found in their experiment. Consider a branched polymer with N monomers where a fraction f of the monomers leads to multifunctional branches. A path spanning the whole structure has m < N monomers from which fm arms, each on average m monomers long, branch off; hence m∝N1/2. For an ideal chain this path forms a random walk leading to an overall size of the branched polymer proportional to N1/4; in a good solvent the path is substantially stretched and the overall size scales like N1/2 (Cunha et al., 2001). The observation by Hong and McAdams of a value of about 1/3 might thus very well reflect a branched structure of the C. crescentus genome in the 30 kb large region directly adjacent to parS. Since it is not straightforward to unambiguously assign a certain type of folding based purely on the value of the exponent, more data will be necessary to come to a definite conclusion about this point.
Interestingly, earlier measurements by Shapiro and coworkers on longer distances (up to 2 Mbp, the distance between ori and ter), have reported linear organization of the C. crescentus genome between ori and ter, i.e. the physical distance scales linearly with genomic distance (Viollier et al., 2004). Although the data by Hong and McAdams on the much shorter region relate to a period of active transport, it implies different types of organization at different length scales. An illuminating recent study on the overall three-dimensional organization of the C. crescentus genome relying on a Chromosome Conformation Capture-based approach (Dekker et al., 2002) by Church and coworkers (Umbarger et al., 2011) revealed that the chromosome is folded as an ellipsoid, with the arms between ori and ter spiralling around each other. To date that approach, however, does not have the resolution to provide information on the folding at short-length scales, the scale at which architectural proteins are likely to act and that is investigated here.
Studies in which the in vivo contribution of an architectural protein to folding of a genomic region is evaluated using polymer physics models have not been reported to date for any organism. As the authors focused their studies on short-length scales they were in an ideal position to investigate the role of a bacterial architectural protein in vivo. HU was one of the first proteins proposed to have an architectural role in the folding of the bacterial genome (Rouviere-Yaniv and Gros, 1975). The protein is present in several tens of thousands copies per cell, it binds DNA with low sequence specificity, is associated with the folded bacterial genome in vivo (Shellman and Pettijohn, 1991; Wery et al., 2001; Lee et al., 2011; Wang et al., 2011) and has been shown to compact DNA in vitro (Skoko et al., 2004; van Noort et al., 2004). An effect of HU on genome folding has however not been directly demonstrated in vivo. The role of the two C. crescentus HU proteins was investigated here by assessing their effect on the folding of 30 kb regions on both sides of parS using single and double knockout strains. The authors convincingly show that the measured distance between loci along the genome and the parS site is larger in the HU-1, HU-2 double knockout strain than in the wild type. Similarly, at all but one of the probed loci deletion of HU-2 results in decompaction of the genome. Deletion of HU-1 does not result in any measurable change in compaction. This observation may at first sight seem surprising, but might be accounted for by in vitro data on the E. coli HU proteins, demonstrating that one of the HU homodimers has much lower DNA binding affinity than the other type of homodimer or the heterodimer (Pinson et al., 1999).
The availability of techniques that permit assessing genome folding in the cell has been driving the progress in understanding chromatin organization and dynamics in eukaryotic organisms. The type of data generated by quantitative Chromosome Conformation Capture and imaging techniques is very well suited to interpretation using polymer physics models (Emanuel et al., 2009; Mirny, 2011; Tark-Dame et al., 2011). However, it is hard to pinpoint exactly which model best suits the data as the value of the scaling exponent alone does not permit conclusive assignment of the ‘best’ polymer model. The bacterial chromatin organization field is now catching up, providing the first glimpses into local (Hong and McAdams, 2011) as well as global genome folding (Viollier et al., 2004; Wiggins et al., 2010; Umbarger et al., 2011). It will be very insightful to assess and validate the roles of other potential chromatin organizing proteins in C. crescentus and to apply and extend these approaches to other bacterial model organisms.