Turing
Turing
Turing
Motivated by recent observations in biological cells, we study Turing patterns in bounded regions where the
nonlinear chemical reactions occur on the boundary and where reagent transport occurs in the bulk. Within a
generic model, we formulate the stability problem and discuss the conditions for the occurrence of a Turing
instability. By choosing other model parameters to be unequal, we find that Turing patterns exist even in the
case of equal diffusion constants. Finally, a recently introduced computation technique is utilized to follow the
nascent pattern into the highly nonlinear regime.
I. INTRODUCTION 关12,13兴兲, but there has not to date been a general investiga-
One of the by-now familiar paradigms of pattern forma- tion of the Turing possibility. We will demonstrate the con-
tion relies on the Turing instability which can occur in a ditions for the existence of the basic instability and will pro-
variety of reaction-diffusion systems 关1兴. This mechanism vide a computational paradigm 共based on the phase-field
requires an autocatalytic activation process coupled to a method兲 for investigating the resultant nonlinear state.
longer-ranged inhibition. The instability itself is character-
ized by having a finite wave-vector mode go unstable even
as the steady-state solution remains stable to uniform pertur- II. RESULTS
bations. Experimental realizations of Turing pattern forma- A. Stability analysis
tion via the use of gel reactors and comparisons to models
thereof have been extensively discussed 关2–4兴. Our model is a standard two-component reaction-
One of the primary motivations for Turing’s original sug- diffusion system where the reaction takes place on the 共inner
gestion was that these models might provide insight into the part of兲 the bounding surface of some bulk region. On this
formation of biological patterns. Although it remains unclear surface, we have
how widespread the Turing paradigm really is, there have
u̇m = − rdum + rau + f共um, vm兲,
been several examples of multicellular systems which have
been argued to lie in this pattern-formation class 关5,6兴. Our
interest here, however, is to begin the investigation of Turing v̇m = − pdvm + pav + g共um, vm兲. 共1兲
patterns on the single-cell scale. A specific impetus is the The processes whereby the bulk species u, v exchange with
suggestion that reaction-diffusion patterns along the cell surface-resident ones um, vm are described by the desorption
membrane may be implicated in the gradient-sensing ma- coefficients rd and pd and adsorption coefficients ra and pa. A
chinery employed by eukaryotic cells to perform directed specific example of the nonlinear production functions f, g
motion 共chemotaxis兲. One line of evidence for this idea will be given later. The equations in the bulk, where no non-
comes from the recent observation of spontaneous symmetry linear chemistry takes place, are
breaking in amoeboid cells exposed to uniform chemoattrac-
tant signals 共see Fig. 1兲 关7兴. If instead the cell is presented u̇ = Duⵜ2u − uu,
with a gradient, only one such spot forms and it is always at
the anterior. The simplest interpretation of this data is that v̇ = Dvⵜ2v − vv
these cells use the external gradient to modulate an internal
Turing-unstable system. This idea was originally suggested with as boundary condition for the normal derivatives
by Meinhardt 关8兴, but he did not explicitly formulate a
u
membrane-bound model. Du = rdum − rau,
Our goal here is not to propose a detailed model for this n
specific system; this will be presented elsewhere. Instead, we
use a generic model to investigate the concept that Turing v
Dv = p dv m − p av .
patterns can appear on membranes wherein there occur non- n
linear reactions involving bulk-diffusing species. Note that
this is different from previous studies of three dimensional For simplicity, we will first work in a two dimensional
Turing patterns which modeled uniform reaction-diffusion circular region of radius R. The steady-state solution of the
equations 关9,10兴. Models which couple intracellular diffusion model is straightforward. In the absence of bulk decay 共u
to surface reactions have appeared in several biological con- = v = 0兲 we have constant u and v in the bulk, equal, respec-
共0兲 共0兲
texts 共e.g., calcium waves 关11兴 and MIN oscillations tively, to rdum / ra and pdvm / pa; the membrane concentra-
␦ v m = − p d␦ v m + v aB n + g u␦ u m + g v␦ v m . 共3兲
The solvability of this system determines the growth rate .
To illustrate this result, we focus on the specific case of FIG. 2. The largest growth rate as a function of pd for the first
f共um , vm兲 = a共um
2
vm − um兲 and g共um , vm兲 = 1 − um
2
vm introduced four modes. Parameter values, in arbitrary dimensionless units, are
by Gierer and Meinhardt 关14兴. In general, the pair of coupled ra = rd = pa = Du = 1, a = 0.4, Dv = 10, u = v = 0.02, and R = 8.
061912-2
MEMBRANE-BOUND TURING PATTERNS PHYSICAL REVIEW E 72, 061912 共2005兲
061912-3
H. LEVINE AND W.-J. RAPPEL PHYSICAL REVIEW E 72, 061912 共2005兲
FIG. 5. The system can exhibit Turing patterns even in the case
of equal diffusion constants. Shown here is um along the membrane
of a circular region, parametrized by the angle . Parameter values
are Du = Dv = 2, ra = pa = 1, rd = 0.1, pd = 0.3, a = 0.05, u = 1, v = 0.3,
and R = 2. The simulation utilized the phase field with a width of
= 0.2 and a computational box containing a rectangular grid with
dimensions 120⫻ 120. For plotting purposes, the membrane is de-
fined as the points of the rectangular grid that are closest to the
perimeter of the disk. Consequently, the plotted field exhibits some
variation which has been reduced by a smoothing procedure.
061912-4
MEMBRANE-BOUND TURING PATTERNS PHYSICAL REVIEW E 72, 061912 共2005兲
061912-5