Multidisciplinary perspectives on bacterial genome organization and dynamics
For correspondence. E-mail firstname.lastname@example.org; Tel. (+31) 20 598 7838; Fax (+31) 20 598 7991.
Bacterial genomes are organized by a plethora of chromatin proteins and physical mechanisms. This organization appears to be hierarchical with DNA folding events at the nm scale influencing higher levels of chromosome organization. Besides acting in shaping the genome these factors also play important regulatory roles in numerous DNA transactions. While DNA folding mechanisms operating at the nm scale are fairly well understood, it has been hard to translate this knowledge into accurate models that describe the complete dynamics of the genome. In recent years new techniques have evolved that are key to filling the current gaps in understanding. Particularly insightful in this light appear techniques that probe architectural properties of chromatin proteins on single molecules, techniques that map the binding of protein components and spatial structure on a genome-wide basis and improved imaging techniques that provide resolutions capable of resolving substructures/heterogeneities in the nucleoid. Moreover, bioinformatic and polymer physics approaches are starting to provide novel insights. In our opinion, an important aim in the field is to generate an accurate and complete description of the nucleoid and its dynamics at all scales. A first step towards this aim has now been set by bringing together people from diverse disciplinary backgrounds at the Lorentz centre workshop ‘Biology and Physics of Bacterial Genome Organization’ in Leiden, the Netherlands from 18 to 22 June 2012.
Provide feedback or get help You are viewing our new enhanced HTML article.
If you can't find a tool you're looking for, please click the link at the top of the page to "Go to old article view". Alternatively, view our Knowledge Base articles for additional help. Your feedback is important to us, so please let us know if you have comments or ideas for improvement.
Defining the compaction and functional organization of genomes is a question of longstanding biological interest in bacteria as well as in other organisms. Nevertheless, we remain far from an integrated model that describes the interactions that organize the bacterial nucleoid. A key tenet of the workshop was that understanding chromosome structure and function requires knowledge spanning different ‘length scales’. On the nanometre scale, the configuration of the DNA is modulated by the action of small chromatin proteins. At an intermediate scale, the genome has been proposed to fold into loops on the order of 10 kbp in size. On the micrometer scale, genomes are divided into independently structured domains on the order of 1 Mbp in size. An interconnecting model describing all of these length scales remains elusive (Espeli and Boccard, 2006; Dame et al., 2011).
During the last decade numerous researchers, applying novel methodology have entered the field of bacterial genome organization. Thus the study of bacterial chromosomes is in a stage of revival and novel excitement. For instance, the application of single-molecule imaging and manipulation techniques in vitro has provided new insights into the architectural properties of individual chromatin proteins acting on DNA. Mapping genome-wide binding of individual chromatin proteins using Chromatin Immuno Precipitation (ChIP)-based approaches has provided two-dimensional information in relation to the proteins that shape the genome and can be implemented in models of the nucleoid (Noom et al., 2007). A next step is to generate spatial models of nucleoid organization. Currently this can be done by probing interaction frequencies between sites along the genome using Chromosome Conformation Capture (3C)-based approaches. With these types of information becoming available the nucleoid has also become an exciting playground for polymer physicists and bioinformaticians that model its organization based on physical principles or structural codes within the genome. Finally, important developments have occurred in relation to direct visualization of the nucleoid using super-resolution microscopy (PALM/STORM) with which resolutions on the scale of several tens of nm's can now be obtained. This makes it possible to resolve substructures and heterogeneities in the nucleoid.
Representatives of all different disciplines in the bacterial chromatin field gathered at the Lorentz workshop ‘Biology and physics of bacterial genome organization’ organized in Leiden, the Netherlands from 18 to 22 June 2012. The workshop gathered about 55 scholars with diverse disciplinary backgrounds and both experimental and theoretical approaches. The workshop was thematically organized around the three different levels of organization mentioned above. Thus, the participants in this workshop ventured into a multidisciplinary journey to explore different levels of organization.
The nanometre scale: the architectural properties and roles of individual chromatin proteins
Ever since the isolation of the first small, basic, abundant DNA-binding proteins in the 1970s it has been believed that these proteins are involved in genome organization. These proteins (e.g. HU, IHF, FIS, H-NS, Dps and CbpA) are to date generally referred to as nucleoid-associated proteins (NAPs) (Dame, 2005; Browning et al., 2010). It has been proposed that these proteins structure DNA by bending the double helix or forming bridges between separate DNA segments (Luijsterburg et al., 2006; 2008). Expression of nucleoid-associated proteins is temporally regulated with different proteins being abundant depending on the physiological state of the cell. The changes in the expression levels of the individual NAPs likely provides a means to remodel the genome and adapt it to growth condition-specific requirements. Important proteins expressed at high levels during exponential growth include FIS, HU and H-NS, whereas in stationary-phase proteins such as Dps and CbpA become dominant (Ali Azam et al., 1999).
The DNA bridging H-NS and SMC proteins were the topic of several contributions. An important question in relation to the mechanism of SMC action relates to the interaction of SMC with accessory proteins (in Bacillus subtilis ScpA and ScpB). This question was addressed by Stephan Gruber (MPI of Biochemistry, Martinsried, Germany) who generated an extensive collection of ScpA and ScpB cysteine derivatives to probe the interactions of these proteins in vivo and in vitro. His experiments revealed that SMC–ScpA–ScpB forms rings that might embrace multiple DNA molecules (submitted). Valentin Rybenkov from the University of Oklahoma, Norman, USA reported studies on the Escherichia coli equivalent of these proteins: MukB and the interaction partners MukE and F (together referred to as MukBEF). His biochemical and single-molecule studies suggest a somewhat different mode of DNA bridging in which the two DNA binding domains of an individual MukB dimer interact directly with (rather than embrace) two separate DNA duplexes. The exact role of MukEF is still undefined, but it appears to modulate the strength of MukB mediated DNA bridging (Petrushenko et al., 2010). Stephan Gruber concluded that the B. subtilis complex closely resembles eukaryotic SMC complexes (including cohesin and condensin), whereas the architecture of the E. coli complex is substantially different.
Jocelyne Vreede (University of Amsterdam, the Netherlands) and William Navarre (University of Toronto, Canada) reported studies into structural and functional aspects of the DNA bridging protein H-NS (Dorman, 2007). For a number of years the structures of parts (the N-terminal dimerization domain and the C-terminal DNA-binding domain) of the H-NS protein have been available, but a full-length structure of the H-NS protein has not. Available structures determined by NMR and X-ray crystallography of the H-NS dimerization domain exhibit striking differences. Thus, the coiled-coil of the dimerization domain has been observed in both parallel and anti-parallel orientations. These structures have been the subject of discussion and it has been proposed that the two structures are both relevant and sensitive to environmental conditions (salt, temperature etc.). Using Molecular Dynamics simulations Jocelyne Vreede showed that both structures are very stable and that conformational switching within the dimer is not likely to occur. The anti-parallel structure is stable under both low- and high-salt conditions, whereas the parallel structure is only stable under high-salt conditions. This suggests that the anti-parallel conformation is predominant in vivo. If transitions between the two conformations occur the dimer needs to pass through monomeric intermediates (Vreede and Dame, 2012). William Navarre elaborated on new insights into the DNA-binding domain of E. coli H-NS and the H-NS-like protein Lsr2 from Mycobacterium tuberculosis. It has already been known for many years that H-NS preferentially binds to AT-rich DNA. Using protein binding microarrays William Navarre assessed on microarrays the binding of both proteins to thousands of different DNA octamers and found that H-NS and Lsr2 recognize a similar feature: narrow minor groove width. Thus, they prefer AT-rich sequences with T-A steps. The proteins bind DNA through an AT-hook like motif, that despite the lack of overall sequence homology appears conserved among different H-NS and H-NS-like proteins (Gordon et al., 2011). In the light of evolution it is very interesting that distinct organisms have independently evolved proteins with similar function and DNA recognition properties. It is exactly the ability of recognizing AT-rich DNA (which has in many genomes been acquired by horizontal gene transfer) that provides ‘xenogeneic silencing’ (Navarre et al., 2007; Ali et al., 2012) and facilitates the uptake and incorporation of foreign DNA.
The large changes in expression levels of NAPs upon transition from exponential into stationary phase coincide with dramatic compaction (and remodelling) of the genome. Two proteins are expressed at particularly high levels during stationary phase, Dps and CbpA (Yamada et al., 1990; Almiron et al., 1992), and might drive these dramatic changes. It is interesting to note that the expression of both proteins is induced by decreasing levels of the exponential phase ‘specific’ FIS protein when growth ceases albeit by different mechanisms (Grainger et al., 2008 and see below). Dps induces the formation of liquid crystalline structures when expressed at high levels and protects DNA against oxidative stress (Nair and Finkel, 2004). At the molecular level the action of Dps is poorly understood. Anne Meyer (Delft University, the Netherlands) addressed which residues are important for DNA binding and for its protective role against hydroxyl radicals. She demonstrated that key residues for efficient (i.e. cooperative) binding of Dps to DNA are in the N-terminal region of the protein. Another important region in the protein is the site of Fe2+ co-ordination, as Dps-catalysed oxidation of Fe2+ by H2O2 prevents the formation of harmful hydroxyl radicals. Surprisingly, mutation of certain residues in the DNA binding, but not the Fe2+ co-ordination, domains results in loss of protective activity against H2O2 in vivo. Even later in stationary phase the CbpA protein becomes highly abundant (Ali Azam et al., 1999). This protein has to date received relatively little attention. Studies discussed by David Grainger (University of Birmingham, UK) suggested that CbpA actually might fulfil a function similar to that of Dps in terms of DNA protection. In addition to extensive biochemical and single-molecule studies addressing the DNA binding and protective activity of the protein, he reported a transcription regulatory mechanism unknown in bacteria that couples decreasing levels of the FIS protein upon transition to stationary phase to high expression of CbpA (K. Chintakayala et al., submitted).
An alternative – complementary – view on changes induced upon transition from fast to slow growth was presented by Ding Jin from the Frederick National Laboratory for Cancer Research, National Cancer Institute, NIH, USA. He reported studies on the role of RNA polymerase in the organization of the nucleoid. Previously, he used fluorescence microscopy to show that clustering of the highly transcribed rRNA operons contributes to genome compaction(Cabrera et al., 2009). In his current unpublished work he showed that depletion of RNA polymerase from the nucleoid by using both competitor DNA and osmotic shock (resulting in dissociation in response to the change in ionic strength) promotes compaction. Other factors that are believed to affect nucleoid compaction such as supercoiling and NAPs have no visible effect on microscopy images. The effects seen for RNA polymerase lead him to suggest that RNAP has an anti-condensation function (unpublished observations).
The intermediate scale: interplay of genome organization and DNA transactions
Bacterial gene regulatory systems have been studied for decades and provide the basis for our understanding of gene regulation in all organisms. However, such systems are often studied in isolation, with little consideration for the wider effects of chromosome organization and DNA topology. Steve Busby, University of Birmingham, UK began proceedings by elegantly illustrating the dramatic effects of chromosomal location on gene expression in E. coli, showing that transcriptional output from promoters varies considerably with chromosomal context. Moreover, when the activity of a given promoter is induced, the location of that DNA segment within the cell is altered (Sanchez-Romero et al., 2012). While currently unexplained this observation hints at a mechanism for transcribed genes to colocalize in time and space, as also suggested by the studies reported by Ding Jin (see above). Charles Dorman, Trinity College Dublin, Ireland delved deeper into the effects of DNA topology on global control of transcription, presenting a fascinating comparison between E. coli and Salmonella enterica serovar Typhimurium (S. typhimurium). Differences in the superhelical density of DNA in the two organisms are modulated by growth conditions and the FIS protein (Cameron et al., 2011). The promoter of ssrA, the gene that controls transcription in the SPI-2 pathogenicity island of S. typhimurium, showed abnormal activity when transferred to E. coli but adopted a Salmonella-like expression pattern upon deletion of E. coli's fis gene and relaxation of DNA supercoiling. Data were also presented showing that interactions between H-NS, OmpR, Fis and their DNA targets are differently affected by changes in DNA supercoiling. Thus, the interplay between global regulatory proteins and their DNA targets is highly dynamic and is modulated by the topology of the DNA. These factors at least partly account for the novel expression patterns exhibited by a given gene following its horizontal transfer to a related bacterial species, illustrating the important role played by regulation in bacterial evolution. Pat Higgins (University of Alabama, USA) further dissected the relationship between DNA supercoiling and transcription on a genome-wide basis. Using a ‘supercoiling sensor’ relying on local recombination efficiency inserted at different sites along the genome he measured supercoiling levels and transcription elongation rates along the genome. He showed that supercoiling levels change directly adjacent to highly transcribed genes [which confirms the Liu and Wang model of transcription-driven changes in supercoiling (Wu et al., 1988)] and that vice versa transcription elongation rates are heavily reduced by supercoil losses (Rovinskiy et al., 2012). Taku Oshima (Nara Institute of Technology, Japan) explored the relationship between genome evolution and nucleoid-associated proteins. H-NS binding profiles between different E. coli strains appear to be conserved, despite differences in underlying DNA sequences. Strikingly, regions of DNA bound by H-NS were found to be more variable than H-NS free sections of the chromosome consistent with recent observations regarding genome evolution and gene expression (Martincorena et al., 2012). Bernt Eric Uhlin (Umeå University, Sweden), Sylvie Rimsky (CNRS, France) and Antonio Juarez (University of Barcelona, Spain) all sought to address the complex relationship between H-NS and the many H-NS-like proteins expressed in pathogenic enterobacteria. These talks highlighted the remarkable complexity of gene regulatory systems co-regulated by H-NS and related proteins such as Hha and Ler. Given that H-NS also modulates housekeeping genes, a relevant issue going forward is to identify the differential sequence motifs to which H-NS binds with or without its partners.
The micrometer scale: folding the chromosome
Two pioneers in the field of super-resolution microscopy (STORM/PALM), Xiaowei Zhuang (Harvard University, USA) and W.E. Moerner (Stanford University, USA) presented studies on bacterial chromosome structure. W.E. Moerner presented studies on the localization of HU in Caulobacter crescentus, which – as expected – is mostly uniformly distributed in the nucleoid (Lee et al., 2011). Similar observations on the localization of HU, FIS and IHF in E. coli were reported by Xiaowei Zhuang. Different from these three proteins the H-NS protein was found in foci. The number of H-NS foci correlated with the number of chromosome equivalents: two foci per chromosome (Wang et al., 2011). The meaning and significance of these foci is not yet understood. In combination with 3C experiments her imaging studies strongly support spatial (re)organization of a number of chromosomal regions as a function of H-NS binding. These results are in line with the notion that H-NS is a DNA bridging protein (Dame et al., 2000; 2006). A similar role in nucleoid organization (i.e. genomic looping) has been recently proposed for genes controlled by the DNA looping transcription factor GalR (Qian et al., 2012), suggesting that gene clustering might be a common occurrence in bacteria. Whereas the 3C studies of Xiaowei Zhuang were limited to defining pairwise interactions between a small number of sites along the genome, it is also possible to do such studies on larger scales using a high-throughput variant of 3C, 5C (Dostie et al., 2006). With the 5C approach all interactions for a large number of genomic sites can be determined. Studies employing a genome-wide 5C approach for the first time in bacteria were reported by Davide Bau from the group of Marc Marti-Renom (University of Barcelona, Spain). They determined the interaction profile of 30 000 sites along the C. crescentus genome. This profile was used to build a 3D model of the genome using the Integrative Modelling Platform (IMP – http://www.integrativemodeling.org) developed at the University of Barcelona. This model confirmed that the C. crescentus nucleoid has an ellipsoidal shape (with the origin and terminus at opposed locations) and that the left and right arms are helically folded around each other. One might expect to be able to detect looped structures (loops of on average 10 kbp in size by analogy with E. coli) or to observe macrodomains, which differ in the level of compaction, using this approach. Unfortunately, the current resolution of ∼ 13 kbp does not permit detection of the small loops. Although a division of the genome in multiple macrodomains – which by analogy with E. coli might exist in C. crescentus – can also not be seen in the data, the terminus region appears to be more compactly folded than other regions of the genome (Umbarger et al., 2011).
Analysis of Chromosome Conformation Capture data was also presented by Leonid Mirny, MIT, USA. Decomposing eukaryotic Hi-C (Chromosome Conformation Capture at the whole genome scale) data revealed that most of the inter-chromosomal interactions identified occur between chromatin regions with similar transcriptional activity, i.e. highly transcribed regions with highly transcribed regions and lowly transcribed regions with lowly transcribed regions. His analysis also revealed that telomere-telomere and centromere–centromere interactions are the second most prevalent type of interactions (Imakaev et al., 2012). The noted importance of interactions between centromeric and telomeric areas is in line with recent studies by Chen and co-workers (Kalhor et al., 2012). Finally, he discussed polymer models that can describe chromatin and emphasized that the appropriate model is in part determined by the density of the packed chromatin. For instance, in human and mouse the genome could be packed as a compact fractal globule, whereas in yeast the packing density is lower and the data are best fit by a random coil model or a mildly confined equilibrium globule (Fudenberg and Mirny, 2012).
A number of participants presented experiments that probed the physics, dynamics, and structure and segregation of bacterial chromosomes. Nancy Kleckner (Harvard University, USA) reported a study in which the E. coli nucleoid was visualized by wide field epifluorescence microscopy in living cells. Findings presented included information about the shape and organization of the nucleoid, identification of unexpected short timescale dynamics, and discovery of longer timescale dynamics related to sister chromosome separation (unpublished observations).
A second perspective on E. coli chromosome segregation was presented by Conrad Woldringh (University of Amsterdam, the Netherlands) who had proposed that one of the forces involved in chromosome segregation relies on anchoring of the genome to the membrane by coupled transcription, translation and translocation of transmembrane proteins (transertion) (Woldringh, 2002). This appealing model has been controversial in the field and clear evidence to prove or disprove it has been lacking. New data presented at the workshop revealed that segregation continued during run-off replication in the absence of growth and unambiguously demonstrated that transertion is not a driving force in chromosome segregation.
Suckjoon Jun (Harvard University, USA) is a leading proponent of a model whereby entropic forces are responsible for the ‘demixing’ of chromosomes during segregation (Jun and Mulder, 2006). Experiments probing the mechanical properties of nucleoid and testing some of the scaling predictions of his polymer models for nucleoid structure were presented. He demonstrated a microfluidic device that allowed single cells to be captured and gently lysed. Upon lysis nucleoids expand within the microchannel to occupy a volume much larger than their in vivo dimensions. Interestingly, nucleoids isolated from exponential-phase cells expanded much more than stationary-phase cells, which might be related to the presence of stationary phase-specific proteins (such as Dps and CbpA – see above). Next he was able to apply forces with a piston on the isolated nucleoid. These compression measurements revealed that bacterial chromosomes are ‘soft’, meaning that a 1000 times less force is needed for their compression than the turgor pressure in a cell. Suckjoon Jun also discussed studies on the effect of macromolecular crowding, a physical compaction force present inside cells due to the high concentration of proteins. Addition of polyethyleneglycol (generating a crowded environment) to nucleoids that were initially expanded after lysis, induced compaction to its in vivo dimensions (Pelletier et al., 2012).
In C. crescentus, and some plasmid systems, the molecular players involved in driving segregation appear to have been identified. W.E. Moerner, Stanford University, USA showed a super-resolution reconstruction of a ParA filament, which seems to drive the segregation of oriC in C. crescentus (Ptacin et al., 2010). The work was complemented by Kiyoshi Mizuuchi, NIH, USA who presented beautiful results demonstrating the reconstitution of the plasmid segregation system from both P1 and the F plasmid from purified components (manuscript in preparation). His work demonstrated processive motion in the apparent absence of a traditional filament-like spindle.
In addition to segregation, micron-scale nucleoid structure was addressed by a number of other speakers. Although it has long been accepted that the bacterial nucleoid appears to be divided into macrodomains, a mechanistic understanding of these domains has only begun to emerge. The group of Frédéric Boccard, CNRS, France recently discovered that the protein MatP is key to the organization of the Ter macrodomain in E. coli (Mercier et al., 2008). He presented follow-up work on the mechanism of organization. MatP must somehow exert its effects on Ter organization over large distances because its target matS sites are found every 35 kbp on average. Genetic, biochemical and structural studies revealed the molecular basis for the structural role of MatP. MatP consists of three functional domains: a large N-terminal helical domain required for specific matS binding, a central domain required for MatP dimerization and a C-terminal domain involved in Ter MD condensation. The MatP protein binds simultaneously to two matS sites (that can be separated by tens of kbp) and this binding provokes looping of the intervening DNA and condensation of the Ter MD in vivo (Dupaigne et al., 2012). Remarkably, MatP also appears to interact with the septum associated ZapB protein and this specific interaction is responsible for the extensive localization of the Ter MD at mid-cell (Dupaigne et al., 2012; Espeli et al., 2012).
The macrodomain organization of the nucleoid has physical consequences on nucleoid dynamics as well as structure. Marco Cosentino-Lagomarsino (CNRS, France) presented a high time-resolution tracking of E. coli chromosomal loci on short timescales (0.1–10 s). This work emphasized that the behaviour of chromosomal loci is very heterogeneous and this might be connected to a heterogeneous physical organization of the nucleoid, varying with time and cell state, and from cell to cell. Interestingly, mostly for the Ter macrodomain, short-time diffusivity correlates with their subcellular localization, with a lower apparent diffusion for loci near cell poles or mid-cell (unpublished observations).
Alessandra Carbone, CNRS, France approached the nucleoid from a different perspective and her work illustrated the importance of powerful bioinformatic tools. She showed that genes with a highly biased codon composition in E. coli are periodically distributed, suggesting an encoded three-dimensional genomic organization helping functional activities among which are translation and, possibly, transcription (Mathelier and Carbone, 2010).
The described studies highlighted the need to investigate the nucleoid as a dynamic entity, within which many DNA-based transactions coexist. It is thus of crucial importance to understand genome organization and its interplay with other processes at different scales of organization. This requires an integrative approach, nice examples of which came forth in the contributions by Georgi Muskhelishvili, International University Bremen, Germany and Dieter Heermann, Heidelberg University, Germany. Key to the work reported by both is a proposed order in the organization of genes along the genome and in space. Both speakers relied on knowledge of transcription regulatory networks from RegulonDB (http://regulondb.ccg.unam.mx/). A starting point in the studies of Georgi Muskhelishvili was the notion that the expression of NAPs, sigma factors and gyrase is temporally regulated and correlates with the relative position along the left and right replichores between Ori and Ter. In addition to the gene position also the regulatory targets of NAPs and that of stationary-phase sigma factors associated with RNA polymerase are biased to different parts of the genome and roughly localized to different macrodomains. Analogous to the studies of Pat Higgins that suggested differences in local superhelical density, he inferred the existence of a gradient in superhelical density along the replichores from origin to terminus (Sobetzko et al., 2012). Dieter Heermann discussed different types of polymer physics models that may be applied to describe genome organization. In particular, the introduction of loops into classical polymer models has been instrumental in describing folding of chromatin fibres in the eukaryotic nucleus (Mateos-Langerak et al., 2009). Such models might also be used to describe folding of the bacterial genome in loops, but ideally would include experimental information in relation to spatial folding and or the presence of loops (at defined sites?) formed by H-NS and SMC proteins. Here, he presented a model for the folded genome that relies on the assumption that the genes encoding transcription factors are spatially close to their targets (analogous to the ideas underlying the studies of Georgi Muskhelishvili). In addition, it uses spatial markers from fluorescence microscopy as constraints for the model (Wiggins et al., 2010). Following this approach they manage to reproduce the overall structure and observed ordering in chromosomes assuming the existence of domains of 85 kbp on average (Fritsche et al., 2012).
Generally, it was acknowledged that the nucleoid is a dynamic entity, that its components interact with each other differentially over time and that these interactions are encoded within the genome itself. As one of the participants, Georgi Muskhelishvili, put it: ‘the nucleoid is a living breathing thing – the brain of the cell’. This notion underscores the importance of approaches that are capable of addressing and integrating these different aspects. It seems a feasible issue to tackle for the community present at the workshop and first steps in this direction were indeed already taken. It also brings to the light aspects that were not or hardly addressed by participants during this meeting. For instance, what is the interplay of genome folding with replication? How does the structure of the genome affect damage induction, repair and mutagenesis and vice versa? What role does genome folding play in evolution, incorporation of foreign DNA by horizontal gene transfer etc.? Finally, it is fair to conclude that bringing together people from different fields at a meeting like this is not enough. No doubt the workshop has been able to generate more openness to (and understanding of) the types of data generated by the diverse approaches. However, at this stage it is hard to integrate much of the available data as the employed growth conditions are so widely different. Clearly new collaborative efforts are required especially in order to compare and integrate data from super-resolution imaging, ChIP and 3C. A next workshop, to be organized in 2 or 3 years, will show the outcome of newly established collaborations and hopefully address many of the outstanding questions.
The authors thank all meeting participants for their excellent contributions and the Lorentz Centre staff for their superb planning and organization. The workshop was financially supported by the Netherlands Organisation for Scientific Research (NWO), the Foundation for Fundamental Research on Matter (FOM), Leiden University and the University of Birmingham. The authors thank Mariliis Tark-Dame for comments on this report.