Stability and Hopf Bifurcation in a
Pluripotent Stem Cell Dynamics
We study a mathematical model describing the dynamics of a pluripotent stem cell population involved in the blood production process in the bone marrow. This model is a differential equation with a time delay. The delay describes the cell cycle duration and is uniformly distributed on an interval. We obtain stability conditions independent of the delay. We also show that the distributed delay can destabilize the entire system. In particularly, it is shown that Hopf bifurcations can occur.
Laboratoire de Mathématiques Appliquées, FRE 2570,
Université de Pau et des Pays de l’Adour,
Avenue de l’université, 64000 Pau, France.
E-mail: email@example.com, firstname.lastname@example.org
Department of Mathematics, University of Miami,
P. O. Box 249085, Coral Gables, FL 33124-4250, USA.
Keywords: Blood production system, stem cells, delay differential equations, stability, Hopf bifurcation.
Blood production process, called hematopoiesis, is one of the major biological phenomena occurring in human body. It takes place in the bone marrow where pluripotent stem cells give birth to mature cells. After ejecting their nuclei, these cells enter the bloodstream and become blood cells.
According to the study of Burns and Tannock , the population of pluripotent stem cells can be divided into two distinct groups: quiescent cells and proliferating cells. Mathematical models describing the dynamics of this cell population have been studied since the end of the seventies, in particularly by Mackey [9, 10]. We refer to the review articles by Haurie et al.  and Mackey et al.  for further study and more references on this topic. More recently, Pujo-Menjouet et al.  and Pujo-Menjouet and Mackey  proved the existence of a Hopf bifurcation for the hematopoiesis model proposed in . In all these works, the authors assumed that the proliferating phase duration is constant. Mathematically, this means that the delay in their models is a discrete delay. However, experimental data (see Bradford et al. ) indicate that cells do not spend the same time in the proliferating phase.
In this paper, taking into account this assumption, we assume that the delay (or proliferating phase duration) is uniformly distributed on an interval. The main objective is to investigate the effect of time delay on the dynamical solutions. It is shown that there exist some critical values of time delay such that a local Hopf bifurcation occurs at the non-trivial equilibrium.
The paper is organized as follows. In section 2, we present our model, which is given in equation (1). In section 3, we derive stability conditions for the two equilibria of equation (1) which do not depend on the delay. We show the existence of Hopf bifurcations at the non-trivial equilibrium in section 4. A brief discussion is given in section 5.
2 The Model
Pluripotent stem cells can be either in a resting phase, also known as -phase, or in a proliferating phase. In the resting phase, they can die at a constant rate , which also includes the cellular differentiation, or be introduced in the proliferating phase at a rate . According to the work of Sachs , is assumed to depend on the resting phase population.
In the proliferating phase, which is in fact the so-called cell cycle, pluripotent stem cells are committed to divide and give birth to two daughter cells at the end of this phase. The two daughter cells enter directly the resting phase and complete the cycle. We assume that proliferating cells divide according to a uniform law on an interval with . This assumption comes from the fact that, even if only a little is known about phenomena involved in hematopoiesis, there are strong evidences (see Bradford et al. ) indicating that cells do not divide at the same age. The function is then defined by
Let denote the pluripotent stem cell population density (cells/kg) at time . It satisfies the nonlinear delay differential equation
The first term in the right-hand side of equation (1) accounts for the cellular loss due to mortality and cellular differentiation, , and introduction in the cell cycle, . The second term is for the division of proliferating cells into two daughter cells during mitosis. Proliferating cells are in fact resting cells introduced in the proliferating phase one generation earlier, so that the quantity appears with a time delay. The factor 2 is, of course, for the division of each proliferating cell into two daughter cells.
In the following, the rate of reintroduction in the proliferating compartment is taken to be a monotone and decreasing Hill function, given by
The coefficient is the maximum rate of reintroduction, is the -phase population density for which the rate of re-entry attains its maximum rate of change with respect to the resting phase population, and describes the sensitivity of with changes in the population. This function was firstly used in hematopoiesis models by Mackey  in 1978.
The value of is usually However, since we shall study the qualitative behavior of the pluripotent stem cells population, the value of is not really important and could be normalized without loss of generality.
Now if we consider an initial continuous nonnegative function defined on , then the equation (1) has a unique continuous and nonnegative solution , defined for , such that
This can be obtained by using the results in Hale and Verduyn Lunel .
Notice that equation (1) has at most two equilibria, the trivial equilibrium and a non-trivial positive equilibrium . The trivial equilibrium always exists and corresponds to the extinction of the population.
Equation (1) has a non-trivial positive equilibrium if and only if
In this case, is explicitly given by
Let be an equilibrium of equation (1). Then satisfies
Consequently, equation (1) has a non-trivial equilibrium if and only if the equation
In the next section, we shall study the stability of the two equilibria of equation (1).
Throughout this section, we are interested in the stability of the equilibria of equation (1), in particularly the stability of the non-trivial equilibrium . We start by giving a result on the global stability of the trivial equilibrium of (1).
The trivial equilibrium of equation (1) is globally stable if
The proof uses a similar technique employed by Adimy and Crauste . It is based on the construction of a Lyapunov functional.
Denote by the space of all continuous nonnegative functions on . Let be the function defined by
Consider the mapping defined, for , by
we obtain that
Let be the function defined, for , by
Assume that Since is a decreasing function, it follows that the function is positive for . Hence, is nonnegative on and if and only if . Consequently, the mapping is a Lyapunov functional when . We then deduce that the trivial equilibrium of (1) is globally stable. ∎
We do not expect to obtain conditions for the global stability of . However, local stability results can be obtained by linearizing equation (1) about . Set
The linearization of equation (1) at is
The characteristic equation of (1) is given by
We now state and prove our first result on the stability of .
Then the non-trivial equilibrium of equation (1) is locally asymptotically stable.
We show, in fact, that is stable when . By the definition of given by (4), it follows that
So we assume that .
We first assume that , given by (5), is a real function. Then, is continuously differentiable and its first derivative is given by
One can see that is positive for as soon as . Moreover,
Consequently, has a unique real root . Since
we deduce that .
Now, we show that if is a characteristic root of equation (5), then . By contradiction, we assume that there exists a characteristic root of equation (5) such that . By considering the real part of , we obtain that
This yields a contradiction. We conclude that every characteristic root of (5) is such that . Hence, all characteristic roots of (5) have negative real parts and the equilibrium is locally asymptotically stable. ∎
When , that is, when
the stability cannot occur for all values of and . In particularly, we shall show that a Hopf bifurcation can occur (see Theorem 4.1). However, we can still have the stability of the non-trivial equilibrium for values of , and if is not too large. This will be considered in the next theorem.
To present the results, without loss of generality we assume that
We want to point out that the results we are going to show remain true when , but the proof is more complicated.
Define a function , for , by
and let be the unique solution of the equation
We have the following local stability theorem.
Then the non-trivial equilibrium of equation (1) is locally asymptotically stable.
Let us assume that (9) holds. Then , and
By contradiction, assume that there exists a characteristic root of (5) with . Then,
Integrating by parts, we obtain that
If is such that
So we obtain a contradiction.
Similarly, if is such that
It implies that
Again we obtain a contradiction. Hence, all characteristic roots of (5) are such that .
Now, we assume that (5) has a purely imaginary characteristic root . Then and satisfy
Using (8) and (10), we obtain a contradiction. Consequently, (11) has no solution and equation (5) does not have purely imaginary roots. We conclude that all characteristic roots of (5) have negative real parts and is locally asymptotically stable. ∎
We are going to show that as soon as condition (12) does not hold, then the equilibrium can be destabilized. In the next section, we shall show that if condition (10) does not hold, then a Hopf bifurcation indeed occurs at .
4 Hopf Bifurcations
In this section we are going to show that the non-trivial equilibrium of equation (1) can be destabilized via Hopf bifurcations. The time delay will be used as a bifurcation parameter. This result is obtained in Theorem 4.1.
Recall that the non-trivial equilibrium of equation (1) exists if and only if In the following, without loss of generality we assume that
Again the results still hold when , but the proof is easier to understand when .
We look for purely imaginary roots of . Of course, we assume that , otherwise is locally asymptotically stable. Let , with , be a purely imaginary characteristic root of equation (5). Then, and satisfy the following system
First, one can see that cannot be a solution of (13). Otherwise
One can check that and are given, for and , by
Consider the sequence
In fact, one can check that
Moreover, for all ,
Define two sequences and , for , by
Using the definition of , one can see that
Thus, the sequence is increasing with and the sequence is decreasing with and for (see Figure 1).
Furthermore, one can check that, as soon as ,
Finally, define a function , for , by
We have the following results about the properties of the function
(i) If , then there exists such that
(ii) If , then there exists such that
Since the function is increasing on the interval , we can see that
is equivalent to
is equivalent to
The lemma now follows. ∎
We only prove when . The other cases can be deduced similarly. Assume that
This is equivalent to
The function is strictly negative and decreasing on with (see Figure 1). So the equation
has a unique solution on the interval . Set
Moreover, the function is strictly negative and increasing on with , so the equation has a unique solution on the interval . Set
Furthermore, the function is nonnegative on and
Finally, using the fact that
we obtain that
This completes the proof. ∎
Lemma 4.1 and Proposition 4.1 give conditions for the existence of pairs of purely imaginary roots of equation (5). In the next proposition, we study the properties of the purely imaginary roots of (5).
We will show that
A simple computation shows that
where the function is defined by
One can check that if and only if there exists a such that . Moreover,
This yields that
Hence, are simple characteristic roots of (5).
Let be a characteristic root of (5) such that . By separating the real and imaginary parts, we obtain that
We denote by (respectively ) the first derivative of (respectively ) with respect to . For , we obtain that
We consider two cases. First, assume that
One can verify that
Consequently, (19) is equivalent to
Then, it follows from equation (18) that
Hence, (20) implies that
Since we have
Furthermore, the sign of is the same as the sign of