MultiCell Modeling of Biological Development Using the GGH Model and CompuCell3DApplications, Techno

1
Multi-Cell Modeling of Biological Development
Using the GGH Model and CompuCell3DApplications,
Technology and Open Problems

• James A. Glazier
• Biocomplexity Institute
• Indiana University
• Bloomington, IN 47405
• USA

Understanding Complex Systems UIUC,
Urbana-Champaign, IL Monday, November 09, 2009
Collaborators S. Schnell, M. Alber, G. Forgacs,
G. Hentschel, J. Izaguirre, R. Merks, M. Swat, S.
Newman
Support NIH, NSF, NASA, IBM Innovation
Institute, State of Indiana, Indiana University.
For Papers on these projects, please visit
http//www.biocomplexity.indiana.edu
2
What does Understand Mean?

• It means different things to different people
• Genomicists say We understand a car, when they
  have a list of all the parts and their names.
• Proteomicists say We understand a car, when
  they have a list of all the parts and their names
  and a list of which connects to which.
• Engineers say We understand a car, when they
  have a description of the structure and function
  of each subsystem.
• Physicists say We understand a car, when they
  understand thermodynamics, materials science...

3
Biology is Full of Complexity

• Nervous system.
• Cardio-vascular system.
• Immune system.
• Genetic Regulatory networks.
• Metabolic networks.
• Cochlea.
• Cytoskeleton.
• Ant colonies.
• Biofilms.
• Wound Healing.
• Development.
• Tissue function.
• ButModern Biology Tends to Look at Biological
  Problems from a Genetic/Reductionist Perspective.

4
Development

• What is Development?
• Biological Process by Which a
• Fertilized Egg ? Organism
• Physical Process Which Translates
• Genetic Information (Genotype)
• ?
• Structure and Behavior (Phenotype)

5
The Classic Problem of Development

• How does gene activity translate into phenotype?
• The classical genomic/analytic approach says x
  happens because gene y is expressed
• But how to go from expression to organism? Why
  are we not just blobs of cells expressing
  different genes?

Image Courtesy of Prof. Sima Setayeshgar (Indiana
University)
6
Development as Self-Organization

• About 5104 genes and 109 cells. Genome does not
  contain enough information to specify each cell.
• Even if it did, development would be fragile if
  completely specified.
• Instead, highly robust at all levels.
• Not Simply Signal?Differentiation?Pattern (Known
  as Prepatterning).
• Cells Create Their Own Environment, by Moving and
  Secreting New Signals, so Signaling Feeds Back on
  Itself.
• Hence Self-Organization and Robustness.

7
Computational Biology Goals
Why Simulation?

• A quantitative computational model is the best
  way to embody and test what we understand when we
  write down biological models.

• To explain biological processes that result in an
  observed phenomena.
• To predict previously unobserved phenomena.
• To identify key generic reactions.
• To guide experiments
• Suggest new experiments.
• Eliminate unneeded experiments.
• Help interpret experiments.

8
Why Needed?

• A huge gap between level of molecular data and
  observed patterns.
• Most Modern Biology is descriptive rather than
  predictive.
• Simplify impossible complexity by forcing a
  hierarchy of importance identifying key
  mechanisms.
• In a model know what all processes are.
• Failure of models can identify missing components
  or concepts.

9
Cell-Centered Modeling

• Genetics primarily drives the individual cell
• Response to extracellular signals secretion of
  signaling agents and extracellular matrix
  proteins.
• To understand how genetics drive multicellular
  patterning, distinguish two questions
• How does genetics drive cell phenomenology?
• How does cell phenomenology drive multicellular
  patterning?
• Todays Talk Completely Ignores All Intracellular
  Regulatory and Signaling Networks Except at the
  Simplest Level.
• Just as Sequencing the Genome Didnt
  Automatically Tell Us How Cells Work, Even if We
  Understood Everything About How Individual Cells
  Worked, We Wouldnt Automatically Understand How
  Organisms Worked.
• We Are Not Just Blobs of Cells Lying on the Floor
  Expressing Genes.

10
Why a Cell Level Model?

• Most mammalian cells are fairly limited in their
  behaviors. They can
• Grow
• Divide
• Change Shape
• Move Spontaneously
• Move in Response to External Cues (Chemotaxis,
  Haptotaxis)
• Stick (Cell Adhesion)
• Absorb External Chemicals (Fields)
• Secrete External Chemicals
• Exert Forces
• Change their Local Surface Properties
• (Send Electrical Signals)
• A long list, but not compared to 1010 gene
  product interactions.
• Many cells have relatively simple
  phenomenological behaviors most of the time.

11
Philosophy

• Fundamental Entities are Cells and Generalized
  Cells (e.g. mesenchymal cells, epithelial cells,
  ECM, medium)
• Cells have Internal states and Types which
  describe their properties and interactions
• Cells interact via an Effective Energy
• Additional Cell Properties described as
  Constraints
• Include External Chemical Fields
• Many possible ways to implement (CA, PDE, Finite
  Element)

12
Glazier-Graner-Hogeweg Model
x 20

• Energy Minimization Formalism Developed by
  Graner and Glazier, 1992
• DAH Contact Energy Depending on Generalized Cell
  Types (Differentiation States or Material
  Identity)

13
MonteCarlo Evolution Dynamics
Nearest Neighbor
14
CompuCell3D Model Sharing Environment
All simulation parameters are controlled by the
config file. The config file allows you to only
add those features needed for your current
simulation, enabling better use of system
resources.
ltPottsgt ltDimensions x"71" y"36"
z"211"/gt ltStepsgt10lt/Stepsgt
ltTemperaturegt2lt/Temperaturegt
ltFlip2DimRatiogt 2 lt/Flip2DimRatiogt lt/Pottsgt
ltPlugin Name"Chemical"gt ltThresholdgt0.8lt/Thresh
oldgt ltLambdagt10lt/Lambdagt ltConcentrationFilegt
field.dat lt/ConcentrationFilegt
lt/Plugingt ...
15
Recent Improvements to CompuCell3D

• Full Python Scripting Capability.
• Improved Performance (x10).
• Greatly Improved 3D graphics.
• Cell Anisotropy.
• Compartmental Cell Modeling.

16
Python Scripting inside CompuCell3D

• Most of the CompuCell simulations are currently
  run through its Python interface. Users can model
  complicated cell behaviors directly in Python
• Visualization can be extended by either using
  visualization libraries through Python or by
  interfacing to the Player through Python module
• Public components of computational CompuCell
  kernel (C) are accessible from Python
• The ability to change cell behavior during the
  simulation through Python opens up possibilities
  to include sub-cellular models in the cell-level
  simulation. Such approach is much more flexible
  than hard-coding sub-cellular models directly in
  C/C . It also allows for an easy addition of
  models written in other programming languages.
• Python scripting provides a level of flexibility
  comparable to Mathematica or Matlab.
• Writing CompuCell3D extension modules in Python
  is easy even for non-programmers

17
Example of CompuCell3D Python extension module
An increase of the target volume could be a
complicated function of morphogens or can be
calculated by a complex sub-cellular model. The
only change in the code required to get this
complicated behavior could look like below def
step(self,mcs) iterate over all cells
and increase target volume
invItrCompuCellPython.STLPyIteratorCINV()
invItr.initialize(self.inventory.getContainer())
invItr.setToBegin() cellinvItr.getCurr
entRef() while (1) if
invItr.isEnd() break
cellinvItr.getCurrentRef()
cell.targetVolume IncreaseVolumeSubCellModel(cel
l) increase target volume
invItr.next() Users need to supply the
implementation of IncreaseVolumeSubCellModel(cell)
function which can be in essentially in any
programming language. IncreaseVolumeSubCellModel(
cell) can be a part of the external library or
can be implemented by a user who runs the
simulation.
18
Applications

• Cell Sorting
• Gastrulation
• Somitogenesis
• Vasculogenesis
• Biofilms
• Tumor Growth

19
Engulfment
20
Development of Body Plan

• Specification of Body Axes
• Cleavage
• Gastrulation (Formation of Primitive
  StreakAnterior-Posterior)
• Somitogenesis (Formation of AP compartments)
• Organogenesis

21
GastrulationFormation of Main Body Axis
22
Our Gastrulation Model

• 4 types of cells Area Pellucida, Area Opaca,
  Kohlers Sickle, Streak Tip
• 5 cell behaviors Adhesion, Volume control,
  Secretion, Chemotaxis, (velocity correlation)
• Auxiliary mechanisms diffusion, decay

23
Various Mechanisms will Produce a Streak
ST secretes chemo-repellant for AO
ST secretes chemo-attractant for KS
KS secretes chemo-repellant for ST
24
Which Chemotactic Mechanism?
25
Large Scale MotionChemotaxis and adhesion not
enough
26
Velocity Correlations Help

• With Planar Polarity

27
Somitogenesis
c-hairy-1 oscillations
Somitogenesis is controlled by a clock, whose
gene expression periodicity corresponds to the
formation of one somite. Hairy is a homologue of
a Drosophila segmentation gene.
Signaling and gene pathways involved in
somitogenesis segmentation clock. Dashed lines
indicate interactions with conflicting findings,
double arrow represent interactions with more
protein components than are represented here.
28
Eph knockout /Initialization
29
N-cadherin knockout
30
PSM-ECM Strong Interaction Parameter Optimized
Adhesion-Repulsion /Control
PSM-ECM Strong Interaction
31
Endothelial cells aggregate into vascular
networks in Matrigel cultures
Courtesy Luigi Preziosi
32
Hypothesis Chemotaxis (Gamba et al. 2003 Serini
et al., 2003)
Chemotaxis cells migrate to higher
concentrations of VEGF-A

• Saturation of VEGF-A gradients inhibits
  directional cell migration
• ECs produce VEGF-A during first hour of vascular
  development
• Lateral Inhibition of Cell movement on Cell-Cell
  contact

33
Sprouting Angiogenesis

• Reproduces aspects of both capillary formation
  and sprouting
• How can we explain constant width of the cords?
• What are the scaling properties of the vascular
  network?

34
Biofilms (Bacterial Colonies)

• Formation of structures (mushrooms, etc)
• Why do biofilms collapse?
• Differentiation within Biofilm
• Competition, cooperation and patterning within
  multispecies biofilms

• How do biofilms form?
• Cell attachment
• Reorganization inside cell
• Spreading on surfaces
• Cell motility
• Cell division
• Secretion of Extracellular material
• Antimicrobial Resistance
• Mutation and competence

35
Biofilms

• Biofilm growth begins with few
• colonists adhering to a solid surface
• Biofilm cells grow at a rate
• which is a function of the concentration
• of oxygen (and/or nutrients)
• We assume that the growth rate is
• proportional to the local oxygen
• concentration (no saturation and death)
• The oxygen diffuses from the air
• to water the O2 concentration at the
• upper surface remains constant
• We use periodic boundary conditions
• along the horizontal axis

• The crucial factor in the biofilm growth is the
  oxygen/nutrient consumption by the growing cells.
  If the uptake is too small, the results of the
  simulations are similar to those without
  consumption. If the uptake is too large, there is
  no growth.
• The uptake of the growth factor causes
  instabilities in the biofilm growth the first
  little mushrooms grow faster than the other
  cells (due to the oxygen gradients) to consume
  more oxygen and prevent the surrounding cells
  from growing (a similar phenomenon occurs in
  reaction-diffusion systems with an inhibitor and
  activator)

36
TWO TYPES OF TUMOR

• MALIGNANT
• Cancerous, has the potential to invade and
  destroy neighboring tissues and create metastases
  (spread of cancer to other parts of the body).
  Very bad.
• BENIGN
• Compact, may locally grow to great size. Does not
  invade neighboring tissues and does not seed
  metastases. It usually does not return after
  surgical removal.

37
MOTIVATION

• What is physical difference between benign and
  malignant tumors?
• What mechanisms determine formation of benign vs.
  malignant tumors?

Clue Biofilms Benign tumors smooth
interface Malignant tumors rough interface
38
SIMPLE MODEL OF TUMOR COMPONENTS
Cell types normal, quiescent, mutated and
necrotic. Fields nutrient, extracellular matrix
(ECM) (non-cellular material supporting cells)
and matrix degradative enzyme (MDE) (degrades ECM
increasing tumor motility).
A. R. A. Anderson, Math. Med. Biol. 22, 163
(2005).
39
SIMPLE MODEL OF TUMOR COMPONENTS
Cell types normal, quiescent, mutated and
necrotic. Fields nutrient, extracellular matrix
(ECM) (non-cellular material supporting cells)
and matrix degradative enzyme (MDE) (degrades ECM
increasing tumor motility).
Cells motility, chemorepulsion from ECM
gradients, growth, division. Reaction-diffusion
equations for the fields Nutrient
concentration change diffusion production by
ECM uptake by tumor cells decay MDE
concentration change diffusion production by
cells decay ECM concentration change
degradation by MDE
A. R. A. Anderson, Math. Med. Biol. 22, 163
(2005).
40
CONSTANT GROWTH RATE ABOVE THRESHOLD
Benign Tumor Sufficient supply of nutrients
41
CONSTANT GROWTH RATE ABOVE THRESHOLD
Benign Tumor Sufficient supply of nutrients
42
GROWTH RATE PROPORTIONAL TO NUTRIENT
CONCENTRATION ABOVE THRESHOLD
Malignant Tumor Sensitivity of growth to
nutrient supply leads to fingering instabilities
(and possibly metastasis)
43
GROWTH RATE PROPORTIONAL TO NUTRIENT
CONCENTRATION ABOVE THRESHOLD
Malignant Tumor Sensitivity of growth to
nutrient supply leads to fingering instabilities
(and possibly metastasis)
44
Tumor Games
45
Plans

• Integrate Reaction Kinetics Models of Biochemical
  Networks into CompuCell Framework.
• Additional quantitative experiments on chick
  embryo cell movement, disruption and mechanical
  properties.
• Parallel version of CompuCell3D to allow
  1,000,000 cell models (in 3D).

46
CompuCell3D Collaboration
People Mark Alber, Alexander Anderson, Ariel
Balter, Mark Chaplain, Rajiv Chaturvedi, Nan
Chen, Trevor Cickovski, Jeff Coffland, Michael
Crocker, Rita deAlmeida, Andreas Deutsch, Gabor
Forgacs, James Glazier, Tilmann Glimm, François
Graner, Randy Heiland, George Hentschel,
Chengbang Huang, Jesus Izaguirre, Yi Jiang,
Charles Little, Roeland Merks, Charles Moad,
Chris Mueller, Stuart Newman, Nikodem Poplawski,
Herbert Sauro, Abbas Shirinifard, Maciej Swat,
Gilberto Thomas, Bakhtier Vasiev, Cornelis
Weijer, Benjamin Zaitlen
Funding Agencies and Institutions NSF, ICSB
Notre Dame, The Biocomplexity Institute Indiana
University, Los Alamos National Laboratory, Keck
Graduate Center, Technical University Dresden,
Universidade Federal do Rio Grande do Sul,
University of Grenoble, NIH, University of Notre
Dame, Kansas University Medical Center, Dundee
University
Websites https//simtk.org/home/compucell3d http
//www.biocomplexity.indiana.edu http//www.nd.edu/
lcls/compucell/ Google keyword CompuCell3D
47
Invitations

• CompuCell3D Users.
• Collaborators for Modeling of New
  Morphogenetic/Other Problems.
• Collaborations on Parameter Optimization/Inverse
  Problem.
• Collaborations on Developing Metrics to Describe
  Patterns.
• Collaborations on an SMBL-like Cell-level Model
  Sharing Framework.

48
CompuCell 3D)Allows You Easily to Reproduce any
of the Cell-Based Models in this Talk, or to
Develop Your OwnWe are always looking for new
groups to work with and can subsidize training at
IUB of people interested in learning to use
CompuCell3D

• If You Want to Learn More Please Visit Our Web
  Location at https//simtk.org/home/compucell3d
  and Try Downloading it Yourself!