Mechanical aspects of the mesenchymal influence on epithelial branching morphogenesis of mouse salivary gland

Branching morphogenesis of mouse salivary epithelium does not proceed without mesenchyme of salivary glands, accessory sexual glands or lungs (Grobstein, 1953; Cunha, 1972; Lawson, 1974). The mouse salivary mesenchyme is able to support branching morphogenesis of lung epithelium (Lawson, 1983)’, to induce mammary epithelium to branch in salivary-like fashion (Kratochwil, 1969) and further to induce nonbranching epithelium of quail anterior submaxillary glands to branch (Nogawa & Mizuno, 1981). These results suggest that mechanisms of branching morphogenesis of salivary glands are never understood without making the nature of instructive influences of the mesenchyme clear.

There are some different explanations for the processes of branching morphogenesis of salivary glands. The first model by Spooner & Wessells (1972) and Spooner (1973) showed that the contraction of epithelial microfilaments, which causes changes in cell shapes of epithelia, takes place in a specific area and forms a cleft. According to this model, the mesenchymal cells that stimulate epithelial microfilaments to contract should be prelocalized in the specific area, but no evidence has been given to show the prelocalization of the specific mesenchymal cells. Second, Bemfield (1981) and Bernfield & Banerjee (1982) reported that the epithelium whose basal lamina was degraded by the mesenchyme had a higher cell proliferation rate than the other epithelium whose basal lamina was stabilized by collagen, and they discussed a possibility that the differential cell proliferation rates cause cleft formation. However, recently, it was proved by Nakanishi, Morita & Nogawa (1987) that clefts were initiated and deepened in vitro when cell proliferation was inhibited with X-ray irradiation and aphidicolin treatment of salivary rudiments. The third model by Nogawa (1983) showed that the mesenchyme has an ability to determine the curvature of epithelial surfaces, and clefts are formed on the epithelial surface when the curvature increases. In the last model by Nakanishi, Sugiura, Kishi & Hayakawa (1986c), mesenchymal cells exert traction forces on collagen bundles at the epitheliomesenchymal interface and clefts are formed on the consequently deformed epithelial surface. In the first two models, it is epithelial cells that generate shape-changing forces and mesenchymal cells only modify the way the epithelial cells work. In the latter two models, in contrast, mesenchymal cells generate shape-changing forces and exert them on the epithelial surface. To determine which mechanism mainly functions in the branching morphogenesis of salivary glands, epithelial or mesenchymal forces, it is necessary to understand the physiological characteristics of the salivary mesenchyme.

In the present study, using mesenchymes of salivary gland, lung, stomach, mandible and skin, we comparatively studied relations between the branchinducing activity and two mechanical activities of these mesenchymes, and tried to elucidate mechanical aspects of the mesenchymal influence on salivary branching morphogenesis. One is the ability to contract collagen gels and the other is the movement of plastic beads that are placed in mesenchymal masses instead of epithelial lobes. Collagen-gel contraction is rather commonly possessed by fibroblastic cells: human skin fibroblasts (Bell, Ivarsson & Merrill, 1979; Grinnell & Lamke, 1984), rat skin fibroblasts (Buttle & Ehrlich, 1983), chick embryonic ventricle, skeletal muscle and dermis (Stopak & Harris, 1982) and bovine vascular smooth muscle cells (Ehrlich, Griswold & Rajaratnam, 1986). However, it is important to study this activity of the salivary mesenchyme, because collagens have a crucial role in salivary branching morphogenesis (Nakanishi, Sugiura, Kishi & Hayakawa, 1986a,b,c).

ICR mice were mated during the night and the day of the discovery of vaginal plug was counted as day 0. Rudiments of lung and stomach were isolated from 11-day fetuses in Hanks’ balanced salt solution (HBSS). Rudiments of submandibular gland, sheets of lateral body skin and mesenchymes of mandible were isolated from 13-day fetuses in HBSS.

Four kinds of rudiments excluding mandibular mesenchyme were treated with dispase (1000 protease units ml⁻¹ in HBSS; Godo Shusei Co., Tokyo, Japan) at 37·5°C for 30 min and epithelia and mesenchymes were separated with fine forceps. After the separated tissue fragments were washed with HBSS, one part of the mesenchymes was submitted to collagen-gel-contraction experiments, and the other part of the mesenchymes and lobes of submandibular epithelium were stored in HBSS with 20% horse serum (Gibco Lab.) at room temperature for recombination experiments.

To equalize the volume within five kinds of mesenchymes, we cut mesenchymal sheets of skin and stomach down to the same size as an isolated submandibular mesenchyme, mesenchymal pieces of lung and mandible being as large as a submandibular mesenchyme when isolated. Three homotypic pieces of each mesenchyme were assembled on semisolid medium composed of medium 199 Earle’s BSS (Gibco Lab.) with 20% horse serum, 0·5% agar (Difco Lab.) and penicillin G potassium (100 units ml⁻¹), and these mesenchymal assemblies were preincubated at 37·5°C in 5 % CO₂ for 4h,to allow compaction to occur. Three submandibular epithelial lobes with diameters from 140 to 160 μm, keeping mutual contact, were placed on the mesenchymal mass and cultivated.

In other recombination experiments, plastic beads were used in place of epithelial lobes. Cytospheres (Lux Scientific Corp.) with specific gravity 1-04 were available. Beads with a diameter of 150 μm were picked out and stored in HBSS with 20 % horse serum. Three beads were placed on a mesenchymal mass in the same manner as with epithelial lobes and incubated.

Mesenchymes were recombined with epithelial lobes or plastic beads on the above-described agar media, which were prepared with HBSS instead of Earle’s BSS in the hollow (diameter 31mm and depth 4 mm) of a thick glass slide. The hollow was covered with a cover-glass whose inner surface was coated with horse serum to avoid collecting moisture and sealed up with paraffin. The glass slide was set on the stage of an Olympus BHS microscope in a warm box adjusted to 37·5°C and recombinates were photographed with Kodak Plus-X 16 mm film at intervals of 2 min using an Olympus PM-16mm time-lapse cinematographic apparatus.

Mesenchymal pieces were dispersed in trypsin solution (0·25% in Ca²⁺-, Mg²⁺-free HBSS; Difco Lab., 1:250), and magnetically stirred for 30 min at 20°C. Suspensions of single mesenchymal cells were obtained through a nylon mesh with 20. μm square pores, and horse serum was added to it in order to stop the remaining activity of trypsin. Larger mesenchymal fragments that were caught in nylon mesh were re-treated in the same manner and the obtained cell suspension was added to the former. After the cell number was counted with a haemocytometer, mesenchymal cells were washed twice with a basal medium (medium 199 Earle’s BSS with 20% horse serum and 100 units ml⁻¹ penicillin G potassium) and finally suspended at l·0×10⁶ cells ml⁻¹ in the basal medium. Five kinds of mesenchymal cells prepared by this method consisted of more than 95 % single cells and seemed intact since the aggregated mass of the submandibular mesenchymal cells was able to support branching morphogenesis of a submandibular epithelial lobe (Nogawa & Nakanishi, 1986).

A collagen solution was purchased from Nitta Gelatine Co. (Osaka, Japan; Cellmatrix type I-A: acid-soluble fraction of type I collagen from bovine tendon). The collagen solution, 10× medium 199 Earle’s BSS, 200 mM-Hepes buffer solution and horse serum were mixed in a ratio of 16:2:2:5, and kept on ice to prevent immediate gelation. 4 vol. of the collagen mixture were added to 1 vol. of the cell suspension. The final concentrations of collagen and cells were l·5 mg ml^-1 and 2·0×10⁵ cells ml⁻¹, respectively. 0·5 ml of the final mixture was allowed to gel in each well of Falcon 24-well tissue-culture plate for 1 h of incubation at 37·5°C in 5 % CO₂. After 2 ml of the basal medium was added to each well, the gels were let float by inserting a needle around the comer of the well and incubated. Changes of the diameters of the gels were measured with the eye-piece micrometer of a dissection microscope.

One gel or 0·5 ml of the medium was placed in a screwcapped tube, hydrolysed with 2·5 ml of 6M-HC1 at 110°Cfor 24 h and neutralized with 10M-KOH. Hydroxyproline contents of a gel or the medium were assayed by the method of Kivirikko, Laitinen & Prockop (1967).

Gels were fixed in a 10% neutral formalin solution, embedded in paraffin and sectioned at 5 gm thickness. Sections were stained with Fast green FCF for cells and Sirius red F3BA (Schmidt GmbH) for collagen fibrils according to the procedure of López-De León & Rojkind (1985).

Cytochalasin B (Aldrich Chemical Co.) was dissolved at 10 mg ml^-1 in dimethylsulfoxide (DMSO) and stored. When used, cytochalasin B was added to the basal medium or the agar medium at a concentration of 10 mg ml⁻¹ with DMSO at 0-1%. In the control experiments, a medium containing only DMSO at 0·1% was used. The concentration of cytochalasin B in the present experiments was similar to that in experiments by Spooner & Wessells (1970, 1972). Cytochalasin D (Aldrich Chemical Co.), which was known not to inhibit hexose transport of cells, in contrast to cytochalasin B (Jung & Rampai, 1977), was also used at a concentration of l μg ml⁻¹.

A set of three lobes of submandibular epithelium was cultivated in recombination with five kinds of mesenchymes (Figs 1–7; Table 1). When three epithelial lobes were cultivated without any mesenchymes, they fused to form one spherical lobe within 8h of cultivation (Fig. 2). In all the recombinates with submandibular mesenchymes, the three lobes fused mutually and further branched typically 1 day after cultivation (Fig. 3). Lung mesenchyme had the ability to induce submandibular epithelial lobes to branch, though a little inferior to submandibular mesenchyme (Fig. 4). Half of the recombinates with stomach mesenchymes showed signs of branching morphogenesis 1 day after cultivation when four or five clefts were present (Fig. 5A) and all the recombinates conspicuously branched on the 2nd day (Fig. 5B). The submandibular epithelium never branched in recombinates with mandibular or dermal mesenchyme for 2 days of cultivation, and three clefts that had been formed by the fusion of three lobes (Fig. 2A) disappeared in recombinates with mandibular mesenchyme (Fig. 6), while they were partly present in recombinates with dermal mesenchyme (Fig. 7). The area of epithelium in recombinates with mandibular or dermal mesenchyme 1 day after cultivation did not expand larger than those at the beginning of cultivation (compare Fig. 6 or 7 with Fig. 1).

The diameters of collagen gels (1·5 mg ml⁻¹) containing submandibular mesenchymal cells (2·0×10⁵cells ml⁻¹) became 88 ± 3 % of the starting diameter 1 day after incubation, 69 ± 7 % after 2 days and 59 ± 6 % after 3 days. Two possibilities were considered as the cause of the decrease in size of the collagen gels. One was the decomposition of collagen molecules by the collagenase activity of mesenchymal cells and the other was changes in the meshes of the collagen lattice. The hydroxyproline contents were assayed in noncontracting gels without cells and contracting gels with 2×10⁵ cells ml⁻¹ 3 days after incubation (Table 2). Although the contracting gels were three fifths as large as the noncontracting gels in diameter, they had almost the same contents of hydroxyproline. The slightly lower value of the contracting gels than the noncontracting gels seemed to be due mainly to the fact that the contracting gels had less intragel space which retained the hydroxyproline-containing medium. Next, paraffin sections of the gels were stained with Fast green FCF and Sirius red F3BA to show the orientation of collagen fibres (Fig. 8). A noncontracting gel without mesenchymal cells was constituted with a loose meshwork of thin collagen fibrils. In a contracting gel, thick collagen fibres were observed tightly gathering around mesenchymal cells. These results indicate that the decrease in size of collagen gels was caused not by the degradation of collagen matrix but by the traction of collagen fibres by mesenchymal cells.

Collagen gels (1·5 mg ml⁻¹) containing respective mesenchymal cells at 2·0×10⁵ml^−l were incubated and activities of gel contraction were expressed as the percentage of the reduced length to the starting diameter (Table 3). Dermal mesenchymal cells, which failed to induce submandibular epithelium to branch in the state of a cell mass, had the strongest gel-contracting activity. The gel-contracting activity of submandibular or lung mesenchymal cells was two thirds as strong as that of the dermal mesenchymal cells and that of stomach or mandibular mesenchymal cells was much weaker. Five kinds of dispersed mesenchymal cells contracted collagen gels in different degrees irrespective of the degrees of their branching-morphogenesis-inducing activities (Table 1). These results suggest that the collagen-gel-contracting activity of cells is a necessary but insufficient condition for the branch-inducing mesenchyme.

We also tried to study specificity of the branchinducing mesenchymes when in a mesenchymal mass. A set of three plastic beads as large as the epithelial lobes was recombined with the mesenchymes and the movement of the beads in the mesenchymal mass was examined (Figs 9–14). The three beads were separated well in submandibular or lung mesenchyme (Figs 10, 11), but never separated in dermal mesenchyme (Fig. 14). Beads-separating activities of the mesenchymes were quantified both by counting the number of the separated points among three contact points and by measuring the sum of three distances away from each bead in individual expiants (Table 3). The beads-separating activity proved to correlate positively with the branching-morphogenesis-inducing activity. The values of ‘separated distance’ had larger deviations, but we observed in some recombinates that the beads moved closer to each other after once parting, which would explain the larger deviations.

The behaviour of mesenchymes in living recombinates was checked with time-lapse cinematography. Submandibular and lung mesenchymal cells were recognized moving (or flowing) around in groups not only near the surface of the beads but also far away from the beads. Since this type of cell movement was observed in the mesenchymes recombined with epithelial lobes, it was not abnormal cell movement caused by the plastic beads as foreign matter. In contrast, most dermal mesenchymal cells moved little, but some cells moved at random in short steps.

In the presence of cytochalasin B (CB, 10 μg ml⁻¹), neither contraction of collagen gels (Fig. 15) nor separation of plastic beads by submandibular mesenchyme were observed. Since the same results were obtained when cytochalasin D (1 μg ml⁻¹) was used, inhibitory effects of CB on the submandibular mesenchymal behaviour seemed to be caused not by inhibition of hexose transport but by disruption of mesenchymal microfilaments. We then examined the effect of CB on the gels that had contracted to 84 % of the starting diameter (Fig. 15). When the medium was replaced with the CB medium, the gels were observed increasing in size as early as 1 h after treatment and they relaxed to 86% of the starting diameter 6h after treatment, and continuing to relax gradually thereafter. The gels, however, resumed contracting when CB was washed out with the control medium 6 h or 1 day after treatment.

Earlier studies of mesenchymal influences on epithelial morphogenesis have mainly taken account of the biochemical aspect of the mesenchymal influences, for instance, the growth factor that stimulates proliferation of epithelial cells (Ronzio & Rutter, 1973; Goldin & Wessells, 1979; Goldin & Opperman, 1980) and the activity of degrading basal lamina (Bernfield & Banerjee, 1982; Smith & Bernfield, 1982), and have not considered the mechanical aspect. As for mechanical influences, Oster, Murray & Harris (1983) and Harris, Stopak & Warner (1984) pointed out the possibility that traction of collagen fibrils by dermal cells may cause the clumping of dermal fibroblasts and the interconnecting polygonal network of collagen bundles in the pattern formation of avian feather papillae. Nogawa (1983) and Nakanishi et al. (1986a,b,c, 1987) have suggested the possibility that the mesenchyme may exert shapechanging forces on the epithelial surface and clefts may be initiated on the deformed surface in the branching morphogenesis of mouse salivary gland. The present study submits more reliable evidence that shows the importance of the mechanical influences of the mesenchyme on branching morphogenesis of salivary epithelium.

The present study showed that the mesenchymes that were able to induce the salivary epithelium to branch had a higher beads-separating activity and that the mesenchymal cells moved (or flowed) around in groups. The separation of beads by the mesenchymes seems to come from the fact that mesenchymal cell current pushes or pulls the beads and from the invasion of mesenchymal cells into gaps between the beads. Since the flowing movement of mesenchymal cells was also observed in the mesenchymes recombined with epithelial lobes, these forces are thought to work in the process of cleft widening in which each lobe is separated while keeping mutual connections.

Lung mesenchyme induced the submandibular epithelium to branch, which is consistent with Lawson’s (1974) results. Stomach mesenchyme proved to be able to induce the submandibular epithelium to branch, though inferior to lung mesenchyme, and to separate plastic beads away to a degree corresponding to its branch-inducing activity. Often clefts were observed in recombinates of epithelial lobes with dermal mesenchyme, which never separated plastic beads. We, however, confirmed with time-lapse cinematography that the clefts were not newly formed but remained because the lobes had not fused, suggesting that immotile, stiff dermal mesenchymal cells physically stopped the lobes fusing into one spherical lobe. This may be supported by the fact that most of the lobes recombined with mandibular mesenchyme, which was slightly motile, fused to give one spherical lobe. Furthermore, when the epithelium was tightly packed with stiff mesenchyme, the epithelium without any space to grow might cease cell proliferation, probably due to ‘postconfluence inhibition of cell division’ (Martz & Steinberg, 1972). We observed that the area of submandibular epithelium did not expand in recombination with dermal and mandibular mesenchyme which were both stiff. The present study, however, gives no evidence of whether the epithelial expansion was inhibited mechanically by the stiff mesenchymes or by lack of growth factors produced by these mesenchymes. Lawson (1984) reported that the lung epithelium that failed to branch in recombination with submandibular mesenchyme began to proliferate actively and to branch when the mesenchyme was removed from it and then recombined with it. This recovery can be explained by the change of the surrounding mesenchyme from a stiff to a flowing mass, which was discussed as the third among four possibilities by Lawson (1984): she referred to ‘packed’ and ‘diffuse’ mesenchymes.

All the five kinds of mesenchymal cells examined had the activity of contracting collagen gels, and the dermal mesenchyme, which failed to induce the epithelium to branch, had the highest gel-contracting activity. The lack of correlation between branchinducing and gel-contracting activities appears to contradict our suggestion for the mechanisms of branching morphogenesis (Nakanishi et al. 1986c), but it is highly probable that the values of gelcontracting activities may not always reflect the strength of traction forces of mesenchymes in vivo because the density of mesenchymal cells and the content of extracellular matrix components containing collagen vary with each organ in vivo. Bernfield & Wessells (1970) and Nakanishi et al. (1986c) reported that collagen bundles are aligned at the epithelio-mesenchymal interface along shallow and deep clefts. Since aligned collagen fibres guide moving cells and conversely moving cells align collagen fibres (Oster et al. 1983; Harris et al. 1984), traction forces generated by the mesenchyme can be amplified to shapechanging forces where collagen fibres happen to be aligned and bundled by moving mesenchymal cells. Initial shallow clefts may be deepened with both the traction force and flowing movement of the mesenchyme. However, there is another possibility that, at the basal lamina, collagen may cross-link proteoglycans which associate with the actin cytoskeleton in the epithelium, and modify the morphology of the epithelium through this cross-linking (Rapraeger, Jalkanen & Bernfield, 1986).

In the present study, we examined the interaction of mesenchymal cells with gels of type I collagen. Recently, Kratochwil et al. (1986), using mouse embryos deficient in type I collagen, reported that salivary rudiments undertook branching morphogenesis normally in vitro without detectable synthesis of type I collagen. Since collagens have a crucial role in salivary branching morphogenesis (Nakanishi et al. V)9f>a,b,c), type III or V collagen, which is fibrillar (Mayne, 1984), may substitute for type I collagen in normal branching morphogenesis. In addition to a collagen-mediating mechanism, there can be another mechanism in which mesenchymal cells exert shapechanging forces on the epithelial surface by interacting with some insoluble materials to which they are attached, as shown by Harris, Wild & Stopak (1980) that chick heart fibroblasts were able to exert traction forces on the silicon rubber and make it wrinkle.

Spooner & Wessells (1970, 1972) and Spooner (1973) reported that shallow clefts disappeared and deep clefts, where collagen fibres were abundantly present, remained when submandibular rudiments were treated with cytochalasin B (CB). From the results, they presented the model of branching morphogenesis in which epithelial microfilaments contribute to the formation of clefts and accumulating collagen fibres stabilize the clefts. However, the present study, demonstrating inhibitory effects of CB on both the gel contraction and beads separation by submandibular mesenchyme, necessitates taking into account the effects of CB on mesenchymal as well as epithelial microfilaments. If the stability of shallow clefts is balanced with the traction forces of the mesenchyme, shallow clefts will disappear with CB in the same way as the contracting gels relax with CB, and deep clefts may remain due to the presence of rearranged collagen fibres and invading mesenchymal cells which become immotile with CB.

Epitheliomesenchymal specificity in salivary branching morphogenesis (Grobstein, 1953) may originate from the mesenchymal nature of motile tissue, but the question how the flowing movement of cells occurs in the specific mesenchymes is unanswered. Comparative studies of submandibular mesenchyme with dermal mesenchyme will give clues for approaching the problem. Pieces of dermal mesenchymes, which were incubated for 4h before recombination in the present experiments, seemed to contact more tightly with one another than pieces of submandibular or lung mesenchyme (data not shown), and the area of dermal mesenchyme became smallest during 1 day of incubation (Figs 7, 13). The cell-to-cell adhesion may have an important role in the flowing movement of the cell mass, and so may the adhesion of cell to extracellular matrices of the mesenchymal tissue (Funderburg & Markwald, 1986). It remains to be determined what cell adhesion molecules and extracellular matrix materials are essential to the characteristic movement of the salivary mesenchyme.

The authors wish to express their gratitude to Prof. T. Mizuno of the University of Tokyo for his continuous encouragement during the course of this work, to Dr S. Takeuchi (the University of Tokyo) for his helpful advice on time-lapse cinematography, and to Dr J. Enami (Dokkyo University School of Medicine) and Dr N. Shiojiri (Shizuoka University) for their technical advice. This work is supported in part by grants to Y. Nakanishi from the Ministry of Education, Science and Culture of Japan and from the Mitsubishi Foundation.