1: %
2: % MANUSCRIPT STARTS HERE
3: %
4: % Revision July 7, 2004
5: %%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%%
6: %
7: %\documentstyle[aps,prl]{revtex}
8: %\documentstyle[twocolumn,prl,aps]{revtex}
9: %\documentclass[aps,preprint,eqsecnum]{revtex}
10: \documentstyle[12pt]{article}
11: %\documentstyle[12pt,doublespace]{article}
12: % if you want enumerated equations according to sections insert: ``eqsecnum''
13:
14: \begin{document}
15: %\draft
16:
17: % below commands are for enumerating eqs according to sections
18: %\newcounter{eq}[section]
19: %\newcommand{\set}{\stepcounter{eq}
20: %\renewcommand{\theequation}{\mbox{\arabic{section}.\arabic{eq}}}}
21:
22:
23:
24:
25: \hsize=6.15in
26: \vsize=8.2in
27: \hoffset=-0.42in
28: \voffset=-0.3435in
29:
30: \normalbaselineskip=24pt\normalbaselines
31:
32:
33: \begin{center}
34: {\large \bf Model of the early development of thalamo-cortical
35: connections and area patterning via signaling molecules }
36: \end{center}
37:
38:
39: \vspace{0.15cm}
40:
41: \begin{center}
42: {Jan Karbowski$^{1,*}$ and G.B. Ermentrout$^{2}$}
43: \end{center}
44:
45: \vspace{0.05cm}
46:
47: \begin{center}
48: $^{1}$ {\it Sloan-Swartz Center for Theoretical Neurobiology,
49: Division of Biology 216-76, \\
50: California Institute of Technology,
51: Pasadena, CA 91125, USA } \\
52: $^{2}$ {\it Department of Mathematics,
53: University of Pittsburgh,
54: Pittsburgh, PA 15260, USA }
55: \end{center}
56:
57: %\date{\today}
58:
59:
60: \vspace{0.1cm}
61:
62:
63: %\widetext
64: \begin{abstract}
65: The mammalian cortex is divided into architectonic and functionally
66: distinct areas. There is growing experimental evidence
67: that their emergence and development is controlled by both
68: epigenetic and genetic factors. The latter were recently implicated
69: as dominating the early cortical area specification.
70: In this paper, we present a theoretical model that explicitly
71: considers the genetic factors and that is able to
72: explain several sets of experiments on cortical area regulation
73: involving transcription factors Emx2 and Pax6,
74: and fibroblast growth factor FGF8.
75: The model consists of the dynamics of thalamo-cortical connections
76: modulated by signaling molecules that are regulated genetically, and
77: by axonal competition for neocortical space.
78: The model can make predictions and provides a basic mathematical
79: framework for the early development of the thalamo-cortical connections
80: and area patterning that can be further refined as more experimental
81: facts become known.
82: \end{abstract}
83:
84: %\pacs{PACS Nos. 87.18.Sn, 87.19.La, 87.80.Xa, 89.40.+k, 87.23.Ge}
85: %\narrowtext
86: %\maketitle
87:
88: %\begin{narrowtext}
89:
90: %\maketitle
91:
92:
93: \noindent {\bf Keywords}: cerebral cortex, development, thalamo-cortical
94: connections, cortical area patterning, signaling molecules.
95:
96: \noindent {\bf Running Title}: cortical patterning via signaling
97: molecules
98:
99: \vspace{0.2cm}
100:
101: \noindent $^{*}$ Corresponding author at:
102:
103: \noindent Email: jkarb@cns.caltech.edu
104:
105: \noindent Phone: (626)-395-5840
106:
107: \noindent Fax: (626)-795-2397
108:
109: \vspace{0.5cm}
110:
111:
112:
113:
114:
115:
116: \newpage
117:
118: \section{Introduction}
119:
120:
121: Neocortex is organized into many functional subdivisions called areas
122: that have sharp boundaries. The areas can be identified anatomically
123: by investigating their distinct cytoarchitectonic properties and
124: unique connectivity patterns. One interesting feature is that the
125: neocortical map composed of these areas is highly conserved within the
126: same species, and has common properties across different species
127: with different brain sizes (Nauta and Feirtag, 1986; Hofman, 1989;
128: Finley, 1995; Northcutt and Kaas, 1995; Krubitzer, 1995;
129: Karbowski, 2003).
130:
131: One of the main questions in the development of the mammalian cortex is
132: what factors control the specification and differentiation of
133: cortical areas. In the past, there were two opposing views.
134: One proposition was that areas are specified by intrinsic genetic
135: factors (Rakic, 1988) - the so-called protomap model.
136: Another proposition was that areas are specified by extrinsic
137: influence, i.e. by thalamo-cortical inputs (O'Leary, 1989).
138: In recent years, however, the consensus has been growing that both
139: of these factors, i.e. genetic and epigenetic contribute to the
140: cortical area patterning (Krubitzer and Huffman, 2000;
141: O'Leary and Nakagawa, 2002).
142: The genetic effects are thought to dominate the early stages of
143: the development, while the epigenetic effects influence the later
144: stages. This view
145: has been stimulated by recent experiments demonstrating a direct
146: genetic involvement in cortical arealization (Bishop et al, 2000;
147: Mallamaci et al, 2000; Fukuchi-Shimogori and Grove, 2001, 2003;
148: Bishop et al, 2002; Muzio et al, 2002; Garel et al, 2003).
149: In particular, it has been found that the genetic transcription
150: factors Emx2 and Pax6 control area specification even before
151: thalamo-cortical input arrives (Bishop et al 2000; Mallamaci et al
152: 2000; Muzio et al, 2002; Bishop et al, 2002). Both of these factors
153: are expressed
154: in a graded and complementary manner along antero-posterior (A/P)
155: axis in the neocortical ventricular zone.
156: In mice with Emx2 mutation, anterior areas expand
157: and posterior areas shrink, in Pax6 mutants the opposite is observed.
158: Another set of experiments has found that cortical maps can be
159: disrupted by modifying fibroblast growth factor FGF8
160: (Fukuchi-Shimogori and Grove, 2001, 2003; Garel et al, 2003),
161: which is also involved in embryonic patterning.
162: In wild type mouse neocortex, the FGF8 source is located in the
163: anterior pole giving rise to an expression concentration decaying
164: towards the posterior end. Under such regular conditions, the barrel field
165: (S1 area) is positioned in the center along A/P axis. However, increase
166: in the expression of FGF8 at the anterior pole, displaces the barrel
167: field more posteriorly. In contrast, blocking the FGF8 activity with
168: a soluble FGF8 receptor moves the field in the opposite direction.
169: Moreover, introduction of an extra source of FGF8 at the posterior
170: pole creates an additional barrel field that partly duplicates the
171: original field. Taken together, all these results suggest that
172: genetic perturbations can have a profound effect on the development
173: of thalamo-cortical connections and area patterning. Additionally,
174: these results are consistent with a hypothesis that the above signaling
175: molecules provide positional information for neuroepithelial cells
176: (Wolpert, 1969, 1996; Grove and Fukuchi-Shimogori, 2003).
177:
178:
179: Area-specific TC projections are probably controlled
180: by axon guidance molecules, similar to what happens in the retino-tectal
181: system (O'Leary et al, 1999). Although several molecules
182: have been found that are able to guide TC axons, there are some
183: experimental indications that the plausible candidates
184: for the late phase of accurate TC targeting are ephrins and their
185: receptors that are expressed both in the neocortex and in the thalamus
186: (Vanderhaeghen et al, 2000; Mackarehtschian et al, 1999; Mann et al,
187: 2002; Uziel et al, 2002; Takemoto, 2002). It is well established that
188: these molecules
189: direct axons to appropriate locations in many systems, in particular,
190: in the retino-tectal system (Goodhill and Richards, 1999),
191: which, like the neocortex, preserves
192: the topography of projections.
193:
194: The relationship between the axon guidance molecules and Emx2, Pax6,
195: and FGF8 is not known at present. It is likely, however, that the
196: latter control the expression level and gradients of the guiding
197: molecules. Thus, if the transcription factors and FGF8 concentrations
198: are modulated, then it should affect the concentration pattern of
199: the axon guidance molecules. This hypothesis is consistent with the
200: experiments on cortical areas shifting described above.
201:
202:
203: The purpose of this article is to provide a theoretical model of
204: the early TC projections and cortical area patterning.
205: The key assumption we make is that the signaling molecules Emx2,
206: Pax6, and FGF8 control the axon guidance molecules, which in turn
207: control TC pathfinding. We assume that there are three main types
208: of the axon guidance molecules, which we call A, B and C, that
209: are expressed
210: in the neocortical ventricular zone in a graded and complementary
211: manner along A/P axis. In the model, these molecules repel and
212: attract different TC axons, which branch diffusively along the cortex,
213: with specific forces, different for different classes of axons.
214: Additionally, axons interact among themselves by competing for
215: neocortical space. As a consequence of these
216: interactions, a pattern of TC connectivity emerges that divides the
217: neocortex into regions with sharp boundaries defined by distinct TC
218: axon types. In this framework, area shifting experiments described
219: above can be understood as a result of shifts in the patterns of
220: expression of the signaling molecules.
221:
222:
223:
224: \section{Model}
225:
226: We assume that the early TC connectivity and cortical regionalization
227: along A/P axis is established by chemo-interaction between axons
228: originating from the thalamus and axon guidance molecules located on the
229: cortex (Sperry, 1963). In this paper, we assume that
230: there are $N=5$ main axon types. Under normal conditions,
231: type $i=1$ corresponds to the axons coming from ventrolateral thalamus and
232: terminating in primary motor area M1, type $i=3$ corresponds to
233: the axons originating from thalamic ventrobasal complex and
234: terminating in S1, type $i=5$ relates to the axons connecting
235: thalamic LGN with the primary visual area V1.
236: The remaining axons of type 2 and 4 terminate between areas M1 and
237: S1 (type $i=2$), and between S1 and V1 (type $i=4$), in order to ensure
238: a topographic TC connectivity.
239:
240: In the model, it is assumed that there are three types of the axon
241: guidance molecules A, B, and C. In our choice of the number of guiding
242: molecules, we were motivated by experimental indications that
243: the best candidates
244: for the late phase of precise TC pathfinding are ephrins and their
245: receptors (O'Leary and Nakagawa, 2002; Lopez-Bendito and Molnar, 2003).
246: Up to now, 3 types of ephrins have been implicated as important:
247: ephrin-A5 expressed at high levels in the central part of the cortex,
248: ephrin-A4 expressed in the cortical intermediate zone with a gradient
249: decaying from the anterior towards the posterior end, and ephrin-B3
250: with a complementary gradient to ephrin-A4 (Lopez-Bendito and Molnar,
251: 2003). In the model, the molecule A may mimic the action of ephrin-A4,
252: the molecule B may act analogously to
253: ephrin-A5, and the molecule C may mimic ephrin-B3. The molecules A, B, and
254: C have a mixed effect on TC axons, i.e. they both repel and attract them
255: and this is encoded in the interaction matrix $\gamma$. Experimentally,
256: ephrin-As and -Bs ligands have been shown to exhibit some selective
257: behavior, too. Ephrin-B expressing cells tend to attract axonal growth
258: cones with EphB receptors (McLaughlin et al, 2003), and repulse axons
259: with EphA4 receptors (Takemoto et al, 2002). The prevailing action
260: of ephrin-As expressing
261: cells is repulsion, which is well documented with axons expressing EphA
262: (e.g., Mann et al, 2002; McLaughlin et al, 2003).
263: However, when axonal cones express both EphA and ephrin-A as receptors,
264: then cells with ephrin-As can attract them (Knoll and Dreschner, 2002).
265:
266: In addition to the chemo-affinity, we assume that axons branch randomly
267: along A/P axis. Axonal branching has been found in the retino-tectal
268: (Yates et al, 2001) and retino-collicular (Simon and O'Leary, 1992)
269: systems, and there are some indications (Mann et al, 2002) that similar
270: effects are present in the
271: establishment of TC connectivity. The basics of the process of random
272: growth and decay of branches along axonal length can be mathematically
273: described by some stochastic component acting on a local branching
274: density (see below). The combined effect of the chemo-interactions
275: and stochasticity on axonal terminals is mathematically equivalent to
276: their biased random growth on the cortical surface. The bias in growth
277: is caused by the gradients of the guidance molecules. For attractive
278: interactions, the larger the gradient the larger the growth rate. For
279: repulsive interactions, the larger the gradient the larger the decay rate.
280:
281: Another crucial assumption is that there exists
282: some sort of repulsive interaction
283: between TC axons of different types that leads to their competition for
284: neocortical space.
285: This constraint implies that the density of TC projections is limited,
286: because not all axon types are allowed to terminate
287: simultaneously at a single point in the cortex. The idea of axon-axon
288: competition was used
289: in the past in the context of retino-tectal mapping (Prestige and
290: Willshaw, 1975; Fraser and Hunt, 1980) and it is consistent with
291: recent experiments in retino-collicular system (Feldheim et al, 2000).
292:
293: In the mathematical model, the development of TC connections ($c_{i}$)
294: between axons of type $i$ (with the branching density $a_{i}$) and cortical
295: neurons $n$ is represented by a simple kinetic reaction scheme:
296: $n + a_{i}\stackrel{\longrightarrow}{\leftarrow}^{\beta_{i}}_{\alpha_{i}}
297: c_{i}$ (e.g. Murray, 1993). Connections are destroyed
298: with the rate $\alpha_{i}$, and created with a rate proportional
299: to the product of the fraction of available neurons $n(x)$ and some
300: power of branching density, i.e. $\beta_{i}n(x)[a_{i}(x)]^{k}$,
301: where $\beta_{i}$ is some positive constant. When the exponent $k > 1$, then
302: TC connections are created with a higher rate in regions with high
303: concentration of axonal branches, so that there is some cooperativity.
304: (In our numerical simulations we take $k=3$.)
305:
306: The full set of differential equations describing
307: the above processes is given by:
308:
309:
310: \begin{equation}
311: \frac{\partial{c_{i}(x)}}{\partial{t}} = -\alpha_{i} c_{i}(x)
312: + \beta_{i} n(x)[a_{i}(x)]^{k},
313: \end{equation}
314: \begin{equation}
315: \frac{\partial{a_{i}(x)}}{\partial{t}} = \frac{\partial{J_{i}(x)}}
316: {\partial{x}} + \alpha_{i} c_{i}(x) - \beta_{i} n(x)[a_{i}(x)]^{k},
317: \end{equation}
318: \begin{equation}
319: n(x,t) + \sum_{i=1}^{N} c_{i}(x,t) = 1,
320: \end{equation}\\
321: with the flux current
322:
323: \begin{equation}
324: J_{i}(x)= D\frac{\partial{a_{i}}}{\partial{x}}
325: - a_{i}(x)\left( \gamma_{Ai}\frac{\partial{\rho_{A}}}{\partial{x}}
326: + \gamma_{Bi}\frac{\partial{\rho_{B}}}{\partial{x}}
327: + \gamma_{Ci}\frac{\partial{\rho_{C}}}{\partial{x}} \right),
328: \end{equation}\\
329: where $c_{i}(x)$ is a fraction of TC connections of type $i$ at position
330: $x$ on the developing cortex, $n(x)$ is a fraction of cortical neurons at
331: point $x$ available for a TC connection (i.e. not already connected by
332: other TC axons), $a_{i}(x)$ is the branching density of axons of type $i$
333: at point $x$ on the cortex that originate from the thalamus, and $N$
334: is the number of axon types. Our model is a population model, which deals
335: with densities and fractions, and does not take directly into account
336: the fine structure of axonal branching on a single axon level, such as
337: complex arborization patterns; instead we consider population branching
338: density. (For detailed modeling of branching structures and their influence
339: on the retino-topic map, see Yates et al, 2004).
340: Eq. (3) is a mathematical consequence of axonal competition
341: for neocortical space, and it provides
342: a conservation law for the fractions of available (unconnected) neurons
343: and all the connection types for every point on the cortex and for all times.
344: The flux $J_{i}(x)$ is associated with the growth and decay of axonal
345: branches of type $i$ at point $x$. It is composed of the following
346: contributions: the diffusive axonal branching that is proportional
347: to the diffusion constant $D$, and chemo-interaction between
348: axonal branches and the axon guidance molecules that have concentrations
349: $\rho_{A}(x)$, $\rho_{B}(x)$, and $\rho_{C}(x)$ (Sperry, 1963).
350: The nature of this interaction is additive, that is, the guiding molecules
351: do not interact with themselves, and it can be either repulsive or attractive,
352: which is controlled by the sign and strength of the parameters
353: $\gamma_{Ai}$, $\gamma_{Bi}$, and $\gamma_{Ci}$.
354:
355:
356: Our model focuses on how the anterior/posterior boundaries of cortical
357: areas are formed. Thus, from a theoretical point of view, this
358: biological problem has a one-dimensional character and
359: the spatial variable $x$ measures distances along
360: the A/P axis. In our model, the anterior pole corresponds to $x\approx 0$,
361: and the posterior pole to $x\approx L$, where $L$ is the linear
362: size of the cortex in that direction.
363:
364:
365: We adopt sealed-end boundary conditions for the flux current $J_{i}(x)$,
366: i.e. $J_{i}(L)= J_{i}(0)= 0$, since in our system axonal branches cannot
367: grow outside the boundaries of the cortex. This choice implies
368: that the total number of axonal branches
369: and connections of any given type in the system is constant, i.e.
370: $\int_{0}^{L} dx [a_{i}(x)+c_{i}(x)]$ is time independent (this follows
371: from adding eqs. (1) and (2)). Thus initially, there are many branches
372: but no connections. However, as time progresses, the total number of
373: branches decreases at the expense of the formed connections. In a real
374: biological system, however, this constraint may be satisfied only
375: approximately.
376:
377:
378: It is assumed that the axon guidance molecules located on the cortex are
379: regulated by the transcription factors Emx2 and Pax6, and the fibroblast
380: growth factor FGF8. Recent experiments (Fukuchi and Grove 2003; Garel
381: et al, 2003) show that
382: (i) FGF8 and Emx2 mutually inhibit each other. The Fukuchi and Grove
383: (2003) results also suggest that (ii) FGF8 directly
384: controls the location of cortical maps, and additionally that
385: (iii) Emx2 indirectly controls the positioning of cortical maps by
386: acting upstream of FGF8 and regulating it.
387: Another set of experiments (Muzio et al, 2002) shows that (iv) the two
388: transcription factors, Emx2 and Pax6, mutually repress each other.
389: Based on these facts, one can construct a minimal signaling pathway
390: (Fig. 1) that is responsible for regulating expression levels of the
391: guiding molecules A, B, and C. In this model, both Emx2 and Pax6 act
392: upstream of FGF8, but only Emx2 regulates it directly.
393: FGF8 and Pax6 have similar gradients, i.e. their concentrations
394: decay from the anterior towards the posterior end, while Emx2
395: has an opposite gradient to them.
396: The output signal $f$ from the pathway can be viewed as a renormalized
397: (by interactions with other signaling molecules) concentration of FGF8.
398: We assume that either this signal or some other $f$-activated signal
399: serves as a morphogen
400: signal (Grove and Fukuchi-Shimogori, 2003) that provides a positional
401: information for cortical cells, similar to what happens in other
402: developing systems (Wolpert, 1969, 1996).
403: In response to the signal $f$, cells express different levels of the
404: guiding molecules, which in turn control TC axons.
405:
406:
407: The diagram in Fig. 1 can be described mathematically in the
408: following way.
409: The transcription factors Emx2 and Pax6 both repress each other
410: with strengths $v_{1}$ and $v_{2}$, respectively ($v_{1}, v_{2} > 0$).
411: Emx2 negatively controls FGF8 with a strength $w_{1}$, and FGF8
412: in turn inhibits Emx2 with a strength $w_{2}$ ($w_{1}, w_{2} > 0$).
413: From all of these relationships, it follows that FGF8 ($f$)
414: is regulated negatively by Emx2 and positively by Pax6.
415: Mathematical equations describing the dynamics of the expression of Emx2,
416: Pax6, and FGF8 are given by
417:
418:
419: \begin{equation}
420: \tau_{s}\frac{ds(x)}{dt}= - s(x) +
421: \frac{\eta_{emx}(x)}{1+w_{2}f(x)+v_{2}r(x)}
422: \end{equation}
423: \begin{equation}
424: \tau_{r}\frac{dr(x)}{dt}= - r(x) +
425: \frac{\eta_{pax}(x)}{1+v_{1}s(x)}
426: \end{equation}
427: \begin{equation}
428: \tau_{f}\frac{df(x)}{dt}= - f(x) +
429: \frac{\eta_{fgf}(x)}{1+w_{1}s(x)},
430: \end{equation}\\
431: where $s(x)$, $r(x)$, and $f(x)$ are expression levels (renormalized
432: concentrations) of Emx2, Pax6, and FGF8, respectively, in the presence
433: of interactions between these molecules.
434: The parameters $\tau_{s}$, $\tau_{r}$, and $\tau_{f}$
435: are signaling time constants, and finally $\eta_{emx}(x)$,
436: $\eta_{fgf}(x)$, $\eta_{pax}(x)$ are uncoupled (i.e. without interactions)
437: concentrations of Emx2, FGF8, and Pax6.
438: We assume the following
439: forms of the uncoupled concentrations for Emx2, Pax6, and FGF8:
440: $\eta_{emx}(x)= A_{emx} \exp[-(x-L)^{2}/\zeta_{emx}^{2}]$,
441: $\eta_{pax}(x)= A_{pax} \exp[-x^{2}/\zeta_{pax}^{2}]$, and
442: $\eta_{fgf}(x)= A_{fgf} \exp[-x^{2}/\zeta_{fgf}^{2}]$, where
443: $A_{emx}$, $A_{pax}$, $A_{fgf}$ are amplitudes of expression, and
444: $\zeta_{emx}$, $\zeta_{pax}$, $\zeta_{fgf}$
445: are constants characterizing the ranges of expression with values in
446: the interval $(0,L)$.
447: When two sources of FGF8 are present, one in the anterior and
448: second in the posterior end, then the uncoupled concentration
449: of FGF8 takes the form:
450: $\eta_{fgf}(x)= A_{fgf} \exp[-x^{2}/\zeta_{fgf}^{2}] +
451: A_{fgf}' \exp[-(x-L)^{2}/\zeta_{fgf}'^{2}]$, where
452: $A_{fgf}'$ and $\zeta_{fgf}'$ are the amplitude and the range of
453: the posterior source, respectively.
454: These uncoupled
455: concentrations act as signals produced at boxes denoted
456: by Emx, FGF, and Pax in Fig. 1. We assume that the time constants
457: $\tau_{s}$, $\tau_{r}$, and $\tau_{f}$ are much smaller than
458: characteristic time
459: constants associated with the processes involving TC projections.
460: That is, at the time when TC axons arrive to the cortical surface,
461: all these molecules have already reached their steady-state, which is
462: consistent with experimental data (Cohen-Tannoudji et al, 1994;
463: Miyashita-Lin et al, 1999; Nakagawa et al, 1999).
464: For this reason, we are interested in a steady-state
465: solution for $f(x)$.
466: Its plot is displayed in Fig. 2a.
467: Under normal conditions, its spatial profile displays a monotonic decay
468: from the anterior towards the posterior end (Fig. 2a), i.e. it has
469: a similar gradient to the uncoupled FGF8.
470:
471:
472: We assume
473: that the stationary distributions of the guiding molecules are
474: as follows: the molecule A is mostly expressed in those regions
475: of the cortex for which the signal $f(x)$ is high (anterior end),
476: B is expressed mostly in regions where $f(x)$ is moderate (center),
477: and C has high concentration in locations where $f(x)$ is weak
478: (posterior end). These relationships can be represented mathematically
479: in the following way:
480:
481:
482: \begin{equation}
483: \rho_{A}(x)= G_{A}\left(f(x)-\theta_{1}\right),
484: \end{equation}
485: \begin{equation}
486: \rho_{B}(x)= G_{B}\left(\theta_{2}-f(x)\right)
487: G_{B}\left(f(x)-\theta_{3}\right),
488: \end{equation}
489: \begin{equation}
490: \rho_{C}(x)= G_{C}\left(\theta_{4}-f(x)\right),
491: \end{equation}\\
492: where $\rho_{A}$, $\rho_{B}$, and
493: $\rho_{A}$ are stationary concentrations of the guiding
494: molecules A, B, and C. The thresholds
495: $\theta_{i}$ $(i=1,...,4)$ are some positive constants such
496: that $\theta_{4} < \theta_{3} < \theta_{2} < \theta_{1} < f(0)$, and
497: $G_{j}(y)= (\kappa_{j}/2)[1 + \tanh(y/\sigma_{j})]$, where
498: $\kappa_{j}$ and $\sigma_{j}$ (for $j=A, B, C$) are some positive
499: constants controlling the amplitude and the slope of the
500: concentration $\rho_{j}(x)$. The parameters $\sigma_{j}$ are
501: chosen such as to ensure that the expressed levels of the guiding
502: molecules A, B, and C have graded concentrations (Fig. 2b).
503:
504:
505:
506:
507: \section{Results}
508:
509: All the results reported in this section were obtained by numerically solving
510: eqs. (1)-(10). We used a second-order Runge-Kutta method
511: for the ordinary differential equations and a Crank-Nicholson method
512: (second order accurate in time and space) for the partial
513: differential equations. The values of the parameters used are given
514: in the legend of Fig. 2.
515:
516:
517: \noindent {\bf Normal area positioning.}
518:
519: Under normal conditions (wild type) cortical areas in the mouse neocortex
520: are located such that motor area M1 occupies anterior part, sensory area
521: S1 occupies the central part, and visual area V1 is positioned in the
522: posterior end. Areas are defined as a spatial
523: pattern of TC connectivity fraction $c_{i}(x)$ for $i=1,...,5$.
524: A high value of $c_{1}(x)$ corresponds to the M1 field, a high value of
525: $c_{3}(x)$ corresponds to the S1 area, and a high value of $c_{5}(x)$
526: corresponds to V1. For the remaining two connection types $c_{2}(x)$
527: and $c_{4}(x)$, their high values correspond to areas between M1
528: and S1, and between S1 and V1, respectively. Both, the connectivity
529: fraction $c_{i}(x,t)$ and the axonal branching density $a_{i}(x,t)$ evolve
530: in time to a steady-state according to eqs. (1)-(4) (see Figs. 3 and 4).
531: Initial conditions are chosen such that
532: $a_{i}(x,t=0)$ is uniformly (although with some noise) distributed
533: in space, and $c_{i}(x,t=0)= 0$, for $i=1,...,5$. This choice is motivated
534: by experimental data (Simon and O'Leary, 1992; Yates et al, 2001),
535: suggesting that the early TC connectivity in mammals is established in
536: three phases. The first phase is axonal overshooting along A/P axis
537: beneath the cortical plate. The second phase is composed of axonal
538: branching along their length, and the final phase provides stabilization
539: of topographically correct axonal collaterals and elimination of distant
540: branches. Our modeling starts from the second phase. As time progresses
541: both of the above distributions evolve into spatially heterogeneous
542: state with a close relationship between patterns of $a_{i}(x)$
543: and $c_{i}(x)$ (compare Figs. 3c and 4c). The reason for this relationship
544: can be disclosed if we use eqs. (1) and (3). Then one can
545: derive a formula relating the stationary distribution of the connections
546: $\overline{c_{i}}(x)$ with the stationary branching densities
547: $\overline{a_{i}}(x)$:
548:
549: \begin{equation}
550: \overline{c_{i}}(x)= \frac{\beta_{i}[\overline{a_{i}}(x)]^{k}}{\alpha_{i}
551: + \sum_{j=1}^{N} \beta_{j}[\overline{a_{j}}(x)]^{k} }.
552: \end{equation}\\
553: From this equation, it follows that the area of type $i$ emerges in regions
554: with high concentration of the branches of type $i$. However, one should
555: note that not all diffusive branches form connections.
556: This can be seen in Fig. 3c, where branches of type 2 and 4 have rather
557: broad distributions, which are absent in the connectivity pattern in
558: Fig. 4c.
559:
560: The locations of cortical regions are shaped
561: by different forces coming from the axon guidance molecules A, B and C.
562: These forces activate the growth of axonal branches in some locations
563: and inhibit their growth in other locations.
564: To have, at a steady-state, a high density of axons of type $i=1$ in the
565: anterior end, it is assumed that they are moderately/strongly attracted
566: by the molecules A,
567: repulsed strongly by the molecules C, and repulsed moderately by B.
568: Similarly, to have high density of
569: axons of type $i=5$ in the posterior end, it is assumed that they are
570: moderately/strongly attracted by C, strongly repulsed by the molecules A,
571: and moderately repulsed by B.
572: Axons of type $i=3$ are forced to aggregate in the center by assuming
573: that they are attracted by B, and repulsed, either strongly or moderately,
574: with approximately the same strength from the guiding molecules A and C.
575: Elements of the interaction matrix $\gamma_{\alpha i}$ for axons of type
576: $i=2$ and $i=4$ are all negative (repulsion),
577: and are chosen such that these axons
578: terminate in between axons 1, 3 and 3, 5 in the wild type case.
579: Although the concentrations of A, B, and C are graded (Fig. 2b),
580: the spatial pattern of the fractions of TC connections $c_{i}(x)$ is almost
581: exclusive with sharp borders (Fig. 4c), which leads to the emergence of
582: cortical areas (Fig. 4d). The exclusiveness comes from the
583: axonal competition constraint we imposed (see eq. (3)).
584: The width of the areal border
585: is directly proportional to the diffusion constant $D$, and inversely
586: proportional to the product of the coefficient $k$ and
587: the magnitude of the interaction between guiding molecules and axons.
588: Purely repulsive
589: interactions increase the border width in comparison to cases with mixed
590: repulsive/attractive interactions (see Fig. 9 in the discussion).
591:
592:
593:
594:
595: \noindent {\bf Area shifting.}
596:
597: In the area shifting experiments expression levels of Emx2, Pax6, and
598: FGF8 were affected
599: (Bishop et al, 2000; Mallamaci et al, 2000;
600: Fukuchi-Shimogori and Grove, 2001; Bishop et al, 2002; Muzio et al
601: 2002; Garel et al 2003, Fukuchi-Shimogori and Grove, 2003).
602: We model this effect by reducing or amplifying
603: the appropriate concentration amplitude ($A_{emx}, A_{pax}, A_{fgf}$)
604: and the range of expression
605: of one of these molecules. In Fig. 5, we present posteriorly shifting
606: areas as we decrease the amplitude $A_{emx}$ to zero. This case corresponds
607: to loss-of-function Emx2 mutants for which such shifting has been
608: reported (Bishop et al, 2000; Mallamaci et al, 2000; Muzio et al, 2002;
609: Bishop et al, 2002).
610: The interpretation of this effect in our model
611: is as follows. Reducing the concentration of Emx2 effectively increases
612: the output signal $f$ (see, Fig. 1), since then the Emx2 $\rightarrow$
613: FGF8 inhibition decreases.
614: This leads effectively to shifting posteriorly the expression profile
615: of $f$ (Fig. 5a), and consequently shifting posteriorly the expression pattern
616: of the guiding molecules (Fig. 5b).
617: That new pattern biases axonal branching growth to the right to new
618: locations, where balanced forces act on them (Fig. 5c).
619: Again, there is a close relationship between distributions of axons and
620: TC connectivity patterns (Figs. 5c,d). All areas are shifted posteriorly
621: (Fig. 5e). In particular, a clear shift in the position of
622: the S1 area can be seen by comparing figures 4c and 5d.
623:
624: Fig. 6 displays anteriorly shifting areas as we decrease the amplitude
625: $A_{pax}$ to zero. This case corresponds to loss-of-function Pax6 mutants
626: for which shifting in this direction has been experimentally
627: reported (Bishop et al, 2000; Muzio et al, 2002; Bishop et al, 2002).
628: By the same token, reduction in Pax6 concentration decreases
629: effectively the output signal $f$ (Fig. 6a), which in turn shifts
630: anteriorly the concentrations of the molecules A, B and C (Fig. 6b).
631: This rearrangement pushes TC axons more into anterior end to new
632: equilibrium positions (Fig. 6c) that leads to areas shift in this
633: direction (Figs. 6d,e).
634:
635: The action of FGF8 is only slightly different. When we increase
636: the expression level of FGF8 by amplifying its amplitude $A_{fgf}$
637: and its range, we also increase the signal $f$ that leads to posterior
638: shift in expression levels of the molecules A, B and C (Fig. 7a).
639: As a consequence of this, TC connectivity (Fig. 7b) and cortical
640: areas (Fig. 7c) shift posteriorly, similarly to what happens in the case
641: of reduced Emx2. In contrast, if we decrease the expression of
642: FGF8 by decreasing $A_{fgf}$ and the range of its expression,
643: then we obtain pattern shifting in the opposite direction by the same
644: mechanism as above (Figs. 7d,e,f). Both types of shifting have been
645: experimentally observed recently (Fukuchi-Shimogori and Grove, 2001, 2003;
646: Garel et al, 2003).
647:
648:
649:
650:
651: \noindent {\bf Generation of two separate S1 fields.}
652:
653: In experiments with an extra source of FGF8 at the posterior end,
654: in some cases, a second entirely separate S1 barrel field can be
655: generated (Fukuchi-Shimogori and Grove, 2001).
656: We can mimic this effect of ``mirror symmetry'' in our model (Figs. 8a,b).
657: The key observation is that by modifying the expression level of FGF8,
658: one changes the spatial shape of the signal $f(x)$. Without the
659: second ectopic source of FGF8 at the posterior end, this function
660: decays monotonically from the anterior to the posterior end. However,
661: when this second source is present, the function $f(x)$ can
662: have a minimum at central part under some conditions (Fig. 8c).
663: This happens if the diffusion range of the ectopic FGF8 is not too
664: large and if its amplitude is sufficiently strong to overcome
665: inhibition of Emx2 in the posterior end. From eq. (9), then it follows
666: that $\rho_{B}(x)$ is expressed at high levels at two separate
667: central locations which partly overlap (Fig. 8d). Thus, instead of
668: growing to one central location, axons of type $i=3$ grow
669: to two slightly separated locations that correspond to
670: two partly symmetric S1 areas (Figs. 8a,b). However, we are unable, within
671: our approach, to distinguish between two possibilities: one, that the
672: same group of axons target two S1 fields by branching in two separate
673: locations, and second, that axons subdivide into two groups, each
674: targeting only one S1 field. These two possibilities await experimental
675: teasing apart.
676:
677: Surprisingly, this is not the only modification that the second
678: ectopic source of FGF8 brings. It turns out that the whole pattern
679: of both TC connectivity (Fig. 8a) and cortical fields (Fig. 8b)
680: has a mirror symmetry.
681: Area V1 ($c_{5}$) is positioned in the gap between the two S1
682: ($c_{3}$) areas, and area M1 ($c_{1}$) has two branches:
683: one in the regular anterior end, and another in the posterior
684: end that under normal conditions would be occupied by V1. Also,
685: area $c_{2}$ has two branches, and area $c_{4}$ overlaps S1 and V1.
686: These results can be understood by invoking again Figs. 8c,d and
687: eqs. (8)-(10). The guiding molecule C that acts as an attractant for
688: axons terminating in V1 area is expressed now only slightly in the central
689: part due to the modified shape of the signal $f(x)$ (only for
690: $x$ located in the center, $f(x)$ is relatively close to the threshold
691: $\theta_{4}$). Axonal branches of type $i=5$ gathered in the center due
692: to attraction to C, are additionally repulsed by two regions of high
693: concentration of B (Fig. 8d), and that effectively pushes them to the gap
694: between the two S1 fields. Similar arguments hold for the M1 ($c_{1}$)
695: fields. The molecule A that attracts axons of type 1 terminating
696: in M1 is expressed now not only in the anterior end but also partially
697: in the posterior end (Fig. 8d). This leads to the type $i=1$ axons
698: terminating in each of these locations (Fig. 8a).
699: Similarly for the area $c_{2}$.
700: Axonal branches of this type are the most repulsed by the molecules B and C,
701: and only weakly repulsed by A. As a result, they grow in the regions
702: where the molecule A has a moderate expression, that is, also in two
703: symmetric locations.
704:
705: From the above it is apparent that the ectopic FGF8 can profoundly
706: affect the architecture of the neocortex by creating and destroying
707: areas at different locations. It would be interesting to verify
708: these predictions experimentally.
709:
710:
711:
712:
713: \section{Discussion}
714:
715: In this paper, a mathematical model of the early development
716: of TC connections and cortical area patterning is presented.
717: Cortical patterning is achieved in the model by allowing TC axonal
718: branches to undergo a combination of biased random growth with their mutual
719: competition for neocortical space.
720: The model captures the essential components, which have
721: been experimentally implicated as the important genetic factors
722: (Bishop et al, 2000; Mallamaci et al, 2000;
723: Fukuchi-Shimogori and Grove, 2001, 2003; Bishop et al, 2002; Muzio et al
724: 2002; Garel et al 2003).
725: Its strength is in its ability to reproduce several recent
726: experiments within a single theoretical framework,
727: and also in its potential to predict the outcomes of some experiments.
728: In this regard, one of the tests of this model can be experimental
729: verification of the theoretical findings related to duplication of S1
730: field. In particular, to test (i) if cells in the gap between two
731: symmetric S1 fields really acquire the properties typical for neurons in
732: the visual cortex, and (ii) if cells in the
733: posterior end acquire properties typical for motor neurons.
734:
735:
736: Other tests of the model that are perhaps more feasible experimentally
737: could include the influence of the transcription factors Emx2 and Pax6,
738: and FGF8 on the guidance molecules. If the guiding molecules A, B, and C
739: really correspond to ephrins A4, A5, and B3, respectively, then by
740: manipulating Emx2, Pax6, and FGF8 one should be able to observe changes
741: in ephrins concentrations. Specifically, in shifting experiments,
742: we predict that for Emx2 mutants the concentrations of all three ephrins
743: should shift posteriorly (Fig. 5b), and for Pax6 mutants they should
744: shift anteriorly (Fig. 6b) as compared to the wild-type case. Partial
745: confirmation of this prediction is provided by the data of Bishop et al
746: (2002), who found that ephrin-A5 concentration shifts posteriorly in Emx2
747: mutants and it shifts anteriorly in Pax6 mutants. In the case of FGF8,
748: we predict that when FGF8 is overexpressed in the anterior pole,
749: then all ephrins should shift posteriorly (Fig. 7a), and when FGF8 is
750: underexpressed, they should shift in the opposite direction (Fig. 7d).
751: In the case of two sources of FGF8 present at the two poles, a crucial
752: experiment would be to verify Fig. 8d. In particular, the model predicts
753: that the molecule A (ephrin-A4) should be expressed not only in the
754: anterior end (regular location) but also in the posterior end, and
755: additionally that the molecule C (ephrin-B3) should be minimally expressed
756: in the center of the cortex instead of the posterior end. Such experiment
757: would prove or disprove the correctness of our assumption on the relationship
758: between FGF8 and the axon guidance molecules.
759:
760:
761: Our model is robust to changes in the interaction parameters between
762: transcription factors ($v_{1}, v_{2}, w_{1}, w_{2}$). However, these
763: parameters cannot be too small, since then the effects of area shifting
764: are small too. In general, the larger these parameters the more pronounced
765: the shiftings.
766:
767: The most crucial interactions in the model are the chemo-interactions
768: between the axon guidance molecules and TC axons, i.e. the matrix $\gamma$.
769: Together with axonal competition (eq. (3)), the elements of this matrix
770: control the location and border sharpness of
771: cortical areas. The results do not depend dramatically on their precise
772: values and selectivity in sign, but rather they depend on their relative
773: magnitudes. In order to obtain the correct TC projections in wild-type
774: conditions, it is essential to have the following pattern in the
775: chemo-interactions: axons that are supposed to terminate in the anterior
776: part of the cortex should be strongly repulsed by the guidance molecules
777: located in the posterior end, moderately repulsed by the centrally located
778: guidance molecules, and either attracted or very weakly repulsed by the
779: anterior guidance molecules. Analogically, for the axons that are supposed
780: to terminate in the posterior part, the reverse pattern of interactions
781: is necessary, i.e. these axons should be strongly repulsed by the anterior
782: guidance molecules, moderately repulsed by the central guidance molecules,
783: and either attracted or very weakly repulsed by the posterior molecules.
784: For the axons that are supposed to terminate in the central part of the
785: cortex, it is important that they interact with the posterior and anterior
786: guidance molecules in a symmetric manner, i.e. they should be repulsed
787: by them with approximately the same strength. The larger this strength
788: the more confined the axonal projections are to the central part. The
789: interaction with the central guidance molecules is not very important,
790: i.e. it can be attractive (of any strength) or repulsive (but the strength
791: cannot exceed the repulsion from the lateral molecules).
792:
793:
794: In this paper we assumed that the axon guidance molecules act selectively
795: such that they repulse some classes of TC axons and attract other. However,
796: this assumption is not crucial for the results. Even, if it turns out that
797: the net effect of chemo-interaction between cortical ephrins
798: (and possibly other guidance molecules) and TC axons is repulsion,
799: the appropriately modified
800: interaction matrix would not qualitatively change the results presented
801: here. As an example, in Fig. 9 we show a spatial
802: profile of TC connectivity when the matrix $\gamma$ has only negative
803: elements but with the pattern of their inter-relationships discussed above.
804: In this case one can still generate separate areas, although
805: with much less sharp borders. They shift as Emx2, Pax6 or FGF8 are up- or
806: downregulated, similarly as it happens in the case with the interaction
807: matrix that has negative and positive elements (Fig. 9b). Also, one
808: can generate a mirror symmetry effect when two sources of FGF8 are
809: present at two separate poles (Fig. 9c).
810:
811:
812: It is interesting to note that an analog of the mirror symmetry effect
813: occurs also in the retino-tectal system (Fig. 2c,d in McLaughlin et al,
814: 2003). This takes place in mice with ephrin-A5 and ephrin-A2/ephrin-A5
815: mutants. There is some analogy with our case of the double FGF8 source,
816: because in our model its second ectopic source also modifies
817: the guiding molecules, especially their posterior distributions.
818:
819:
820: In the model we consider, motivated by experimental data, the three types
821: of the guidance molecules. However, one can obtain qualitatively similar
822: results with only one type of guiding molecule located primarily either
823: in the posterior or anterior part of the cortex. Regionalization of the
824: cortex then would be obtained by appropriately choosing the elements of
825: the matrix $\gamma$ in the similar way as we did in this paper. The seemingly
826: uneconomical number of the axon guidance molecules chosen by nature may be
827: necessary for a finer targeting of TC axons within cortical areas.
828:
829:
830: Our model is the first study, which takes into account genetic factors
831: (Emx2, Pax6, FGF8) and their influence on the gradients of the guiding
832: molecules. However, some aspects of the axon guidance part have been
833: modeled previously in the context of retino-tectal projections
834: (Prestige and Willshaw, 1975; Fraser and Hunt, 1980;
835: Whitelaw and Cowan, 1981; Gierer, 1983). The model of Whitelaw and Cowan
836: (1981) puts a greater emphasis on synaptic/neural activity than on
837: guidance molecules. The model of Gierer (1983) seems to be somewhat abstract;
838: it uses minimization procedure of some abstract function related to
839: retinal and tectal gradients in order to obtain topographic mapping.
840: The axonal guidance part of our model is
841: closest in spirit to the models of Prestige and Willshaw (1975), and
842: Fraser and Hunt (1980), which also used the idea of axonal competition.
843: However, there are several significant differences between these models
844: and our model. Our main objective was to generate topographic arealization
845: with sharp borders, while their goal was to generate continuous topography
846: without areas. In terms of mathematics, our model uses the reaction
847: kinetics equations (1-4) on a population level, while Prestige and Willshaw
848: (1975) use algorithmic approach with a discrete time on a single axon/neuron
849: level. Fraser and Hunt (1980), on the other hand, impose several constraints
850: on axonal dynamics that guide them to the correct locations. Also,
851: we consider explicitly the forces between TC axons and the guidance molecules,
852: and additionally stochastic branching effects employed by the diffusion term
853: (see, Eq. (4)); these features are absent in their models.
854:
855:
856: The present model can be also modified by including more biophysical details,
857: more signaling molecules in the pathway, more pathways, and
858: cross-interactions between pathways, if necessary.
859: For example, it is likely that more signaling molecules
860: are involved in regulating axon guidance molecules and that the simple
861: scheme in Fig. 1 will be expanded as more experimental data becomes
862: available. Other candidates playing a role similar to FGF8 but providing
863: positional information from another cortical end, could be
864: WNT and BMP molecules located on the cortical hem (O'Leary and Nakagawa,
865: 2002; Grove and Fukuchi-Shimogori, 2003).
866: However, if those molecules act in coordination with FGF8, then
867: it is unlikely that such an expanded model would
868: qualitatively change the present results.
869:
870:
871:
872: Throughout the paper, we assume that genetic effects determine the very
873: early stage of cortical development and set the basic parcelation of
874: the neocortex (Miyashita-Lin et al, 1999; Nakagawa et al, 1999).
875: Thus, we model only the early cortical development,
876: before and right after the arrival of TC input. Later stages of the
877: development are probably controlled by TC input (i.e. molecules diffused
878: by arrived TC axons), and by activities of neurons and their synapses
879: (Katz and Shatz, 1996), and therefore require a different approach.
880: However, it is likely that those late stage activities only refine that basic
881: patterning plan set by genetic factors
882: (O'Leary and Nakagawa, 2002; Lopez-Bendito and Molnar, 2003).
883: In this respect, it is probable that neural activities can additionally
884: reduce a partial overlap of the areas that is due to molecules diffusion
885: (e.g. see Fig. 4c). Also those axons/branches diffusing around the cortex that
886: do not establish connections, will likely die out at later stages of the
887: development. That effect is not included in our model.
888:
889:
890:
891: \noindent{\bf Acknowledgments}
892:
893: We thank Elizabeth Grove for comments on a draft of this paper.
894: The work was supported by the Sloan-Swartz fellowship at Caltech (J.K.)
895: and the National Science Foundation (G.B.E.).
896:
897:
898:
899:
900: \vspace{1.5cm}
901:
902: \noindent {\bf References}
903:
904: \noindent Bishop, K.M., Goudreau, G., and O'Leary, D.D. (2000).
905: Regulation of area identity in the mammalian neocortex by Emx2 and
906: Pax6. Science 288, 344-349.
907:
908:
909: \noindent Bishop, K.M., Rubenstein, J.L.R., and O'Leary, D.D.
910: (2002). Distinct actions of Emx1, Emx2 and Pax6 in regulating the
911: specification of areas in the developing neocortex. J. Neurosci.
912: 22, 7627-7638.
913:
914:
915: \noindent Cohen-Tannoudji, M., Babinet, C., and Wassef, M. (1994).
916: Early determination of a mouse somatosensory cortex marker.
917: Nature 368, 460-463.
918:
919:
920: \noindent Feldheim, D.A., et al. (2000). Genetic analysis of ephrin-A2
921: and ephrin-A5 shows their requirement in multiple aspects of
922: retinocollicular mapping. Neuron, 25, 563-574.
923:
924:
925: \noindent Finlay, B.L., and Darlington, R.B. (1995). Linked regularities
926: in the development and evolution of mammalian brains. Science
927: 268, 1578-1584.
928:
929:
930: \noindent Fraser, S.E., and Hunt, R.K. (1980). Retinotectal specificity:
931: models and experiments in search of a mapping function. Annu. Rev.
932: Neurosci. 3, 319-352.
933:
934:
935: \noindent Fukuchi-Shimogori, T., and Grove, E.A. (2001). Neocortex
936: patterning by the secreted signaling molecule FGF8. Science
937: 294, 1071-1074.
938:
939:
940: \noindent Fukuchi-Shimogori, T., and Grove, E.A. (2003).
941: Emx2 patterns the neocortex by regulating FGF positional signaling.
942: Nature Neurosci. 6, 825-831.
943:
944:
945: \noindent Garel, S., Huffman, K.J., and Rubenstein, J.L.R. (2003).
946: Molecular regionalization of the neocortex is disrupted in FGF8
947: hypomorphic mutants. Development 130, 1903-1914.
948:
949:
950: \noindent Gierer, A. (1983). Model for the retino-tectal projection.
951: Proc. R. Soc. Lond. B 218, 77-93.
952:
953:
954: \noindent Goodhill, G.J., and Richards, L.J. (1999). Retinotectal
955: maps: molecules, models and misplaced data. Trends Neurosci. 22, 529-534.
956:
957:
958: \noindent Grove, E.A., and Fukuchi-Shimogori, T. (2003). Generating
959: the cerebral cortical area map. Annu. Rev. Neurosci. 26,
960: 355-380.
961:
962:
963: \noindent Hofman, M.A. (1989). On the evolution and geometry of
964: the brain in mammals. Prog. Neurobiol. 32, 137-158.
965:
966:
967: \noindent Karbowski, J. (2003). How does connectivity between
968: cortical areas depend on brain size? Implications for efficient computation.
969: J. Comput. Neurosci. 15, 347-356.
970:
971:
972: \noindent Katz, L.C., and Shatz, C.J. (1996). Synaptic activity
973: and the construction of cortical circuits. Science 274,
974: 1133-1138.
975:
976:
977: \noindent Knoll, B., and Drescher, U. (2002). Ephrin-As as receptors
978: in topographic projections. Trends Neurosci. 25, 145-149.
979:
980:
981: \noindent Krubitzer, L. (1995). The organization of neocortex in mammals:
982: Are species differences really so different? Trends Neurosci.
983: 18, 408-417.
984:
985:
986: \noindent Krubitzer, L., and Huffman, K.J. (2000). Arealization
987: of the neocortex in mammals: genetic and epigenetic contributions to the
988: phenotype. Brain Behav. Evol. 55, 322-335.
989:
990:
991: \noindent Lopez-Bendito, G., and Molnar, Z. (2003). Thalamo-cortical
992: development: How are we going to get there? Nature Rev. Neurosci.
993: 4, 276-289.
994:
995:
996:
997:
998: \noindent Mackarehtschian, K., Lau, C.K., Caras, I., and
999: McConnel, S.K. (1999). Regional differences in the developing
1000: cerebral cortex revealed by $Ephrin-A5$ expression. Cereb.
1001: Cortex 9, 601-610.
1002:
1003:
1004: \noindent Mallamaci, A., Muzio, L., Chan, C.H., Parnavelas, J.,
1005: and Boncinelli, E. (2000). Area identity shifts in the early cerebral
1006: cortex of $Emx2^{-/-}$ mutant mice. Nature Neurosci. 3,
1007: 679-686.
1008:
1009:
1010: \noindent Mann, F., Peuckert, C., Dehner, F., Zhou, R.,
1011: and Bolz, J. (2002). Ephrins regulate the formation of terminal
1012: axonal arbors during the development of thalamocortical projections.
1013: Development 129, 3945-3955.
1014:
1015:
1016: \noindent McLaughlin, T., Hindges, R., and O'Leary, D.D.M. (2003).
1017: Regulation of axonal patterning of the retina and its topographic
1018: mapping in the brain. Curr. Opin. Neurobiol. 13, 57-69.
1019:
1020:
1021: \noindent Miyashita-Lin, E.M., Hevner, R., Wassarman, K.M.,
1022: Martinez, S., and Rubenstein, J.L.R. (1999). Early neocortical
1023: regionalization in the absence of thalamic innervation. Science
1024: 285, 906-909.
1025:
1026:
1027: \noindent Murray, J.D. (1993). {\it Mathematical Biology. \/}
1028: Springer-Verlag, Berlin, pp. 109-139.
1029:
1030:
1031: \noindent Muzio, L. et al (2002). Emx2 and Pax6 control
1032: regionalization of the pre-neuronogenic cortical primordium.
1033: Cereb. Cortex 12, 129-139.
1034:
1035:
1036: \noindent Nakagawa, Y., Johnson, J.E., and O'Leary, D.D.M. (1999).
1037: Graded and areal expression patterns of regulatory genes and cadherins
1038: in embryonic neocortex independent of thalamocortical input.
1039: J. Neurosci. 19, 10877-10885.
1040:
1041:
1042: \noindent Nauta, W.J.H., and Feirtag, M. (1986). {\it
1043: Fundamental Neuroanatomy. \/} (Freeman, New York).
1044:
1045:
1046: \noindent Northcutt, R.G., and Kaas, J.H. (1995). The emergence
1047: and evolution of mammalian neocortex. Trends Neurosci.
1048: 18, 373-379.
1049:
1050:
1051: \noindent O'Leary, D.D.M. (1989). Do cortical areas emerge from
1052: a protocortex? Trends Neurosci. 12, 400-406.
1053:
1054:
1055: \noindent O'Leary, D.D.M., Yates, P.A., McLaughlin, T. (1999).
1056: Molecular development of sensory maps: representing sights and smells
1057: in the brain. Cell 96, 255-269.
1058:
1059:
1060: \noindent O'Leary, D.D.M., and Nakagawa, Y. (2002). Patterning
1061: centers, regulatory genes and extrinsic mechanisms controlling
1062: arealization of the neocortex. Curr. Opin. Neurobiol. 12,
1063: 14-25.
1064:
1065:
1066: \noindent Prestige, M.C., and Willshaw, D.J. (1975). On a role for
1067: competition in the formation of patterned neural connexions.
1068: Proc. Roy. Soc. Lond. B 190, 77-98.
1069:
1070:
1071: \noindent Rakic, P. (1988). Specification of cerebral cortical
1072: areas. Science 241, 170-176.
1073:
1074: \noindent Simon, D.K., and O'Leary, D.D.M. (1992). Development of
1075: topographic order in the mammalian retinocollicular projection.
1076: J. Neurosci. 12, 1212-1232.
1077:
1078:
1079: \noindent Sperry, R.W. (1963). Chemoaffinity in the orderly
1080: growth of nerve fibers patterns and connections. Proc.
1081: Natl. Acad. Sci. USA 50, 703-710.
1082:
1083:
1084: \noindent Takemoto, M., et al (2002). Ephrin-B3-EphA4
1085: interactions regulate the growth of specific thalamocortical
1086: axon populations in vitro. Eur. J. Neurosci. 16, 1168-1172.
1087:
1088:
1089: \noindent Uziel, D. et al (2002). Miswiring of limbic
1090: thalamocortical projections in the absence of ephrin-A5.
1091: J. Neurosci. 22, 9352-9357.
1092:
1093:
1094: \noindent Vanderhaeghen, P. et al (2000). A mapping label
1095: required for normal scale of body representation in the cortex.
1096: Nature Neurosci. 3, 358-365.
1097:
1098:
1099: \noindent Whitelaw, V.A., and Cowan, J.D. (1981). Specificity and
1100: plasticity of retinotectal connections: a computational model.
1101: J. Neurosci. 1, 1369-1387.
1102:
1103:
1104: \noindent Wolpert, L. (1969). Positional information and spatial
1105: pattern of cellular differentiation. J. Theor. Biol. 25, 1-47.
1106:
1107:
1108: \noindent Wolpert, L. (1996). One hundred years of positional
1109: information. Trends Genet. 12, 359-364.
1110:
1111:
1112: \noindent Yates, P.A., Roskies, A.L., McLaughlin, T., and O'Leary, D.D.M.
1113: (2001). Topographic-specific axon branching controlled by ephrin-As is
1114: the critical event in retinotectal map development. J. Neurosci. 21,
1115: 8548-8563.
1116:
1117: \noindent Yates, P.A., Holub, A.D., McLaughlin, T., Sejnowski, T.J.,
1118: and O'Leary, D.D.M. (2004). Computational modeling of retinotopic map
1119: development to define contributions of EphA-EphrinA gradients, axon-axon
1120: interactions, and patterned activity. J. Neurobiol. 59, 95-113.
1121:
1122:
1123:
1124: \newpage
1125:
1126: {\bf \large Figure Captions}
1127:
1128: Fig. 1\\
1129: Schematic diagram depicting the pathway influencing the axon
1130: guidance molecules A, B, and C. The transcription factors Emx2
1131: and Pax6 repress each
1132: other with strengths $v_{1}$ and $v_{2}$. Emx2 represses FGF8 with
1133: strength $w_{1}$, and FGF8 represses Emx2 with strength $w_{2}$.
1134: The renormalized expression concentration $f$ of FGF8 or some other
1135: signal activated by it serves as
1136: a morphogen for the guiding molecules A, B, and C.
1137:
1138:
1139: \vspace{0.3cm}
1140:
1141: Fig. 2\\
1142: Stationary spatial profiles of the signal $f$ and the guiding molecules
1143: under normal conditions (wild type). (A) Renormalized concentration $f(x)$
1144: of FGF8. (B) Concentrations of the guiding molecules A, B, and C. Dashed
1145: line represents the concentration of A, solid line corresponds to B, and
1146: dashed-dotted line represents the molecule C.
1147: Parameters used: $A_{emx}= 1.34$, $A_{pax}= 1.4$, $A_{fgf}= 0.9$,
1148: $\zeta_{emx}=25.6$, $\zeta_{pax}=27.3$, $\zeta_{fgf}=26.4$,
1149: $w_{1}=2.4$, $w_{2}=2.1$, $v_{1}=2.6$, $v_{2}=2.7$,
1150: $\gamma_{A1}= 1.6$, $\gamma_{A2}= -0.4$, $\gamma_{A3}= -2.21$,
1151: $\gamma_{A4}= -2.1$, $\gamma_{A5}= -2.45$,
1152: $\gamma_{B1}= -0.6$, $\gamma_{B2}= -0.5$, $\gamma_{B3}= 0.4$,
1153: $\gamma_{B4}= -0.5$, $\gamma_{B5}= -1.0$,
1154: $\gamma_{C1}= -2.9$, $\gamma_{C2}= -2.5$, $\gamma_{C3}= -2.23$,
1155: $\gamma_{C4}= -0.6$, $\gamma_{C5}= 1.7$,
1156: $\theta_{1}= 0.77$, $\theta_{2}= 0.5$,
1157: $\theta_{3}= 0.39$, $\theta_{4}= 0.08$, $\kappa_{A}=0.58$,
1158: $\kappa_{B}=0.9$, $\kappa_{C}=0.55$, $\sigma_{A}= \sigma_{B}= \sigma_{C}=
1159: 0.2$. Parameters for the dynamics of TC connections: $D= 0.1$, $L= 40$,
1160: grid size $dx= 0.25$, $\alpha_{i}=3.0$, $\beta_{i}=3.0$ for $i=1,2,3$.
1161:
1162:
1163:
1164:
1165: \vspace{0.3cm}
1166:
1167:
1168: Fig. 3\\
1169: Temporal evolution of the pattern of axonal densities around the cortical
1170: surface under normal conditions. (A) Initial distribution,
1171: (B) distribution after $t= 6$,
1172: (C) steady-state. Note the emergence of the heterogeneous pattern.
1173: Solid line represents the profile of $a_{1}$, dashed-dotted line
1174: correspond to $a_{2}$, solid line with open circles represents $a_{3}$,
1175: dashed line corresponds to $a_{4}$, and dotted line corresponds to $a_{5}$.
1176: Parameters are the same as in Fig. 2.
1177:
1178:
1179: \vspace{0.3cm}
1180:
1181:
1182: Fig. 4\\
1183: Temporal evolution of the pattern of TC connectivity under normal conditions.
1184: (A) Initial distribution with no TC connectivity
1185: (all lines collapse onto one), (B) distribution after
1186: $t= 6$, (C) steady-state. Note the emergence of sharp bordered areas with
1187: different axon types. Regions of high values of $c_{1}$, $c_{3}$, and $c_{5}$
1188: correspond to areas M1, S1, and V1, respectively.
1189: Solid line represents the profile of $c_{1}$, dashed-dotted line
1190: correspond to $c_{2}$, solid line with open circles represents $c_{3}$,
1191: dashed line corresponds to $c_{4}$, and dotted line corresponds to $c_{5}$.
1192: (D) Schematic stationary pattern of emerged areas on the cortical surface,
1193: corresponding to the TC connectivity in (C).
1194: Parameters are the same as in Fig. 2.
1195:
1196:
1197: \vspace{0.3cm}
1198:
1199:
1200:
1201: Fig. 5\\
1202: Stationary spatial profiles of the signal $f$, the axon guidance molecules,
1203: axonal densities, TC connectivity, and cortical area pattern
1204: when the transcription factor Emx2 is not expressed. For Emx2 mutants
1205: the distributions of signal $f(x)$ (A), guiding molecules (B),
1206: density of axons (C), TC connectivity (D), and cortical areas (E)
1207: all shift posteriorly. In (A) the dashed line corresponds to the control
1208: distribution of $f(x)$ from fig. 2A. Parameters are the same as in Fig. 2,
1209: except for $A_{emx}= 0$.
1210:
1211:
1212:
1213: \vspace{0.3cm}
1214:
1215:
1216: Fig. 6\\
1217: Stationary spatial profiles of the signal $f$, the axon guidance molecules,
1218: axonal densities, TC connectivity, and cortical area pattern
1219: when the transcription factor Pax6 is not expressed. For Pax6 mutants the
1220: distributions of signal $f(x)$ (A), guiding molecules (B), density of axons
1221: (C), TC connectivity (D), and cortical areas (E) all shift anteriorly.
1222: In (A) the dashed line corresponds to the control distribution of $f(x)$
1223: from fig. 2A. Parameters are the same as in Fig. 2, except for $A_{pax}= 0$.
1224:
1225:
1226: \vspace{0.3cm}
1227:
1228:
1229: Fig. 7\\
1230: The influence of FGF8 on the cortical architecture and guiding molecules.
1231: (A), (B), and (C) Overexpression of FGF8. (D), (E), and (F) underexpression
1232: of FGF8. Note the opposite shifts of the areas in these cases (compare
1233: (B), (C) with (E), (F), respectively).
1234: Line convention is the same as in Figs. 2 and 4.
1235: Parameters used: (A), (B) $A_{fgf}= 1.6$, $\zeta_{fgf}= 32.8$, (C), (D)
1236: $A_{fgf}= 0.6$, $\zeta_{fgf}= 19.0$. Other parameters are the same as
1237: in Fig. 2.
1238:
1239:
1240: \vspace{0.3cm}
1241:
1242:
1243: Fig. 8\\
1244: Generation of two partly symmetric S1 areas when two sources of FGF8
1245: are present. (A) Stationary spatial pattern of TC connectivity and
1246: (B) corresponding area pattern on the cortical surface.
1247: Note the mirror symmetry effect, i.e. not only two S1 areas ($c_{3}$)
1248: are generated, but
1249: also two M1 fields ($c_{1}$) are present (one regular, second in
1250: the posterior end), and V1 area ($c_{5}$) is located in between
1251: two S1 areas. (C) Spatial profile of the signal $f$. Note a change
1252: of shape and the appearance of a minimum. (D) Stationary concentrations
1253: of the guiding molecules. Note that the molecule B is expressed broadly
1254: and it has two maxima, the molecule A is also partly expressed in the
1255: posterior end, and C is weakly expressed in the central part corresponding
1256: to minimum in $\rho_{B}$. The figures are generated by assuming that
1257: there is a second source of FGF8 at the posterior end with the amplitude
1258: $A_{fgf}'= 1.5$ and the range $\zeta_{fgf}'= 12.0$. Other parameters
1259: are the same as in Fig. 2.
1260:
1261:
1262:
1263:
1264: \vspace{0.3cm}
1265:
1266:
1267: Fig. 9\\
1268: Stationary spatial patterns of TC connectivity in the case of purely
1269: repulsive interactions between guiding molecules and axons. (A)
1270: Pattern of TC connectivity under normal conditions. (B) Pattern of
1271: TC connectivity with Emx2 mutation, i.e. $A_{emx}= 0$.
1272: (C) Pattern of TC connectivity when
1273: two sources of FGF8 are present similarly as in Fig. 8. Note a mirror
1274: symmetry effect. Parameters used are the same as in Fig. 2 except:
1275: $\gamma_{A1}= -0.08$, $\gamma_{B3}= -0.1$, $\gamma_{C5}= -0.06$.
1276:
1277:
1278:
1279: %\end{narrowtext}
1280: \end{document}
1281:
1282:
1283:
1284:
1285:
1286:
1287:
1288:
1289:
1290:
1291:
1292:
1293:
1294:
1295:
1296:
1297:
1298:
1299:
1300:
1301:
1302:
1303:
1304:
1305:
1306:
1307:
1308:
1309:
1310:
1311:
1312:
1313:
1314:
1315:
1316:
1317:
1318:
1319:
1320:
1321:
1322:
1323:
1324:
1325:
1326:
1327:
1328:
1329:
1330:
1331:
1332:
1333:
1334:
1335:
1336:
1337:
1338:
1339:
1340:
1341:
1342:
1343:
1344:
1345:
1346:
1347:
1348:
1349:
1350:
1351:
1352:
1353:
1354:
1355:
1356:
1357:
1358:
1359:
1360:
1361:
1362:
1363:
1364:
1365:
1366:
1367:
1368:
1369:
1370:
1371:
1372:
1373:
1374:
1375:
1376:
1377:
1378:
1379:
1380: