What's left in asymmetry?


  • Sherry Aw,

    1. Biological and Biomedical Sciences Program, Harvard Medical School, Boston, Massachusetts
    2. Center for Regenerative and Developmental Biology, Forsyth Institute, Boston, Massachusetts
    Search for more papers by this author
  • Michael Levin

    Corresponding author
    1. Center for Regenerative and Developmental Biology, Forsyth Institute, Boston, Massachusetts
    2. Developmental Biology Department, Harvard School of Dental Medicine, Boston, Massachusetts
    • Center for Regenerative and Developmental Biology, Forsyth Institute, and Developmental Biology Department, Harvard School of Dental Medicine, 140 The Fenway, Boston, MA 02115
    Search for more papers by this author


Left–right patterning is a fascinating problem of morphogenesis, linking evolutionary and cellular signaling mechanisms across many levels of organization. In the past 15 years, enormous progress has been made in elucidating the molecular details of this process in embryos of several model species. While many outside the field seem to believe that the fundamental aspects of this pathway are now solved, workers on asymmetry are faced with considerable uncertainties over the details of specific mechanisms, a lack of conceptual unity of mechanisms across phyla, and important questions that are not being pursued in any of the popular model systems. Here, we suggest that data from clinical syndromes, cryptic asymmetries, and bilateral gynandromorphs, while not figuring prominently in the mainstream work on LR asymmetry, point to crucial and fundamental gaps of knowledge about asymmetry. We identify 12 big questions that provide exciting opportunities for fundamental new advances in this field. Developmental Dynamics 237:3453–3463, 2008. © 2008 Wiley-Liss, Inc.

“The great field for new discoveries … is always the unclassified residuum. Round about the accredited and orderly facts of every science there ever floats a sort of dust-cloud of exceptional observations … Any one will renovate his science who will steadily look after the irregular phenomena.”

— William James, 1890


Left–right (LR) asymmetry is an important and puzzling feature of embryogenesis (Burdine and Schier,2000; Levin,2005), also having important implications for human health and development (Cohen,2001; Ramsdell,2005). There now seems to be a perception outside the field that the question of LR patterning is largely solved. Cell biology textbooks and reviews now often summarize their view of the situation with one sentence, such as “… and ciliary movement also initiates Left-Right patterning”. This is unfortunate, because it leads young developmental biologists (and reviewers of grant proposals) to think that there is little interesting science left in this field. However, those working on LR patterning know that there is considerable uncertainty about the early mechanisms of symmetry breaking, the evolutionary origins and degree of conservation of LR mechanisms at all levels, and the clinical implications of what we know and do not know about asymmetry (Tabin,2005; Levin,2006; Raya and Belmonte,2006; Speder et al.,2007).

The study of LR patterning has largely been focused on the asymmetric gene cascades (Levin,1998; Whitman and Mercola,2001), understanding the dynamics of node cilia using technologically intricate approaches (Cartwright et al.,2004; Buceta et al.,2005), and modeling the properties of morphogen gradients of Nodal (Nakamura et al.,2006). However, a constellation of classic and recent observations suggests that, contrary to popular opinion, we are still missing major components of this field, and, in many ways, have been looking “where the light is good”.

Here, we briefly sketch 12 major open questions in an effort to catalyze work on these challenging and novel directions. We chose these specific issues to be distinct in nature from the many details that remain to be fleshed out regarding the known pathways. The major areas of ignorance are revealed because they are not currently being addressed, have no established powerful model for their investigation, and/or involve a complex phenomenon not centered on a specific gene product. While we highlight some of the most difficult issues facing the field, these are also the areas of the greatest opportunity for discovery of new biology in this field.

(1) What is the molecular mechanism of “concordance” that determines whether a population is 50% situs inversus and 50% situs solitus vs. heterotaxic (random and independent sidedness of each organ)? Perturbations, like the spontaneous mouse mutants iv and lgl (Layton,1976; McNeish et al.,1990; Singh et al.,1991), and patients with immotile cilia syndrome (Afzelius,1976), exhibit 50% situs inversus and 50% situs solitus, while almost all of the known manipulations of early pathways (and mammalian mutations of Zic3, Lefty, cryptic/CFC, and ActRIIb) result in heterotaxia. None of the existing models (ciliary or intracellular) specifically address this; nor do they provide an explanation for how heterotaxia occurs. Thus, a more careful characterization of known mutations as to their degree of concordance (by scoring additional organ readouts and paying attention to subtle organ malformations as well as overall reversals), and the development of models that make predictions about the timing and nature of independent vs. linked situs decisions, will be essential.

(2) What is the base state of the LR axis? Complete inversions are induced by partial deletion of inv (Mochizuki et al.,1998), overexpression of mature Vg1 (Hyatt and Yost,1998), and inhibition of serotonin receptor 3 (albeit with a lower penetrance; Fukumoto et al.,2005). The inv mouse is especially curious, because its mutation results in the majority of its offspring exhibiting situs inversus (although there is a detectable level of heterotaxia; Morgan et al.,1998; Otto et al.,2003). Unfortunately, the true genetic deletion has not been analyzed and all published experiments appear to have been carried out with the spontaneous mutant, in which a small part of the protein still exists (and hence may not be a true null). Thus, it remains to be determined whether this fragment is enough to bias asymmetry, refuting the bizarre possibility that the default state (without INV activity) is biased asymmetry, but in the reverse orientation, and that the INV gene product is needed to simply flip everything in the complete opposite direction. The inv mutation appears to affect a very early step, because a cross between an inv mouse and a Foxj1 knockout mouse (knockout Foxj1 mice are 50% situs inversus and 50% situs solitus; Chen et al.,1998; Brody et al.,2000; Zhang et al.,2004) produce mice that more closely resemble the Foxj1 phenotype (Tamakoshi et al.,2006). An understanding of the definitive inv loss-of-function should be very informative.

Recent studies show that retinoic acid signaling isolates somites from asymmetric cues (Vermot et al.,2005; Vermot and Pourquie,2005). Symmetrical structures may inadvertently be susceptible to the effects of LR signaling molecules, because many developmentally important factors are used in patterning programs that overlap in space and time. The advantages of symmetry, however, have selected for the evolution of processes that shield the developing symmetric structures from asymmetric cues (Vermot et al.,2005; Vermot and Pourquie,2005). Along with the orientation, it should be determined whether the base state of symmetrical structures is symmetry or asymmetry (Jefferies et al.,1996; Palmer,2004).

(3) Are the breaking and orientation steps truly distinct and sequential? It is commonly thought that to pattern the LR axis, symmetry must first be broken and then oriented with respect to the other two axes (Brown and Wolpert,1990). However, sometimes, perturbation of the same molecule in different locations can lead to predictable, differential laterality phenotypes. Ventral injection of chimeric BVg1 into the R3 cells causes almost complete situs inversus, while injection into the R1 cell causes heterotaxia, and injections into the left side do not cause any significant laterality defects (Hyatt and Yost,1998). These data may indicate a close relationship between the mechanisms for the breaking of symmetry and the biasing of asymmetry, and could serve as an entry point into this process. Precedents for such a bistable switch include the Notch/Delta system (Raya et al.,2004). Regardless of whether the breaking of symmetry and biasing of asymmetry is one step or two steps, the iv mutant mouse, which exhibits random 50% situs inversus and 50% situs solitus, indicates that the next processes of amplification and restriction of the initial asymmetric signal may still be able to receive cues from, or act downstream of the symmetry breaking point, and that the orientation mechanism may not be necessary to bridge the symmetry breaking point with concordant downstream organ morphogenesis. This makes sense evolutionarily; the direction of asymmetry is not as important as concordance between the organs, because situs inversus individuals exhibit few clinical symptoms, but heterotaxic individuals can often have severe medical problems (Burn,1991; Peeters and Devriendt,2006).

A related aspect is the precise characterization of phenotypes reported for the various genetic and pharmacological perturbations. To arrive at a complete model for LR patterning mechanisms, detailed characterization of the phenotypes caused by each experimental system is essential. Indeed, both iv and inv mouse phenotypes are not purely 50%/50% and fully inverted respectively, but include a significant degree of heterotaxias, isomerisms, and cardiac malformations (Hummel and Chapman,1959; Morgan et al.,1998; Morishima et al.,1998; McQuinn et al.,2001). Likewise, situs inversus individuals are not really normal with respect to cardiac function because of discordances of orientation of structures at different scales (Delhaas et al.,2004; Ramsdell,2005). Thus, the distinctions between symmetry breaking, orientation, and concordance/morphogenesis may be more artifacts of our conceptual analysis of LR patterning phenotypes than a reflection of truly separate mechanisms during embryogenesis.

(4) Probably related is a set of puzzling inconsistencies about the universally conserved Left determinantNodal, other asymmetric genes, and downstream readouts. If Nodal is indeed a Left determinant, why does bilateral Nodal not lead to left isomerism, but to randomization (Levin et al.,1997; Sampath et al.,1997)? What kind of mechanism could explain cases where the upstream markers exhibit a higher level of incorrectly sided expression then the downstream targets they induce in the LR cascade (Tsukui et al.,1999; Kitaguchi et al.,2000; Kelly et al.,2002; Levin et al.,2002; Morokuma et al.,2002), which may even become completely uncoupled from expression of Nodal (Cota et al.,2006)? Because the iv mouse exhibits 50% situs inversus and 50% situs solitus, we would expect nodal expression in the iv mouse population to be 50% left-sided, and 50% right-sided. Instead, Nodal in iv mice populations is expressed on the left-side, right-side, bilateral or absent, at statistically equal probabilities (Lowe et al.,1996). This finding indicates that, in iv mice, bilateral or absent Nodal does not prevent organ concordance. Yet, when left-sided Nodal is experimentally removed, heterotaxia does result (Brennan et al.,2002)! In zebrafish, over 30% of cup mutant embryos show complete situs inversus, but only a small percentage of them show right-sided spaw expression (Schottenfeld et al.,2007), even though cup is a homolog of Pkd2 which is thought to be upstream of Nodal signaling. Another related puzzle: why do as many as 20% of siblings from a batch of embryos with 99% normal organ situs exhibit no expression of Nodal whatsoever (Levin and Mercola,1998a; Takano et al.,2007)?

(5) How conserved are the early, intermediate, and late mechanisms in LR patterning? The various intracellular, biophysical, and ciliary mechanisms have each only been probed in a small number of representative model systems (Palmer,2004; Levin,2006; Levin and Palmer,2007). While cilia appear to be involved in mouse (Nonaka et al.,1998; McGrath et al.,2003), zebrafish (Otto et al.,2003; Amack and Yost,2004), and Xenopus (Schweickert et al.,2007), numerous other mechanisms such as cytoskeletal polarity (Thitamadee et al.,2002; Qiu et al.,2005), serotonergic signaling (Fukumoto et al.,2005), gap junctions (Levin and Mercola,1998a,1999; Chuang et al.,2007), and ion flows (Adams et al.,2006) are conserved in LR patterning among other phyla, including Arabidopsis, Drosophila, Caenorhabditis elegans, and ciliates (Levin and Palmer,2007; Oviedo and Levin,2007). The recent demonstration that individual cells can align the LR axis from intracellular polarity cues (Xu et al.,2007) suggests that subcellular and ancient chiral components of all cells may underlie an origin of LR patterning existing before the evolution of multicellularity. This powerful idea was presciently suggested in 1977, before the identification of any of the molecular aspects of asymmetry: “one may begin to look for a relationship between the bilateral symmetry of organisms and the mirror symmetry between daughter cells at an early state of the embryo” (Albrecht-Buehler,1977).

The question of which, if any, of the early physiological mechanisms are used in mouse asymmetry is still open (see Levin and Palmer,2007 for a discussion of alternative interpretations of the genetic data vis-a-vis the proposal of cilia as originators of asymmetry). Because of the ubiquitous compensation and redundancy that exists among ion transport machinery, a truly satisfactory approach to answering this would require knock-in of dominant negative and constitutively active transporter mutants that will change cell membrane potential in prestreak embryos in well-defined ways, rather than knockout of individual channel/pump genes. Related to this is the need to re-examine roles of “ciliary” proteins in mammals, as they are likely to have other functions in addition to ciliary roles, and hence, knockouts do not cleanly implicate a specific function of ciliary proteins in LR patterning. For example, LRD is expressed in a random manner in head folds (Supp et al.,1997), and, intriguingly, has been shown to be important for biased chromatid segregation (Armakolas and Klar,2007).

(6) What explains the relatively low “penetrance” of some laterality-perturbing reagents and mutations? In a population where the background of defects is ∼1%, a 25% incidence of heterotaxia in 1,000 embryos is unquestionably significant (Levin and Mercola,1998a; Levin et al.,2002; Bunney et al.,2003). But, are the remaining embryos insensitive to the reagent? Are they somehow regulating to normalize asymmetry despite the effect of the reagent? Or are they truly part of an “affected” group of which 75% are developmentally determined to be situs solitus? With respect to pharmacological inhibition of ion channels and transporters, a possible reason for this may relate to the existence of robust physiological homeostasis mechanisms that can compensate for the loss of channel function, or even that the precise repertoire of functioning transporters expressed in a given individual may be a complex function of heredity, diet, and culture conditions that remain to be worked out. For example (unpublished observations in our lab), a pharmacological reagent from the same aliquot, on the same day, when applied to embryos of two different mothers, can cause heterotaxia rates of 3% to one group, and 55% in the other! This type of result reveals the complexity underlying the simple question of whether a particular pathway is or is not “involved in LR patterning.”

One possible explanation for the variability in phenotypes may involve the early embryonic fate-map. Some organisms (e.g., C. elegans) have a very mosaic and stereotypical development; others (e.g., Xenopus) use a basic scheme in which blastomeres give rise to which tissues, but the exact cell boundaries vary considerably among individuals of normal egg batches. This is clearly seen in the original data sets of fate mapping studies (Dale and Slack,1987; Moody and Kline,1990; Chalmers and Slack,2000). One prediction of this model is that a meta-analysis of LR perturbations among phyla should show more variability in phenotypic outcome in model species whose development is less stereotypical. If true, this would also suggest the existence of endogenous robustness mechanisms compensating for high levels of plasticity in embryonic cell fate (Palmer,2004). Importantly, an understanding of these robustness mechanisms will also improve our ability to investigate LR patterning, because it is still not known why laboratory culture of rodent and flatfish embryos automatically results in a significant background of LR defects (Fujinaga et al.,1990; Fujinaga and Baden,1991; Schreiber,2006). The understanding of stability against, and susceptibility to, environmental factors in LR patterning will be a key component of a complete appreciation of the stochasticity underlying consistent asymmetry.

(7) What is the nature of “subtle” asymmetries? The field is almost entirely focused on pathways determining situs of heart and visceral organs, with some attention now being paid to laterality of the brain (Toga and Thompson,2003; Halpern et al.,2005; Sun and Walsh,2006; Hendricks and Jesuthasan,2007). Central nervous system asymmetry may be patterned differently from that of body asymmetry, along pathways that may either be unrelated but parallel or, more likely, divergent from a common early step: although humans with situs inversus exhibit reversal of some anatomical brain asymmetries, they do not show reversals in functional asymmetries, and establish language dominance on the left cerebral hemisphere and a strong bias for right-handedness, just like in the general population (Kennedy et al.,1999; Tanaka et al.,1999; McManus et al.,2004). The zebrafish model system has not provided an ideal model for this discordance, because changes in canonical visceral situs pathways do seem to randomize fish brain asymmetry (Bisgrove et al.,2000; Concha et al.,2000; Barth et al.,2005; Essner et al.,2005), although point mutations of cyclops (the zebrafish Nodal homolog) result in normal heart looping and visceral situs but perhaps anatomical randomization in the brain because asymmetric gene expression is altered (Saude et al.,2005). However, the clinical data indicate that numerous other body structures and processes pay attention to LR cues, in pathways not predicted by or included in any of the current models (ciliary or intracellular).

There are several human syndromes that affect symmetrical or paired structures like the limbs (Smith et al.,1979), face or hips (Paulozzi and Lary,1999), that consistently present more strongly on one side than the other. Facial clefting presents a consistent asymmetry (Delaney and Boyd,2007), and this condition has recently become experimentally tractable, with the identification of molecular markers of asymmetry in symmetrical embryonic structures (somites; Golding et al.,2004a,b), and the discovery of a genetic entry point into zebrafish craniofacial asymmetry (Albertson and Yelick,2005). Hemihyperplasia (Fraumeni et al.,1967; Clericuzio,1993; Leung et al.,2002), a rare phenomenon where one side of the body to begins to grow in adulthood, is right-biased (Hoyme et al.,1998), and not only suggests that symmetric structures (knee joints, long bones of the legs) have LR identity, but that, astonishingly, they are able to retain that information for decades after embryogenesis.

Monozygotic twins are discordant for hemihypertrophy (West et al.,2003), and they display opposite-sided hair whorl direction, unilateral tooth defects, and other “bookending” phenomena (reviewed in Levin,1999), strengthening the link between very early chiral decisions and subtle asymmetries in man. Is this related to the mirroring (Fig. 1) of cytoskeletal structures and migration paths observed when cells split (Albrecht-Buehler,1977)? What is the molecular nature of the asymmetrical information that exists at early cleavages of vertebrate embryos and that results in asymmetry defects in one of the twins when they are separated early? Why does this separation result in opposite sidedness of subtle characteristics such as hair whorls in man (Golbin et al.,1993Aa, b; Morison et al.,1994) but also affect visceral and cardiac situs in some vertebrates (Mangold,1921; Takano et al.,2007)? An entry point into the mechanisms of these subtle asymmetries may be provided by the study of hair whorls on the face of cattle produced by egg splitting (in the context of on-going in vitro production of valuable animals; Ozil et al.,1982; Gatica et al.,1984; Lanier et al.,2001; Meola et al.,2004). Does the fact that these subtle asymmetries in human monozygotic twins exist almost entirely above the neck suggest a separate LR-organizer for the head and brain, distinct from that for the body?

Figure 1.

Mirroring among single cells. A–C: Actin cytoskeletal elements of daughter cells in culture showing a mirroring of the angle of filament orientation. D,E: Sample migration tracks of cells in culture showing mirroring of cell paths after splitting. These data indicate the presence of chirality in single cells that is preserved at splitting, possibly providing a model for understanding enantiomer (bookending) of symmetry properties in monozygotic twins derived from early splitting of the egg. Images are used with permission of Albrecht Guenther-Buehler (Albrecht-Buehler,1977).

The link to sex determination pathways (which have no obvious asymmetry) is fascinating, revealed by the consistently sided presence of testes vs. ovaries in hermaphrodites (van Niekerk and Retief,1981; Krob et al.,1994; Mittwoch,2000,2001,2008). There is a cryptic asymmetry aspect to sex determination; for example, the consistent asymmetry in foot size is opposite between women and men (Levy and Levy,1978). A model system may be available, because, in the Zic3 mutant, all affected males are heterotaxic, while females are wild-type or situs inversus (Gebbia et al.,1997). Finally, there are associations between asymmetry and the immune response (Renoux et al.,1983; Fride et al.,1990; Betancur et al.,1991; Neveu,2002; Fu et al.,2003; Meador et al.,2004; Quaranta et al.,2004; Shen et al.,2005), and cancer (McManus,1992; Sandson et al.,1992). The origin and implications of these links remain completely mysterious.

(8) “What is randomization”? This is a term often used in the field but attempting to construct a logical, mechanistic model of randomization quickly reveals our lack of conceptual clarity. For example, the normal chick embryo has a depolarization of cells on the left side of the streak followed by expression of Sonic hedgehog on the left side of the node. The simple way of modeling this would be to hypothesize that depolarization induces Shh expression. However, experimental induction of bilateral depolarization (Levin et al.,2002) does not result in bilateral Shh, but instead gives a spectrum of left-sided, right-sided, bilateral, and absent Shh expression. A mechanistic understanding of randomization would not only have to explain this bizarre result (which also occurs for other LR components), but would also have to show how cells are synchronized (so that the L and R domains of Hensen's node act as a unit, and do not exhibit a salt-and-pepper pattern on each side, reflecting cells' individual randomization). It would also need to predict the nonequiprobable outcomes (right-sided Shh is less frequent than bilateral Shh for example).

(9) What happens upstream of asymmetric gene expression in the chick? Before the right-sided expression of Activin Receptor IIa in the early chick streak (Levin et al.,1995; Stern et al.,1995), there are no nodal cilia, and the subsequent serotonin localization does not match the asymmetric pattern described in Xenopus (Fukumoto et al.,2005). Yet, many of the same molecular components are involved—serotonin and its receptors, gap junctions, and the H,K-ATPase/V-ATPase system (Levin,2005). How are the same molecular signaling elements implemented in a body plan with a completely different large-scale architecture and cell size? Is the differential timing of similar physiological mechanisms among Xenopus and chick an instance of evolutionary heterochrony? Might the temporal sliding of molecular LR components relative to other embryonic events explain the evolutionary differences and similarities observed in asymmetry across phyla? Finally, what is the role of molecular chirality in amniotes, given that application of nonbiological stereoisomers of simple molecules induces heterotaxia in chick (Gray et al.,1942)?

(10) Is there a LR coordinator or organizer? In many types of embryos, LR patterning first takes place in a sheet of cells; it has been suggested that the information originates locally—in a kind of organizer or LR coordinator that is a single cell or small group of cells. This organizer is suggested to be in early blastomeres in Xenopus (Hyatt and Yost,1998), in the node in mice (Pagan-Westphal and Tabin,1998) and in the base of the primitive streak in chick (Levin and Mercola,1998b). One proposal has been that LR patterning is a kind of planar cell polarity (Aw et al.,2008), as occurs in Drosophila wings and hair cells of the inner ear in mammals (Axelrod and McNeill,2002; Zallen,2007), and is a mechanism by which an organizer imposes a coherent LR directionality on the rest of the blastoderm. However, the ability of individual mammalian cells to set up a consistent LR polarity within a single cell (Xu et al.,2007), and the long-known ability of ciliates to epigenetically establish handedness (Nelsen et al.,1989) suggest that perhaps cells can independently orient their intracellular directionality along the LR axis. It will be crucial to determine whether LR information (including direction of the L and R sides, and position with respect to the midline) is imposed from without or generated internally in embryonic cells, or maybe even an interplay of both.

(11) Why can LR phenotypes be dissociated from essential housekeeping functions for numerous pathways? Several pathways that have been implicated in asymmetry also have important and profound roles in basic cell physiology. These include the V-ATPase H+ pump (Adams et al.,2006), which acidifies vacuoles, gap junctions (Levin and Mercola,1998a; Chuang et al.,2007), which regulate numerous aspects of small metabolites between cells, tight junctions (Brizuela et al.,2001; Eckert et al.,2005; Vanhoven et al.,2006), which provide epithelial integrity, K+ channels (Levin et al.,2002), that regulate transmembrane potential, microtubules, and actin filaments (Thitamadee et al.,2002; Abe et al.,2004; Qiu et al.,2005; Adams et al.,2006; Aw et al.,2008), which are required for cell division and intracellular transport, and Ca++ signaling (McGrath et al.,2003; Raya et al.,2004), which regulates numerous second-messenger pathways. Despite the pleiotropic nature of these mechanisms, it has proven possible to analyze clean LR asymmetry phenotypes after loss-of-function perturbations of key members of these pathways. A priori, this is highly unexpected; blocking all K+ channels with barium chloride or disrupting actin organization with Latrunculin would be expected to result in massive defects or toxicity, not the observed heterotaxia with normal dorsoanterior development. While this is not always possible, in the above cases a moderate level of abrogation could be identified and achieved, at which normal development and cell health was maintained, but the LR pathway was specifically disrupted. What signaling mechanisms or network properties in development allow LR patterning to be intimately dependent upon, and yet separable from, effectors of fundamental cell functions?

(12) When is the midline really determined? The left–right axis is defined with respect to the midline—a reflection line across which overall symmetry is preserved. There are many open questions about the mechanisms of midline determination, and this is clearly central to the question of LR patterning. In Xenopus, while the plane of first cleavage can be experimentally repositioned (Black and Vincent,1988; Danilchik and Black,1988), in normal embryos, the cleavage furrow usually corresponds to the future midline of the embryo (Klein,1987; Masho,1990). This makes it a natural candidate for LR mechanisms that rely on chirality (Danilchik et al.,2006) or redistribution of maternal intracellular components to the prospective L and R sides (Esser et al.,2006; Aw et al.,2008; Morokuma et al.,2008) during early cleavages. In contrast, the midline is thought to be established quite late in amniotes, when thousands of small cells exist and none have the advantage of spanning the midline of the whole embryo. If the midline (and thus position with respect to the LR axis) is not determined until cells are too small to enable transport to the whole embryo's L or R side, then initiation models based on multicellular signaling must be sought. Thus, knowing when the midline is truly set up is crucial for being able to gauge the plausibility of intracellular transport and planar cell polarity models in embryos of zebrafish, chick, and mouse.

The remarkable phenomenon of gynandromorphy throws doubt onto the certainty of late establishment of the midline in a wide range of phyla. In orthoptera, bilateral gynandromorphs are thought to result when one of the X chromosomes in an XX zygote is eliminated at the first cleavage division (Barranco et al.,1995), resulting in an animal that is part male and part female. The human cases revealed by X-linked genetic pigmentation syndromes (Happle et al.,1995) resemble examples of gynandromorphs found throughout phyla (Fig. 2), including butterflies, ants, crabs, crustaceans, and chickens (Farmer,1972; Dang and Peterson,1979; Homsher and Yunker,1981; Sivaradjam and Bierne,1981; Mey,1982; Taber and Francke,1986; Taylor,1986; Micheli,1991; Stevens and Munk,1991; Barranco et al.,1995; Sagi et al.,1996,2002; Moriyasu et al.,1998; Zou and Fingerman,2000). What is truly remarkable is that the male–female division takes place precisely at the LR midline of the animal.

Figure 2.

Bilateral gynandromorphy. A:Papilio glaucus, used with permission from James K. Adams, showing boundary between male and female cells exactly down the midline. B: A bilateral gynandromorph lobster, used with permission of the Bangor Daily News; photo by Abigail Curtis. C: A CHILD syndrome patient with harlequin pigmentation, used with permission of John Wiley and Sons from the American Journal of Medical Genetics, 2000;90:340. D: A gynandromophic blue crab, Callinectes sapidus (ventral view; left side of the body exhibiting the narrower abdomen typical of males), with permission from Rom Lipcius (VIMS); taken from Kleps et al., (2007). E: A section of the brain of a gynandromorphic finch processed for in situ hybridization with probes to sex-specific transcripts (dark signal = female chromosome, no signal = male chromosome), where expression of a molecular marker in the brain of a gynandromorphic finch reveals that the separation between male and female cells is exactly down the midline. Image taken from Fig. 6 of (Agate et al.,2003), copyright held by National Academy of Sciences of the United States.

This is most strikingly seen in bird embryos, where the midline was always thought to be determined in a blastodisc of many thousands of cells. When failure of polar body extrusion occurs during meiosis (producing a binucleate ovum with Z and W nuclei), the resulting birds are chimeras of male and female cells (Lillie,1931). Amazingly, the separation (revealed by the sex-specific pigmentation pattern) is exactly down the midline, and in roosters, is respected even by the comb on top of the head of the male half. One recent example is shown in Figure 2E, where expression analysis of a molecular marker in the brain of a gynandromorphic finch reveals that the separation between male and female cells is exactly down the midline (Agate et al.,2003). A late definition of the midline would predict a random, mosaic distribution of male/female regions with respect to the axes of the adult. In contrast, the sharp midline separation suggests that the first cell cleavage may give rise to the L and R halves of the animal that can then inherit differential chiral information from the egg.

The main axes of rabbit embryos are known to be established before the appearance of the streak (Viebahn et al.,1995), and a similar situation has been proposed in the case of mouse development (Gardner,2001; Piotrowska and Zernicka-Goetz,2001; Plusa et al.,2002). Indeed, bilateral symmetry is said to be discernible in untreated living eggs and early embryos of the rat (Jones-Seaton,1950; cited in Gardner,1996). In human embryogenesis, the sharply unilateral pigmentation patterns that occur in humans with X-linked diseases such as CHILD syndrome requires unilateral X-inactivation (or the asymmetric activation or accumulation of some factor responsible for subsequent X-inactivation) to occur at two- or four-cell stage in such patients. This also suggests that the midplane is defined quite early in human embryos—a possibility consistent with the “bookending” phenomenon observed in monozygotic twinning (Levin,1999). More research is clearly needed, to determine whether gynandromorphy is always bilateral, and to understand the mechanisms of midline determination in different amniote embryos. Mice also exhibit the strong sidedness of hermaphroditic organs observed in human cases, (Eicher and Washburn,1983; Ward et al.,1987; Biddle et al.,1994), although the consistent laterality of testes' versus ovaries' placement is opposite of that observed in humans (van Niekerk and Retief,1981; Krob et al.,1994).

If the chick embryo indeed sets up its midline at first cleavage, some of the puzzles regarding early mechanisms in the chick may need to be reformulated (part 9); the induction of situs inversus by division of the preincubation chick blastoderm (Lepori,1967) is consistent with this model. However, even if set up very early, the midline can be experimentally re-specified at later stages in mice and birds (Mintz,1971; Eyal-Giladi and Fabian,1980). A detailed study of the LR asymmetry phenotype of such animals has not been performed, and it remains to be determined exactly how heavily LR initiation relies on alignment between blastomere cleavage and the prospective midline in various model species. Indeed, even though Xenopus normally sets midline and asymmetry by the second cleavage, frog embryos can initiate correct LR patterning in organizers induced during blastula stages (Nascone and Mercola,1997). Thus, while it is possible that some embryos establish a midline far earlier than previously thought, the models relying on intracellular transport across the prospective midline may need to be extended, to explain how LR patterning still occurs correctly when a new primary axis is initiated in fields of many small cells.


Left–right patterning is a truly fascinating aspect of developmental biology. Its implications extend to behavior (Klar,1999; Crow,2002; Bisazza and de Santi,2003), immunology (Wise et al.,1993), and human culture (McManus,2002). While tremendous progress has been made over the past 15 years, it is clear that major questions and fundamental areas of ignorance remain (especially with respect to subtle asymmetries and the timing of midline definition). There will be important and far-reaching discoveries to be made here for decades to come, and we hope that young scientists from physics, engineering, and cell biology enter the field, bringing fresh approaches to the contradictory data and brittle models existing today.

At the turn of the past century, Lord Kelvin identified “two small dark clouds on the horizon of classical physics”—at the time, minor observations that did not fit neatly into the powerful, unified paradigm of classical physics. Investigations of these anomalies gave rise to quantum theory and relativity. We hope that these open questions indeed contain several clouds of this type, leading to new discoveries, better questions, and perhaps ultimately a profound unification across scales of organization from quantum parity violations (Kondepudi,1987; Mason,1991) to large-scale patterning of a whole embryonic axis.


We gratefully acknowledge the many helpful discussions with Richard Palmer, Chris McManus, other members of the Levin lab, and the LR asymmetry community. M.L. was supported by NIH grants and American Heart Association Established Investigator grant. S.A. was supported by the A*STAR overseas National Science Scholarship from the Agency for Science, Technology, and Research (Singapore). This article was written in a Forsyth Institute facility renovated with support from Research Facilities Improvement Grant from the National Center for Research Resources, National Institutes of Health.