Toward a Multiscale Model of Cell Migration

1
Toward a Multiscale Model of Cell Migration

• Michelle Wynn
• Indiana University

Spring 2008
School of Informatics
2
Why is Cell Migration Important?

• Embryo Development
• Congenital Defects
• Wound/Tissue Repair
• Immune System
• Tumor Formation
• Cancer

Vital to life but can also make us very sick
(video www.biochemweb.org/neutrophil.shtml --
made in 1952!)
3
Neural Crest Cells migrate from the neural tube
and form vital structures
Brain Region
in vivo video of NCCs as they migrate to the site
of spinal ganglia development
(video Kasemeier-Kulesa et al, 2005 , image
www.med.unc.edu/embryo_images)
4
Objective why do NCCs form chains during
migration?
Hypothesis
Contact Guidance
contact between protrusions plays an important
role in guiding cells toward other cells, in the
direction of migration (but may not be unique
mechanism)
(image from Rupp and Kulesa, 2007 experiments
performed on chick embryos)
5
Testing the Contact Hypothesis
Use chick as animal model to study NCC
migration. Analyze live video microscopy data of
developing chick embryos and derive possible
model parameters from video data and published
literature. Computationally model via a rules
based simulation to test mechanism of cell
contact.
Identify those parameters which create the most
stable chains
Want to determine the physical mechanism before
modeling the molecular and gene level components
(image Teddy and Kulesa, 2004 experiments
performed on chick embryos)
6
Stable chain must have a net positive
displacement toward target, over time

7
How I built the model framework
DESIGN PATTERNS! Built extensible, non-specific
and reusable simulation framework with O.O.
design patterns (Strategy, Observer, Factory,
Memento, others). Fully configurable at runtime
with dynamic class loading (limited compile time
dependency) Can easily change out various
strategies (update, rules, output, etc) Suite of
over 20 tests are run whenever model is altered
to make sure model is not broken!
Algorithm/Strategy Factory
Grid Update Algorithm
Neighborhood Algorithm
Rule Algorithm
Output Algorithm
Proliferation Algorithm
Parameter Factory
Probability Distributions
Discrete Values
Boolean Values
Can, for example, chose to change output strategy
(file or graphic viewer), or change grid update
strategy all at run time.
8
Contact Guidance Computational Model Rules
Cells protrude and retract their extensions
randomly (via an internal clock)
Follower cells may be directionally polarized and
have up to two protrusions
Follower cells randomly change polarization
direction when not in contact (via an internal
clock)
Leader cells are not polarized and can have many
protrusions
Leader cells move according to a probability
distribution (assumed to be biased toward target)
Follower cells move randomly, unless in contact,
then preferential movement toward a contacted
cell is assumed
9
Use Two Main Types of Parameters
Protrusion Parameters
Directionality Parameters
protrusion count
Polarization
direction change clock
protrusion / retraction clock
Movement
Likelihood to move in direction of contact
protrusion length
Directionality Bias (chemotaxis?)
Likelihood to retract when in contact
10
Contact Guidance Preliminary Results
Computationally intensive still running
parameter analysis
Current results indicate that single most
important parameter is relative bias of the
leader cell to move toward target. Suggests
chemotaxis plays a role.
11
Current Model Limitations
Not considering the importance of synchronization
of internal clocks
Contacts between cells do not stretch if one
cell moves away (probably not biologically
realistic)
Not incorporating other plausible mechanistic
hypothesis
Molecular and gene expression parameters not yet
included
12
Alternate HypothesisPath of Least Resistance
Red leader cells are biased to move toward
target. Blue followers are not directionally
biased. Follower cells move randomly but prefer
to move to an open site over a closed one.
Red leader cells form clear channel which the
blue follower cells seem to follow but they do
not keep up with the red cells.
TODO combine this model with contact guidance
model
13
Acknowledgements
Thank you to
Professor Santiago Schnell
Schnell Systems Biology Research Group
Professor Gregory Rawlins
Bioinformatics Group at SOI
Linda Hostetter
Linda Roos
14
Questions?