Retinoic acid-binding protein in the chick limb bud: identification at developmental stages and binding affinities of various retinoids

Vitamin A and its analogues, the retinoids, have long been a subject of research in the fields of vision, nutrition, metabolism and differentiation. Most recently they have aroused great interest because of their unique effects on pattern formation in developing and regenerating limbs. These compounds have the ability to change the positional information of cells in the limb in a very precise manner.

When retinoic acid (RA) is applied locally to the anterior side of the developing chick limb bud it causes mirror-image duplications in the anteroposterior axis such that a 6-digit double posterior limb develops instead of the normal 3-digit one (Tickle, Alberts, Wolpert & Lee, 1982; Summerbell, 1983). In developing amphibian limbs, retinyl palmitate causes duplications in both the anteroposterior axis and the proximodistal axis (Niazi & Saxena, 1978; Maden, 1983), the net result of which can be the production of a complete pair of limbs including the girdle growing out from the foot. The regenerating limbs of amphibians also duplicate in the proximodistal axis after retinoid administration (Maden, 1982; Thoms & Stocum, 1984) and can be made to reveal anteroposterior effects after appropriate surgery (Stocum & Thoms, 1984).

As far as mechanisms are concerned we have no idea of precisely what retinoids are doing to limb cells when they are respecified in this manner. However, in the chick limb bud, the striking similarity between retinoic acid effects and the result of grafting the zone of polarizing activity (Tickle, Summerbell & Wolpert, 1975) led to the suggestion that either retinoic acid itself is a morphogen or that it is acting locally to convert adjacent tissue into a signalling region (Tickle, 1983; Summerbell, 1983). The similarities extend not only to the obvious morphological structure, but also to dose dependency, position effects and time dependency. Most recently, data showing that retinoic acid becomes distributed across the anteroposterior axis of the chick limb bud in the same graded fashion as the putative morphogen released by the zone of polarizing activity have provided support for the former suggestion (Tickle, Lee & Eichele, 1985; Eichele, Tickle & Alberts, 1985). More than this we do not yet know, but some insights may be gained by looking at how retinoids are thought to act in other systems.

One view of the action of retinoids on cells is the steroid model whereby retinoids enter the cell, bind to a specific cytoplasmic receptor, are transported to the nucleus and change the pattern of gene transcription, which results in altered cell morphology (Roberts & Sporn, 1984). The involvement of specific cytoplasmic receptors, the retinoid-binding proteins, is central and considerable evidence has accumulated in support of this contention. For example, the distribution of cytoplasmic retinol-binding protein (CRBP) and cytoplasmic retinoic acid-binding protein (CRABP) in various tissues, mostly of mammalian origin, correlates with the requirement of that tissue for vitamin A (Chytil & Ong, 1984). CRABP is usually present at high levels in tumours, but at low levels or absent in the corresponding normal tissues (Ong, Page & Chytil, 1975; Palan & Romney, 1980; Mehta, Cerny & Moon, 1980; Ong, Goodwin, Jeese & Griffin, 1982). Terato-carcinoma cells and other cancer cell lines which normally respond to retinoic acid contain CRABP, but mutants selected for their lack of responsiveness have lost their binding proteins (Schindler, Matthaei & Sherman, 1981; Wang & Gudas, 1984; Marth, Mayer & Daxenbichler, 1984).

Thus an obvious question to ask with regard to the action of retinoic acid (RA) on the chick limb bud is whether it contains CRABP at stages that are susceptible to RA-induced pattern alterations. We report here the detection and partial characterization of CRABP from the cytoplasm of the cells of the chick limb bud, the binding of various analogues to this receptor and discuss its possible role in RA-induced pattern respecification.

Fertilized chick eggs from a mixed local flock (Needle Farm) were incubated to stages 20–35 (Fig. 1). At stages 20, 21, 22, 25, 30 and 35 forelimbs were removed from embryos and frozen in liquid nitrogen until enough material had been accumulated.

[ll,12-³H]all-/rans-retinoic acid (30 Ci mmol^-1) and the synthetic analogues Ro 10-9359 (etretinate: ethyl all-trans-9-(4-methoxy-2,3,6-trimethylphenyl)-3,7-dimethyl-2,4,6,8-nonatetra-enoate), Ro 10-1670 (all-irans-9-(4-methoxy-2,3,6-trimethylphenyl)-3,7-dimethyl-2,4,6,8-nona-tetraenoic acid), Ro 13-7410 (p-[(E)-2-(5,6,7,8-tetrahydro-5,5,8,8-tetramethyl-2-naphthyl)-l-propenyl] benzoic acid) and Ro 13-6298 (arotinoid: ethyl p-[(E)-2-(5,6,7,8-tetrahydro-5,5,8,8-tetramethyl-2-naphthyl)-l-propenyl] benzoate) were kindly provided by Hoffman-La Roche & Co. Ltd, Basel. 13-cis-retinoic acid was a gift from Dr C. Tickle, The Middlesex Hospital Medical School, London, and all other retinoids were purchased from Sigma. The purity of all-trans-RA and 13-cis-RA was determined by reverse phase HPLC using a μBondapak C₁₈ column with 0·01 M sodium acetate methanol: water 80:20 as the buffer (McCormick, Napoli & DeLuca, 1978).

Limb buds were homogenized in 1 ml of ice-cold phosphate-buffered saline (PBS = 0·03 M sodium phosphate buffer, pH7·2 in 0·1 M NaCl) in a Dounce homogenizer. Samples were then spun at 100000g for 1 h at 4 °C in a Spinco SW50.1 rotor. The protein content of the supernatant was measured with the Pierce protein assay reagent (Pierce Chemical Company, Rockford, Illinois) using bovine serum albumin as the standard. The preparations were then aliquotted and stored at –80°C until use.

Samples for assays contained about 400 μg of protein in 40–125 μl. Each was diluted to 500 with PBS, 66nM of [³H]all-ira7W-retinoic acid was added in 10μl dimethylsulphoxide (DMSO) with or without 100 x molar excess of unlabelled all-tras-RA, also in 10 gl DMSO. The mixtures were incubated at 4°C for 4h. They were then dialysed against PBS for 24h at 4°C, placed on 4·5 ml 5–20% sucrose gradients and spun at 180000g for 17 h in an SW50.1 rotor at 4°C using a Beckman L2-65B Ultracentrifuge. Gradients were separated into five drop fractions (180μl), the refractive index of each fraction was checked and radioactivity counted in 10 ml Liquiscint (National Diagnostics).

Bovine serum albumin (BSA) and myoglobin were used as relative molecular mass markers (4·6 S and 2S) on separate gradients, their positions on the gradient being determined by the OD₂₈₀ reading. Specific binding was demonstrated with the addition of 100 × molar excess cold all-iram-RA and quantified by determining the drop in cts min^-1 in the 2 S region of the gradient (between fractions with refractive index 1·3425 and 1·3475). The values obtained assumed a single binding site for retinoic acid and were for comparisons between stages and not intended to be absolute measures of the amount of CRABP in limb buds.

Protein samples (400 μg in 500μl PBS) were incubated at 4°C for 4h with increasing concentrations of ³H-RA from 1–220 nM with or without 100 × molar excess of unlabelled all-trans-RA. They were dialysed, spun and fractionated as described above. Specific binding was calculated and a Scatchard analysis performed.

For competition studies, protein samples (400 μg protein in 500μl PBS) were incubated for 4 h at 4 °C with 130 nM ³H-RA in 10 μl DMSO and from 0·33 to 500 × molar excesses of unlabelled analogues also in 10 μl DMSO. The analogues added were retinol, retinyl palmitate, retinyl acetate, retinal, 13-cw-RA, Ro 10-9359, Ro 10-1670, Ro 13-7410 and Ro 13-6298. Samples were then dialysed for 24 h, spun on a sucrose gradient and fractionated as described above. The decrease in binding caused by the analogues was quantified and plotted against concentration of ligand. The 50% intercept (IC₅₀) was used as a comparative value.

We began by examining stage 20 limb buds (Fig. 1), which are still susceptible to pattern changes induced by retinoic acid (Summerbell & Harvey, 1983). At this stage the limb bud forms a small protrusion from the flank of the embryo, consisting of an apparently homogeneous mass of mesenchymal cells surrounded by a thin layer of ectoderm. Cytoplasmic protein preparations incubated with 66 nM [³H]ZraHs-retinoic acid (³H-RA) showed a single peak of radioactivity on a 5–20 % sucrose gradient (Fig. 2A). The peak cosedimented with myoglobin run on a separate gradient, suggesting an approximate relative molecular mass of 17000, and could be abolished when incubated in the presence of 100 × molar excess cold trans-R A (Fig. 2A). This indicated the presence of a specific, saturable receptor for retinoic acid.

At later stages of limb development, differentiation begins and the characteristic shape of the limb is manifest (Fig. 1). These stages also have CRABP present (Fig. 2B–D). Indeed, at all stages tested (20, 21, 22, 25, 30, 35) the receptor was present at approximately similar levels per mg of cytosolic protein (Table 1), although there is some suggestion of a decrease at stages 25 and 30. These levels are similar to those found in embryonic chick skin (Maden & Summerbell, unpublished data), a tissue that is also susceptible to retinoid-induced pattern changes (Dhouailly, 1982). Thus CRABP is present during the early stages of limb development that are sensitive to the patterning effect of RA (around stage 20) and also at later stages when teratogenic effects generally appear after RA administration (Summerbell, 1983).

Scatchard plots of several sets of data from stage 25 and 30 CRABP gave estimates of 140–280 nM for the dissociation constant (K_d), which is at the upper end of the range of K_d values from other tissues. These estimates vary from 2UM (mouse limb buds: Kwarta, Kimmel, Kimmel & Slikker, 1985) and 4·2 nM (rat testis: Ong & Chytil, 1978) to 318 nM (rat testis: Bonelli & DeLuca, 1985) and 400 nM (human skin: Siegenthaler & Saurat, 1985). The total specific binding capacity was estimated to be 14–28 pmoles CRABP mg^-1 cytosol protein, the same value obtained from mouse limb buds (Kwarta et al. 1985). The plots were consistent with a single binding site.

The relative binding affinities of various vitamin A analogues to this chick limb bud CRABP were determined in competition experiments. We first examined the competition with 100 × molar excesses of each of nine analogues. Retinol, retinyl palmitate, retinyl acetate, etretinate (Ro 10-9359) and arotinoid (Ro 13-6298) did not compete for binding to CRABP (Fig. 3A; Table 2). Increasing the molar excess of arotinoid to ×300 or etretinate to ×500 did not reveal any competition either. The terminal end group of etretinate and arotinoid is –COOC₂H₅: thus neither esters nor the alcohol can compete for binding to CRABP.

On the other hand, retinal did compete, but not as effectively as all/rans-RA, whereas 13-cis-RA, Ro 13-7410 and Ro 10-1670 did compete as effectively as all-trans-RA at 100 × molar excess (Fig. 3A).Ro 10-1670 is the free acid of etretinate and Ro 13-7410 is the free acid of arotinoid. Thus the binding to CRABP is largely specific for an acid end group although the aldehyde can compete to a certain extent.

We then examined in more detail the relative efficacies of the binding of retinal, 13-cw-RA, Ro 13-7410 and Ro 10-1670. We began by determining the 50% displacement level (IC₅₀) of all-iraftf-RA. Low excesses (×0·33 or 1) decreased significantly the binding of label and the IC₅₀ was reached at a value of about × 3·5. When the displacement curve was plotted the molarity of all-irajw-RA at IC₅₀ was 1·9×10⁻⁷M (Fig. 3B; Table 2). The same experiment using Ro 13-7410 gave a slightly lower value of 1·0×10⁻⁷M, suggesting that this analogue may be more effective at binding to CRABP than its natural ligand. 13-cis-RA, the isomer of all-trans-RA, gave an IC₅₀ of 6·6×10⁻⁷M (Fig. 3B). This means that cis-RA is three to four times less effective at competing for sites on CRABP than all-trans-RA. Ro 10-1670 gave a value of 1·8×10⁻⁶M, which is about ten times less effective than all-irans-RA, and retinal gave a value of 2·2×10⁻⁶M, which is about twelve times less effective (Table 2).

We can therefore rank these analogues in order of binding efficacy to CRABP thus: Ro 13-7410 > all-irans-RA > 13-cw-RA > Ro 10-1670 > retinal and those with low or no affinity are retinol, retinyl palmitate, retinyl acetate, etretinate and arotinoid.

The graded response of the chick limb bud to RA, whereby more duplicated digits are produced as the dose increases suggested to Tickle et al. (1985) that a specific receptor protein may mediate the acquisition of positional information. Here we have demonstrated that there is indeed a specific cellular receptor for retinoic acid in the developing chick limb bud. It is present at least from stage 20 to stage 35, has an apparent K_d of 140–280 UM and shows specificity for an acid terminal group. But does this mean that CRABP plays a role in the mechanism of RA-induced pattern alterations? Several points can be made to support this contention.

First, we can estimate how the concentration of CRABP in the limb bud is related to the amount of RA needed to cause duplications. If RA is in an excess then it would be unlikely to act via CRABP because the receptor-ligand complexes would be saturated. Conversely, if the CRABP concentration is in excess compared to RA then it would be unlikely to be acting primarily as a receptor for RA. Tickle et al. (1985) have deduced that 7·6 pg of RA per limb bud gives a full duplication, which is 25 fmoles of RA. We estimate above that there is about 5 pmoles of CRABP per mg cytosol protein, which is 125fmoles CRABP per limb bud. Therefore the amount of RA needed to cause duplications lies on the linear part of the binding curve where the percentage of sites occupied is proportional to the concentration of RA.

Second, CRABP was present at high levels during those stages of limb development when RA changes the positional information of cells in the anteroposterior axis (around stage 20: Summerbell & Harvey, 1983). It is also present at similar levels in embryonic chick skin (Sani & Hill, 1976; Maden & Summerbell, unpublished data) and in this system RA changes the pattern of development by causing feathers to form instead of scales (Dhouailly, 1982). In the regenerating amphibian limb RA changes the positional information of blastemal cells (see Introduction) and the levels of CRABP are four times higher at RA-susceptible stages (Keeble & Maden, 1986a). These correlations suggest a role for CRABP in the mechanism of action of RA. Of course, it must be said that the mere presence of CRABP does not imply that pattern changes will, of necessity, occur. In the chick limb bud the capacity to regulate in the anteroposterior axis has almost disappeared by stage 24/25 (Summerbell, 1974) so despite the presence of CRABP at stages 25+, RA could not be expected to cause pattern alterations. CRABP has been identified in the mouse limb bud (Kwarta et al. 1985), but RA administration to pregnant mice only produces teratogenic effects on these embryonic limbs (Kochhar, 1977). Similarly in culture systems some cell lines are refractory to the effects of RA despite having normal levels of CRABP (Matthaei, Andrews & Bronson, 1983; Lacroix, L’Heureux & Bhat, 1984). Nevertheless, the evidence presented in the Introduction suggests that the presence of CRABP may well be a precondition for the action of RA.

Third, in both the chick limb bud (Maden & Summerbell, unpublished data) and the regenerating amphibian limb (Keeble & Maden, 1986b) we have shown that ³H-RA enters the nuclei of limb cells. Thus the suggested mode of action of RA, whereby it enters the cells, binds to CRABP, the CRABP-RA complex is transported to the nucleus and changes gene transcription, may well be operating to change the positional information of limb cells. Indeed, changes in protein synthetic profiles in RA-treated regenerating blastemas have been identified on 2D gels (Slack, 1983). We are currently exploring in greater detail this postulated mode of action.

Fourth, competition studies with retinoid analogues are performed to correlate binding activity with biological activity (Lotan, Neumann & Lotan, 1980; Trown etal. 1980; Sani et al. 1984). Such correlations, if they exist, provide evidence that CRABP mediates the biological effects of RA. Thus with chick limb bud CRABP we obtained the following binding efficacy: Ro 13-7410 > all-trans-RA > 13-cis-RA > Ro 10-1670 > retinal. Retinol, retinyl palmitate, retinyl acetate, etretinate and arotinoid showed little or no affinity (Table 2). So far we only know that Ro 13-7410 is more effective than all-trans-RA (Eichele et al. 1985), that all-trans-RA is more effective than 13-cis-RA (Tickle et al. 1985) and that retinyl palmitate does not appear to be active in the chick limb bud (Summerbell & Harvey, 1983). Ro 13-7410 is also the most active analogue in altering feather morphogenesis in chick skin (Cadi, Pautou & Dhouailly, 1984) and 13-cis-RA is less active than all-trans-RA as a teratogen of the mouse limb bud (Kochhar, Penner & Tellone, 1984). These limited biological efficacy data that we have fit well with the binding studies reported here although they ignore the possibility of differential uptake and differential metabolism of these retinoids by the cells of the limb bud (e.g. Eichele et al. 1985). They may be converted to active or inactive metabolites before binding to CRABP and thus invalidate the above comparisons. For example, etretinate can alter pattern formation in the regenerating amphibian limb, although it is less active than all-trans-RA (Kim & Stocum, unpublished data) but etretinate shows little or no affinity for amphibian blastemal CRABP (Keeble, unpublished data). However, etretinate is known to be metabolized to its free acid Ro 10-1610 (Paravicini et al. 1984) and this compound does bind to CRABP. With these provisos in mind the analogues used here for binding studies are currently being tested on the chick limb bud for their biological efficacy in inducing pattern alterations.

In conclusion, the above considerations provide evidence to support the steroid model as a possible explanation for the mode of action of retinoic acid on the chick limb bud. If this is the case and RA acts via the genome then we may be able to identify the genes responsible for the control of pattern formation during limb development.

We thank Hoffmann-La Roche, Basel, for many generous gifts of labelled and unlabelled retinoids and Katriye Mustafa for excellent technical assistance.