Differential expression of TGF β1, β2 and β3 genes during mouse embryogenesis

Transforming growth factor beta (TGF β) is the name given to a family of polypeptides that have multifunctional regulatory activities. To date, this family consists of five closely related members of which three mammalian isoforms exist: TGF β1, β2and β3 (see recent reviews by Roberts and Spom, 1990a; Lyons and Moses, 1990). TGF βs, are synthesized as inactive precursors, consisting of a latent associated protein (LAP) and an active mature form which is cleaved from the carboxy-terminal of the precursor to yield the biologically active dimeric molecule. Comparison of the mature human TGF β1, TGF β2 and TGF β3 peptides reveals identities of between 75 and 80%, while the LAP sequences are only 25–35 % conserved (Dernyck et al. 1988). The diverse biological effects of TGF β1 in vitro and in vivo have been well described in the literature and are detailed in the above review articles and will be briefly discussed below. In vitro TGF β1 has mitogenic effects on bone, cartilage and connective tissue fibroblasts. It inhibits the proliferation of epithelial cells and stimulates the expression of extracellular matrix proteins and the integrin class of adhesion molecules. In vivo TGF β1 stimulates the formation of granulation tissue during wound healing and also induces bone formation. The data available about the bioactivities of TGF β2 and TGF β3 indicate that these molecules are qualitatively similar although quantitative differences appear to exist in some systems (see proceedings of the Ciba Foundation Symposium Number 157, 1990, in press). Northern blot analysis has shown that all three TGF β genes are expressed during embryonic development and that the total levels of their specific mRNAs increase with the age of the embryo (Heine et al. 1987; Miller et al. 1989a,b). Previous in situ hybridisation studies have shown that TGF β1 is very abundant in fetal bone and megakaryocytes (Lehnert and Akhurst, 1988; Wilcox and Der-ynck, 1988), which correlates with immunohistochemical investigation showing that TGF β1 is closely associated with connective tissue, cartilage, bone and tissues derived from neural crest mesenchyme (Heine et al. 1987). In situ hybridisation studies have demonstrated TGF β2 mRNA expression in various embryonic tissues of mesenchymal origin (Pelton et al. 1989). In the present study, we have investigated the spatial and temporal expression patterns of TGF β1, TGF β2 and TGF β3 by in situ hybridisation in histological sections of mouse embryos from day 10.5 p.c. to day 16.5 p.c. We have used homologous ³⁵S-labelled riboprobes, which were identical in length and were complementary to the DNA sequences encoding the mature forms of all three TGF β peptides.

Embryos, placentas and whole deciduas were isolated from RB (4.15) 4 RM A mice (Jackson Laboratories) at the times indicated in the text. Midday of the day of vaginal plug appearance was considered as day 0.5 post coitum (p.c.). Samples were fixed overnight in PBS at 4°C in a freshly prepared solution of 4 % paraformaldehyde and were then placed overnight at 4 °C in 0.5 M sucrose in PBS before storage in liquid nitrogen. Prior to sectioning, embryos were embedded in OCT compound (Miles). 10 /mt cryostat sections were placed on 3-aminopropyltriethoxy-saline-treated slides (Rentrop et al. 1986) and stored at –70 °C. Prior to hybridisation with RNA probes, the sections were dried for 5 min on a heated plated at 50°C, postfixed with 4 % paraformaldehyde in PBS for 5 min, rinsed in PBS and H₂O, depurinated for 20min with 0.2 N HC1 at room temperature, treated for 30 min with 2×SSC at 70 °C, dehydrated with increasing ethanol solutions and finally air dried. All solutions were treated with 0.1% diethylpyrrocarbonate and autoclaved.

‘Sense’ and ‘antisense’ RNA probes were labelled with α³⁵S-UTP (1200 Ci mmol^-1, New England Nuclear) to a specific activity of >10⁹disints min^-1 μg using SP6 or T7 RNA polymerase and according to the suppliers directions (Boehringer Mannheim). The TGF β riboprobe templates were 339-nucleotide long fragments, subcloned into pGEM5 (Promega, Biotec) and corresponded to the cDNA sequences encoding the mature forms (plus stop codon) of murine TGF β1, TGF β2, and TGF β3. The α-fetoprotein riboprobe template was a 900-nucleotide long Pstl fragment of the murine α-fetoprotein cDNA subcloned into pSPT18 (Boehringer Mannheim).

Prehybridisation was performed at 54 °C for 3h in 50% formamide, 10% dextransulfate, 0.3M NaCl, 10mM Tris, 10 mM sodium phosphate pH 6.8, 20 HIM dithiothreitol, 0.2×Denhardt’s reagent, 0.lmgml^-1E. coli RNA, and cold 0.2 μM α-S-UTP. Hybridisation was carried out overnight in the same mix supplemented with 2×10^scts min^-1/μl^-1 αS-UTP-labelled RNA probe in a humidified chamber at 54°C. Slides were washed in hybridisation solution without dextransulfate, RNA and ‘cold’ UTP containing 50% formamide and 10 mM dithiotreithol at 55 °C two times for 1 h and equilibrated for 15 min in a buffer solution consisting of 0.5 M NaCl, 10mM Tris, ImM EDTA pH 7.5. Sections were then treated with 50 μg RNA ase A in equilibration buffer for 30 min at 37°C to remove any non-specifically bound probe. Slides were washed in 2×SSC for 1h and then in 0.lxSSC for 1h at 37°C. Sections were then sequentially dehydrated in 65 %, 85 % and 95% (v/v) ethanol solutions containing 300 mM ammonium acetate and absolute ethanol before being air dried. Following X-ray autoradiography, the sections were coated with a 1:2 dilution of Ilford K5 photoemulsion, air dried and exposed for two weeks in a light-safe box containing silica gel at 4°C. Slides were developed in D19 developer (Kodak), fixed in AGEFIX LIQUID (AGFA) and stained either with Giemsa or with haematoxylin/eosin.

RNA was isolated from day 13, 14 and 16 p.c. embryos and newborn mice by homogenisation of tissues in 4 M guanidinium thiocyanate, 0.5% sarkosyl, 0.1M mercaptoethanol and 25 mM sodium citrate pH 7 using a Brinkman Polytron (Chomczynski and Sacchi, 1987). Poly(A)⁺ RNA was purified by oligo(dT)-cellulose chromatography (Maniatis et al. 1982). Samples of mRNA (10 μg) were electrophoresed through 0.8% agarose-formaldehyde gels and blotted onto Zetra-probe membranes (BioRad) according to the protocol of the manufacturer. Hybridisations were performed with TGF β1, β2, and β3, cDNAs corresponding to the mature form peptide of the mouse. Hybridisations were performed at 42°C in 40 % formamide, 5×SSC, 1% SDS, 5× Denhardt’s reagent, 200μgml^-1 tRNA (Maniatis et al. 1982). After each hybridisation the blot was stripped by heating to 90°C in 0.2×SSC/l% SDS for 15 min and rehybridised with the next cDNA probe.

Northern blot analyses of mouse embryo polyadenylated RNAs for TGF β1, β2 and β3 are shown in Fig. 1. The following characteristic bands were seen: TGF β 1, 2.4 kb; TGF β2, multiple bands from 6 to 3 kb (a predominant band at 4.4kb); TGF β3, 3.5 kb, thereby confirming the specificity of the probes for each of the three isoforms. The difference in signal intensity of embryo day 16 RNA is due to a smaller quantity being loaded as determined by u.v. shadowing and by probing a constitutively expressed mRNA, the eukaryotic protein synthesis initiation factor gene eIF-4A (Nielsen et al. 1985).

In situ hybridisation analysis performed with antisense RNA probes revealed complex hybridisation patterns indicating that all three TGF β genes are expressed in a distinct and unique pattern in the developing mouse embryo. In contrast, the sense probes that were used as negative controls produced a weak and uniformly distributed non-specific background signal (data not shown). At day 10.5 p.c., TGF β1 transcripts were clearly discernible in different tissues. We also detected moderate levels of TGF β2 expression which were confined to the myocardium and the endocardial cushion tissue (data not shown). No TGF β3 hybridisa-tion signals were observed; although, as reported by Miller et al. (1989b), Northern blots of polyadenylated RNA from day 10.5 p.c. embryos revealed low signals.

In the placenta, only TGF β1 and TGF β3 were expressed at all developmental stages investigated. At day 10.5 p.c. TGF β1 and TGF β3 transcripts revealed distinct patterns of expression (Fig. 2A – C). TGF pl expression was strong in mesenchymal cells forming the vascular zone near the central artery of the maternal part of the placenta (Fig. 2A). In the fetal part of the placenta, high levels of TGF β 1expression were detected in a small number of cells scattered throughout the connective tissue of the chorion. These strongly labelled cells had the appearance of blood cells and were larger in size than the connective tissue cells of the chorion (Fig. 2D). In contrast, high levels of TGF β3 transcripts were found exclusively in the spongiotro-phoblast, a cell layer of fetal origin that forms the junctional zone between the chorionic villi and the maternal blood vessels (Fig. 2C). These patterns of expression did not change during placental development although there was an apparent decrease in signal intensity (data not shown).

From day 10.5 p.c. to day 16.5 p.c., TGF β1expression was strongest in the developing liver. To more clearly define the cells expressing high levels of TGF β, we compared the pattern of TGF β expression with the pattern of α-fetoprotein expression. During embryogenesis, formation of the liver becomes evident around day 9 as a thickening and progressive stratification of a region on the ventral side of the foregut, at which point high levels of α-fetoprotein transcripts are present (Schmid and Schulz, 1990), but no TGF β1transcripts are detected (data not shown). During the next stage of liver development, strands of epithelial cells invade the surrounding mesodermal mesenchyme and eventually form a three-dimensional network. Hematopoietic precursor cells invade the liver at this stage and become established (Moore and Johnson, 1976). At day 10.5 p.c., both α-fetoprotein and TGF β1 were expressed at high levels but showed a distinct cellular distribution (Fig. 3D-E). Both a-fetoprotein mRNA (Schmid and Schulz, 1990) and protein (Dziadek and Adamson, 1978) are confined to the endodermal population of prehepatocytes whereas TGF β1signals were restricted to the mesoderm. Our findings strongly suggest that the

TGF β-expressing cell population represents mesodermal hematopoietic precursor cells, which arise in the liver at this stage. Later in liver development the pattern of TGF β1 expression became more dispersed. On day 12.5 p.c., megakaryocytes scattered throughout the fetal liver were labelled strongly by the TGF β1 probe (Fig. 7A), but not by either TGF β2 or TGF β3 probes. Other hematopoietic precursor cells showed moderate levels of expression. The endodermal pre-hepatocytes, which represent about 40% of the liver cells at this stage, were only weakly labelled. Between days 14.5 and 16.5 p.c., TGF β1 expression remained very strong in the megakaryocytes but decreased in other cells of the fetal liver until on day 18.5 p.c. TGF β1 transcripts were detectable only in the large megakaryocytes (data not shown). In contrast, TGF β2 and TGF β3 mRNAs were not expressed either in the hematopoietic precursor cells or in the prehepatocytes of the fetal liver at all stages investigated; however, moderate levels of TGF β3 transcripts were visible in the mesenteric epithelium surrounding the fetal liver on day 12.5 p.c. (Fig. 4C).

At day 10.5 p.c., TGF β1expression was visible in blood corpuscles of the visceral yolk sac. The first blood cells are produced in extraembryonic sites as small groups of mesodermal cells located next to the endodermal wall of the visceral yolk sac are thought to constitute a primitive population of erythrocytes (Die-terlen-Lievre, 1984). Afterwards, a new population of hematopoietic stem cells arises from the intraembryonicmesoderm and produces a population of blood cells which progressively substitutes for the yolk sac-derived blood cells. The early embryonic visceral yolk sac shares some properties with the embryonic liver during later stages of development such as hematopoiesis and production of α-fetoprotein. TGF β1is expressed in the hematopoietic islands of the yolk sac but only in a small number of cells; however, in these cells there is an extremely high level of expression (Fig. 2E). It therefore appears that TGF β1expression is restricted to a specific cell type during early hematopoiesis. TGF β2 and TGF β3 transcripts were not detected.

At day 10.5 p.c., we found TGF β1expressing cells scattered throughout the embryonic mesenchyme with transcripts most abundant in the mandibular and maxillary arches and in single cells surrounding the somites (Fig. 3A) and the nervous tissue (Fig. 3F-G). Due to their distribution we assume that these strongly labelled cells are neural crest cells, which are endowed with the ability to undertake extrensive but tightly controlled migrations throughout the embryo. TGF β2 and TGF β3expression was not observed in the neural crest mesenchyme.

The embryonic thymus contains a heterogenously distributed population of stromal cells and T-cells at various stages of maturation. TGF β1transcripts were seen in all cells of the embryonic thymus but the hybridisation signals were interspersed with T-cell precursors, which were apparently stronger labelled than the endodermal cell population (Fig. 7B). TGF β2and TGF β3 transcripts were not found in the developing thymus.

On day 12.5 p.c., high levels of TGF β2and TGF β3transcripts were visible in the primordia of the vertebral column and were strongest in the caudal sclerotomic halves. TGF β3 was expressed in all prevertebral segments (Fig. 4C). In contrast, high levels of TGF β2transcripts were visible only in the thoracic sclerotomes (Fig. 4B). TGF β1transcripts were detected in intersegmental cells of the early spinal column and in cells lining the cerebral ganglia (Fig. 4A). On day 14.5 p.c., the TGF β1probe strongly labelled the narrow bands of osteoblasts adjacent to the vertebrae and ribs (Fig. 5A). However, less differentiated elements of the developing axial skeleton, such as sternum and limb skeleton, were only weakly labelled by the TGF β1probe (Fig. 5A). In contrast, TGF β3 expression was observed in the perichondrium of all cartilaginous elements of the axial skeleton but the levels of expression varied (Fig. 5C). Within the vertebral column, TGF β3 expression was strongest in the intervertebral discs (data not shown). Only low levels of TGF β2 mRNA were visible in perichondreal mesenchyme (Fig. 5B).

On day 16.5 p.c., TGF β1expression was strong in the vertebrae and ribs and was detected both in the periosteal layer and in the ossification centres (Fig. 6A) where only very low levels of TGF β3 transcripts were visible (Fig. 6B). On the other hand, TGF β2 was expressed in the perichondrium of the limb skeleton (Fig. 6B). TGF p2 transcripts were only visible in growth zones of the limb plates (data not shown).

All components of the skeleton are derived from mesenchyme. Mesenchyme can be converted into skeletal elements by first forming a cartilaginous framework, which is subsequently replaced by true bone (endochondral bone formation), or by forming bone directly (intramembranous ossification). Most elements of the facial skeleton, such as mandible and maxilla, are formed by intramembranous ossification of the cranofacial mesenchyme. Highly contrasting patterns of distribution of TGF βl and TGF β3 transcripts were observed during formation of the facial skeleton elements. TGF β1 was expressed at very high levels in ossifying tissues of the upper and lower jaw (Fig. 5A, 6A and 8A) whereas TGF β3expression was strongest in the undifferentiated mesenchymal cell layers adjacent to the ossification centres (Figs 5C, 6B and 8C). TGF β2 expression was not observed in the upper and lower jaw regions that were undergoing intramembranous ossification. However, TGF β2 transcripts were visible in the perichondreal layers of certain cartilaginous elements of the facial skeleton and also in the nasal and mandibular mesenchyme that forms the soft tissue components of the face (Fig. 5B and 8B).

Formation of the lung begins around day 10 p.c. with the outgrowth of the foregut-derived endodermal trachea into the bronchial mesoderm. As the trachea lengthens, it bifurcates at its caudal end to form two lung buds. These in turn continue to grow and branch, giving rise to the bronchial trees of the lung. All three TGF β genes were expressed at high levels and showed distinct spatial and temporal patterns of mRNA distribution. At all stages investigated, TGF β1expression was strongest in the bronchial mesoderm (Figs 4A, 5A and 6A) whereas TGF β2 transcripts were found exclusively in the endodermal bronchiolar epithelia where the signal became stronger in later stages of development (Figs 4B and 5B). The TGF β3 expression pattern changed during lung development. At day 12.5 p.c. TGF β3 transcripts were found predominantly in the tracheal mesenchyme (Fig. 4C) but at day 14.5 p.c. TGF β3 signals were visible in the endodermal epithelia cells of the growing bronchioles (Fig. 5C) although by day 16.5 p.c. TGF pi expression was no longer detectable (Fig. 6B). TGF β3 transcripts were also expressed in mesodermal epithelial cells, which later give rise to the visceral pleura (Fig. 5C).

TGF β1 expression was strongest on day 14.5 p.c. in the mesodermal cell layers of the submucosa but not in the intestinal epithelia (Fig. 5A) although it was visible at later stages of gut development at decreased levels. In contrast, TGF β2 was expressed exclusively in the single-cell layer forming the mesodermal mesentera (Figs 4B and 5B). On day 12.5 p.c. moderate levels of TGF β3 transcripts were visible in both the submucosal mesenchyme and mesenteric mesoderm (Fig. 4C) but later became confined to the mesentera (Figs 5C and 6B).

Kidney development is the result of reciprocal inductive interactions between the metanephric duct and the surrounding metanephrogenic tissue. The terminal portions of the metanephric duct induce the formation of metanephric tubules. TGF β1 expression was detected in the surrounding stromal mesenchyme during formation of the metanephros at day 14.5 p.c. (Fig. 5A). The epithelial tubules deriving from the metanephrogenic mesenchyme showed only low levels of TGF β2 transcripts although no TGF β3 expression was observed.

From day 12.5 p.c. to day 16.5 p.c., the most prominent site of TGF β2 expression was the cochlear epithelium (Figs 5B and 7C). At day 16.5 p.c., TGF β2 hybridisation signals were also visible in the olfactory epithelium of the nasal chamber (Fig. 8B). No expression of TGF β1or TGF β3 was found in these tissues.

Early mouse embryo skin has a simple epithelium. By 16 days p.c. a second skin layer, the stratum germinativum, arises followed by an interstitial basal layer. At this stage, only TGF β1 expression was detected in the loose mesenchyme underlying the epidermis (Fig. 6A). Within the dermis a small number of cells showed high levels of TGF β1 expression and were scattered between moderately or weakly labelled mesenchymal cells. TGF β2or TGF β3 signals were not detectable in the skin.

Formation of the facial hair follicles is visible as early as day 13 p.c. when the vibrissae papillar are still invaginated and ectoderm is proliferating in the external nares. On day 14.5 p.c., all three TGF β transcripts were expressed at high levels in the ectodermal epithelium of the external root sheath (Fig. 5A-C). In the primordia of the dermal sheaths only moderate TGF β1 expression was detected. On day 16.5 p.c., TGF β1, β2 and β3 transcripts were visible both in the inner and outer root sheath of the whisker hair follicles but with striking differences in expression levels. TGF β3 was expressed stronger than TGF β1 and β2. Low levels of TGF β1 were also visible in the connective tissue sheath of the hair follicle; however, TGF β2 and β3 were not expressed.

In this report, we have described the first comparative study of TGF β gene expression during mouse embryogenesis where homologous gene probes of the same length have been used. We clearly demonstrate that TGF β1, β2 and β3 show distinct patterns of expression during development.

TGF β1transcripts are expressed predominantly in hematopoietic cells of the fetal liver, in fetal bone and in the mesenchymal compartments of several internal organs. These results are in agreement with the work of Lehnert and Akhurst (1988). In addition, we have observed on days 10.5 p.c. and 12.6 p.c. expression of TGF β1transcripts in mesenchymal cells of probable neural crest origin that line the nervous tissue and somites. Immunohistochemical studies of Heine et al. (1987) have shown a similar distribution pattern for the protein. In addition, a strong TGF β1 signal was clearly seen in the maternal decidua. In the fetal part of the placenta and in the haematopoietic islands of the yolk sac, a small number of cells showed very strong TGF β1 expression. This cell population probably represents precursors of the Hofbauer cells, which are believed to be primitive macrophages. Since peripheral blood monocytes and mature macrophages produce only the TGF pi isoform (G. Bilbe, unpublished results), it is likely that TGF β1 is switched on at an early stage in myeloid differentiation. Interestingly in vitro and in vivo studies on bone marrow preparations have shown that TGF β1 inhibits the proliferation of myeloid progenitor cells while more differentiated cells of this lineage are not inhibited (Keller et al. 1990; Ruscetti et al. 1990).

For the first time, our study demonstrates that TGF β2 is expressed in the mesenchymal components of the facial region and also in bronchial epithelia. Moreover,

TGF β2 was the only isoform expressed in the cochlear epithelium. We were unable to confirm the general findings of Pelton et al. (1989). The differences in expression patterns can most likely be explained by the choice of probes. Pelton and co-workers performed their in situ hybridisation experiments with a probe corresponding to the 5’ untranslated sequences and part of the LAP region of the human TGF β2 gene. Because the 5’ untranslated region of the human TGF β2 transcript is very rich in A and T residues we used a riboprobe that is complementary to the mature mouse TGF β2 peptide and which in Northern analysis gave a typical TGF β2 banding pattern. To confirm our results, we compared by in situ hybridisation the TGF β2 expression patterns revealed by the ‘mature form’ probe with those shown by a probe covering either the entire coding region or the LAP region. All probes showed identical patterns of hybridisation but with different intensities because of the varying lengths of the probes. We propose that the 5’ untranslated region of the human TGF β2 gene as used by Pelton et al. (1989) contains cross-hybridizing sequences responsible for these contrasting results. It is interesting to note that a comparison of the TGF β1, β2 and β3 DNA sequences encoding the mature peptide forms reveals homologies of less than 75 % between the three isoforms where the longest contiguous stretch of nucleotides is 14 bases long.

Our study also describes the expression patterns of TGF β3 by RNA in situ hybridisation. Northern blot analysis with mRNA from day 13 to day 16.5 p.c. hybridised with the mouse TGF β3 probe complementary to the mature peptide detected a single band. This finding is in agreement with Miller et al. (1989b) who found weak TGF β3 expression at day 10.5 p.c. although at this early stage we were not able to detect TGF β3 transcripts by in situ hybridisation. These transcripts are probably expressed at low levels and are dispersed throughout the embryo and therefore cannot be detected by in situ hybridisation. A unique TGF β3 hybridisation signal was also observed in the spongio-trophoblast of the placenta as early as day 10.5. p.c. and was predominantly restricted to the reticular zone between the fetal and maternal placenta. TGF β3 may therefore be involved in the regulation of cell proliferation and matrix formation during growth and stabilisa-tion of the chorionic villi in the placenta.

A comparison of the expression patterns for all three TGF βs strongly suggests a coordinated role for these proteins in mesenchymal-epithelial interactions during embryonic development. For example, there are marked differences in the pattern of TGF β1, β3 2 and β3 3 expression during branching morphogenesis of the lung. Endodermal branching, i.e. cleft formation, is one of the major events in lung morphogenesis. The characteristic budding pattern of the endodermal bronchial tree is the result of continuous interactions between endoderm and the surrounding mesoderm. The contrasting expression patterns of TGF β3s in the developing lung strongly suggest that these genes are involved in the budding process probably by controlling formation and degradation of extracellular matrix components. In vitro studies have shown that TGF β1 regulates the formation of extracellular matrix either by inducing synthesis of matrix proteins, such as collagen and fibronectin (Ignotz and Massague, 1986; Roberts et al. 1986; Varga and Jimenez, 1986; Fine and Goldstein, 1987; Ignotz et al. 1987) glycosaminoglycans (Rasmussen and Rapraeger, 1988) and cell adhesion receptors (Heino et al. 1989; Ignotz and Massague, 1987) or by controlling proteolytic degradation of matrix proteins (Laiho et al. 1986, Edwards et al. 1987; Keski-Oja et al. 1988). Recently colocalisation of TGF β1 protein and matrix proteins by a TGF β1 antibody was described in mesenchymal cells during lung development of the mouse, particularly at times when cell-cell interactions between mesenchyme and epithelium are important for normal epithelial cell differentiation (Heine et al. 1990). We have confirmed the localisation of TGF β1 mRNA in lung mesenchymal cells, which suggests that this peptide may induce matrix proteins in an autocrine fashion. Since exogenous TGF β1 has been shown to block the maturation of bronchial epithelial cells (Masui et al. 1986), this peptide may also regulate differentiation of the lung epithelia in a paracrine fashion. Whereas TGF β1 transcripts are localized in lung mesenchymal cells, TGF β2 and TGF β3 expression occurs only in the epithelial linings of the bronchii but TGF β3 not in the tips of actively growing ducts. A comparative study by Graycar et al. (1989) has recently shown more potent inhibitory effects of TGF β2 and β3 on lung epithelial cell proliferation than TGF β1 suggesting that the observed expression patterns in the developing lung may simply reflect differences in potency of all three isoforms.

We have shown that all three TGF β isoforms are expressed during endochondrial bone development but only TGF β1 and β3 are involved in intramembranous bone formation. Recent in vivo studies have shown that TGF β1 and TGF β3 induce the differentiation of periosteal mesenchymal cells of long bones into osteoblast and chondrocytes (Joyce et al. 1990). They also stimulate these cells to proliferate and synthesize the extracellular matrix proteins characteristic for bone and cartilage. These workers have found that TGF β2 is a more potent stimulator of osteogenesis and chondrogenesis than TGF β1 in vivo’, however, both isoforms share the ability to influence the route of tissue differentiation into bone or cartilage in a dosedependent way. In the present study, we have shown strong TGF β3 expression in precartilaginous masses and the perichondrium early in bone development. Therefore, we propose that TGF β3 may be involved in early differentiation processes in endochondral bone formation. In contrast, TGF β3, which is expressed predominantly at later stages of vertebral development, may induce matrix formation in differentiated skeletal elements. Since expression of TGF β2 transcripts were only observed at a very specific, early stage of bone development, this peptide possibly induces differentiation of mesenchymal cells into a chondriocytic phenotype. Absence of TGF β2 at this stage of development may result in ossification without a prior cartilage matrix.

Most of the flat bones of the face are derived from neural crest mesenchyme and their morphogenesis and growth is a very complex process involving the interaction of many factors (Noden, 1984). The present study shows that both TGF β1 and TGF β3 are involved in this process. Possibly TGF β3 induces differentiation of neural crest mesenchyme whereas TGF β1 expression is necessary for inducing bone matrix proteins during further maturation and ossification. Interestingly TGF β2 transcripts were never detected at any stage during intramembraneous bone formation although they were coexpressed with TGF β3 in some tissues of the axial skeleton.

The present report describes the coexpression of TGF β2 and TGF β3 in certain tissues, e.g., in lung epithelia, and coordinated expression of all three isoforms in others, e.g., in root sheath epithelia. The differential expression of the three genes may be explained by an analysis of their promoter regions (Roberts and Sporn, 1990b). There are two major classes of TGF β transcriptional promoters. The TGF β1 promoter contains and AP-1 binding site, responds strongly to phorbol ester induction but does not contain a TATA A box consensus sequence. On the other hand, TGF β2 and β3 have TATAA boxes, AP-2 binding sites and cyclic AMP responsive elements which can be induced by forskolin, an activator of adenylate cyclase. In addition, TGF β2 also has an AP-1 binding site (Roberts and Spom, 19906) suggesting that TGF β2 gene expression could be regulated differently from that of TGF β3. Since we do find expression of all three isoforms transcripts, one can hypothesise that coordinate binding’toAPl and AP2 sites must occur if all three genes are to be expressed.

What is the biological significance of the complex patterns of TGF β expression? Our overall impression is that TGF β1 is involved in early (major) inductive event(s) during interaction of mesenchymal and epithelial cell layers. TGF β1 is the only isoform expressed in early mesenchyme at enhanced levels and appears to regulate the layout of the embryo by stimulating the formation of extracellular matrix components which are important in cell migration and/or cell-cell interaction (Bemfield, 1981). For instance, in early fetal liver, lung and gut, TGF β3 expression is confined to mesenchymal cells. Heine et al. (1990) have shown that TGF β1 regulates the expression of collagen I and III, fibronectin and glycosaminoglycans, which are all components of the extracellular matrix. Production of these components is important for interaction with cell surfaces via specific receptor complexes and subsequent initiation of inductive processes resulting in differentiation events. In vivo experiments have demonstrated that TGF β1 has potent angiogenic activity (Roberts et al. 1986, Cox et al. unpublished data), which partly explains the extensive vascularisation and invasion by lymphatics and nerve fibers that occur during organogenesis. Further evidence for a pivotal role for TGF β1 in early mesenchymal-epithelial interactions has been provided by Antonelli-Orlidge et al. (1989) who have shown that TGF β1 mediates inhibition of cell growth in a capillary-endothelial cell coculture system.

As development enters a phase of morphogenesis characterized by more complex differentiation events, all three TGF β isoforms are expressed in differential patterns. This phenomenon can be seen clearly in developing bone where distinct TGF βs are expressed in a controlled fashion as described above. The differential expression of TGF βs at specific times during bone formation indicates differences in their biological functions. TGF β1 has been reported to stimulate or inhibit growth and differentiation in vitro depending on the cell type examined (Spom et al. 1987) and more specifically has been shown to inhibit osteoclast (Oreffo et al. 1989), and to stimulate osteoblast, proliferation (Centrella et al. 1986; Gehron-Robey et al. 1987). Antonelli-Orlidge et al. (1989) have demonstrated that conversion of the precursor TGF β molecule to the mature form is essential for its activity and that this process is mediated by the responder cells in a paracrine fashion. Thus, during bone development the differential expression of TGF βs affects the developmental fates of cells partly as a result of precursor conversion. However, Graycar et al. (1989) have demonstrated that the potency of the mature peptides differs depending on the cell type affected. This suggests that those cells under the regulation of TGF βs have receptors or receptor-coupled second messenger systems that determine the inductive or inhibitory effect of TGF βs. For example, a guanine nucleotide-binding protein-dependent pathway is involved in transmission of the signal for at least one TGF β -induced response (Howe and Leof, 1989). Finally TGF β can control morphogenesis at three different levels. Firstly, at the transcriptional level via transcription factor-promoter interactions. This determines the type of TGF β isoform expressed and is dependent on the cell type and its differentiation stage. Secondly, recent studies have indicated that the responding cell is crucial for activation of TGF β to its active form. Thirdly, the response of the affected cell is dependent on the TGF β receptors and their second messenger systems.

Using our in situ hybridisation conditions, the human TGF β1 probe, chosen by Pelton et al. (1989), revealed a very high unspecific background, probably due to the reasons stated in the text.