Gap junctions relay FGF8-mediated right-sided repression of Nodal in rabbit


  • Kerstin Feistel,

    1. University of Hohenheim, Institute of Zoology, Stuttgart, Germany
    Current affiliation:
    1. Division of Neuroscience, Oregon National Primate Research Center, Oregon Health and Science University, Portland, OR 97006
    Search for more papers by this author
  • Martin Blum

    Corresponding author
    1. University of Hohenheim, Institute of Zoology, Stuttgart, Germany
    • University of Hohenheim, Institute of Zoology, Garbenstrasse 30, D-70593 Stuttgart, Germany
    Search for more papers by this author


In vertebrate gastrula/neurula embryos, a cilia-driven leftward flow asymmetrically activates the Nodal cascade in the left lateral plate mesoderm (LPM). In frog embryos left–right axis formation was postulated to depend on gap junctions (GJs) during cleavage. Here, we show that GJs cooperate with fibroblast growth factor-8 (FGF8) to specify asymmetric Nodal in the rabbit embryo at gastrula/neurula. GJs and FGF signaling were manipulated in whole embryo and explant cultures of rabbit blastodiscs. These experiments demonstrate that right-sided inhibition of Nodal by FGF8 depended on intercellular communication by means of GJs, and that left-sided induction of Nodal required attenuation of gap junctional communication (GJC). Before flow, the left and right side were equally competent but actively prevented from Nodal induction through FGF8/GJ. Our data suggest that flow unilaterally attenuates FGF8/GJ-mediated repression of Nodal on the left side, integrating GJC and FGF8 into the flow-based mechanism of symmetry breakage in the vertebrate embryo. Developmental Dynamics 237:3516–3527, 2008. © 2008 Wiley-Liss, Inc.


Vertebrates display an internal asymmetric organization of most organs, superimposed on the overt bilateral symmetry of their body plans (Capdevila et al.,2000). The apex of the heart, which itself is positioned in the ventral midline, points to the left side, lungs differ with respect to lobation, stomach and spleen are found on the left, the liver on the right side, and the small and large intestines coil asymmetrically (Moore and Persaud,2003). An asymmetric module consisting of the growth factor Nodal, its feedback inhibitor Lefty, and the homeobox transcription factor Pitx2 has been identified in deuterostomes (Boorman and Shimeld,2002; Hamada et al.,2002; Duboc et al.,2005). In most vertebrates, a cilia-driven leftward flow of extracellular fluid acts as a left-specifying determinant at early neurula stages, resulting in the left-asymmetric activation of the so-called Nodal cascade (Nonaka et al.,1998; Hirokawa et al.,2006). Flow occurs over ciliated epithelia at the embryonic midline, namely the posterior notochord (PNC; Blum et al.,2006) in mouse (Nonaka et al.,1998) and rabbit (Okada et al.,2005), Kupffer's vesicle (KV) in teleost fish (Essner et al.,2005; Kramer-Zucker et al.,2005; Okada et al.,2005), and the gastrocoel roof plate (GRP) in amphibian embryos (Schweickert et al.,2007). We have previously argued that the ciliated structure commonly referred to as “node“ in mouse in fact represents the homolog of the rabbit PNC (Blum et al.,2006). PNC, KV, and GRP represent homologous structures, as, besides leftward flow, they share bilateral Nodal expression domains on either side, and the positioning at the posterior pole of the notochord (Blum et al.,2006).

In rabbit embryos, we have shown that fibroblast growth factor-8 (FGF8), although expressed symmetrically in the caudal midline, acts asymmetrically in the process of laterality determination, namely as a repressor of the Nodal cascade on the right side (Fischer et al.,2002). Ectopic FGF8 applied on the left side of the PNC invariably prevented Nodal transcription in the left LPM, while inhibition of FGF8 signaling on the right side resulted in ectopic right-sided activation of the Nodal cascade (Fischer et al.,2002). The time window of ectopic FGF8 action on the left side closed concomittantly with the onset of cilia-driven leftward flow (one- to two-somite stage; Fischer et al.,2002; Okada et al.,2005; and Rietema, Bitzer, and MB, unpublished observations). We have previously reasoned that gap junctional communication (GJC) might relay FGF8 action from the midline to the lateral plate (Fischer et al.,2002).

FGF-signaling has been known to interact with GJC in other contexts during development. In chick lens cultures, FGF controls the opening status of gap junctions (GJs) and enhances GJC (Le and Musil,2001). In chick limb bud cultures, application of FGF4/8 augments expression of the GJ gene connexin 43 (Cx43; Makarenkova et al.,1997). Down-regulation of Cx43 protein level by introduction of antisense oligonucleotides reduced expression of FGF4/8 in the chick limb bud (Makarenkova and Patel,1999). The involvement of GJs in laterality determination was experimentally demonstrated in frog and chick embryos. GJ agonists and antagonists as well as mutant Cx43 gene constructs and antisense approaches resulted in altered left–right (LR) gene expression and organ placement (Levin and Mercola,1998,1999). Despite their role in Xenopus and chick, genetic evidence for a link of GJC and LR axis formation in mammals is scarce and controversial. No laterality defects were detected in Cx43 knockout mice (Reaume et al.,1995). In humans, one study described Cx43 mutations in laterality patients (Britz-Cunningham et al.,1995), while a second study disputed this finding, as no Cx43 mutations were recorded in 38 cases of sporadic and familial heterotaxia in humans (Gebbia et al.,1996). Thus, a firm genetic basis for the involvement of GJC in mammalian laterality decisions has yet to be established.

Experimental data from frog and chick were integrated into the ion flux model, which postulated symmetry breakage to occur at cleavage in frog and the onset of gastrulation in chick embryos (Levin et al.,2002; Levin,2005). According to this hypothesis, GJC is required before flow for the asymmetric distribution of (a) small molecule(s) along an electrochemical gradient generated by asymmetrically localized ion pumps (Levin et al.,2002; Levin,2005). In contrast to this proposal, we show in the rabbit embryo that asymmetric regulation of GJC is required for the relay of laterality cue(s) from the midline (primitive streak/Hensen's node/PNC) to the LPM. Right-sided repression of Nodal by FGF8 depended on permeable GJs, while left-sided induction of Nodal required attenuated GJC. Our data suggest that before flow FGF8 represses the Nodal cascade in a bilaterally symmetrical manner. We propose that leftward flow at the PNC results in unilateral repression of GJ conductance and, consequently, release of FGF8-based repression, integrating GJC and FGF8 into the flow-based mechanism of symmetry breakage.


Coupling of Epithelial Cells in the Rabbit Blastodisc Before and During Somitogenesis

To follow up on the hypothesis that FGF8-mediated repression of Nodal in the right LPM might be implemented through GJC, we investigated the expression pattern of the GJ gene Cx43. This member of the connexin gene family was chosen because it was previously shown to be actively transcribed in the chick blastoderm (Levin and Mercola,1999). In addition, interference with Cx43 perturbed laterality in chick and Xenopus (Levin and Mercola,1998,1999). Embryos of the relevant stages, i.e., up to the onset of LPM Nodal expression, were subjected to whole-mount in situ hybridization with a rabbit Cx43 specific probe (for staging of rabbit embryos, see Blum et al.,2006). A sense control probe did not result in any signal (not shown). Figure 1 demonstrates that, from the formation of Hensen's node at stage 3 to early somite stages, Cx43 mRNA was abundant in most embryonic tissues. Noteworthy features of Cx43 expression include its absence from Hensen's node (Fig. 1A,A′), the expression from the midline to the lateral plate in all three germ layers up to the onset of somitogenesis (Fig. 1B–D,D′), and a lack of signal in the condensing somites (Fig. 1F′). Furthermore, a distinct LPM-specific expression domain appeared from the two-somite stage onward (Fig. 1E,F,E′,F′), which coincided with the pattern of Nodal expression in its dynamic rostrocaudal expansion. Thus, potential GJ-dependent signals could emanate from the midline and reach the lateral plate by means of GJC from stage 5 to the one-somite stage. Additionally, the coupling of the LPM in a domain reminiscent of Nodal expression at the very stages indicated possible LPM-intrinsic intercellular communication in that process.

Figure 1.

Cx43 expression in the rabbit embryo. Whole-mount in situ hybridization of rabbit blastodiscs of defined stages using an antisense probe specific for Cx43. Embryos are shown in ventral view with anterior to the top. A: When Hensen's node (n) forms at stage (st.) 4, strong Cx43 signals were detected in the posterior-most portion of the primitive streak (ps), and in the anterior third of the blastodisc with the marked exception of Hensen's node itself, which was devoid of transcripts (A,A′). Differences between germ layers were obvious such that expression in the hypoblast (hy) was significantly weaker as compared to the overlying epiblast (ep; A′). B,C: Regional differences persisted when the notochordal plate (np) emerged and lengthened at stage 5 (B) and 6 (C). D: The beginning of somitogenesis marked a transition, as the laterally continuous expression from midline to the LP (A–D,A′,D′) became discontinuous from the two-somite (som) stage onward (E,F). E,F: Ectodermal expression was now restricted to the neural plate (E,E′, F). In addition, a new domain developed in the lateral plate mesoderm (lpm), which formed at the level of the first pair of somites (E) and spread cranially and caudally thereafter (F,F′). ec, ectoderm; en, endoderm; fp, floor plate; me, mesoderm

GJ Are Required for Right-Sided Repression and Left-Sided Induction of LPM Nodal in the Rabbit Blastodisc

A possible role of GJs in laterality determination in the rabbit was tested in experiments, in which embryos were cultured in the presence or absence of known regulators of intercellular communication. Initially, we used lindane to block GJC. Because of the toxicity of this agent, we switched to heptanol in subsequent experiments, an aliphatic alcohol known to block GJC, presumably by intercalation into the membrane and thus mechanically squeezing channels closed (Guan et al.,1997; Rose and Ransom,1997; Cotrina et al.,1998). As heptanol proved to be an efficient agent in our experiments, other known effectors of gap junctional conductance such as octanol were not investigated (Spray et al.,1985; Harris,2001; Rozental et al.,2001). Embryos were taken into culture from stage 5 to the three-somite stage, and exposed to control medium or medium supplemented with 0.07 mM heptanol. Specimens were cultured until they reached the four- to six-somite stage, and, after fixation, analyzed for Nodal expression by whole-mount in situ hybridization. In the course of these experiments we noted that effects were highly dependent on embryonic age at the beginning of culture. Data therefore are presented in three sets, representing embryos taken into culture before the two-somite stage, exactly at the two-somite stage, and from the three-somite stage onward. Control cultures showed that stable left-sided Nodal expression in the LPM was present from the two-somite stage onward (Fig. 2A). Earlier cultures resulted in predominantly bilateral expression (Fig. 2A), confirming our previous results (Fischer et al.,2002). In heptanol-exposed cultures, no differences were observed in embryos before and after the two-somite stage (Fig. 2B), as compared to control untreated specimens (Fig. 2A).

Figure 2.

Gap junctions (GJs) are required for laterality determination in the rabbit embryo. Embryos of defined stages were cultured up to the four- to six-somite (som) stage in the presence or absence of GJ modulators and assayed for Nodal transcription by whole-mount in situ hybridization. A–C,E: The GJ blocker heptanol induced ectopic right-sided Nodal expression specifically in embryos explanted at the two-somite stage. Before and after the two-somite stage, no significant deviations between control and treated cultures were recorded (A,B), while highly significant differences were detected in the right (r) lateral plate mesoderm (LPM) of specimens taken into culture exactly at the two-somite stage (C; four representative examples shown in E). D,F: The enhancer of GJ permeability EM12 repressed left (l) LPM Nodal transcription in a highly significant proportion of embryos (11/15) explanted exactly at the two-somite stage. The two representative examples depicted in F demonstrate that PNC Nodal was unaffected by EM12.

Strikingly differing patterns between control and heptanol treated embryos, however, were noted when embryos were taken into culture exactly at the two-somite stage (Fig. 2A,B). While in control cultures more than 90% of specimens displayed normal left-sided Nodal expression (Fig. 2A), heptanol treatment resulted in bilateral expression (i.e., ectopic right-sided induction) in more than 60% of cases (Fig. 2B). Remarkably, the initial experiments in which lindane was applied showed the same results as heptanol (5/7 embryos treated at the two-somite stage displayed bilateral Nodal expression; 23 embryos treated before the two-somite stage showed Nodal expression indinstinguishable from control cultures; not shown). For a statistical analysis of data, the right LPM was considered separately (Fig. 2C). While right-sided expression in control cultures was below 10%, this fraction raised to more than 60% upon heptanol exposure (Fig. 2C). This effect was statistically highly significant (P < 0.01). A close examination of the predominant bilateral phenotype of heptanol specimens cultured at the two-somite stage revealed that the ectopic right expression domain consistently lagged behind the endogenous left domain, as judged by differences between left and right domains (four representative embryos shown in Fig. 2E). This finding was true both for signal intensity (Fig. 2E1) as well as craniocaudal expansion of Nodal expression (Fig. 2E2-4). Notably, heptanol treatment did not alter expansion or intensity of the endogenous left-sided domain, as could have been anticipated by the correlating expression domains of Cx43 and Nodal in the LPM (c. Fig.1E,F,E′,F′). The heptanol experiments thus demonstrated that GJC was essential for right-sided repression of Nodal transcription in the LPM, while the left LPM, in contrast, was unaffected by blockage of GJ conductance.

A reciprocal scenario was therefore tested in a second set of experiments, in which EM12, a thalidomide analog known to enhance GJ conductance, was used (Nicolai et al.,1997; Levin and Mercola,1999). Embryos were treated at the two-somite stage with 1 mg/ml EM12 in medium, cultured, and analyzed for Nodal transcription as described above. The most pronounced effects were now observed in the left LPM. In contrast to the invariant expression of Nodal in the left LPM of control cultures, EM12 resulted in absence of Nodal transcripts in >70% of treated specimens (Fig. 2D). This result was statistically very highly significant (P < 0.001). The two typical examples of treated embryos depicted in Figure 2F further demonstrate that the PNC Nodal domain was not affected, that is, that the EM12-induced repression of Nodal transcription was restricted to the LPM tissue.

Taken together, these experiments establish a role of GJC in laterality determination in mammalian embryos. Contrary to chick and frog, where GJs were required before and during gastrulation, our data show that in the rabbit GJC is involved shortly before induction of Nodal in the left LPM at the three-somite stage, and thus concomitant with flow.

FGF8 Antagonizes the Inductive Effect of Heptanol in the Right LPM

The heptanol phenotype, that is, ectopic right-sided induction of Nodal (Fig. 3B), was reminiscent of the previously recorded effect of blocking FGF8 signal transduction with the FGF receptor specific inhibitor SU5402 (Fig. 3A; Fischer et al.,2002). We wondered whether GJ/heptanol and FGF8 interacted in the rabbit embryo. Embryos were explanted at the two-somite stage, beads soaked in 1 mg/ml bovine serum albumin (BSA) or human recombinant FGF8 were placed on the right side of the PNC, i.e., close to the endogenous Ffg8 expression domain, and specimens were cultured in the presence or absence of 0.07 mM heptanol. While control cultures treated with BSA beads and heptanol showed the heptanol-specific right ectopic Nodal domain in the LPM (Fig. 3C), ectopic Nodal was absent in eight of nine specimens exposed to FGF8 and heptanol at the same time (Fig. 3D,E). Effects were statistically significant (Fig. 3E; P < 0.05), demonstrating that ectopically applied FGF8 was able to effectively override the heptanol-induced blockage of GJs. These data show that the ectopic right-sided induction of Nodal transcription by blockage of GJs was rescued by parallel administration of FGF8, establishing the interaction of FGF8 and GJC in this process.

Figure 3.

Fibroblast growth factor-8 (FGF8) and gap junctions (GJs) interact to repress Nodal in the right lateral plate mesoderm (LPM). Embryos were explanted at the two-somite (som) stage, treated with beads placed on the right (R) side of the posterior notochord (PNC) as indicated, and—following culture to the four- to six-somite stage in the absence (A) or presence (B–D) of 0.07 mM heptanol—analyzed for Nodal expression. A: Inhibition of FGF8 signaling by SU5402 resulted in ectopic right-sided induction of Nodal in all cases (n = 5). B: Heptanol treatment induced bilateral Nodal expression in approximately two-thirds of cases (13/21). C: Culture of embryos with a control bovine serum albumin (BSA) bead in the presence of heptanol yielded bilateral Nodal signals in four of four specimens. D: FGF8 repressed right-sided Nodal induced by heptanol in eight of nine cases. E: Effects in the right LPM were statistically significant (P < 0.05). Arrowheads indicate presence (black) or absence (white) of Nodal transcripts in right LPM tissue. L, left.

FGF8 Is Required at the Midline

The FGF8-dependent repression of Nodal could be the result of a relay of downstream mediators from the midline, where Fgf8 is transcribed, to the LPM, where Nodal is either induced or repressed. Alternatively, FGF8 might act by means of an unknown long-range mechanism directly in the LPM itself. To distinguish between these possibilities, we applied control BSA- or FGF8-soaked beads to left LP tissue explanted at different time points, and analyzed cultures for Nodal transcription. Figure 4A shows a typical example of such an explant at stage 6, in which Nodal was unaffected by FGF8, demonstrating that FGF8 was unable to interfere with induction of Nodal transcription in the left LPM. Likewise, the rapid craniocaudal spreading of the Nodal domain, which occurs between the two and five somite stage (Blum et al.,2006), was not altered by an FGF8 bead placed directly into the Nodal domain (representative example shown in Fig. 4B). Statistical analysis of 35 FGF8-treated and 65 untreated explants revealed virtually identical expression ratios (P = 0.9, Fig. 4C). FGF8 thus did not act on Nodal transcription at the very site where Nodal was induced, i.e., in the LPM tissue itself. These experiments therefore demand a mechanism to relay FGF8-specific signals from the midline to the lateral periphery, where Nodal is repressed in the right LPM.

Figure 4.

Fibroblast growth factor-8 (FGF8) does not repress Nodal in isolated left lateral plate mesoderm (LPM) tissue. Left LPs were cut from stage 5 to the three-somite (som) stage, an FGF8-loaded bead was placed into the LP tissue, and explants were cultured until the contralateral side had reached the four- to six-somite stage. A–C: Following fixation, explants were analyzed for Nodal expression. Explants isolated at stage 6 (A) and during somitogenesis (example of three-somite specimen shown in B) revealed Nodal signals in ratios indistinguishable from control explants cultured in the absence of FGF8 (C). Dashed lines demarcate boundaries between embryonic and extraembryonic tissue.

Gap Junctions Relay Repression From the Midline to the LPM

The interaction between FGF8 and GJC (Fig. 3) suggested that such a relay mechanism depended on GJs. In whole embryos at the two-somite stage, blockage of GJs by heptanol had resulted in ectopic right-sided induction of Nodal (Fig. 2A–C,E). To test whether this was the result of autonomous regulation of GJ conductance in the LPM, or whether this effect was controlled by the midline as the source of endogenous FGF8, we performed LP explant cultures at the two-somite stage with or without adhering midline. Removal of midline tissue included the bilateral Nodal domain on both sides of the PNC. Right LPs explanted with adhering midline at the two-somite stage resulted in Nodal induction through heptanol-mediated blockage of GJs (Fig. 5A). While in control cultures 3/15 specimens (20%) displayed Nodal signals, the frequency raised to >50% (5/9) upon heptanol treatment. In contrast, right LP explants at the two-somite stage cultured in isolation did not express Nodal, irrespective of the presence of heptanol (Fig. 5B). The low frequency expression in control cultures was not significant compared with the absence of Nodal in LP explants without midline (P = 0.4), while the midline effect in heptanol cultures was statistically significant (P < 0.05). This result showed that blockage of GJC between midline and LPM was a prerequisite of right-ectopic expression of Nodal, demonstrating that GJC served as a link between LP and midline FGF8.

Figure 5.

Gap junctions relay repression from the midline to the lateral plate mesoderm (LPM). A,B: Right LP tissue was isolated with (A) or without (B) adhering midline at the two-somite (som) stage, cultured in the presence or absence of heptanol, and analyzed for Nodal transcripts after explants reached the four- to six-somite stage. A: Explants with midline showed low-frequency expression of Nodal in control cultures dissected at the two-somite stage (3/15). Heptanol treatment significantly increased the frequency of Nodal expression (5/9). B: Without midline isolated right LP explants never expressed Nodal when isolated at the two-somite stage. A,B: Heptanol effects were statistically significant between specimens cultured with (A) and without (B) midline (P < 0.05), while differences between control cultures were not significant (P = 0.4). Dashed lines demarcate boundaries between embryonic and extraembryonic tissue.

Maintenance of Right-Sided Nodal Repression Requires Continued Presence of Midline

Next we wondered whether the midline was required subsequent to flow-mediated symmetry breakage for the maintenance of molecular asymmetry after the two-somite stage. Midline tissue might be involved in maintenance of left-sided and/or absence of right-sided Nodal transcription. We investigated the midline as the source of both FGF8 and flow in the context of LPM Nodal transcription. Nodal was systematically analyzed in LP explants cultured in the presence or absence of adhering midline tissue (Fig. 6). Embryos were dissected either at the two-somite stage, that is, before induction of LPM Nodal transcription but after development of robust leftward flow (Okada et al.,2005; and our unpublished observations), or at the three- to five-somite stage, that is, following left-sided LPM Nodal induction. Left LP explants displayed no significant midline-dependent differences after the two-somite stage, demonstrating that following Nodal induction the midline was dispensible for continued Nodal transcription (P = 0.3; Fig. 6A). Differences were, however, significant at the two-somite stage, reflecting the presence or absence of flow at the time point of explantation (P < 0.05; Fig. 6A). Examination of corresponding right LP explants in contrast revealed absence of Nodal irrespective of the presence of midline when tissue was explanted at the two-somite stage (P = 0.4; Fig. 6B), proving that flow did not affect absence of right-sided Nodal. Remarkably, explants cultured after the two-somite stage displayed significant midline effects, namely induction of Nodal in six of nine specimens cultured without adhering midline (P < 0.05; Fig. 6B). This data set shows that the midline was required for maintenance of right-sided repression of Nodal, while left-sided transcription was independent of midline following its initial induction.

Figure 6.

Maintenance of right-sided repression of Nodal requires continued presence of midline. Embryos between the two- and five-somite stage were dissected such that one lateral plate (LP) was isolated away from the midline, leaving the contralateral LP with adhering midline and paraxial tissue. Following culture, the four- to six-somite stage specimens were fixed and analyzed for Nodal transcription. A: Left-sided explants. Midline was required for Nodal induction at the two-somite stage but dispensible for continued expression thereafter. B: Right-sided explants. Midline was required for maintenance of Nodal repression following the two-somite stage. L, left; R, right.


In the present study, we show that GJC is required for LR axis formation in mammalian embryos. Experimental manipulation of GJ conductance demonstrated that GJC was required to relay laterality cues from the midline to the periphery (LPM). Our analysis revealed that the two-somite stage was particularly sensitive to GJC manipulation. Blockage of GJs in whole embryos resulted in ectopic right-sided induction of Nodal, without affecting the left side. Enhancement of GJ conductance, in contrast, repressed left-sided Nodal. Confirming our initial working hypothesis, we further show that FGF8 and GJs interact in the process of right-sided Nodal repression.

Experimental manipulations of Xenopus and chick embryos have suggested that GJC plays a pivotal role upstream of the Nodal cascade. In frog embryos, GJ blockers and dominant-negative connexin gene constructs resulted in randomization of organ situs and altered marker gene expression (Levin and Mercola,1998). The time window, however, during which GJ modulators affected laterality in frog embryos closed at stage 12 (late gastrula), that is, clearly before the recently described cilia-driven leftward flow initiates in the frog neurula at stage 15 (Schweickert et al.,2007). Thus, GJC and flow might not be involved in the same process in LR axis determination in Xenopus. This time window, however, could also reflect a temporal limitation of accessibility of drugs to the relevant tissue. Contrary to the situation in rabbit, where the ventral surface of the embryo is freely accessible in cultured embryos, the homologous tissue in Xenopus, the GRP (i.e., the PNC equivalent) and the lateral endodermal cells, invaginate around stage 12 and, consequently, localize inside the embryo as part of the dorsal roof of the archenteron (Shook et al.,2004; Schweickert et al.,2007).

The ion flux model of symmetry breakage in frog proposed that unidirectional electrophoresis through open GJ channels resulted in asymmetric segregation of charged low molecular weight determinants and thus asymmetric gene expression (Levin et al.,2002; Levin,2003,2005). We have investigated whether early rabbit embryos displayed voltage or pH gradients similar to those described in Xenopus and chick (Levin et al.,2002). No evidence for pH or voltage imbalances across the midline were recorded (K.F. and M.B., unpublished observations). Even though the present study clearly establishes a role of GJC in LR development of rabbit embryos, it does not lend support to an ion flux based mechanism driving symmetry breakage during primitive streak stages in mammals.

During our analysis of the spatial and temporal expression of Cx43, we observed dynamically changing patterns of Cx43 transcription in mesodermal cells. In presomite stages, expression at the level of the PNC was continuous between midline and LPM. With the onset of somitogenesis, expression receded in the condensing somites, and signals reminiscent of Nodal expression appeared in the LPM. The Cx43 domains thus correlated with two essential processes in laterality determination: (1) the process of LR signal transfer from the midline to the LPM; (2) the propagation of the Nodal cascade in the LPM itself. The involvement of GJC in the first process has clearly been demonstrated by our heptanol experiments (Figs. 2, 5). Heptanol, however, had no effect on the propagation of the Nodal signal within the LPM. Interestingly, a recent report indicates that this second process may indeed also be regulated by Cx43, as it demonstrated a requirement of Cx43 for transforming growth factor-beta (TGF-β) signaling (Dai et al.,2007). In cardiomyocytes, Cx43-regulated TGF-β activity by triggering the release of Smads from microtubules into the cytosol. This competition for microtubule binding sites facilitated signal transduction from the TGF-β receptor to the nucleus and, thus, activation of target gene expression (Dai et al.,2007). Such a mechanism would be unaltered by heptanol treatment, because heptanol sterically alters the pore size of the GJ complex rather than acting on the individual connexin protein (Guan et al.,1997; Rose and Ransom,1997; Cotrina et al.,1998). If a similar mechanism were active in the LPM, Cx43 expression in the Nodal expression domain would be involved in the activation of the Nodal target genes Nodal, Lefty, and Pitx2. Further experiments such as an LPM-targeted gene knockdown of Cx43 might establish an interaction of TGF-β signaling and Cx43 in the progression of the Nodal cascade.

Our data not only involve GJC in laterality determination in rabbit, but more specifically show that permeable GJs are required precisely during the two-somite stage. This finding raises the question about the specific role of this very narrow time window. The two-somite stage is characterized by two major events in LR axis specification: it opens with the full development of a laminar cilia-driven leftward flow above the rabbit PNC (Okada et al.,2005; Rietema and M.B., unpublished results). The time window for GJC manipulation closes with the end of the two-somite stage, which coincides with the onset of Nodal expression in the left LPM from the three-somite stage onward. Embryos taken into culture during that critical period and treated with the GJ blocker heptanol developed ectopic expression of Nodal in the right LPM. We have observed that the right ectopic domain varies in size and is either equivalent to the endogenous left-sided Nodal expression in its rostrocaudal extension or lags behind the endogenous domain (Fig. 2). Because each cycle of somitogenesis lasts approximately 1.5 to 2 hr, it seems reasonable to assume that differences between ectopic and endogenous domains reflect the time point of explantation and beginning of treatment. GJ blockage starting at the onset of the two-somite stage would result in an early ectopic activation of Nodal and lead to bilaterally symmetric domains. A treatment starting midway through the two-somite stage would still be capable of inducing ectopic expression of Nodal, but would result in a retarded ectopic domain compared with the endogenous one.

Explant cultures of LP tissue with or without adhering midline also pointed to a mechanism of LR-determination active specifically in two-somite stage embryos. While right LP tissue taken from two-somite stage embryos never expressed Nodal when cultured without the midline (Fig. 6B, left bar in upper panel), the presence of midline structures induced ectopic Nodal expression (Fig. 6B, left bar in lower panel). These results are in perfect agreement with data from Hamada and colleagues (Fig. 2A–D in Nakamura et al.,2006) who cultured right LP tissue from mouse embryos with or without adhering PNC (“node”). Although in their experiments only the PNC was included and notochord and primitive streak were removed, they essentially observed the same result, that is, an ectopic induction of Nodal expression only when the midline was present. A 8223μSelf Enhancement and Lateral Inhibition” (SELI) mechanism for the establishment of robust asymmetric expression of Nodal in the left LPM and its continued repression in the right LPM was proposed by Hamada and colleagues and has recently been demonstrated in Xenopus as well (Ohi and Wright,2007). According to this model, Lefty1 expressed in the midline acts as a long-range inhibitor of Nodal expression on the embryo's right side. Our experiments in which we explanted LP tissue after the two-somite stage (Fig. 6B, right columns of both panels) match the predictions of the SELI model: presence of the midline (Fig. 6B, lower panel) was required to maintain right-sided repression of Nodal, and removal of midline tissue (Fig. 6B, upper panel) resulted in ectopic expression of Nodal, i.e. a loss of Nodal repression in the right LPM.

A SELI mechanism is thus very likely to play an important role in maintaining LR asymmetry in the rabbit as well. The SELI model, however, does not integrate the fact that FGF8 acts unilaterally at an earlier stage already. This right-specific function of FGF8 is all the more surprising as not only was symmetrical expression of Fgf8 described at PNC and Hensen's node of rabbit and mouse embryos (Meyers and Martin,1999; Fischer et al.,2002; Sirbu and Duester,2006; Rietema and M.B., manuscript in preparation), but RTK-receptor activation around the PNC/node in mouse (Tanaka et al.,2005) is entirely symmetrical as well. To account for these data, we like to propose a release-of-repression mechanism based on our data and in agreement with the published literature (schematically depicted in Fig. 7).

Figure 7.

Release-of-repression (R-o-R) model of symmetry breakage. A: Gene expression and inductive/repressive interactions superimposed on scanning electron micrographs of rabbit embryos from stage (st.) 5 to three-somites (som), shown to scale in ventral view. Morphological features of embryos are indicated in drawing on right. B–E: Schematic depiction of midline–lateral plate mesoderm (LPM) interactions. For detail see text. Dashed ovals indicate competence for Nodal transcription, outlined and filled ovals demarcate established Nodal domains. GJC, gap junctional communication; HN, Hensen's node; L, left; NO, notochord; PNC, posterior notochord; PS, primitive streak; R, right.

Before we elaborate on the different phases, we like to outline the main features of this model: (1) before flow, the left and right side of the embryo are kept in symmetrical balance by two counteracting mechanisms: the FGF8/GJC module prevents induction of Nodal bilaterally from the outgrowth of the primitive streak at stage 3 onward. Competence for Nodal induction emerges on both sides between stage 5 and stage 6, likely conferred upon the LPMs by the bilaterally symmetrical expression of Nodal at the PNC. (2) The leftward flow at the PNC, which sets in at the two-somite stage, results in unilateral release of repression, that is, acts as a permissive signal toward the left side. (3) Maintenance of right-sided repression of Nodal requires midline Lefty from the three-somite stage onward, as the FGF8 and LPM Nodal domains undergo craniocaudal displacement relative to each other.

Before the two-somite stage, Fgf8 is expressed in the caudal midline, i.e, primitive streak and Hensen's node, and all three germ layers are coupled by GJs (Fig. 7B,C). The PNC has formed, cilia have started to elongate and become motile, but no flow has developed as yet (Rietema, Bitzer, and M.B., unpublished observations). We propose that during this period FGF8 acts as a repressive signal, which is relayed bilaterally into both left and right LPM through a GJ-based mechanism (Fig. 7B). Thus, FGF8—in agreement with its symmetrical mode of expression—executes a symmetrical long-range effect on the LPM tissues. How could FGF8 signaling be relayed in a GJ-dependent manner? A recent study by Stains and Civitelli (2005) proposed a mechanism by which GJC can regulate spreading and amplification of RTK-mediated signals through coupled cells. In osteoblast tissue culture, approximately half of the cell's activated ERK was shown to arise from an amplification of ERK-signaling by Cx43-mediated GJC (Stains and Civitelli,2005). Inhibition of GJC resulted in a 50% reduction of the level of activated ERK, as well as down-regulation of target gene expression (Stains and Civitelli,2005). It was suggested that, as a primary response to growth factor stimuli, ERK was activated as well as a second (unknown) signal which spreads through GJs to coupled neighboring cells. This second messenger passing through the GJ pore would in response activate ERK in the adjacent cells, thus spreading and amplifying the growth factor response from the source toward the periphery. Such a mechanism would obviate the need for long-range diffusion of a growth factor. It is tempting to speculate that such a mechanism may account for the observed GJ-dependent long-range effects of FGF8 during LR axis formation.

By what means might the relayed FGF8 signal repress Nodal in the LPM tissue? First, it is well documented that the LPM tissue in principle is competent to initiate Nodal transcription on both sides. The many bilateral gene expression patterns encountered upon experimental manipulation of cultured embryos as well as in knockout mouse mutants underscore this notion (Bisgrove et al.,2003). Likely, this competence is conferred upon the LPM by the reaction–diffusion mechanism recently demonstrated in the mouse and Xenopus (Fig. 7C), in which Nodal protein synthesized in the bilateral PNC domain diffuses as a morphogen to the LPM (Nakamura et al.,2006, Ohi and Wright,2007). Rabbit LP explant cultures isolated before the onset of PNC Nodal transcription never initiated LPM Nodal (not shown), in agreement with corresponding genetic data in mouse (Brennan et al.,2002; Saijoh et al.,2003). In that setting, the role of an inhibitor on both sides falls upon FGF8, which acts to prevent the initiation of the autoregulatory Nodal feedback loop, in which Nodal signaling induces Nodal transcription (Saijoh et al.,2000; Hamada et al.,2002, Ohi and Wright,2007). Mechanistically, this process might be accomplished by ERK-mediated phosphorylation of the linker domain of SMAD2, which inhibits Nodal signal transduction (Massague,2003; Pera et al.,2003). Although we consider this an attractive scenario, the rabbit system seems less suited to investigate this question in greater depth, as the repertoire of manipulative techniques is limited, and large numbers of embryos cannot be processed.

During the two-somite stage, cilia-driven leftward flow sets in (Fig. 7D). We propose that as one result of flow GJC becomes specifically attenuated on the left side, i.e., that as net result of flow repression of Nodal signaling by FGF8 is released unilaterally. Several lines of evidence from our data as well as from the literature support this proposal. First, our experiments show that, at the two-somite stage, GJC endogenously has to become attenuated as a prerequisite of left-sided Nodal transcription. Therefore, the GJ blocker heptanol had no effect on endogenous left-sided LPM-Nodal at and after the two-somite stage. Second, the enhancer of GJ conductance EM12 promoted loss of endogenous left LPM Nodal. Third, in mouse and zebrafish a transient left-asymmetric calcium signal was reported as consequence of leftward flow (McGrath et al.,2003; Sarmah et al.,2005; Tanaka et al.,2005). Elevated cytoplasmic calcium levels are known negative regulators of GJ conductance (for a recent review, see Bukauskas and Verselis,2004). And finally, it has been reported in mouse that cilia-driven leftward flow transports vesicles which contain retinoic acid (RA) and Sonic hedgehog (Shh) to the left margin of the PNC (Tanaka et al.,2005). RA has been linked to LR pattern formation in several studies. Notably, systemic application of RA in mouse, chick, frog, and zebrafish resulted in bilateral LPM Nodal expression (Chazaud et al.,1999; Tsukui et al.,1999; Wasiak and Lohnes,1999). We have noted the same effect of RA in rabbit as well (not shown). RA is a known antagonist of FGF8 signaling (Diez del Corral et al.,2003; Diez del Corral and Storey,2004; Sirbu and Duester,2006) and has been reported to act as a negative regulator of GJ conductance (Zhang and McMahon,2001). These data open the possibility that RA asymmetrically distributed by cilia-driven leftward flow may contribute to the attenuation of GJ permeability and thus counteract the long-range effect of FGF8 on Nodal.

One of the unresolved issues in LR axis determination relates to the opposing roles of FGF8 described in different vertebrate models (Tabin,2005). In chick and rabbit, FGF8 promoted right-sidedness (Boettger et al.,1999; Fischer et al.,2002), while in mouse a hypomorphic FGF8 allele failed to induce left-sided LPM Nodal (Meyers and Martin,1999). The FGF-dependent release of RA/Shh-loaded vesicles at the PNC further supported the view of FGF8 as a left determinant in mouse (Tanaka et al.,2005). In the light of our experiments, we suggest that FGF8 plays a dual role in LR axis determination. First, in the primitive streak, FGF8 is required for bilateral repression of the Nodal cascade before flow, acting as a right determinant (Fig. 7B). The FGF-dependent vesicle release may be considered as a negative feedback mechanism by which FGF8 by means of flow-calcium/RA-GJs attenuates its own effects selectively at the left margin of the PNC (Fig. 7D). A hypomorphic FGF8 mouse strain progresses normally through gastrulation, demonstrating that primitive streak FGF8 function is not compromized in these embryos (Meyers and Martin,1999). The laterality phenotype suggests, however, that vesicle release does not occur, a prediction that might be tested in future analyses. In rabbit, FGF8 beads placed on the left side of the PNC likely override flow-mediated antagonism of endogenous FGF8 signaling. In the chick, no flow has been detected nor has a PNC-homologous structure been identified. As both instructive (Nodal) and repressive (Fgf8) signaling molecules are expressed asymmetrically at Hensen's node (Levin et al.,1995; Boettger et al.,1999), flow may have become obsolete during evolution, and thus the left-specific function of FGF8 been lost in chick.

After the two-somite stage, Nodal signal transduction in the left LPM ensures stable expression until the feedback inhibitor Lefty-2 terminates the Nodal cascade at around the six- to eight-somite stage (Nakamura et al.,2006, Ohi and Wright,2007). Our model states that maintenance of right-sided Nodal repression still requires the midline, however, GJC is not involved in this process (Fig. 7). The data presented in this study support this view, as right LP explants after the two-somite stage expressed Nodal when the midline was absent (Fig. 6B), and heptanol in whole embryos had no effect after the two-somite stage, proving the independence of this induction from GJC (Fig. 2B). In addition, a repressive role of the midline on right-sided Nodal transcription in the LPM was already inferred from experiments in Xenopus, in which competent LP tissue was cocultured with midline and LPM Nodal consequently lost (Lohr et al.,1998). The regression of Hensen's node and ensuing shortening of the Fgf8 expression domain in the primitive streak in early somite embryos result in a segregation of LPM Nodal and primitive streak Fgf8 in craniocaudal dimension (see 3-somite embryo in Fig. 7A), necessitating the implementation of an effective means anterior of the Fgf8 expression domain to maintain repression of Nodal in the right LPM (Nakamura et al.,2006; Tabin,2006). This repressive function likely falls on Lefty1, which in rabbit is expressed along the entire length of the elongating midline until after disappearance of Nodal from the left LPM (Anja Rietema and M.B., manuscript in preparation).

In summary, the central feature of the release-of-repression model is constituted by the flow-mediated left-sided abolition of bilateral repression, enabling the induction of the Nodal cascade. This model thus integrates GJC and FGF8 into the flow mechanism of symmetry breakage and may provide a step toward a unifying model of LR development in the vertebrates.



Female rabbits were purchased from a commercial breeder (Haussler, Germany) and mated in the animal facility at Hohenheim University.

Embryo Culture and Dissection

Embryos from timed matings were dissected from uteri in sterile PBS at room temperature and either fixed in 4% PFA for in situ hybridization or further processed for in vitro culture. For LP explants, tissue was removed with fine iridectomy scissors by cutting a straight line along the paraxial mesoderm. Care was taken to leave the presomitic and somitic mesoderm attached to the midline. Whole embryos as well as LP explants were placed ectoderm-side down on a clot of agarose (0.5% in PBS) in a 5-cm Petri-dish in Ham's F10 medium supplemented with 20% fetal calf serum, such that only the extraembryonic portion of the embryo was submerged in medium (semi-dry culture). Beads were applied under a dissecting microscope by using fine tungsten needles with a small hook. For treatment with 1 mg/ml mouse recombinant FGF8b (R&D Systems), heparinized acrylic beads (Sigma) were used, and Affigel Blue beads (Bio-Rad) were chosen for application of 1 mg/ml SU5402 (Calbiochem). For treatment with heptanol, the hydrophobic alcohol was diluted 1:1,000 in culture medium by vigorous vortexing and ultrasonication. This preparation was further diluted 1:100 directly into the culture medium, to yield a 0.07 mM solution. Cultures were incubated at 38°C – 5% CO2 in a Gasboy C40 incubator (Labotect) for 4–24 hr. When embryos reached the four- to six-somite stage, cultures were terminated by fixation in 4% paraformaldehyde for 1 hr at room temperature or overnight at 4°C.

Whole-Mount In Situ Hybridization and Histology

Nonradioactive whole-mount in situ hybridization of wild-type and cultured rabbit embryos with digoxigenin-labeled probes was performed following standard methodology. For histological examination, embryos were embedded in gelatin/albumin and sectioned at 30 μm using a Vibratome (Leica).

Statistical Analysis of Data

For statistical evaluation of data, P values were calculated using the χ2 test.


We thank Christoph Viebahn for many valuable suggestions and discussions, Jan Idkowiak and Christoph Viebahn for the rabbit Cx43 plasmid, Grünenthal GmbH (Aachen, Germany) for the kind donation of EM12, and Verena Mauch for help with the in situ analysis of cultures. The critical reading of the manuscript by Thomas Weber is gratefully acknowledged, as well as his help with the preparation of Figures.