- Email: [email protected]
Depletion of FGF acts as a lateral inhibitory factor in lung branching morphogenesis in vitro
Mechanisms of Development 116 (2002) 29–38 www.elsevier.com/locate/modo
Depletion of FGF acts as a lateral inhibitory factor in lung branching morphogenesis in vitro Takashi Miura a,*, Kohei Shiota a,b a
Department of Anatomy and Developmental Biology, Kyoto University Graduate School of Medicine, Kyoto 606-8501, Japan b Congenital Anomaly Research Center, Kyoto University Graduate School of Medicine, Kyoto 606-8501, Japan Received 31 January 2002; received in revised form 28 March 2002; accepted 11 April 2002
Abstract Previous studies have shown that the interaction of positive and inhibitory signals plays a crucial role during lung branching morphogenesis. We found that in mesenchyme-free conditions, the lung epithelium exerted a lateral inhibitory effect on the neighbouring epithelium via depletion of fibroblast growth factor 1 (FGF1). Contrary to previous suggestions, bone morphogenetic protein 4 could not substitute for the inhibitory effect. Based on of this observation, we used a reaction–diffusion model of the substrate-depletion type to represent the initial phase of in vitro branching morphogenesis of lung epithelium, with depletion of FGF playing the role of lateral inhibitor. The model was able to account for the effects of the FGF1 concentration, extracellular matrix degradation and different subtypes of FGF on morphogenesis of the lung bud epithelia. These results suggest that the depletion of FGF may be a key regulatory component in initial phase of branching morphogenesis of the lung bud epithelium in vitro. q 2002 Elsevier Science Ireland Ltd. All rights reserved. Keywords: Lung; Branching morphogenesis; Theoretical model; Fibroblast growth factor depletion
1. Introduction During mammalian development, the lung primordium first appears as an epithelial outgrowth from the ventral surface of the embryonic foregut and undergoes branching morphogenesis to form a bronchial tree (Gilbert, 1997). Its branching morphogenesis has been extensively studied experimentally in the field of developmental biology with the aid of organ culture systems, which reproduce branching morphogenesis in vitro. Some classical experiments revealed the importance of extracellular matrices in lung morphogenesis. For example, when collagenase was added to the culture medium, lung bud explants showed a cystic morphology instead of branches (Ganser et al., 1991). Recent studies with molecular techniques suggest that diffusible signaling molecules such as fibroblast growth factor 10 (FGF10), Sonic hedgehog (Shh) and bone morphogenetic protein 4 (BMP4) play key roles in branching morphogenesis (Bellusci et al., 1997a,b, 1996; Lebeche et al., 1999; Park et al., 1998; Pepicelli et al., 1998). Two schemes have been proposed to explain the lung branching morphogenesis (Hogan, 1999). One is the ‘hardwired’ scheme, which hypothesizes that a prepattern of gene * Corresponding author. Tel.: 181-75-753-4341; fax: 181-75-751-7529. E-mail address: [email protected] (T. Miura).
expression exists in the lung bud. Recently some supportive evidence for this scheme was provided by a chick experiment where the Hox gene cluster was shown to be unevenly expressed in the lung mesenchyme (Packer et al., 2000; Sakiyama et al., 2000). The other is the ‘generative’ scheme, in which the branching pattern is supposed to be determined dynamically by epithelial–mesenchymal interaction. In this scheme, the lateral inhibitory effect of BMP4 secreted from the distal tip of the lung bud endoderm is assumed to play a major role in determining the budding pattern (Hogan, 1999). These two schemes are not mutually exclusive, and recent studies on somite and limb development suggest that these two may act simultaneously (Richardson et al., 1998). From the theoretical point of view, several mathematical models have been proposed for branching morphogenesis in non-living systems such as crystal formation and viscous fingering, and the dynamics of these systems are well understood (Ball, 1999). Up to the present, experimental justification is lacking for these models to be applied to lung branching morphogenesis in the embryo. For example, although branching morphogenesis in the lung has been hypothesized to be homologous to viscous fingering (Lubkin and Murray, 1995), this bulk transport model does not take diffusible signals such as FGF (Bellusci et al., 1997b; Park et al., 1998) into account. Moreover, our recent finding on dynamic movement of the lung bud epithe-
0925-4773/02/$ - see front matter q 2002 Elsevier Science Ireland Ltd. All rights reserved. PII: S 0925-477 3(02)00132-6
30
T. Miura, K. Shiota / Mechanisms of Development 116 (2002) 29–38
lium is inconsistent with this hypothesis (Miura and Shiota, 2000). On the other hand, a physiology-based model for the human bronchial tree has been proposed (Kitaoka et al., 1999) in which the branch pattern of the lung is related to the efficiency of gas exchange. However, it is difficult to apply this model to lung morphogenesis in embryos because the branched structure is established well before respiration starts. Therefore, no theoretical models for lung branching morphogenesis proposed so far are adequately supported by experimental data. Both experimental schemes and theoretical models have their own strengths and limitations. Experimental schemes can include interaction of many molecules which can be tested by actual experiments and suitable to explain in vivo situation, but usually the scheme becomes too complex and it is difficult to understand the qualitative morphological change (such as transformation from branches to cysts) only by the schemes. On the other hand, theoretical models can explain a wide variety of morphologies by a single mechanism and helpful in understanding and predicting the qualitative morphological changes, but it usually contains only one or two factors and sometimes seems too simple to be applied to the in vivo situation. To utilize the advantages of both, we first chose the simplest in vitro experimental system which retains a key aspect of morphogenesis in vivo, and then applied a theoretical model to in vitro pattern formation, and finally discussed the relationship between in vitro and in vivo situations. It has been shown that isolated lung epithelia retain a capacity to undergo branching morphogenesis in vitro. Nogawa and Ito (1995) showed that when the lung epithelia were detached from the surrounding mesenchyme and cultured in Matrigel in the presence of FGF1, the epithelia underwent branching morphogenesis. The mesenchymefree culture is a common experimental system to study the mechanism of branching morphogenesis in vitro (Bellusci et al., 1997b; Park et al., 1998), and Cardoso et al. (1997) showed that FGF7 exerts a similar effect on the lung epithelial explant irrespective of the coexistence of mesenchyme. These studies strongly imply that the in vitro culture system reproduces the generative aspect of lung branching morphogenesis occurring in vivo. In the present study, we chose this mesenchyme-free culture as a simple experimental system of the earliest stage of the branching morphogenesis and examined the mechanism of morphogenesis in this culture system using both experimental and theoretical approaches. We found that in a mesenchyme-free condition, the isolated lung explant exerts a lateral inhibitory effect on the neighbouring lung epithelium. However, BMP4, which was assumed to be a lateral inhibitory factor in a previous experimental scheme (Hogan, 1999), could not reproduce this result in the mesenchyme-free culture system. Since the FGF protein concentration was lower around the cultured epithelia than in the other area of the gel, we proposed an alternative mechanism that epithelial cells consume FGF and the result-
ing depletion of FGF exerts a lateral inhibitory effect. We constructed a reaction–diffusion model of substrate-depletion type which could reproduce branching pattern through computer simulation, and showed that the model could explain the lung pattern changes observed under various experimental conditions. 2. Results 2.1. The neighbouring lung epithelia exert inhibitory effect in mesenchyme-free culture To examine whether the lateral inhibitory mechanism proposed by Hogan (1999) actually works in cultured embryonic lung epithelia, we cultured two epithelia in close vicinity and observed their budding pattern. If the site of bud formation is predetermined or intrinsically determined inside the lung epithelium, the location of buds should not be influenced by the nearby epithelium. In our experiment, budding of the epithelium was suppressed between the two epithelial explants (Fig. 1a,b), which means that the location of bud formation is not predetermined in the epithelium but is regulated by some extrinsic inhibitory factors in the extracellular environment. To confirm this, we counted the number of buds which grew toward the other epithelial explant and that of the buds growing in the opposite direction. The former was smaller than the latter (Fig. 1c) and a statistically significant difference was detected between the two values. 2.2. The inhibitory effect of neighbouring epithelia is not reproduced by local application of BMP4 In the scheme proposed by Hogan (1999), a key factor for the ‘generative’ aspect of branching morphogenesis is supposed to be the lateral inhibitory action of BMP4. To examine whether the morphological change described above can be reproduced by a BMP4 gradient, we applied BMP4-soaked bead near the cultured epithelia and observed its effect on branching morphogenesis. BMP-soaked beads reduced the size of the explants (Fig. 1d,e), and the size of the epithelia was significantly smaller than phosphatebuffered saline (PBS)-soaked beads (Fig. 1f). However, the inhibition of growth appeared not to be local and budding was not specifically inhibited near the bead. The number of the buds which grew toward the beads and that of the buds growing in the opposite direction were not statistically different (Fig. 1g). Although we have tried various BMP4 concentrations (0–100 mg/ml, n ¼ 72), we could not detect any tendency indicating that budding was suppressed near the bead. Moreover, a time-lapse observation of the lung bud epithelia cultured with a BMP4-soaked bead showed no local effect of a BMP4-soaked beads on budding (Fig. 1h). Therefore, BMP4 was shown to inhibit growth of lung bud epithelial cells, but we could not detect any local effect in our culture condition.
T. Miura, K. Shiota / Mechanisms of Development 116 (2002) 29–38
31
2.3. Depletion of FGF mediates the lateral inhibition by the neighbouring epithelia
Fig. 1. Lateral inhibition caused by the neighbouring epithelium and its relationship with BMP4. (a) Two lung epithelia cultured in close vicinity. Before the culture. (b) Two lung epithelia cultured for 24 h in the medium supplemented with 100 ng/ml FGF1. Budding was suppressed in the area between the two explants and they usually did not fuse with each other. (c) Morphometrical analysis of the direction of bud formation. Lung epithelia were cultured in close vicinity and the number of the buds which grew toward the other epithelial explant and the number of those growing in the opposite direction were counted (n ¼ 35). More buds were formed in the latter direction, and a statistically significant difference was detected between the two values by Student’s t-test (P , 0:01). (d) A lung epithelium cultured for 48 h near a control bead soaked with PBS. (e) A lung epithelium cultured for 48 h near a bead soaked in 100 mg/ml BMP4. The growth was significantly inhibited but no directional growth was observed as detected in (b). (f) The growth of the epithelium was inhibited by BMP4soaked beads. The relative growth of epithelia was shown by the area of the epithelial explant in captured images. A statistically significant difference was observed between the control and BMP-treated groups (n ¼ 6 for each point). (g) A morphometrical analysis of the direction of bud formation. The number of the buds which grew toward the other epithelial explant and that of the buds growing to the opposite direction were compared (n ¼ 6 for each group). No statistically significant difference was observed both in the control and BMP4-treated cultures. (h) Contour plots showing the growth of the epithelium near the BMP4-soaked bead. The contour line of the same cultured epithelium was plotted every 2 h. Specific growth arrest was not observed near the BMP4-soaked bead. Scale bars, 100 mm.
One possible reason for lack of the local inhibitory effect of BMP4 is insufficient sensitivity of our assay system, but another possibility is the existence of some other inhibitory factors that works with BMP4. Theoretical studies have shown that depletion of a positive regulatory factor can also act as a lateral inhibitory factor (Meinhardt, 1976; Meinhardt, 1995). Therefore, we hypothesized that depletion of FGF1 can act as a lateral inhibitory factor. To test this, we observed the distribution of FGF1 protein in the culture. First, we added fluorescent dye-conjugated FGF1 in the culture medium and observed the distribution during the culture period. After 24 h of culture, a region of weaker fluorescence than other regions was observed around the cultured epithelia and strong fluorescence was observed inside the epithelium (Fig. 2a,b), indicating that FGF was taken up inside the lung epithelial cells as observed in other cell cultures (Citores et al., 2001; Citores et al., 1999). As a negative control, we added an unconjugated fluorescent dye in the culture medium and observed its distribution in the culture (Fig. 2c,d). No specific distribution was observed around lung epithelial explant, suggesting that the result described in Fig. 2b was not an artefact. An immunohistochemical study also showed that region of low FGF1 concentration was found in the Matrigel around cultured lung explant and strong signal was detected inside the epithelia (Fig. 2e). To examine whether the depletion of FGF1 plays a role in the lateral inhibitory mechanism in this culture system, we cultured two epithelia in close vicinity under a high FGF1 concentration and observed their budding pattern. If the lateral inhibition is induced by some factors other than FGF, the suppression of budding in the area between the two epithelia should not be influenced by the concentration of FGF1. On the contrary, if depletion of FGF1 causes a lateral inhibition, a saturating amount of FGF1 should suppress the inhibitory effect on budding between the two explants. Our result showed that under a high FGF1 concentration (500 ng/ml), the inhibitory effect was suppressed and the two explants fused with each other (Fig. 2f,g). This was quantitatively confirmed by the fact that the number of fused samples significantly increased under high FGF concentrations (Fig. 2h). Although we cannot rule out the contribution of other unknown inhibitory signals to bud formation, our experimental data strongly imply that depletion of FGF exerts a lateral inhibitory effect on budding of the lung epithelium in vitro. 2.4. A reaction–diffusion model of the substrate-depletion type can reproduce the formation of branched structure and lateral inhibitory effect From the experiments described above, we inferred that depletion of FGF can act as a lateral inhibitory factor, and
32
T. Miura, K. Shiota / Mechanisms of Development 116 (2002) 29–38
therefore explored whether a reaction–diffusion model of the substrate-depletion type (Kawasaki et al., 1997; Matsushita et al., 1998; Meinhardt, 1976; Meinhardt, 1995) could describe branching morphogenesis in vitro. First, we made the following assumptions concerning the
Fig. 2. Depletion of FGF1 protein acts as a lateral inhibitory factor in the mesenchyme-free culture system. (a) Phase-contrast image of the cultured lung epithelium with fluorescent dye-conjugated FGF1. (b) Fluorescent image of the same epithelia. The area of lower fluorescent intensity was observed around the epithelia (arrowheads) and high intensity areas were observed inside the lung epithelia (arrow). (c) Phase-contrast image of the cultured lung epithelium with fluorescent dye and FGF1. (d) Fluorescent image of (c). No specific distribution was observed around epithelia, indicating images obtained in (b) was not an artefact caused by autofluorescence of surrounding gel. (e) Immunohistochemistry against FGF1 of mesenchyme-free culture. The staining intensity near the cultured epithelia (arrowhead) was lower than the surrounding Matrigel (arrow). (f) Budding of two lung epithelia cultured in close vicinity in the medium supplemented with 500 ng/ml of FGF1. At the beginning of culture. (g) After 24 h of culture. The lung epithelia frequently fused with each other. (h) A statistical analysis of the rates of fusion between two epithelia cultured for 48 h. Under the FGF-rich condition (500 ng/ml), the epithelia fused with each other in most cases, while under a low FGF1 condition (100 ng/ml), the two explants did not usually fuse with each other. A statistically significant difference was detected between the two groups by Chi-square test (P , 0:05). Scale bars, 100 mm.
interaction of epithelial cells and FGF1: 1. FGF1 is consumed by epithelial cells. 2. Epithelial cell growth depends on the concentration of FGF1. 3. FGF1 molecules diffuse inside the Matrigel. 4. Each cell keeps a given cell size, and therefore the cell density in epithelial explant tends to remain constant. Some experimental evidence supports these assumptions. Assumption 1 came from our finding that the FGF1 protein concentration was lower around the cultured epithelial tissue (Fig. 2a–e) as well as from the fact that FGF is taken up into cells (Citores et al., 2001; Citores et al., 1999). For assumption 2, we confirmed that growth of the lung epithelial explant depends on the FGF concentration (data not shown). Assumption 3 was experimentally confirmed by the gradient of FGF inside Matrigel (Figs. 2b and 5c). It has been reported that the cell density within the lung epithelium is not very different between the tip and cleft regions (Miura and Shiota, 2000), which supports assumption 4. Complex events happen at the interface between growing lung epithelia and surrounding Matrigel such as extracellular matrix (ECM) degradation and chemotaxis, but these events do not affect the result as long as the growth of lung explant increases according to the amount of FGF. Based on these assumptions, we made a discrete model and undertook a computer simulation (the details of this model are described in Appendix A) and confirmed that the branched structure was produced spontaneously by this model, which was similar to the pattern observed in actual experiments (Fig. 3b,c). We also confirmed that the lateral inhibitory effect of neighbouring epithelia and its suppression by saturating concentration of FGF1 can be reproduced by the simulation based on our model (Fig. 3d,e). The similarity of the results of experiments and simulations strongly suggests that the model reproduces the mechanism actually working in vitro. The qualitative explanations for these results are described in Section 3. 2.5. Effects of the FGF1 concentration on lung bud morphology can be predicted and explained by the reaction–diffusion model To validate the model for applying to other experimental conditions, we examined whether the initial concentration of FGF1 affects the resulting branch pattern. The morphological pattern of the epithelial explant was altered by different FGF concentrations in the same way as predicted by the simulation (Fig. 4a–f). Under a low FGF concentration, the epithelial explant simply did not grow (Fig. 4a,b). Under an intermediate FGF1 condition (500 ng/ml), several buds with a similar diameter were formed (Fig. 4c,d) whereas under a high FGF1 concentration (1000 ng/ml), the shape of the lung epithelium was more like a balloon with some
T. Miura, K. Shiota / Mechanisms of Development 116 (2002) 29–38
Fig. 3. Description of the model. (a) The model for lung branching morphogenesis. cðx; y; tÞ represents the epithelial cell density and nðx; y; tÞ the FGF1 concentration in a given grid. f ðc; nÞ is an increasing function for both c and n. In this model, cells are supposed to proliferate according to the FGF1 concentration and move passively, and FGF1 is supposed to be consumed by the cells. (b) The initial condition of the cell density cðx; y; 0Þ. The black area indicates the high cell density area. (c) Distribution of the cell density generated by numerical simulation. The pattern resembles that observed in actual experiments. (d) A computer simulation result of budding of two lung epithelia placed in close vicinity. Budding is suppressed between the two epithelia, which resembles the pattern observed in actual experiments (Fig. 1b). (e) A computer simulation result of the budding of two epithelia under a high FGF1 concentration. The final pattern is similar to that observed experimentally (Fig. 2h).
randomly located clefts (Fig. 4e,f). The variance in the bud diameter was larger in the FGF1-rich condition both in the experimental data and in computer simulation (Fig. 4g,h). The similarity of experimental and simulation results strongly suggests that the reaction–diffusion model properly represents the mechanism of pattern formation in the experimental system. 2.6. Effects of the degradation of extracellular matrices on the lung bud pattern can be explained by the reaction– diffusion model We examined whether the model accounts for the results of the classical experiments which dealt with ECMs.
33
Fig. 4. Effects of the FGF1 concentration on the pattern of the lung epithelium. (a) A lung epithelium cultured under a low FGF1 concentration (100 ng/ml) for 48 h. The epithelial explant did not grow properly and no budding could be observed. (b) A computer simulation result of the epithelial growth under a low initial FGF1 concentration. The pattern resembles that observed in experiments (a). (c) A lung epithelium cultured under an intermediate FGF1 concentration (500 ng/ml) for 48 h. Bud formation was more evident and the buds had similar diameters. (d) A computer simulation result of the epithelial growth under an intermediate initial FGF1 concentration. The shape resembles that observed in experiments (c). (e) A lung epithelium cultured under a high FGF1 concentration (1000 ng/ml) for 48 h. The epithelium became a balloon-like shape with randomly located clefts. (f) A computer simulation result of the epithelial growth under a high initial FGF1 concentration. The pattern resembles that observed in experiments (e). (g) A morphometrical analysis of the bud diameter obtained by actual experimental result. The diameter of buds was measured manually and its distribution was shown in a histogram. The diameter of buds was more variable when the epithelia were cultured under the higher FGF1 concentration. A statistically significant difference was observed in the variance of the bud diameter distribution by F-test (P , 0:05) . (h) Morphometrical analysis of the bud diameter obtained by computer simulation. The diameter of buds was measured manually and its distribution was shown in a histogram. The diameter of buds was more variable in the simulation for the higher initial FGF1 concentration. A statistically significant difference was observed in the variance of the bud diameter distribution by F-test (P , 0:05). Scale bars, 100 mm.
34
T. Miura, K. Shiota / Mechanisms of Development 116 (2002) 29–38
Previous studies have shown that the degradation of ECM results in inhibition of branching morphogenesis in various experimental systems (Ganser et al., 1991; Gilbert, 1997), most of which were done in organ culture of the lung bud with mesenchymal tissue. Our experiments in mesenchymefree culture reproduced their results, and ECM degradation by collagenase caused cystic growth of the mesenchymefree epithelium (Fig. 5a,b). Next, we examined whether similar patterns could be generated by computer simulation if a plausible parameter of the model was changed. We assumed that the diffusion
coefficient of FGF1 (d2) should be increased when the surrounding Matrigel was degraded. To experimentally verify this assumption, we implanted a heparin bead soaked with fluorescent dye-labelled FGF1 in Matrigel and examined the distribution of FGF around the bead after 48 h in culture with and without collagenase I. Fluorescent dyeconjugated FGF1 could be detected around control beads but not so around beads in collagenase-treated gel (Fig. 5c,d), indicating that the surrounding Matrigel lost the ability to retain FGF molecules and resulted in an increased diffusion coefficient. Computer simulations with increased diffusion coefficients reproduced the pattern changes which were actually observed in the in vitro experiment (Fig. 5e,f). Then we examined whether the model could explain the cystic pattern induced in the epithelia cultivated with FGF7 (Fig. 5g,h). FGF7 has been shown to induce cystic patterns instead of branches of the lung epithelium (Cardoso et al., 1997) . Such a morphological difference has been attributed to the localization of FGF receptors (Cardoso et al., 1997), but this explanation does not answer the question why the initial distribution of FGF receptors is not homogeneous. Based on the observation described above, we hypothesized that FGF7 promotes the ECM-degrading activity of the epithelium because the pattern change induced by FGF7 is similar to those produced by collagenase I treatment. Gelatin enzymography revealed that FGF7 increased the gelatinolytic activity around 83 kDa (Fig. 5i) indicating that an activated form of matrix metalloproteinase 9 (MMP9) was upregulated by FGF7. Blocking the MMP activity by ilomastat caused budding even in the presence of the FGF7 (Fig. 5j), which suggests that the cystic pattern induced in the culture with FGF7 may have been produced via MMPs. MMP9 has been shown to be expressed in the developing lung epithelium (Canete-Soler et al., 1995; Fukuda et al., 2000), and is known to degrade the basement membrane. Since the composition of the Matrigel is similar to that of the basement membrane, MMP9 induced by FGF7 should have
Fig. 5. Effects of ECM degradation and FGF7 treatment on the pattern of the lung epithelium cultured in vitro. (a) A lung epithelium cultured for 24 h in the medium without collagenase. Budding morphogenesis was evident. (b) A lung epithelium cultured with 2 mg/ml collagenase I for 24 h. The epithelium became cyst-like. (c) Beads soaked with fluorescent-labelled FGF1 implanted in Matrigel in the absence of collagenase I for 48 h. Labelled FGF remains around beads. (d) The same beads implanted in Matrigel in the presence of collagenase I for 48 h. Labelled FGF could not be observed around beads. (e) A computer simulation result for a low FGF1 diffusion coefficient (d2). Budding morphogenesis is evident. (f) A computer simulation result for a high FGF1 diffusion coefficient (d2). The resulting pattern is more spherical than the simulation result under the low d2 (c) and looks more like the actual experimental result (b). (g) A lung epithelium cultured for 48 h with FGF1. Several buds were formed. (h) A lung epithelium cultured for 48 h with FGF7. The shape was cyst-like. (i) Gelatin enzymography. A strong gelatinolytic activity was detected at 83 kDa in the FGF7-supplemented culture medium, indicating that FGF7 induced an activated form of MMP9. (j) A lung epithelium cultured for 48 h with FGF7 and an MMP inhibitor ilomastat. Several buds were formed even at the presence of FGF7. Scale bars, 100 mm.
T. Miura, K. Shiota / Mechanisms of Development 116 (2002) 29–38
35
degraded the surrounding Matrigel and changed the shape of the lung epithelial explant cystic as predicted by simulation.
is that the depletion of FGF acts directly as a lateral inhibitory factor.
3. Discussion
3.2. Relationship between in vitro and in vivo pattern formation
3.1. Mechanism of pattern formation in the reaction– diffusion model Although the results of the computer simulations might seem counterintuitive, each result has simple qualitative explanation. The lateral inhibition caused by the neighbouring epithelia (Figs. 1b and 3d) can be explained by the model as follows: in the area between two explants, the FGF1 concentration is lower than the other areas because FGF1 is consumed by both of the opposing epithelia, resulting in the suppression of budding. The suppression does not occur if a saturating amount of FGF1 was added inside culture medium (Figs. 2g and 3e) because there is enough FGF1 in the area between two explants even if FGF1 is consumed by both of the epithelia. Spontaneous branching morphogenesis reproduced by the model in Fig. 3c can be explained as follows: the initial form of epithelial explants is not ideally spherical, and epithelial cells at slightly protruded regions should be exposed to higher concentrations of FGF and grow more rapidly than those at the other regions because protruded regions are distant from other FGF-consuming cells. When the protruded regions grow further, they become more distant from other regions and should be exposed to more FGF in the gel. This mechanism should result in the ‘protrusion grows faster’ tendency, and such positive feedback makes the subtle fluctuations in the initial shape of the explant grow further and produce a branched pattern (Ball, 1999). Moreover, depletion of FGF around the protruded regions should inhibit the formation of new buds near the preexisting buds, and this lateral inhibition mechanism should make the diameter of the buds rather constant. We can also explain the pattern change induced by the change in initial concentration of FGF (Fig. 4) as follows: in the reaction–diffusion model, the diameter of buds is determined by lateral inhibition caused by the depletion of diffusible substances (FGF1 in this case). Therefore, under high FGF1 concentrations, the initial fluctuation of in the epithelial shape should become dominant and the location of clefts should become random, while under low FGF1 concentrations, the effect of lateral inhibition by the consumption of FGF1 should become dominant and the diameter of buds should become rather constant (Dr A. Mochizuki, personal communication). Since currently proposed candidates for lateral inhibitory factors such as BMP4 and sprouty2 are upregulated by FGF (Mailleux et al., 2001; Weaver et al., 2000), it is difficult to explain the observed pattern change by the secondary induction of these factors by FGF because the resulting pattern implies a loss of the lateral inhibitory mechanism. Therefore, the simplest explanation in this case
We have shown that the reaction–diffusion model of substrate-depletion type can explain the pattern formation in mesenchyme-free condition, but we should be careful about to what extent this model can be applied to in vivo situation. Some correlation exists between in vitro and in vivo patterns because basement membrane components and FGFs exist both in in vitro and in vivo conditions (Ohtsuka et al., 2001) and collagenase and FGF7 treatments produce similar patterns in cultures with and without mesenchyme (Cardoso et al., 1997; Ganser et al., 1991). Moreover, recently it was shown that mouse lacking TIMP-3, which is able to inhibit MMP-1, -2, -3 and -9, shows spontaneous air space enlargement (Leco et al., 2001). This report also strengthens the possibility that a similar mechanism is working both in vitro and in vivo. However, other reports have suggested that the mechanism in vivo seems to be more complicated than that in vitro. For example, several subtypes of FGF are expressed during lung development and they seem to play different roles. The FGF10 molecule was shown to be expressed in the mesenchyme surrounding the tip of the growing epithelium at GD 11.5, while FGF1 and FGF7 distributed ubiquitously in the mesenchyme at later stages of development (Bellusci et al., 1997b), and these molecules differently regulate the expression of other genes (Lebeche et al., 1999). Moreover, we could not rule out the possibility that BMP4 acts as a lateral inhibitory factor via mesenchyme. Further studies are needed to elucidate to what extent the in vitro situation can be extrapolated to that in vivo. The difference between in vitro and in vivo conditions raises a number of interesting questions. For example, it needs to be investigated whether the existence of mesenchyme influences the pattern change observed in mesenchyme-free condition. It is possible that the lateral inhibitory mechanism is different between in vitro and in vivo, and the application of BMP4 or a saturated amount of FGF1 under existence of mesenchyme may cause quite different morphological changes as compared with those occurring under mesenchyme-free conditions. A method to directly assay the response of the mesenchymal tissue on various FGF molecules has been established (Lebeche et al., 1999), so the effect of mesenchymal tissue could be incorporated in the model in future. Moreover, the relationship between the ‘hardwired’ and ‘generative’ aspects of morphogenesis should be examined. Although the bifurcated pattern of the bronchial tree is strictly determined in vivo, the branching pattern does not seem to follow a specific rule under mesenchyme-free conditions and the identity of each bud is difficult to recognize. In a recent report, Hoxb gene cluster was shown to be expressed in the lung bud
36
T. Miura, K. Shiota / Mechanisms of Development 116 (2002) 29–38
mesenchyme, suggesting that the involvement of Hox genes in regionalization of the bronchial tree (Sakiyama et al., 2000). Since mesenchyme possibly contains information for determining the identity of each bud, only the generative aspect of branching morphogenesis should become apparent in mesenchyme-free conditions. The relationship needs to be clarified between the signals specifying the identity of each branch and those which produce a branched structure in a generative way. 3.3. The cystic and branched patterns in the bronchial tree can be explained by a single mechanism The model can explain the formation of both cystic and branched patterns, which is useful to explain cystic transformation of airway seen in various animals. For example, many fishes have swimbladder instead of lung, which apparently lacks branched structure (Wake, 1979). In the chick lung, the bronchial airway consists of two kinds of tissues, i.e. the cyst-like air sac and the bifurcated bronchial airway as seen in mice and humans. The sites of Hoxb6 expression in the chick lung primordium coincide with the that of air sacs, and a transplantation experiment revealed that the Hoxb6-expressing mesenchyme has a capacity to induce a cyst-like structure (Sakiyama et al., 2000). The model can explain the formation of both cystic and branched structures by a single mechanism and can predict which factors affect the branch pattern. It should be interesting to see whether Hoxb6 upregulates ECM-degrading enzymes or FGF7 which contribute to the formation of the cystic epithelial structure in vitro. 3.4. Possible application of the model to branching morphogenesis in other organs It remains to be determined whether the model described above can be applied to branching morphogenesis in other organs such as the kidney, salivary gland, prostate gland and pancreas. Since their duct systems are endodermal in origin and form a tree-like structure, it is likely that a similar mechanism works during morphogenesis of these organs. Several studies showed that these organs also utilize FGF during morphogenesis (Miralles et al., 1999; Morita and Nogawa, 1999; Thomson, 2001), therefore the model could contribute to the elucidation of the branching mechanism in these embryonic organs. For example, there is a clinically important disease called polycystic kidney, in which mutation in a membrane glycoprotein causes numerous cyst formations in the kidney (Peters and Breuning, 2001). The application of the model to the branching morphogenesis of kidney may elucidate the mechanism of cyst formation and can be useful in predicting possible modifying factor of the progression of the disease. Further interaction between experimental and theoretical studies is definitely needed to elucidate the generative aspects of branching morphogenesis in developing organs.
4. Experimental procedures 4.1. Lung epithelium culture Lung epithelium culture was carried out as described previously (Cardoso et al., 1997; Miura and Shiota, 2000; Nogawa and Ito, 1995). Gestational day 11.5 mouse embryos (plug day ¼ day 0) were removed from the uterus, and their lungs were aseptically collected and treated with 1 unit/ml dispase (Invitrogen, Corp., Carlsbad, California, USA) for 20 min at 378C. Then the epithelia of the lung were separated from the surrounding mesenchyme using sharp tungsten needles. Isolated epithelia were injected into a 20 ml drop Matrigel/Dulbecco’s modified Eagle medium (DMEM)/F-12 placed on a tissue culture insert (Nalge Nunc International, Naperville, IL, USA) or a four-well dish (10 mm in diameter) and incubated for 30 min at 378C in a humidified atmosphere. Then 400 ml of the culture medium (DMEM/F-12 1 0.1% bovine serum albumin (BSA)) containing various concentrations of FGF1 or FGF7 (Peprotech, Rocky Hill, NJ) was added and the explants were incubated for 24–48 h in a humidified atmosphere. In some cases, the culture was treated with 2 mg/ml collagenase I (Sigma, St Louis, MO) or 0.5 mM ilomastat (Carbiochem, San Diego, CA). 4.2. Time-lapse recording of lung epithelia cultured with BMP4 beads Time-lapse observation of explanted lung epithelia was carried out as described previously (Miura and Shiota, 2000). Lung epithelia were prepared as described above, and BMP4-soaked beads were implanted near the cultured epithelia. The culture dish was then placed on an inverted microscope (Nikon TMD) covered with an incubation chamber. The atmosphere in the chamber was kept at 378C and the chamber was filled with air containing 5% CO2 to maintain the culture medium around pH 7.0. The images of the lung explants were captured with a chargecoupled device (CCD) camera (Hamamatsu Photonics C2400-77) every 20 min, and stored in a magneto-optical disk. All procedures were controlled using a Macintosh Quadra 700 computer and the NIH Image program (developed at US National Institutes of Health and available on the Internet at http://rsb.info.nih.gov/nih-image). 4.3. Immunohistochemistry against FGF1 Cultured lung epithelia were fixed overnight with 4% paraformaldehyde. Then the samples were embedded in paraffin, sectioned and dewaxed. After blocked by 1.5% normal donkey serum in PBS for 1 h, an anti-FGF1 antibody (Santa Cruz, Santa Cruz, CA) was applied to the samples and incubated at 48C overnight. For negative controls, PBS was used instead of primary antibody. The samples were washed three times with PBS for 5 min and treated for 1 h with the biotin-conjugated secondary antibody at 378C.
T. Miura, K. Shiota / Mechanisms of Development 116 (2002) 29–38
Then the samples were washed three times with PBS for 5 min and the distribution of FGF1 was visualized using the Elite ABC kit (Vectastain, Burlingame, CA, USA). 4.4. Preparation of fluorescent dye-labelled FGF1 Fluorescent dye-labelled FGF1 was prepared as described previously (Citores et al., 1999). Five micrograms of FGF1 protein (Peprotech, Rocky Hill, NJ) diluted in 50 ml of 0.1 M sodium carbonate–sodium bicarbonate buffer was incubated with Cy3 dye (Amersham Pharmacia, Uppsala, Sweden) for 30 min at room temperature. Then the remaining dye was removed by washing with PBS two times using Microcon YM-10 (Millipore, Bedford, MA). For negative controls, PBS was used instead of FGF1 protein solution. For FGF1-soaked beads, 50 ml of Heparin beads solution (Sigma, St. Louis, MO) was added in the Cy3–FGF1 solution and incubated overnight at 48C. The beads were washed three times with PBS and used as a source of FGF. 4.5. Gelatin zymography The medium for the lung epithelium culture was collected after 48 h and concentrated approximately ten-fold using Microcon YM-10 (Millipore, Bedford, MA). The samples were mixed with a 2 £ sample buffer without reducing reagents and loaded onto a 10% polyacrylamide gel containing 0.75 mg/ml bovine skin gelatin. After electrophoresis, the gel was washed twice with 2.5% Triton X-100 and incubated overnight with a substrate buffer (5 mM CaCl2, 50 mM Tris–HCl pH 8.0) (Ganser et al., 1991). The gel was stained with Commassie blue and destained with methanol/acetic acid. The image of the gel was captured using a flatbed scanner and the band was visualized using the NIH-Image software. Acknowledgements The authors especially thank Professor Stuart Newman for critical reading of the manuscript and helpful comments. Helpful discussions with Professor Masayasu Mimura, Professor Hisao Honda, Professor Mitsugu Matsushita, Dr Atsushi Mochizuki, Dr Mikiko Kobayashi-Miura and technical assistance by Dr Hiroyuki Nogawa and Ms Takako Shibayama are also gratefully acknowledged. This work was sponsored by the Japanese Ministry of Education, Culture, Sports, Science and Technology and the Japanese Ministry of Health, Labour and Welfare. Appendix A. Theoretical model of pattern formation in vitro The theoretical model we used is a modified form of a reaction–diffusion model of the substrate-depletion type (Meinhardt, 1995), which was originally proposed by Meinhardt (1976) and has been utilized to explain the pattern
37
formation in bacterial colonies (Kawasaki et al., 1997; Matsushita et al., 1998). We used a discrete model instead of a continuous model because discrete model can deal with discontinuous distribution of cell density. Unlike bacterial colonies, the spatial distribution of cell density is apparently discontinuous at the interface between the epithelial explant and the surrounding Matrigel and it is difficult to prove that Laplacian can be always defined at this point if we use continuous model (Dr A. Mochizuki, personal communication). At first we divided the whole culture area into small grids of the same size and defined the epithelial cell density and FGF1 concentration of each grid at time t as cðx; y; tÞ and nðx; y; tÞ (Fig. 3a). Then we defined the changes in the epithelial cell density and FGF1 concentration after a short length of time as described in the following equations, which were based on the assumptions in the Section 2.4: cðx; y; t 1 1Þ 2 cðx; y; tÞ ¼ pf ½cðx; y; tÞ; nðx; y; tÞ 1 d1 JðcÞ nðx; y; t 1 1Þ 2 nðx; y; tÞ ¼ 2f ½cðx; y; tÞ; nðx; y; tÞ 1 d2 DðnÞ where f ðc; nÞ represents the consumption ratio of FGF1 by epithelial cells and is an increasing function for both c and n. We included saturation kinetics because cell proliferation appeared to be saturated at high FGF1 concentrations. pf ðc; nÞ is the growth rate of the cell where p is the conversion rate of consumed FGF1 to epithelial growth. d1 and d2 are the diffusion coefficients for cells and FGF1, respectively. D represents diffusion of the FGF molecule and we defined the function as:
DðnÞ ¼ nðx 1 1; y; tÞ 1 nðx 2 1; y; tÞ 1nðx; y 1 1; tÞ 1 nðx; y 2 1; tÞ 2 4nðx; y; tÞ where J is the passive movement of epithelial cells. Simple Laplacian is not suitable for describing the movement of epithelial cells because they adhere tightly to each other and do not diffuse unlimitedly. We defined this function as: JðcÞ ¼ D½Fðc 2 c 0 Þ
½FðxÞ ¼ 0ðx , 0Þ; xðx $ 0Þ
A qualitative explanation of the function J(c) is as follows: if the cell density in a given grid is lower than a certain value c0, the cells do not cover the whole grid and should not affect cells in the neighbouring grids. However, if the cell density becomes higher than c0 by proliferation, the cells should repulse each other and push toward the neighbouring grids, which results in diffusion-like dynamics of the cell density. This function stabilizes the movement of cells and limits the variation of patterns, which were obtained by original reaction–diffusion model (Matsushita et al., 1998). Their model used simple Laplacian for cell movement and generate complex patterns such as concentric ring, which are unlikely to be reproduced in lung epithelia. We defined the initial cell density c(x,y,0) as c ¼ c 0 in a certain area and c ¼ 0 in other areas (Fig. 3b) and the initial FGF1 concentration as nðx; y; 0Þ ¼ n0 (constant
38
T. Miura, K. Shiota / Mechanisms of Development 116 (2002) 29–38
value). Numerical calculation of the model was carried out using Macintosh G4/350 with Metrowerks CodeWarrior. The simulation parameters are as follows: f(c,n) ¼ 5cn/ (1 1 3\ n), p ¼ 1:0, c0 ¼ 1:0, d1 ¼ 4:0, d2 ¼ 3:4–27:0, n0 ¼ 0:5–0:8, grid number ¼ 200 £ 200. References Ball, P., 1999. The Self-Made Tapestry, Oxford University Press, Oxford. Bellusci, S., Furuta, Y., Rush, M.G., Henderson, R., Winnier, G., Hogan, B.L.M., 1997a. Involvement of Sonic Hedgehog (Shh) in mouse embryonic lung growth and morphogenesis. Development 124, 53–63. Bellusci, S., Grindley, J., Emoto, H., Itoh, N., Hogan, B.L., 1997b. Fibroblast growth factor 10 (FGF10) and branching morphogenesis in the embryonic mouse lung. Development 124, 4867–4878. Bellusci, S., Henderson, R., Winnier, G., Oikawa, T., Hogan, B.L., 1996. Evidence from normal expression and targeted misexpression that bone morphogenetic protein (Bmp-4) plays a role in mouse embryonic lung morphogenesis. Development 122, 1693–1702. Canete-Soler, R., Gui, Y.H., Linask, K.K., Muschel, R.J., 1995. Developmental expression of MMP-9 (gelatinase B) mRNA in mouse embryos. Dev. Dyn. 204, 30–40. Cardoso, W.V., Itoh, A., Nogawa, H., Mason, I., Brody, J.S., 1997. FGF-1 and FGF-7 induce distinct patterns of growth and differentiation in embryonic lung epithelium. Dev. Dyn. 208, 398–405. Citores, L., Khnykin, D., Sorensen, V., Wesche, J., Klingenberg, O., Wiedlocha, A., Olsnes, S., 2001. Modulation of intracellular transport of acidic fibroblast growth factor by mutations in the cytoplasmic receptor domain. J. Cell Sci. 114, 1677–1689. Citores, L., Wesche, J., Kolpakova, E., Olsnes, S., 1999. Uptake and intracellular transport of acidic fibroblast growth factor: evidence for free and cytoskeleton-anchored fibroblast growth factor receptors. Mol. Biol. Cell 10, 3835–3848. Fukuda, Y., Ishizaki, M., Okada, Y., Seiki, M., Yamanaka, N., 2000. Matrix metalloproteinases and tissue inhibitor of metalloproteinase-2 in fetal rabbit lung. Am. J. Physiol. Lung Cell Mol. Physiol. 279, L555–L561. Ganser, G.L., Stricklin, G.P., Matrisian, L.M., 1991. EGF and TGF alpha influence in vitro lung development by the induction of matrix-degrading metalloproteinases. Int. J. Dev. Biol. 35, 453–461. Gilbert, S.F., 1997. Developmental Biology, Sinauer, Massachusettes, MA. Hogan, B.L., 1999. Morphogenesis. Cell 96, 225–233. Kawasaki, K., Mochizuki, A., Matsushita, M., Ueda, T., Shigesada, N., 1997. Modeling spatio-temporal patterns generated by Bacillus subtilis. J. Theor. Biol. 188, 177–185. Kitaoka, H., Takaki, R., Suki, B., 1999. A three-dimensional model of the human airway tree. J. Appl. Physiol. 87, 2207–2217. Lebeche, D., Malpel, S., Cardoso, W.V., 1999. Fibroblast growth factor interactions in the developing lung. Mech. Dev. 86, 125–136. Leco, K.J., Waterhouse, P., Sanchez, O.H., Gowing, K.L., Poole, A.R., Wakeham, A., Mak, T.W., Khokha, R., 2001. Spontaneous air space enlargement in the lungs of mice lacking tissue inhibitor of metalloproteinases-3 (TIMP-3). J. Clin. Invest. 108, 817–829. Lubkin, S.R., Murray, J.D., 1995. A mechanism for early branching in lung morphogenesis. J. Math. Biol. 34, 77–94.
Mailleux, A.A., Tefft, D., Ndiaye, D., Itoh, N., Thiery, J.P., Warburton, D., Bellusci, S., 2001. Evidence that SPROUTY2 functions as an inhibitor of mouse embryonic lung growth and morphogenesis. Mech. Dev. 102, 81–94. Matsushita, M., Wakita, J., Itoh, H., Rafols, I., Matsuyama, T., Sakaguchi, H., Mimura, M., 1998. Interface growth and pattern formation in bacterial colonies. Physica A 249, 517–524. Meinhardt, H., 1976. Morphogenesis of lines and nets. Differentiation 6, 117–123. Meinhardt, H., 1995. The Algorythmic Beauty of Sea Shells, Springer, Berlin. Miralles, F., Czernichow, P., Ozaki, K., Itoh, N., Scharfmann, R., 1999. Signaling through fibroblast growth factor receptor 2b plays a key role in the development of the exocrine pancreas. Proc. Natl Acad. Sci. USA 96, 6267–6272. Miura, T., Shiota, K., 2000. Time-lapse observation of branching morphogenesis of the lung bud epithelium in mesenchyme-free culture and its relationship with the localization of actin filaments. Int. J. Dev. Biol. 44, 899–902. Morita, K., Nogawa, H., 1999. EGF-dependent lobule formation and FGF7dependent stalk elongation in branching morphogenesis of mouse salivary epithelium in vitro. Dev. Dyn. 215, 148–154. Nogawa, H., Ito, T., 1995. Branching morphogenesis of embryonic mouse lung epithelium in mesenchyme-free culture. Development 121, 1015– 1022. Ohtsuka, N., Urase, K., Momoi, T., Nogawa, H., 2001. Induction of bud formation of embryonic mouse tracheal epithelium by fibroblast growth factor plus transferrin in mesenchyme-free culture. Dev. Dyn. 222, 263–272. Packer, A.I., Mailutha, K.G., Ambrozewicz, L.A., Wolgemuth, D.J., 2000. Regulation of the Hoxa4 and Hoxa5 genes in the embryonic mouse lung by retinoic acid and TGFbeta1: implications for lung development and patterning. Dev. Dyn. 217, 62–74. Park, W.Y., Miranda, B., Lebeche, D., Hashimoto, G., Cardoso, W.V., 1998. FGF-10 is a chemotactic factor for distal epithelial buds during lung development. Dev. Biol. 201, 125–134. Pepicelli, C.V., Lewis, P.M., McMahon, A.P., 1998. Sonic hedgehog regulates branching morphogenesis in the mammalian lung. Curr. Biol. 8, 1083–1086. Peters, D.J., Breuning, M.H., 2001. Autosomal dominant polycystic kidney disease: modification of disease progression. Lancet 358, 1439– 1444. Richardson, M.K., Allen, S.P., Wright, G.M., Raynaud, A., Hanken, J., 1998. Somite number and vertebrate evolution. Development 125, 151–160. Sakiyama, J., Yokouchi, Y., Kuroiwa, A., 2000. Coordinated expression of Hoxb genes and signaling molecules during development of the chick respiratory tract. Dev. Biol. 227, 12–27. Thomson, A.A., 2001. Role of androgens and fibroblast growth factors in prostatic development. Reproduction 121, 187–195. Wake, M.H., 1979. Hyman’s Comparative Vertebrate Anatomy, The University of Chicago Press, Chicago, IL pp. 768. Weaver, M., Dunn, N.R., Hogan, B.L., 2000. Bmp4 and Fgf10 play opposing roles during lung bud morphogenesis. Development 127, 2695– 2704.
Depletion of FGF acts as a lateral inhibitory factor in lung branching morphogenesis in vitro Takashi Miura a,*, Kohei Shiota a,b a
Department of Anatomy and Developmental Biology, Kyoto University Graduate School of Medicine, Kyoto 606-8501, Japan b Congenital Anomaly Research Center, Kyoto University Graduate School of Medicine, Kyoto 606-8501, Japan Received 31 January 2002; received in revised form 28 March 2002; accepted 11 April 2002
Abstract Previous studies have shown that the interaction of positive and inhibitory signals plays a crucial role during lung branching morphogenesis. We found that in mesenchyme-free conditions, the lung epithelium exerted a lateral inhibitory effect on the neighbouring epithelium via depletion of fibroblast growth factor 1 (FGF1). Contrary to previous suggestions, bone morphogenetic protein 4 could not substitute for the inhibitory effect. Based on of this observation, we used a reaction–diffusion model of the substrate-depletion type to represent the initial phase of in vitro branching morphogenesis of lung epithelium, with depletion of FGF playing the role of lateral inhibitor. The model was able to account for the effects of the FGF1 concentration, extracellular matrix degradation and different subtypes of FGF on morphogenesis of the lung bud epithelia. These results suggest that the depletion of FGF may be a key regulatory component in initial phase of branching morphogenesis of the lung bud epithelium in vitro. q 2002 Elsevier Science Ireland Ltd. All rights reserved. Keywords: Lung; Branching morphogenesis; Theoretical model; Fibroblast growth factor depletion
1. Introduction During mammalian development, the lung primordium first appears as an epithelial outgrowth from the ventral surface of the embryonic foregut and undergoes branching morphogenesis to form a bronchial tree (Gilbert, 1997). Its branching morphogenesis has been extensively studied experimentally in the field of developmental biology with the aid of organ culture systems, which reproduce branching morphogenesis in vitro. Some classical experiments revealed the importance of extracellular matrices in lung morphogenesis. For example, when collagenase was added to the culture medium, lung bud explants showed a cystic morphology instead of branches (Ganser et al., 1991). Recent studies with molecular techniques suggest that diffusible signaling molecules such as fibroblast growth factor 10 (FGF10), Sonic hedgehog (Shh) and bone morphogenetic protein 4 (BMP4) play key roles in branching morphogenesis (Bellusci et al., 1997a,b, 1996; Lebeche et al., 1999; Park et al., 1998; Pepicelli et al., 1998). Two schemes have been proposed to explain the lung branching morphogenesis (Hogan, 1999). One is the ‘hardwired’ scheme, which hypothesizes that a prepattern of gene * Corresponding author. Tel.: 181-75-753-4341; fax: 181-75-751-7529. E-mail address: [email protected] (T. Miura).
expression exists in the lung bud. Recently some supportive evidence for this scheme was provided by a chick experiment where the Hox gene cluster was shown to be unevenly expressed in the lung mesenchyme (Packer et al., 2000; Sakiyama et al., 2000). The other is the ‘generative’ scheme, in which the branching pattern is supposed to be determined dynamically by epithelial–mesenchymal interaction. In this scheme, the lateral inhibitory effect of BMP4 secreted from the distal tip of the lung bud endoderm is assumed to play a major role in determining the budding pattern (Hogan, 1999). These two schemes are not mutually exclusive, and recent studies on somite and limb development suggest that these two may act simultaneously (Richardson et al., 1998). From the theoretical point of view, several mathematical models have been proposed for branching morphogenesis in non-living systems such as crystal formation and viscous fingering, and the dynamics of these systems are well understood (Ball, 1999). Up to the present, experimental justification is lacking for these models to be applied to lung branching morphogenesis in the embryo. For example, although branching morphogenesis in the lung has been hypothesized to be homologous to viscous fingering (Lubkin and Murray, 1995), this bulk transport model does not take diffusible signals such as FGF (Bellusci et al., 1997b; Park et al., 1998) into account. Moreover, our recent finding on dynamic movement of the lung bud epithe-
0925-4773/02/$ - see front matter q 2002 Elsevier Science Ireland Ltd. All rights reserved. PII: S 0925-477 3(02)00132-6
30
T. Miura, K. Shiota / Mechanisms of Development 116 (2002) 29–38
lium is inconsistent with this hypothesis (Miura and Shiota, 2000). On the other hand, a physiology-based model for the human bronchial tree has been proposed (Kitaoka et al., 1999) in which the branch pattern of the lung is related to the efficiency of gas exchange. However, it is difficult to apply this model to lung morphogenesis in embryos because the branched structure is established well before respiration starts. Therefore, no theoretical models for lung branching morphogenesis proposed so far are adequately supported by experimental data. Both experimental schemes and theoretical models have their own strengths and limitations. Experimental schemes can include interaction of many molecules which can be tested by actual experiments and suitable to explain in vivo situation, but usually the scheme becomes too complex and it is difficult to understand the qualitative morphological change (such as transformation from branches to cysts) only by the schemes. On the other hand, theoretical models can explain a wide variety of morphologies by a single mechanism and helpful in understanding and predicting the qualitative morphological changes, but it usually contains only one or two factors and sometimes seems too simple to be applied to the in vivo situation. To utilize the advantages of both, we first chose the simplest in vitro experimental system which retains a key aspect of morphogenesis in vivo, and then applied a theoretical model to in vitro pattern formation, and finally discussed the relationship between in vitro and in vivo situations. It has been shown that isolated lung epithelia retain a capacity to undergo branching morphogenesis in vitro. Nogawa and Ito (1995) showed that when the lung epithelia were detached from the surrounding mesenchyme and cultured in Matrigel in the presence of FGF1, the epithelia underwent branching morphogenesis. The mesenchymefree culture is a common experimental system to study the mechanism of branching morphogenesis in vitro (Bellusci et al., 1997b; Park et al., 1998), and Cardoso et al. (1997) showed that FGF7 exerts a similar effect on the lung epithelial explant irrespective of the coexistence of mesenchyme. These studies strongly imply that the in vitro culture system reproduces the generative aspect of lung branching morphogenesis occurring in vivo. In the present study, we chose this mesenchyme-free culture as a simple experimental system of the earliest stage of the branching morphogenesis and examined the mechanism of morphogenesis in this culture system using both experimental and theoretical approaches. We found that in a mesenchyme-free condition, the isolated lung explant exerts a lateral inhibitory effect on the neighbouring lung epithelium. However, BMP4, which was assumed to be a lateral inhibitory factor in a previous experimental scheme (Hogan, 1999), could not reproduce this result in the mesenchyme-free culture system. Since the FGF protein concentration was lower around the cultured epithelia than in the other area of the gel, we proposed an alternative mechanism that epithelial cells consume FGF and the result-
ing depletion of FGF exerts a lateral inhibitory effect. We constructed a reaction–diffusion model of substrate-depletion type which could reproduce branching pattern through computer simulation, and showed that the model could explain the lung pattern changes observed under various experimental conditions. 2. Results 2.1. The neighbouring lung epithelia exert inhibitory effect in mesenchyme-free culture To examine whether the lateral inhibitory mechanism proposed by Hogan (1999) actually works in cultured embryonic lung epithelia, we cultured two epithelia in close vicinity and observed their budding pattern. If the site of bud formation is predetermined or intrinsically determined inside the lung epithelium, the location of buds should not be influenced by the nearby epithelium. In our experiment, budding of the epithelium was suppressed between the two epithelial explants (Fig. 1a,b), which means that the location of bud formation is not predetermined in the epithelium but is regulated by some extrinsic inhibitory factors in the extracellular environment. To confirm this, we counted the number of buds which grew toward the other epithelial explant and that of the buds growing in the opposite direction. The former was smaller than the latter (Fig. 1c) and a statistically significant difference was detected between the two values. 2.2. The inhibitory effect of neighbouring epithelia is not reproduced by local application of BMP4 In the scheme proposed by Hogan (1999), a key factor for the ‘generative’ aspect of branching morphogenesis is supposed to be the lateral inhibitory action of BMP4. To examine whether the morphological change described above can be reproduced by a BMP4 gradient, we applied BMP4-soaked bead near the cultured epithelia and observed its effect on branching morphogenesis. BMP-soaked beads reduced the size of the explants (Fig. 1d,e), and the size of the epithelia was significantly smaller than phosphatebuffered saline (PBS)-soaked beads (Fig. 1f). However, the inhibition of growth appeared not to be local and budding was not specifically inhibited near the bead. The number of the buds which grew toward the beads and that of the buds growing in the opposite direction were not statistically different (Fig. 1g). Although we have tried various BMP4 concentrations (0–100 mg/ml, n ¼ 72), we could not detect any tendency indicating that budding was suppressed near the bead. Moreover, a time-lapse observation of the lung bud epithelia cultured with a BMP4-soaked bead showed no local effect of a BMP4-soaked beads on budding (Fig. 1h). Therefore, BMP4 was shown to inhibit growth of lung bud epithelial cells, but we could not detect any local effect in our culture condition.
T. Miura, K. Shiota / Mechanisms of Development 116 (2002) 29–38
31
2.3. Depletion of FGF mediates the lateral inhibition by the neighbouring epithelia
Fig. 1. Lateral inhibition caused by the neighbouring epithelium and its relationship with BMP4. (a) Two lung epithelia cultured in close vicinity. Before the culture. (b) Two lung epithelia cultured for 24 h in the medium supplemented with 100 ng/ml FGF1. Budding was suppressed in the area between the two explants and they usually did not fuse with each other. (c) Morphometrical analysis of the direction of bud formation. Lung epithelia were cultured in close vicinity and the number of the buds which grew toward the other epithelial explant and the number of those growing in the opposite direction were counted (n ¼ 35). More buds were formed in the latter direction, and a statistically significant difference was detected between the two values by Student’s t-test (P , 0:01). (d) A lung epithelium cultured for 48 h near a control bead soaked with PBS. (e) A lung epithelium cultured for 48 h near a bead soaked in 100 mg/ml BMP4. The growth was significantly inhibited but no directional growth was observed as detected in (b). (f) The growth of the epithelium was inhibited by BMP4soaked beads. The relative growth of epithelia was shown by the area of the epithelial explant in captured images. A statistically significant difference was observed between the control and BMP-treated groups (n ¼ 6 for each point). (g) A morphometrical analysis of the direction of bud formation. The number of the buds which grew toward the other epithelial explant and that of the buds growing to the opposite direction were compared (n ¼ 6 for each group). No statistically significant difference was observed both in the control and BMP4-treated cultures. (h) Contour plots showing the growth of the epithelium near the BMP4-soaked bead. The contour line of the same cultured epithelium was plotted every 2 h. Specific growth arrest was not observed near the BMP4-soaked bead. Scale bars, 100 mm.
One possible reason for lack of the local inhibitory effect of BMP4 is insufficient sensitivity of our assay system, but another possibility is the existence of some other inhibitory factors that works with BMP4. Theoretical studies have shown that depletion of a positive regulatory factor can also act as a lateral inhibitory factor (Meinhardt, 1976; Meinhardt, 1995). Therefore, we hypothesized that depletion of FGF1 can act as a lateral inhibitory factor. To test this, we observed the distribution of FGF1 protein in the culture. First, we added fluorescent dye-conjugated FGF1 in the culture medium and observed the distribution during the culture period. After 24 h of culture, a region of weaker fluorescence than other regions was observed around the cultured epithelia and strong fluorescence was observed inside the epithelium (Fig. 2a,b), indicating that FGF was taken up inside the lung epithelial cells as observed in other cell cultures (Citores et al., 2001; Citores et al., 1999). As a negative control, we added an unconjugated fluorescent dye in the culture medium and observed its distribution in the culture (Fig. 2c,d). No specific distribution was observed around lung epithelial explant, suggesting that the result described in Fig. 2b was not an artefact. An immunohistochemical study also showed that region of low FGF1 concentration was found in the Matrigel around cultured lung explant and strong signal was detected inside the epithelia (Fig. 2e). To examine whether the depletion of FGF1 plays a role in the lateral inhibitory mechanism in this culture system, we cultured two epithelia in close vicinity under a high FGF1 concentration and observed their budding pattern. If the lateral inhibition is induced by some factors other than FGF, the suppression of budding in the area between the two epithelia should not be influenced by the concentration of FGF1. On the contrary, if depletion of FGF1 causes a lateral inhibition, a saturating amount of FGF1 should suppress the inhibitory effect on budding between the two explants. Our result showed that under a high FGF1 concentration (500 ng/ml), the inhibitory effect was suppressed and the two explants fused with each other (Fig. 2f,g). This was quantitatively confirmed by the fact that the number of fused samples significantly increased under high FGF concentrations (Fig. 2h). Although we cannot rule out the contribution of other unknown inhibitory signals to bud formation, our experimental data strongly imply that depletion of FGF exerts a lateral inhibitory effect on budding of the lung epithelium in vitro. 2.4. A reaction–diffusion model of the substrate-depletion type can reproduce the formation of branched structure and lateral inhibitory effect From the experiments described above, we inferred that depletion of FGF can act as a lateral inhibitory factor, and
32
T. Miura, K. Shiota / Mechanisms of Development 116 (2002) 29–38
therefore explored whether a reaction–diffusion model of the substrate-depletion type (Kawasaki et al., 1997; Matsushita et al., 1998; Meinhardt, 1976; Meinhardt, 1995) could describe branching morphogenesis in vitro. First, we made the following assumptions concerning the
Fig. 2. Depletion of FGF1 protein acts as a lateral inhibitory factor in the mesenchyme-free culture system. (a) Phase-contrast image of the cultured lung epithelium with fluorescent dye-conjugated FGF1. (b) Fluorescent image of the same epithelia. The area of lower fluorescent intensity was observed around the epithelia (arrowheads) and high intensity areas were observed inside the lung epithelia (arrow). (c) Phase-contrast image of the cultured lung epithelium with fluorescent dye and FGF1. (d) Fluorescent image of (c). No specific distribution was observed around epithelia, indicating images obtained in (b) was not an artefact caused by autofluorescence of surrounding gel. (e) Immunohistochemistry against FGF1 of mesenchyme-free culture. The staining intensity near the cultured epithelia (arrowhead) was lower than the surrounding Matrigel (arrow). (f) Budding of two lung epithelia cultured in close vicinity in the medium supplemented with 500 ng/ml of FGF1. At the beginning of culture. (g) After 24 h of culture. The lung epithelia frequently fused with each other. (h) A statistical analysis of the rates of fusion between two epithelia cultured for 48 h. Under the FGF-rich condition (500 ng/ml), the epithelia fused with each other in most cases, while under a low FGF1 condition (100 ng/ml), the two explants did not usually fuse with each other. A statistically significant difference was detected between the two groups by Chi-square test (P , 0:05). Scale bars, 100 mm.
interaction of epithelial cells and FGF1: 1. FGF1 is consumed by epithelial cells. 2. Epithelial cell growth depends on the concentration of FGF1. 3. FGF1 molecules diffuse inside the Matrigel. 4. Each cell keeps a given cell size, and therefore the cell density in epithelial explant tends to remain constant. Some experimental evidence supports these assumptions. Assumption 1 came from our finding that the FGF1 protein concentration was lower around the cultured epithelial tissue (Fig. 2a–e) as well as from the fact that FGF is taken up into cells (Citores et al., 2001; Citores et al., 1999). For assumption 2, we confirmed that growth of the lung epithelial explant depends on the FGF concentration (data not shown). Assumption 3 was experimentally confirmed by the gradient of FGF inside Matrigel (Figs. 2b and 5c). It has been reported that the cell density within the lung epithelium is not very different between the tip and cleft regions (Miura and Shiota, 2000), which supports assumption 4. Complex events happen at the interface between growing lung epithelia and surrounding Matrigel such as extracellular matrix (ECM) degradation and chemotaxis, but these events do not affect the result as long as the growth of lung explant increases according to the amount of FGF. Based on these assumptions, we made a discrete model and undertook a computer simulation (the details of this model are described in Appendix A) and confirmed that the branched structure was produced spontaneously by this model, which was similar to the pattern observed in actual experiments (Fig. 3b,c). We also confirmed that the lateral inhibitory effect of neighbouring epithelia and its suppression by saturating concentration of FGF1 can be reproduced by the simulation based on our model (Fig. 3d,e). The similarity of the results of experiments and simulations strongly suggests that the model reproduces the mechanism actually working in vitro. The qualitative explanations for these results are described in Section 3. 2.5. Effects of the FGF1 concentration on lung bud morphology can be predicted and explained by the reaction–diffusion model To validate the model for applying to other experimental conditions, we examined whether the initial concentration of FGF1 affects the resulting branch pattern. The morphological pattern of the epithelial explant was altered by different FGF concentrations in the same way as predicted by the simulation (Fig. 4a–f). Under a low FGF concentration, the epithelial explant simply did not grow (Fig. 4a,b). Under an intermediate FGF1 condition (500 ng/ml), several buds with a similar diameter were formed (Fig. 4c,d) whereas under a high FGF1 concentration (1000 ng/ml), the shape of the lung epithelium was more like a balloon with some
T. Miura, K. Shiota / Mechanisms of Development 116 (2002) 29–38
Fig. 3. Description of the model. (a) The model for lung branching morphogenesis. cðx; y; tÞ represents the epithelial cell density and nðx; y; tÞ the FGF1 concentration in a given grid. f ðc; nÞ is an increasing function for both c and n. In this model, cells are supposed to proliferate according to the FGF1 concentration and move passively, and FGF1 is supposed to be consumed by the cells. (b) The initial condition of the cell density cðx; y; 0Þ. The black area indicates the high cell density area. (c) Distribution of the cell density generated by numerical simulation. The pattern resembles that observed in actual experiments. (d) A computer simulation result of budding of two lung epithelia placed in close vicinity. Budding is suppressed between the two epithelia, which resembles the pattern observed in actual experiments (Fig. 1b). (e) A computer simulation result of the budding of two epithelia under a high FGF1 concentration. The final pattern is similar to that observed experimentally (Fig. 2h).
randomly located clefts (Fig. 4e,f). The variance in the bud diameter was larger in the FGF1-rich condition both in the experimental data and in computer simulation (Fig. 4g,h). The similarity of experimental and simulation results strongly suggests that the reaction–diffusion model properly represents the mechanism of pattern formation in the experimental system. 2.6. Effects of the degradation of extracellular matrices on the lung bud pattern can be explained by the reaction– diffusion model We examined whether the model accounts for the results of the classical experiments which dealt with ECMs.
33
Fig. 4. Effects of the FGF1 concentration on the pattern of the lung epithelium. (a) A lung epithelium cultured under a low FGF1 concentration (100 ng/ml) for 48 h. The epithelial explant did not grow properly and no budding could be observed. (b) A computer simulation result of the epithelial growth under a low initial FGF1 concentration. The pattern resembles that observed in experiments (a). (c) A lung epithelium cultured under an intermediate FGF1 concentration (500 ng/ml) for 48 h. Bud formation was more evident and the buds had similar diameters. (d) A computer simulation result of the epithelial growth under an intermediate initial FGF1 concentration. The shape resembles that observed in experiments (c). (e) A lung epithelium cultured under a high FGF1 concentration (1000 ng/ml) for 48 h. The epithelium became a balloon-like shape with randomly located clefts. (f) A computer simulation result of the epithelial growth under a high initial FGF1 concentration. The pattern resembles that observed in experiments (e). (g) A morphometrical analysis of the bud diameter obtained by actual experimental result. The diameter of buds was measured manually and its distribution was shown in a histogram. The diameter of buds was more variable when the epithelia were cultured under the higher FGF1 concentration. A statistically significant difference was observed in the variance of the bud diameter distribution by F-test (P , 0:05) . (h) Morphometrical analysis of the bud diameter obtained by computer simulation. The diameter of buds was measured manually and its distribution was shown in a histogram. The diameter of buds was more variable in the simulation for the higher initial FGF1 concentration. A statistically significant difference was observed in the variance of the bud diameter distribution by F-test (P , 0:05). Scale bars, 100 mm.
34
T. Miura, K. Shiota / Mechanisms of Development 116 (2002) 29–38
Previous studies have shown that the degradation of ECM results in inhibition of branching morphogenesis in various experimental systems (Ganser et al., 1991; Gilbert, 1997), most of which were done in organ culture of the lung bud with mesenchymal tissue. Our experiments in mesenchymefree culture reproduced their results, and ECM degradation by collagenase caused cystic growth of the mesenchymefree epithelium (Fig. 5a,b). Next, we examined whether similar patterns could be generated by computer simulation if a plausible parameter of the model was changed. We assumed that the diffusion
coefficient of FGF1 (d2) should be increased when the surrounding Matrigel was degraded. To experimentally verify this assumption, we implanted a heparin bead soaked with fluorescent dye-labelled FGF1 in Matrigel and examined the distribution of FGF around the bead after 48 h in culture with and without collagenase I. Fluorescent dyeconjugated FGF1 could be detected around control beads but not so around beads in collagenase-treated gel (Fig. 5c,d), indicating that the surrounding Matrigel lost the ability to retain FGF molecules and resulted in an increased diffusion coefficient. Computer simulations with increased diffusion coefficients reproduced the pattern changes which were actually observed in the in vitro experiment (Fig. 5e,f). Then we examined whether the model could explain the cystic pattern induced in the epithelia cultivated with FGF7 (Fig. 5g,h). FGF7 has been shown to induce cystic patterns instead of branches of the lung epithelium (Cardoso et al., 1997) . Such a morphological difference has been attributed to the localization of FGF receptors (Cardoso et al., 1997), but this explanation does not answer the question why the initial distribution of FGF receptors is not homogeneous. Based on the observation described above, we hypothesized that FGF7 promotes the ECM-degrading activity of the epithelium because the pattern change induced by FGF7 is similar to those produced by collagenase I treatment. Gelatin enzymography revealed that FGF7 increased the gelatinolytic activity around 83 kDa (Fig. 5i) indicating that an activated form of matrix metalloproteinase 9 (MMP9) was upregulated by FGF7. Blocking the MMP activity by ilomastat caused budding even in the presence of the FGF7 (Fig. 5j), which suggests that the cystic pattern induced in the culture with FGF7 may have been produced via MMPs. MMP9 has been shown to be expressed in the developing lung epithelium (Canete-Soler et al., 1995; Fukuda et al., 2000), and is known to degrade the basement membrane. Since the composition of the Matrigel is similar to that of the basement membrane, MMP9 induced by FGF7 should have
Fig. 5. Effects of ECM degradation and FGF7 treatment on the pattern of the lung epithelium cultured in vitro. (a) A lung epithelium cultured for 24 h in the medium without collagenase. Budding morphogenesis was evident. (b) A lung epithelium cultured with 2 mg/ml collagenase I for 24 h. The epithelium became cyst-like. (c) Beads soaked with fluorescent-labelled FGF1 implanted in Matrigel in the absence of collagenase I for 48 h. Labelled FGF remains around beads. (d) The same beads implanted in Matrigel in the presence of collagenase I for 48 h. Labelled FGF could not be observed around beads. (e) A computer simulation result for a low FGF1 diffusion coefficient (d2). Budding morphogenesis is evident. (f) A computer simulation result for a high FGF1 diffusion coefficient (d2). The resulting pattern is more spherical than the simulation result under the low d2 (c) and looks more like the actual experimental result (b). (g) A lung epithelium cultured for 48 h with FGF1. Several buds were formed. (h) A lung epithelium cultured for 48 h with FGF7. The shape was cyst-like. (i) Gelatin enzymography. A strong gelatinolytic activity was detected at 83 kDa in the FGF7-supplemented culture medium, indicating that FGF7 induced an activated form of MMP9. (j) A lung epithelium cultured for 48 h with FGF7 and an MMP inhibitor ilomastat. Several buds were formed even at the presence of FGF7. Scale bars, 100 mm.
T. Miura, K. Shiota / Mechanisms of Development 116 (2002) 29–38
35
degraded the surrounding Matrigel and changed the shape of the lung epithelial explant cystic as predicted by simulation.
is that the depletion of FGF acts directly as a lateral inhibitory factor.
3. Discussion
3.2. Relationship between in vitro and in vivo pattern formation
3.1. Mechanism of pattern formation in the reaction– diffusion model Although the results of the computer simulations might seem counterintuitive, each result has simple qualitative explanation. The lateral inhibition caused by the neighbouring epithelia (Figs. 1b and 3d) can be explained by the model as follows: in the area between two explants, the FGF1 concentration is lower than the other areas because FGF1 is consumed by both of the opposing epithelia, resulting in the suppression of budding. The suppression does not occur if a saturating amount of FGF1 was added inside culture medium (Figs. 2g and 3e) because there is enough FGF1 in the area between two explants even if FGF1 is consumed by both of the epithelia. Spontaneous branching morphogenesis reproduced by the model in Fig. 3c can be explained as follows: the initial form of epithelial explants is not ideally spherical, and epithelial cells at slightly protruded regions should be exposed to higher concentrations of FGF and grow more rapidly than those at the other regions because protruded regions are distant from other FGF-consuming cells. When the protruded regions grow further, they become more distant from other regions and should be exposed to more FGF in the gel. This mechanism should result in the ‘protrusion grows faster’ tendency, and such positive feedback makes the subtle fluctuations in the initial shape of the explant grow further and produce a branched pattern (Ball, 1999). Moreover, depletion of FGF around the protruded regions should inhibit the formation of new buds near the preexisting buds, and this lateral inhibition mechanism should make the diameter of the buds rather constant. We can also explain the pattern change induced by the change in initial concentration of FGF (Fig. 4) as follows: in the reaction–diffusion model, the diameter of buds is determined by lateral inhibition caused by the depletion of diffusible substances (FGF1 in this case). Therefore, under high FGF1 concentrations, the initial fluctuation of in the epithelial shape should become dominant and the location of clefts should become random, while under low FGF1 concentrations, the effect of lateral inhibition by the consumption of FGF1 should become dominant and the diameter of buds should become rather constant (Dr A. Mochizuki, personal communication). Since currently proposed candidates for lateral inhibitory factors such as BMP4 and sprouty2 are upregulated by FGF (Mailleux et al., 2001; Weaver et al., 2000), it is difficult to explain the observed pattern change by the secondary induction of these factors by FGF because the resulting pattern implies a loss of the lateral inhibitory mechanism. Therefore, the simplest explanation in this case
We have shown that the reaction–diffusion model of substrate-depletion type can explain the pattern formation in mesenchyme-free condition, but we should be careful about to what extent this model can be applied to in vivo situation. Some correlation exists between in vitro and in vivo patterns because basement membrane components and FGFs exist both in in vitro and in vivo conditions (Ohtsuka et al., 2001) and collagenase and FGF7 treatments produce similar patterns in cultures with and without mesenchyme (Cardoso et al., 1997; Ganser et al., 1991). Moreover, recently it was shown that mouse lacking TIMP-3, which is able to inhibit MMP-1, -2, -3 and -9, shows spontaneous air space enlargement (Leco et al., 2001). This report also strengthens the possibility that a similar mechanism is working both in vitro and in vivo. However, other reports have suggested that the mechanism in vivo seems to be more complicated than that in vitro. For example, several subtypes of FGF are expressed during lung development and they seem to play different roles. The FGF10 molecule was shown to be expressed in the mesenchyme surrounding the tip of the growing epithelium at GD 11.5, while FGF1 and FGF7 distributed ubiquitously in the mesenchyme at later stages of development (Bellusci et al., 1997b), and these molecules differently regulate the expression of other genes (Lebeche et al., 1999). Moreover, we could not rule out the possibility that BMP4 acts as a lateral inhibitory factor via mesenchyme. Further studies are needed to elucidate to what extent the in vitro situation can be extrapolated to that in vivo. The difference between in vitro and in vivo conditions raises a number of interesting questions. For example, it needs to be investigated whether the existence of mesenchyme influences the pattern change observed in mesenchyme-free condition. It is possible that the lateral inhibitory mechanism is different between in vitro and in vivo, and the application of BMP4 or a saturated amount of FGF1 under existence of mesenchyme may cause quite different morphological changes as compared with those occurring under mesenchyme-free conditions. A method to directly assay the response of the mesenchymal tissue on various FGF molecules has been established (Lebeche et al., 1999), so the effect of mesenchymal tissue could be incorporated in the model in future. Moreover, the relationship between the ‘hardwired’ and ‘generative’ aspects of morphogenesis should be examined. Although the bifurcated pattern of the bronchial tree is strictly determined in vivo, the branching pattern does not seem to follow a specific rule under mesenchyme-free conditions and the identity of each bud is difficult to recognize. In a recent report, Hoxb gene cluster was shown to be expressed in the lung bud
36
T. Miura, K. Shiota / Mechanisms of Development 116 (2002) 29–38
mesenchyme, suggesting that the involvement of Hox genes in regionalization of the bronchial tree (Sakiyama et al., 2000). Since mesenchyme possibly contains information for determining the identity of each bud, only the generative aspect of branching morphogenesis should become apparent in mesenchyme-free conditions. The relationship needs to be clarified between the signals specifying the identity of each branch and those which produce a branched structure in a generative way. 3.3. The cystic and branched patterns in the bronchial tree can be explained by a single mechanism The model can explain the formation of both cystic and branched patterns, which is useful to explain cystic transformation of airway seen in various animals. For example, many fishes have swimbladder instead of lung, which apparently lacks branched structure (Wake, 1979). In the chick lung, the bronchial airway consists of two kinds of tissues, i.e. the cyst-like air sac and the bifurcated bronchial airway as seen in mice and humans. The sites of Hoxb6 expression in the chick lung primordium coincide with the that of air sacs, and a transplantation experiment revealed that the Hoxb6-expressing mesenchyme has a capacity to induce a cyst-like structure (Sakiyama et al., 2000). The model can explain the formation of both cystic and branched structures by a single mechanism and can predict which factors affect the branch pattern. It should be interesting to see whether Hoxb6 upregulates ECM-degrading enzymes or FGF7 which contribute to the formation of the cystic epithelial structure in vitro. 3.4. Possible application of the model to branching morphogenesis in other organs It remains to be determined whether the model described above can be applied to branching morphogenesis in other organs such as the kidney, salivary gland, prostate gland and pancreas. Since their duct systems are endodermal in origin and form a tree-like structure, it is likely that a similar mechanism works during morphogenesis of these organs. Several studies showed that these organs also utilize FGF during morphogenesis (Miralles et al., 1999; Morita and Nogawa, 1999; Thomson, 2001), therefore the model could contribute to the elucidation of the branching mechanism in these embryonic organs. For example, there is a clinically important disease called polycystic kidney, in which mutation in a membrane glycoprotein causes numerous cyst formations in the kidney (Peters and Breuning, 2001). The application of the model to the branching morphogenesis of kidney may elucidate the mechanism of cyst formation and can be useful in predicting possible modifying factor of the progression of the disease. Further interaction between experimental and theoretical studies is definitely needed to elucidate the generative aspects of branching morphogenesis in developing organs.
4. Experimental procedures 4.1. Lung epithelium culture Lung epithelium culture was carried out as described previously (Cardoso et al., 1997; Miura and Shiota, 2000; Nogawa and Ito, 1995). Gestational day 11.5 mouse embryos (plug day ¼ day 0) were removed from the uterus, and their lungs were aseptically collected and treated with 1 unit/ml dispase (Invitrogen, Corp., Carlsbad, California, USA) for 20 min at 378C. Then the epithelia of the lung were separated from the surrounding mesenchyme using sharp tungsten needles. Isolated epithelia were injected into a 20 ml drop Matrigel/Dulbecco’s modified Eagle medium (DMEM)/F-12 placed on a tissue culture insert (Nalge Nunc International, Naperville, IL, USA) or a four-well dish (10 mm in diameter) and incubated for 30 min at 378C in a humidified atmosphere. Then 400 ml of the culture medium (DMEM/F-12 1 0.1% bovine serum albumin (BSA)) containing various concentrations of FGF1 or FGF7 (Peprotech, Rocky Hill, NJ) was added and the explants were incubated for 24–48 h in a humidified atmosphere. In some cases, the culture was treated with 2 mg/ml collagenase I (Sigma, St Louis, MO) or 0.5 mM ilomastat (Carbiochem, San Diego, CA). 4.2. Time-lapse recording of lung epithelia cultured with BMP4 beads Time-lapse observation of explanted lung epithelia was carried out as described previously (Miura and Shiota, 2000). Lung epithelia were prepared as described above, and BMP4-soaked beads were implanted near the cultured epithelia. The culture dish was then placed on an inverted microscope (Nikon TMD) covered with an incubation chamber. The atmosphere in the chamber was kept at 378C and the chamber was filled with air containing 5% CO2 to maintain the culture medium around pH 7.0. The images of the lung explants were captured with a chargecoupled device (CCD) camera (Hamamatsu Photonics C2400-77) every 20 min, and stored in a magneto-optical disk. All procedures were controlled using a Macintosh Quadra 700 computer and the NIH Image program (developed at US National Institutes of Health and available on the Internet at http://rsb.info.nih.gov/nih-image). 4.3. Immunohistochemistry against FGF1 Cultured lung epithelia were fixed overnight with 4% paraformaldehyde. Then the samples were embedded in paraffin, sectioned and dewaxed. After blocked by 1.5% normal donkey serum in PBS for 1 h, an anti-FGF1 antibody (Santa Cruz, Santa Cruz, CA) was applied to the samples and incubated at 48C overnight. For negative controls, PBS was used instead of primary antibody. The samples were washed three times with PBS for 5 min and treated for 1 h with the biotin-conjugated secondary antibody at 378C.
T. Miura, K. Shiota / Mechanisms of Development 116 (2002) 29–38
Then the samples were washed three times with PBS for 5 min and the distribution of FGF1 was visualized using the Elite ABC kit (Vectastain, Burlingame, CA, USA). 4.4. Preparation of fluorescent dye-labelled FGF1 Fluorescent dye-labelled FGF1 was prepared as described previously (Citores et al., 1999). Five micrograms of FGF1 protein (Peprotech, Rocky Hill, NJ) diluted in 50 ml of 0.1 M sodium carbonate–sodium bicarbonate buffer was incubated with Cy3 dye (Amersham Pharmacia, Uppsala, Sweden) for 30 min at room temperature. Then the remaining dye was removed by washing with PBS two times using Microcon YM-10 (Millipore, Bedford, MA). For negative controls, PBS was used instead of FGF1 protein solution. For FGF1-soaked beads, 50 ml of Heparin beads solution (Sigma, St. Louis, MO) was added in the Cy3–FGF1 solution and incubated overnight at 48C. The beads were washed three times with PBS and used as a source of FGF. 4.5. Gelatin zymography The medium for the lung epithelium culture was collected after 48 h and concentrated approximately ten-fold using Microcon YM-10 (Millipore, Bedford, MA). The samples were mixed with a 2 £ sample buffer without reducing reagents and loaded onto a 10% polyacrylamide gel containing 0.75 mg/ml bovine skin gelatin. After electrophoresis, the gel was washed twice with 2.5% Triton X-100 and incubated overnight with a substrate buffer (5 mM CaCl2, 50 mM Tris–HCl pH 8.0) (Ganser et al., 1991). The gel was stained with Commassie blue and destained with methanol/acetic acid. The image of the gel was captured using a flatbed scanner and the band was visualized using the NIH-Image software. Acknowledgements The authors especially thank Professor Stuart Newman for critical reading of the manuscript and helpful comments. Helpful discussions with Professor Masayasu Mimura, Professor Hisao Honda, Professor Mitsugu Matsushita, Dr Atsushi Mochizuki, Dr Mikiko Kobayashi-Miura and technical assistance by Dr Hiroyuki Nogawa and Ms Takako Shibayama are also gratefully acknowledged. This work was sponsored by the Japanese Ministry of Education, Culture, Sports, Science and Technology and the Japanese Ministry of Health, Labour and Welfare. Appendix A. Theoretical model of pattern formation in vitro The theoretical model we used is a modified form of a reaction–diffusion model of the substrate-depletion type (Meinhardt, 1995), which was originally proposed by Meinhardt (1976) and has been utilized to explain the pattern
37
formation in bacterial colonies (Kawasaki et al., 1997; Matsushita et al., 1998). We used a discrete model instead of a continuous model because discrete model can deal with discontinuous distribution of cell density. Unlike bacterial colonies, the spatial distribution of cell density is apparently discontinuous at the interface between the epithelial explant and the surrounding Matrigel and it is difficult to prove that Laplacian can be always defined at this point if we use continuous model (Dr A. Mochizuki, personal communication). At first we divided the whole culture area into small grids of the same size and defined the epithelial cell density and FGF1 concentration of each grid at time t as cðx; y; tÞ and nðx; y; tÞ (Fig. 3a). Then we defined the changes in the epithelial cell density and FGF1 concentration after a short length of time as described in the following equations, which were based on the assumptions in the Section 2.4: cðx; y; t 1 1Þ 2 cðx; y; tÞ ¼ pf ½cðx; y; tÞ; nðx; y; tÞ 1 d1 JðcÞ nðx; y; t 1 1Þ 2 nðx; y; tÞ ¼ 2f ½cðx; y; tÞ; nðx; y; tÞ 1 d2 DðnÞ where f ðc; nÞ represents the consumption ratio of FGF1 by epithelial cells and is an increasing function for both c and n. We included saturation kinetics because cell proliferation appeared to be saturated at high FGF1 concentrations. pf ðc; nÞ is the growth rate of the cell where p is the conversion rate of consumed FGF1 to epithelial growth. d1 and d2 are the diffusion coefficients for cells and FGF1, respectively. D represents diffusion of the FGF molecule and we defined the function as:
DðnÞ ¼ nðx 1 1; y; tÞ 1 nðx 2 1; y; tÞ 1nðx; y 1 1; tÞ 1 nðx; y 2 1; tÞ 2 4nðx; y; tÞ where J is the passive movement of epithelial cells. Simple Laplacian is not suitable for describing the movement of epithelial cells because they adhere tightly to each other and do not diffuse unlimitedly. We defined this function as: JðcÞ ¼ D½Fðc 2 c 0 Þ
½FðxÞ ¼ 0ðx , 0Þ; xðx $ 0Þ
A qualitative explanation of the function J(c) is as follows: if the cell density in a given grid is lower than a certain value c0, the cells do not cover the whole grid and should not affect cells in the neighbouring grids. However, if the cell density becomes higher than c0 by proliferation, the cells should repulse each other and push toward the neighbouring grids, which results in diffusion-like dynamics of the cell density. This function stabilizes the movement of cells and limits the variation of patterns, which were obtained by original reaction–diffusion model (Matsushita et al., 1998). Their model used simple Laplacian for cell movement and generate complex patterns such as concentric ring, which are unlikely to be reproduced in lung epithelia. We defined the initial cell density c(x,y,0) as c ¼ c 0 in a certain area and c ¼ 0 in other areas (Fig. 3b) and the initial FGF1 concentration as nðx; y; 0Þ ¼ n0 (constant
38
T. Miura, K. Shiota / Mechanisms of Development 116 (2002) 29–38
value). Numerical calculation of the model was carried out using Macintosh G4/350 with Metrowerks CodeWarrior. The simulation parameters are as follows: f(c,n) ¼ 5cn/ (1 1 3\ n), p ¼ 1:0, c0 ¼ 1:0, d1 ¼ 4:0, d2 ¼ 3:4–27:0, n0 ¼ 0:5–0:8, grid number ¼ 200 £ 200. References Ball, P., 1999. The Self-Made Tapestry, Oxford University Press, Oxford. Bellusci, S., Furuta, Y., Rush, M.G., Henderson, R., Winnier, G., Hogan, B.L.M., 1997a. Involvement of Sonic Hedgehog (Shh) in mouse embryonic lung growth and morphogenesis. Development 124, 53–63. Bellusci, S., Grindley, J., Emoto, H., Itoh, N., Hogan, B.L., 1997b. Fibroblast growth factor 10 (FGF10) and branching morphogenesis in the embryonic mouse lung. Development 124, 4867–4878. Bellusci, S., Henderson, R., Winnier, G., Oikawa, T., Hogan, B.L., 1996. Evidence from normal expression and targeted misexpression that bone morphogenetic protein (Bmp-4) plays a role in mouse embryonic lung morphogenesis. Development 122, 1693–1702. Canete-Soler, R., Gui, Y.H., Linask, K.K., Muschel, R.J., 1995. Developmental expression of MMP-9 (gelatinase B) mRNA in mouse embryos. Dev. Dyn. 204, 30–40. Cardoso, W.V., Itoh, A., Nogawa, H., Mason, I., Brody, J.S., 1997. FGF-1 and FGF-7 induce distinct patterns of growth and differentiation in embryonic lung epithelium. Dev. Dyn. 208, 398–405. Citores, L., Khnykin, D., Sorensen, V., Wesche, J., Klingenberg, O., Wiedlocha, A., Olsnes, S., 2001. Modulation of intracellular transport of acidic fibroblast growth factor by mutations in the cytoplasmic receptor domain. J. Cell Sci. 114, 1677–1689. Citores, L., Wesche, J., Kolpakova, E., Olsnes, S., 1999. Uptake and intracellular transport of acidic fibroblast growth factor: evidence for free and cytoskeleton-anchored fibroblast growth factor receptors. Mol. Biol. Cell 10, 3835–3848. Fukuda, Y., Ishizaki, M., Okada, Y., Seiki, M., Yamanaka, N., 2000. Matrix metalloproteinases and tissue inhibitor of metalloproteinase-2 in fetal rabbit lung. Am. J. Physiol. Lung Cell Mol. Physiol. 279, L555–L561. Ganser, G.L., Stricklin, G.P., Matrisian, L.M., 1991. EGF and TGF alpha influence in vitro lung development by the induction of matrix-degrading metalloproteinases. Int. J. Dev. Biol. 35, 453–461. Gilbert, S.F., 1997. Developmental Biology, Sinauer, Massachusettes, MA. Hogan, B.L., 1999. Morphogenesis. Cell 96, 225–233. Kawasaki, K., Mochizuki, A., Matsushita, M., Ueda, T., Shigesada, N., 1997. Modeling spatio-temporal patterns generated by Bacillus subtilis. J. Theor. Biol. 188, 177–185. Kitaoka, H., Takaki, R., Suki, B., 1999. A three-dimensional model of the human airway tree. J. Appl. Physiol. 87, 2207–2217. Lebeche, D., Malpel, S., Cardoso, W.V., 1999. Fibroblast growth factor interactions in the developing lung. Mech. Dev. 86, 125–136. Leco, K.J., Waterhouse, P., Sanchez, O.H., Gowing, K.L., Poole, A.R., Wakeham, A., Mak, T.W., Khokha, R., 2001. Spontaneous air space enlargement in the lungs of mice lacking tissue inhibitor of metalloproteinases-3 (TIMP-3). J. Clin. Invest. 108, 817–829. Lubkin, S.R., Murray, J.D., 1995. A mechanism for early branching in lung morphogenesis. J. Math. Biol. 34, 77–94.
Mailleux, A.A., Tefft, D., Ndiaye, D., Itoh, N., Thiery, J.P., Warburton, D., Bellusci, S., 2001. Evidence that SPROUTY2 functions as an inhibitor of mouse embryonic lung growth and morphogenesis. Mech. Dev. 102, 81–94. Matsushita, M., Wakita, J., Itoh, H., Rafols, I., Matsuyama, T., Sakaguchi, H., Mimura, M., 1998. Interface growth and pattern formation in bacterial colonies. Physica A 249, 517–524. Meinhardt, H., 1976. Morphogenesis of lines and nets. Differentiation 6, 117–123. Meinhardt, H., 1995. The Algorythmic Beauty of Sea Shells, Springer, Berlin. Miralles, F., Czernichow, P., Ozaki, K., Itoh, N., Scharfmann, R., 1999. Signaling through fibroblast growth factor receptor 2b plays a key role in the development of the exocrine pancreas. Proc. Natl Acad. Sci. USA 96, 6267–6272. Miura, T., Shiota, K., 2000. Time-lapse observation of branching morphogenesis of the lung bud epithelium in mesenchyme-free culture and its relationship with the localization of actin filaments. Int. J. Dev. Biol. 44, 899–902. Morita, K., Nogawa, H., 1999. EGF-dependent lobule formation and FGF7dependent stalk elongation in branching morphogenesis of mouse salivary epithelium in vitro. Dev. Dyn. 215, 148–154. Nogawa, H., Ito, T., 1995. Branching morphogenesis of embryonic mouse lung epithelium in mesenchyme-free culture. Development 121, 1015– 1022. Ohtsuka, N., Urase, K., Momoi, T., Nogawa, H., 2001. Induction of bud formation of embryonic mouse tracheal epithelium by fibroblast growth factor plus transferrin in mesenchyme-free culture. Dev. Dyn. 222, 263–272. Packer, A.I., Mailutha, K.G., Ambrozewicz, L.A., Wolgemuth, D.J., 2000. Regulation of the Hoxa4 and Hoxa5 genes in the embryonic mouse lung by retinoic acid and TGFbeta1: implications for lung development and patterning. Dev. Dyn. 217, 62–74. Park, W.Y., Miranda, B., Lebeche, D., Hashimoto, G., Cardoso, W.V., 1998. FGF-10 is a chemotactic factor for distal epithelial buds during lung development. Dev. Biol. 201, 125–134. Pepicelli, C.V., Lewis, P.M., McMahon, A.P., 1998. Sonic hedgehog regulates branching morphogenesis in the mammalian lung. Curr. Biol. 8, 1083–1086. Peters, D.J., Breuning, M.H., 2001. Autosomal dominant polycystic kidney disease: modification of disease progression. Lancet 358, 1439– 1444. Richardson, M.K., Allen, S.P., Wright, G.M., Raynaud, A., Hanken, J., 1998. Somite number and vertebrate evolution. Development 125, 151–160. Sakiyama, J., Yokouchi, Y., Kuroiwa, A., 2000. Coordinated expression of Hoxb genes and signaling molecules during development of the chick respiratory tract. Dev. Biol. 227, 12–27. Thomson, A.A., 2001. Role of androgens and fibroblast growth factors in prostatic development. Reproduction 121, 187–195. Wake, M.H., 1979. Hyman’s Comparative Vertebrate Anatomy, The University of Chicago Press, Chicago, IL pp. 768. Weaver, M., Dunn, N.R., Hogan, B.L., 2000. Bmp4 and Fgf10 play opposing roles during lung bud morphogenesis. Development 127, 2695– 2704.