Zebrafish tinman homolog demarcates the heart field and initiates myocardial differentiation

The cardiac field, a demarcated region of the embryo which provides progeny to the heart, was first defined as part of the lateral plate mesoderm which, when explanted at gastrulation or neurulation stages, generates beating tissue (Rawles, 1943; DeHaan, 1965). Cells in this region express cardiac markers, including myosin heavy chains (Litvin et al., 1992). By single cell lineage analysis in the zebrafish, we have found that a demarcated zone for generating heart is in evidence even earlier than can be marked by current probes and prior to cardiac specification. This field is in the ventral hemisphere of the late blastoderm, with the ventral cells having peak propensity to form heart which tapers off laterally (Lee et al., 1994).

What transcription factors initiate cardiomyogenic differentiation are not known. The skeletal myogenic factors do not play a role in heart development (Olson, 1993). Recently discovered basic HLH factors, dHAND and eHAND, may be important, but their reduction, even in combination, does not appear to eliminate initiation of heart differentiation, although it perturbs later development (Srivastava et al., 1995). If heart cell differentiation really begins in the field, it is quite likely that the molecules that begin to establish the cardiac cell fate are not the same as those that complete the process. This is because the fate of cells in the field is plastic, such that removal or introduction of a few cells does not perturb the eventual organ (Huxley and De Beer, 1934). Furthermore, not all cells in the field contribute progeny to the heart. These features are true for the heart field of zebrafish (Lee et al., 1994).

Tinman is a homeodomain protein of Drosophila which is genetically directly downstream of the mesoderm determining gene, twist, and which is responsible for development of the visceral and cardiac subdivisions of the mesoderm (Bodmer, 1993). A similar homeodomain has been noted in vertebrate genes, mNkx2.5 (or Csx) in the mouse (Lints et al., 1993; Komuro and Izumo, 1993), XNkx2.5 and XNkx2.3 in the frog (Tonissen et al., 1994; Evans et al., 1995), and cNkx2.5 in chicken (Schultheiss et al., 1995), all of which are expressed in the precardiac mesoderm. In the mouse, mNkx2.5 is first detected in myocardiogenic progenitor cells as early head fold stage and its expression remains in myocardial cells to adult (Lints et al., 1993; Komuro and Izumo, 1993). In the frog, XNkx2.5 and XNkx2.3 transcripts can be detected by whole-mount in situ hybridization in the cardiac precursors from neurula stage (Tonissen et al., 1994; Evans et al., 1995). In chicken, cNkx2.5 is first detected in the cardiac progenitors from stage 5 (Schultheiss et al., 1995). Targeted gene mutation of Nkx2.5 causes arrest of cardiac development at the looping stage (Lyons et al., 1995).

We find that zebrafish Nkx2.5 is expressed in a spatial and temporal fashion strongly suggesting that it plays a role early in cardiac myocyte differentiation, and it is capable of causing cells to assume some, but not all, attributes of heart cells, both in vivo and in vitro. In vivo, it can regulate both heart size and at higher doses, ventralize the animal. It is a candidate for initiating cell fate and for tying organotypic size to the essential body plan.

We generated a zebrafish adult heart cDNA library in λZAPII (Stratagene) using poly(A)⁺ RNA from adult hearts. We screened the library with probes generated by RT-PCR, using degenerate oligonucleotides based on the tinman domain and homeodomain of Drosophila tinman and mouse Nkx2.5 as primers and RNA from 24-hour zebrafish embryos or adult hearts as templates. Initial denaturation was at 94°C for 1 minute, and was followed by three temperature cycles at 94°C for 1 minute, 56°C for 2 minutes, and 72°C for 2 minutes, with final extension at 72°C for 2 minutes.

The digoxigenin-labeled antisense full-length Nkx2.5 RNA probe was transcribed using T7 RNA polymerase (Promega). Whole-mount in situ hybridization was carried out essentially as described by Oxtoby and Jowett (1993). In brief, embryos were fixed with 4% paraformaldehyde, digested with proteinase K, and hybridized with the zebrafish Nkx2.5 probe at 67°C. Alkaline-phosphatase-conjugated anti-digoxigenin antibody (Boehringer Mannheim) was used to detect the zebrafish Nkx2.5 signals. After staining with NBT/X-phosphate (Boehringer Mannheim), embryos were refixed with 4% paraformaldehyde and stored in PBS.

Whole-mount immunohistochemistry was carried out as described by Chen et al. (1996). In brief, embryos were fixed in 4% formalde-hyde in PBS for 1 hour, blocked in 0.3% saponin, 10% goat serum, in PBS, for 1 hour, and incubated with anti-myosin heavy chain antibodies at 4°C overnight. Alkaline phosphatase or fluorescein-conjugated anti-mouse IgG antibody (Boehringer Mannheim) were used as secondary antibody. MF20 reacts with both the skeletal and cardiac myosin heavy chain in zebrafish. S46 is specific to the atrial-myosin heavy chain in zebrafish (Stainier and Fishman, 1992). MF20 was purchased from the Hybridoma Bank. S46 is a kind gift from Dr J. Miller.

For injection, full length, capped RNA was generated from the Nkx2.5 plasmid (or, for control, lacZ) or Nkx2.5 lacking the homeodomain (ΔHD)) and injected into zebrafish embryos at the 1- to 2-cell stage. The injected embryos were raised at 28.5°C. Embryos used for analyzing the expression of dorsal and ventral markers were fixed 6 hours after fertilization in 4% paraformaldehyde. Otherwise, embryos were scored after 2 days of development for late effects. Approximately 100 pg of Nkx2.5 RNA was used for the lower dose injection and 250 pg was used for the higher dose injection.

250 pg of Nkx2.5 RNA was mixed with tetramethylrhodamine dextran and biotin dextran (Molecular Probes) in 0.2 M KCl, and injected into the donor embryos at the 1- to 2-cell stage and blastomeres from the donor were transplanted to the host embryos using methodology as described by Lee et al. (1994), but in an animal pole position. After 36 hours of development at 28°C, embryos were fixed with 4% paraformaldehyde and analyzed with the anti-myosin heavy chain antibodies, MF20 or S46.

Full-length zebrafish Nkx2.5 and an internal deletion construct which lacks the homeodomain (ΔHD) were subcloned to a eucaryotic expression vector, pBMN (Messersmith et al., 1995) (a kind gift from Dr Alex L. Kolodkin), in frame with the myc-epitope. The constructs were then transfected to the zebrafish fibroblast cell line, PAC2 (Lin et al., 1994; a kind gift from Dr Nancy Hopkins), using lipofectin (Gibco). The transfected cells were fixed for immunohistochemistry 3 days after transfection.

We cloned a zebrafish homolog of Drosophila tinman, termed zebrafish Nkx2.5 because of its homology to Nkx2.5 genes. Zebrafish Nkx2.5 shares high homology with Drosophila tinman only in the N-terminal tinman domain (TN domain) and the C-terminal homeodomain, which is the DNA binding domain. Its sequence is conserved with other vertebrate homologues in the NK2 domain, which may have transcriptional repression activity (Chen and Shwartz, 1995) (Fig. 1). The predicted amino acid sequence of zebrafish Nkx2.5 is 98%, 98%, 91%, 92% and 63% identical in the homeodomain to XNkx2.5, cNkx2.5, XNkx2.3, mNkx2.5 and Drosophila tinman, respectively, but is only 73%, 81% and 73% identical to XeNk2, TTF-1 (Nkx2.1) and mNkx2.2. In the Nk2 domain, zebrafish Nkx2.5 is 94%, 94%, 83%, and 83% identical to XNkx2.5, cNkx2.5, XNkx2.3 and mNkx2.5, but is only 78%, 78% and 72% identical to XeNk2, TTF-1 (Nkx2.1) and mNkx2.2, respectively. In the TN domain, there is an amino acid change, from lysine to arginine, in zebrafish compared to other vertebrate Nkx2.5 genes and Drosophila tinman (Fig. 1B).

Zebrafish Nkx2.5 expression correlates with the position of cardiac precursors. In zebrafish, cardiac progenitors reach the embryonic axis by the 8-somite stage and form bilateral tubes by the 18-somite stage. These tubes then migrate towards, and fuse, at the midline (Stainier and Fishman, 1993). Bilateral Nkx2.5-expressing cells are noticeable by the 10-somite stage (Fig. 2A) and form bilateral tubes ventral to the neural tube by the 18-somite stage, as shown in transverse section in Fig. 2B. These tubes migrate towards the midline, where they begin to fuse at the 20-somite stage (Fig. 2C). The fusion is complete by 24-somite stage (Fig. 2D). The expression of zebrafish Nkx2.5 in the myocardial cells continues to the adult heart (data not shown). In contrast to other vertebrate Nkx2.5 genes, where Nkx2.5 appears to be present in both cardiac precursors and endoderm (Evans et al., 1995; Lints et al., 1993; Schultheiss et al., 1995; Tonissen et al., 1994), we find no evidence of Nkx2.5 expression other than in the mesoderm in the region of cardiac precursors and in defined heart tissue (Fig. 2E,F).

Zebrafish Nkx2.5 expression also differs from other described vertebrate Nkx2.5 genes, and more resembles Drosophila tinman, in that it is expressed earlier in embryogenesis. In Drosophila, tinman is expressed during involution of the mesoderm and then becomes progressively restricted to the cardiac mesoderm (Bodmer et al., 1990). In mouse and frog, tinman homologs are not detectable by whole-mount in situ hybridization during gastrulation. However, low levels of XNkx2.5 and XNkx2.3 transcripts are detected by RNase protection assay at the end of gastrulation (Evans et al., 1995; Lints et al., 1993; Tonissen et al., 1994). In chick, cNkx2.5 transcripts can be detected during gastrulation, but only after stage 5 (Schultheiss et al., 1995). In the zebrafish, Nkx2.5 expression is ubiquitous prior to gastrulation (not shown). This expression is likely to be predominantly maternal in origin because zygotic transcription begins in the zebrafish only at about the 1000-celled stage (Kane and Kimmel, 1993). Nkx2.5 becomes regionalized abruptly to half of the margin of the embryo at 50% epiboly, just at the onset of gastrulation (Fig. 3A). This crescent is at the ventral margin, as shown in Fig. 3B, using goosecoid as a dorsal reference. The Nkx2.5 expressing cells then migrate in the hypoblast (the term used to refer to deeper cell layers during zebrafish gastrulation; Fig. 3C). This position correlates with the position of cardiac progenitors on the fate map. Lineage tracing shows heart precursors to originate from cells in the ventral marginal zone of the blastula, where they involute and converge towards the embryonic axis (Lee et al., 1994). There is a gradient of propensity to generate heart progeny, which peaks at the margin at the ventral axis, and diminishes in a gradient dorsally (Lee et al., 1994). Nkx2.5 expression is in a matching gradient with a ventral peak. The correlation between the pattern of Nkx2.5 expression and the propensity to become heart cells suggests that there may be a threshold level of Nkx2.5 which initiates the cardiac program and that more cells achieve this threshold in the ventral and lateral regions than at the dorsal margin. If so, then increased Nkx2.5 level should bring more cells to threshold and thereby affect heart formation.

We increased the Nkx2.5 level by injection of zebrafish Nkx2.5 RNA (100 pg) at the 1-2 cell stage. The resulting embryos develop normally, but have a visibly enlarged heart (29.5% of surviving injected embryos, n=112). The embryos have mild edema, but otherwise appear unperturbed. The heart beat and circulation appear normal. Both chambers are proportionally bigger. Comparison of atrial sizes of control and Nkx2.5 enlarged hearts are shown in Fig. 4A and B, respectively. Uninjected embryos (n=65), as well as the embryos injected with lacZ (n=86), or ΔHD, a Nkx2.5 RNA in which the homeodomain is deleted (n=72), are not so affected.

To quantitate the effect and to examine if the enlargement is due to an increase in cell number, we counted cell nuclei on histological sections by DAPI staining. The cells of embryonic zebrafish heart appear mononuclear. Therefore, we assume that the number of nuclei represents the number of cells of the embryonic zebrafish heart. The enlarged hearts from Nkx2.5-injected embryos have 30% more cells than do those of embryos injected with lacZ (Nkx2.5-injected mean heart cell number = 2438±171.4, s.e.m., n=8 hearts, 55±3 sections per heart; lacZ-injected mean heart cell number = 1885±38.5, s.e.m., n=5 hearts, 52±5 sections per heart; significant, P<0.05.) We assume that the variability reflects different effectiveness of injection or retention of RNA.

Embryos injected with higher levels of Nkx2.5 RNA (250 pg) manifest more generalized deficits, which have characteristics referred to as ventralization (Graff et al., 1994), with diminution of dorsal and anterior structures, including the head, eyes, notochord and somites (64%, n=104). Fig. 5B shows a severe case, in which the head and eyes are missing, the notochord is diminished, and the somites are fused. More blood cells are observed in the injected embryos. These dorsal-ventral axial effects are evident by the onset of gastrulation, with reduction of the expression of the dorsal marker, goosecoid (Fig. 5C,D) and expansion of the domain of a ventral marker, eve1 (not shown), and of a ventral lateral marker, gata2 (Fig. 5E,F). Because Nkx2.5 expression is expressed in the hypoblast and gata2 and eve1 in the epiblast (Detrich et al., 1995; Joly et al., 1993), it is likely that these effects are cell non-autonomous.

As described for other manipulations which reduce dorsal-anterior structures in Xenopus (Danos and Yost, 1995), there is an accompanying randomization of cardiac looping in the Nkx2.5 injected fish (55% of the surviving injected embryos loop incorrectly, n=104). Such effects are not observed in the control embryos (uninjected, lacZ injected or ΔHD injected; n=54, 67 and 98, respectively).

Many fish injected with these higher levels of zebrafish Nkx2.5 have cells which express myosin heavy chain in ectopic locations (22%; n=104). In nearly all cases such cells are scattered. About half of the time (12.5%; n=104), these are accompanied as well by the presence of additional beating structures in the vicinity of the normal heart (Fig. 6A,B). These embryos are quite deformed and have axial defects, so this effect may represent a block to fusion of cardiac progenitors (De Haan, 1959) rather than generation of new heart tissue. Rarely, there are patches of cells expressing myosin heavy chain which can be in the brain (2%; n=104) (Fig. 6C, D) or on the yolk (4%; n=104) (Fig. 6E,F). These patches are immunoreactive for MF20, which also recognizes skeletal muscle, or, in separate experiments, for S46 which, in zebrafish, is specific for the cardiac atrium (Stainier and Fishman, 1992) (not shown).

The generation of ectopic cardiac tissue in these deformed embryos could be due to cell-autonomous activation of cardiac genes, but, alternatively, could be the result of disruption of the embryonic axis or the interruption of the normal migratory paths of precardiac cells. To examine whether zebrafish Nkx2.5 can autonomously initiate cardiac differentiation, we transplanted Nkx2.5 overexpressing cells, marked by tetramethylrhodamine-conjugated dextran, into late blastrula unmarked hosts, at positions where they do not contribute to the heart (Kimmel et al., 1995, 1990; Lee et al., 1994) (Fig. 7A). Embryos subject to this procedure appear normal and, in all embryos raised successfully (n=26), all of the transplanted cells migrate to the brain or trunk, as the fate map predicts. This suggests that the migratory path of the transplanted cells is not distorted by overexpression of zebrafish Nkx2.5. In most of the embryos (20/26), a subset of transplanted cells become myosin heavy chain-immunoreactive in the ectopic locations, such as the brain and the trunk. Interpretation of MF20 labeling in the trunk is complicated by its concomitant marking of the skeletal muscle along with cardiac muscle. However, because a subset of the transplanted cells also express the atrial-specific marker, S46, it is clear that zebrafish Nkx2.5 can initiate adoption of the cardiac fate, even in ectopic locations (Fig. 7B-D).

We were curious as to why not every transplanted cell expresses myosin heavy chain. In the transplantation experiment, we are not able to distinguish whether the variability is due to differences of protein levels within transplanted cells or modification by the environment. To address this issue, we transfected the zebrafish fibroblast cell line, PAC2, with zebrafish Nkx2.5 tagged with the myc epitope, which can be recognized by the antibody, 9E10. About 20% of the transfected cells are positive for 9E10, and hence are presumably generating Nkx2.5 protein. MF20 and 9E10 have different IgG subtypes, which allows us to do double staining with these anti-bodies. 90% of zebrafish Nkx2.5/myc-expressing fibroblasts adopt these attributes of cardiac cell fate, defined by MF20 labeling (n=234) (Fig. 8A,B). In other experiments, about the same ratio of the Nkx2.5 transfected cells are positive for S46 (not shown; S46 has the same IgG isotype as 9E10, so cannot be simultaneously compared). myc-tagged constructs in which zebrafish Nkx2.5 lacks the homeodomain do not cause myosin heavy chain expression (n=147) (Fig. 8C,D). These data indicate that zebrafish Nkx2.5 acts cell-autonomously with regard to myosin heavy chain expression, and that this effect requires the homeodomain.

The zebrafish Nkx2.5 gene appears tied to generation of the heart. From the site of the early field in the ventral marginal zone, to the lateral plate, and then to the heart tube, its pattern of expression corresponds tightly to the location of heart precursors. Its overexpression enlarges heart size, an effect which can occur in the absence of other effects upon the embryo. Transplanted cells expressing Nkx2.5 can assume a cardiac fate, even in distant ectopic sites such as the brain, and it causes fibroblasts in culture to express myosin heavy chain. However, these cells do not beat, suggesting that Nkx2.5 is one step, perhaps the first, in determining cardiac cell fate, but that sub-sequent signals are needed to complete differentiation.

In Drosophila, tinman is believed to function just downstream of the maternal mesoderm determining genes, including the transcription factor, twist. tinman is expressed first throughout the mesoderm, and then restricted to visceral and cardiac mesoderm (Bodmer, 1995; Bodmer et al., 1990). Embryos mutant in tinman lack visceral mesoderm derivatives and heart (Bodmer, 1993). The target genes for tinman are unknown. Vertebrate homologs are related to Drosophila tinman in the homeodomain and in an 11 amino acid stretch near the amino terminus. Unlike tinman, they also share a short peptide domain of homology downstream of the homeodomain (Lints et al., 1993). In other vertebrates, they are largely, but not totally, restricted to cardiogenic areas and the heart. In the mouse, mNkx2.5 is expressed in the pharyngeal endoderm and the tongue, lingual muscle, spleen and stomach, in addition to myocardial cells (Lints et al., 1993). In the frog, both XNkx2.5 and XNkx2.3 are expressed in the pharyngeal endoderm (Tonissen et al., 1994; Evans et al., 1995). In chick, cNkx2.5 is also expressed in the pharyngeal ectoderm and endoderm (Shultheiss et al., 1995). The functions of Nkx genes in these tissues are not yet clear. The heart of mice mutant in Nkx2.5 fails to develop through looping, a stage apparently sensitive to loss of several genes (Lyons et al., 1995), so that, although clearly critical to heart development, the stage at which Nkx2.5 acts has not been defined by the targeted mutation. If Nkx genes are essential for initiating the cardiac cascade, continued differentiation in mice with the Nkx2.5 mutation may reflect the presence of redundant pathways. Indeed, Nkx2.3 is expressed in a similar domain, at least in frog (Evans et al., 1995). Therefore, the loss of function experiments do not preclude a critical role of Nkx2.5 in early steps of cardiomyogenesis.

Is the zebrafish Nkx2.5 really essentially distinct from other vertebrate homologs? This seems unlikely. The pattern of zebrafish Nkx2.5 expression is similar to other vertebrate homologs in that it is expressed in the cardiac precursors. It differs from other vertebrate Nkx2.5 genes in cardiac specificity and early onset of expression. Although the role of fish Nkx2.5 might differ from other vertebrates, for example because the dynamics of fish gastrulation are distinctive, it is equally plausible that the embryonic transparency of the zebrafish facilitates visualization of low levels of expression. Along these lines, it is interesting that the expression of XNkx2.5 is evident during gastrulation by RNase protection at times prior to evident expression by whole mount in situ hybridization (Tonissen et al., 1994; Evans et al., 1995).

One proposed role of Nkx2.5 is in the cascade which drives assumption of the cardiomyogenic phenotype (Bodmer, 1995; Olson and Srivastava, 1996). We provide three lines of evidence to support this. (1) Overexpression of Nkx2.5 causes ectopic expression of myosin heavy chain immunoreactivity, including that identified by the cardiac-specific antibody S46. Such cells are generally scattered, but occasionally cluster on yolk or brain. (2) Cells transplanted from embryos injected with Nkx2.5 can express cardiac markers even in ectopic locales, suggesting that the effect is cell-autonomous. (3) Fibroblasts transfected with Nkx2.5 express cardiac markers.

Nkx2.5 does not appear sufficient to cause expression of the entire repertoire of cardiac genes. Although cardiac-specific reagents for the zebrafish are limited, it is clear that differentiation is incomplete because ectopic cells expressing Nkx2.5 do not beat. The two hearts noted in some of the deformed embryos are as likely to represent secondary effects of perterbation as they are induction of an ectopic heart. Interuption of midline signals can prevent fusion of left and right primordia, an abnormality termed cardia bifida (DeHaan, 1959). As suggested (Nascone and Mercola, 1995), Nkx2.5 could be the first step in cardiac differentiation, with subsequent signals needed to finish the process, some presumably from endoderm. This model also is compatible with the observation that Nkx2.5 is expressed by cells in the early gastrula at a time when cells are still plastic and can assume fates dictated by host locale after transplantation. It is likely that persistence of Nkx2.5 expression also is needed for cardiac differentiation. In Drosophila and vertebrates, not all the cells which express tinman or Nkx2.5 adopt the cardiac fate. In Drosophila, tinman is first expressed in all mesoderm and gradually becomes restricted to the cardiac precursors (Bodmer et al., 1990). In frog, chick and mouse, it is expressed transiently in both endoderm and cardiac mesoderm (Lints et al., 1993; Schultheiss et al., 1995; Tonissen et al., 1994). In zebrafish, its expression is restricted to the ventral margin at the onset of gastrulation, a position which includes the cardiac precursors but is not limited to them (Lee et al., 1994). If a threshold level of Nkx2.5 is needed to initiate cardiac differentiation, it may be that not every cell within the field expresses levels high enough or persistent enough to initiate cardiac cell fate decisions.

The expression of zebrafish Nkx2.5 is restricted to the ventrallateral margin at the onset of gastrulation. The expansion of ventral determinants and reduction of dorsal and anterior structures caused by zebrafish Nkx2.5 injection suggests that it interacts with the network of localized factors which set the axial form of the embryo. In all likelihood, this ventralization is responsible for abnormal cardiac looping in these embryos. Reduction of dorsoanterior structures in Xenopus by UV or Xwnt-8 causes a randomization of looping (Danos and Yost, 1995), a condition referred to as situs inversus. Hence, although the effects of Nkx2.5 on looping may reflect direct effects upon the heart, we believe them more likely to be secondary to reduction in the dorsoanterior structures. How and when the left-right signal is transmitted is unknown, although asymmetrical expression of activin receptor IIa, sonic hedgehog and cNR are evident as early as stage 4 of the chick, and their perturbation can cause situs inversus (Levin et al., 1995).

Nkx2.5 overexpression causes a large hyperplastic heart. The chamber morphology and function are well-retained, and, aside from some mild edema, the circulation appears to be normal in these embryos. Cardiac dilation is a response to failure or increased load, an adaptation termed Starling’s law. However, the enlarged hearts noted here after Nkx2.5 injection include more cells, indicating that even if there is subtle dysfunction, the enlargement cannot be attributed solely to dilation but also includes an element of hyperplasia. Cleaver, Patterson and Krieg have found recently that Xenopus Nkx2.5 and Nkx2.3 cause hyperplasia in injected Xenopus embryos (Cleaver, Patterson and Krieg, 1996). It is as though the heart of the Nkx2.5 overexpressing embryo is designed for a larger animal. In transplantation experiments between newts of different sizes, using tissue from the heart field of the gastrula as donor, Copenhaver noted that heart size did not adjust completely to the size of the host (Copenhaver, 1939). It is possible that over-expression of Nkx2.5 increases the proliferation rate of cardiac cells. However, a more plausible explanation is that information has been specified by the gastrula stage as to the overall size of the heart. The effect of Nkx2.5 overexpression could be to raise Nkx2.5 levels diffusely and thereby bring cells on the edges of the heart field to threshold, or bring more cells in the heart field to persistently express Nkx2.5 above the threshold. If their position is near or in the field, such cells presumably could receive the additional signals needed to commit them to the heart cell fate.

Organs develop in a highly regulative fashion in vertebrates, as though the precursor regions can adapt to longer-range embryonic determinants of form and size. The region denoted as capable of forming an organ, the field, is larger than that

which eventually does so, and normal organs form even after introduction or removal of cells from the field (Huxley and De Beer, 1934). We speculate that one role for Nkx2.5 could be as a molecular device to make heart growth proportional to that of the embryo. The expansion of ventral determinants and reduction of dorsal and anterior structures caused by Nkx2.5 injection could mean that it interacts with the network of localized factors which set the axial form of the embryo so that the Nkx2.5-expressing zone would be controlled by the signals which ensure a balance of the relative axial determinants. In some cases, this regulation may be from another germ layer. For example, since BMP4 (Graff et al., 1994), like dpp, is a TGFβ family member, and appears to be a ventral determinant in vertebrates, it might play a role in Nkx2.5 regulation analogous to that of dpp in Drosophila (Frasch, 1995; Staehling-Hampton et al., 1994). Therefore, inter-germ layer and axial information could be coordinated and integrated with regard to organ size in a regulative manner beginning early in embryogenesis. The heart is the first organ to form. It will be of interest to see whether other genes could play sizing roles for other organs.

This work is supported by NIH RO1-HL49579 and NIH RO1-RR08888 to M. C. F. Thanks to Brant Weinstein for help with the initial transplantation, to Chris Simpson for help with histology, to Nadia Rosenthal and members of the Fishman laboratory for thoughtful comments on the manuscript.