Effects of localized application of transforming growth factor beta 1 on developing chick limbs.

The effects of exogenous transforming growth factor fl (TGF-P) on chick limb development in z$vo were studied by implanting carriers of TGF-@1 into developing wing buds. Agarose beads were soaked in solutions containing TGF-Pl and implanted into wing buds at stages 18 to 27. Localized application of TGF-01 to distal regions of the wing bud caused specific skeletal elements in the limb to be reduced or absent. The particular proximal-distal limb element affected depended on the stage at which the bead was implanted. Position of the bead in the anterior-posterior axis also influenced the pattern of affected structures. Experiments in which TGF-fil beads were implanted and then removed at 24- and 48-hr intervals indicate that there are specific periods during which a skeletal element appears to be sensitive to the effects of exogenous TGF-01. In a few cases, beads placed in proximal positions in later staged limbs resulted in formation of ectopic cartilage near the bead. These results suggest that exposure to exogenous TGF-61 in t:i~o influences the development of skeletal elements in the chick limb in a stage- and position-dependent manner.


Effects of Localized Application of Transforming Growth
The effects of exogenous transforming growth factor fl (TGF-P) on chick limb development in z$vo were studied by implanting carriers of TGF-@1 into developing wing buds. Agarose beads were soaked in solutions containing TGF-Pl and implanted into wing buds at stages 18 to 27. Localized application of TGF-01 to distal regions of the wing bud caused specific skeletal elements in the limb to be reduced or absent. The particular proximal-distal limb element affected depended on the stage at which the bead was implanted. Position of the bead in the anterior-posterior axis also influenced the pattern of affected structures. Experiments in which TGF-fil beads were implanted and then removed at 24-and 48-hr intervals indicate that there are specific periods during which a skeletal element appears to be sensitive to the effects of exogenous TGF-01. In a few cases, beads placed in proximal positions in later staged limbs resulted in formation of ectopic cartilage near the bead. These results suggest that exposure to exogenous TGF-61 in t:i~o influences the development of skeletal elements in the chick limb in a stage-and position-dependent manner.

INTRODUCTION
The transforming growth factor ps (TGF-0s) are multifunctional peptides that have been shown to regulate cellular activities, including cell proliferation and cytodifferentiation, in a wide spectrum of normal and neoplastic cell types in vitro Rizzino, 1988;. Several distinct forms have so far been identified including TGF-Pl (Derynck et ah, 1985), TGF-P2 (Madisen et al., 1988), TGF-03 (ten Dijke et ab, 1988;Derynck et ah, 1988), TGF-P4 (Jakowlew et al., 1988), and TGF-05 (Kondaiah et ah, 1990). The effects of TGF-/31 and TGF-/32 have been most extensively studied. TGF-P effects can be stimulatory or inhibitory and appear to be dependent both on cell type and on culture conditions, especially the "milieu" of other peptide growth factors present. TGF-/3s have also been shown to have dramatic effects on extracellular matrix components (Rizzino, 1988;Roberts et al., 1990).
It has been proposed that members of the TGF-fi superfamily may have a role in vertebrate limb development (Newman, 1988;Kulyk et al., 1989a;Lyons et al., 1989b). Temporally and spatially distinct patterns of expression of TGF-P-like molecules have been seen in the developing mouse limb (Heine et ah, 1987;Sandberg et al., 1988;Lehnert and Akhurst, 1988;Flanders et al., 1989;Pelton et al., 1989;Lyons et al., 1989b;Lyons et ab, 1990;Pelton et al., 1990). In vitro, TGF-@l has been shown to have effects on proliferation and differentiation of cells within the chondrocyte lineage Seyedin et ah, 1988;Kulyk ef al., 1989a;Solursh and Reiter, 1989;Carrington and Reddi, 1990). TGF-P-like molecules may exert their effects either by directly affecting growth and differentiation, in concert with other regulators, or through effects on extracellular matrices (Newman, 1988;Kulyk et ab, 198913). BMP-2A has been shown to induce ectopic bone formation in rats (Wozney et al, 1988;Wang et al, 1990), but so far there have been no studies looking at the TGF-p effects in the developing limb bud in viva The goal of this study was to evaluate the response of developing chick limbs to locally applied TGF-fil. We found that localized application of TGF-01 to chick wing buds via agarose beads alters limb pattern in a concentration-, stage-, and position-dependent manner.

MATERIALS AND METHODS
Fertile White Leghorn chicken eggs (K&R Enterprises, Westminster, CA) were incubated at 38°C in a standard unit providing the necessary humidity and egg rotation. On the fourth day of incubation, eggs were prepared by withdrawing some of the albumen and creating a window in the shell overlying the embryo. Embryos were staged according to the criteria of Hamburger and Hamilton (1951) and then used immediately or reincubated at 38°C without rotation until the appropriate stages were achieved.
The Affi-Gel Blue beads (Bio-Rad, ZOO-250 pm diameter) used in this study are routinely used for protein purification and are composed of crosslinked agarose mesh with covalently attached Cibachrome blue F3GA dye. These carriers have the ability to absorb peptides in solution, to concentrate them within the matrices, and to slowly release them. Agarose beads have been used in previous studies (Schreiber ef al., 1986;Hayek et al., 1987) as a system for the slow release of peptide growth factors ij~ C+IY) and irr vitro.
Recombinant TGF-@l (kindly provided by A. F. Purchio, Oncogen, Seattle, WA), with bovine serum albumen (BSA) as a carrier, was lyophilized and stored at -20°C then reconstituted in 4 mM HCl to achieve concentrations of 0.004 to 4.0 pg/pl, aliquoted, and stored at 4°C. For these studies, 5 ~1 of agarose beads (100-200 beads) in phosphate-buffered saline (PBS) were added to a 5-~1 aliquot of the TGF-/!?l solution (final TGF-fll concentrations of 0.002 to 2.0 pg/pl). These were then incubated at 38°C for 30 min. Beads used as controls were incubated in a solution of BSA which had been lyophilized, reconstituted in 4 mM HCl, and then mixed with the beads in PBS.
A soaking solution of 0.02 @g/p1 is equivalent to 0.2 pg of TGF-0 adsorbed onto 100-200 beads, or l-2 ng per bead, assuming 100% adsorption. However, in a previous experiment, it was estimated that only 75%1 of peptide was adsorbed onto beads contained in 10 ~1 of buffer solution (Hayek et ul., 1987). In addition, the soaking solutions used in our experiments contained BSA which would also be bound, thereby reducing the amount of growth factor adsorbed onto the bead.

Implantation of Beads
Working through the window in the shell, we incised the vitelline and extraembryonic membranes overlying the right wing bud. Manipulations were performed using microsurgical instruments and fine glass probes. Individual agarose beads were inserted through incisions in the dorsal ectoderm of right wing buds. Following experimental manipulation, windows were sealed with removable adhesive tape (Scotch brand) and the eggs were returned to the incubator for reincubation at 38°C without rotation. Embryos were examined daily for at least 2 days following the operations to assess viability, bead position, and responses of limbs. Dead embryos were discarded.

Skeleta,l Anulysis
On the 10th day of incubation, embryos were removed from the shells, rinsed in saline, and eviscerated. The upper torsos with attached wings were fixed in alcoholic Bouin's solution for at least 2 days, then rinsed, and stored in 70% ethanol. The pattern of cartilage structures in the limbs were determined in Victoria bluestained whole mounts (Bryant and Iten, 1974). Left wings were used as controls for assessing the effects on the treated right wings. Some specimens were harvested 24 and 48 hr following bead implantation and were prepared by paraffin embedding, sectioning, and staining with hematoxylin and eosin for histological examination.

Eflects of TGF-01 Concentration on Survival
To evaluate the effects of different doses of exogenous TGF-61 on developing chick limbs, agarose beads that had been soaked in solutions of TGF-fll at concentrations ranging from 2.0 to 0.002 pg/pl were implanted into chick wing buds at stages 18 to 27. The survival of embryos following TGF-fi bead implantation is shown in Table 1. Beads soaked in a 2.0 pg/pl solution gave no surviving embryos. Decreasing the TGF-01 concentration resulted in significant increases in embryo survival. Survival following implantation of beads soaked in 0.02 and 0.002 pg/pl TGF-/31 was similar to survival following implantation of beads without TGF-01.
The experiments reported below were all performed with beads soaked in solutions containing 0.02 and 0.002 pg/pl TGF-@l. The maximum amounts of TGF-Pl that could be released from single beads soaked in these concentrations are calculated to be in the range of l-2 ng and 0.1-0.2 ng of growth factor per bead, respectively (see Materials and Methods).

Eflects of TGF-pl on Limb Cartilage FormatiorL
TGF-/31-containing beads implanted distally at all stages resulted in limbs in which one or more of the skeletal elements either failed to form or were extremely reduced (Fig. 1). The frequency with which limbs were affected correlated with the concentration of TGF-01 in the bead soaking solution (Table 1). Beads soaked in 0.02 pg/pl TGF-01 resulted in abnormal limbs in all embryos in which the bead was retained, while about half of the limbs implanted with beads soaked in 0.002 pg/J TGF-PI were affected. In all cases, implantation of beads without TGF-01 resulted in the development of normal limbs.
In each case, the entire segment of the limb containing the affected skeletal element was extremely reduced, indicating complete absence of part or all of an element, rather than its failure to react with the cartilage-specific stain. For example, in the specimen shown in Fig. lB, extreme reduction of the ulna has led to gross distortion of the radius.

Eflects of TGF-/31 at DiSferent Stages of Limb Development
Implantation of TGF-@l beads along the distal margin of chick wing buds resulted in localized reductions or deletions of skeletal structures within the limb. The particular proximal-distal (PD) limb segment affected by the TGF-01 treatment depended on the stage at which the bead was implanted (Table 2). In wing buds treated at stages 19-21, the element most frequently affected was the humerus (Fig. lA), and among limbs with affected humeri, the majority had normal elements distal to the humerus. At stages 22-24, the majority of limbs showed defects in the forearm (radius/ulna) (Fig. 1B). At stages 25-27, the majority of limbs showed defective elements in the autopod (Fig. lC), with no instances of defective humeri. By stage 27, all defects were localized to the autopod. Hence, the TGF-fll-sensitive target is shifted distally as limb outgrowth progresses. In addition to the correlation between the stage of bead implantation and the PD position of the affected skeletal element, there was also a correlation between stage and final location of the bead. As limb outgrowth progresses, implanted beads are displaced proximally away from the distal margin of the elongating limb bud. Hence, beads implanted distally at stages 19-21 were eventually located in the humerus region, beads implanted at stages 22-24 in the radius/ulna region, and those implanted at stages 25-26 in the autopod (data not shown).

TGF-/31 Bead Position
To evaluate whether the position of the bead in the anterior-posterior (AP) axis would also influence the pattern of affected structures, TGF-fil-containing beads were implanted along the distal margin of the limb bud in either anterior (A), apical or middle (M), or posterior (P) locations. As shown in Table 3, the position of TGF-/31 bead implantation along the AP limb bud axis also affected the position of the abnormal elements, regardless of stage. Anterior beads usually resulted in defects localized to the anterior elements of the limb (radius, anterior digits). When beads were placed in a posterior position, posterior elements (ulna, posterior digits) were affected to a greater extent (Fig. 1B). In the majority of cases, beads placed at the apex affected both sides of the limb to a similar extent.

Efeects of 2.&-azd &'-hr TGF-Pl Exposure Intervals
To learn more about the TGF-@l-sensitive step in the development of cartilage elements, we implanted TGF-@l-containing beads into limb buds and then removed them at 24-and 4%hr intervals.
Beads implanted distally at stages 19-21 and removed 24 hr later at stages 22-24 did not cause limb abnormalities (Fig. 2a). However, beads left in place for 48 hr prior to removal at stages 25-26 resulted in typical humeral abnormalities in 88% of cases (Fig. 2b). These data suggest either that a prolonged exposure to TGF-01 is necessary for an effect or that there is a sensitive period for humerus development that is between 24 and 48 hr after implantation at stages 19-21.
To distinguish between these possibilities, we performed an additional experiment in which beads were placed directly into a proximal (humerus-forming) position at stages 22-24. This is the position in which a bead implanted distally at stages 19-21 would be found 24 hr later. When these proximally implanted beads were re- moved after 24 hr, 73% resulted in defective humeri (Fig. 2d). These results indicate that proximal limb cells exposed to TGF-/31 between stages 22-24 and stages 25-26 are sensitive to the inhibitory effects of the growth factor. Furthermore, limbs in which beads were implanted in proximal positions at stages 25-26 show a greatly reduced incidence of affected skeletal elements (Fig. 2e).
Finally, in a small number of limbs (7.3%) in which beads were implanted proximally at stages 25-26 (i.e., after the period during which cells of the presumptive humerus are sensitive to the inhibitory effects of TGF-Pl), extra cartilage was formed next to the bead (Fig. 1D).
Eflects of TGF-@I Beads on Cell Viability bud near the bead, we examined histological sections of wing buds which had been fixed during the TGF-Pl sensitive period (i.e., between 24 and 48 hr following distal implantation of TGF-P-containing beads at stages 19-21). As demonstrated in Fig. 3, there was no evidence of cell death in the tissue surrounding the bead releasing TGF-Pl when examined 24 (Figs. 3A and 3C) and 48 hr (Figs. 3B and 3D) after implantation.
Rather, the limb cells are healthy and many are actively dividing. DISCUSSION Our results show that exogenous TGF-/31 released from agarose beads causes loss or extreme reduction of specific skeletal elements in the developing chick limb bud. When beads soaked in 0.02 pg/gl TGF-@l (with maximal loading of l-2 ng of growth factor per bead; see To evaluate the possibility that TGF-@l might be hav-Materials and Methods) were implanted distally in ing its effect by causing cell death in regions of the limb chick wing buds at stages 19-21 and then removed 24 hr DEVELOPMENTAL BIOLOGY VOLUME 145, lY!X later, no defects were observed. On the other hand, beads left in place for 48 hr or longer resulted in a high frequency of affected limbs, indicating that an effective dose of TGF-@l is delivered over a period of time in excess of 24 hr. The release of peptide growth factors from agarose beads has been shown to be time dependent. Hayek and colleagues (1987) reported that 50% of basic fibroblast growth factor (bFGF) adsorbed onto beads is released into solution over the first 24 hr, with the remaining growth factor slowly released over the next 3 days. Similar kinetics have been described for the release of epidermal growth factor (EGF) and transforming growth factor 01 (TGF-a) from agarose beads in viva (Schreiber et ab, 1986). In experiments similar to those described above using TGF-Pl, we have also looked at the effects of other peptide growth factors, including nerve growth factor (NGF), EGF, TGF-cu, platelet-derived growth factor (PDGF), acidic and basic FGF, and tumor necrosis factor (TNF) on developing chick wing buds. At the levels used in these studies, so far only TGF-/31 has caused alterations in the limb pattern. The release kinetics for TGF-/31 are expected to be comparable to those shown for other growth factors. Based on the amount loaded (see Materials and Methods), we conclude that less than 1 ng of TGF-01 is slowly released into the limb bud over the course of 24 hr following bead implantation, with the remainder subsequently coming off the bead over the next few days. We also expect that diffusion of TGF-P through tissue, as well as removal via blood perfusion, further reduces the local concentrations of TGF-/3 to which the cells were exposed.

Estimates
of the local concentrations of TGF-fil released from beads are not available. However, similar microrelease methods have been used to study the release of retinoids from beads implanted into developing chick limbs (Tickle et al., 1985). Although the particular details of compound delivery in tissue will differ for each molecule/carrier system studied, some general features apply. For example, when beads were soaked in retinoic acid (RA) at concentrations of 100 pg/ml, approximately 3 ng was adsorbed onto each bead (Eichele et al., 1984). In another study (Tickle et ab, 1985), it was demonstrated that beads soaked in 50 pug/ml RA resulted in local tissue concentrations of about 7.6 pg RA per bud (25 nlV). Using these data, we predict that the amount of TGF-/3 released from a bead soaked in a 20 /*g/ml solution would be in the range of l-10 pg per bud. It has been estimated that volume of a stage 21 chick limb bud is about 0.87 ~1 (Eichele and Thaller,198'7). Thus, we estimate that the resulting concentration of exogenous TGF-/3 in the limb bud is around l-10 rig/ml, which is comparable to the levels of TGF-@ used in in vitro studies. The levels of endogenous TGF-P-like molecules in developing limbs have not been described.
The effect of TGF-6 released by agarose beads remains fairly localized to the region of the bead. When beads were placed on either the anterior or the posterior side of the distal tip of the limb bud, the majority of defects were confined to the side of bead placement. We conclude that TGF$l is present in effective amounts only in the vicinity of the bead and that it does not spread to all regions of the limb bud. In addition, since the effect of localized application of TGF$l is to cause limb defects, we considered the possibility that the exogenous TGF$l was causing localized cell death. We have looked at numerous histological sections of limbs fixed 24 and 48 hr after distal implantation of TGF-P beads (Figs. 3A-3D) and have detected no evidence of cell death. Cells located in the immediate vicinity of the bead exhibit no apparent morphological changes, and mitotic figures are evident in many of the sections. By implanting TGF-/31 beads at different stages, we have shown a stage-dependent sequence of defects along the proximal-distal (PD) axis. Beads implanted distally at stages 19-21 lead primarily to defects in the humerus, at stage 22-24 to defects in the radius and/or ulna, and at stages 25-26 to defects of the hand skeleton. The pattern of defects along the PD axis could also be correlated with the eventual PD position of the bead at later stages in development. Beads implanted at stages 19-21 are found in presumptive upper arm regions 24 hr later at stage 22-24; beads implanted distally at stages 22-24 are later found in forearm regions. Hence, beads are found in the affected segment by 24 hr after distal implantation. Taken together with the implantation/removal experiments, which show that the TGF-@l-sensitive window for the humerus begins at 24 hr and ends at 48 hr after stages 19-21, these data suggest that the humerus anlagen is sensitive to the effects of TGF-@l at stages 22-24.
The timing of the TGF-/31 effect is consistent with the idea that exogenous TGF-@l is influencing the early events of chondrogenesis. Prior to stage 22, the start of the TGF$l-sensitive period for the humerus, the limb mesenchyme appears homogeneous and extracellular matrix components are thought to be evenly distributed. Starting at this stage, fibronectin (FN) and type I collagen expression increase in prechondrogenic areas (Dessau et al., 1980;Melnick et al., 1981). Newman and colleagues (Newman and Frisch, 1979;Tomasek et al., 1982) have suggested that FN is involved in promoting the formation of prechondrogenic aggregates at the onset of chondrogenesis. It is during this time period that we obtain our TGF-/3 effects. At about stages 24-25, at the end of the sensitive period, distinct cellular condensations within the humeral region can be observed. As chondrogenesis begins, levels of FN and type I collagen start to decline (Dessau et al., 1980), and expression of cartilage-specific type II collagen and sulfated cartilage proteoglycan core protein increase (Kosher et ul., 1986a,b).
From i?~ ?>itro studies, TGF-p-like molecules are known to be able to affect many of the processes involved in cartilage formation. However, the reported effects of TGF-/3 are variable and cover the range from stimulation to inhibition of chondrogenesis. Studies showing stimulation include the report by Seyedin and colleagues (1985) that rat mesenchymal cells can be stimulated by TGF-/31 to synthesize collagen type II and cartilage proteoglycan core protein. Kulyk et al. (1989a) demonstrated that TGF-01 promotes chondrogenic differentiation in both high and low density micromass cultures of stage 23-24 whole bud and stage 25 distal limb bud cells. These investigators also reported that TGF-,!3 did not appear to affect cell proliferation or cell viability i?r vitro, as they observed little difference in total DNA content between TGF-P-treated and control cultures. In addition, Carrington and Reddi (1990) reported that, when micromass cultured stage 24-25 chick wing bud cells were treated with TGF-PI for 1-2 days, an increase in alkaline phosphatase activity and [""S]-sulfate incorporation into proteoglycans was observed. In our lab, we have also found that TGF-01 (10 rig/ml) added to the culture medium during the first 24 hr of culture promotes accumulation of Alcian blue-stainable matrix in micromass cultures of stage 23-24 chick wing bud cells (unpublished results). In contrast, there are several reports in the literature of inhibition of chondrogenesis in response to TGF-P exposure. In monolayer cultures grown on plastic, TGF-/31 has been shown to inhibit expression of type II collagen and cartilage proteoglycan by chondroblasts in a dose-dependent manner . Solursh and Reiter (1989) also reported that TGF-P had an inhibitory effect on chondrogenesis (evaluated by Alcian blue staining, as well as collagen type II and cartilage proteoglycan expression) in micromass cultures of stage 23-24 chick wing bud cells. Furthermore, in addition to the results mentioned above, Carrington and Reddi (1990) also showed that the effects of TGF-/31 on limb bud cells in micromass culture depended on the timing and duration of treatment. Exposure of stage 24-25 chick wing bud cells to TGF-01 over the course of 4-i' days in culture, or during days 3-4, resulted in a reduction of both alkaline phosphatase activity and [35S]sulfate incorporation into proteoglycans. It has also been reported that TGF-0 inhibited chondrogenesis by mesenchymal cells grown on collagen gels (Solursh and Reiter, 1989). Preliminary experiments in our lab have shown that TGF-81 delivered via agarose beads to micromass cultured limb bud cells in collagen gels, at the same concentrations used in our in viuo studies, neither stimulated nor inhibited cartilage differentiation (unpublished observations).
One possible explanation for the variety of in vitro results is the diversity of culture conditions, limb bud stages, and limb bud regions used in the various studies. Carrington and Reddi (1990) have suggested that the effects of TGF-fll on chick limb bud cells in vitro may also depend on the state of differentiation of the cells at the time of exposure to the growth factor. Our results support this idea. Whereas implants at early stages lead to inhibition of cartilage formation, proximal implants at later stages (when cells are in a different developmental state) stimulate the formation of extra cartilage at low frequency (4 of 55 cases).
The demonstration that TGF-01 has effects on chick limb bud cells at comparable stages both in vitro and in llim appears to indicate that these cells have receptors that can be influenced by TGF-01 binding. Hence, exogenous TGF-01 can be viewed as interfering with the normal function of an endogenous TGF-P-like molecule(s). At present, there are no reports of the expression of TGF-P-like gene products in early chick limb buds. However, several TGF-/3s (l-4) have been identified in developing chicks (Jakowlew et al., 1991), and several members of the TGF-0 superfamily (BMP-2A and Vgr-1, as well as TGF-P2 and another "TGF$-like" molecule) have been shown to be expressed in spatially and temporally distinct patterns in the developing mouse limb (Lyons et al., 1989b).
In conclusion, both the present in vivo study and previous in vitro studies suggest that TGF-P-like molecules are involved in the development of the limb skeleton and that the variable responses of cells to exogenous TGF-8 can be attributed to the state of differentiation of the cells at the time of exposure. Clarification of the way in which molecules like TGF-P affect limb bud cells in v+vo as well as in vitro will be an important part of understanding the way in which these factors might function in normal limb outgrowth.
We thank Tony Purchio (Oncogen) for the gift of TGF-01. We also thank David Gardiner and the Muneoka lab at Tulane Iiniversitg for helpful discussions and for criticisms of the manuscript. This research was supported by PHS Grant HD20662 and by a gift from the Monsanto Co.