ABSTRACT
Primordial germ cells are the stem cells that provide the functional gametes of adult animals. In many animal groups they are set aside at the earliest stages of development, and migrate from their sites of first appearance to the sites where the gonad will form, the genital ridges. During this migration they proliferate. In the mouse embryo their numbers increase from less than one hundred to approximately four thousand during the period of their migration. In a previous paper we showed that both the proliferation and the direction of migration of mouse PGCs in culture were influenced by soluble factors released from their target tissue, the genital ridges. Studies on other stem cell populations have shown that complex combinations of growth factors control their proliferation, migration and differentiation. In this paper, we show that TGFβ1 inhibits proliferation of PGCs taken from 8.5 day old embryos and cultured on embryonic fibroblast feeder layers. We also show that the previously reported chemotropic effect of genital ridges in this culture system is mediated by TGFβ1, or a closely related molecule, released from the genital ridges.
Introduction
Primordial germ cells (PGCs) first appear in mouse embryos at the end of gastrulation, during the seventh day of development (Ginsburg et al. 1990). Initially they form a small population of alkaline phosphatasepositive cells posterior to the primitive streak, at the root of the developing allantois. From here they become incorporated into the wall of the developing hind gut (at 8.5 days post coitum, dpc), from which they actively migrate into the dorsal mesentery (9.5dpc), and then to the genital ridges (10.5–12.5dpc). During this period their numbers increase from less than 100 to approximately 4000, representing 5-6 division cycles (Tam and Snow, 1981).
Factors that control the survival, proliferation and migration of the germ line population are largely unknown. It has recently been shown that the Steel gene product is an essential survival factor for migrating PGCs (Godin et al. 1990; Dolci et al. 1991). Proliferation of other, better-studied stem cell populations is known to be controlled by complex mixtures of growth factors (see Dexter et al. 1990 for review). Migration is probably controlled by a number of factors, acting in concert. It is known that axonal outgrowth in different neuronal populations in vertebrate embryos is controlled by chemotropic factors. Examples include trigeminal sensory neurons (Davies, 1987) and commissural neurons (Tessier-Levigne et al. 1988; Placzek et al. 1990). There is evidence from grafting experiments that PGCs in birds (Dubois, 1968) and amphibians (Gipouloux, 1970; Giorgi, 1974) may respond to chemotropic signals. Interaction with their substratum is also important (Wylie et al. 1979; Heasman et al. 1981; Alvarez-Buylla and Merchant-Larios, 1986; De Felici and Dolci, 1989).
In previous work we showed that PGCs can be isolated from their migratory route, and cultured on irradiated feeder layers, which support their proliferation and migration for approximately one week (Donovan et al. 1986). On feeder layers of STO fibroblasts, 10.5dpc PGCs exhibit the phenotype of actively motile cells (Stott and Wylie, 1986) when taken from their migratory route. This phenotype is progressively lost, so that once the PGCs have colonised the genital ridges, they are no longer migratory in the same culture conditions (Donovan et al. 1986). More recently, we have shown that the behaviour of early migratory (8.5dpc) PGCs in culture is influenced by diffusible factors released by cultured genital ridges. Medium conditioned by 10.5dpc genital ridges stimulated PGC numbers in culture. Isolated genital ridges also exerted a chemotropic effect on PGCs (Godin et al. 1990).
We have been looking for purified growth factors that mimic the actions of genital ridge-conditioned medium, in order to identify the normal environmental factors that control this early stem cell population. In this manuscript, we present evidence that TGFβ is an inhibitor of PGC proliferation in culture, and mimics the chemotropic effect of genital ridges. Furthermore, we show that an antibody against TGFβ1 blocks the chemotropic effect of whole genital ridges in culture, suggesting that TGFβ1, or a related molecule, may perform this function in vivo.
The TGFβ family is large and diverse. Its actions are complex, and vary, depending both upon the responding cell type, and upon the conditions used (Spom and Roberts, 1990; Moses et al. 1990; Hampson et al. 1989; Dexter et al. 1990, for reviews). TGFβ has a concentration-dependent chemotactic effect on both monocytes (Wahl et al. 1987) and fibroblasts (Postlethwaite et al. 1987). It also has an inhibitory effect on the profiferation of at least two stem cell populations; haemopoietic cells and intestinal crypt cells (Migdalska et al. 1991). The results presented here show that TGFβ1 inhibits PGC proliferation in culture, suggesting that this may be a general mechanism in the complex control of stem cell proliferation.
Materials and methods
Proliferation assays
PGC-containing fragments (allantois and posterior primitive streak) were dissected from 8.5dpc MFI strain embryos, and disaggregated to form a single-cell suspension as described previously (Godin et al. 1990). 50ul aliquots (0.3–1.0 embryo equivalents) were added to STO-cell monolayers in 96-well plates, containing 100μl of a defined serum-free medium (SF-1, Northumbria Biologicals) with or without added human platelet TGFβ1 (ICN). Cultures were fed every day with fresh medium. After fixation and staining for alkaline phosphatase, PGCs in five replicate wells were counted for each treatment and time point. The blocking antibody against TGFβ1 was raised against porcine TGFβ1 in turkeys by Danielpour et al. (1989), and was a generous gift of Michael Spom. The antibody blocks receptor binding and biological activity of TGFβ1; but not β2 This antibody, and control serum, were used, as recommended, at dilutions of 1:1500.
Chemotropic assay
Three wells were cut in an agarose layer in a gridded Petri dish as shown in Fig. 2A. The central well was seeded with a suspension of y-irradiated STO cells and kept overnight to form a monolayer. 8.5dpc PGCs (2–3 embryo equivalents of PGC suspension) were then seeded over the centre of the well. Test tissues or solutions were put in the lateral wells I and II, and the cultures left for 24 h. TOFft solutions were changed five times in the 24h period. Each assay was performed on 14–17 replicate cultures, and the resulting distribution of PGCs shown as a group of three bars, representing the percentage of PGCs in end-zone A, middle zone, and end-zone B, respectively, of the central well. Each outer bar is shaded according to the content of the lateral well nearest to it. The ratios of the percentages of the total PGC number in zones A and B were also calculated for each separate experiment and its respective control. These ratios were then compared using unpaired t-tests.
Distribution of TGFβ1 in mouse embryos
10.5dpc mouse embryos were fixed in 2% trichloracetic acid in water for 6h, dehydrated in ethanol and embedded in polyethylene glycol distearate wax (Koch Chemicals Ltd.) plus 1% cetyl alcohol. Sections were stained with a rabbit antiserum raised against a synthetic peptide identical to the N-terminal 30 amino acids of bovine TGFβ1 (Ellingworth et al. 1986). This antibody and its control serum were a kind gift of Larry Ellingworth of Collagen Corporation, and have been characterised previously by Flanders et al. (1989) and Heine et al. (1987). The antibody was used in this study at a dilution of 10μgml-1. Sections were dewaxed in an acetone series, washed in washing buffer (1 % normal goat serum in PBS), incubated in blocking buffer (10% goat serum, 4% bovine serum albumin in PBS) for 10 min, before incubating overnight at 4°C in primary antibody diluted in washing buffer. After three five-minute washes in washing buffer, sections were incubated in 1:50 dilution of fluorescein-conjugated goat anti-rabbit Ig (Nordic Immunochemicals), washed three times again, stained for five minutes with 0.01 % Eriochrome black in PBS before washing briefly in distilled water and mounting in 90% glycerol containing 100 mg ml-1 1,4 diazabicyclo (2,2,2) octane (DABCO).
Results
PGCs were isolated from 8.5 dpc embryos from MFI mice, and seeded onto feeder layers of irradiated STO embryonic fibroblasts. PGCs were identified by alkaline phosphatase staining and their numbers counted at days 1, 3 and 5 of culture. In serum-free medium, the number of PGCs rose to a maximum between days three and five, and then decreased again, so that PGCs could no longer be identified after about seven days. Serum-free medium conditioned by 10.5 dpc genital ridges caused an increase in the numbers of PGCs over the first five days of culture, compared to serum-free medium alone (Fig. 1A). However, when TGF/Ji was added to the medium, the increase in PGC number was inhibited in a dose-dependent manner (Fig. 1B).
(A) Time-course of the numbers of PGCs, isolated at 8.5dpc and cultured on irradiated STO feeder layers. PGC numbers rise to a maximum at 3–5 days. Genital ridge-conditioned medium increases this maximum. Bars each represent the mean PGC number from 5 duplicate experiments, plus standard errors. (B) shows that TGFβ1 inhibits the increase in PGC numbers seen with SF-1 alone and with genital ridge-conditioned medium, in a dose-dependent manner. In this timecourse, five treatments are shown (see bar code for details). TGFβ1 inhibits the increase in PGC numbers seen between days 1 and 3. Anti-TGFβ but not normal turkey serum (NTS) blocks this effect, and restores PGC numbers to control levels.
(A) Time-course of the numbers of PGCs, isolated at 8.5dpc and cultured on irradiated STO feeder layers. PGC numbers rise to a maximum at 3–5 days. Genital ridge-conditioned medium increases this maximum. Bars each represent the mean PGC number from 5 duplicate experiments, plus standard errors. (B) shows that TGFβ1 inhibits the increase in PGC numbers seen with SF-1 alone and with genital ridge-conditioned medium, in a dose-dependent manner. In this timecourse, five treatments are shown (see bar code for details). TGFβ1 inhibits the increase in PGC numbers seen between days 1 and 3. Anti-TGFβ but not normal turkey serum (NTS) blocks this effect, and restores PGC numbers to control levels.
Chemotropic assay: (A) Diagram of chemotropic assay. PGCs are counted after 24 h in end-zones A and B and the middle zone. Results are shown as trios of bars in (B) – (D). In each trio, the left-hand and right-hand bars represent the % of the total number of PGCs in end-zones A and B respectively, and the central bar the percentage in the middle zone. Each bar is the mean of 15– 17 duplicate experiments, plus S.E.S (B) shows the effect of 25 ng ml−1 TGFβ1 and (C) the effect of 25 ng ml−1. (D) shows the effect of whole 10.5dpc genital ridges in the presence of either normal turkey serum or the blocking antibody. The antibody, but not normal turkey serum, blocks the chemotropic effect of the whole genital ridges.
Chemotropic assay: (A) Diagram of chemotropic assay. PGCs are counted after 24 h in end-zones A and B and the middle zone. Results are shown as trios of bars in (B) – (D). In each trio, the left-hand and right-hand bars represent the % of the total number of PGCs in end-zones A and B respectively, and the central bar the percentage in the middle zone. Each bar is the mean of 15– 17 duplicate experiments, plus S.E.S (B) shows the effect of 25 ng ml−1 TGFβ1 and (C) the effect of 25 ng ml−1. (D) shows the effect of whole 10.5dpc genital ridges in the presence of either normal turkey serum or the blocking antibody. The antibody, but not normal turkey serum, blocks the chemotropic effect of the whole genital ridges.
The specificity of the response to TGFβ1 was shown by blocking its inhibitory effect, using an antibody that blocks its action, (kindly donated by Michael Spom). A 1:1500 dilution of this antibody blocked the effect of the highest concentration (25 ng ml-1) of TGFβ1 whereas the control antiserum had no effect. This blocking antibody has been fully characterised by Danielpour et al. (1989).
In order to test whether TGF & plays a role in PGC migration in culture, we used a chemotropic assay system described by Godin et al. (1990), with the modification that serum-free medium was used throughout, rather than DMEM+foetal calf serum. PGCs were seeded into the central well of a triplet of wells (shown in Fig. 2A) cut into an agar layer in a gridded Petri dish. The central well contained a feeder layer of irradiated STO cells, and the two lateral wells contained test substances, which were replenished five times over a 24 h period. If the contents of the lateral wells exert no chemotropic effect, the distribution of PGCs after 24 h will be random, and the percentages of the total PGC number in the end-zone A, middle zone, and end-zone B, will correspond to their proportional surface areas; 40 %, 20 %, and 40 % respectively, of the central well. If a test substance causes PGCs to migrate towards it, then the percentage of the total number of PGCs in the end-zone nearest it will rise above 40 %, and will show a corresponding fall in the end-zone furthest away. Each experiment is shown as a trio of bars. From left to right, these represent the percentage of the total PGC number in end-zone A, middle zone, and end-zone B, respectively. The height of each bar represents the mean of 14–17 separate repeats of the experiment, together with the standard error of the mean. To compare the effects of different treatments, and to judge their statistical significance, the ratio:% of the total PGC number in end-zone A/% of the total PGC number in B (=A/B) was calculated for each repeat of the experiment. This term is shown for all the experiments in Table 1. Unpaired t-tests were used to compare the A/B ratios of different experiments and their respective controls.
Statistical comparison of chemotropic assays (see Fig. 2A for designation of wells and their regions)

Fig. 2B shows the effect of TGFβ1 in this assay system. After 24 h there was a greater proportion of PGCs in the zones nearest the TGFβ1-containing wells, whereas in the controls an even distribution was seen. Two concentrations of TGFβ1were used, 10 ng ml-1 and 25 ng ml-1. In both cases the growth factor exerted a chemotropic effect. However, this was considerably reduced in the higher concentration. Comparison of the A/B ratios shows that there is a statistically significant difference in the distribution (at the 5 % level) of PGCs at the 10 ng ml-1 concentration, but not at the 25 ng ml-1 concentration (see Table 1).
In a previous paper (Godin et al. 1990) we showed that whole genital ridges in the lateral well exert a chemotropic effect on PGCs. In order to test whether TGFβ1 contributes to this in vitro effect, we tested the chemotropic effects of whole genital ridges in the presence or absence of the blocking anti-TGFβ1 antibody. The results are shown in Fig. 2D, where it can be seen that a 1:1500 dilution of the antibody blocks the effect of whole genital ridges.
In the originally described chemotropic assays, we tested the effects of three tissues isolated from 10.5 day mouse embryos. These tissues, termed genital ridge, mesentery, and limb bud, are shown diagrammatically in Fig. 3. ‘Genital ridge’ included contaminating tissues from the dorsal end of the hind-gut mesentery, dorsal aorta and developing mesonephric kidney. An actual genital ridge in culture is shown in Fig. 3B. ‘Mesentery’ included the hind gut and part of its dorsal mesentery. ‘Limb bud’ included the whole forelimb bud. Only the genital ridges exerted a chemotropic effect. We therefore studied the distribution of TGF-β1 in these tissues to see if it corresponded to that expected from the chemotropic assays. Sections of 10.5 dpc mouse embryos were stained with a rabbit antibody raised against the N-terminal 30 amino acids of mature TGFβ1, and which was kindly donated by Michael Spom and Kathleen Flanders. This antibody has been characterised previously in early mouse embryos (Flanders et al. 1989; Heine et al. 1987). It does not cross-react with TGFβ1 in ELISA, radioimmunoassay, or western blots. However, we cannot exclude the possibility of cross reaction with other members of the TGFβ class. Fig. 3C-G show the distribution of TGFβ1 at the hindgut level of 10.5dpc mouse embryos, and in the forelimb bud. No staining at all was seen in epithelial tissues, including the central nervous system, mesonephric ducts, hindgut and epidermis. Mesenchymal tissues in the dorsal body wall of the embryo stained strongly (Figs 3D and E). The dorsal mesentery of the gut was only partially stained, with staining concentrated in its most dorsal parts (Fig. 3D). Further caudally, little staining was seen in any part of the gut mesentery (Fig. 3E). The forelimb bud showed only small scattered foci of staining, seen between cells at high magnification (Fig. 3G)
Distribution of TCFβ1 (A) Diagrammatic cross-section through the hindgut region of a 10.5dpc embryo, showing the areas used previously in chemotropic assays GR=genital ridge, Mt=mesentery. The limb bud (LB) was taken from a more cranial region of the embryo. (B) A genital ridge piece after 2 days in culture, with the PGCs stained with alkaline phosphatase, to show the distribution of PGCs. (C) – (E) Sections of 10.5dpc embryos. (C) is stained with a control rabbit serum; (D) and (E) are sections from progressively more caudal regions stained with rabbit anti-TGFβ1 (CC1–30) at a concentration of 10μmg ml-1. c=central nervous system, gr=genital ridges, m=mesentery, g=hindgut. (F) A higher magnification picture of staining in the genital ridge region. (G) shows the low level of staining seen in the forelimb bud at this stage. Scale bars: B=300μm, C-E=500μm, F=100μm, G=50μm.
Distribution of TCFβ1 (A) Diagrammatic cross-section through the hindgut region of a 10.5dpc embryo, showing the areas used previously in chemotropic assays GR=genital ridge, Mt=mesentery. The limb bud (LB) was taken from a more cranial region of the embryo. (B) A genital ridge piece after 2 days in culture, with the PGCs stained with alkaline phosphatase, to show the distribution of PGCs. (C) – (E) Sections of 10.5dpc embryos. (C) is stained with a control rabbit serum; (D) and (E) are sections from progressively more caudal regions stained with rabbit anti-TGFβ1 (CC1–30) at a concentration of 10μmg ml-1. c=central nervous system, gr=genital ridges, m=mesentery, g=hindgut. (F) A higher magnification picture of staining in the genital ridge region. (G) shows the low level of staining seen in the forelimb bud at this stage. Scale bars: B=300μm, C-E=500μm, F=100μm, G=50μm.
Discussion
The TGFβ family is diverse and its members have a remarkable range of effects on different cell types (Moses et al. 1990; Spom and Roberts, 1990). TGFβ has been shown to stimulate the profiferation of some cell types, and inhibit others. In general it seems to inhibit the proliferation of stem cell populations (Hampson et al. 1989), and the data presented here support this general view. It has also been shown to be a potent chemoattractant for cells of the haemopoietic (macro-phages and monocytes) and some fibroblastic lineages in culture (Wahl et al. 1987; Postlethwaite et al. 1987).
The results presented here suggest that TGFβ1 has two effects on migrating stage PGCs. Firstly it inhibits PGC proliferation in culture, and secondly it seems to be the active factor in the chemotropic effects of whole genital ridges in culture. We do not know if the TGFβ1 is exerting these effects directly on PGCs in these assays, or indirectly via the contaminating somatic cells or the STO feeder cells. This will remain unknown until we find a way to block its effect on the other cells in the culture system, or until PGCs can be cultured without them. At the moment we cannot purify the very small number of PGCs from the somatic cells around them, and we find that PGCs will only proliferate and migrate on STO feeder layers. We showed previously that neither the STO feeder layer itself, nor the contaminating somatic cells, responded chemotropically to genital ridges (Godin et al. 1990).
We find that TGFβ1 reduces PGC numbers in vitro whilst genital ridge-conditioned medium increases them. There are several possible explanations for this paradoxical observation, which can only be resolved by further experiments. First, we do not know the quantities of growth factor released by genital ridges in culture. It is known that TGFβ1 exerts a chemotropic effect on fibroblasts at concentrations below 50 μg ml-1, whilst a higher concentration is required to inhibit proliferation (Postlethwaite et al. 1987). If the TGFβ1 released by genital ridges in vivo is primarily for chemotaxis, then it may be at concentrations too low to inhibit proliferation. Second, the overall stimulatory effect of genital ridges on PGC proliferation in culture may be a nett effect of both stimulatory and inhibitory growth factors. This can only be tested by identifying the factors responsible, and inhibiting their individual effects. To resolve these possibilities we will need to know the following; the concentration of TGFβ1 released by the genital ridges in culture, the lowest concentration of TGFβ1 which exerts a chemotropic effect, and whether this is below the threshold concentration that also inhibits proliferation. We will also need to know whether other TGFβS are involved.
There are two ways in which TGFβ1 could exert a chemotropic effect in this system. First, the PGCs might carry specific receptors for this growth factor, and respond to a gradient of the signal by recruiting cytoskeletal elements so that the cell polarises and migrates up the concentration gradient. This would be true chemotaxis, and has been shown unequivocally for a rather small number of cell types, for example slime mold amoebocytes, neutrophils, monocytes (Zigmond, 1978; Devreotes and Zigmond, 1988; Wahl et al. 1987). Second, TGFβ1] released from the genital ridges might stimulate the feeder cells to increase the synthesis and release of extracellular matrix molecules, or the regulation of production of other growth factors such as PDGF (Battegay et al. 1990). TGF β1 are known to increase fibronectin secretion by directly upregulating mRNA levels (Ignotz and Massague, 1986) as well as other matrix components (Ignotz and Massague, 1986; Roberts et al. 1986; Ignotz et al. 1987; Bassols and Massague, 1988; Rasmussen and Rapraeger, 1988), and their receptors (Ignotz and Massague, 1987). PGCs in turn could respond to increased levels of these molecules by increased migration rates. We present evidence separately that locally increased levels of fibronectin in these cultures can increase PGC migration rate (ffrench-Constant et al. 1991). Either or both of these mechanisms could happen in vivo, as well as in vitro.
The distribution of TGFβ1 in [uneq] day embryo tissues is puzzling. On the one hand it is consistent with the in vitro effect of genital ridges and purified TGFβ1. On the other hand, the area of the dorsal body wall rich in this growth factor is much larger than the eventual target, the genital ridges. Thus if TGFβ1 does play a role in PGC chemotropism in vivo, then it cannot be the only factor involved. There are several possible explanations. First, the distribution of active TGFβ1 may be less extensive than the staining pattern seen, since forms other than active TGFβ1 may be stained. This problem cannot be resolved at present. Second, several additional factors may be required for correct localization of PGCs to the genital ridges. Despite the likely complexity, this is the first demonstration that the migratory behaviour of this early stem cell population can be controlled by locally produced growth factors.
ACKNOWLEDGEMENTS
We are grateful to the Wellcome Trust for their financial support, and to Kim Goldstone for technical support.