The effect of retinoic acid on chondrogenesis in the fetal hamster tibia in vivo

Since the description by Kochhar (1967) of the teratogenicity of retinoic acid, it has been widely utilized as an experimental teratogen. Kochhar (1973) described the relatively specific teratogenic effect of retinoic acid on the appendicular skeleton of the mouse fetus, while Shenefelt (1972) produced similar defects in the hamster. The mechanism whereby retinoic acid produces these skeletal deformities has been investigated using mainly in vitro techniques. The 24 h immediately following retinoic acid administration have been studied in particular detail. Kwasigroch & Kochhar (1975) showed that retinoic acid administration results in decreased cell motility and Kochhar (1977) demonstrated an increased labelling index and cell death in the limb-bud mesenchymal cells. These effects are produced before the mesenchymal cells have begun to differentiate morphologically into chondroblasts. Kochhar (1977) also described the differentiation of limb chondroblasts and cartilage matrix production in organ culture after exposure to retinoic acid. The morphogenesis of cartilage in limbs from fetuses exposed to retinoic acid, developing in vivo, has yet to be reported.

Because retinoic acid administered orally to pregnant hamsters at the appropriate dose and time results in a high percentage of limb malformations associated with few fetal resorptions (Shenefelt, 1972), it presents an excellent model with which to examine early skeletal dysmorphogenesis in vivo.

Although normal limb-bud cartilage that is developing in vitro when examined with the light microscope, resembles that differentiated in vivo, the former does not proceed to the stage of mineralization (Neubert, Merker & Tapken, 1974). To understand more completely the mechanism of limb maldevelopment associated with retinoic acid administration, it is necessary to compare in vivo development with reported in vitro changes.

In this paper we describe chondroblast development at the ultrastructural level in fetal hamster limbs after maternal exposure to teratogenic doses of retinoic acid. The study concentrated on the 24 h period of development from time of chondroblast differentiation until just prior to calcification of the cartilaginous skeletal limb blastema so that comparisons with previously reported work may be made. Because normal chondrogenesis has been described in detail (Goodman & Porter, 1960; Goel, 1970; Searls, Hilfer & Mirow, 1972; Thorogood & Hinchliffe, 1975) a comprehensive report is not presented here; rather the differences between normal and retinoic acid affected chondroblast differentiation are given.

Female golden Syrian hamsters weighing 90–100 g (High Oak) were housed in a temperature controlled room with a 12 h light cycle. At 20.00 on the predetermined evening of estrous, they were individually mated for 2 h with a single male. This time of observed mating was considered zero time of pregnancy. The hamsters were housed in separate cages for the duration of the experiment. At 20.00 on day 10 of pregnancy, hamsters were administered orally 80 mg/kg body weight retinoic acid (Sigma) suspended in 0·5 ml corn oil (Fisher). Control animals received 0·5 ml corn oil.

At 24, 36 and 48 h after treatment, animals were killed by decapitation. The fetuses were removed from the uterus and fixed in a solution of 2% glutaraldehyde in phosphate buffer (pH 7·4) for 30 min, washed 30 min in phosphate buffer and post-fixed for 2 h in 1% osmium tetroxide in the same buffer. Limbs were dehydrated in ethanol and embedded in Spurr’s low viscosity medium (Spurr, 1969).

Thick sections (0·3–0·5 μm) were cut and stained with methylene blue. It was important to orientate properly the block so similar areas in treated and control tibiae could be compared. Thin sections of silver interference colours were cut on an LKB ultratome III, mounted on copper grids, stained with 3% uranyl acetate and lead citrate (Reynolds, 1963) and examined in an Hitachi HU IIE electron microscope.

The preaxial hindlimb-bud mesenchyme consisted of randomly orientated cells not yet formed into a condensation (Fig. 1).

Cells maintained contact through filopodia or occasionally along surface membranes of contiguous cell bodies. These latter contacts were up to 1–2 μm in length. The nucleus was large, containing a dark-staining nucleolus. Mitochondria were few, RER was sparse and the Golgi complex consisted of a few lamellae and vesicles in a cytoplasm containing many free ribosomes. Fibrils and granules were rarely seen in the intercellular space.

Cells in the central area of the tibial blastema resembled typical chondroblasts (Fig. 2 A) with cell surfaces having an irregular outline. Intercellular contacts were rarely seen, and when present were along short lengths (1 μm) of cell membrane or through pseudopodial processes. The RER was well developed, consisting of dilated cisternae up to 0·3 μm in width containing granular, electron-dense material. The Golgi apparatus consisted of elongated, parallelly arranged lamellae, small vesicles (60–70 nm in diameter) and larger vacuoles up to 300 nm in diameter containing electron-lucent material. Occasionally electrondense osmiophilic material, resembling lipid droplets, was seen in the cytoplasm. Electron-dense granular material, probably glycogen, was abundant in some chondroblasts; however, it was not consistently observed in cells of all tibiae examined.

The matrix contained randomly orientated fibrils varying between 4 and 8 nm in width. Dense granules 4–15 nm in diameter were scattered throughout the intercellular space. They were often associated with, or directly attached along, the length of the fibrils.

The main difference between control and retinoic acid-treated tibial blastema was the nature of the cell-to-cell contacts. The chondroblasts in treated limbs were more closely packed than those of controls, with contiguous cell membranes about 30–40 nm apart over several microns of cell surface (Figure 2B). Also the cells often made contact through pseudopodial processes similar to that seen in day-10 tibiae.

The RER had components similar to that of control chondroblasts, with cisternae containing electron-dense granular material. The Golgi apparatus consisted of lamellae, vacuoles and vesicles, but did not appear as well developed as in controls. Glycogen, represented by masses of electron-dense intercellular granules, was sometimes present in large amounts; in other samples there was a paucity of glycogen. Mitochondria were numerous and occasionally appeared swollen.

Intercellular matrix was not as abundant as in controls, and contained fewer scattered fibrils and granules. Fibrils varied from 4 to 12 nm in width, and were occasionally grouped in bundles up to 80 nm in width. Matrix granules were scarce and not associated with the fibrils to the extent seen in controls.

Cells of the tibial maturation zone rarely contacted each other and had deeply scalloped cell outlines. The cytoplasm contained dilated RER and Golgi consisting of lamellae, vesicles and vacuoles (Fig. 3 A). Golgi vacuoles containing coarse granular material (chondrogenic vacuoles) were often seen (Fig. 4 A). Granular glycogen deposits were present in many cells. The extracellular matrix contained fibrils, 6–20 nm in width, scattered throughout the intercellular space (Fig. 3A, inset). Granules up to 30 nm in diameter were found associated with the fibrils and scattered throughout the matrix.

Cell contacts were often maintained by cell processes (Fig. 3B); in some cases contiguous cell membranes were apposed for several microns. Similar areas of cell contact were not seen in control tibiae. Cell surfaces did not have a deeply scalloped, indented appearance.

Dilated RER containing granular electron-dense material, Golgi vacuoles, vesicles and lamellae (Fig. 4B) were present in the cytoplasm, although not as well developed as in controls. Chondrogenic vacuoles, similar to those seen in control tibial chondroblasts were not often seen. Glycogen, represented by granular deposits, was frequently observed.

The intercellular space contained fibrils and granules. The fibrils, however, were generally wider (average 30 nm) and were often clumped together to form aggregates up to 100 nm wide (Fig. 3B, inset). These thicker fibrils and fibril aggregates were not associated with granules to the same degree as were fibrils seen in control tibia. The granular component of the matrix was reduced in amount with granules ranging in size from 6 to 20 nm scattered throughout the intercellular space.

Cells and matrix in the zone of proliferation and maturation of the tibial blastemata were generally similar to those seen at day (Fig. 5 A). The rough endoplasmic reticulum and Golgi apparatus were well developed, and the cells had an irregular outline. The intercellular space contained fibrils and granules closely associated with each other.

The central, most differentiated cells of the tibial blastemata were in most respects comparable to those in the maturation zone of control tibiae (Fig. 5B). The Golgi apparatus and RER were well developed. Occasionally cells with swollen mitochondria were seen. The cells were surrounded by matrix consisting of fibrils and granules, but not as abundant as in controls. Intercellular contacts, as seen 12 h earlier, were not seen as often.

Retinoic acid, administered in teratogenic doses, results in aberrant mesenchymal condensation and cell death in the skeletal limb blastemata during the 24 h following treatment (Kochhar, 1970, 1977). These detrimental effects occur before large amounts of cartilaginous matrix accumulate. Cell death was not observed to any extent in the present study, probably because this study concentrated on chondrogenesis in the 24–48 h period following retinoic acid administration. Cell death is observed mainly prior to this time (Kochhar, 1977). Results of the present paper show that, in vivo, subsequent chondroblast differentiation and matrix morphology is affected in retinoic acid-treated fetuses. It has been proposed that retinoic acid has a solubilizing effect on cartilage matrix in vitro similar to that of vitamin A, which has been thought to be due to release of lysosomal enzymes into the intercellular space (Fell, Dingle & Webb, 1962; Goodman, Smith, Hembry & Dingle, 1974). In the present study, and those of Kochhar (1970, 1977), the administration of retinoic acid coincided with mesenchymal condensation of the limb skeletal blastemata before matrix had begun to accumulate. Recently, Hassell, Pennypacker & Lewis (1977) presented evidence that glycosylation in limb-bud cartilage cultured in the presence of retinoic acid, is inhibited, resulting in decreased proteoglycan synthesis. They found no evidence of increased proteoglycan degradation. It is therefore improbable that the matrix labilizing effect of retinoic acid mediated by lysosomes is the initial cause of the skeletal dysmorphogenesis in the fetal hamster tibia.

The decreased matrix accumulation, described by Kocchar (1970) and observed ultrastructurally in the present in vivo study 24 h after retinoic acid treatment, appears instrumental in production of the malformation. Both the interaggregate environment of chondroblasts (Ahrens, Solursh Reiter, 1977) and the polarity of the cells in the developing limb (Holmes & Trelstad, 1977) are important in chondrogenesis. In retinoic acid-treated limb-buds, neither the matrix in the intercellular space nor the orientation of the chondroblasts appear normal. Goel (1970) and Gould, Day & Wolpert (1972) suggested that following mesenchymal condensation, accumulation of matrix around the developing chondroblasts pushes the cells apart and may assist the skeletal blastema in assuming its proper shape. The decreased matrix in the extracellular space seen 24 h after retinoic acid treatment is associated with closer cell packing and intercellular contacts. The latter may be a result of insufficient matrix failing to push the chondroblasts apart (Thorogood & Hinchliffe, 1975). This may in turn prevent the cells from assuming their normal spatial relations with each other resulting in an apparent prolongation of the condensation phase of skeletal blastema development.

In the present investigation, 36 h after maternal retinoic acid administration, quantities of matrix had begun to accumulate in the extracellular space between the chondroblasts. However, the fibrillar component of the matrix consisted of wider fibrils than seen in controls and were often clumped together into aggregates which were relatively unassociated with matrix granules. A similar increase in fibril width was not described by Kochhar (1977) during in vitro development. The growth of collagenous fibrils is inhibited in the presence of glycosaminoglycans (Seegmiller, Fraser & Sheldon, 1971). Decreased granularity of the matrix in retinoic acid-treated tibial blastemata, unassociated with matrix fibrils, is morphological evidence of lack of glycosaminoglycan fibril interaction and could result in fibrils observed in the present study attaining a larger than normal size. Linsenmayer & Kochhar (1977) reported an increased fibril diameter in limb-bud matrix when glycosaminoglycan synthesis is inhibited, and suggested that the glycosaminoglycans may be a controlling factor in collagen fibrillogenesis. Because no quantitative results were obtained, one cannot be categorical in assessing the evidence adduced in the present study concerning the selective inhibition of glycosaminoglycan synthesis and the observed increased fibril width.

An alternate explanation to the glycosaminoglycan inhibition theory can be suggested: during chondrogenesis, presumptive chondroblasts are known to change from synthesis of type I to type II collagen. Von der Mark, von der Mark & Gay (1976a) have suggested that thicker collagenous fibrils in some forms of cartilage may be due to the presence of type I collagen. Perhaps the thicker fibrils seen in retinoic acid-treated tibiae are due to the presence of type I collagen in greater than normal amounts. Mesenchymal cells secrete type I collagen before differentiating into chondroblasts (von der Mark, von der Mark & Gay, 1976b) and retinoic acid treatment may slow their differentiation to the extent that thicker fibrils indicative of type I collagen are frequently seen in the matrix. Only biochemical analysis of collagen types, or application of immunofluorescent techniques (von der Mark et al. 1976b) will answer the question of whether an abnormal type of collagen is present in retinoic acidtreated skeletal blastemata and is responsible for the thicker fibrils unassociated with matrix granules.

Chondrogenic vacuoles (Goel, 1970) which in control chondroblasts were round, membrane-bound vesicles containing electron-dense granules, were relatively scarce in retinoic acid-treated limb chondroblasts. Seegmiller, Ferguson & Sheldon (1972) proposed that the relatively smooth surfaced outline of the chondroblasts seen in the mutant mouse chondrodysplasia, indicated lack of secretion by Golgi vacuoles failing to fuse with, and release their contents through, the chondroblast cell membrane. Similarly, smooth surfaced chondroblasts seen in the present study may be indicative of decreased matrix release into the extracellular space of retinoic acid-treated tibiae. The same morphological evidence is presented by Kochhar (1977) to explain decreased glycosaminoglycan production in retinoic acid-treated mouse limb-buds in vitro.

By day 12, 48 h after retinoic acid treatment, the cells of the tibial anlagen appeared to be recovering from the effect of the teratogen. While there was less matrix than in controls, the chondroblasts were isolated from each other and contained relatively normal appearing matrix producing organelles. It was important, however, to compare cells of similar zones in treated and control tibia. The recovery of the skeletal blastema from the teratogenic effect of retinoic acid is consistent with its rapid excretion (Kochhar, 1976).

The authors with to thank Mr H. Verstappen for preparation of the photographs and Mrs C. Cattermole for typing of the manuscript.