This document discusses computational modelling of Epithelial to Mesenchymal Transition (EMT). It presents an international team that performs multiscale modelling of heart development, including EMT simulations. The modelling spans multiple scales from sub-cellular processes to organ-level phenomena. EMT simulations have been conducted both in vitro and in silico using Cellular Potts Models and agent-based modelling software. The effects of factors like Notch and BMP2 signaling on EMT are explored. Future work aims to validate the EMT simulations against experimental data.
Strategize a Smooth Tenant-to-tenant Migration and Copilot Takeoff
RITS, Abdulla
1. Computational Modelling of
Epithelial to Mesenchymal Transition
Tariq Abdulla1, Ryan Imms1, Jean-Louis Dillenseger2,3,
Jean-Marc Schleich2,3, Ron Summers1
1.Dept Electronic & Electrical Engineering, Loughborough University, UK
2. LTSI, University of Rennes 1, France
3. INSERM, U642, Rennes, F-35000, France
R.Summers@lboro.ac.uk
http://www-staff.lboro.ac.uk/~lsrs1
15. In vitro EMT
Wildtype Notch1 BMP2
L. Luna-zurita et al. “Integration of a Notch-dependent mesenchymal gene
program and Bmp2-driven cell invasiveness regulates murine cardiac
valve formation,” The Journal of Clinical Investigation, vol. 120, 2010.
16.
17. CPM Model
E J ( ( x )), ( ( x ')) (1 ( ( x )), ( ( x ')) ) s (s S ) 2 v (v V ) 2
x, x'
Compucell3D
26. Validation Step
Our simulations are not validated
but we intend to do this.
To illustrate how this can be
done, here is an existing example
of a validated simulation.
Hello, The presentation I will give you today is of a computer modelling of Epithelial to Mesenchymal Transition.We are interested in modelling this process because of its central importance in heart development.I am Jean-Marc Schleich. My background is clinical – not modelling – but I will do my best to explain this work to you.
The team is formed from a collaboration between Loughborough University in the UK and Universite de Rennes 1.By bringing together engineers with modelling skills with clinical and imaging experts, we aim to unravel some of the mysteries of heart formation.(click)We illustrate some of the work done at each university, with the Loughborough team focussing on modelling. At the top is a protein level model. In the middle a cell level simulation. They are particularly interested in the linking of models at different levels of scale. The bottom figure shows their use of “ontologies” to annotate model parameters – which helps with linking between scales.(click)In Rennes we are conducting CT acquisition and 3D reconstruction of heart specimens with tetralogy of Fallot. We aim to better understand the spatial arrangement of heart structures in this disease, and to improve understanding of the embryological causes.
The motivation for looking in more detail at heart development is to understand how congenital heart defects occur.Many different mechanisms can interact and contribute towards several different types of congenital heart defect.To really understand the risk factors behind CHD, and provide early warning, it is not sufficient just to identify genes associated with CHD. We need to understand how each of these processes work.(click)Each of these processes are complex in their own right – and the work presented here focuses only on the EMT.
The main question the team is investigating is how the heart changes from a primitive heart tube to the fully formed heart at the end of development.From a single circulation system to a double circulation system.From a hollow tube, to a heart with four chambers.
A great deal of remodelling occurs during development of this organ. But what is interesting here is that – in two restricted regions of the endocardium (the Outflow tract and atrioventricular canal) we have swellings known as endocardial cushions.(click)
The way these cushions grow is by EMT. The heart tube consists of an outer layer of myocardium and inner endocardium, with an exra-cellular matrix known as cardiac jelly in between.During EMT, endocardial cells lose their cohesion, and invade the cardiac jelly. The fields of cells which can do this (in the OFT and AVC) are established by Notch signalling. The endocardial cells in these areas have a higher concentration of Notch protein. One of the effects of this is that they lose their cohesion, ecause notch acts to upregulate other proteins (slug/snail) which in turn inhibit VE Cadherin = the main adhesion molecule for endothelial cells.
This lovely picture shows tissue redistribution after looping of the heart.We can see from this diagram that the Endocardial cushions in the conotruncus/OFT later form the pulmonary and aortic valve, the septum between pulmonary artery and aorta and the membranous portion of the interventricular septum.The membranous portion of the septum is by far the most common location for interventricularseptal defects, and normally morecomplicated – due it’s association with OFT rotation and endocardial cushion growth defects.The cushions in the atrioventricular canal form the mitral and tricuspid valves.
This slide shows what is happening in the heart between embryonic day 24 and embryonic day 32.Much of the inner structure of the heart is the result of the growth of endocardial cushions.These grow in the Outflow tract and the Atrioventricular canal by EMT. In both locations, they fuse by Day 32.Because the Outflow tract rotates, the fusing cushions form a helical shaped septum, which divides the aorta from the pulmonary artery.In the AV canal. Upper and lower cushions fuse to form the AV septum. They also contribute flaps to the Mitral and Tricuspid valves, as shown.
The Physiome Project is an international effort fororganisingmultiscalemodelling of human physiology. Scales of interest range from picometer (genes) to meters (organs and the organ systems).In the time domain, the Physiome covers events from microseconds (molecular events, such as ion channel gating) to processes that take place over long periods in the human lifetime.
We apply this framework for multiscale modelling to heart development – which is illustrated here for endocardial cushion growth. We are interested in processes at the nanometer scale (protein interaction) which take place within each cell, as well as the cell behaviour and tissue interaction at the micrometer scale, and the overall development of the heart at the milimeter scale.There are different modelling approaches applicable to each level of scale, and for this initial work, we are really interested in the middle level (cell-tissue level) (click)
We base our simulations on this in vitro study – which induced EMT in mouse endocardial cells. They cells were taken from mouse ventricles, so are not disposed to undertake EMT. In their wildtype state, they remain in a monolayer floating on the collagen gel.By the introduction of Notch protein, the endocardial cells lose their adhesion as expected. However this is not sufficient to induce a full EMT – instead cells only scatter on the surface of the collagen gel, without invading it.(click)With the introduction of BMP2 – a protein which is secreted by the myocardium in the cushion forming regions – cells both scatter on the surface of the gel, and invade into it.(click)
Without going into too much detail, part of the explanation for this is that Notch is downstream of BMP2, as illustrated here.(click)So the endocardial cells also lose their adhesion to one another, but something else also happens to induce them to invade the collagen gel. The hypothesis tested in these initial simulations is that this could be due to a simultaneous increase in the adhesion between the endocardial cells and the collagen gel / cadiac jelly.
The simulations we used are a type of Cellular Potts model. Cells are defined as being multiple adjacent sites on a lattice. We decide the “surface energy parameter” J between different types of cells. Larger J means less adhesion between the cells.The simulation uses a Monte Carlo algorithm so there is some randomness. It goes through each pixel and attempts to update it, with an adjacent cell. Whether it updates depends on whether this is a valid attempt (the tow pixels belong to two different cells) and the change in energy which would result. If the energy is reduced (change in E less than zero) then it will definitely accept the change (p=1). If it increases the energy, it is accepted with probability e^-/\\E/kT – where kT is a constant.
This is an illustration of the iteration of lattice sites, with both invalid attepmtps (same cell) and valid attempts (different cells). Valid attempts are either rejected or accepted depending on the equation shown before.
So our simulations use a 3D lattice, at the bottom you can see a cross section. In the base case, our simulated endocardial cells float on the collagen gel in a monolayer.
By increasing the surface energy parameter between endocardial cells (e.g. Reducing their cohesion) we have a 2D scattering on the surface of the simulated collage gel.
By simulating a simultaneous reduction in endocardial – to – endocardial adhesion, and an increase in endocardial – collagen gel adhesion, we observe fully invasive behavior. This supports the hypothesis that the mesenchymal phenotype is achieved in part by an increase in adhesion to the extracellular matrix.
VEGF regulates the level of endocardial mitosis. Higher VEGF = more mitosis(click)But the level of VEGF expression is lower in the cushion forming regions.(click)If the level of VEGF is too high, there will be a reduction in EMT – with consequently malformed endocardial cushions.
The hypothesis investigated here (in an abstract way) is this:>Endocardial cells have strong adhesion to each other>With a high rate of mitosis, there is no possibility for small gaps between endocardial cells.>It this factor – no small gaps – that prevents cells from migrating into the ECM.This was simulated in an abstract way. We used the same parameters as the previous simulations.On the left we use the parameters for 2D separation, on the right the parameters for 3D separation . We use contact inhibited mitosis, which means that cells divide when the they have less than a certain percentage of their surface area adjacent to other cells. The inclusion of mitosis preserves the monolayer, and this action could plausibly inhibit EMT, and explain why a higher level of VEGF prevents EMT from happening.
Validation can be achieved by comparing simulation to reality. The right hand side illustrates normal somitogenesis, while the left hand side illustrates a double N-Cadherin knockout (this particular protein is prevented from being made). In the simulation this is achieved by setting one of the parameters to zero. In both the real experiment and the simulation, we see a formation of twice the normal number of somites.So through a comparison of imaging from different experiments, and observing numbers of cells to undergo the change in simulations and reality, and the rate at which they change – this is the way of doing the validation step.