Accumulation of collagen III at the cleft points of developing mouse submandibular epithelium

Collagens are widespread extracellular matrix components in connective and epithelial tissues (Furthmayr & von der Mark, 1982). It has been established that the branching morphogenesis of mouse embryonic submandibular epithelium is under the influence of the mesenchyme (Grobstein, 1954; Bernfield et al. 1972; Kratochwil, 1969; Lawson, 1974; Nogawa & Mizuno, 1981; Nogawa, 1983) and it has been postulated that it is regulated by collagen (Grobstein & Cohen, 1965; Wessells & Cohen, 1968; Spooner & Faubion, 1980; Bernfield et al. 1984; Nakanishi et al. 1986a,b,c); Clostridial collagenase in culture impairs the epithelial branching and a collagenase inhibitor from bovine dental pulp has the reverse effect. Electron microscopic studies have revealed the presence of bundles of collagen fibrils at the epithelial-mesenchymal interfaces at the cleft point of the epithelium (Bernfield & Wessells, 1970; Nakanishi et al. 1986c; Fukuda et al. 1988).

More than ten genetically distinct types of collagen have been characterized (Martin et al. 1985), among which collagens I, III, IV and V are known to be present in developing mouse submandibular glands (Bernfield et al. 1984; Kratochwil et al. 1986). Previously these collagens, except IV, had been shown to form fibres or to associate with fibrillar structures (Mayne, 1984) and, recently, Chen & Little (1987) suggested that collagen IV also appeared to produce such structures in lung rudiments. The fact that an interstitial collagenase from bovine dental pulp, which degrades collagens I and III, but not IV or V, inhibited in vitro branching, as did the bacterial collagenase (Fukuda et al. 1988), suggested that the collagen required for cleft initiation could be type I and/or III.

However, salivary glands as well as pancreas, kidneys and lung from collagen-I-deficient Movl3 mutant mouse embryos underwent normal epithelial morphogenesis in vitro (Kratochwil et al. 1986; Dziadek et al. 1987), which raised an unavoidable question on the importance of interstitial collagens in cleft formation. We therefore tried to determine the localization of these types of collagen during early branching morphogenesis immunohistochemically and found that collagen III accumulated preferentially at the epithelial-mesenchymal interfaces of the cleft points of the epithelium. In addition, we will show that collagen III is regularly found at sites of cleft of the glands cultured in the presence of bovine dental pulp collagenase inhibitor.

An inhibitor for interstitial collagenases was described previously (Kishi & Hayakawa, 1984) and a bacterial collagenase (Clostridium hystolyticum;Peterkofsky, 1982), which is free of caseinolytic, hyaluronidase or chondroitinase activities (Nakanishi et al. 1986a), was kindly donated by Dr T. Ohya and Dr N. Yokoi of Amano Pharmaceutical Co., Japan.

Various types of collagen for immunoblot analysis except IV were obtained from DDY mouse skin. Collagens were solubilized by limited pepsin digestion and were precipitated by the addition of NaCl to a concentration of 2 ·0 M. The precipitates were dissolved in 0 ·5M-acetic acid and dialysed against 0 ·1 M-acetic acid and then 50mm-Tris-HCl buffer, pH 7 ·4 containing 2 ·5m-NaCl. Collagens I and III were collected by centrifugation (Glanville et al. 1979) (fraction 1). The supernatant contained collagens V and VI (fraction 2). Both fractions were dialysed against 0 ·1 M-acetic acid and lyophilized. Pepsinized collagen IV from human placenta was obtained from Seikagaku Kogyo Co., Japan.

Rabbit antiserum directed against collagen I from rat tail tendon was kindly provided by Mr M. Yoneda and Dr K. Kimata of Aichi Medical University. Goat antibodies raised against human and bovine collagens III, IV and V were obtained from Southern Biotechnology Associates (Alabama, USA). Peroxidase-conjugated anti-goat immunoglobulins and StraVigen immunostaining kit containing peroxidase-conjugated avidin were purchased from Cappel (Pennsylvania, USA) and BioGenex (California, USA), respectively.

12-day and 13-day embryos were obtained from DDY strain mice. The day of discovery of the vaginal plug was designated as day 0. Salivary glands were dissected from 12-day and 13-day embryos in Tyrode’s solution. Glands at three different stages were used: mid 12-day glands having a distal epithelial bulb without any indication of clefts, late 12-day glands with signs of initial clefts and early and mid 13-day glands with definite clefts. The method of organ culture was described previously and the concentrations of both Clostridial collagenase and dental pulp collagenase inhibitor used were 5 ·0 μg ml ⁻¹ (Nakanishi et al. 1986a).

For the determination of specificities of antibodies, collagens were subjected to interrupted electrophoresis (Sykes et al. 1976) on SDS-8% polyacrylamide gel with some modifications (Hashimoto et al. 1986). Gels were electroblotted and nitrocellulose sections were incubated with antibodies against various types of collagen. After washing, bound first antibodies were detected with horseradish peroxidase-conjugated second antibodies, 3,3 ′-diamino-benzidine (DAB) and H₂O₂.

Intact 12-day and 13-day glands and 12-day glands cultured for 10 h under various conditions were fixed in icecold Bouin’s fluid for 30 –60 min. Fixed tissues were immediately washed in cold 70% ethanol, dehydrated with cold ethanol and embedded in paraffin. Sections (5;zm) were deparaffinized, treated with methanol containing H₂O₂ for 30 min and washed in phosphate-buffered saline. After treatment of the sections with trypsin (2 μg ml^-1) for 10 min at room temperature, the first antibodies were applied at room temperature in a wet box. Dark-brown colour was developed by the use of the peroxidase-conjugated second antibodies or avidin, DAB and H₂O₂ (Graham & Kamovsky, 1966). Control sections without first antibodies always gave negative staining patterns. It is of note that 2h fixation of tissues in Bouin’s fluid almost abolished the antigenicities of collagens. Fixation of tissues in 10 % neutral formalin and 95 % ethanol did not give satisfactory results. Collagen antigenicities were also impaired by treatment of deparaffinized sections with Clostridial collagenase (1 mg ml ⁻¹) for 3 h at room temperature but not with chondroitinase ABC (Yamagata et al. 1968). To determine the distribution of collagens, we examined more than 30 intact glands at days 12 and 13 and more than 15 glands cultured in each condition.

Fig. 1 shows the specificities of the antibodies used in this study which were raised against collagens I, III, IV and V, respectively. Coomassie blue staining clearly indicated that interrupted electrophoresis was successful in separating collagen I from III (lane 1) and collagen V from I, III and VI (lane 3), respectively. The rabbit antibody directed against rat collagen I was highly specific to α1(I) and α2(I) chains of collagen I (Fig. 1B, lane 1). The goat antibody against human and bovine collagen III reacted only with mouse skin α1(III) chain (Fig. 1C, lane 1) and exhibited no cross-reactivities with other types of collagen. The goat antibodies against human and bovine collagens IV and V were also found to be satisfactory in specificity for the immunohistochemical examination (Fig. 1D and E). It should be stressed that none of the antibodies used reacted with mouse skin collagen VI (Fig. 1, lane 3).

The immunoperoxidase staining patterns of the 12-day submandibular glands with antibodies against collagens I and III are shown in Fig. 2. In the mid 12-day gland with a round lobule, the anti-collagen I antibody strongly stained the extracellular matrix of the mesenchyme and the epithelial-mesenchymal interfaces, where a uniform and continuous dye distribution was observed (Fig. 2A). The staining pattern with the anti-collagen III antibody was similar to that of collagen I in the mesenchyme. However, a stronger staining was observed at the epithelial-mesenchymal interfaces (Fig. 2B,C). It is notable that staining was discontinuous at basal surfaces of the epithelial layer and that the dye was intensely deposited at every slightly indented point.

At the late 12-day stage, some basal epithelial sites begin to dent inward in several positions (Spooner & Faubion, 1980; Nogawa, 1983; Nakanishi et al. 1986b). The strongest staining of collagen III was seen at these sites, but the patchy staining pattern was preserved (Fig. 2E,F). Collagen I was, however, found to be present rather uniformly in the mesenchyme and at the epithelial-mesenchymal interface and no definite, reproducible staining was observed in initial narrow clefts (Fig. 2D).

At late 12-day to early 13-day stages, branching morphogenesis begins with appearance of a few definite V-shaped surfaces of the basal epithelial layer of the lobule. Intense immunoreaction of collagen III was observed in these areas, especially in the extracellular spaces of the mesenchyme (Fig. 3A,B). The bottom of the narrow clefts was strongly positive for collagen III (Fig. 3C,D). In these sections, the basal epithelial surfaces between cleft points exhibited weak but patchy staining intensity (Fig. 3A,B), as seen in the 12-day gland (Fig. 2). It should be emphasized that all the initial clefts reacted positively with the anti-collagen III antibody. This type of collagen also accumulated at the joining points of the lobule and stalk (arrowheads in Fig. 3A,B), suggesting that these sites correspond to clefts. Time-lapse cinematographic studies supported that this is indeed the case (Nogawa & Nakanishi, unpublished observations).

On the other hand, collagen I distribution in the early 13-day gland was as uniform as in the 12-day gland (Fig. 3E). However, relatively strong staining at the site of initial clefts was sometimes observed, though much less often than that of collagen III (see an arrow and arrowheads in Fig. 3E). The staining with anti-collagen V antibody showed a similar staining pattern to that with the anti-collagen I antibody (Fig. 3G). The basement membrane collagen IV was uniformly distributed along the basal epithelial surface (Fig. 3F). No specific accumulation in the early clefts was observed. However, considerable immunoreaction was seen in the mesenchyme where capillary endothelial cells were located (Fukuda et al. 1988).

Initial narrow clefts, such as those in Fig. 3, become wide at the mid 13-day stage (Fig. 4). Staining of collagen I at the basal epithelial layers was not strong except for that in the cleft (Fig. 4A). However, the relative intensity of immunoreaction of collagen III at the interfaces of both the distal aspects of the lobule and clefts was stronger than that in the mesenchyme (Fig. 4B,C). In particular, it is evident that accumulation of collagen III was also found in newly formed shallow indentations and clefts (see small arrows in Fig. 4B). However, patchy staining was less significant than that in 12-day glands (Fig. 4C). Collagen IV encircled the lobules as in the 12-day glands and was found to be abundant in the clefts (Fig. 4D), where basal laminae were often folded (Spooner & Faubion, 1980; Fukuda et al. 1988). It was always observed that staining of collagens I, III and IV in wide clefts of mid 13-day gland was prominent at the lateral sides of the lobules close to the bottom of clefts, as seen in the photograph by Ekblom (1984).

We have established that branching of the mouse embryonic submandibular epithelium is enhanced when the 12-day glands are cultured in the presence of the collagenase inhibitor and that it is halted when cultured in the presence of the bacterial collagenase (Nakanishi et al. 1986a). If collagen III was a crucial collagen type as in the intact glands, its accumulation would be expected at every initial cleft point. Fig. 5B indicates that the number of clefts in the gland cultured with the inhibitor for 10 h increased as compared with the control gland (Fig. 5A), and that collagen III stained strongly in the clefts. It should be emphasized that in some glands the bottom of every cleft showed clear dye deposition (Fig. 5C) while, in other cases, there were a few dotted stains in one cleft (Fig. 5D). Collagen I was seen in the clefts of the treated glands (Fig. 5G,H), but not every cleft was stained for collagen I (see arrows in Fig. 5H).

Clostridial collagenase in the culture medium totally inhibited cleft initiation of the 12-day glands, as described previously (Nakanishi et al. 1986a). Nevertheless, staining of both collagens I and III was observed in the experimental glands (Fig. 5E,I). However, periodicity of the staining pattern of collagen III was almost lost. This suggested that fragments derived from intact collagen molecules by the collagenase treatment were still associated with the tissues.

Our previous studies suggested that the epithelial branching of mouse embryonic submandibular gland is influenced by interstitial, fibril-forming collagens (Nakanishi et al. 1986 a,b,c, Fukuda et al. 1988) and that it is independent of epithelial cell proliferation (Nakanishi et al. 1987). Although collagen types known to be present in submandibular glands include not only I and III but also IV and V (Bernfield et al. 1984; Kratochwil et al. 1986), none of the experiments reported so far determined which type of collagen was essential for cleft initiation. Studies using Movl3 mutant mouse embryos proved that collagen I was not a candidate involved in this phenomenon (Kratochwil et al. 1986; Dziadek et al. 1987). It was therefore a reasonable assumption that collagen III would be involved in cleft initiation. This is consistent with the present observation that there is an accumulation of collagen III at every cleft point. Furthermore, it was surprising that the staining pattern of collagen III was discontinuous and almost periodic, particularly at the early stages of branching, the pattern becoming less significant at the mid 13-day stage. Even at this stage, newly formed, shallow clefts of the lobule were stained with the anti-III antibody much stronger than the other distal aspects of the lobule (Fig. 4).

On the other hand, no specific accumulation of collagen I in clefts was demonstrated immunohistochemically at the early 13-day stage. In some cases, collagen I was detected at the branch points more clearly than at the tips of the lobule, but this was not reproducible. Since the transmission electron microscopic studies (Bernfield & Wessells, 1970; Fukuda et al. 1988) revealed that matrix components including mesenchymal cells were found even in narrow clefts, the detection in clefts of collagen I, which might not be actively involved in cleft formation, was not unlikely. Collagen types IV and V did not appear to be concentrated in narrow clefts. These results are perfectly consistent with our previous findings that the glands contain an interstitial-type collagenase(s) and that the fibrillar structures participate in epithelial branching (Nakanishi et al. 1986a,c; Fukuda et al. 1988).

Examination of rudiments cultured in the presence of the collagenase inhibitor also supported the above findings. As shown in Fig. 5, the inhibitor-treated gland underwent typical branching morphogenesis with the formation of a hand-like epithelial shape during 10 h of cultivation. Importantly, the bottom of every narrow cleft strongly reacted with anti-III antibody. As for collagen I, some of the narrow clefts were stained but the others were not, as shown in the intact 13-day gland (Fig. 3). It is possible that a portion of the collagen I accumulated in the treated glands was mobilized together with collagen III for cleft initiation.

Summarizing these observations, we conclude that collagen III, not I, is actively involved in branching of the epithelium. It is also worthy of mention that the presence of collagen III at the interfaces of 12-day glands appeared to be restricted to indented areas between neighbouring epithelial cells of the lobules without clefts (Fig. 2B,C). The scanning electron microscopic studies revealed the presence of collagen-like fibrils between epithelial cells at the tip of the 12-day lobule (see figs 3, 4 in Nakanishi et al. 1986c). The reason for the collagen III accumulation at the epithelial-mesenchymal interfaces is unknown, but may partly be ascribed to the restricted existence of some receptor for collagen III at the basal epithelial surface. One of the possible candidates is cell surface heparan sulphate proteoglycan which has been shown to bind to native fibrillar collagens including collagen III (Koda et al. 1985). There is another possibility that traction forces exerted by the mesenchyme through collagen fibrils (probably type III) may press over the lobular surface randomly, eventually resulting in their discontinuous localization at the interfaces. It is also possible that specific cells apposed to the epithelial surfaces synthesize and secrete collagen III preferentially. We have no evidence which mechanism works for the patchy distribution of collagen III. However, it should be stressed that the basal surfaces of the collagenase-treated glands became smooth, and no specific localization of collagen III was observed (Fig. 5E).

The change in distribution of collagen III, from even to uneven, could be crucial for cleft initiation. One possible explanation is the change of the activity and the location of the interstitial collagenase(s) in the gland. Additionally, we have recently identified a characteristic cell movement of submandibular mesenchyme, in which cells with branch-inducing activity move around in groups (Nogawa & Nakanishi, 1987). Assuming that mesenchymal cells have some, as yet unknown, receptor for collagen III, such cell movement might concentrate collagen fibrils into bundles and also change their distribution. In these respects, studies on the metabolism of collagen III and its interaction with the epithelium and the mesenchyme could be an essential prerequisite for understanding the mechanism of epithelial shape change.

There is good evidence that collagen III is abundant in embryonic tissues (Furthmayr & von der Mark, 1982). Collagens I and III were once considered to codistribute in connective tissues, but recent studies have shown that both collagen types are expressed separately during avian neural crest development (Duband & Thiery, 1987) and during morphogenesis of fetal rat testis and ovary (Paranko, 1987). However, no definite biological function has yet been established. This study provides an important clue concerning one role of collagen III in epithelial-mesenchymal interactions during organogenesis.

We thank Mr M. Yoneda and Dr K. Kimata for the anticollagen I antibody. Y.N. wishes to express his thanks to Dr S. Suzuki for the critical discussion. This work was partly supported by grants to Y.N. from the Ministry of Education, Science and Culture of Japan (No. 61540518 and 62540544) and a grant from The Mitsubishi Foundation.