bFGF as a possible morphogen for the anteroposterior axis of the central nervous system in Xenopus

The development of the well-patterned vertebrate nervous system begins with the inductive differentiation of the neural plate from the dorsal side of flat ectoderm at the end of gastrulation. Transplantation experiments performed by Spemann and Mangold first demonstrated that the dorsal mesoderm is responsible for the induction of neural plate and the subsequent patterning of the nervous system (Spemann and Mangold, 1924). They showed that transplantation of the dorsal lip mesoderm of early gastrula (organizer) to the ventral side of the host embryo induced ventral ectoderm to form a secondary nervous system that displayed large scale anteroposterior (A-P) organization. Based on the detailed experiments that followed, it has long been believed that the differentiation of cells in the ectoderm into well-organized neural tissues is brought about by a two-step induction by Spemann’s organizer at the appropriate developmental time (Saxen, 1989; Sive et al., 1989; Nieuwkoop and Albers, 1990; Slack and Tannahill, 1992). The initial step is a general induction of neural development that alone causes the anterior differentiation of the dorsal ectoderm, and the second step involves the further posteriorization of the induced tissue that establishes the A-P axis in the nervous system. However, in spite of intensive research on the molecular mechanism underlying the neural induction, the nature of the specific signal(s) that emanate from Spemann’s organizer has not fully been understand for nearly 70 years since the discovery of the induction phenomenon (Spemann and Mangold, 1924; Warner, 1985).

Very recently, two endogenous molecules isolated from Xenopus embryos have been put forward as potential candidates for the neural inducers. One is the secreted protein encoded by the noggin gene, which is expressed in the organizer region and has neuralizing activity on ectoderm at high concentrations (Lamb et al., 1993; Smith and Harland, 1992). The other substance is follistatin, a protein that specifically inhibits the action of activin, a highly potent mesoderm inducer in Xenopus embryo (Hemmati-Brivanlou et al., 1994). Noggin and follistatin both separately are capable of mimicking the first-step of organizer activity that gives rise to anterior neural tissue. However, the molecular identity and mode of action of the presumptive second-step morphogen that causes regional specification of neural tissue are largely unknown to date.

We have previously demonstrated that basic fibroblast growth factor (bFGF) is another promising candidate for the neural inducer released from the organizer (Kengaku and Okamoto, 1993). Using monoclonal antibodies against various Xenopus embryonic tissues, we showed that physiological concentrations of bFGF changed the developmental fate of the ectoderm from epidermis to central nervous system (CNS) neurons when applied to a culture of dissociated animal cap cells from Xenopus early gastrula. An obvious question is whether bFGF acts solely as a first-step inducer like noggin and follistatin, or whether it can also contribute to the antero-posterior patterning of the CNS from forebrain to spinal cord, a characteristic feature of the second-step activity of Spemann’s organizer. To explore this issue, we cultured animal cap ectoderm cells in the presence of various concentrations of bFGF and examined the expression of position-specific neural markers of the anteroposterior axis. Here we present evidence that bFGF can also drive the second-step in concentration-dependent manner, suggesting that it is identical to or at least can mimic the endogenous neural morphogen(s) released from Spemann’s organizer. A preliminary account of this study has been presented (Kengaku and Okamoto, 1994).

Dissociation and culture of ectoderm cells from gastrula embryos of Xenopus laevis were performed as previously described (Mitani and Okamoto, 1989, 1991). Briefly, animal cap tissue was dissected from dejellied embryos in modified Barth solution (MBS; Gurdon, 1977) and dissociated by incubating in Ca²⁺-, Mg²⁺-deficient MBS containing 1% BSA at room temperature. The dispersed cells were suspended in standard MBS containing 1% BSA and inoculated into plastic culture wells of Terasaki plates (Nunc) at 250 cells/well. After completion of reaggregation by brief centrifugation, cells were treated with recombinant bovine bFGF (Progen Biotechnik, Heidelberg) and incubated in the presence of bFGF at 22.5°C in humidified air until control embryos reached tailbud stage (stage 25; Nieuwkoop and Faber, 1967).

RNA was extracted from 20 cultures and subjected to reverse transcription with oligo(dT)_12-18 as a primer. For quantitative analysis, PCR was performed as described (Kinoshita et al., 1992) with slight modifications. In brief, one-twentieth of the reverse-transcribed mixture was used as template DNA, which was amplified in a reaction volume of 50 μ l containing 10 μ Ci of [α-³²P]dCTP (ICN Biomedicals Inc., Costa Mesa, CA) and 5 units of Taq polymerase (Perkin-Elmer Cetus, Norwalk, CT). Amplification was initiated with the primers for the marker sequence of interest. The primers for the internal control, the EF1α sequence, were added after the first 6 cycles to avoid the possible interference of large amounts of its PCR product with the amplification efficiency of the sequence of interest. After the completion of the amplification PCR products were then separated on a 4% polyacrylamide gel and radioactivity of each of the PCR products were estimated using a laser image analyzer (Fujix BAS 2000, Fuji Film). The primers used are shown in Table 1.

We have previously shown that bFGF could mimic the action of Spemann’s organizer to induce CNS neurons from gastrula ectoderm cells in culture by utilizing a monoclonal antibody N1 as a general marker for CNS neurons. Since it has been hypothesized that anterior and posterior neural tissues are independently induced by different inducers from the organizer (Saxen, 1989; Sive et al., 1989; Nieuwkoop and Albers, 1990; Slack and Tannahill, 1992), it is important to determine what type of neural tissue is induced by bFGF. To this end, we assayed mRNA levels of anteroposterior markers using RTPCR as described in Fig. 1A. For quantitative analysis, PCR products were estimated within the exponential phase of amplification and were standardized against a co-amplified internal control, elongation factor 1α (EF1α; Krieg et al., 1989). As shown in Fig. 1B,C, the amplified products originated from both the marker transcript and the EF1α transcript increased exponentially until the 30th cycle. We therefore decided to quantify the amount of PCR products after the 26th cycle of amplification.

To determine the types of neural tissue induced by bFGF, we assayed the expression of four position-specific homeobox genes in animal cap ectoderm cells from stage 10 embryos treated with various concentrations of bFGF. In intact embryos, the marker genes are expressed within specific regions along A-P axis in the developing CNS: XeNK-2 is expressed in the anterior part of the brain (Saha et al., 1993); En-2 at the midbrain-hindbrain boundary (Hemmati-Brivanlou et al., 1991); XlHbox1 in the anterior spinal cord (Oliver et al., 1988), and XlHbox6 throughout the spinal cord (Wright et al., 1990). As shown in Fig. 2, the anterior markers XeNK-2 and En-2 were detected at doses as low as 0.05-0.1 ng/ml of bFGF. As the bFGF dose increased, En-2 peaked at 0.2 ng/ml and then sharply repressed, whereas XlHbox1 and XlHbox6 became apparent in this order. The expression of the posteriormost marker XlHbox6 was detected at 1.0-2.0 ng/ml and peaked at 5.0-12.5 ng/ml, a concentration about 25-fold higher than the dose required for maximal En-2 expression. XeNK-2 and XlHbox1 were repressed at doses above 2.0 ng/ml. Similar sets of experiments were repeated several times and we obtained essentially the same results, except for one case in which XeNK-2 was detected at a lower concentration (0.05 ng/ml) of bFGF than En-2 (0.1 ng/ml). From these results, we conclude that bFGF can affect the fate of ectoderm cells in culture in a dosedependent manner. The fact that the dose-response profiles for four independent neural markers are distinct from each other within the dose range 0.05-12.5 ng/ml suggests that there may be at least four states of neural cell differentiation within this dose range. The most interesting feature of these profiles is that at lower doses of bFGF the cell state tends to have an anterior characteristics, while at higher doses, the cell becomes to have more posterior characteristics. These results suggest that a concentration gradient of bFGF could contribute to A-P patterning during early neural morphogenesis.

A similar dose-dependent specification has been reported for the induction of mesodermal lineages from blastula animal cap cells by activin (Green et al., 1992). However, it has subsequently been suggested that the graded induction of mesodermal lineages by activin depends on interactions within populations of the induced cells, which initially involve only a few different cell states (Green et al., 1994). These observations raise the possibility that the graded response of gastrula ectoderm cells to bFGF is not directly induced, but rather develops as the induced cells interact with each other during later stages following induction. To test this possibility, we assayed the expression of XeNK-2, En-2 and XlHbox6 in ectoderm cells cultured in the presence of low (0.25 ng/ml) and high (5.0 ng/ml) concentrations of bFGF every 3rd hour after the start of culture (Fig. 3). When the markers were initially detected, each of them was induced within the same dose ranges observed in cultures assayed at tailbud stage (stage 25; Fig. 2): XeNK-2 was detected after 12 hours of culture (stage 19.5) at both high and low doses; En-2 expression was detected after 9 hours of incubation (equivalent to stage 15.5) in cultures treated with low bFGF, but was never detected in those treated with high bFGF; XlHbox6 was first detected after 6 hours of incubation (stage 13) only in cultures treated with a high dose of bFGF. It is thus likely that bFGF determines the four different cell states at the initial stage of the induction.

The posterior neural markers XlHbox1 and XlHbox6 used here are known to be expressed in some mesodermal derivatives in addition to neural tissue (Oliver et al., 1988; Wright et al., 1990; Green et al., 1992). It is possible that bFGF at higher doses induces the formation of posterior mesodermal cells instead of neural cells, although previous studies have shown that the competence of animal cap cells to form mesoderm in response to bFGF was lost before the stage that we examined in the present experiments (Slack et al., 1988). To test this possibility, we analyzed the expression of a neural-specific marker NCAM (Kintner and Melton, 1987) and four mesodermal markers that are expressed in a mesoderm-specific manner at mid-gastrula stage; both gsc (Blumberg et al., 1991) and noggin (Lamb et al., 1993) expressed in dorsal mesoderm, Xwnt-8 (Christian et al., 1991) expressed in lateral and ventral mesoderm and Xbra (Smith et al., 1991) expressed throughout dorsal and ventral mesoderm. None of the four mesodermal markers were detected at a high dose of bFGF, where the posterior markers and NCAM were fully induced (Figs 2B,C, 4). We also examined the expression of follistatin, another candidate of neural inducer expressed in the dorsal mesoderm besides noggin. However, the endogenous expression level of follistatin in stage 12 embryos is very low, and we did not detect any visible bands from cultured ectoderm cells with or without bFGF (data not shown). These results are consistent with our previous observation that myocytes are never induced in stage 9+ to 10G gastrula ectoderm cells by bFGF, even at concentrations as high as 100 ng/ml. It is thus highly likely that the expression of posterior markers at higher doses of bFGF are not of mesodermal, but largely of neural origin.

To verify that bFGF exhibits the properties required for an authentic neural inducer, we investigated whether bFGF can induce a neural-specific marker in ectoderm cells at the time of normal neural induction. Ectoderm cells from gastrula of increasing ages were treated with various concentrations of bFGF and examined for levels of mRNA encoding NCAM, a cell adhesion molecule specifically expressed throughout the nervous system in intact embryos. In cultured cells from stage 10 ectoderm, NCAM mRNA levels increased sharply at the same low doses that increased XeNK-2 and En-2 mRNAs and plateaued with increasing doses (Fig. 5A, upper panel). The induction of NCAM mRNA levels by bFGF was correlated with the suppression of epidermal keratin mRNA levels (Jonas et al., 1989) in a dose-dependent manner (Fig. 5A, lower panel), which is consistent with our previous results utilizing monoclonal antibodies as markers. However, the responsiveness of ectoderm cells to bFGF in neural induction and epidermal suppression decreased as gastrula stage proceeded (Fig. 5B). At stage 10G, there appears to be a number of cells that express neither NCAM nor epidermal keratin. These cells are thought to be differentiating into melanophores and/or possibly other neural crest derivatives (Kengaku and Okamoto, 1993). These findings agree with the observations demonstrating that the responsiveness of ectoderm cells to the action of the organizer to induce neural differentiation and suppress epidermal differentiation decreases during gastrulation (Holtfreter, 1938; Godsave and Shiurba, 1992).

To investigate whether there is a similar temporal change in the ability of ectoderm cells to differentiate into anterior and posterior neural tissues in response to bFGF, ectoderm cells from late blastula (stage 9+) and gastrulae of increasing ages (stage 10 to stage 11) were cultured in the presence of bFGF, and analyzed for the expression of XeNK-2, En-2 and XlHbox6 (Fig. 6A,C). XeNK-2 was maximally induced in ectoderm cells isolated from stage 10 embryos, but exhibited rapidly decreased expression by stage 10S. En-2 expression plateaued between cells from stage 10 and stage 10S embryos and decreased by stage 10G. It should be noted that the presumptive anterior neural region of the dorsal ectoderm has not yet received direct contact from the involuted organizer mesoderm at stage 10G. By contrast, the ability to express XlHbox6 plateaued between cells from stage 10-10G embryos and remained at high levels until stage 11.

The animal cap tissue dissected from later gastrula could contain a smaller population of cells than the tissue from earlier stages because of the epiboly of animal cap cells during gastrulation. To test whether this epiboly might affect our results, we isolated ectoderm cells at stage 10 and cultured them to allow autonomous development. bFGF was added after 3 hours of preincubation and the expression of En-2 and XlHbox6 was assayed at the tailbud stage (Fig. 6B). The ability of ectoderm cells to express En-2 was lost during 3 hours of preincubation, whereas the ability to express XlHbox6 was retained, indicating that ectoderm cells first lose anterior-forming competence, then lose posterior-forming competence. A similar autonomous change of ectoderm competence was observed in the induction of neural tube and neural crest lineages (Kengaku and Okamoto, 1993). Thus, as gastrulation proceeds, the number of neural-competent cells in ectoderm decreases (Fig. 5B) and their fates become progressively restricted to more posterior neural ones (Fig. 6).

We have presented evidence that bFGF can induce gastrula ectoderm cells to differentiate into neural tissue from forebrain to spinal cord in a dose-dependent manner (Fig. 2). Further, bFGF induced neural differentiation from ectoderm cells at physiological low concentrations. The effective doses of bFGF for the induction of both anterior neural markers En-2 and XeNK-2 (2.5 pM) and the posteriormost marker XlHbox6 (125 pM) are within the range of Kd value reported for a FGF receptor in animal cap ectoderm cells of Xenopus blastula (Gillespie et al., 1989). We also showed that bFGF induced general neural marker NCAM and the position-specific markers in ectoderm cells at the time of normal neural induction (Figs 4, 6). These results indicate that bFGF or its homologue is a promising candidate for the neural morphogen derived from Spemann’s organizer.

We previously showed that ectoderm cells of early- to mid-gastrula in culture never differentiated into myocytes, the main dorsal type derivative of mesoderm induced by bFGF, even at concentration of bFGF as high as 5 nM (Kengaku and Okamoto, 1993; Slack et al., 1987; Kimelman and Kirschner, 1987; Godsave and Slack, 1991; Kimelman and Maas, 1992). However, other mesodermal tissues could still be induced from a fraction of ectoderm cells, which in turn induce the remaining cells to differentiate into neural tissues. To address this question, we scored for the expression of several immediate early mesodermal markers in ectoderm cells treated with bFGF.

We showed that neither the organizer mesoderm, in agreement with previous observations (Green et al., 1992; Cho et al., 1991), nor the lateral and the ventral mesoderm were induced by bFGF at 250 pM, a dose where the posterior neural markers and NCAM were fully induced. We therefore conclude that each of the position-specific homeobox genes expressed in cultured ectoderm cells was largely of neural origin, and period (Godsave and Slack, 1991; Grunz and Tacke, 1989). However, this does not seem to be the case for animal cap cells that are immediately reaggregated prior to the start of culture (Grunz and Tacke, 1989). We also observed that gastrula ectoderm cells reproducibly differentiated into epidermal cells when cultured without bFGF (Fig. 5). Further, bFGF could is directly induced by bFGF in the absence of the detectable organizer cells. In addition, the observation that bFGF does not induce the expressions of noggin and follistatin suggests that the neural inducing pathways mediated by these molecules and by bFGF are independent.

It has been shown that animal cap cells from blastula- and early gastrula-stage embryos have a tendency to differentiate into neural tissues when they are kept dispersed for a considerable induce neural lineages from gastrula ectoderm explants, when the outer pigmented layer was removed from the explant (Kengaku and Okamoto, 1993). It is thus unlikely that the neural inducing activity of bFGF is an artifact of the culture condition. However, we cannot exclude the possibility that the ectoderm cells in culture could be induced to form neural tissues by lower concentrations of bFGF than those present in vivo, due to the dilution of endogenous inhibitor(s) such as activin during cell dissociation process (Hemmati-Brivanlou and Melton, 1994).

We showed that specific dose ranges of bFGF specify at least four different neural cell states along the anteroposterior axis of the CNS. Low doses of bFGF induced gastrula ectoderm cells to express anterior neural markers, and increasing doses of bFGF induced progressively more posterior neural markers (Fig. 2).

Induction and regionalization of the vertebrate nervous system has been thought to require two distinct inducers originating in the dorsal mesoderm. In this model, the first-step inducer is distributed uniformly along the A-P axis and induces anterior neural development. The second-step inducer distributes in an A-P gradient with a high point in the posterior mesoderm and causes posterior neural development. In contrast, our present data suggest that bFGF alone could drive both of steps by establishing a concentration gradient.

bFGF and other related molecules have been shown to be expressed at the right time and place to be the authentic neural inducer(s). bFGF is prepositioned in mesoderm-forming regions intracellularly throughout cleavage and gastrula stages, and, during the gastrula stage, it is also detected extracellularly between embryonic tissues (Shiurba et al., 1991; Godsave and Shiurba, 1992). It has been shown that XeFGF and Int-2, other members of FGF family, are expressed in the organizer region of early gastrula (Isaacs et al., 1992; Tannahill et al., 1992). It is probable that the FGF-like molecule(s) is distributed along the A-P axis with the highest point in the posteriormost (i.e., organizer) region. Such a concentration gradient could be formed within the involuted organizer mesoderm that induces the A-P pattern through a vertical action (Sharpe and Gurdon, 1990; Hemmati-Brivanlou et al., 1991; Saha and Grainger, 1992) to the overlying ectoderm. Our present results suggest, however, that bFGF is not likely to act as the vertical inducer, at least for anterior neural tissue, since ectoderm cells have almost lost their ability to respond to bFGF to form anterior neural tissue by mid-gastrula, i.e., prior to the stages at which they receive vertical contacts from the invaginated organizer mesoderm (Fig. 6). Rather, it is likely that FGF is released during early gastrula stages when the anterior-forming competence is available, forming a gradient within the plane of ectoderm with high doses inducing posterior patterns and low doses anterior patterns. Thus the action of bFGF fits the role of an initiator of planar induction, which was originally proposed by Spemann (1938) and has recently been confirmed (Kintner and Melton, 1987; Ruiz i Altaba, 1990; Doniach, 1992; Doniach et al., 1992).

Induction of the anteriormost marker XeNK-2 within a broad dose range does not necessarily conform to the prediction of the above model, which suggests that XeNK-2 should peak at the lowest dose of bFGF (Fig. 2). This apparent contradiction may partly be explained by the fact that XeNK-2 is expressed in the anterior neural crest lineages (Saha et al., 1993). We have previously shown that neural crest-derived melanophores were induced by high doses of bFGF (Kengaku and Okamoto, 1993), suggesting that XeNK-2 induced by higher doses of bFGF may originate from neural crest lineages. It is thus plausible that XeNK-2 expression in the CNS peaks and declines at lower doses of bFGF than En-2 expression.

Alternatively, differential regulation of anterior markers may be accomplished by factors other than FGF-like molecules at later stages, possibly through vertical transmission. If this were the case, our model should be modified as follows: general anterior induction would occur at low doses of FGF driven by planar transmission and with early-receding ectoderm competence. Then posterior respecification, for which late ectoderm competence is available, would be driven by a gradient of higher doses of FGF transmitted through planar and/or vertical contact.

Our previous study showed that bFGF induced melanophores from late-gastrula ectoderm cells when they began to lose their competence to form posterior neural tissue. Melanophores originate from the lateral edges of the neural plate (neural fold) which simultaneously gives rise to the peripheral nervous system. Taken together with the present results, it is strongly suggested that bFGF or other related molecules can contribute to the two-dimensional patterning of the nervous system along the anteroposterior and the mediolateral axes in both dose- and time-dependent manners. It is shown that two distinct types of FGF receptor named XFGFR-1 (Musci et al., 1990; Friesel and Dawid, 1991) and XFGFR-2 (Gillespie et al., 1989; Friesel and Brown, 1992) are expressed in Xenopus embryo. Further, we have recently cloned three new FGF receptor cDNAs from Xenopus gastrula, which we designated XFGFR-3, XFGFR-4 and XFGFR-4′ based on their homology to the human FGF receptors FGFR-3 and FGFR-4, respectively (XFGFR-4′ is highly homologous, but distinct from XFGFR-4). It might be that these receptor molecules with differential affinities to FGF molecules mediate the neural patterning by their appropriate spatiotem-poral expressions in the dorsal ectoderm. Detailed studies using the dominant-negative molecule for each of those five types of FGF receptors are in progress to reveal the precise role of FGF signalling in neural induction.

Our present data do not necessarily exclude the contribution of FGF to other early developmental processes such as mesoderm induction (Amaya et al., 1991, 1993; Cornell and Kimelman, 1994; LaBonne and Whitman, 1994), nor do they exclude a contribution to early neural development from other molecules. There is some evidence suggesting that animal cap cells of late blastula have received an endogenous signal from vegetal blastomeres, which cooperates with Xwnt-8 in mesoderm induction (Sokol, 1993; Doniach et al., 1992). This raises the possibility that gastrula animal cap cells used in this study had already been prepatterned anteriorly by a similar endogenous signal, although those cells does not include the presumptive neural region according to Keller’s fate map (Keller, 1976). Low doses of bFGF might cooperate with the signal to form anterior neural tissue. It should be noted, however, that increasing doses of bFGF changed the cell fate to progressively posterior ones, suggesting that FGF acted as a A-P patterning signal of the nervous system.

Recent studies have shown that noggin (Lamb et al., 1993) and follistatin (Hemmati-Brivanlou et al., 1994), secreted proteins expressed in organizer, are potential candidates for anterior neural inducers. These molecules might cooperate with FGF to accomplish anterior neural induction and A-P patterning in the CNS. Both the gain-of-function and the loss-of-function studies of FGF signalling appear to agree with this idea. Isaacs et al. (1994) have recently demonstrated that over-expression of XeFGF during gastrulation suppressed anterior head development and instead produced expanded posterior tissues derived from ectoderm. Their observation is consistent with our idea that ectoderm cells are induced to form posterior neural tissues by concentrated FGF emanated from the organizer mesoderm. In contrast, Amaya et al. (1993) have shown that inhibition of the normal function of a fraction of FGF receptors in Xenopus early embryos by overexpression of a dominant-negative form of XFGFR-1 disrupted development of trunk axial structures such as notochord and spinal cord, but not of anterior head structures. Lack of axial mesoderm in the mutant embryos may indicate some important role of FGF in mesoderm induction, but it is not inconsistent with the idea that an FGF-like molecule also acts as an endogenous inducer for posterior neural development. Rather, their results may mean that the anterior neural tissue is induced by either noggin or follistatin, or both, in the dominant-negative mutant embryos. However, this idea does not exclude the possible involvement of FGF-like molecules in anterior neural development via parallel complementary pathway with those of noggin and follistatin.

Alternatively, it is also possible that a concentration gradient of FGF-like molecule roughly determines the A-P polarity during early gastrulation stages and that some other molecules from the dorsal mesoderm bring about additional fine tuning of the A-P axis formation. Retinoic acid (RA) has been put forward as a potential candidate for the second-step inducer for the anteroposterior axis of the developing CNS (a ‘neural morphogen’ in Wolpert model reviewed by Green and Smith, 1991). RA can transform anterior embryonic axial tissue containing both neurectoderm and mesoderm to a more posterior specification in a dose-dependent manner (Boncinelli et al., 1991; Durston et al., 1989; Dekker et al., 1992; Ruiz i Altaba and Jessell, 1991).

In conclusion, although many aspects of the function of FGFs in neural induction remain to be elucidated, the dosedependent induction of neural markers by bFGF suggests that gradients of FGF-like molecules may be important for A-P neural patterning and other morphogeneses during early vertebrate development.

We are grateful to Ikuko Hongo and Susumu Ijuin for help with PCR reactions. We thank Prof. Kunitaro Takahashi for advice and discussions and Dr David W. Saffen for critical reading of the manuscript. This work is supported by a grant-in-aid for science research from the Ministry of Education, Science and Culture, Japan and by a fellowship from Japan Society for the Promotion of Science to M. K.