A CENTURY OF EXPERIMENTAL ANALYSIS OF LATERALITY IN AMPHIBIANS
Hans Spemann and his co-workers initiated experimental analysis of body plan specification in the early 20th century. To date, probably one of the best-cited embryological works is the famous 1924 Spemann and Mangold paper, “On the Induction of Embryonic Primordia by Implantation of Heterologous Organizers” (“Über Induktion von Embryonalanlagen durch Implantation artfremder Organisatoren”), the molecular analysis of which is still keeping countless scientists busy in the different vertebrate model organisms (Spemann and Mangold,1924). It is less well known, however, that Spemann, before he turned to the organizer, was the first to manipulate the left-right (LR) axis in a defined and predictable manner, i.e., that experimental analysis of laterality was established in the amphibian embryo.
In 1904, Spemann reported inversions of the normal asymmetric arrangement of inner organs (situs inversus) in twinned embryos, which he had created by partial or complete ligature of newt embryos (Triturus taeniatus, Triturus cristatus; Spemann,1904). His observation was in agreement with the then-known fact that situs inversions occur frequently in identical and Siamese human twins. Fifteen years later, Spemann and Falkenberg concluded this work with an in-depth study, in which they described that it was always the right twin that was affected, and that situs inversions occurred in about 50% of cases (Spemann and Falkenberg,1919). A second experimental approach was to rotate the mid-dorsal part of neurula embryos by 180°, including all three germ layers, i.e., neural plate, notochord, somites (including perhaps intermediate and some LPM as well), and the archenteron roof (Fig. 1A, B). These experiments, initiated by Spemann and followed up by his students Kurt Pressler and Rudolph Meyer using yellow-bellied toad (Bombinator pachypus, renamed Bombina variegata), common water frog (Rana esculenta), and common toad (Bufo bufo) embryos, provided an amazing result: a few days later, operated tadpoles revealed a complete inversion of organ placement in most cases (>80%; Fig. 1C; Spemann,1906; Pressler,1911; Meyer,1913).
Hilde Wilhelmi followed up on Falkenberg's ligature experiments to create twins and occasionally found embryos with alterations of organ situs, which were not the result of twinning but of tissue ablations. Her careful analysis revealed that laterality defects only occurred when she induced ablations on the left side of the embryo at mid-neurula stages. Wilhelmi concluded from her experiments and from earlier ones from Spemann and colleagues that “…situs inversion in general was explained by the fact that the left side of the germ has something that the right half does not have” (Wilhelmi,1921). For a good decade now we have known that this “something” is represented by the Nodal signaling cascade in the left LPM, which was uncovered over a 4-year period in chick, frog, mouse, and zebrafish (Levin et al.,1995; Lowe et al.,1996; Logan et al.,1998; Rebagliati et al.,1998; Ryan et al.,1998; Yoshioka et al.,1998; Campione et al.,1999).
CILIA, FLOW, AND SYMMETRY BREAKAGE
Evidence that motile cilia are involved in symmetry breakage has accumulated from 1976 onwards, when Afzelius reported the absence of dynein arms on axonemes of human Kartagener patients (Afzelius,1976). This syndrome comprises, among other cilia-related malfunctions, inversion of organ situs (Fliegauf et al.,2007). Schoenwolf and colleagues first speculated about cilia at the distal pit of the mouse egg cylinder as a determinant of symmetry breakage (Sulik et al.,1994). We have previously argued that this indentation in fact is distinct from the organizer/node, and represents the posterior-most aspect of the notochordal plate (posterior notochord, PNC; Blum et al.,2007). The distinguishing features of node vs. PNC are summarized in Table 1. Hirokawa and colleagues in their ground-breaking work on the Kif3B knockout mouse provided compelling evidence that (1) cilia at the PNC are motile, (2) cilia produce a leftward flow of extracellular fluid, and (3) lack of cilia and/or flow results in laterality defects (Nonaka et al.,1998). Meanwhile, knockout mice and spontaneous mutants have proven that in mouse cilia-driven leftward flow presents the decisive step upstream of the Nodal cascade (Okada et al.,1999; Murcia et al.,2000).
Table 1. Criteria Distinguishing Between the Primary Embryonic Organizer (Spemann's Organizer) and the Ciliated Epithelium, Which Produces a Vectorial Leftward Flow (Left-Right Coordinator)a
The chick, in which the Nodal cascade was initially characterized (Levin et al.,1995), presents an exception in that no flow has been reported thus far. We have previously hypothesized that chick embryos may have lost flow secondarily, as they lack superficial mesoderm and represent a rather derived taxon (Shook et al.,2004; Blum et al.,2009). Flow was, however, present in basal vertebrates; in Kupffer's vesicle (KV) of teleost fish embryos, motile cilia produce a vectorial fluid flow from right to left, and this flow is instrumental for symmetry breakage (Essner et al.,2005; Kramer-Zucker et al.,2005). Evolutionary conservation of this mechanism was demonstrated by the recent finding that amphibian embryos also possess a ciliated epithelium at the gastrocoel roof (gastrocoel roof plate, GRP), where flow develops during neurula stages shortly before the left-sided onset of asymmetric gene expression (Essner et al.,2002; Shook et al.,2004; Schweickert et al.,2007; Blum et al.,2009). PNC, GRP, and KV should represent homologous structures, as they share a great number of structural, molecular, and functional features, which are summarized in Table 2. In particular, these epithelia are always monociliated, derived from superficial mesoderm, bordered by bilateral expression of Nodal, and produce a leftward flow of extracellular fluids (Blum et al.,2007,2009).
Table 2. Determinants and Models of Symmetry Breakage in Frog, Mouse, Zebrafish, and Chick
–, Not reported thus far.
Schweickert et al. (2007); Nonaka et al. (1998); Essner et al. (2005); Kramer-Zucker et al. (2005)
Shook et al. (2004); Sulik et al. (1994); Essner et al. (2002)
Given the evolutionary conservation of flow, the frog offers the opportunity to tackle the major open questions in the field in a highly accessible and potent system, namely (1) What is transferred by flow? (2) How and where is such a determinant perceived? (3) How and by what route is an asymmetric signal transported from the midline to the left LPM? (4) How is asymmetric organ morphogenesis executed downstream of the Nodal cascade? (5) How do early (pre-flow) determinants of laterality connect to flow?
FAST AND RELIABLE ASSESSMENT OF FLOW PARAMETERS IN XENOPUS
Flow in frog was detected many years later than in mouse and fish. The reason for this delay was the hidden nature of the ciliated epithelium in the dorsal gastrocoel roof, although Essner et al. (2002) and Shook and colleagues (2004) had already pinpointed the potential of the GRP and their cilia for LR development. Once exposed in dorsal explants, which can be easily prepared (Fig. 2), flow is readily detected upon addition of fluorescent beads. To better visualize flow, we process time-lapse videos of bead movements to yield gradient time trails (GTTs), i.e., color-coded tracks of beads that reveal direction of transport and velocity of particles (from green to red; 25 sec). Undirected particle movement is eliminated from the analysis to filter out particles moved by Brownian motion. When GTT trajectories are compared using frog, mouse, and rabbit (Fig. 3G–I; and see Suppl. Movie S1, which is available online), flow is clearly directed to the left in all cases. As a qualitative measure, we have introduced the dimensionless number rho, which is the mean resultant length of flow. Rho is calculated by performing a Rayleigh's test of uniformity over the mean angles of bead trajectories. In lay terms, when rho equals 1, all beads project into the same direction. A rho value of 0 reflects movement of beads into all different directions. A robust leftward flow in frog (i.e., at stage 17) is typically characterized by rho values between 0.7–0.9, with similar characteristics in mouse and rabbit. To illustrate the directedness of all beads in an explant, we draw wind roses such as the ones depicted in Figure 3G′–I′, in which bead trajectories are grouped into sectors of 45° each in a circular histogram, and the size of petals reflects the relative frequency with which trajectories are found in a given sector.
Parameters, which are not directly accessible in the frog, are beat pattern and beat frequency of cilia. This is due to the high yolk content of cells, which results in scattering of polarized light. In contrast, mouse and rabbit cilia can be directly observed and videographed. This apparent disadvantage of the frog system can be overcome by fluorescent labeling of axonemal proteins. Park et al. (2008), for example, injected a tau-GFP fusion construct targeted to the epidermis, which labeled the axonemes of epidermal cilia in vivo. In summary, the basic characteristics of leftward flow in frog are conserved compared to mouse, rabbit, and zebrafish embryos.
A number of differences, however, apply as well. In frog, rabbit, and mouse, the epithelia themselves are basically flat, while the fish KV may be dome- (medaka) or sphere-shaped (zebrafish). In mouse, the margins of the ciliated epithelium, i.e., the transit zone to the flanking endoderm, is elevated towards the ventral side (Fig. 3E), a feature not seen in Xenopus. It has been argued that this elevation may be of functional relevance, i.e., to provide a barrier, at which vesicles transported by flow would be smashed (Nakamura et al.,2006). The GRP topology, however, does not support this notion (see Fig. 3D). A second noteworthy feature is the great variability of dimensions of the respective structures. While the PNC in mouse measures 50 μm from left to right, this amounts to about 100 μm in rabbit, and approximately 200 μm across the Xenopus GRP, and KVs in fish range between 50–150 μm (Cooper and D'Amico,1996; Melby et al.,1996; Okada et al.,2005; Blum et al.,2007,2009; Schweickert et al.,2007). The large size of the GRP not only allows for easier imaging, but also simpler preparation and handling compared to the PNC in mouse and rabbit. The KV in teleost fish embryos, representing the remnant of the archenteron (Cooper and Virta,2007), presents itself as a closed sphere that very much limits access to cilia, requiring technically challenging 3D imaging equipment.
Comparing all systems used to date, the frog GRP offers distinct advantages: (1) a great number of explants can be prepared and analyzed in a short time, facilitating statistical analysis; (2) qualitative and quantitative parameters can be determined with statistical significance, such as velocity, directionality, and the ratio of directed (flow) vs. random (Brownian) movements; and (3) stages unequivocally predict the status of flow, with very little variability. Stage 17/18 neurulae will in all cases reveal robust leftward flow in dorsal explants. In contrast, the impossibility to stage mouse and rabbit embryos in utero with the required accuracy results in the frequent recovery of pre- or post-flow embryos, necessitating large numbers of animals to be processed in order to receive statistically valid data. Taken together, the frog system offers a highly standardized assay system for the analysis of cilia-driven leftward flow.
EXPERIMENTAL MANIPULATIONS OF GRP AND FLOW
A straight-forward approach to manipulate flow is its direct elimination by placing a droplet of methylcellulose (MC), i.e., a highly viscous liquid, onto the ciliated epithelium of the GRP by injection into the archenteron at stage 14–17. In such experiments, we achieved on average 50% of embryos without induction of Xnr1 or Pitx2 in the left LPM (Schweickert et al.,2007; and our unpublished data). This manipulation established leftward flow as the decisive step in setting up embryonic laterality.
Shook et al. (2004) described the derivation of the GRP from superficial mesoderm (SM), i.e., surface epithelial cells that give rise to the outer cell layer of the dorsal lip of the blastopore at stage 10–10.5. These cells invaginate during gastrulation and form a distinct epithelium of small ciliated cells from stage 12/13 onwards. They are positioned in the posterior roof of the archenteron, from where they fold off successively from stage 15 onwards to integrate into deep mesoderm (notochord, hypochord, and somites; Shook et al.,2004). In order to analyze whether SM is required for laterality development, we used classical microsurgery to remove this tissue at stage 10 (Fig. 4A). Ablated embryos developed through gastrulation and neurulation without apparent dorso-anterior axis defects (DAI = 5); however, organ situs at stage 45 was randomized. Positioning of the gall bladder and heart and gut looping were inverted (situs inversus) or developed discordantly (heterotaxia; Fig. 4B). This rather simple experiment demonstrates that the SM-derived GRP is strictly required for LR development. In zebrafish, a comparable experiment has been reported, in which the mature KV was mechanically destroyed by injection of water (Bajoghli et al.,2007). In these experiments, no other defects except laterality disturbances resulted, strongly suggesting that KV and GRP indeed are functionally homologous.
As a prerequisite of further manipulations, the GRP lineage needs to be known. For this purpose, we injected LacZ mRNA at the 4–32 cell stage. The dorsal-marginal region of the 4-cell embryo (Fig. 5A,B) or the dorsal-marginal blastomere C1 of the 32-cell embryo define the GRP lineage most accurately (Fig. 5C, D). Depending on the stage of injection, the floor plate is targeted with a certain likelihood as well. This should be kept in mind when ciliary determinants are addressed by experimental manipulations, as in these cases neural tube closure defects may ensue, possibly resulting in misjudgment of embryonic stages. A unique possibility of the Xenopus system is the opportunity to target the GRP in a sided manner, such that the left or right half is specifically hit (Fig. 5E, F). Such manipulations are not available in any other vertebrate model organism in which flow was characterized.
We have started to use these injection schemes to specifically knock down axonemal proteins and other known left-right determinants. As an example, we have targeted a dynein heavy chain gene, because the human Kartagener syndrome and the mouse iv mutant both are caused by mutations in axonemal dyneins. In agreement with the genetic mutants, we observed laterality defects and abnormal flow (our unpublished results), demonstrating that frog morphants can effectively mimic mutations generated in mouse or zebrafish. The frog system offers the opportunity to perform multiple tissue-specific gene knock-downs as well as combinations of loss- and gain-of-function experiments, schemes that are much more complicated to set up in mouse or zebrafish. In addition, epistasis experiments can be designed by combining knock-downs and experimental manipulations such as MC treatment or microsurgery. Finally, ablations and transplantations (Spemann and Mangold,1924; Borchers et al.,2000; Ohi and Wright,2007) can be combined with knock-downs and drug treatments, to avoid deleterious effects due to collateral damage (i.e., targeting of unwanted tissues along the same lineage).
The perspective for such approaches is bright, as every new player published in any system can be assessed directly in the frog. The recent advent of transgenic frog technologies in Xenopus tropicalis, which has a GRP and flow just like Xenopus laevis (Blum et al.,2009), provides further options for generating and analyzing mutant lines using all of the above-mentioned techniques. The first three examples of mutants with LR defects have already been published (Noramly et al.,2005), and more should become available as mutagenesis projects proceed (Amaya,2005; Grammer et al.,2005; http://tropicalis.berkeley.edu/home).
THE POTENTIAL OF THE FROG TO ADDRESS PRESSING LEFT-RIGHT ISSUES
Two models are currently being discussed for flow-mediated symmetry breakage. The morphogen model postulates that a factor is transported by flow to the left side of the ciliated epithelium (Nonaka et al.,1998; Hirokawa et al.,2006). At the left margin, it is perceived by an unknown receptor, from where the signal is transferred by an unknown route to the left LPM. A morphogen could either be released from the margins, or arise from the ciliated epithelium itself (Okada et al.,1999), for example via FGF-triggered extrusion of factor-bearing vesicles, as has been described in mouse where vesicles were shown to harbor retinoic acid and SHH (NVP model; Tanaka et al.,2005). The alternative two-cilia hypothesis postulates that cilia-driven flow results in left-sided bending of a second class of cilia, i.e., mechanosensory cilia, which in turn via secondary signals such as calcium would lead to left-sided Nodal transcription in the LPM (McGrath et al.,2003; Tabin and Vogan,2003). While none of the components of these two models have been analyzed in the frog thus far, the experimental repertoire outlined above should have a great potential to address both models. For example, unilateral knock-down of flow could answer the question whether the right side of the GRP is necessary for symmetry breakage, i.e., whether a morphogen enters the GRP from the right side. Methylcellulose should block transport of vesicles, provided they exist in frog. Combinations of MC-treatment and left- or right-sided supplementation with retinoic acid and/or SHH should potentially restore the Nodal cascade on the left or right side. Such experiments may thus have the power to distinguish between the two models.
Transfer of Asymmetric Cue(s) From the GRP to the Left LPM
The sole readout of flow upstream of the induction of Nodal in the left LPM is an asymmetric calcium wave at the left margin of PNC in mouse and KV in zebrafish (McGrath et al.,2003; Sarmah et al.,2005; Hadjantonakis et al.,2008). How this calcium signal acts on LPM Nodal is currently unknown. An attractive alternative way has been proposed by Hamada and colleagues (Oki et al.,2007). They suspect that Nodal itself reaches the LMP in a flow-dependent manner from the left margin of the PNC via the extracellular space between endoderm and mesoderm (Fig. 6). In support of this notion, Brivanlou and colleagues have shown that Nodal at the left GRP margin is required for induction of asymmetric genes in the LPM (Vonica and Brivanlou,2007). In principle, a signal in the frog could also travel through the endoderm or outside of the endoderm through the gastrocoel, through the somitic and intermediate mesoderm, or even through the notochord and ectoderm (Fig. 6). All of these options can be tested in the frog, for example by combining MC-mediated blockage of flow with sided release of calcium (caged calcium; ionophore), or by using a tagged Nodal construct.
Asymmetric Organ Morphogenesis
A prerequisite for asymmetric organ morphogenesis is the sustained maintenance of the Nodal cascade on the left side. A model was proposed in mouse and Xenopus to account for this fact, involving positive (Nodal) and negative (Lefty) feedback loops (Lohr et al.,1998; Meinhardt and Gierer,2000; Nakamura et al.,2006; Ohi and Wright,2007). As the Nodal cascade has been known for >10 years in frog, it is surprising that organ morphogenesis has been investigated only rarely (Breckenridge et al.,2001; Gormley and Nascone-Yoder,2003; Muller et al.,2003). The heart lineage in frog has been well studied and described in depth, as well as morphogenesis of the gastrointestinal tract. The molecular mechanisms of asymmetric morphogenesis, for example how the Nodal-induced transcription factor Pitx2 exerts its function, i.e., what the transcriptional targets are, have largely remained unresolved. It may be worthwhile to explore the experimental toolbox for the analysis of organ morphogenesis, which in frog as in all other vertebrates involves bending of linear tubes in the first place. Thus, despite the three-chambered nature of the adult frog heart, the initial steps of asymmetric morphogenesis should be conserved (Ramsdell et al.,2006).
A number of additional components with influence on the LR pathway have been identified in Xenopus without obvious link to cilia and flow. Brief pharmacological disruption of cortical actin during the first cell cycle randomizes the LR orientation of tadpole heart and gut, representing the earliest acting process affecting laterality reported thus far (Danilchik et al.,2006). The maternal growth factor Vg1 has been described as a left-right coordinator, as its overexpression in a single right ventro-lateral blastomere at the 16-cell stage can (unlike other TGFβ growth factors) fully invert the organ situs in > 80% of cases (Hyatt et al.,1996; Hyatt and Yost,1998). The proteoglycan Syndecan-2, initially expressed in the animal cap of Xenopus blastula and gastrula embryos, is thought to act as a co-receptor for left asymmetric Vg1 signaling in the involuting mesoderm. Supporting this notion, asymmetric phosphorylation of Syndecan-2 in the right involuting mesoderm was reported during early gastrulation, and thus still upstream of flow stages. This phosphorylation, which depended on protein kinase C γ (PKCγ), was instrumental for laterality determination, as both loss- and gain-of-function of both factors effectively altered heart looping and asymmetric marker gene transcription (Kramer et al.,2002; Kramer and Yost,2002). It should be noted that none of these three determinants was reported to be relevant for LR asymmetry in any other vertebrate system so far.
Gap junctional communication (GJC) was originally suspected in the LR pathway due to connexin 43 mutations in human laterality patients (Britz-Cunningham et al.,1995). Gain- and loss-of-function experiments have indicated that GJC may act during cleavage (frog) or gastrulation (chick) (Levin and Mercola,1998,1999). In rabbit embryos, however, we recently showed that GJC act downstream of flow in the transfer of asymmetric cue(s) from the midline to the LPM (Feistel and Blum,2008). Our preliminary experiments confirm this observation in Xenopus as well (T.B. and M.B., unpublished). It is currently unclear whether GJC in addition act at earlier stages, and if and how such earlier functions impact on GRP and flow. Another component of cell-cell interactions has been identified with the tight-junction protein claudin, originally identified in a screen for proteins enriched in microsomes in the blastula (Brizuela et al.,2001). Upon Xcla overexpression, Xnr1 was bilaterally expressed in about 50% of cases, and the visceral situs was randomized (Brizuela et al.,2001). It will be interesting to analyze whether flow is affected in these experiments, particularly as Xcla may be expressed in the GRP (see fig. 2G,H in Brizuela et al.,2001). Claudin may be a particularly important factor, because it has been involved in laterality determination in the chick upstream of the Nodal cascade as well (Simard et al.,2006).
Other presumably early acting determinants (i.e., during cleavage) comprise the ion pumps H+-K+-ATPase and V-ATPase as well as the monoamine Serotonin and the 14-3-3 family member E (Levin et al.,2002; Bunney et al.,2003; Adams et al.,2006). These factors have been identified using pharmacological inhibitors. In addition, H+-K+-ATPase mRNA and Serotonin were reported to be asymmetrically localized in the 4- and 64-cell embryo, respectively (Levin et al.,2002; Fukumoto et al.,2005). The relevance of asymmetric localization of H+-K+-ATPase mRNA seems questionable, however, as Levin and colleagues recently reported significant variabilities of in situ signals in embryos from different females (Aw et al.,2008). Given the robustness of the LR pathway, occasionally encountered asymmetries in mRNA localization cannot possibly account for the asymmetric induction of the Nodal cascade. However, H+-K+-ATPase activity certainly influences LR asymmetry, particularly because pharmacological inhibitors also affect laterality in chick (Levin et al.,2002). As the specificity of pharmacological drugs may be limited due to solubility, half-life, diffusion rate, and accessibility of target tissue (e.g., SM following invagination), to name just a few critical parameters, all of these factors should be addressed by gene-specific knock-downs in the tissue of interest in the future. GRP morphogenesis and flow as readout should be given special emphasis, as these stages are reached several hours before marker genes can be assessed, and 3–4 days before organ situs becomes evident. In summary, linking up early determinants to GRP cilia and flow will uncover whether some or all of these factors act in parallel pathways, or, as we would expect from our preliminary experiments, play important roles during flow stages by influencing different kinds of flow-related steps: GRP morphogenesis, ciliogenesis, ciliary beat frequency, and cilia polarization.
Over the past decade, experimental investigation of LR asymmetry in the frog Xenopus was governed by the conviction that mechanisms differ greatly from mouse and fish. With the discovery of flow as the evolutionary conserved common denominator between these three species and, thus, between fish, amphibians, and mammals, future work should clearly have the potential to develop a unifying hypothesis for symmetry breakage. If this hope comes true, this will be particularly satisfying because it was difficult in the past to assume radically different modes of symmetry breakage upstream of a highly conserved module, i.e., the Nodal cascade. We, therefore, expect a new focus on Xenopus as a model organism to study LR asymmetry, particularly with respect to the most pressing issues of symmetry breakage, the transfer of asymmetric cues to the left LPM, and the linking-up of early determinants with flow stages and flow as mechanism. Maybe even Spemann's experiments, which in the light of our present day knowledge still await molecular elucidation, can be tackled anew using the repertoire of modern Xenopus technologies.
We apologize to those colleagues whose work we could not cite due to space limitations. We thank all members of the Blum lab for lively discussions, valuable suggestions, and critical reading of the manuscript. T.B., T.W., and P.V. were recipients of Ph.D. fellowships from the Landesgraduiertenstiftung Baden-Wuerttemberg. Frog work in the Blum lab was funded by a grant from the Deutsche Forschungsgemeinschaft.