TGF-β Superfamily Signaling and Left-Right Asymmetry

See allHide authors and affiliations

Science's STKE  09 Jan 2001:
Vol. 2001, Issue 64, pp. re1
DOI: 10.1126/stke.2001.64.re1


Despite an outwardly bilaterally symmetrical appearance, most internal organs of vertebrates display considerable left-right (LR) asymmetry in their anatomy and physiology. The orientation of LR asymmetry with respect to the dorsoventral and anteroposterior body axes is invariant such that fewer than 1 in 10,000 individuals exhibit organ reversals. The stereotypic orientation of LR asymmetry is ensured by distinct left- and right-side signal transduction pathways that are initiated by divergent members of the transforming growth factor-β (TGF-β) superfamily of secreted proteins. During early embryogenesis, the TGF-β-like protein Nodal (or a Nodal-related ortholog) is expressed by the left lateral plate mesoderm and provides essential LR cues to the developing organs. In chick embryos at least, bone morphogenetic protein (BMP) signaling is active on the right side of the embryo and must be inhibited on the left in order for Nodal to be expressed. Thus, at a key point in the determination of LR asymmetry, left-sided signaling is mediated by the transcription factors Smad2 and Smad3 (regulated by Nodal), whereas signaling on the right depends on Smad1 and Smad5 (which are regulated by BMP). This review summarizes the considerable progress that has been made in recent years in understanding the complex network of feedback and feedforward circuitry that regulates both the left- and right-sided pathways. Also discussed is the problem of how signal transduction mediated by the Smad proteins can pattern LR asymmetry without interfering with coincident dorsoventral patterning, which relies on the same Smad proteins.


Distinct left- and right-sided programs of gene expression in the early embryo ensure concordant left-right (LR) asymmetric anatomy and physiology of the visceral organs. Disruption of LR patterning, in particular at early points in the programs, causes organ arrangement (heterotaxia) syndromes characterized by striking mirror-image reversals of one or more organs, or in some cases, situs inversus, in which the LR asymmetry of the entire viscera is inverted. Advances in the understanding of LR asymmetry point to the central roles played by two sets of transforming growth factor-β-like (TGF-β-like) proteins and their intracellular signal transducers, the Smad proteins [reviewed in (1)]. The TGF-β-like factors [bone morphogenetic proteins (BMPs), acting on the right] and Nodal and its orthologs (acting on the left) transduce early LR cues to the developing organs. A unifying picture of how LR information is propagated seems tantalizingly close to realization yet remains obscured by differences in experimental approaches and species studied. What emerges, however, is a remarkable network of feedback and feedforward circuitry.

Mechanistically, the acquisition of LR pattern can be considered to occur in three broad steps [for review, see (2, 3)]. First is the initial polarization of LR asymmetry with respect to the dorsoventral (DV) and anteroposterior (AP) axes of the embryo. TGF-β-like factors then function in the second step to reinforce LR asymmetric polarity and to transmit this information to the developing organ primordia. Third, individual organs undergo asymmetric morphogenesis. Little is known about the step that first orients LR asymmetry relative to the embryo's DV and AP body axes, except that the proteins and processes that organize LR asymmetry are quite different from those that establish the other axes. This probably reflects the unique problem of orienting orthogonal body axes rather than defining a particular position along a fixed axis, as is the case for DV, AP, and mediolateral patterning. An early step in the orientation of LR asymmetry in mice appears to involve monocilia that have been shown to be capable of propelling particles leftward over the ventral surface of the node (4, 5). A role for cilia provides a satisfying explanation for clinical and experimental data that link defects in LR asymmetry to ciliary dyskinesia and mutations in motor or other axonemal proteins (6-10). However, whether directional flow or some other aspect of ciliary motility is necessary to polarize LR asymmetry is not entirely clear [for review, see (11)].

Regardless of how orientation is achieved, its direction must be converted into a signal that can be interpreted by the organ primordia and propagated throughout the embryo. Experiments in Xenopus and in chicks implicate gap junctional communication in this process (12, 13). In chicks, such early signals result in asymmetric expression of sonic hedgehog (Shh) on the left side of Hensen's node at Hamburger-Hamilton (HH) stages 5 and 6 (12, 14-18). Although in Xenopus, zebrafish, and mice (19-21), hedgehog genes are bilaterally transcribed, and a Shh mutation (sonic you) in zebrafish does not affect laterality [although other hedgehog homologs could be involved (22)]. Nonetheless, experimental perturbation of Hedgehog signaling does alter laterality in these species [for example, see (19, 21, 23-25)]. However, potential secondary effects caused by altered midline patterning have made it difficult to determine the actual role of Hedgehog signaling in LR axis orientation in any species other than chick.

Downstream from signals that initiate LR polarization, LR information is propagated by BMPs, acting on the left side of the embryonic midline, and Nodal orthologs, acting on the right. Experiments with mice, frogs, fish, and chicks have uncovered many aspects of how these factors reinforce the initial LR information and transmit it to the developing organs [reviewed in (2, 3)]. Observations of asymmetric expression of mRNAs encoding several TGF-β ligands and their regulators provided the first evidence of the central role played by these proteins in the propagation of LR asymmetric signals. Furthermore, perturbing the TGF-β-like signals can lead to disrupted or reversed asymmetric gene expression and altered morphogenesis (Fig. 1). In the chick, suppression of BMP function on the left establishes the conditions for signaling by Nodal orthologs. In contrast, BMP signals on the right side of the embryonic midline suppress the expression of Nodal, Lefty, and, consequently, downstream genes such as Pitx2. Thus, left-sided signaling is mediated by the transcription factors Smad2 and Smad3 (which are regulated by Nodal), whereas signaling on the right depends on Smad1 and Smad5 (which are regulated by BMPs).

Left-Sided Signaling

nodal is expressed early in the developing node and primitive streak in gastrulating mouse embryos and later in the left-side lateral plate mesoderm (LPM) during neural tube stages (20, 26). Left-sided expression of nodal orthologs in post-neurula embryos appears to be a general feature of chordate development and has been reported in mouse, chick, zebrafish, frog, and ascidian embryos (14, 19, 20, 26-28). lefty1 and lefty2 also encode divergent TGF-β superfamily ligands and are expressed in the left LPM and floor plate coincident with nodal in mouse embryos [(29, 30); for review, see (31)]. lefty homologs with similar expression patterns have been identified in chicks, frogs, zebrafish, mice, and humans (30, 32-40). The structural features of Lefty proteins in mice and humans suggest that the corresponding genes arose by duplication independently after the divergence of human and mouse lineages (40); thus, there is not a one-to-one correspondence between the human and murine orthologs. Mutations in mouse and zebrafish that alter organ location (situs) also alter expression of nodal and lefty genes, suggesting the involvement of these genes in LR determination (4, 23-25, 29, 30, 32, 41-44).

Although nodal and lefty share the property of being expressed on the left side of the embryo, both gain and loss of function experiments indicate that they have radically different roles in the specification of LR pattern. Ectopic expression or application of Nodal in the right side of the frog or chick embryo is sufficient to create unbiased asymmetric gene expression and organ situs, suggesting that Nodal generates an essential left-determining signal (15, 19). However, loss of function experiments have not clearly demonstrated that Nodal's role in LR patterning is necessary. This is most likely because of confounding effects: loss of Nodal disrupts midline patterning and may thereby perturb LR asymmetric gene expression and development (20, 38, 45, 46). Insight into the function of Nodal in LR asymmetry, however, is provided by analyses of laterality disturbances caused by mutation of cryptic and OEP (one-eyed pinhead) genes, which encode essential cofactors for Nodal signaling in mouse and zebrafish (47). Cryptic/OEP mutant embryos exhibit heterotaxia and a loss of left LPM expression of nodal, lefty2, and Pitx2, supporting an essential role for Nodal signaling in left-sided specification (22, 27, 41, 42). In addition, the loss of left LPM nodal expression in these mutants suggests that Nodal feeds back to elevate its own expression. Whereas Nodal promotes leftness, Lefty appears to inhibit it. Loss of Lefty1 function leads to bilateral expression of left-side markers, including nodal (32). Chick, zebrafish, mouse, and Xenopus Lefty homologs (the Xenopus protein is known as antivin) antagonize Nodal signaling in embryological assays (33, 38, 48). Thus, the expression of Lefty1 in the midline is presumed to provide a barrier to prevent the propagation of Nodal signals to the right side. Because Nodal induces the expression of lefty2, a negative feedback loop likely occurs in the left LPM where lefty2 and nodal overlap.

Characterization of the signal transduction pathway regulated by Nodal and its orthologs has begun to clarify how these factors control left-sided gene expression. Mapping of enhancer elements from the nodal and lefty2 genes that drive left-side specific gene expression in the mouse revealed the presence of two binding sites for the winged helix transcription factor FAST (49). In Xenopus, FAST regulates mesodermal gene transcription in response to Nodal and activin-like signals by interaction with Smad2 and the common cofactor Smad4 (50-52). Although FAST interactions with Smad2 have been studied most extensively, FAST also interacts with Smad3 to transduce activin-like signals (53), and both Smad2 and Smad3 can function similarly downstream of Nodal (54). On some promoters, however, Smad2 and Smad3 might function differently (55). Importantly, the FAST sites in the nodal and lefty2 enhancers are both necessary and sufficient for left-side specific gene expression in the mouse (49). Moreover, in the Xenopus animal cap assay, the FAST sites are sufficient to confer Nodal or activin responsiveness on a luciferase reporter. Thus, Nodal acts through the FAST-Smad pathway to regulate both its own expression and that of lefty2 in the left LPM. In addition, the presence of a FAST-regulated module is conserved for nodal orthologs from Xenopus ascidians, indicating that the Nodal → FAST-Smad2-Smad3 pathway is an ancient feature of left-sided signaling that has been conserved throughout chordate evolution (56).

Identification of nodal-regulated FAST sites in the nodal and lefty2 genes indicates that the FAST-Smad2-Smad3 pathway positively regulates nodal and lefty2 expression and that expression of nodal and lefty2 is involved in both positive and negative feedback loops in the left LPM. In addition to these feedback loops, does this pathway function downstream of Nodal to convey LR cues to developing organs as well? Pitx2 encodes a homeobox gene expressed on the left side of the diencephalon, heart tube, and gut and is implicated as an effector of asymmetric morphogenesis (57-65). As with nodal orthologs, Pitx2 genes in both mouse and frog have FAST regulatory modules (66). Additional downstream effectors of asymmetric Nodal signaling have not yet been isolated but are likely to exist. It will be interesting to learn whether they also contain FAST regulatory modules and whether the FAST-Smad2-Smad3 pathway is a common mechanism for transducing left-sided Nodal signals to asymmetric organs. Finally, in addition to being downstream of Nodal, ectopic expression studies showed that Pitx2 can activate expression of both its own gene and that of nodal (67, 68), suggestive of another positive feedback loop that could act to reinforce left-sided signaling.

Right-Sided Signaling

Whereas activation of Smad2 and Smad3 signal transduction promotes or sustains left-sided-specific gene expression, signaling mediated by Smad1 and Smad5 is implicated downstream of BMPs in suppression of left-sided gene expression. Moreover, inhibition of Smad1-Smad5 signals may be a critical event in the onset of Nodal signaling on the left side. In the chick, mRNA encoding Caronte (also known as cCerberus1), a member of the Cerberus-DAN family of secreted proteins that inhibits actions of BMP (as well as Nodal and Wnt), is expressed unilaterally on the left side of the embryo beginning at HH stage 5 (34, 37, 69). Ectopic inhibition of BMP signaling on the right by either Caronte or another BMP antagonist, Noggin, is sufficient to induce expression of nodal, whereas overexpression of BMP on the left suppresses normal left-sided gene expression. Because Nodal cannot induce expression of Caronte in the right LPM, the apparent function of Caronte is to permit Nodal expression adjacent to the node and later in the left LPM (Fig. 1).

The model in which BMP signaling is permitted on the right and inhibited on the left is consistent with the finding that mice lacking Smad5 have bilateral expression of nodal (70). Similarly, expression of a dominant-negative BMP receptor on the right side of Xenopus embryos also induced the expression of nodal ortholog XNr1 (71). However, an activated BMP receptor expressed on the left not only prevented XNr1 expression on the left side, as predicted, but also led to aberrant bilateral and right-only XNr1 expression, which would not be predicted by the chick model. Nonetheless, in an animal cap assay, ectopic activation of BMP signaling suppressed the ability of XNr1 to activate the FAST regulatory cassette, indicating that BMPs can interfere with the positive feedback loop that maintains left-sided XNr1 expression (56). Although these data generally support the idea that BMPs antagonize Nodal expression and function, so far only the identification of Caronte/cCerberus1 in the chick provides direct evidence that inhibition of BMP signaling initiates the left-sided signaling cascade. Thus, it will be important to learn how widespread inhibition of BMP signaling is in chordates as a means to initiate left-sided gene expression.

Timing and the Problem of Keeping LR and DV Signals Distinct

The emerging picture is that left-sided signaling by Nodal, mediated through Smad2 and Smad3, is set into motion by the inhibition of BMP signaling and then is limited in extent and duration by negative feedback through Lefty. On the right side, in contrast, BMP signaling through the Smad1-Smad5 pathway is permitted, suppressing the left-sided program and perhaps initiating a poorly characterized right-sided signaling cascade. The interpretation of experiments involving ectopic manipulation of TGF-β superfamily signaling pathways is complicated, however, by technical limitations associated with how ectopic signals are introduced. Injection of synthetic mRNAs encoding Vg1, XNr1 (the Nodal ortholog), or activin have quite distinct effects on LR pattern in Xenopus embryos (72). These data are currently difficult to reconcile from a signaling perspective: if each of these ligands activates the Smad2-Smad3 pathway, then why does Vg1 cause greater LR reversals than do the other ligands? Although the possibility cannot be excluded that one or another of these ligands activates pathways in addition to the Smad2-Smad3 pathway, several recent experiments suggest an alternative possibility. When activin is delivered as protein by microinjection to Xenopus tissues after the neural plate stage (stage 13), LR asymmetry is reversed (73), even though it does not do so effectively when synthesized chronically from mRNA injected at early cleavage stages (72). This observation suggests that activation of Smad2 and Smad3 at inappropriately early times may disrupt or fail to support left-side determination. But why do activin and Vg-1 elicit different phenotypes when misexpressed early on? Recent experiments showed that ectopically expressed activin was found to activate Smad2 before the mid-blastula transition (MBT), whereas activated Vg-1 only activated Smad2 after the MBT (74). The basis for this differential responsiveness is not known, but it may involve differential expression of ligand binding (type II) receptors or accessory molecules. These data raise the possibility that differential responsiveness, rather than differences in transduction pathways, account for the distinct effects of activin and Vg1 on LR pattern.

Although it is not surprising that the timing of signaling is critical, timing is often difficult to mimic experimentally, especially when relying on chronic misexpression of DNA or mRNA, or the chronic and systemic effects of germ line mutations. Differences in cellular responsiveness, ligand diffusion, and signal down-regulation may all be critical determinants of the distinctive phenotypes elicited by different TGF-β superfamily ligands, even if they signal through a common Smad. An excellent example of the effect of timing and context of signals on patterning is provided by the differential effects of Nodal and activin-like signals (both presumed to be mediated by Smad2 and Smad3) in the chick. Before HH stage 5, endogenous activin-receptor signaling appears to suppress Shh and Nodal expression on the right side of the node, and this can be mimicked by application of activin-saturated beads on the left side of the embryo at HH stage 4 (14), indicating that Smad2-Smad3 signaling provides a very early right-sided signal. However, application of Nodal later (at HH stages 6 and 7) induces leftness, regardless of the side of the embryo to which it is applied (15), demonstrating that ligand or context or both alters the outcome of Smad2-Smad3 activation. Timing also appears to be critical in the case of Lefty proteins: when applied to the right LPM of chick embryos at HH stage 4, Lefty represses expression of the left-sided genes nodal and Pitx2 (34), but when added to the right LPM later (at HH stage 5), Lefty paradoxically induces the expression of nodal and Pitx2 on the right (34, 61). Although the current model emphasizes Lefty's ability to inhibit expression of Nodal, it is important to remember that, at high doses or at different developmental stages, Lefty proteins might have quite different activities (e.g., inhibit BMPs or activate receptors). An important consideration, therefore, is that ectopic (and often chronic) manipulation of ligands at times, doses, or locations that are not physiologically appropriate can be misleading, particularly in a system controlled by a complex set of positive and negative feedback loops of structurally similar regulators.

Understanding the influence of timing is closely tied to the fundamental question of how the vertebrate embryo uses the same signaling pathways for patterning along different body axes. FAST-Smad2-Smad3 and Smad1-Smad5 pathways mediate signals that are important for both LR and DV patterning. Moreover, FAST-Smad2-Smad3 interactions induce early mesoderm-specific gene expression (50). Even expression of Pitx2 can be induced in pre-gastrula prospective ectoderm and in post-neurula LPM by inappropriate activation of Nodal-like signaling (67, 68). Presumably, the context in which TGF-β superfamily signals are being received prevents cells from confounding mesoderm-inducing, DV- and LR-patterning signals. Determining the molecular basis of this contextual information, and how it changes during development, will be critical for understanding how the ultimate readout of TGF-β signaling is determined.

Evolutionary Similarity Between LR and DV Patterning

The antagonism between Smad2-Smad3 (Nodal/activin-like) and Smad1-Smad5 (BMP) signals that characterize the left and right sides, respectively, of the streak stage chick embryo is mirrored by a similar dichotomy between the dorsal and ventral sides of pre-gastrula stage embryos. In all vertebrate embryos so far examined, Smad2-Smad3 signals induce dorsal mesodermal genes and dorsal structures, whereas Smad1-Smad5 signals work in opposition to induce mesoderm with ventral characteristics. Does this indicate a remarkable parsimony, or does it reflect the evolutionary origin of LR asymmetry in chordates?

One theory on the origin of the bilateral symmetry and LR asymmetry maintains that LR asymmetries of chordates is due to a 90° rotation of the body axis (75). In this view, known as "dexiothetism," a presumed LR-symmetrical (but DV-asymmetrical) ancestor, which may have resembled the hemichordate Cephalodiscus, came to lie on its right side on the ocean floor. Over time, its descendants that evolved into chordates underwent a shift in anatomy that converted the originally symmetrical left and right sides into new and distinct dorsal and ventral sides. Although the originally asymmetrical DV axis acquired substantial bilateral symmetry (evident in the outward appearance of human LR bilateral symmetry), the underlying patterning mechanisms might have been preserved to coordinate the LR asymmetry characteristic of internal organs and ontogeny of extant descendants. The evolutionary aspect of this model is controversial because it predicts that echinoderms (which have LR asymmetries in their early development) and chordates are derived from this common ancestor but that hemichordates, which lack LR asymmetry, are more distantly related. Although this model does not predict many of the molecular aspects of LR asymmetry, it is nonetheless intriguing to consider that the reliance of both left and dorsal signaling on Smad2 and Smad3 and right and ventral signaling on Smad1 and Smad5 might have first occurred in the present LR axis and was reiterated upon the creation of a new DV axis. Although the parallels between DV and LR patterning are superficial and might turn out to be coincidental, the dexiothetism model is worth bearing in mind as further studies are carried out on a molecular level to see if they provide support for this evolutionary scenario. Regardless of how DV and LR patterning are linked evolutionarily, it will be interesting to see if the parallels in DV and LR signaling for TGF-β-like factors also hold for other signaling molecules. For instance, Wnt signaling generates patterning along the DV axis, but its role in LR asymmetry awaits description.


Additional research is needed to understand the complex regulatory circuits that initiate and stabilize signaling by Nodal orthologs and Smad2 and Smad3 on the left and BMPs and Smad1 and Smad5 on the right sides of the embryo. It is a perplexing problem to understand how similar signal transduction machinery (e.g., Smad2) can relay distinct LR asymmetry signals and also convey DV information at different times or in different tissues within the developing embryo. Part of the answer may lie in the identification of novel factors and/or regulatory mechanisms that modulate signaling by TGF-β-like factors. Precedents exist for the modulation of signaling at the levels of ligand processing [for instance, see (76-80)] and receptor-ligand interaction (81-83) in addition to the regulation of signal transduction and gene expression discussed here. It will be of great interest to understand how these mechanisms provide the embryological context that governs a cell's response to a particular TGF-β-like factor.

Fig. 1.

Signaling pathways of LR development [redrawn from (2)]. The dashed line indicates the midline of each organism, with the left side to the reader's left. Known interactions are illustrated. Gray shading indicates that a pathway or expression of a particular gene or protein is suppressed (e.g., Nodal on the right side of chick embryos). Particular Smad proteins are indicated where their involvement has been implicated experimentally. Lack of an indicated Smad reflects an absence of knowledge rather than a lack of involvement. For clarity, certain genes that are involved in LR asymmetry but not positioned in a particular signaling cascade are omitted (e.g., cryptic is symmetrically expressed in the mouse, and a null mutation affects LR patterning). Moreover, individual Pitx2 isoforms are not indicated but are themselves differentially regulated and differentially regulate downstream targets (67, 68). FGF8, fibroblast growth factor 8; CSnR, chick snail-related.


  1. 1.
  2. 2.
  3. 3.
  4. 4.
  5. 5.
  6. 6.
  7. 7.
  8. 8.
  9. 9.
  10. 10.
  11. 11.
  12. 12.
  13. 13.
  14. 14.
  15. 15.
  16. 16.
  17. 17.
  18. 18.
  19. 19.
  20. 20.
  21. 21.
  22. 22.
  23. 23.
  24. 24.
  25. 25.
  26. 26.
  27. 27.
  28. 28.
  29. 29.
  30. 30.
  31. 31.
  32. 32.
  33. 33.
  34. 34.
  35. 35.
  36. 36.
  37. 37.
  38. 38.
  39. 39.
  40. 40.
  41. 41.
  42. 42.
  43. 43.
  44. 44.
  45. 45.
  46. 46.
  47. 47.
  48. 48.
  49. 49.
  50. 50.
  51. 51.
  52. 52.
  53. 53.
  54. 54.
  55. 55.
  56. 56.
  57. 57.
  58. 58.
  59. 59.
  60. 60.
  61. 61.
  62. 62.
  63. 63.
  64. 64.
  65. 65.
  66. 66.
  67. 67.
  68. 68.
  69. 69.
  70. 70.
  71. 71.
  72. 72.
  73. 73.
  74. 74.
  75. 75.
  76. 76.
  77. 77.
  78. 78.
  79. 79.
  80. 80.
  81. 81.
  82. 82.
  83. 83.
  84. 84.
View Abstract

Navigate This Article