1: % todo: Rever biografia Raya:2003
2:
3: \documentclass[aps,twocolumn,showpacs,preprintnumbers,amsmath,amssymb]{revtex4}
4:
5: %\documentclass[twocolumn,showpacs,preprintnumbers,amsmath,amssymb]{revtex4}
6: %\documentclass[preprint,showpacs,preprintnumbers,amsmath,amssymb]{revtex4}
7: %\documentclass[aps,pre,preprint,groupedaddress,floatfix]{revtex4}%,showpacs
8: \usepackage{graphicx}% Include figure files
9: \usepackage{dcolumn}% Align table columns on decimal point
10: \usepackage{bm}% bold math
11: \usepackage{amssymb}
12: \usepackage{amsmath}
13: %\nofiles
14:
15: \bibliographystyle{apsrev}
16:
17: \begin{document}
18:
19: %\preprint{APS/123-QED}
20:
21: \title{A Novel Field Approach to 3D Gene Expression Pattern Characterization}
22:
23: \author{L. da F. Costa}
24: \email{luciano@if.sc.usp.br}
25: \author{B. Traven\c{c}olo}
26: \email{bant@if.sc.usp.br}
27: \author{A. Azeredo}
28: \affiliation{Instituto de F\'{\i}sica de S\~{a}o Carlos,
29: Universidade de S\~{a}o Paulo, Av. Trabalhador S\~{a}o Carlense
30: 400, Caixa Postal 369, CEP 13560-970, S\~{a}o Carlos, SP, Brazil}
31:
32: \author{M. E. Beletti}
33: % \email{mebeletti@ufu.br}
34: \affiliation{Instituto de Ci\^{e}ncias Biom\'{e}dicas, Universidade
35: Federal de Uberl\^{a}ndia, Av. Par\'{a} 1720, CEP 38400-902, Uberl\^{a}ndia,
36: MG, Brazil}
37:
38: \author{D. Rasskin-Gutman}
39: \author{G. Sternik}
40: % \altaffiliation[Also at ]{Salk Institute}
41: \author{J. C. I. Belmonte}
42: \author{M. Iba\~{n}es}
43: % \altaffiliation[Also at ]{Salk Institute}
44: \affiliation{Salk Institute, 10010N Torrey Pines Road, La Jolla,
45: USA CA 92037}
46:
47: \author{G. B. M\"{u}ller}
48: \affiliation{Department of Zoology, University of Vienna,
49: Althaustrasse 14, A-1090 Vienna Austria.}
50:
51: \date{\today}
52:
53: \begin{abstract}
54: We present a vector field method for obtaining the spatial
55: organization of 3D patterns of gene expression based on gradients
56: and lines of force obtained by numerical integration. The
57: convergence of these lines of force in local maxima are identified
58: as centers of gene expression, providing a natural and powerful
59: framework to characterize the organization and dynamics of
60: biological structures. We apply this novel methodology to analyze
61: the expression patterns of light chain myosin II protein linked to
62: enhanced green fluorescent protein (EGFP) during zebrafish heart
63: formation.
64: \end{abstract}
65:
66: %\pacs{Valid PACS appear here}
67:
68: \maketitle
69:
70:
71: Animal development involves synchronized gene activation modulated
72: by environmental influences \cite{Carroll:2001,Gilbert:2003}. Far
73: from being uniform, such a gene expression gives rise to
74: structured spatial and temporal patterns of varying protein
75: concentration. Recent advances in biochemical and imaging methods
76: have paved the way to obtaining 3D reconstructions of spatial gene
77: activation \cite{Streicher:2000} which can be analyzed in order to
78: better understand the intricate mechanisms governing tissue, organ
79: and member formation \cite{Gilbert:2003}. Among the several
80: currently available methodologies allowing characterization of 3D
81: gene expression, special attention has been given to EGFP
82: (Enhanced Green Fluorescence Protein). The EGFP is used as a
83: marker. Its expression is controlled by the promoter of the gene
84: of interest creating a fluorescent fusion protein that maintains
85: the normal functions and localization of the wild type protein.
86: This methodology can be used to demonstrate gene activity in
87: intact cells and organisms, while taking into account the fact
88: that the host protein is continuously synthesized, degraded, and
89: suffering alterations within cells
90: \cite{Tisen:1998,Patterson:2003}. As such a type of gene
91: expression data becomes available, it is important to identify and
92: develop mathematical methologies for measuring and modeling
93: spatial gene activation. In addition to traditional approaches
94: (e.g. density or dispersion estimation), it is important to
95: consider more sophisticated methods capable of addressing more
96: directly aspects related to the dynamics of the involved
97: biological processes, such as cell communication and migration
98: \cite{Schock:2002,Kuure:2000}, which play an important role during
99: both embryonic development and pathological processes.
100:
101: In this article we characterize the spatial organization of gene
102: expression patterns in order to assess the geometrical basis of some
103: dynamical processes during morphogenesis. To this end, we compute a
104: ``gene expression landscape'' as a scalar field $\omega = g(x, y, z)$,
105: where $\omega$ is interpreted as the amount of expression of the
106: protein in the spatial position $(x, y, z)$. The same approach can be
107: used to model and predict the dissemination of cell signalling or
108: other influence factors emmanating from the cell under analysis which,
109: combined with the possibility of adopting varying values of the
110: parameters affecting the field (e.g. the dielectric constant), defines
111: a truly general framework for expressing field influences. In analogy
112: with the potential dynamics of dissipative systems, we obtain the
113: spatial trajectories (lines of force) corresponding to maximizing the
114: gradient of gene expression. Such trajectories tend to converge to
115: local peaks of activity, defining gene expression centers. It is
116: proposed in this article that the distribution of such centers provide
117: a natural framework for characterizing and analyzing the spatial
118: interactions between the involved developmental rudiments. The
119: potential of such a methodology is illustrated with respect to the
120: analysis of zebrafish heart formation from 3D gene expression data.
121:
122: Zebrafish embryos have been widely used in order to study heart
123: formation, due to their transparency and its partial independence
124: from the cardiovascular system. For vertebrates, the heart is the
125: first organ that forms and starts operating~\cite{Stainier:2001}.
126: Constrictions and bending (folding) are key elements in the early
127: morphogenetic shaping of the heart tube. The spatial gene
128: expression data considered in this work was acquired through the
129: observation of 42-hour post-fertilization transgenic zebrafish
130: embryos expressing EGFP specific for heart mesoderm myosin light
131: chain (mlc2a-EGFP)~\cite{Raya:2003}. The zebrafish embryos were
132: anesthetized and kept fixed, and live-images of the heart were
133: taken at ambient room temperature. The image recordings were made
134: using a Nikon Eclipse TE300 inverted microscope using 20x/0.75 NA
135: magnification. The microscope is coupled to a Bio-Rad Radiance MP
136: 2100 scanning multiphoton confocal system (Cambridge,MA) with a
137: two-photon Tsunami laser (Spectra Physics, CA). The GFP was
138: excited with the two-photon laser, at 900 nm. The total dataset is
139: composed of 110 confocal sections.
140:
141: All the 110 confocal slices were combined so as to obtain the
142: three-dimensional volume of the heart, from which the gene
143: expression landscape was computed as described above. It is
144: interesting to note that this scalar field can be visualized with
145: direct volume rendering algorithms (DVR)~\cite{Schroeder:1996}. In
146: order to minimize the spatial quantization noise implied by
147: digital image representation, gaussian smoothing was applied over
148: the gene concentration data. This is done through the discrete
149: convolution of a three-dimensional Gaussian kernel {\it k(x,y,z)}
150: with the scalar field {\it w}, as expressed in Eq.
151: (\ref{Eq_GaussSmooth})
152:
153: \begin{eqnarray} \label{Eq_GaussSmooth}
154: w(x,y,z)*k(x,y,z) &=& \sum_{i,j,k}w(i,j,k) \nonumber \\ &\times&
155: k((x-i),(y-j),(z-k))
156: \end{eqnarray}
157:
158: The smoothed reconstruction of the 3D gene activity pattern is
159: shown in Figure~\ref{Fig_Cluster}a. The gradient of this scalar
160: field was estimated by using the enhanced finite differences
161: scheme described in \cite{Zucker:1981}, by convolving the gene
162: expression concentration with three-dimensional masks. Next, we
163: compute the lines of force by calculating the trajectories that
164: maximize the gradient starting form arbitrary spatial positions
165: sampled as points uniformly distributed through spheres centered
166: at the three-dimensional volume.
167:
168: The considered lines of force would correspond, for instance, to
169: the putative path (set of 3D coordinates) followed by an object at
170: position $\vec{r}=(x,y,z)$ with gradient dissipative dynamics:
171:
172: \begin{equation}
173: \frac{\partial \vec{r}}{\partial
174: t}=\vec{\nabla}\{\omega(x,y,z)*k(x,y,z)\}\,,
175: \end{equation}
176:
177: standard numerical integration was used in order to estimate such
178: lines of force, which are illustrated in
179: Figure~\ref{Fig_Cluster}b. The sampling criteria removed the lines
180: whose scalar value of its end point were less than 10 (from a
181: range of 0 to 255), eliminating those that do not reach the
182: regions where mlc2a was being expressed. Small and too long
183: trajectories were also removed, because they were influenced by
184: noises. As expected, these lines converge to local maxima of the
185: scalar gene expression field, which could be considered as
186: \emph{gene expression centers}. In analogy to graph theory, the
187: total number of sampled lines of force converging to a specific
188: center is referred to as the center \emph{degree}. A total of 734
189: lines and 89 centers were obtained for the considered 3D gene
190: expression data.
191:
192:
193: \begin{figure*}
194: \includegraphics[scale=1]{DVRandCluster.eps}
195: \caption{\label{Fig_Cluster}(a) Visualization of the smoothed and
196: reconstructed gene activity pattern of mlc2a during zebrafish
197: heart formation. The inflow pole is on the upper left. Arrows~1
198: and 2 indicates the constriction$/$bending regions. The respective
199: lines of force are shown in (b), segregated into black and white
200: as described in the text.}
201: \end{figure*}
202:
203:
204:
205:
206: Figure ~\ref{Fig_Cluster}b shows the sampled lines of force
207: obtained by using the above described methodology, drawn in black
208: or white according to thresholding criteria: the lines
209: corresponding to gene expression activity centers with degrees
210: smaller than 14 have been marked in white. Such threshold value
211: was defined with basis on the relative frequency histogram of the
212: distribution of centers degree, showed in Figure~\ref{Fig_Hist}.
213: It can be seen from Figure~\ref{Fig_Cluster}b that the genic
214: activity centers exhibiting higher numbers of converging lines of
215: force (marked black) tend to concentrate along the regions
216: subjected to the constriction and folding implied by the heart
217: formation dynamics (marked by arrow~$1$ in Figure
218: ~\ref{Fig_Cluster}a) as well as the sinus venosus (marked by
219: arrow~$2$ in Figure ~\ref{Fig_Cluster}a). The following
220: biological interpretation are suggested in order to account for
221: such result.
222:
223: \begin{figure}[t]
224: \includegraphics[scale=1]{histDegree.eps}
225: \caption{\label{Fig_Hist} Relative frequency histogram of the
226: distribution of centers degree}
227: \end{figure}
228:
229: The heart forms from a tube of epimyocardial cells that express,
230: among other genes, mlc2a. This gene is expressed uniformly
231: throughout the heart, with the possibility of a weaker expression
232: in the inflow pole, i.e. the region of the venous sinus and the
233: atrium (Figure~\ref{Fig_Cluster}a). It is suggested here that the
234: distribution of active cells could be determined by a gene
235: activity field in such a way that the higher degree activity
236: centers positively regulates the activation patterns of
237: surrounding cells. The line of force pattern indicates that the
238: expression of these cells coincides with morphogenetic events of
239: heart formation, in particular the characteristic constrictions
240: and bendings of the heart tube at the atrio-ventricular and the
241: ventriculo-bulbar borders (arrows in Figure~\ref{Fig_Cluster}a),
242: which are sites composed by high degree activity centers,
243: involving cells more actively producing mlc2a. This process might
244: be affected by the differential distribution of gene activation
245: centers, as indicated by the respective numbers of lines of force
246: which tended to be smoother at these locations.
247:
248:
249: While such hypotheses can only be verified through further
250: experimental investigations, a novel methodology for 3D gene
251: activity characterization has been shown to provide a natural and
252: effective means for quantifying the spacial interactions between
253: the biological structures involved in gene expression. Unlike
254: differential measurements such as gradients or divergent
255: magnitudes, the estimation of the lines of force and activity
256: centers are integral features, indicating spatial interactions
257: over substantial distances. It is expected that the proposed
258: framework will prove to be useful in a number of other gene
259: expression investigations, paving the way to a more objective
260: understanding of the dynamics governing animal development and its
261: pathologies.
262:
263:
264: \begin{acknowledgments}
265:
266: The authors thank HFSP RGP39/2002 for funding this project.
267: A. Azeredo are grateful to FAPESP (02/09149-2), and B. Traven\c{c}olo
268: is grateful to CAPES and FAPESP (03/13072-8) for financial
269: support. M. Iba\~nes is partially supported by the Fulbright Program
270: and Generalitat of Catalunya. Luciano da F. Costa is grateful to
271: FAPESP (proc. 99/12765-2) and CNPq (proc. 301422/92-3) for financial
272: support.
273:
274:
275: \end{acknowledgments}
276:
277:
278:
279: %\newpage %Just because of unusual number of tables stacked at end
280: \bibliography{pre3}% Produces the bibliography via BibTeX.
281:
282:
283: \end{document}
284: