Skip to main content

Systems Biology of Mitosis

Final Report Summary - MITOSYS (Systems Biology of Mitosis)

Executive Summary:
The MitoSys project has generated a comprehensive mathematical understanding of mitotic division in human cells, a process of fundamental importance for human health. To create the critical mass and multidisciplinarity that was needed to achieve this ambitious goal, internationally leading mathematicians, biochemists/biophysicists and biologists working at twelve universities, research institutes, international organizations and companies in eight different European countries have collaborated. MitoSys focused on four biological modules that represent the most important aspects of the mitotic cell division process; (i) spindle assembly, (ii) the spindle assembly checkpoint (SAC) and kinetochores, (iii) segregation of mitotic chromosomes and (iv) mitotic exit. Computational models of these separate modules have been established. These individual models have been integrated into a comprehensive model of mitosis that combines these steps of mitosis with regulation by several key cell cycle regulators. To accomplish these tasks, the modellers were supported by biologists and physicists who used microscopic imaging, biochemistry, biophysics, single-molecule techniques and proteomics to generate kinetic and other quantitative data suitable for model building. To validate these mathematical models and to evaluate the relevance of the models for human health and disease, the MitoSys biologists have subjected selected key predictions from these models to rigorous in vivo tests by using conditional “knock-out” mice or other systems. MitoSys has compiled and disseminated its own data and models in a web-database, which will serve as a systems biology resource for the scientific community. To train other scientists in systems biology, MitoSys organized a training course on mathematical modelling. Finally, MitoSys produced three public exhibitions “Lens on Life” along with a documentary film “Meetings of Minds” that were displayed in three European cities (Rome, London and Heidelberg) to inform the general public about systems biology of mitosis and its relevance to health and disease.

Project Context and Objectives:
MitoSys proposed to generate a comprehensive mathematical understanding of mitotic division in human cells, a process of fundamental importance for human health. To create the critical mass and multidisciplinarity that is needed to achieve this ambitious goal, internationally leading mathematicians, biochemists /biophysicists and biologists working at twelve universities, research institutes, international organizations and companies in eight different European countries have collaborated on this project. Our mathematical modeling efforts focus on four biological modules that represent the most important aspects of the mitotic cell division process; (i) spindle assembly, (ii) the spindle assembly checkpoint and kinetochores, (iii) segregation of mitotic chromosomes and (iv) mitotic exit. Computational models of these separate modules have been integrated into a comprehensive model of mitosis that combines these steps of mitosis with regulation by protein kinases and ubiquitin ligases. To be able to achieve these tasks, the modellers are supported by biologists and physicists who use microscopic imaging, biochemistry, biophysics, single-molecule techniques and proteomics to generate kinetic and other quantitative data suitable for model building. To evaluate the relevance of the mathematical models for human health and disease, other biologists subject selected key predictions from these models to rigorous in vivo tests in conditional “knock-out” mice and in mouse oocytes and pre-implantation mouse embryos.
(i) Spindle Assembly
At the time when MitoSys started (June 2010) only very basic theories of the mitotic spindle were available. One of the most interesting models for spindle assembly at that time was the “slide-and-cluster” model (Burbank et al. Curr. Biol. 2007) from Tim Mitchison’s lab. Their model postulated that microtubules were of constant length, were only nucleated near the chromosomes, and considered their motion in 1D by two kinds of motors: Eg5 and dynein. One object of MitoSys was to conceive more accurate models that incorporate some key elements that are known to determine the organization of microtubules within the mitotic spindle, models that faithfully reproduce aspects of spindle assembly, dynamics and positioning as well as cytokinesis using combined biochemical, biophysical and imaging approaches.
(ii) Spindle Assembly Checkpoint (SAC) and Kinetochores
Workpackage 2 aimed to develop a quantitative model of the molecular mechanism of the spindle assembly checkpoint (SAC). The SAC controls the state of kinetochore microtubule attachment, delaying anaphase until bi-orientation of all sister chromatid pairs. WP2 aimed to formalize the dynamics of the checkpoint response, from its activation at unattached kinetochores to its “satisfaction” at sister kinetochores that have achieved bi-orientation. The workpackage consists of four distinct activities. The first activity consisted of measurements of protein interactions and protein dynamics in cells, in particular through the application of fluorescence correlation (or cross-correlation) spectroscopy. The second activity consisted of biochemical reconstitutions with the aim of reconstituting the complex interactions and reactions that take place at kinetochores. The third activity consisted of validation approaches aiming at testing model predictions in genetically tractable systems. Finally, the fourth activity aimed to sustain modeling the SAC.
(iii) Segregation of Mitotic Chromosomes
Human development and health critically depend on the ability of cells to pass their genome properly from one cell generation to the next. This is achieved by first duplicating the genome during DNA replication, and then segregating the two copies of the genome via chromosome segregation in mitosis. The latter process depends on physical connections between duplicated “sister” DNA molecules, called sister chromatid cohesion. Cohesion is mediated by cohesin complexes and enables the bi-orientation of chromosomes on the mitotic spindle, a pre-requisite for proper chromosome segregation in mitosis. The major objective of workpackage 3 was to describe how cohesin complexes interact with DNA to mediate cohesion and how the step-wise loss of cohesin from DNA enables chromosome segregation in mitosis. To achieve this objective, we have combined mathematical modelling, biochemical reconstitution, quantitative in vivo imaging and genetic validation experiments.
(iv) Mitotic Exit
Mitotic exit involves an extensive internal re-organization of cells, resulting in the disassembly of the mitotic apparatus and a re-assembly of interphase organelles. Mitotic exit is known to proceed by ubiquitin-proteasome-mediated degradation of mitotic determinants and by reversal of mitotic phosphorylations, yet the exact role of candidate regulators and the signaling kinetics underlying mitotic exit were not well understood before this project. Workpackage 4 aimed at identification, quantification, and modeling of the regulatory networks underlying mitotic exit in human cells. This requires a detailed biochemical and cell biological characterization of the candidate regulators, and development of automated microscopy techniques and computational image analysis software for kinetic measurements of mitotic exit. Chemical and biophysical perturbation assays enable to probe for regulatory network responses after specific signaling events. The experimental data generated in this workpackage serve as a basis to implement a mathematical model that recapitulates hallmarks of mitotic exit, including unidirectionality, irreversibility, sequential substrate dephosphorylation, and feed-back-loops between kinases and phosphatases. Model tissue culture cell lines are suitable for initial characterization of mitotic exit signaling, but key findings need to be validated in vivo using genetic mouse models.
A Comprehensive Model of Mitosis
Mitosis is characterized by intense interactions of the biochemical regulatory network controlling cell cycle progression with chromosomes and their segregation apparatus (mitotic spindle). At the beginning of mitosis, the activation of the cyclin-dependent protein-kinase (Cdk1) instructs the spindle to be formed and chromosomes to be condensed, which are essential prerequisites of the chromosome segregation process. A feedback signal from the uncaptured chromosomes blocks the biochemical device until all the chromosomes are aligned along the spindle. Once chromosome congression is completed a proteolytic machinery is activated that dissolves the cohesive glue holding chromosomes together, and inactivates Cdk1 at the same time. Finishing mitosis also requires the activation of phosphatases that revert protein-phosphorylations caused by the Cdk1 protein-kinase. The ultimate goal of MitoSys was to construct a comprehensive mathematical model for mitotic progression of mammalian cells by using differential equations that capture the interactions of the molecular components to describe the intrinsically dynamic process of mitosis.
Development of New technologies
Obtaining quantitative data on relevant proteins is a prerequisite for systems biology. Workpackage 5 aimed at developing new technologies which enable the generation of quantitative data on human proteins, in particular on their abundance and functional dynamics in high spatial and temporal resolution in living cells, and the stoichiometry of their subunits in protein complexes.
Dissemination activities
MitoSys compiles and disseminates not only its own data and models, but also datasets from other sources, in a web-database, which will serve as a systems biology resource for the scientific community. To train other scientists in systems biology, MitoSys organized a one-week long training course on mathematical modelling. Finally, MitoSys organized three public exhibitions in three European cities with the aim of informing the general public about systems biology of mitosis and its relevance to health and disease.

Project Results:
WP1: Spindle Assembly
The mitotic spindle is a self-organized structure of dynamic microtubules, which nucleate at the centrosomes and the chromosomes, interact with the chromosomes at the kinetochores and with the cytoskeleton at the cell cortex. It must ensure that kinetochores attached correctly so that spindle assembly checkpoint (SAC) is satisfied, and that sister kinetochores separate so that one of each of the sister chromatids goes to each pole of the spindle. Mitotic spindles function as a consequence of the temporal and spatial interactions of a large set of proteins. These interactions take place not only through signaling mechanisms such as kinases and small G-protein pathways, but also as a result of the forces generated by the dynamics of the filaments and motor proteins. We have taken multidisciplinary approaches to study spindle assembly, dynamics, and positioning.
Modelling:
Together with the Joanny group (Institut Curie, Partner 4), the Jülicher group (MPG-PKS, Partner 5) has developed a coarse grained model for spindle assembly, which is based on the idea that the spindle has liquid-like material properties. These material properties allow us to separate the problems of how shape and size of the spindle are determined. Spindle size is determined by material balance in the spindle, while spindle shape is determined by force balance. Mass balance, which sets spindle size, is governed by microtubule nucleation, polymerization, and depolymerization. The spindle shape is set by the force balance and involves stresses generated by motors as well as surface stresses related to surface tension. The source term of the mass balance – the nucleation of microtubules – is not yet fully understood. In accordance with experimental observations we assume that microtubules are mainly nucleated by chromatin-dependent pathways. This model was tested by perturbations of the growth velocity. We extended our passive model to study force balances in metaphase spindles under conditions with motor-generated forces and couple it with the mass balance description. The model is based on the theory of active nematic liquid crystals. The resulting partial differential equation we solve with the finite-element method, which is a suitable tool for arbitrary geometries. This allows us to formulate the problem as a shape optimization problem.
To take the right modeling assumptions it is crucial to have detailed information about the microtubule arrangement, especially at the centrosomes where microtubules are being nucleated and the forces are concentrated. For this purpose we analyzed the formation of centrosomes. The concept of phase separation allows us to describe the growth dynamics of centrosomes with an autocatalytic chemical transition between the centrosome material in a form that is soluble in the cytosol and a form that tends to phase separate. The model describes data of various cell sizes and perturbed cell volumes and centrosomes. The microtubule arrangement we modeled is based on data from high-pressure freezing. Although the temporal information is lost in this preparation, we have detailed information about the microtubules in terms of their local density, amount of ordering, and length distribution.
The Nedelec group (EMBL, Partner 2) took a complementary approach to model spindle assembly taking into consideration the elements such as the assembly dynamics of microtubules, the activity of cutting and chewing enzymes, and more realistic molecular pathway for microtubule nucleation. Over the course of the project, we have proposed increasingly refined models of the mitotic spindle assembly.
Model 2010:

Our model published in 2010 was the first to include the assembly dynamics of the microtubules, with realistic growth and shrinkage rate. Each microtubule would have a different length, and eventually would completely disassemble, as observed in real spindles. Constant nucleation was modeled in a crude way to keep the number of microtubules in the spindle at steady state (Loughlin et al. JCB 2010).
Model 2012:

In 2012, a refined 2D model was proposed. Microtubule catastrophe was simulated using a detailed stochastic GTP hydrolysis scheme (Brun et al. PNAS 2009). In addition, a continuous scalar field was implemented to keep track of the accumulation of material at the poles of the spindle (this is represented as a blue-white gradient on the picture).
Model 2013:

In 2013, we made a 3D model that included chromosomes for the first time. It was intended to capture the organization of the microtubules in the directions orthogonal to the pole-to-pole axis. The computational cost were however expensive, which limited our ability to explore the model to its full extent. This prompted us to focus temporarily on a smaller system: the yeast spindle.
Model 2014:

In 2014, we published a geometrically realistic 3D model of a yeast spindle at anaphase. It accurately captured the organization of the microtubules in the directions orthogonal to the spindle axis. The model was used to decipher the link between form and function, more specifically to understand how the organization of the microtubules within the spindle determines its ability to produce the forces necessary to segregate the chromosomes (Ward, Roque et al. eLife 2014).
Model 2015:

We are now working on a composite model in which the microtubule motions are simplified, allowing us to simulate 10000 microtubules, which is the quantity estimated for spindles in Human cells. The dynamics of microtubule assembly, and disassembly from the minus-ends are faithfully modeled. The two pathways generating new microtubules are realistic, with nucleation occurring near the chromosomes, and microtubule amplification occurring in a more dispersed manner across the spindle.
Biochemistry / Biophysics
In order to provide experimental data for theoretical modelling, protein required for spindle assembly, dynamics, function and positioning are produced and analyzed in vitro.
The Howard lab (MPG-CBG, Partner 5) isolated two spindle assembly proteins XMAP215/ch-TOG and EB1/MAPRE1. These two proteins are microtubule +Tip interacting proteins and their interaction is shown to be important for proper spindle assembly in Xenopus egg extract. We have found that EB1 and XMAP215 act synergistically on growth speed. Size exclusion chromatography experiment revealed that these two proteins do not directly interact with each other and therefore we conclude that the observed synergetic effect of EB1 and XMAP215 on microtubule growth rates is not a consequence of formation of a complex of the two proteins, which then binds to the microtubule end. Rather, we think that the effect comes through the combined action of the two proteins that takes place only at the growing microtubule end.
EB1 is a potent catastrophe factor alone and in the presence of XMAP215. Our results show that the growth-promoting and catastrophe-inducing properties of EB1 regulate the mitotic progression by modulating the microtubule dynamics together with chTOG and Kif18A.
Using purified recombinant XMAP15, the main microtubule polymerase in Xenopus egg extract, the Hyman lab (MPG-CBG, Partner 5) has been able to show that spindle length correlates linearly with XMAP215 activity. However, the microtubule lifetime and density as well as spindle shape are not modulated by XMAP215 activity. Based on in vitro characterization of XMAP215, spindle reconstitution, quantitative analysis of microtubule dynamics, and theory we are able to describe spindle length by a mass balance model assuming liquid-like properties of the spindle (Reber et al., 2013).
For the generation and validation of various models established by the consortium labs, a large number of human BACs (bacterial artificial chromosomes) expressing key players of mitosis including spindle components, kinetochore-assembling proteins, phosphatases, and kinases involved in cell cycle regulation have been tagged and transfected into HeLa and mouse embryonic stem cells, respectively. During the entire project time, the Hyman lab has provided close to 300 stable BAC transgenic cell lines to the consortium labs. In addition, the Hyman group has generated a comprehensive resource to characterize the superfamily of motor proteins in more detail. This protein family consists of multiple kinesin and myosin motors that are known to be required for intracellular transport, cell motility and mitosis. However, their complex interplay as well as their role in cell division have not been characterized so far. For this purpose, a BAC transgeneOmics resource for 71 of 82 annotated human kinesin and myosin genes have been established that consists of 227 stable BAC cell lines (Hela) expressing N- and C-terminal tagged mouse and human BACs. All the lines have been characterized by expression and localization analyses. Further investigation by affinity-purification mass spectrometry revealed several new candidate protein–protein interactions (Maliga et al., 2013).
The Schuh group (MRC-LMB, Partner 11) found that asymmetric spindle positioning in mouse oocytes is driven by a cytoplasmic actin network that is nucleated by cooperation between the actin nucleation factors Formin-2 and Spire1/2 (Pfender et al., Curr. Biol. 2011). The Carlier group (CNRS, Partner 8) has elucidated the molecular mechanism by which Formin 2 synergizes with Spire to stimulate the assembly of a dynamic cytoplasmic actin meshwork, which is required for translocation of the meiotic spindle toward the cortex of the oocyte in assymetric division.
Model validation:
The Schuh group in vivo validated key aspects of models for spindle organization, function and positioning in mouse oocytes or embryos.
In collaboration with Takashi Hiiragi’s and Jan Ellenberg’s labs (EMBL, Partner 2) the Schuh group analysed how the spindle is assembled and organized in preimplantation mouse embryos (Courtois et al., J. Cell Biol. 2012). This study confirmed several key aspects of existing models for spindle organization and function from other cell types and in vitro data. For instance, it showed that bipolar spindle assembly in early mouse embryos also relies on the motor protein kinesin-5; it confirmed that the spindle length decreases as the cells of the embryo get smaller; and that the size of the microtubule organizing centres (MTOCs) at the spindle poles correlate with cell size. The latter two observations confirmed models that were derived based on work in C. elegans embryos in the Hyman group.
The Schuh group also studied how the spindle is organized in mouse oocytes. Their work revealed that spindle assembly in oocytes involves extensive MTOCs, which is essential for timely spindle assembly and chromosome alignment (Clift and Schuh, Nat. Comm. 2015). A similar restructuring of MTOCs has previously been observed in mitotic cells at the transition to interphase, when the interphase microtubule network is reestablished.
The Schuh group also carried out the first systematic screen for meiotic genes in mammals, and identified several new genes essential for spindle assembly and function in oocytes (Pfender et al., Nature 2015).
In addition, the Schuh group studied how the spindle is positioned in cells. They found that asymmetric spindle positioning in mouse oocytes is driven by a cytoplasmic actin network that is nucleated by cooperation between the actin nucleation factors Formin-2 and Spire1/2 (Pfender et al., Curr. Biol. 2011). Spindle movement along this network is facilitated by myosin-dependent pulling from the spindle poles and outward-directed myosin Vb-dependent movement of Rab11a-positive vesicles, which keep the network dynamic and help to modulate its density by localizing and transporting the actin nucleation factors (Schuh, Nat. Cell Biol. 2011; Holubcova et al., Nat. Cell Biol. 2013).
The Malumbres group (CNIO) focused on studying in vivo the function of major kinases - Aurora and Polo kinases in regulating spindle maturation and dynamics. First we demonstrated that Aurora kinases, Aurora A and Aurora B, are regulated by SUMOylation during mitosis, adding a new level of regulatory mechanisms to the function of these enzymes and spindle dynamics. We also generated knockout models for Aurora A and Tpx2, its major activator during spindle function. These models corroborated the relevance of the Aurora A/Tpx2 in spindle dynamics and in maintaining genomic stability. Furthermore, conditional ablation of Aurora A in tumors resulted in tumor regression thus supporting the use of this holoenzyme as a cancer target. A similar model for Plk1 function confirmed the relevance of this molecule in centrosome separation and spindle function. In addition, we identified Plk1 as a major kinase regulating the elasticity of cells, a new function that may have specific implications in mitosis but that also has broader implications as Plk1 heterozygous mice died of cardiovascular defects due to problems in regulating tension in the major blood vessels.

WP2: Spindle assembly checkpoint (SAC) and the kinetochores
The SAC signal originates at kinetochores, multi-subunit assemblies that connect chromosomes to microtubules of the mitotic spindles. All SAC proteins are recruited to kinetochores during late prophase and early prometaphase. SAC activation at kinetochores is demonstrated by the fact that mutations or depletions of kinetochore proteins prevent SAC activation. SAC function converges on the production of an effector named the mitotic checkpoint complex (MCC), which halts anaphase progression by targeting the anaphase promoting complex or cyclosome (APC/C), an ubiquitin (Ub) ligase whose function is required for mitotic exit (Figure 2.1). By targeting the APC/C, the SAC prevents mitotic exit, ultimately providing the kinetochore with more time to complete attachment to microtubules. Thus, the SAC is essentially a synchronization device that couples the timing of mitotic exit to the completion of kinetochore-microtubule attachment.

The MCC consists of the four proteins Mad2, Cdc20, BubR1, and Bub3, with the latter two forming a tight constitutive complex. Among the best-guarded and crucial secrets of the SAC is whether the generation of the MCC by SAC proteins is catalytic, and whether kinetochores act as catalysts in the production of the SAC effector. The question can be re-formulated as follows: Is the rate at which the MCC assembles to reach its steady-state concentration in the mitotic cell determined solely by the on-rate of association of the MCC subunits, or is this rate accelerated by other SAC proteins and kinetochores? The question is highly relevant, because our previous studies demonstrated that the on-rate of the interaction of Mad2 with Cdc20, which leads to the formation of the MCC sub-complex Mad2:Cdc20, is extremely small, likely rate-limiting, and unlikely to allow the timely generation of the SAC effector in the amounts required for halting mitotic progression (Simonetta et. al. PloS Biol., 2009). Thus, catalytic activation of MCC formation is expected, but the detailed molecular mechanism had been missing.
To help answering this question, we proposed within MitoSys to generate a system to measure the rate of MCC production in vitro by means of FRET sensors responding to the interaction of two MCC components. We had proposed that this tool would have allowed us to measure the basal rate of MCC production, and therefore to ask whether the addition of other SAC components or kinetochores had any influence on the basal rate of MCC production. Thus, the ambition of this WP was to be able to reconstitute SAC signaling in vitro with purified components, ultimately allowing answering the fundamental questions delineated above. We proposed that answering this crucial question would allow realistic and predictive modeling of the SAC pathway.
To achieve our goals, we initially developed recombinant expression systems for SAC and kinetochore components, in line with the activities initially described in Deliverables 2.1 2.5 and 2.6. Figure 2.2 summarizes our considerable achievements in the expression of kinetochore and MCC proteins that were made available to the project.

To accomplish Deliverable 2.3 we developed several fluorescence labeling schemes for Mad2, BubR1:Bub3, and Cdc20, using both genetically encoded tags (e.g. mTurquoise) and organic dies (e.g. TAMRA). The latter were inserted in the proteins of interest by protein-peptide fusions with the Sortase system (Popp et. al., Curr Protoc Protein Sc 2009). Inspired by the crystal structure of the MCC (Chao et. al., Nature, 2012) we were able to generate two fluorescence resonance energy transfer (FRET) sensors to monitor MCC assembly. Titrations of the FRET signal as a function of concentration allowed the derivation of binding isotherms for each of the two sensors, which, as expected, converged on the same equilibrium point. This indicated that MCC establishment requires all three components (Mad2, Cdc20, and the BubR1:Bub3 sub-complex) and behaves therefore as a typical cooperative system. The overall dissociation constant (Kd) was 3 nM (Figure 2.3) a number that needs to be compared with the cellular concentration of the MCC components, estimated to be in the order of 100 nM. Thus, our measurements suggest that formation of MCC may occur spontaneously in the mitotic cytosol.

With FRET sensors in hand, we could proceed to answer the first of our two crucial questions, namely what is the rate of MCC establishment and whether other SAC components contribute to the modulation of such rate. Figure 2.4 shows experiments in which the components of one of the two MCC sensors described in Figure 2.3 were mixed in a fluorimeter and the accumulation of the (FRET signal indicating formation of the) complex monitored as a function of time. At the specific concentration of MCC components used in this experiment (100 nM), the half time of MCC formation is 150 minutes (‘) (Figure 2.4). Very excitingly, the halftime (t1/2) of the reaction reduces to 30 seconds (‘’) when a defined set of additional unlabeled SAC components is added to the reaction together with ATP. The reaction is powered by ATP, and therefore probably involves the kinase activity of Mps1 and Bub1.

The dramatic acceleration of MCC formation revealed by these experiments provides evidence that SAC components working upstream of MCC (e.g. the Mps1 and Bub1 kinases and the Mad1:Mad2 complex) function as catalysts for MCC assembly. Previously, we had proposed that the Mad1:Mad2 complex may act as a catalyst to assemble Mad2:Cdc20 complexes (Simonetta et. al. PloS Biol., 2009), but the low rate of acceleration observed in the much simpler in vitro reconstitution described in these previous studies had led us to predict that other contributions might exist. The stunning 300-fold acceleration of MCC assembly observed in our studies demonstrates the correctness of this prediction and anticipates our full characterization of the catalytic mechanism of MCC establishment, which we anticipate to be able to complete in the next 2-3 months.
The studies discussed in the previous paragraph were obtained with soluble versions of the “catalytic” components Bub1, Mps1, and Mad1:Mad2 and depend linearly on the concentration of these components (likely indicating that one of the components, which we suspect to be the Mad1:Mad2 complex, is rate-limiting). Are kinetochores contributing to the catalytic activation of MCC components? Kinetochores that are not attached to microtubules recruit these components and concentrate them, which is likely to favor combination of MCC components by sheer mass action combined with catalysis. It is also possible that kinetochores trigger conformational changes of the catalysts or the substrates that further accelerate MCC establishment.
The availability of recombinant kinetochore components will allow us to answer this question in vitro. For instance, we will add “kinetochore beads” in reactions in which we added rate-limiting concentrations of SAC catalysts and ask if the rates of MCC establishment are accelerated by kinetochore proteins. We had hoped to be able to carry out these experiments already within the time limit of MitoSys. While this has not been possible, due to the extreme technical challenges associated with these experiments, which caused some delay in the achievement of some deliverables, it is fair to say that MitoSys has allowed achieving a truly fundamental result: the reconstitution of SAC signaling in vitro. This, in turn, allows disclosing the molecular basis of its mechanism of action, until now a well-guarded secret of life.
To verify various aspects of the SAC model, we generated Aurora B conditional knockout mice. We first confirmed the relevance of this kinase in modulating the proper attachment of microtubules to kinetochores and the error correction mechanism. Our data also suggested that Aurora B is required, but not essential, for the SAC function with independence of the error correction function. We have also uncovered an essential role for Aurora C, an additional member of the Aurora family poorly characterized, during early embryonic development. We also generated Cdc20 conditional knockout mice and used these animals to demonstrate the essential role of Cdc20 for anaphase onset as the major and unique target of the SAC. In the absence of Cdc20, cells arrest in metaphase until they die. These data indicate that mitotic slippage is triggered by slow but significant activation of the APC/C-Cdc20 even in conditions where the SAC is not satisfied. These data also opened new directions to investigate how cells survive upon prolonged mitotic arrest, a process that may have relevant implications in cancer therapy.

WP3: Chromosomes
Eukaryotes pass their genomes from one cell generation to the next by first replicating DNA in S-phase, resulting in chromosomes containing two identical DNA molecules (sister chromatids). Cells then bi-orient these chromosomes on the mitotic spindle and segregate sister chromatids in anaphase, enabling the generation of daughter cells with correct ploidy. This mechanism of genome inheritance critically depends on the ability of cells to generate physical linkages between sister chromatids during DNA replication. This sister chromatid cohesion resists the pulling forces of the spindle, thereby enables bi-orientation of chromosomes and is thus essential for chromosome segregation. In the 1990s, genetic screens in budding yeast identified seven genes essential for cohesion(Guacci, Koshland et al. 1997, Michaelis, Ciosk et al. 1997). Four of these encode subunits of a complex called cohesin (Smc1, Smc3, Scc1/Mcd1, Scc3, (Losada, Hirano et al. 1998)), two encode subunits of a complex needed to load cohesin onto DNA (Scc2, Scc4) and one encodes an acetyltransferase (Eco1/Ctf7) that enables cohesin to maintain cohesion by acetylating Smc3 (reviewed in (Nasmyth 2011)). Smc1 and Smc3 associate with each other via hinge domains, which are connected via 50 nm long coiled coils with globular ATPase domains (Figure 3.1: A-C). These associate with Scc1 to form tripartite rings that in budding yeast interact with yeast mini-chromosomes by entrapping DNA inside the cohesin ring(Haering, Farcas et al. 2008). Cohesin is therefore thought to mediate cohesion by entrapping both sister DNAs. Once chromosome bi-orientation is complete, the protease separase cleaves Scc1. This opens cohesin rings (Figure 3.2) and releases DNA from them, so that sister chromatids can be separated(Uhlmann, Wernic et al. 2000).

Figure 3.1: (A) Illustration of cohesin subunit topology (SA1 and SA2 are vertebrate orthologs of yeast Scc3). (B) Subunits of recombinant human cohesin separated by SDS-PAGE and stained with silver. (C) Electron micrograph of cohesin as in B (from (Huis in 't Veld, Herzog et al. 2014)). (D) Illustration of how cohesin is thought to entrap DNA and how these interactions are regulated (for simplicity, only cohesin’s three ring-forming subunits are shown; gray line, DNA).
Cohesin can also be released from DNA by a separase-independent mechanism which depends on the protein Wapl (Figure 3.1 D (Kueng, Hegemann et al. 2006)). During S-phase, “cohesive” cohesin complexes become stabilized on DNA by an inhibitor of Wapl, called sororin(Schmitz, Watrin et al. 2007, Nishiyama, Ladurner et al. 2010). Sororin is recruited to cohesin, which has been acetylated on Smc3 by the Eco1 orthologs Esco1 and Esco2(Lafont, Song et al. 2010, Nishiyama, Ladurner et al. 2010). Sororin prevents Wapl-mediated release of cohesin from DNA, which could otherwise destroy cohesion precociously(Nishiyama, Ladurner et al. 2010). In early mitosis, sororin is removed from chromosomes and Wapl-mediated cohesin release is activated, resulting in loss of cohesin from chromosome arms(Liu, Rankin et al. 2013, Nishiyama, Sykora et al. 2013), a process that is thought to facilitate sister chromatid separation by promoting de-catenation of sister chromatids(Tedeschi, Wutz et al. 2013, Haarhuis, Elbatsh et al. 2014). At centromeres, cohesin is protected from Wapl, and these complexes are destroyed in metaphase by separase, as in yeast(Hauf, Waizenegger et al. 2001).

Figure 3.2: Cohesin is thought to be released from DNA via ring opening. (A) Cohesin containing Scc1 with a PreScission protease site, treated without or with protease, and analyzed by SDS-PAGE-silver staining. (B, C) Models and EM micrographs of cohesin cleaved as in A (B) or containing an Scc1 mutant defective in DNA “exit gate” closure (C). (D) Chromosome spreads from HeLa cells expressing wild-type or only the DNA “exit gate” mutant Scc1 (as in C); bar 10 μm17.
The main objective of the MitoSys WP3 ‘Chromosomes’ was to provide a systems level description of how stepwise loss of sister chromatid cohesion from prophase until metaphase leads to the synchronous separation of sister chromatids in anaphase. To achieve this objective, we took four different approaches: (i) mathematical modelling, (ii), biochemical reconstitution, (iii) quantitative in vivo imaging, and (iv) in vivo validation of models and hypotheses through genetic experiments.
The mathematical modelling was initially based on data curated from the literature and was continuously refined as additional data were generated during the course of the project. Mathematical models were generated which describe how cohesin interacts with DNA dynamically before DNA replication and how inhibition of Wapl by sororin during DNA replication leads to stabilization of cohesin on DNA and an increase in the number of DNA bound cohesin complexes (unpublished). This model is partly based on measurements of absolute copy numbers of cohesin complexes in cells which were determined by quantitative imaging and mass spectrometry techniques during this project (see below). Models were also generated which describe how the cohesin protease separase is activated during mitosis, an event that leads to complete removal of cohesin from DNA, loss of sister chromatid cohesion and chromosome segregation. These results were published in Cundell et al., Mol. Cell, 2013; Rattani et al., Curr. Biol. 2014 and Hellmuth et al., Mol. Cell 2015.
To support the mathematical modelling with experimental data, we performed biochemical reconstitution experiments. For this purpose we generated various recombinant forms of cohesin complexes and its regulatory proteins. To facilitate the generation of different recombinant protein complexes, we first developed a method for the rapid assembly of multiple cDNAs into one Baculoviral genome, which can be used for protein expression by infecting cultured insect cells. We call this method biGBac (manuscript in preparation). We then used this approach for the generation of wild type and mutated cohesin complexes and used these samples for enzymatic and structural analyses of cohesin (Figure 3.2; Gligoris et al., Science 2014; Huis in ‘t Veld et al., Science 2014; Ladurner et al., Curr. Biol. 2014) and to reconstitute cohesin-DNA interactions (unpublished). We also developed assays for measuring biophysical properties of chromatin (Vlijm et al., Cell Rep. 2015) and will in the future use related approaches to measure physical properties of cohesin-chromatin interactions.
To complement the biochemical work done in vitro, we performed quantitative in vivo imaging experiments in cells. For this purpose we developed cell lines and techniques, in which the interactions between cohesin and DNA could be analyzed in living cells by a technique called fluorescence recovery after photobleaching (FRAP). We used these FRAP assays to analyze how Wapl, sororin and the ATPase activity of cohesin control the interactions between cohesin and DNA (Huis in ‘t Veld et al., Science 2014; Ladurner et al., Curr. Biol. 2014; Ladurner, Kreidl et al., manuscript in preparation). We also used another technique called fluorescence correlation spectroscopy (FCS) which allows the direct measurement of absolute copy numbers of cohesin complexes and other molecules in cells. We controlled the validity of these results obtained with FCS by comparing them to cohesin copy number data measured independently by quantitative mass spectrometry. We are currently using data obtained by both techniques to improve our mathematical models.
Finally, to analyze if our hypotheses are correct and to test predictions made by our mathematical models, we performed genetic validation experiments. For this purpose we generated mouse models in which subunits of cohesin or cohesin regulatory proteins can be inactivated by conditional deletion of the corresponding gene. We then performed biological experiments to analyze the loss-of-function phenotypes observed in cells derived from these models and compared them to our hypotheses and models. By using this approach we could confirm that Wapl has an essential role in controlling cohesin-DNA interactions (Tedeschi et al., Nature 2013) and that soroin is required for the maintenance of sister chromatid cohesion (Ladurner, Kreidl et al., manuscript in preparation).

WP 4: Mitotic Exit
In a collaborative effort involving biochemists, biophysicists, cell biologists, mouse geneticists, and mathematical modelers, we achieved a detailed characterization of mitotic exit regulators and developed a mathematical model recapitulating hallmarks of mitotic exit. Key experimental findings and predictions from computer simulations were validated by genetic mouse models. The model was further linked to mitotic subprocesses studied in other WPs to achieve an integrated mathematical model of mitosis. In the following, we summarize the most important results.
Regulation of the human mitotic exit phosphatase PP2A-B55a
An image-based RNAi screen in the Gerlich laboratory had previously identified PP2A-B55alpha as a key mitotic exit phosphatase counteracting Cdk1 kinase in human cells. In the MitoSys project, regulatory mechanisms of PP2A-B55alpha were investigated. By purification of tagged PP2A-B55alpha subunits, expressed from endogenous promoters, and mass spectrometric analysis, the laboratories of Jan-Michael Peters (IMP), Anthony Hyman (MPG), and Daniel Gerlich (ETH Zurich/IMBA) identified mitosis-specific phosphorylation sites on all subunits of the PP2A-B55alpha heterotrimeric complex (Figure 4.1). Biochemical analysis of a phospho-mimicking mutation on a Cdk1 consensus phosphorylation site on B55alpha revealed that it negatively regulates the efficiency of B55alpha binding to the PP2A core dimeric complex, indicating a potential mechanism of mitotic PP2A inactivation. This regulatory mechanism was incorporated into a model of mitotic exit (Figure 4.1) which the Novak laboratory (UOXF) implemented as a set of ordinary differential equations (ODEs). Simulations derived from this model reflect key features of mitotic exit, including unidirectionality and sequential dephosphorylation of mitotic substrates during mitotic exit.

Figure 4.1: Cell cycle-dependent regulation of PP2A-B55α. (a) Phosphorylation sites on PP2A-B55α identified by mass spectrometry highlighted in red on the 3D structure of PP2A-B55α. The abundance of phosphorylated peptide in the mitotic sample was estimated by peak area quantification of the elution profiles. The mitotic increase of phosphorylation abundance (indicated in brackets) was estimated by comparing the normalized peak area quantifications of phosphorylated peptides in interphase and mitotic samples. (b) Wiring diagram of mitotic exit control.

Identification and functional analysis of the Aurora B-counteracting mitotic exit phosphatases
Aurora B kinase controls several important events at different cellular locations during mitotic progression. Its function during mitotic exit depends on relocalization from chromosomes to the central spindle. How Aurora B-counteracting phosphatases remove Aurora B-dependent phosphorylations from chromosomes during mitotic exit was incompletely understood. The Gerlich laboratory (ETH/IMBA) conducted an image-based RNAi screen using a biosensor to identify phosphatases counteracting Aurora B during mitotic exit. Two PP1 regulatory subunits were found to be rate-limiting for the removal of Aurora B-dependent phosphorylations from anaphase chromosomes, and their depletion perturbed faithful anaphase chromosome segregation (Figure 4.2). Aurora B substrate dephosphorylation was not affected by PP2A depletion, and conversely, depletion of the PP1 regulatory subunits did not affect mitotic exit events that depend on PP2A. Thus, two distinct regulatory networks drive different mitotic exit events. These findings were published in Wurzenberger et al., JCB, 2012.
Figure 4.2: PP1-regulatory subunits Sds22 and RepoMan counteract Aurora B to promote faithful chromosome segregation during mitotic exit. (a) Llive HeLa cells expressing Aurora B FRET biosensor, 42h after transfection of control siRNA, or siRNAs targeting Repo-Man or Sds22. (b, c) Kinetochore tracking in HeLa cells stably expressing EGFP-CENP-A. (d) Quantification of transient segregation pausing. Bars: 10 μm. Wurzenberger et al., JCB, 2012.

Kinetics of spindle assembly checkpoint signaling during mitotic exit
Mitotic exit is normally initiated only after all chromosomes have correctly attached to the mitotic spindle, which satisfies the spindle assembly checkpoint (see WP 2). To study the signaling networks that initiate mitotic exit, the Gerlich laboratory (ETH/IMBA) developed a biophysical perturbation method: using a high-energy pulsed laser, individual chromosomes were detached from the metaphase spindle, which activates spindle assembly checkpoint (Figure 4.3).
Figure 4.3: Laser microsurgery to detach individual chromosomes. (a) Live metaphase HeLa cell expressing H2B-mCherry and mEGFP-α-tubulin. Kinetochore fiber microtubules were cut with a pulsed 915 nm laser at the area indicated by the white line. (b) Quantification of Mad2 accumulation of detached chromosomes, in cells expressing H2B-mCherry and Mad2-EGFP. Bar, 10 μm. Adapted from Dick and Gerlich, NCB 2013.

The majority of cells re-aligned chromosomes after their detachment from the spindle and then initiated mitotic exit after some delay, as expected based on our knowledge about the spindle assembly checkpoint. Surprisingly, about one third of the cells entered anaphase in the presence of unattached chromosomes, indicating that the spindle checkpoint may not completely suppress mitotic exit. Direct measures of mitotic exit progression by fluorescent substrate degradation kinetics showed that a residual activity of the anaphase promoting complex was substantially higher in the presence of a single unattached chromosome, as opposed to a complete detachment of all chromosomes from the spindle. Thus, the spindle assembly checkpoint generates a graded signal that correlates with the number of unattached chromosomes. Cells with only one unattached chromosomes are therefore prone to exit mitosis prior to correcting the attachment defect, which would result in the generation of an aneuploid karyotype. Further characterization of the spindle assembly checkpoint signaling kinetics at distinct stages of mitotic exit revealed that active Cdk1 is required for spindle assembly checkpoint signaling. These findings were published in Dick and Gerlich, NCB, 2013, Vazquez-Novelle et al., Curr Biol, 2014, and Rattani et al., Curr Biol, 2014.

An integrated mathematical model recapitulates key features of mitotic exit
The Novak lab (UOXF) has developed mathematical models for the Spindle Assembly Checkpoint (SAC) and for the regulation of the Cdk1 counter-acting phosphatase, PP2A-B55. The SAC-model is based on a positive feedback loop by which Cdk1 promotes checkpoint activation, which stabilizes its activator, cyclin-B (He et al. PNAS, 2011). This assumption has been experimentally confirmed by the Gerlich Lab. The regulation of PP2A-B55 is described by the recently discovered BEG (B55/Endosulfine/Greatwall) pathway by which Cdk1-CycB inactivates its counter-acting phosphatase, PP2A-B55. The combination of the SAC- and the BEG-model (reported in detail in Workpackage 6) provided us with a detailed kinetic description of the mitotic exit process of mammalian cells. Mitotic exit is initiated at the end of prometaphase by satisfaction of the SAC, which activates the Anaphase Promoting Complex (APC/C). The APC/C promotes the degradation of its anaphase substrates, securin and CycB, during metaphase, when SAC can be still reactivated. At the meta/anaphase transition separase is relieved from Securin and CycB inhibition and cleaves cohesins, but the BEG pathway still keeps PP2A-B55 inactive. PP2A-B55 is only relieved from inhibition once Cdk1 activity drops below a lower threshold. PP2A-B55 dephosphorylates Cdk1-targets and initiates the events of cytokinesis as well.

Validation of mitotic exit model by mouse mutants
The generation of Cdc20 conditional knockout mice allowed us to identify a critical function for Mastl/Greatwall to inhibit PP2A complexes and to maintain the mitotic state. In the absence of Cdc20, cells can only exit from mitosis when both Cdk1 and Mastl are inhibited, indicated that the mitotic exit phosphatase, PP2A, must be reactivated during this process.
The Malumbres lab (CNIO) studied the control of mitotic phosphorylations by characterizing the relevance of Mastl using genetic models in the mouse. Genetic ablation of Mastl results in a mitotic collapse in which cells cannot maintain a proper mitotic state due to hyperactivation of these phosphatases. Mastl shuttles from the nucleus to the cytoplasm during prophase in a Cdk-dependent manner to inhibit cytoplasmic PP2A complexes. This kinase is not required for mitotic entry but its absence results in a mitotic collapse due to PP2A-B55 activity after nuclear envelop breakdown. In addition, mitotic exit required the inactivation of Mastl to reactivate PP2A-B55 complexes in a timely fashion. Four different B55 genes exist in mammals and we have also generated conditional knockout models for two of them, B55alpha and B55delta, two forms that are thought to be ubiquitous in mammals, whereas B55beta and gamma are supposed to be neural-specific. Lack of either B55alpha or delta does not alter normal kinetics of mitotic exit and protein dephosphorylation. In the absence of both of these subunits, mitotic exit still occurs although double KO cells display defective behaviour, especially in conditions in which mitotic exit is delayed in a SAC-dependent manner.
Finally, we have also generated and studied mouse models deficient in Cdc14 phosphatases. Cdc14 is an essential mitotic exit phosphatase in yeast but its function in mammals is not well understood. Lack of the two members of this family, Cdc14a or Cdc14b does not result in significant deficiencies in mitotic exit, neither in individual knockout nor in combination. Yet, Cdc14b seems to be crucial to regulate transcription and chromatin conformation during the mammalian cell cycle, as well as specific aspects during meiotic checkpoints.

WP5: New technologies to generate quantitative data on protein function in human cells
Obtaining quantitative and dynamic data on relevant proteins is a prerequisite for systems biology. Workpackage 5 aimed at developing new technologies which enable the generation of quantitative data on human proteins, in particular on their abundance and functional dynamics in high spatial and temporal resolution in living cells, and the stoichiometry of their subunits in protein complexes, in order to establish mathematical models of various aspects of mitosis in human cells.
An integrated quantitative imaging system for systems biology
In the past 5 years, the Ellenberg Lab (EMBL) has collaborated with Carl Zeiss Microcopy GmbH (Partner 12) to develop novel imaging technology allowing quantitative detection of subcellular localization and fluxes of proteins during cell cycle. The resulting information provided the basis for predicting the potential functions and interactions of mitotic proteins. Our new imaging system allows the integration of single molecule FCS data in an automated and high-throughput manner to enable absolutely quantitative confocal imaging of fluorescently tagged cellular proteins. This work has involved the development of innovative solutions for both the hardware and the software. The automation of FCS data acquisition and their visualization through the Zeiss confocal system (LSM780/Confocor3) has been achieved and is of great utility for the prediction of interactions and functions of novel protein complexes. During the period of the project, several software packages (ZEN2, MicroscopyPipelineConstructor, Fluctuation Analyzer) were developed for fully automatic acquisition and analysis of F(C)CS data. Using automated quantitative imaging on two Zeiss LMS780 microscopy systems and the developed software packages, we were able to successfully produce quantitative in vivo imaging data on the absolute concentration changes of key mitotic proteins, for example of cohesin subunits in living cells in cytoplasm and nucleus during cell division (Figure 5.1).

D
Figure 5.1: Workflow of automated acquisition of time series of FCS measurements (A) Software package combining automated FCS and image analysis. (B) Workflow consisting of autofocus, segmentation of nuclei to select points for FCS measurements in cytoplasm and nucleus. (C) Automated analysis of the FCS raw data using the ‘Fluctuation Analyzer’ software package. (D) Quantitative in vivo imaging of cohesin subunits.
Measuring and quantifying mitotic protein complexes are crucial for establishing valid mathematical models describing how the mitotic proteins function/interact during cell cycle. To ensure that the imaging-based measurements are physiologically relevant, it is necessary to replace the endogenous human genes with fluorescently tagged transgenes that preserve all aspects of expression regulation to achieve physiological concentration and localization. This can be accomplished by either (i) silencing the untagged gene using RNA interference and expressing fluorescently tagged transgenes via the siRNA/BAC technology (Hyman lab, WP1), or (ii) tagging proteins endogenously and homozygously with fluorescent tags via genome-editing methods (Ellenberg Group). The Ellenberg Group has assessed mainly two genome-editing techniques, ZFN and CRISPR/Cas9, and developed a pipeline for the production of homozygous human cell lines in which all alleles of the endogenous gene are fluorescently tagged. The paired CRISPR/Cas9 Nickase based approach is now used routinely to produce homozygous cells expressing endogenously fluorescently tagged proteins of interest for the acquisition of automated FCS-calibrated and therefore absolutely quantitative imaging data (Figure 5.2).
A B

Figure 5.2: Genome editing produced HeLa cells homozygously expressing AURKB-mEGFP. (A) Junction PCR was performed to test for correct integration of mEGFP at the C-terminus of the AURKB gene resulting in 3.1kb PCR product. The asterix indicates a cell line in which all alleles were labeled with mEGFP (=homozygous). (B) Live cell imaging with HeLa cells homozygously expressing AURKB-mEGFP demonstrated correct localization of AURKB-mEGFP during mitosis.

The genome-editing method and the automated imaging system we developed allowed us to establish a robust pipeline to produce quantitative data on abundance, localization and dynamics of proteins of interest in 3D with high spatial and temporal resolution in living cells. The pioneer data sets on mitotic proteins that we obtained through this pipeline have been incorporated into the MitoSys database. More data will be added once the validation procedure has been completed. These datasets promise to be of great importance for systems biology of mitosis and beyond.
Quantitative mass spectrometry to determine the stoichiometry of multisubunit protein complexes
We have developed quantitative mass spectrometry (qMS) methods to determine the stoichiometry and absolute cellular copy numbers of multi-subunit protein complexes such as the anaphase promoting complex/cyclosome (APC/C), the mitotic checkpoint complex (MCC) and cohesion in different cell cycle stages. We further validated these numbers by determining cellular copy numbers of cohesin complexes using an independent method, namely fluorescence correlation spectroscopy (FCS). Automated acquisition of time series of FCS was performed with HeLa cells expressing histone H2B-mCherry as a chromosome marker and the cohesin subunits STAG1-EGFP or STAG2-EGFP labeled endogenously. Absolute concentrations and kinetic data in different cell cycle stages (G2, G1, Prometaphase) were obtained in the cytoplasm and nucleus. The resulting FCS data have confirmed the data we obtained previously with the qMS methods. These data have been used to establish and improve the mathematical models in WP3.

WP6: A comprehensive model of mitosis
Regulation of the Cdk1 counter-acting phosphatase
Mitosis is brought about by activation of the cyclin-dependent kinase-1 B-type cyclin complex (Cdk1:CycB) in all eukaryotes. Cdk1:CycB activation is amplified by two positive feedback loops by which Cdk1:CycB activates the Cdc25 phosphatase and inhibits its inhibitory kinase (Wee1). Phosphorylation of proteins during mitosis requires that the activity of Cdk1 overcomes its counter-acting protein-phosphatases. Therefore, mitotic phosphorylation is facilitated if the activation of the protein-kinase is accompanied by the simultaneous inhibition of its counter-acting phosphatase. The Cdk1:CycB reduces the activity of some counter-acting phosphatases in mitosis. For example, Cdk1:CycB phosphorylates Greatwall-kinase (Mastl in human), which in turn phosphorylates and thereby activates two small-molecular inhibitors of PP2A:B55 (Endosulfine and Arpp19, collectively referred to as ENSA). Phosphorylated ENSA (pENSA) binds to and inhibits PP2A:B55; thereby the BEG PP2A:B55/ENSA/Greatwall) pathway facilitates the complete phosphorylation of certain Cdk1:CycB substrates at mitotic entry and their timely dephosphorylation at mitotic exit (Figure 6.1).

Down-regulation of the phosphatase dephosphorylating Cdk1 substrates creates a coherent feedforward loop and suppresses futile cycling of the Cdk1 targets. Since both Wee1 and Cdc25 are PP2A:B55 substrates, Cdk1 and the phosphatase are antagonistic enzymes as well. We have incorporated these new network motifs into our mitotic control model and analyzed their dynamic consequences (Domingo-Sananes, Kapuy et al. 2011, Krasinska, Domingo-Sananes et al. 2011). We have also proposed that PP2A:B55 could be a Greatwall-inactivating phosphatase, thereby creating another double-negative feedback loop that creates a bistable switch controlled by Cdk1:CycB (Tuck, Zhang et al. 2013, Zhang, Tyson et al. 2013). As a cell enters mitosis, PP2A:B55 activity is high, and high MPF activity is required to flip the switch to the low B55-activity state. Once the switch is flipped, then moderate or even low activity of Cdk1 is enough to keep B55 activity ‘off’. We argued that the BEG switch is important to control the temporal order of anaphase (separase activation, Figure 6.2) and mitotic exit (Cundell, Bastos et al. 2013).
Recent work has illustrated that the reactivation of PP2A:B55 depends on its own intrinsic activity. In the pENSA-phosphatase complex the phosphate group of ENSA can only be removed by the ‘inhibited’ PP2A:B55. Therefore pENSA is not only an inhibitor, but also a substrate, of the phosphatase by a ‘unfair competition’ mechanism. Recently we have incorporated this mechanism into the model and explored its dynamic consequences (Vinod and Novak 2015).
Modelling the mitotic spindle and error correction mechanism
Cdk1 activation causes mitotic spindle formation and nuclear envelope breakdown; thereby the spindle microtubules starts to capture chromosomes at their kinetochores. Successful completion of mitosis requires that each chromosome is bi-oriented along the mitotic spindle, which is called amphitellic attachment. Bi-orientation takes place during prometaphase of mitosis. The formation of amphitellic KT-MT attachments is essentially a stochastic, and thus error-prone process. How error-free chromosome bi-orientation robustly emerges from inherently error-prone biochemical and mechanical reactions is a long-standing puzzle that is not yet fully resolved. The prevalent hypothesis in the field is that inter-kinetochore tension distinguishes correct from incorrect KT-MT attachment, allowing a detaching activity (mediated by AuroraB kinase) to discern between correct and incorrect attachments through an error-correction mechanism. Since KT-MT attachments are necessary for KT-stretch (tension) and vice versa, the error-correction mechanism creates a so called “initiation problem of bi-orientation” illustrated in Figure 6.3. We have analyzed this problem in a stochastic modelling framework (Zhang, Oliveira et al. 2013), and experiments suggest that conversion of lateral to end-on attachment provides a solution (Kalantzaki, Kitamura et al. 2015).
Modelling the Spindle Assembly Checkpoint (SAC)
The biochemical network controlling mitotic progression is regulated by the KT-attachments and/or KT-tension. Eukaryotic cells stop (or delay) segregation of chromosomes by activating the SAC until all chromosomes become amphitellically attached to the mitotic spindle. Mitotic progression after prometaphase requires the activation of the Anaphase Promoting Complex/Cyclosome (APC/C), which promotes degradation of its substrates by ubiquitination. Tensionless/unattached KTs catalyze the formation of the Mitotic Checkpoint Complex, which is a diffusible APC/C inhibitor that blocks the activation of the ubiquitin-ligase. The two crucial substrates of APC/C are CycB and securin, which is an inhibitor of the protease (separase) responsible for cleaving sister-chromatid cohesins. Therefore, APC/C promoted separase activation leads to initiation of anaphase, while degradation of CycB inactivates Cdk1 and causes mitotic exit (Figure 6.4).
Since centromeric cohesins are essential for KT-tension, the loss of KT-stretch can reactivate EC and the SAC. During normal mitotic progression the checkpoint is not reactivated, suggesting irreversible inactivation of EC and the SAC, which is known as the ‘anaphase problem’.
This fundamental observation suggests that the mitotic checkpoint behaves as an irreversible, bistable switch (He, Kapuy et al. 2011). The dynamic characteristics of the mitotic checkpoint:
• Sensitivity: the SAC is responsive to a single unattached chromosome (Rieder; Gerlich)
• Reversibility: the SAC can be reactivated during metaphase
(Pines – spindle poison; Gerlich – laser cut of MTs)
• Irreversibility: neither the EC nor the SAC is reactivated during anaphase progression despite loss of KT tension (‘anaphase problem’)
are consistent with the concept of a bistable switch (Figure 6.5).

Our model proposes that both modules (EC and SAC) are characterized by positive feedback loops (Figure 6.6). In addition to that, the two modules have an inverse input-output relationship:
• Cdk1 is the output of the SAC module and the input of the EC module
• uKT is the output of the EC module and the input of the SAC module.

The interaction of the two modules is best illustrated on a phase plane (Figure 6.7). At anaphase, the EC is reactivated but only at high Cdk1 activity. The bistability allows the system to exit smoothly from mitosis despite loss of KT-tension caused by cohesion cleavage (Rattani, Vinod et al. 2014).

The model has been used to simulate the quantitative live-cell imaging experiments of partner 7 (IMBA). Securin levels have been measured with fluorescent-tag in real time during mitotic progression in HeLa cells. The state of chromosomes was visualized by H2B-Cherry. Kinetic parameters were estimated by fitting the model to the experimental data by assuming that the degradation of Securin (in the experiment) and cycB (in the model) are similar (Figure 6.8).

Figure 6.8: mitotic progression of HeLa cells followed by securin degradation (left) and simulated by the EC-SAC feedback model.
The prediction of the integrated EC-SAC model predicts that irreversible inactivation of the mitotic checkpoint at anaphase is the consequence of the drop of Cdk1 activity (Rattani, Vinod et al. 2014). This prediction of the model has been experimentally validated by silencing the mitotic checkpoint using different Cdk-inhibitors. Cells were arrested in mitosis using a high concentration of spindle poison (nocodazole) and treated with different concentrations of flavopiridol or roscovitine (Figure 6.9).
Both experimental data and model simulation show a rate-change point during silencing (Figure 6.9) that is the consequence of the positive feedback within the SAC module.

Figure 6.9: Silencing of the SAC by Cdk1 inhibition (roscovitine-left, flavopiridol-right). Securin-eGFP levels are measured in HeLa cells (dotted curves) after inhibitor addition (top panels). The results of numerical simulations are shown by solid curves. The rates of Securin degradation were calculated and simulated by the model (bottom panels).
The deterministic model described above has been converted into a stochastic model and analyzed by the Gillespie’s Stochastic Simulation Algorithm. Model simulations were compared with the experimental data on Drosophila neuroblast cells’ progression through mitosis without cohesins (Figure 6.10).

WP7: Bioinformatics resource
A bioinformatics resource for quantitative data about mitotic protein is an essential prerequisite for a joint systems biology effort of several experimental and theoretical groups as they are assembled in the MitoSys consortium. This resource serves as a central hub for data sharing and evaluation among the consortium groups as well as data dissemination to the outside of the consortium. The MitoSys database housed in the project website (www.mitosys.org) has been developed on the basis of the existing MitoCheck bioinformatics platform (www.mitocheck.org). New functionality has been added to the MitoCheck database in order to incorporate new data generated by the MitoSys labs. For instance, quantitative data of some mitotic proteins from the FCS imaging experiment carried out in the Ellenberg group (EMBL) have been added to the MitoSys database and are publicly available (http://www.mitosys-org/cgi-bin/mtc?action=FCS). These data are of great value to the modeling groups within and outside MitoSys for establishing models of several key aspects of mitosis and eventually a comprehensive model of mitosis in human cells. The quantitative data on its own is also a valuable resource for the scientific community. The MitoSys database is undergoing continuous update to add new data and functionality. We plan to continue this effort after the MitoSys project if the necessary resources can be secured.

Potential Impact:
Mitosis is a fundamental process in biology. Studying mitosis systematically is the goal of MitoSys. Over the five-year period, MitoSys has achieved the following scientific, technological as well as social results:
• Simulations and dynamic models of steady state spindle assembly and microtubule-cortical interactions; mathematical models describing the spindle assembly checkpoint (SAC), the regulation of cohesion activity and function by other cell cycle regulators, the dynamic interaction of cohesin with DNA, and the roles of cohesin in chromosome cohesion and separation during cell cycle, and mitotic exit. These models have been refined and validated by biochemical, biophysical, imaging and validation experiments carried out within MitoSys.
• A first comprehensive model of mitosis in human cells.
• Technology advancement such as methods for replacement of endogenous proteins by tagged proteins in human cells, a microscope system for quantitative imaging in living cells, improved methods for quantitative mass spectrometry.
• A project database which stores and disseminates data; a training course to train other scientists in systems biology; a public exhibition to inform the general public about systems biology of mitosis and its relevance to health and disease
We reckon that these results will have a tremendous conceptual and technological impact on the study of mitosis and more generally of cell division and cell physiology as well as broad scientific community and the society at large.
Impact on cell cycle studies
The results MitoSys obtained have generated a wealth of knowledge which significantly advanced our understanding of mitosis at molecular as well as systems level. This is illustrated by more than 130 publications in peer-review journals, many of which are of high profile and impact. The MitoSys database has becoming a valuable resource for researchers in the cell cycle field worldwide.
Impact on basic research
Mitosis is fundamental for all forms of life. Studying the organization of mitosis has therefore a profound intrinsic value to biological and medical research.
MitoSys focused on the quantitative description of how cells progress through mitosis during the cell cycle. The underlying mechanisms are relevant for cellular decision making in general in cell and developmental biology. The unique approach we took in model building which combines biochemistry, biophysics, mathematics, molecular biology, imaging technology proved to be very successful and can be applied in studying other complex biological processes.
The work performed in this project has led to the development of several new assays and techniques that are broadly applicable in other research areas, for example, the biGBac method for the rapid assembly of baculoviral genomes for recombinant expression of protein complexes, mass spectrometric and imaging techniques for the determination of absolute protein copy numbers in cells, the optical microscopy for live cell imaging and genetic manipulation techniques. All of these methods and techniques are of potential interest to a large community of researchers involved in the study of cell and organismal physiology, and some of them are amenable to industrial partnership.
Over the periods of MitoCheck and MitoSys, we have established hundreds of stable cell lines expressing GFP-tagged proteins from BACs (bacterial artificial chromosomes). These cell lines have been of great importance to the member labs of MitoSys. More importantly, they have been shipped worldwide upon request for search in many areas - far beyond cell cycle field.
The MitoSys database has proved to be a valuable data resource for biology as well as medical research communities.
We believe that MitoSys can provide a credible example for how leading researchers with complementary expertise but with a common interest (specifically, cell division) can work together as a delocalized, transnational, integrated “Department of Systems Biology” to crack a daunting scientific and technological problem. Throughout the MitoSys project, a team of biochemists, cell biologists, biophysicists, computer vision experts, engineers and mathematical modelers collaborated closely. This involved not only regular communications, but also actual interdisciplinary activities. Computer scientists were exposed to the complexity of diverse biological data coming from biochemical assays, live cell imaging etc. Some even participated directly in wet-lab bench work. Biologists on the other hand programmed computational analysis routines and gained experience in mathematical modelling of highly dynamic biological control systems. This framework provided a culturally rich and innovative environment much appreciated by and beneficial to researchers of all stages. In particular, the young generation of scientists trained this way with interdisciplinary skills will be highly relevant to foster Europe’s advancement in basic research and industry, and thus create strong socio-economic impact.
Impact on medical research and human health
Mitosis is fundamental for human development and health. Its correct execution is essential for the propagation of life, whereas mistakes during mitosis are frequently associated with human diseases such as cancer. For instance, mutations in formin 2 (studied in details in workpackage 1) account for the majority of fertility defects. Malfunctions of the cohesin complex (studied in Work Package 3) contribute to several human diseases and syndromes. A gradual loss of cohesin from DNA in ageing oocytes is thought to be a main reason for the increased frequencies with which trisomy 21 (Down Syndrome), other trisomies and spontaneous abortions are observed in ageing women. Furthermore, mutations in cohesin and cohesin regulatory proteins are known to be the cause of a group of rare genetic diseases, called “cohesinopathies”. Finally, recent cancer genome sequencing projects have revealed that genes encoding cohesin subunits are frequently mutated in several human cancers, such as acute myeloid leukemia, bladder cancer, glioblastoma and a few others. The knowledge created in this project about mitosis will therefore contribute to a better understanding of human diseases and may also contribute to the development of better diagnostic, prognostic or therapeutic approaches in the future.
Impact on pharmaceutical industry
Mitosis is considered an ideal phase of the cell cycle for targeting cancer cells, due to their high mitotic index and (frequently) very derailed genome. In fact, compounds that suppress mitotic exit are already widely used as cancer therapeutics (e.g. paclitaxel). The knowledge MitoSys created about mitosis at the molecular and systems levels may influence pharmaceutical industry, in particular in the field of chemotherapeutics in several ways.
(1) The detailed analyses of several key mitotic proteins may set the basis for new drug discovery.
(2) Our study of crucial mitotic steps (spindle assembly and dynamics, spindle assembly checkpoint (SAC), chromosome cohesion and separation, mitotic exit) may provide new concepts on developing anti-mitotic therapeutics. This is particularly evident with our new results on the relevance of preventing mitotic exit as a new strategy in cancer therapy (Manchado et al., 2010). This finding has elicited significant interest for new treatment strategies aiming at preventing mitotic exit and killing cells in mitosis. We are in the discussions with pharmaceutical companies to promote further research and development in this area.
(3) The quantitative models of mitotic progression MitoSys established will impact on drug discovery processes and will help predicting the effects of drugs with much higher accuracy.
(4) The various mouse models we established over the course of the project can be used to validate cancer drugs in vivo. For instance, mitotic kinases are currently considered as cancer drug targets. Specific small-molecule inhibitors against these mitotic kinases are currently under clinical evaluation for cancer therapy. The mouse models we have generated for mitotic kinases Aurora A, Aurora B or Plk1 can be employed to test in vivo the efficacy of these inhibitors as potential cancer drugs.
Training activities
Systems biology is gaining momentum as an approach of formalization of complex and dynamic biological systems. Mathematical models, i.e. computational descriptions of biological systems and processes, are bound to result in a new level of biological understanding. This calls for a new generation of biologists equipped with not only fundamental knowledge of biology but also deep understanding and grasp of modern techniques. MitoSys attempted to meet this requirement by organizing the first cellular modeling course in the field-“Modeling Cellular Processes in Space and Time”. 19 researchers at different career levels from 7 European countries as well as USA, India, Canada and Brazil attended the one-week training course. The background of the participants varies significantly – computer engineering, chemical engineering, mathematics, physics, synthetic biology as well as traditional biology. This training course shall equip the students with new tools for advancing their research work further and broader.
Dissemination activities
The results MitoSys obtained have mainly been disseminated to the scientific community via the established mechanisms of presentations at international conferences and publications in peer reviewed scientific journals. In addition, the MitoSys database (http://www.mitosys.org/cgi-bin/mtc) has becoming a valuable resource for researchers in the cell cycle field worldwide. Furthermore, MitoSys has made several software packages for the quantitative imaging systems publicly available. This includes the micropilot software and Zeiss ZEN macros (www.embl-em.de/software.php?lang=de&hid=8&sw_group=12&Sub=Imaging) the ‘Fluctuation Analyzer’ software package (M. Wachsmuth et al. 2015, www.embl.de/~wachsmut/downloads.html,) and the Cellgonition software (www.cellcognition.org/downloads). A series of videos describing the model of the yeast spindle were released by the EMBL with the title “under pressure (http://news.embl.de/science/1502_spindles_under_pressure/).
Knowledge about systems biology is very much limited in the general public despite the fact that modern biology approaches have an increasingly important impact on medical, ethical, economic and socio-cultural aspects of our society. MitoSys took the challenge and presented three public exhibitions “Lens On Life” in three European cities (Rome, London, Heidelberg) with the aim of disseminating the scientific knowledge about mitosis and systems biology and their relevance to health and disease in innovative and exciting ways. This attempt turned out rather successful. More than 3,000 people visited the exhibitions, some of which interacted directly with the artists, scientists and the curators involved in the exhibitions in accompanying satellite events.
One unique aspect of the MitoSys project is the collaborations between the MitoSys scientists and several highly acclaimed contemporary artists in exploring mitosis from inspiring and innovative angles. The collaboration has been documented in a one-hour film “Meetings of Minds”. This film has been screened at the MitoSys exhibitions and on other occasions and will continued to be used in the future to educate general audiences about the importance and fascination of research about cell division.
The MitoSys project website (www.mitosys.org) not only disseminates the results and the publications MitoSys has been generating to the scientific community, it also provides the general public with the knowledge and images about mitosis, work performed in the MitoSys project and its relevance to society. The documentary film “Meetings of Minds” is publicly available on the project website (http://www.mitosys.org/exhibitions). It shall reach out to many more people and will last beyond the duration of the project.
Exploitation results
Carl Zeiss Microscopy GmbH (Partner 12) and the Ellenberg group at EMBL (Partner 2) together have developed a quantitative FC(C)S confocal imaging set up for generating quantitative data on the abundance and dynamics of labeled proteins with high spatial and temporal resolution in living cells. Combined with the genome-editing method we developed for labeling proteins of interest, this set up can be broadly used in other biological and biomedical studies for acquiring quantitative and dynamic data of proteins in living cells. This quantitative FC(C)S confocal imaging set up will be exploited by Call Zeiss Microscopy GmbH in 2016.
Fully motorized microscope systems are a prerequisite for automatic high throughput data acquisition and long-term quantitative live cell imaging experiments. Carl Zeiss Microscopy GmbH has developed and optimized motorized objective lenses and water dispenser allowing software controlled automated imaging thereby avoiding complex manual manipulations of microscope components. Further research is undergoing so that the systems can take other immersion fluids such as silicon oil, oil or glycerol. This system is expected to be commercially available by the end of 2015.
Customized software is crucial for any automated system. For the quantitative confocal imaging systems for the MitoSys work, Carl Zeiss Microscopy GmbH developed the “Experiment Designer” module in the ZEN2 imaging software. This module is complemented by a number of hardware and software options that allow repetitive imaging of a large number of samples with different imaging set-ups in an automated fashion. This flexible automation setup increases the throughput, and can be used by customers with no special expertise in microscopy or programming. The “Experiment Designer” and the ZEN2 software have already been commercialized by Carl Zeiss Microscopy GmbH in 2014.

List of Websites:
www.mitosys.org

Project coordinator:
Jan-Michael Peters, Ph.D
Research Institute of Molecular Pathology
Dr. Bohrgasse 7
A-1030, Vienna
Austria
Email: peters@imp.ac.at

Project manager:
Yan Sun, Ph.D
Research Institute of Molecular Pathology
Dr. Bohrgasse 7
A-1030, Vienna
Austria
Email: sun@imp.ac.at
Tel: +43 1 797 30 3299
Fax: +43 1 798 7153