Abstract
We have examined directly the expression of one collagen gene (DCgl) during Drosophila melanogaster embryogenesis by means of in situ hybridization. Transcripts of this gene, which were demonstrated to encode a basement membrane type IV collagen chain, began to accumulate specifically in mesodermal de-rivatives at stages 12-13 of embryogenesis, and not before. Cells expressing this gene overlap, or are closely intermingled with, somatic and visceral mesoderm in stages 12-14. In stages 15-17, in addition to the strongly positive fat bodies, highly labelled cell spots are found scattered around all the parts of the gut and symmetrically on each side of the ventral nerve cord. They correspond to circulating mesodermal cells which we consider to be haemocytes or mesoblasts.
Introduction
The role of collagenous matrices during developmental processes is now better documented and it is evident that they are involved in cytodifferentiation, cell adhesion, cell migration and cell microenvironment (Hay, 1984; Bissell et cd. 1982). It is now a current concept that basement membrane components, including type IV collagen, are important guiding factors in morphogenetic movements of cells and tissues (Reddi, 1976; Hay, 1984), but little is known about their specific role. The importance of this function was thought to be crucial during embryogenesis since, in early embryos, basement membranes segregate into layers of cells that have separate lineages and different biochemical properties (ectoderm, endoderm, mesoderm).
Type IV collagen, a major structural component of basement membranes, appears first in the mouse embryo at the 2-cell stage (Sherman et al. 1980). According to Leivo et al. (1980), its appearance in the mouse blastocyst coincides with differentiation of the primitive endoderm and assembly of the first embryonic basement membrane. It forms a specific lattice network by end-to-end interactions, the C-terminal noncollagenous domain (NCI) having been shown to be one of the major cross-linking sites (Timpl et al. 1981; Kühn et al. 1981; Yurchenko et al. 1986; Weber et al. 1984).
By means of sequence analysis of cDNA and genomic DNA clones for the 3’ end of the collagen gene DCgl, characterized solely in Drosophila melanogaster (Monson et al. 1982; Le Parco et al. 1986ü,b), we have given the first direct evidence that type IV collagen exists in this organism (Cecchini et al. 1987). The data we have reported revealed a high degree of similarity between the noncollagenous domains (NCI) of Drosophila and of human and mouse al (IV) chains, including an unusual internal homology. The extent of both interspecies and intramolecular homologies strongly suggests the maintenance, both in vertebrates and in invertebrates, of an ancestral specific function.
Taking into account the relative similarity of Drosophila and vertebrate type IV collagens, and the extensive knowledge of the morphogenetic events occurring during Drosophila embryogenesis, we were prompted to study type IV collagen gene expression during the onset of Drosophila life cycle.
The results presented in this paper are based on in situ hybridization experiments on tissue sections probed with labelled coding fragments of the Drosophila DCgl collagen gene. Thus, we examined directly the expression of this gene during embryogenesis by detecting the sites, and the onset, of specific mRNA accumulation. The results suggest a fundamental role of circulating mesodermal cells during morphogenetic processes.
Materials and methods
DCgl collagen gene
We previously reported the identification and characterization of members of the Drosophila collagen gene family (Le Parco et al. 1986a,b). Specifically, we studied the DCgl gene, previously characterized by Monson et al. (1982), and we showed that it is a single-copy gene located at the 2L25C locus. It mainly encodes a 6-4 kb transcript, differentially accumulated during the course of Drosophila development (Le Parco et al. 1986a,b;Knibiehler et al. 1987). By sequence analysis of cDNA and genomic DNA clones for the 3’ end, we were the first to demonstrate that DCgl gene encodes a type IV basement membrane collagen chain (Cecchini et al. 1987).
Preparation of the probes
The DNA sequences used for our in situ hybridization experiments were the restriction fragments of DCgl gene shown in Fig. 1. Since all the results were similar, we have chosen to use routinely the 1-5 kb long EcoRI-BamHI fragment (black boxes). The methods used for fragment DNA preparation were as described by Maniatis et al. (1982). The probe was labelled by the method of ‘randomprimed’ DNA labelling, developed by Feinberg & Vogelstein (1984), which enables labelling to high activities and to an equal degree along the entire length of the input DNA. With all the three 3H-labelled deoxynucleotides (3H-dATP, 3H-dCTP, 3H-dTTP, approximately 50CimM, NEN) we routinely obtained up to SxKdisintsmin-1/zg-1 DNA. This method thus permits shorter exposures (about 10 days), usually only obtained with 32P-labelled probes, but without 32P disadvantages (Knibiehler et al. 1987).
In situ hybridization
Frozen tissue sections were prepared from living embryos which were previously dechorionated and freed of their vitelline membrane before being embedded in OCT compound (Tissue teck). The 6/rm-thick sections were processed as described by Hafen et al. (1983) and finally acetylated with acetic anhydride as described by Hayashi et al. (1978) and dehydrated. All subsequent treatments of the slides were as previously described (Knibiehler et al. 1987).
Embryonic stages
The embryos were harvested every 3h (0-3, 3-6, 6-9, 9-12 h) in early development stages and every 6h in late development (12-18, 18-24h) at 22°C. Numbering of the embryonic stages was that proposed by Campos-Ortega & Hartenstein (1986) and their identification was based on the same morphological criteria.
Results
We have performed in situ hybridization experiments on serial frozen tissue sections of embryos selected at sequential stages of development. They were probed with different restriction fragments of DCgl genomic DNA clone, located upstream and downstream of the 1-5 kb EcoRI-BamHI fragment (black boxes in Fig. 1). The results we obtained with each fragment were strictly comparable, emphasizing the great specificity of DCgl mRNA detection by in situ hybridization.
No signal has been detected during early embryogenesis, even after the longest exposures. That is, the transcripts specified by type IV collagen DCgl gene do not appear to be accumulated anywhere (or at a level below the detection capacity of the technique) (Fig. 2). The first signals appeared abruptly at the end of stage 12 (or at the beginning of stage 13), i.e. approximately in 9-5h-old embryos.
Stage 12-14 embryos (9-5-12 h)
The germ band retracts during stage 12. During this process, the fusion of anterior and posterior midgut takes place, and the definitive segmental furrows become apparent. At the end of stage 13, muscle cells become visible, inserting at incipient apodemes of the lateral epidermis. At completion of germ band shortening (stage 14), pharynx and oesophagus are clearly distinguishable and cytodifferentiation of axonal processes begins in sensory organs.
The autoradiographs corresponding to these stages are shown in Figs 3, 4. Globally, the repartition of hybridization can be divided into three different sets, which overlap or are closely intermingled with mesodermal derivatives.
The first set of results corresponded to strongly labelled, discrete spots each one overlapping one or a few cells. In 10h-old embryos they are either arranged in tightly packed slender layers or are present as single entities, scattered all around the different parts of the gut (Fig. 3A,B,D). These accumulation sites of DCgl transcripts strictly coincide with visceral mesoderm, a derivative of the inner splanchnic layer which, at these stages, detaches as single cells from the mesodermal primordium (Campos-Ortega & Hartenstein, 1985; Beer et al. 1987; Poulson, 1965).
The second set also had the appearance of intensely labelled spots and was well visible, as shown in Fig. 3C. Its segmented nature is sometimes difficult to observe owing to the plane of the sections. When sagittal sections were obtained, the spots appeared to be regularly arranged between the ectoderm and the ventral cord and, also, symmetrically between the ventral cord and endoderm. This second set of labelling corresponds to cells of the segmented somatic mesoderm. This derivative originates from single cells of the somatopleura, contacting both the developing nervous system and ectoderm.
The last set of hybridization signals was mostly visible in frontal and transverse sections (Figs 4A,B, 3D). It consists of a slightly labelled, thin, dorsolateral layer of cells extending on either side along the embryo between the gut and the somatic musculature. It overlaps strictly the mesodermal derivative that will give rise to the fat bodies extending from the gonad to the level of the brain hemispheres.
Stage 15, 16, 17 embryos (12-24 h)
During these stages, dorsal closure and epidermal segmentation are accomplished, the gut completely contains the yolk sac (stage 15), shortening of the ventral cord occurs and the fat body can be seen laterally extending from the gonads to the anterior thoracic levels. The larval pattern of somatic musculature becomes distinguishable (stage 16). During stage 17, no conspicuous differences can be distinguished compared to stage 16.
On the autoradiographs corresponding to these stages, the signals previously detected in younger embryos are reproduced and amplified (Fig. 5); they confirm the mesodermal nature of the cells actively accumulating DCgl transcripts, i.e. (i) strongly labelled single cells of the visceral mesoderm, scattered over the entire embryo, surrounding all parts of the gut, the brain lobes and other organs (Fig. 5); (ii) intense spots on somatic mesoderm cells arranged regularly in each segment beneath the epidermis and also on each side of the ventral nerve cord (Fig. 5D), where they seem to be associated with each of the paired nerves innervating the corresponding segment of the body; (i) lastly, the very active fat bodies extending on each side of the embryo (Fig. 5B,C,F).
If we exclude the fat body cells, we can conclude that DCgl transcripts are specifically accumulated during Drosophila late embryogenesis by single circulating mesodermal cells. These cells do not overlap with the forming muscles and are still present even after muscle formation (stages 15, 16). This type of circulating mesodermal cell corresponds to the definition lato sensu of a haemocyte which, as mentioned by Jones (1962), ‘is a mesodermal cell which sometimes during its life circulates in the haemolymph’ and which originates from the midventral epineural sinus (Mori, 1979).
According to this definition the myocytes, too, might be considered as transformed haemocytes. However, the term ‘haemocyte’ is now mainly used to characterize circulating cells involved in phagocytosis (Crossley, 1975) so we will use the more general term of mesoblast to define all the cells belonging to the family of mesodermal circulating cells (J. A. Hoffmann, personal discussion).
Discussion
We have used a fragment of a DNA sequence encoding a type IV basement membrane collagen chain in Drosophila as probe in in situ hybridization experiments to examine directly the spatiotemporal expression of the corresponding DCgl gene during Drosophila embryogenesis. The main results of this investigation are (i) the DCgl gene only begins to be expressed at developmental stages 12, 13 (9-5 h-old embryos) and (ii) its transcription products are specifically accumulated in mesodermal derivatives, i.e. fat bodies and mesenchymal circulating cells which we call haemocytes or mesoblasts.
The absence of DCgl gene expression during early embryogenesis does not mean that there is no type IV collagen during these stages. Indeed, as in vertebrates, Drosophila type IV collagen molecules consist of triple-helical thread-like particles composed of homotrimers or heterotrimers of two different types of o’chains (al (IV)) and o2 (IV)) (Bomstein & Sage, 1980; Timpl et al. 1981; Miller & Gay, 1982; Qian & Glanville, 1984; Fessler et al. 1984). Among the clones we have isolated (Le Parco et al. 1986a) one might correspond to the other type of a chain that could be involved in basement membrane composition during early embryonic stages. Work is in progress to check this hypothesis. Another possibility might be that type IV collagen has been stocked in the oocyte. The fact that follicular cells actively accumulate DCgl transcripts (Le Parco et al. 1986b) may favour this hypothesis. Work is in progress to obtain specific antibodies raised against a fusion protein containing a portion of the DCgl a IV chain.
Although it is currently admitted that basement membrane type IV collagen is deposited by the underlying epithelial cells (Hay, 1981) the mesodermal expression of DCgl gene during Drosophila late embryogenesis was not surprising since we have previously shown (Le Parco et al. 19866; Knibiehler et al. 1987) that it is also differentially expressed during larval, pupal and adult stages in fat bodies, lymph glands and in circulating mesodermal cells. The mesodermal origin of these tissues and cells was recently confirmed by Lawrence & Johnston (1986) and Beer et al. (1987). The fat-body cells of all insects that have been examined are surrounded by a basement membrane (Ashhurst, 1979), which was assumed to be secreted by each cell. However, the insect fat bodies are responsible for a wide range of roles in the ontogeny of insects (Riski, 1978; Wyatt, 1980). These include the synthesis of the major haemolymph proteins. Thus, it may be possible, although we do not know the exact sites and the time of type IV collagen deposition, that fat bodies secrete into the haemolymph this component of basement membranes during the larval periods of growth.
The circulating mesoblasts are active in DCgl expression only during late embryogenesis and metamorphosis of Drosophila, both developmental stages in which specific morphogenetic processes occur. These ‘fibroblastic-like’ cells are intermingled with somatic and visceral mesoderm before muscle formation during stages 12-13 of embryogenesis, and, in addition, they surround the autolysing and reforming adult muscles during metamorphosis. These observations suggest that the type IV collagen that they synthesize may be incorporated into the forming basal lamina on the plasmalemma of the myotubes. This situation was recently described in vertebrate cell cultures in which fibroblasts were demonstrated to promote the formation of a continuous basal lamina by type IV collagen deposition during myogenesis (Sanderson et al. 1986; Kühl et al. 1984).
However, some mesoblasts, which are located between the nerve cord and the epidermis during late embryogenesis (Fig. 5D), may also be involved in the deposition of type IV collagen in an extracellular material necessary for later development of muscles and motoneurone growth cones. This was thought to be the case in grasshopper embryos in which it was suggested that the large mesodermal ‘muscle-pion-eers’ erect such a scaffold (Ho et al. 1983). The absence of, or a delay in, the deposition of this extracellular material overlapping the period of the first muscular contractions was thought to explain the substantial disorders observed during embryonic development of the mutant lethal myospheroid l(l)mys described by Wright (1960).
We think that the results we report here agree well with this hypothesis and support the notion that the mesoblasts as ‘fibroblastic cells’ seem to be of crucial importance during developmentally critical stages in Drosophila.
Acknowledgements
We thank Dr Michèle Cavallin (LGBC, Marseille) for her help in preparing the embryos and for discussions, and Michel Berthoumieux and Gerard Turini for the difficult photographic work. We are grateful to Serge Long (LGBC, Marseille) for providing living embryos and to Dr Ruth Griffin-Shea (LGBC) for correcting the English. This work was supported in part by a Contrat externe de recherche INSERM no. 854013.