1: \documentclass[fleqn,11pt]{article}
2: \setlength{\textwidth}{14.8cm}
3: \setlength{\oddsidemargin}{0.46cm}
4: %\usepackage{latexsym}
5: \usepackage{graphicx}
6: \usepackage{amsmath}
7: \usepackage{enumerate}
8: \renewcommand{\baselinestretch}{1.6}
9: \raggedright
10:
11:
12: \begin{document}
13:
14: \noindent{\LARGE \textbf{Turing pattern with proportion preservation}}
15: \vspace{5mm}
16:
17: \noindent Shuji Ishihara$^{1}$ and Kunihiko Kaneko$^{1,2}$
18: \vspace{5mm}
19:
20: \noindent
21: $^1$Department of Pure and Applied Sciences, University of Tokyo,
22: Komaba, Meguro-ku, Tokyo 153-8902, Japan \\ $^2$ ERATO Complex Systems
23: Biology Project, JST\\
24: \vspace{5mm}
25:
26:
27: \noindent Correspondence should be addressed to S. Ishihara\\[-2mm]
28: \noindent E-mail : shuji@complex.c.u-tokyo.ac.jp \\[-2mm]
29: \noindent Department of Pure and Applied Sciences, College of Arts and Sciences,\\[-2mm]
30: \noindent The University of Tokyo, 3-8-1 Komaba, Meguro-ku, Tokyo 153-8902, Japan.\\[-2mm]
31: \noindent Tel/Fax : +81-3-5454-6731\\
32: \vspace{5mm}
33:
34:
35: Although Turing pattern is one of the most universal mechanisms for
36: pattern formation, in its standard model the number of stripes changes
37: with the system size, since the wavelength of the pattern is
38: invariant: It fails to preserve the proportionality of the pattern,
39: i.e., the ratio of the wavelength to the size, that is often required
40: in biological morphogeneis. To get over this problem, we show that
41: the Turing pattern can preserve proportionality by introducing a
42: catalytic chemical whose concentration depends on the system size.
43: Several plausible mechanisms for such size dependence of the
44: concentration are discussed. Following this general discussion, two
45: models are studied in which arising Turing patterns indeed preserve
46: the proportionality. Relevance of the present mechanism to biological
47: morphogenesis is discussed from the viewpoint of its generality,
48: robustness, and evolutionary accessibility.
49:
50: \vspace{2mm}
51:
52: {\bf Keywords}~:~Turing~pattern;~Morphogenesis;~size-invariance
53:
54:
55: \newpage
56: \parskip=3.5mm
57: \parindent=5mm
58: \section{Introduction \label{sec:intro}}
59:
60: ~~~~~Since the seminal paper by Turing (1952), a large variety of
61: pattern formation phenomena in nature has been explained by his
62: theory. The original motivation of Turing himself lied in the
63: explanation of biological morphogenesis, as was succeeded to Gierer
64: and Meinhardt (Gierer and Meinhardt, 1972), and others over a half
65: century (Gray and Scott, 1984; Murray 1993; Pearson, 1993; Kondo and
66: Asai, 1995; Meinhardt and Gierer, 2000). Although the Turing pattern
67: is one of the most beautiful and ubiquitous mechanisms for
68: morphogenesis, frequent criticism raised to it is non-adjustability of
69: the characteristic wave length of the pattern against the system size.
70: Because the generation of the Turing pattern comes from instability of
71: an uniform state over a certain range of wavelengths, the possible
72: range of the wavelengths is pre-fixed, and is invariant against the
73: change of the system size. Hence the number of segments or stripes is
74: proportional to the system size as shown in Fig.~\ref{fig:Prop}~(a).
75: In contrast, the number of segments or stripes, rather than the
76: wavelength, is often invariant against the change of the size in many
77: biological systems as shown in Fig.~\ref{fig:Prop}~(b). In a
78: biological system, the scale of pattern is often proportional to the
79: system size, and also it is desirable to have such proportionality in
80: many situations. However, neither the Turing pattern nor the
81: positional information theory by Wolpert (1969) which assumes the
82: morphogen gradient to convey positional information satisfies the
83: scale-invariance.
84:
85: For example, it is observed that the patterns in \textit{Hydra} and
86: \textit{Dictyostelium discoideum} slugs have proportionality with the
87: size. In \textit{Dictyostelium discoideum} the ratio of two different
88: cell types is almost fixed, independent of the size. Transgenic mice
89: indeed preserve the body proportion despite their larger size in
90: phenotype (Palmiter et al., 1982). In the development of
91: \textit{Drosophila Melanogaster}, it was reported that the expression
92: of gap gene \textit{hb} is robust and preserves the proportion over
93: different sizes of individual eggs (Houchmandzadeh et al., 2002). The
94: proportion preservation is important for robust morphogenesis in
95: general.
96:
97:
98: In the present paper we discuss a general mechanism which enables the
99: proportionality of wavelength with the size in Turing patterns as in
100: Fig.~\ref{fig:Prop}~(b).
101:
102: To explain the proportion regulation of different cell types for
103: \textit{Dictyostelium discoideum} slug, Meinhardt (1982) proposed an
104: activator-inhibitor model in which the ratio of two cell types is
105: preserved. The mechanism works well for a pattern with only a single
106: boundary between two types, but is not valid for multiple stripes
107: pattern. Theory based on globally coupled dynamical systems also
108: provides a regulation mechanism for the ratio of different cell types
109: (Kaneko and Yomo, 1994, 1999; Mizuguchi and Sano, 1995; Furusawa and
110: Kaneko, 2001), but a proportion preservation of a pattern with
111: multiple stripes is not discussed yet.
112:
113: As an extension of Turing pattern, Othmer and Pate (1980) showed that
114: if diffusion constants depend on the concentration of auxiliary
115: chemical factor which has size-dependence, then size-invariant pattern
116: formation is possible, where the size-dependence of the auxiliary
117: chemical concentration is provided by choosing a proper boundary
118: condition (see also Pate and Othmer, 1984; Dillon et al.,
119: 1994). Hunding and S\o rensen (1988) also discussed a simple mechanism
120: to explain such concentration-dependent diffusion by an auxiliary
121: chemical factor. In these models, it is necessary that all the
122: diffusion coefficients are regulated in the same manner. Apart from
123: Turing pattern, a model for the proportion regulation in
124: \textit{Drosophila Melanogaster} was proposed by Aegerter-Wilmsen et al.
125: (2005), also by assuming effectively concentration-dependent
126: diffusion. It is not yet sure if such regulation of diffusion is
127: really adopted to control the proportionality.
128:
129: In this paper we discuss mechanisms of scale-invariant Turing pattern
130: without considering any variations of diffusion constants. Instead we
131: seek for a possibility that concentration of some chemical changes
132: with some power of the system size, which influences the rate of
133: reaction for the Turing pattern so that the size-invariance is
134: generated. We introduce a chemical component whose concentration
135: depends on the size of the system. In Section \ref{sec:sdc} we discuss
136: several possibilities for such size-dependent concentration of a
137: chemical. Following this general discussion, we give two specific
138: examples leading to the Turing pattern whose wavelength is
139: proportional to the system size. The first example introduced in
140: Section \ref{sec:Model1} adopts a size-dependent auxiliary chemical
141: component, in the same way as Othmer and Pate (1980) or Hunding and
142: S\o rensen (1988), while we believe it is simpler than the earlier
143: models, and is also plausible biologically because change between
144: active and inactive forms adopted therein is ubiquitous in a
145: biochemical process. The second example introduced in Section
146: \ref{sec:Model2} contains only two chemical components, which is the
147: minimum number for the Turing instability. The model provides a novel
148: mechanism for the proportionality preservation based on the
149: conservation of some quantity. Due to the size dependence of the
150: conserved quantity, the scale-invariance in Turing pattern is
151: resulted. In Section \ref{sec:discussion} we summarize the mechanisms
152: for the proportion preservation and discuss possible relevance of them
153: to biological morphogenesis.
154:
155: Before discussing the mechanisms, we make one remark what we have in
156: mind with the term ``system'' in the present paper. In a class of
157: examples, the system refers to a cell, in which case the boundary of
158: the system is a membrane, while in some other cases, the system refers
159: to cell aggregates (or tissue). As a mathematical expression of
160: reaction-diffusion equation the two cases are treated in the same way,
161: and thus we discuss the two cases together by adopting the term
162: ``system''.
163:
164:
165:
166: \section{Size dependent concentration \label{sec:sdc}}
167:
168: ~~~~~Following Section \ref{sec:intro}, we seek for a mechanism in
169: which concentration of some chemical components changes with the
170: system size. Let us consider the case in which a single component W
171: satisfies the size-dependence in concentration. Here we will discuss
172: several possible mechanisms in which the concentration of W indeed
173: changes with some power of the system size. We take a
174: three-dimensional system with a size scale $\sim L$, and assume that
175: the size of the system (e.g., a cell) varies keeping conformity in
176: shape, so that the volume of the system is proportional to $L^3$,
177: while the area of the boundary increases with $L^2$. We further
178: assume that the diffusion of W is so rapid as $\sqrt{D_w/\gamma_w} >
179: L$ where $D_w$ and $\gamma_w$ are the diffusion coefficient and the
180: degradation rate of W respectively, so that W is distributed almost
181: homogeneously. This condition, however, is not so restrictive to
182: realize the size-dependence, and the argument below can be generalized
183: even by relaxing the condition.
184:
185:
186: \begin{enumerate}[\bfseries (A)]
187:
188: \item
189: Consider the case in which the total quantity of W is conserved
190: against the change of the system size (or through the growth), as
191: shown in Fig. \ref{fig:turingcase} (A). In this case the concentration
192: of W is proportional to $L^{-3}$ due to the conservation of the total
193: amount and the dilution by the increase of the size. Here the case
194: with only a single chemical component (W) is discussed, but $W$ can be
195: a sum of multiple components if the total sum of them is conserved.
196: In Section \ref{sec:Model2}, we discuss an example of such case, which
197: results in the size-invariant Turing pattern.
198:
199: \item
200: Consider the case in which W is generated whole through the system
201: at a rate $g$, while W escapes out of the system only through the
202: boundary, as in Fig. \ref{fig:turingcase} (B-(i)). As another
203: possibility, consider the case in which W is decomposed by enzymes bounded on the
204: membrane (if a system is a cell) or by specific cells that are located
205: at the boundary of a tissue as in Fig. \ref{fig:turingcase} (B-(ii)).
206:
207: The case (i) is the same as that discussed by Othmer and Pate
208: (1980). In this case, the boundary condition is represented by the
209: following equation
210: \begin{eqnarray}
211: -\vec{n}\cdot D_w \nabla w = b w \label{eq:bound}
212: \end{eqnarray}
213: where $\vec{n}$ is an unit vector perpendicular to the boundary, and
214: $b$ is the mass transfer coefficient of W at the surface of the
215: system. In the cases (ii), degradation of W is catalyzed only at the
216: boundary. Then $w(\vec{r},t)$ follows the equation
217: \begin{eqnarray}
218: \frac{\partial w}{\partial t} = D_w \triangle w + g - \gamma w
219: \delta(\vec{r}-\vec{r}_s), \label{eq:wB}
220: \end{eqnarray}
221: where $\vec{r}_s$ denotes the coordinate of the boundary.
222:
223: In both cases, W is distributed almost homogeneously in the system if
224: the diffusion coefficient is sufficiently large. In the steady state,
225: concentration $w$ in the system is evaluated by the integration, where
226: W is synthesized with the rate proportional to $L^3$ and is
227: decomposed in proportion to $L^2$. Thus the abundances of W are
228: proportional to $L$.
229:
230: It is often the case that the generation of some chemical factors are
231: limited at a localized region in a system. Bcd-protein in the embryo of
232: \textit{Drosophila} is an example, where Bcd-mRNA is fixed in
233: the anterior of the cell. In such case, the rate of synthesis of W is
234: independent of the system size ($L^0$), and thus the concentration of W is
235: proportional to $L^{-2}$.
236:
237: \item
238: As is shown in Fig. \ref{fig:turingcase} (C), W flows into a system
239: from (or is synthesized by a chemical factor from) the environment of
240: the system, and is decomposed within the system. Then the former rate
241: is proportional to $L^2$, and the latter to $L^3$, so that the
242: concentration of W is proportional to $L^{-1}$. This situation is
243: typical for morphogenesis where each part of the embryo transmits and
244: receives signals with each other. As another example, cAMP in a cell
245: is synthesized by the membrane-bound enzyme Adenylcyclase, and thus
246: the concentration of cAMP follows the above scaling relation.
247:
248: \item
249: Consider a chemical factor Z that is synthesized whole through the
250: system while it is degraded on the boundary. Then the concentration of Z,
251: $z$, is proportional to the system size ($z \propto L$). Also consider
252: a chemical W that is synthesized whole through the system while it is
253: degraded, catalyzed by two molecules of Z, such as $2Z + W \to 2Z +
254: G$, as shown in Fig.~\ref{fig:turingcase} (D). Then the concentration
255: of W is proportional to $ L^{-2}$. In general, cooperative reactions
256: as in this example can induce various $L$ dependence.
257: \end{enumerate}
258:
259:
260:
261: Of course, some other situations are possible in which $w$ depends on
262: the size of a system. Next we give specific examples of
263: reaction-diffusion equations that leads to the scale-invariant Turing
264: pattern, based on this scaling behavior of a chemical W. In these
265: models, chemical factors U and V regulate each other, which, we
266: assume, are impenetrable through the boundary (membrane).
267:
268:
269:
270:
271: \section{Model~I : Turing model with a size regulator\label{sec:Model1}}
272: \subsection{Controlling proportionality}
273: ~~~~~Here we give an example of size-invariant Turing pattern, based
274: on the chemical $W$ with the size-dependent concentration in Section
275: \ref{sec:sdc}. Consider a reaction-diffusion system composed of three
276: chemical components U, V, W. The concentrations of U and V at time $t$
277: and at position $\vec{r}$, $u(\vec{r},t)$ and $v(\vec{r},t)$, obey the
278: following equations
279: \begin{subequations}
280: \begin{eqnarray}
281: \frac{\partial u}{\partial t} & = & D_u \triangle u +f(u,v;w) \label{eq:Turinga}\\
282: \frac{\partial v}{\partial t} & = & D_v \triangle v +g(u,v;w). \label{eq:Turingb}
283: \end{eqnarray}
284: \end{subequations}
285: W is a factor controlling the reactions, which is a size-dependent
286: component at the same time. The reaction terms are represented by
287: $f(u,v;w)$ and $g(u,v;w)$, and the wavelength of U-V pattern is
288: controlled by the concentration of W ($w$). In general, the change of
289: $w$ is accompanied with the change of the homogeneous steady state
290: itself, and as a result the characteristic wave length at the unstable
291: uniform state may change in a complicated manner. Here, we just give
292: two simple classes of reaction equations that satisfy the
293: scale-invariant pattern formation.
294:
295: In the first case, all the reactions for U and V are homogeneously
296: regulated by W, i.e. $f(u,v) \propto w^{\mu}$ and $g(u,v) \propto
297: w^{\mu}$. Here, by the spatial scale transformation $x \to
298: x/w^{\mu/2}$, the $w$-independent differential equations are obtained
299: for a steady-state pattern. Such regulation was also assumed in
300: earlier study by Saunders and Ho (1995), but it might not be so
301: natural, as W regulates all the reaction process in the same manner.
302:
303: As far as we know, the second case has been slipped over, in
304: which the reactions for U and V have the functional forms
305: \begin{eqnarray}
306: f(u,v;w) = F(w^{\nu}u,w^{\nu}v), ~~~~g(u,v;w) =G(w^{\nu}u,w^{\nu}v). \label{eq:exp}
307: \end{eqnarray}
308: In this case a homogeneous fixed point given by the conditions
309: $f(u,v;w)=g(u,v;w)=0$ is realized at $\hat{u} \equiv
310: w^{\nu}u=\hat{u}^0$ and $\hat{v} \equiv w^{\nu}v=\hat{v}^0$, where
311: $(\hat{u}^0,\hat{v}^0)$ is the solution of $F(u,v)=G(u,v)=0$, so that
312: the linearized partial differentiations around the fixed point are
313: given by $f_u=w^{\nu} F_{\hat{u}}(\hat{u}^0,\hat{v}^0)$ and so forth.
314: By the transformation of $(u,v)$ to $(\hat{u},\hat{v})$, the equations
315: are invariant under the spatial scale transformation $x \to
316: x/w^{\nu/2}$,
317:
318: In the above two cases, the characteristic wavelength for the unstable uniform
319: steady state of U, V is controlled by the concentration of W. In the
320: next subsection, by taking a simple specific reaction-diffusion model
321: we show that this latter case arises rather naturally.
322:
323: \subsection{A reaction-diffusion model}
324: ~~~~~Here we study the following reaction-diffusion model based on
325: Brusselator (Prigogine and Lefever, 1968; Nicolis and Prigogine,
326: 1977), in addition to the size-regulator W;
327:
328: \begin{tabular}{clcccc}
329: \textbf{(\,I\,)} & U is generated by A at a constant rate &:& A $\stackrel{k_A}{\to}$ U \\
330: \textbf{(II)} & U is activated into U$^*$ by W with a reversible reaction &:& U $+$ W $\mathop{\rightleftharpoons}_{k^{-1}_U}^{k_U}$ U$^*$ \\
331: \textbf{(III)} & V is activated into V$^*$ by W with a reversible reaction &:& V $+$ W $\mathop{\rightleftharpoons}_{k^{-1}_V}^{k_V}$ V$^*$ \\
332: \textbf{(IV)} & U$^*$ changes to V$^*$ at a constant rate &:& U$^*$ $\stackrel{k_b}{\to}$ V$^*$ \\
333: \textbf{(V)} & Dimer of U$^*$ catalyzes V$^*$ into U$^*$ &:& 2U$^*$ + V$^*$ $\stackrel{k_a}{\to}$ 3U$^*$\\
334: \textbf{(VI)} & U$^*$ is degraded at a constant rate &:& U$^*$ $\stackrel{k_G}{\to}$ G
335: \end{tabular}
336:
337: \noindent
338: The model is illustrated in Fig.~\ref{fig:Brusselator}. In the model,
339: U and V have active and inactive states, and can react only in its active
340: state ``*''. At the same time, U and V are activated by W.
341:
342: Under a proper rescaling and redefinition of the parameters, the
343: rate equations for the system are given by
344: \begin{subequations}
345: \begin{eqnarray}
346: \dot{u}_i &=& A-k_U\,w u_i + k_U^{-1} u_a \\
347: \dot{u}_a &=& k_U \,w u_i -k_U^{-1} u_a - u_a - B u_a + u_a^2 v_a\\
348: \dot{v}_i &=& - k_V w v_i + k_V^{-1} v_a\\
349: \dot{v}_a &=& k_V w v_i - k_V^{-1} v_a + B u_a - u_a^2 v_a
350: \end{eqnarray}
351: \end{subequations}
352: where $u_i,~u_a,~v_i,~v_a$, and $w$ are concentrations of U, U$^*$, V,
353: V$^*$, and W respectively. Let us assume that the reversible reactions
354: between active and inactive states (II,~III) are sufficiently rapid
355: and in equilibrium. Then, the ratio of U to U$^*$ (V to V$^*$) is
356: given by a constant $u_a = \left(k_U\,w/ k_U^{-1} \right) u_i$ ~($v_a =
357: \left(k_Vb/ k_V^{-1} \right) v_i $), by which their terms
358: with $u_a$ and $v_a$ are replaced. As a result, we obtain the
359: equations of $u \equiv u_a+u_i$ and $v \equiv v_a+v_i$ as:
360: \begin{subequations}
361: \begin{eqnarray}
362: \dot{u} & = & A -m(w)\, u + m(w)^2 n(w) \, u^2 v- B m(w)\, u \\
363: \dot{v} & = & -m(w)^2 n(w) \,u^2 v + B m(w)\, u
364: \end{eqnarray}
365: \end{subequations}
366: where $m(w) = k_U\,w/(k_U\,w + k_U^{-1}) $ and $n(w) = k_V\,w/( k_V\,w
367: + k_V^{-1})$. Note that $u_a = m(w) u $ and $v_a = n(w) v$. In the
368: case $k_U^{-1} \gg k_U $ and $k_V^{-1} \gg k_V $, where inactive
369: states are dominant, $m(w) \sim (k_U / k_U^{-1}) w$ and $n(w) \sim
370: (k_V / k_V^{-1}) w$ approximately. Accordingly, the conditions of
371: Eq.~(\ref{eq:exp}) are satisfied with $\nu=1$. Notice that this
372: satisfaction of Eq.~(\ref{eq:exp}) is not specific to this model, but
373: is general when the chemicals have active and inactive states and only
374: the former participates in the reaction.
375:
376: Now we consider the situation given by \textbf{(B)} in
377: Section \ref{sec:sdc}, where W is synthesized at a limited domain in the
378: system. We just consider an one-dimensional pattern, and study a model
379: represented by the following partial differential equations;
380: \begin{subequations}
381: \begin{eqnarray}
382: \frac{\partial u}{\partial t } &= & D_U \frac{\partial^2 u}{\partial x^2} + A -m(w)\, u + m(w)^2 n(w) \, u^2 v- B m(w)\, u \\
383: \frac{\partial v}{\partial t } &= & D_V \frac{\partial^2 v}{\partial x^2} - m(w)^2 n(w) \,u^2 v + B m(w)\, u \\
384: \frac{\partial w}{\partial t } &= & D_W \frac{\partial^2 w}{\partial x^2} + H(x) - \gamma w \label{eq:PB3}
385: \end{eqnarray}
386: \label{eq:simueq}
387: \end{subequations}
388: \hspace{-1.5mm}where $H(x)=1$ for $0<x<x_0$ and 0 otherwise. To represent the
389: synthesis of W in some definite area of the system irrespective of the
390: system size, the constant $x_0$ is independent of $L$ and is set at
391: $1.0$ in the simulation. The last term $-\gamma w$ represents the
392: escape (or decomposition) of W through the surface. We have carried
393: out simulations on the temporal evolution of $u_n, v_n ~ (n=0 \sim
394: N-1)$ under Dirichlet boundary condition $U(0)=U(L)=V(0)=V(L)=0.0$.
395: Since we consider a one-dimensional direction of a three dimensional
396: system with the size $L$, the flow-out of the escaping chemicals at
397: the boundary should be proportional to $L^2$, so that the $\gamma$
398: term in the above model equation should be scaled by $L^{2}$. Under
399: these assumptions on $H(x)$ and $\gamma$, the discussion in
400: \textbf{(B)} of Section \ref{sec:sdc} is valid. Indeed, we have confirmed
401: that the concentration $w$ is proportional to $L^{-2}$, in the
402: simulation for large $D_W$. Thus we take $w=W_0 L^{-2}$ for the
403: simulations below.
404:
405: In Fig.~\ref{fig:model1x-L}, we plot the wavelength $\xi$
406: corresponding to the wavenumber that leads to the largest eigenvalue
407: in the linear stability analysis, for a given system size $L$. $\xi$
408: increases in proportion to $L$, and thus the generated pattern from
409: this instability is expected to preserve the proportion.
410:
411:
412:
413: The results of the simulation are shown in
414: Fig.~\ref{fig:SimulationTuring}, which clearly show that the number of
415: stripes does not change against the change of system size $L$ as long
416: as it is sufficiently large. For small size, because $w$ is large,
417: the saturation in the terms in $m(w)$ (or $n(w)$) is not negligible,
418: so that the number of stripes is decreased.
419:
420:
421:
422: \subsection{Simplicity of the mechanism \label{sec:CondTuring}}
423: ~~~~~The above example gives us a simple but plausible model for the
424: regulation of the wave length, which leads to the
425: proportion preservation for a pattern generated by Turing instability. As shown in
426: Fig.\ref{fig:TuringMech}, conditions for this proportion preservation
427: are summarized as follows:
428: \begin{enumerate}[\sffamily\bfseries (i)]
429: \item Each chemical component has active and inactive states.
430: The activation is reversible and is catalyzed by a chemical factor W
431: (whose concentration changes with the size). Only chemicals in the
432: active state can participate in reactions.
433: \item The reaction-diffusion system shows Turing instability.
434: \item W is generated at a localized region in the system (cell), diffuses rapidly, and
435: goes out of, or is degraded on, the surface (membrane).
436: \end{enumerate}
437: These are the only conditions for the proportion preservation of a
438: Turing pattern, which works regardless of the specific choice of a
439: model. The condition \textbf{\textsf{(iii)}} can be replaced by some
440: other conditions in which the concentration of the factor is scaled as
441: $w \propto L^{-2}$. Notice that the above conditions are independent
442: of each other so that they would be easily satisfied by combining each
443: process that satisfies each condition. Thus, a size-invariant Turing
444: pattern by the above conditions may be achieved easily through the
445: evolution.
446:
447:
448:
449:
450:
451: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
452: \section{Model~II : Model with a conserved quantity\label{sec:Model2}}
453: \subsection{Size invariant Turing instability by the conserved quantity}
454: ~~~~~Here we give another model with two chemical components, which are
455: regarded as two states of a single chemical species. At the same time
456: both U and V are not synthesized or decomposed, so that the total
457: quantity of the chemical components is conserved. According to the
458: mechanism \textbf{(A)} discussed in Section \ref{sec:sdc}, we seek for the
459: possibility of the proportion preservation in this model. For simplicity,
460: we assume one-dimensional system in this section.
461:
462: Consider two components U and V that regulate the concentration of each other
463: through a reaction, as shown in Fig.~\ref{Fig:Model2}. Then the reaction-diffusion
464: equations are represented by
465: \begin{subequations}
466: \begin{eqnarray}
467: \frac{\partial u}{\partial t} &=& D_u \frac{\partial^2 u}{\partial x^2} + F(u,v) \\
468: \frac{\partial v}{\partial t} &=& D_v \frac{\partial^2 v}{\partial x^2} - F(u,v)
469: \end{eqnarray}
470: \label{eqn:rd}
471: \end{subequations}
472: Hence the total quantity of U and V is conserved;
473: \begin{eqnarray}
474: S \equiv \int dx \left( u+v \right) = \mbox{constant}.
475: \end{eqnarray}
476: As discussed in the case \textbf{(A)} in Section \ref{sec:sdc}, the increase in
477: the system size leads to the dilution of the concentration of $S$.
478:
479: In the steady homogeneous state, the Jacobian matrix for the reaction
480: terms is given by
481: \begin{eqnarray}
482: J=\left(
483: \begin{array}{cc}
484: F_u & F_v \\
485: -F_u & -F_v
486: \end{array}
487: \right)
488: \end{eqnarray}
489: where $F_u$ denotes the partial derivative of $F$ by $u$ at a
490: homogeneous steady state of Eq.~(\ref{eqn:rd}), and so forth. Through
491: the stability analysis of the Fourier transform of the linearized
492: equation around the homogeneous state by using $J$, the wave number
493: that has the largest eigenvalue is obtained, which gives the wavenumber
494: that grows most rapidly from this unstable homogeneous state. This
495: wavenumber is given by
496: \begin{eqnarray}
497: k_m^2 = \frac{-\hat{D}(F_u+F_v)+ (1+\hat{D}) \sqrt{F_u F_v \hat{D}}}{\hat{D} (\hat{D}-1)}
498: \label{eq:Awacond}
499: \end{eqnarray}
500: with $\hat{D}\equiv D_v/D_u$, where $\hat{D}$ is larger than $1$ for
501: Turing instability to occur (Turing, 1952). To preserve the
502: proportional pattern by increasing the length of the system $L$, it is
503: necessary that both $F_u$ and $F_v$ behave so as to $k_m$ scales as
504: $L^{-1}$ in Eq.~(\ref{eq:Awacond}), at least approximately. Below, we
505: give an explicit example corresponding to this case.
506:
507:
508: \subsection{An explicit reaction-diffusion model with a conserved quantity}
509: ~~~~~Consider the following reaction diffusion system corresponding to the
510: reactions shown in Fig.~\ref{Fig:Model2}.
511: \begin{subequations}
512: \begin{eqnarray}
513: \frac{\partial u}{\partial t} &=& D_u \frac{\partial^2 u}{\partial x^2} + u^3v - Bu^2 \\
514: \frac{\partial v}{\partial t} &=& D_v \frac{\partial^2 v}{\partial x^2} - u^3v + Bu^2 \end{eqnarray}
515: \label{eq:simueq2}
516: \end{subequations}
517: This system is a modified version of the Brusselator, so that supplies
518: and degradations of substances are excluded (Awazu and Kaneko,
519: 2004). Although the reaction term $u^3v$ is higher than the original
520: Brusselator, indeed, the reaction with a lower order cannot satisfy
521: the requirement of the last subsection for the proportion
522: preservation. As far as we have examined, this choice is one of the
523: simplest to satisfy the requirement (See Appendix A). Also, in a
524: biological system, such high order catalysis is not so uncommon. Hence
525: we adopt this reaction model.
526:
527: In this system, the corresponding homogeneous fixed point $(u_0,v_0)$
528: is given by $u_0v_0=B$ and $u_0+v_0=S/L$. Note that $u_0$ is almost
529: proportional to $S$ as long as $S$ is sufficiently large. Jacobian
530: around the uniform steady state is given by
531: \begin{eqnarray}
532: J=\left( \begin{array}{cc} Bu_0& {u_0}^3 \\ -Bu_0& -{u_0}^3
533: \end{array}
534: \right)
535: \end{eqnarray}
536: If $\hat{D}$ is large enough, the dominant term in
537: Eq.~(\ref{eq:Awacond}) is the one containing $\sqrt{F_u F_v}$, because
538: other terms are lower order with regards to $\hat{D}$. Thus, $k_m^2
539: \sim \sqrt{{u_0}^4} \sim S^2$ holds in the model, which results
540: in the proportion preservation.
541:
542: We plot the characteristic wavelength $\xi=2\pi/k_m$, corresponding to
543: the most unstable mode given by the linear analysis of $J$, for the
544: system size $L$ in Fig.~\ref{fig:model2anal}. $\xi$ increases in
545: proportion to $L$ over a wide range of $L$ for sufficiently large $S$,
546: thus leading to a size-invariant pattern. We have carried out
547: numerical simulation of Eq.~(\ref{eq:simueq2}). The results are shown
548: in Fig.~\ref{fig:SimulationTuring2}, which support the above
549: estimation to realize the size-invariant Turing pattern formation.
550:
551:
552: The model we give here is one of the simplest, in the sense that it
553: contains the lowest order polynomial reaction term among such
554: equations, as is also explained in the Appendix A, where more detailed
555: estimations as well as some other equations leading to the
556: size-invariant pattern formation.
557:
558:
559:
560: \section{Summary and Discussion \label{sec:discussion}}
561:
562: ~~~~~In this paper, we have discussed a possible mechanism of
563: proportion regulation based on the control of the reaction rate in
564: reaction-diffusion systems. We have introduced a morphogen W which
565: itself does not convey positional information (Wolpert, 1969), but
566: works as a carrier of information on the size of the system. It is
567: important to recognize that the proportion preservation is possible by
568: such simple mechanism. We have discussed several possible schemes
569: that can naturally realize the proposed mechanism in a biochemical
570: system.
571:
572: In some earlier studies and in our model, it is assumed that there are
573: chemical factors whose concentration depends on the system size. As
574: discussed by Hunding and S\o rensen (1988), a candidate of such
575: chemical component is cAMP, which is synthesized by the membrane-bound
576: enzyme Adenylcyclase. Indeed, cAMP is involved in a number of
577: important biological processes.
578:
579: Some proteins can also fit as the size regulator W. A candidate is the
580: product of the gene \textit{staufen} (\textit{stau}) working in the
581: early development of \textit{Drosophila}. Houchmandzadeh et
582: al. (2002) reported that in the \textit{Drosophila} embryo, the domain
583: boundary of zygotic gene \textit{hunchback} (\textit{hb}) expression
584: is tuned precisely at a half of the embryo, despite individually
585: fluctuating embryo size and expression of its direct regulator
586: Bicoid. It is discussed that maternal gene
587: \textit{stau} may play a major role to control such positioning of
588: \textit{hb} expression, as mutants lacking \textit{stau} lose precise
589: expression boundary of
590: \textit{hb} at the half of the embryo. Although a model based on the
591: effective change of diffusion constant was proposed by
592: Aegerter-Wilmsen et al. (2005) recently, it may be interesting to seek
593: for the possibility that \textit{stau} may work as a size-regulator W
594: with the mechanism (A) or (B) in Section \ref{sec:sdc}. In general,
595: it will be interesting to search for some molecules that work as size
596: regulators, or carry out a knock-out experiment on a candidate
597: molecule of such regulator.
598:
599: In Section \ref{sec:Model1}, we have introduced a simple model in
600: which the Turing pattern is size-invariant. The proposed scheme for
601: the proportion preservation there is rather general and robust, and at
602: the same time is naturally realized in a biological system. We give
603: three conditions for the scheme, which are rather simple and plausible
604: to be achieved in biological morphogenesis. Additionally these
605: conditions are independent of each other, which is a good feature from
606: an evolutionary viewpoint, because they can be established one by one
607: through evolution, without any influence with each other. Consider
608: two neighbor species with similar proportional organization, but with
609: different sizes. Most of the genes are common between the two, and
610: they may share the same diffusion coefficients for most of their
611: products. Under these conditions, ordinary Turing pattern or other
612: mechanisms cannot explain conformity in their morphology for two
613: species with different sizes. On the other hand, in the mechanism we
614: proposed in Sec. \ref{sec:Model1}, control of solely a single chemical
615: W can lead to proper adaptive patterning. This is one of evolutionary
616: advantages of the present mechanism.
617:
618: Independence of each condition is also a good feature for an
619: experimetal reailzation of the present mechansim. One can realize the
620: present proportion-prserving Turing pattern based on the established
621: experimental examples (Castets et al., 1990; Ouyang and Swinney
622: 1990). Because ordinary Turing pattern has just one definite
623: wavenumber, it is interesting experimetally to construct a pattern
624: whose intrinsic scale is changed flexibly according to environmental
625: conditions, using the mechanism discussed in this paper.
626:
627: In Section \ref{sec:Model2}, we have given an example for the
628: size-invariant Turing pattern realized by chemical reactions with a
629: conserved quantity as proposed in (\textbf{(A)} in Section
630: \ref{sec:sdc}). Existence of such conserved quantity in morphogenesis
631: may be a rather natural assumption. However, in contrast to the
632: models (conditions) given in Section \ref{sec:Model1}, the model here
633: may lack generality, since the condition for the scale-invariance,
634: i.e., the combination of exponents, may be rather specific. Also,
635: accurate control for the initial value of the conserved quantity $S$
636: (the total quantity of U and V) may be required. In this sense,
637: search for the mechanism in Section \ref{sec:Model1} may be more
638: important in a biological context.
639:
640: As another possible explanation for the size-invarint pattern, one
641: could assume that Turing mechanism works only in certain period of
642: early development, leading to a pattern with differentiated cell
643: types, and then the cells grow at the same rate, keeping the
644: proportionality. In general, this mechanism has a low tolerance for
645: the individual fluctuation of body size at the earlier stage of
646: development (Houchmandzadeh et al., 2002). Also the growth process
647: keeping the proportionality is required, and the mechanism is
648: vulnerable by disturbance through the development. On the other hand,
649: in our mechanism the proportionality is tolerant against size
650: fluctuations, and is recoverable again such disturbances or external
651: manipulation.
652:
653: At last we give a speculation on a relationship between the size
654: regulation and pattern formation. In general the size of an organ is
655: much flexibly regulated by the mutual compensation between cell size
656: and cell number (Frabkenhauser, 1945; Potter and Xu, 2001). However,
657: the mechanism of size control in development is not so clear
658: yet. Potter and Xu (2001) discuss the relationship between size
659: regulation and pattern formation, in which mutations in genes
660: regulating pattern result in the changes in total tissue mass. If the
661: size regulator W discussed in this paper also concerns with size
662: control through its concentration, the pattern formation process is
663: tightly coupled with the organ size. It may be interesting to seek for
664: this possibility, since the present mechanism then allows for adaptive
665: control of the pattern scale as well as the organ size.
666:
667: In conclusion, we have shown that morphogenesis with proportion
668: preservation is possible under Turing instability, by simply utilizing
669: a catalytic molecule whose concentration is properly scaled with the
670: system size.
671:
672:
673: \section*{Appendix A : On the Equations (\ref{eq:simueq2})}
674: ~~~~~Here we explain why we choose Eq.~(\ref{eq:simueq2}) as a model to
675: satisfy $k_m \sim L^{-1}$. First consider the reaction-diffusion
676: equation in a polynomial form
677: \begin{subequations}
678: \begin{eqnarray}
679: \frac{\partial u}{\partial t} &=& D_u \frac{\partial^2 u}{\partial x^2} + u^mv^n - Bu^l \\
680: \frac{\partial v}{\partial t} &=& D_v \frac{\partial^2 v}{\partial x^2} - u^mv^n + Bu^l
681: \end{eqnarray}
682: \end{subequations}
683: Then the steady uniform solution $(u_0,v_0)$ is given by
684: \begin{subequations}
685: \begin{eqnarray}
686: u_0^{m-l}v_0^n&=&B \label{eq:uvst1}\\
687: u_0+v_0&=&S/L \label{eq:uvst2}
688: \end{eqnarray}
689: \label{eq:uvsteady}
690: \end{subequations}
691: and the Jacobian at this solution is given by
692: \begin{eqnarray}
693: J=\left(
694: \begin{array}{cc}
695: (m-l) B u_0^{l-1} & n Bu_0^l v_0^{-1}\\
696: -(m-l) B u_0^{l-1} & -n Bu_0^l v_0^{-1}
697: \end{array}
698: \right)
699: \end{eqnarray}
700: The steady state solution $(u_0,v_0)$ is represented as crossing
701: points of Eq.~(\ref{eq:uvst1}) and (\ref{eq:uvst2}). When $m>l$, the
702: relationship is represented as in Fig. \ref{fig:uvApend1} on the
703: $u$-$v$ plane. Then the crossing point satisfies $u_0 \propto S/L$
704: approximately for large $S$. On the other hand, if $m<l$, $v_0$ is an
705: increasing function of $u_0$ in the relationship
706: Eq.~(\ref{eq:uvst1}). Then the crossing point does not satisfy $u_0
707: \propto S/L$. Hence we assume $m>l$ here.
708: Recall that if $D_v/D_u$ is sufficiently large, the leading term
709: determining the characteristic scale length $\xi$ is given by
710: $\sqrt{-F_vG_u}$ ~(Eq.~(\ref{eq:Awacond})),
711: \begin{eqnarray}
712: \xi^{-2} = k_m^{2} \sim \left( u_0^{2l-1} v_0^{-1} \right)^{1/2} \sim
713: u_0^{\frac{2l-1}{2}+\frac{m-l}{2n}} \sim
714: L^{-\frac{2l-1}{2}-\frac{m-l}{2n}}
715: \end{eqnarray}
716: (note $v_0 \sim u_0^{(l-m)/n}$ from Eq.~(\ref{eq:uvst1})). Since the
717: exponent has to be -2 to sustain the proportionality, we get the condition
718: \begin{eqnarray}
719: n=\frac{m-l}{5-2l}
720: \end{eqnarray}
721: Suppose that $n,~m,~l$ are positive integers. Then because $m-l>0$,
722: $l$ can be only $1$ or $2$. Now by choosing $m=3, n=1, l=2$
723: Eq.~(\ref{eq:simueq2}) (Model II) is derived, which is the system with
724: the lowest degree of exponent. Another choice will be $m=4, n=1, l=1$
725: which leads to the following equation:
726: \begin{subequations}
727: \begin{eqnarray}
728: \frac{\partial u}{\partial t} &=& D_u \frac{\partial^2 u}{\partial x^2} + u^4v - Bu \\
729: \frac{\partial v}{\partial t} &=& D_v \frac{\partial^2 v}{\partial x^2} - u^4v + Bu
730: \end{eqnarray}
731: \label{eq:ModelIIb}
732: \end{subequations}
733:
734:
735: %%%%% Acknowledgment %%%%%
736: \section*{Acknowledgment}
737: The authors are grateful to A. Awazu, K. Fujimoto, T. Shibata, and
738: H. Takagi for discussions.
739:
740:
741:
742:
743: %%%%%%% Refereces %%%%%%%%%%%%
744:
745: \section*{References}
746:
747: \noindent Aegerter-Wilmsen, T., Aegerter, C. M., Bisseling, T., 2005.
748: Model for the robust establishment of precise proportions in the early
749: Drosophila embryo. J. Theor. Biol. 234, 13-19
750:
751: \noindent Awazu, A., Kaneko, K., 2004. Is relaxation to equilibrium
752: hindered by transient dissipative structures in closed systems?
753: Phys. Rev. Lett. 92, 258302
754:
755: \noindent Castets, V., Dulos, E., Boissonade, J., and De Kepper, P., 1990.
756: Experimental evidence of a sustained standing Turing-type
757: nonequilibrium chemical pattern. Phys. Rev. Lett. 64, 2953-2965
758:
759: \noindent Dillon, R., Maini, P. K., Othmer, H. G., 1994. Pattern formation
760: in generalized Turing systems I. Steady-state patterns in systems with
761: boundary conditions. J. Math. Biol. 32, 345-393.
762:
763: \noindent Frankenhauser, G., 1945. The effects of changes in chromosome
764: number on amphibian development. Q. Rev. Biol. 20, 20-78
765:
766: \noindent Furusawa, C., Kaneko, K., 2001. Theory of Robustness of Irreversible
767: Differentiation in a Stem Cell System: Chaos Hypothesis.
768: J. Theor. Biol. 209, 395-416
769:
770: \noindent Gierer, A., Meinhert, H., 1972. A theory of biological pattern
771: formation. Kybernet 12, 30-39
772:
773: \noindent Gray, P., Scott, S. K., 1984. Autocatalytic reactions in the
774: isothermal continuous stirred tank reactor: oscillations and
775: instabilities in the system a + 2b $\to$ 3b; b $\to$
776: c. Chem. Eng. Sci. 39,1087-1097.
777:
778: \noindent Houchmandzadeh, B., Wieschais, E., Leibler, S., 2002.
779: Establishment of developmental precision and proportions in the early
780: Drosophila embryo. Nature 415, 798-802
781:
782: \noindent Hunding, A., S\o rensen, P. G., 1988. Size adaptation of Turing
783: prepatterns. J. Math. Biol. 26, 27-39
784:
785: \noindent Kaneko, K., Yomo, Y., 1994. Cell Division, Differentiation, and
786: Dynamic Clustering. Physica D 75, 89-102
787:
788: \noindent Kaneko, K., Yomo, T., 1999. Isologous Diversification for Robust
789: Development of Cell Society. J. Theor. Biol. 199, 243-256
790:
791: \noindent Kondo, S., Asai, R., 1995. A reaction-diffusion wave on the
792: skin of the marine angelfish Pomacanthus. Nature 376, 765-768.
793:
794: \noindent Meinhardt, H., 1982. Models of Biological Pattern Formation.
795: Academic Press New York
796:
797: \noindent Meinhardt, H., Gierer, A., 2000. Pattern formation by local
798: self-activation and lateral inhibition BioEssays 22, 753-760.
799:
800: \noindent Mizuguchi, T., Sano, M., 1995. Proportion regulation of Biological
801: Cells in Globally Coupled Nonlinear Systems. Phys. Rev. Lett. 75,
802: 966-969.
803:
804: \noindent Murray, J. D., 1993. Mathematical Biology (2nd ed.)
805: Spriger-Verlag
806:
807: \noindent Nicolis, G., Prigogine, I., 1977. Self Organization in
808: Non-Equilibrium Systems. J. Wiley and Sons, New York
809:
810: \noindent Othmer, H. G., Pate, E., 1980. Scale-invariance in
811: reaction-diffusion models of spatial pattern
812: formation.~Proc.~Natl.~Acad.~Sci. 77, 4180-4184
813:
814: \noindent Ouyang, Q. and Swinney, H. L., 1991
815: Transition from a uniform state to hexagonal and striped Turing patterns.
816: Nature 352, 610-611
817:
818: \noindent Potter, C. J. and Xu, T., 2001. Mechanism of size control.
819: Curr. Opin. Gen. Dev. 11. 279-286
820:
821: \noindent Palmiter, R.D., Brinster, R. L., Hammer, R.E., Trumbauer, M.E.,
822: Rosenfeld, M.G., Birnberg, N.C., Evans. R.M., 1982. Dramatic growth of
823: mice that develop from eggs microinjected with metallothionein-growth
824: hormone fusion genes. Nature 300,611-615.
825:
826: \noindent Pate, E., Othmer, H. G., 1984. Application of a model for
827: scale-invariant pattern formation on developing
828: systems. Differentiation~28, 1-8.
829:
830: \noindent Pearson, J. E., 1993.
831: Complex Patterns in a Simple System. Science 261, 189-192
832:
833: \noindent Prigogine, I., Lefever, R., 1968. Symmetry Breaking
834: Instabilities in Dissipative Systems II. J. Chem. Phys. 48,
835: 1695-1700.
836:
837: \noindent Saunders, P. T., Ho, M. W., 1995. Reliable segmentation by
838: successive bifurcation. Bull. Math. Biol. 57, 539-556.
839:
840: \noindent Turing, A. M., 1952. The chemical basis of Morphogenesis.
841: Philos. Trans. Roy. Soc. Lond. B. 237, 37-72.
842:
843: \noindent Wolpert, L., 1969. Positional information and the spatial
844: pattern of cellar differentiation. J. Theor. Biol. 25, 1-47.
845:
846:
847:
848:
849:
850: %%%%%%%% Figure legends %%%%%%%%%%%w
851: \newpage
852:
853: \noindent Figure \ref{fig:Prop} : Pattern formation (a) with a fixed wave
854: length and (b) with a fixed proportion. Ordinary Turing pattern
855: belongs to the class (a), as the pattern arises by instability of a
856: certain range of wave lengths.
857:
858: \noindent Figure \ref{fig:turingcase} : Several cases for the scaling
859: behavior of the concentration of W against the system size, as
860: discussed in the text.
861:
862: \noindent Figure \ref{fig:Brusselator} : Two-state Brusselator model.
863: Only chemical components in the active state can react in the model.
864:
865: \noindent Figure \ref{fig:model1x-L} : Characteristic length $\xi$ of
866: the pattern in the two-state Brusselator model (Eq.~(\ref{eq:simueq}))
867: plotted against the system size $L$. The parameters are $k_U=0.5,
868: k_U^{-1}=10.0, k_V=0.1, k_V^{-1}=2.0, A=2.0, B=4.0, W_0=5.0 \times
869: 10^4$. We plot for $\hat{D} \equiv D_v/D_u = 4.0, 6.0 8.0,
870: 10.0$. Normalized wavelengths by the system size ($\xi/L$) are plotted
871: in the inset.
872:
873: \noindent Figure \ref{fig:SimulationTuring} : Pattern of the two-state
874: Brusselator model in Eq.~(\ref{eq:simueq}) obtained numerically, for
875: the system size $L=$ 64, 128, 256, 512, and 1024. The parameters are
876: $k_U=0.5, k_U^{-1}=10.0, k_V=0.1, k_V^{-1}=2.0, A=2.0, B=4.0, D_U=0.5,
877: D_V=3.0, W_0=5.0 \times 10^4$. The number of stripes are invariant
878: over a wide range of system size, while for too small system size, the
879: proportion is no longer sustained due to the nonlinearity (saturation)
880: in $m(w)$ and $n(w)$. Simulations are carried out with the grid size
881: 1.
882:
883: \noindent Figure \ref{fig:TuringMech} : A scenario for proportion
884: preservation of a Turing pattern.
885:
886: \noindent Figure \ref{Fig:Model2} : A model with two states of morphogen
887: U and V whose sum is conserved, where reaction between them to
888: regulate the concentration brings about Turing instability.
889:
890: \noindent Figure \ref{fig:model2anal} : Characteristic length $\xi$ of
891: the pattern by Eq.~\ref{eq:simueq2} plotted against the system size
892: $L$, for various values of the total quantity $S\,(= \int
893: u(x)+v(x)dx)$. For large $S$, $\xi$ increase in proportion to the
894: system size $L$. $B=0.5, \hat{D}=5.0\times 10^2$. Normalized
895: wavelengths by the system size ($\xi/L$) are plotted in the inset.
896:
897: \noindent Figure \ref{fig:SimulationTuring2} : Simulation results of
898: the modified Brusselator model with a conserved quantity in
899: Eq.~(\ref{eq:simueq2}) for various size $L=$64, 96, 128, 160, 192, 224
900: and 256. The number of stripes are $6$ and invariant for $L \ge 128$.
901: The parameters are $D_u=1.0 \times 10^2 , D_v=1.0 \times 10^4, B=0.5$,
902: $S=2560.0$.
903:
904: \noindent Figure \ref{fig:uvApend1} : Steady state solution plotted in
905: $u$-$v$ plane. The crossing points of the line and the curve give
906: homogeneous steady solutions of reaction equations
907: Eq.~(\ref{eq:uvsteady}).
908:
909:
910:
911:
912:
913: %%%%%%%% Figures %%%%%%%%%%%
914:
915: \newpage
916:
917: \begin{figure}[tbhp]
918: \centering
919: \includegraphics[width=.8\textwidth]{Fig1.eps}
920: \caption{Ishihara and Kaneko}
921: \label{fig:Prop}
922: \end{figure}
923:
924: \begin{figure}[tbhp]
925: \centering
926: \includegraphics[width=1.\textwidth]{Fig2.eps}
927: \caption{Ishihara and Kaneko}
928: \label{fig:turingcase}
929: \end{figure}
930:
931: \begin{figure}[tbhp]
932: \centering
933: \includegraphics[width=1.\textwidth]{Fig3.eps}
934: \caption{Ishihara and Kaneko}
935: \label{fig:Brusselator}
936: \end{figure}
937:
938: \begin{figure}[tbhp]
939: \centering
940: \includegraphics[width=1.\textwidth]{Fig4.eps}
941: \caption{Ishihara and Kaneko}
942: \label{fig:model1x-L}
943: \end{figure}
944:
945: \begin{figure}[tbhp]
946: \centering
947: \includegraphics[width=1.\textwidth]{Fig5.eps}
948: \caption{Ishihara and Kaneko}
949: \label{fig:SimulationTuring}
950: \end{figure}
951:
952:
953: \begin{figure}[tbhp]
954: \centering
955: \includegraphics[width=1.\textwidth]{Fig6.eps}
956: \caption{Ishihara and Kaneko}
957: \label{fig:TuringMech}
958: \end{figure}
959:
960: \begin{figure}[tbhp]
961: \centering
962: \includegraphics[width=.6\textwidth]{Fig7.eps}
963: \caption{Ishihara and Kaneko}
964: \label{Fig:Model2}
965: \end{figure}
966:
967: \begin{figure}[tbhp]
968: \centering
969: \includegraphics[width=.9\textwidth]{Fig8.eps}
970: \caption{Ishihara and Kaneko}
971: \label{fig:model2anal}
972: \end{figure}
973:
974:
975: \begin{figure}[tbhp]
976: \centering
977: \includegraphics[width=1.\textwidth]{Fig9.eps}
978: \caption{Ishihara and Kaneko}
979: \label{fig:SimulationTuring2}
980: \end{figure}
981:
982:
983: \begin{figure}[tbhp]
984: \centering
985: \includegraphics[width=.6\textwidth]{Fig10.eps}
986: \caption{Ishihara and Kaneko}
987: \label{fig:uvApend1}
988: \end{figure}
989:
990:
991: \end{document}
992:
993: