Title: Structured Population Models for Hematopoiesis Marie Doumic with Anna MARCINIAK-CZOCHRA, Beno
1Structured Population Modelsfor
HematopoiesisMarie Doumic with Anna
MARCINIAK-CZOCHRA, Benoît PERTHAME and Jorge
ZUBELLI part of A. Marciniak group
BIOSTRUCT aims
http//www.iwr.uni-heidelberg.de/groups/amj/BioStr
uct/
2Outline
- Introduction biological medical motivation
- Quick review of models of hematopoiesis
- Short focus on I. Roeders model
- The original model a discrete compartment model
- A continuous model link with the discrete model
- boundedness
- steady states
- stability and instability
- Perspectives
3What are stem cells ?
- Functionally undifferentiated
- Able to proliferate
- Give rise to a large number of more
differentiated progenitor cells - Maintain their population by dividing to
undifferentiated cells - Heterogeneous in respect to morphological and
biochemical properties
4Role of (adult) stem cells
- Found in lots of different tissues
- Govern regeneration processes importance in
- Bone marrow transplantation (leukemia), liver
regeneration - Cancerogenesis (cancer stem cells)
5(No Transcript)
6What is hematopoiesis ?
- Formation of blood components
- All derived from Hematopoietic Stem Cells (HSC)
7Open questions
- How is cell differentiation and self-renewal
regulated ? - Which factors influence repopulation kinetics ?
- How cancer cells and healthy cells interact ?
- How drug resistance of cancer cells can appear ?
- How acts a drug therapy (e.g. Imatinib for
leukemia) ? Can it cure the patient completely ? - and many others
8Models of hematopoiesis
- Compartments / quiescence and proliferation
- Maturation discrete or continuous process?
- IBM or PDE/ODE/DDE models
- Modelling the Cell cycle (or simplifications)
- Nonlinearities to regulate the system
- Feedback-loops (A. Marciniaks model)
- competition for space (stem cells niches I.
Roeder) - Choice of a model depends on
- which aim is pursued
9(very partial) short overviewon models of
hematopoiesis
- A good review Adimy et al., Hemato., 2008
- First models MacKey, 1978 Loeffler, 1985
- F. Michor et al (Nature 2005, ) linear ODE and
stochastic - I. Roeder et al (Nature 2006,) IBM model
- Nonlinearity reversible maturation process
- -gt Kim, Lee, Levy (PloS Comp Biol 2007, )
- PDE model based on Roeder IBM model
- Adimy, Crauste, Pujo-Menjouet et al. DDE and
application to chronic leukemia
10Short focus on I. Roeders modelIBM model built
on the following main ideas
- IBM model built on the following main ideas
11Short focus on I. Roeders modelIBM model built
on the following main ideas
- IBM model built on the following main ideas
12Short focus on I. Roeders model
- Goal to model leukemia Imatinib treatment. 2
Main ideas - 1. Reversible maturation process
- 2. Competition for room in stem cell niches
this nonlinearity controls the system - Work of Kim, Lee, Levy
- Write a full PDE model mimicking the IBM model
- Show strictly equivalent (quantitatively
qualitatively) behaviours - -gt very efficient numerical simulations
- Work of MD, Kim, Perthame
- Write successive simplified PDE models, keeping
ideas 1. 2. - Show equivalent qualitative behaviours (stability
or instability) - -gt analytical analysis explaining these
behaviours
13Short focus on I. Roeders model
- Simplest version of I. Roeders model
14Short focus on I. Roeders modelIBM model built
on the following main ideas
- IBM model built on the following main ideas
15Anna Marciniak Czochra sGroup BioStruct
aim
- See http//www.iwr.uniheidelberg.de/groups/amj/Bio
Struct/ - To model hematopoietic reconstitution
- gt model Cytokin control (feedback loop)
- Medical applications
- Stress conditions (chemotherapy)
- Bone marrow transplantation
- Blood regeneration
16Experimental data
17Original model discrete structure
differentiation
proliferation
Marciniak, Stiehl, W. Jäger, Ho, Wagner, Stem
Cells Dev., 2008.
18(No Transcript)
19Regulation and signalling
- Cytokines
- Extracellular signalling molecules (peptides)
- Low level under physiological conditions
- Augmented in stress conditions
- Dynamics
- Quasi steady-state approximation
20Models
- What is regulated?
- Evidence of cell cycle regulation
- Evidence of high self-renewal capacity in HSC
- Regulation modes
- Regulation of proliferation
- Regulation of self renewal versus
differentiation
21(No Transcript)
22(No Transcript)
23(No Transcript)
24(No Transcript)
25Model analysis
- Steady states
- Trivial stable iff it is the only equilibrium
- Semi-trivial linearly unstable iff there exists
a steady state with more positive components - Positive steady unique if it exists (globally)
stable ? -
- -gt Stiehl, Marciniak (2010) T. Stiehls talk on
friday
26(No Transcript)
27(No Transcript)
28PDE model derived from the discrete one(MD,
Marciniak, Zubelli, Perthame,in progress)
- Stem cells w, aw, pw, dw u1, a1, p1, d1
discrete - Maturing cells u(x), p(x,s), d(x) ui,
ai, pi, di discrete - gi-1 ui-1 - gi ui with gi
2(1-ai(s))pi(s)
29PDE model from discrete to continuous
- 1 - We formulate the original model as
30PDE model from discrete to continuous
- 2 We adimension it by defining characteristic
constants - 3 We introduce a small parameter e?0, with nne
? x - 4 To have sums Riemann sums integrals
- differences finite differences
derivatives
31PDE model from discrete to continuous
- 2 We adimension it by defining characteristic
constants - 5 Define
32PDE model from discrete to continuous
- 2 We adimension it by defining characteristic
constants - 6 Continuity assumptions
33- 7 Proposition under the continuity
assumptions, the - Solution to the discrete system converges,
up to a - subsequence, to with
- if moreover the convergence is strong in
- for w lim(u1e) solution of
- we get
- If moreover u is continuous in x and un-1e
converges to u(t,x) - Then une converges to v solution of
34Analysis of the PDE model
- Remark decorrelation between differentiation and
- proliferation is needed, else due to orders of
magnitude - transport becomes a corrective term and we get
35Analysis of the PDE model
- Remark decorrelation between differentiation and
- proliferation is needed, else due to orders of
magnitude - transport becomes a corrective term and we get
- see Grzegorz Jamrozs talk for more insight
36Numerical simulations
Discrete model
Continuous model
Stem cells
Maturing cells
mature cells
37Analysis of the general PDE model
With initial conditions
Cell number balance law
38Analysis of PDE -Assumptions
- Theorem. The unique solution is uniformly bounded
39Analysis of PDE - boundedness
- Main difficulty feed-back loop involves a delay
- Main tool the following lemma
- Sketch of the proof deriving the equation
divided by u
40Analysis of PDE - boundedness
- From boundedness of z we deduce
- 1st and 3rd estimate directly from boundedness
of z - 2nd estimate look at
41Analysis of PDE - boundedness
- Extra estimate, used for non-extinction (see
below) - Proof
42Analysis of PDE steady states
- Solution of
- With .
- Proposition. There exists a steady state iff
- In this case, it is unique.
- Remark similar assumption for the discrete
system - BUT here no semi-trivial steady state.
43Analysis of PDE extinction or persistance
- Theorem.
- extinction with exponential rate
- bounded away from zero
- Proof for extinction uses entropy by calculating
- Proof for positivity
44Analysis of PDE extinction or persistance
- Theorem.
- extinction with exponential rate
- bounded away from zero
- Remark a similar alternative is found
- in many other nonlinear structured models
- (see D, Kim, Perthame for CML
- Calvez, Lenuzza et al. for prion equations
- Bekkal Brikci, Clairambault, Perthame for cell
cycle)
45Analysis of PDE Linearised stability of the
non trivial steady state
- Linearised equation around the steady state
- Method look for the sign of the real part of the
eigenvalues
46Analysis of PDE Linearised stability of the
non trivial steady state
- Eigenvalue problem
- Defining it gives
47Analysis of PDE Linearised stability of the
non trivial steady state
- Simplest case no feed-back on the maturation
process. - The characteristic equation becomes
- Proposition.
- If
- There is a Hopf bifurcation for one value of µ
gt0. - Proof look for purely imaginary solutions, which
are the - places where a bifurcation can occur.
48Analysis of PDE Linearised stability of the
non trivial steady state
- Case derived from the discrete model
- Proposition.
- If the maturation and the proliferation rates
are independent of maturity linear stability. -
- If proliferation rate varies instability may
appear. - Proof same ideas (but longer calculations)
49perspectives
- Comparison discrete continuous
- biological interpretation of analytical
constraints - What could give a measure of differentiation ?
- Opportunity of the discrete vs continuous
modelling ? - Inverse problems
- recover g from data of differentiated cells ?
- Mathematical challenge prove nonlinear
(in)stability by the use of entropy-type
arguments ?