Baixe o app para aproveitar ainda mais
Prévia do material em texto
Developmental mechanism and evolutionary origin of vertebrate left/right asymmetries Jonathan Cooke* Department of Zoology and Museum of Comparative Zoology, University of Cambridge, Downing Street, Cambridge, UK (E-mail : jonathan.cooke5@btinternet.com) (Received 18 October 2002; revised 27 May 2003; accepted 28 May 2003) ABSTRACT The systematically ‘handed’, or directionally asymmetrical way in which the major viscera are packed within the vertebrate body is known as situs. Other less obvious vertebrate lateralisations concern cognitive neural function, and include the human phenomena of hand-use preference and language-associated cognitive partitioning. An overview, rather than an exhaustive scholarly review, is given of recent advances in molecular understanding of the mechanism that ensures normal development of ‘correct ’ situs. While the asymmetry itself and its left/right direction are clearly vertebrate-conserved characters, data available from various embryo types are compared in order to assess the likelihood that the developmental mechanism is evolutionarily conserved in its entirety. A conserved post-gastrular ‘phylotypic ’ stage, with left- and right-specific cascades of key, orthologous gene expressions, clearly exists. It now seems probable that earlier steps, in which symmetry-breaking information is reliably transduced to trigger these cascades on the correct sides, are also conserved at depth although it remains unclear exactly how these steps operate. Earlier data indicated that the initiation of symmetry-breaking had been transformed, among the different vertebrate classes, as drastically as has the anatomy of pre-gastrular develop- ment itself, but it now seems more likely that this apparent diversity is deceptive. Ideas concerning the functional advantages to the vertebrate lifestyle of a systematically asymmetrical visceral packing arrangement, while untestable, are accepted because they form a plausible adaptationist ‘ just-so ’ story. Nevertheless, two contrasting beliefs are possible about the evolutionary origins of situs. Major recent advances in analysis of its developmental mechanism are largely due not to zoologists, comparative anatomists or evolution- ary systematists, but to molecular geneticists, and these workers have generally assumed that the asymmetry is an evolutionary novelty imposed on a true bilateral symmetry, at or close to the origin of the vertebrate clade. A major purpose of this review is to advocate an alternative view, on the grounds of comparative anatomy and molecular systematics together with the comparative study of expressions of orthologous genes in different forms. This view is that situs represents a co-optation of a pre-existing, evolutionarily ancient non-bilaterality of the adult form in a vertebrate ancestor. Viewed this way, vertebrate or chordate origins are best understood as the novel imposition of an adaptively bilateral locomotory-skeletal-neural system, around a retained non-symmetrical ‘visceral ’ animal. One component of neuro-anatomical asymmetry, the habenular/parapineal one that originates in the dien- cephalon, has recently been found (in teleosts) to be initiated from the same ‘phylotypic ’ gene cascade that controls situs development. But the function of this particular diencephalic asymmetry is currently unclear. Other left-right partitionings of brain function, including the much more recently evolved, cerebral cortically located one associated with human language and hand-use, may be controlled entirely separately from situs even though their directionality has a particular relation to it in a majority of individuals. Finally, possible relationships are discussed between the vertebrate directional asymmetries and those that occur sporadically among protostome bilaterian forms. These may have very different evolutionary and mol- ecular bases, such that there may have been constraints, in protostome evolution, upon any exploitation of left and right for complex organismic, and particularly cognitive neural function. Key words : handedness, left-right asymmetry, evolution, development, phylogeny. * Address for correspondence : 10, Danvers Road, London N8 7HH, UK. Biol. Rev. (2004), 79, pp. 377–407. f Cambridge Philosophical Society 377 DOI: 10.1017/S1464793103006298 Printed in the United Kingdom CONTENTS I. Introduction: Scope and level of this review ........................................................................................... 378 II. Overview of the developmental mechanism for visceral situs ................................................................ 379 (1) A vertebrate-conserved ‘phylotypic ’ post-gastrular sector .............................................................. 379 (2) Initial left-right symmetry breaking in earlier development : apparent diversification in timing and in upstream cascade steps ................................................................................................ 382 (3) Is there a unitary mechanism for developmental origin of vertebrate situs? ................................ 387 (4) Changing left-right co-options of gene orthologues during vertebrate diversification, and other puzzles .......................................................................................................................................... 388 III. Viewpoints on the evolutionary origins of situs ....................................................................................... 389 (1) The functional ‘ just-so’ story: situs as a vertebrate (chordate) invention ...................................... 389 (2) A constrasting view: situs as the co-optation of a pre-existing axis ................................................ 391 IV. Visceral major organ, neuro-anatomical and neuro-functional left and right : unitary or independent origins? ................................................................................................................................... 395 (1) Hominoid hand-use/linguistic lateralisation: origin and inheritance ........................................... 395 (2) Hand-use/linguistic lateralisation and situs in development ........................................................... 397 (3) Status of other vertebrate functional forebrain lateralisations ........................................................ 398 V. Implications for directional asymmetries in protostome animals ......................................................... 399 VI. Conclusions .................................................................................................................................................. 400 VII. Acknowledgements ...................................................................................................................................... 402 VIII. References .................................................................................................................................................... 402 I. INTRODUCTION: SCOPE AND LEVEL OF THIS REVIEW Despite the nearly perfect bilateral symmetry of vertebrate axial anatomy, paired sense organs and limbs, it is widely recognised even by non-biologists that the major visceral organs are packed into the human body cavity in a highly asymmetrical manner, and that this is a ‘handed’ asym- metry in that it follows the same left-right direction (technically known as normal situs or situs solitus) in the vast majority of individuals. The rare human situs inversus however, the complete mirror-reversal of normal visceral anatomy, is a perfectly viable state of affairs (Torgerson, 1950), though the decidedly less rare heterotaxias, confusions of left-right anatomy within individuals, are usually patho- genic. It becomes clear, from comparative study of a range of vertebrate embryos, that while the developed viscera and particularly the heart and greatblood vessels have become more complex in function and anatomy, the directionality of the situs asymmetry is vertebrate-conserved. Such directional asymmetry is quite distinct from the phenomena of fluctuating asymmetry and antisymmetry. In the former, limitations of ‘buffering’ of the developmental mechanism against (genetically or environmentally con- ditioned) perturbation limit the precision with which right and left ‘ symmetrical ’ structures can be mirror replicas within individual bodies. In antisymmetry, competitive suppressive mechanisms ensure that right and left structures are very different in size etc. (as in certain arthropod appendage pairs), but with essentially random ‘handedness ’. Certain more subtle directional asymmetries within ver- tebrate structure are reliably co-ordinated with, and could thus be considered part of, situs. The preferred direction of axial torsion within the uterus or egg-shell, seen in most types of vertebrate embryo, is perhaps the clearest example, but others may be the antero-posterior staggering or alternation of the primitive left and right somite boundaries that is clearest in acraniates (Wada, Garcia-Fernandez & Holland, 1999) but detectable in vertebrates ( J. Cooke, unpublished observatons on chick and frog embryos), and left-right differences in both developmental and adult branchial anatomy of some agnathan vertebrates. Less universal but still highly significant is the preponder- ance (around 9:1) of right-hand-use preference for human skilled activites ; the most conventional use of the term ‘handedness ’. This is strongly (but not universally) linked with the phenomenon that most of us have the capacity for language use preferentially located in our left cerebral cortices, with other, harder-to-define but complementary cognitive functions located on the right (e.g. McManus, 1991). Thus it is only mildly incorrect to speak, in humans, of a particular ‘normal ’ relationship between visceral situs, hand-use preference and differential placement of higher cognitive capacities within the hemispheres, although in fact various ‘discordancies ’ among these are common and not strongly dysfunctional. Left-right functional complementarities of brain usage, with a preponderant directionality corresponding to the human one, have recently been found to characterise most of the hominoid (human-ape) clade, and are probably related to the onset of language-like cognitive abilities and associated differential hand use (Annett & Annett, 1991; Cantalupo & Hopkins, 2001). But in addition to this, there is a whole range of phenomena involving differential function- ing, sometimes linked to detectably different anatomy, of the two sides of the brain in fish, reptiles, birds and non- primate mammals (see Section IV and references). For most of these, however, it is unclear whether population levels of directionality are similar to those of human hand preference etc., or show the much higher reliability characterising situs (see later discussion). Well-known adaptive idiosyncracies, such as the asymmetrical aural apparatus of owls, are very 378 Jonathan Cooke probably extensions of, and thus capitalisations on, situs itself although this is not positively known. Molecular understanding of developmental mechanisms has recently made explosive advances for situs itself, and these mechanisms, including data from all the experimental ‘model ’ vertebrate embryo types, have been comprehen- sively reviewed for developmental specialists by several authors (cited in Section II.1). The aim of the present article is not to replicate or even update these reviews but instead to survey the field for a broader readership, and in relation to three key questions of evolutionary interest ; the degree of conservation or universality of the whole situs developmental cascade, the evolutionary origin of situs, and finally the degrees of developmental and thus evolutionary relationship between situs and the various neural lateralisations. There is clear, vertebrate-wide conservation of a post- gastrular, ‘phylotypic ’ left-right organisation of the embryo, with distinctive left-hand and right-hand expressions of a set of gene orthologues. These initiate the differentials in growth and morphogenetic movement that characterise visceral organ primordia on either side of the midline to produce situs anatomy. But dramatic apparent differences in the events that lead up to this stage, among vertebrate embryo types, have led to suggestions that the earlier sector of mechanism, that which breaks the embryo’s apparent bilateral symmetry in the first place, has varied opportun- istically as much as has pre-gastrulation anatomy during vertebrate evolution. While acknowledging that this may turn out to be the case, I nevertheless propose that current data are consistent with in-depth conservation of even this early, so-called symmetry-breaking sector of mechanism. Recently, experimental gene expression data in zebrafish (Danio rerio) have linked the origin of an anatomical brain asymmetry, the diencephalon-associated habenular and parapineal one that is probably of functional importance in fish, amphibians, reptiles and birds, with the phylotypic situs gene cascade itself (Concha et al., 2000; Liang et al., 2000). However, a wider variety of documented left-right par- titionings of vertebrate forebrain function are not necessarily all linked with this particular feature ; they remain to be fitted into what traditional developmental geneticists would have referred to as a ‘pedigree of causes ’ underlying vertebrate asymmetries. Current data seem to link together human hand-use preference and higher cognitive left-right partitioning, as being controlled probabilistically by the same system of left-right developmental information (see Section IV and references). But strikingly, the data also indicate that this particular neural lateralisation system is quite independent from that controlling situs (Torgerson, 1950; Kennedy et al., 1999; Tanaka et al., 1999). Evolutionarily speaking, ‘explanation’ for a consistently lateralised arrangement of vertebrate visceral packing is not hard to find. Given the material possibility, evolution can be expected to have simulated optimal engineering design, and the argument that such an arrangement is indeed optimal for the vertebrate will be laid out in Section III. But, with the interesting exception of at least some bird gastrulae (see Section III.2), extant vertebrates appear bilaterally sym- metrical in cellular structure as early embryos. It has thus been easy to make the assumption that the mechanism for breaking this symmetry with a reliable left-right direction- ality was a novel ‘ invention’, selected for by the distinctive lifestyle requirements of an organism at or relatively close to the origins of the vertebrate clade. The present article re-emphasises data from zoology, comparative anatomy and comparative gene expression studies that advocate a different evolutionary scenario. These data suggest that a vertebrate ancestor possessed an evolutionarily ancient, non-symmetrical organisation, but secondarily re-symmetrised its locomotory, outer body wall and, to a large extent, nervous system as an adaptation to a new lifestyle. The developmental substrate of the previous non-symmetry was however retained, for further elabor- ation of its increasingly complex viscera. Viewed in this way the vertebrate body plan is a secondary re-imposition of bilaterality, in the form of an almost symmetrical locomotor- skeletal outer body wall and paired brain structure, upon a radically non-bilateral adult body plan in the precursor animal (see references cited in Section III.2). According to this scenario, that non-bilaterality is evolutionarily deep, characterising a clade embracing most or all deuterostome animals. It was itselfa secondary departure from original bilaterality, the mode of axial organisation ancestral to all metazoans other than sponges, cnidarians and possibly ctenophores (the clade Bilateria : Finnerty & Martindale, 1998; Holland, 1999). Initial symmetry ‘breaking ’, in vertebrate left-right development, is then seen as the reliable re-evocation of the ancestral adult non-symmetry, within the symmetrical cellular anatomy of the contemporary blastula/ gastrula. On this view, in most vertebrate embryos, the relatively recently imposed locomotory/neural bilaterality has come to dominate the structure of the pre-gastrula so completely that the lack of mirror-symmetry inherent in biomolecular structure – usually referred to as molecular ‘chiral ’ information – must be recruited to initiate the ancestral gene cascades as what is now a left-right ‘axis ’. Tantalising indications do exist, however, of a left-right component at the outset of vertebrate development (see Section III.2 and references). II. OVERVIEW OF THE DEVELOPMENTAL MECHANISM FOR VISCERAL SITUS Fully detailed reviews of this expanding field are widely available (see e.g. Tamura, Yonei-Tamura & Izpisua- Belmonte, 1999; Burdine & Schier, 2000; Capdevila et al., 2000; Schneider & Brueckner, 2000; Wright, 2001; Yost, 2001; Hamada et al., 2002). Section II.1 inevitably involves some detailed consideration of anatomy and molecular developmental genetics ; readers interested in the wider issues might follow the remaining material without this section, especially by referring to Fig. 1. (1 ) A vertebrate-conserved ‘phylotypic’ post-gastrular sector It has become clear that activation of the transforming growth factor b (TGFb)-related intercellular signal gene Evolutionary origin of vertebrate left/right asymmetries 379 ‘Activin’ signal FGF8 BMP4 Shh Vg signal nodal signal lefty signal flow at node Pitx2 (prolonged, extensive ‘left encoding’) BMP4 signal SnR (’right encoding’ for heart loop & embryo torsion) Shh BMP4 BMP4 lateral SnR Pitx2 nodal Fig. 1. Vertebrate left and right gene activity cascades. A generic embryo midline is represented as a ‘primitive streak’ or elongated site of gastrulation as seen from above, with the axis being generated in an anterior (top of page) to posterior sequence from a regressing node. Time progression is also represented from the top to bottom of the page. Thus the later and more lateral (‘phy- lotypic ’ – see text) gene expressions actually occur in post-gastrular structures, anterior to the level of a node more regressed than that shown. The earlier, near-midline ones in the chick occur around the node, before the onset of regression or during its earliest stages. Heavy hatched arrows connecting bolder gene names represent positive control (activation) input ; light stippled connections leading towards obliquely slashed gene names indicate negative control input onto their expression. Where the target of negative 380 Jonathan Cooke nodal, in a broad domain in the left lateral mesoderm, is an essential conserved mechanism imbuing this region with left identity in all studied vertebrate embryos (Levin et al., 1995; Collignon, Varlet & Robertson, 1996; Lohr et al., 1998). Gastrulation proceeds as a sequence, in which mesoderm destined for successively more caudal regions of the body plan moves into place as a middle cell layer ahead of the closing blastopore (in amphibians) or regressing Hensen’s node (avian and other ‘blastoderm type’ and mammalian embryos). Left lateral nodal expression initially follows this wave-like sequence, beginning with material lying within or just posterior to the heart-forming region. The mRNA expression, presumably followed by peak levels of nodal signal, is dynamic and is down-regulated after relatively brief periods at successive antero-posterior levels, to disap- pear at a level near the regressing node after around 10 somites have segmented, for instance, in the chick. This nodal signalling in turn activates transcription of Pitx2, encoding a homeobox-containing transcription factor. Once activated, left Pitx2 expression is both more durable than that of nodal and spreads anteriorly and posteriorly beyond the nodal domain that triggered it. Pitx2, whose left-right role appears to be that of an executive control gene for Left developmental character and tissue contribution to success- ive organs, is ultimately expressed throughout the deriva- tives of the left contribution to the heart tube, with a sharp boundary maintained from the inflow (future atrial) region up into the outflow or ventricular region until advanced stages of looped heart morphogenesis. Over an extended period as the body plan is laid down, Pitx2 expression becomes activated at the origin of the left contribution to successive paired or asymmetrically developing visceral rudiments, such as those of lungs, stomach and other intestinal derivatives (Logan et al., 1998; Piedra et al., 1998; Ryan et al., 1998; St Amand et al., 1998; Yoshioka et al., 1998; Campione et al., 1999). The snail-related zinc-finger transcription factor gene, SnR in chick (Isaac, Sargent & Cooke, 1997), msna in mouse (Sefton, Sanchez & Nieto, 1998), becomes activated in right- lateral splanchnic mesoderm during gastrulation, in a domain that also appears to extend and then move back in a wave-like manner, but beginning slightly earlier than that for left-lateral nodal. Thus having begun in the anteriorly positioned, compressed cardiac territory, right lateral SnR expression is last detected in the posterior heart inflow region only, at around the 12 somite stage in chick. The orthologous snail-related gene is known from the frog (Xenopus laevis) and zebrafish (Danio rerio), though expression studies that would reveal a distinctive right-lateral expression component have not been published. It thus seems reason- able to suggest that right-lateral expression of this gene be considered as an additional conserved component of the gene cascade for vertebrate situs. It should be noted that nodal, Pitx2 and this snail-related gene each have other, conserved but bilateral expression domains in early vertebrate embryos. These expressions, some of which have attested developmental roles, are under separate control (see Isaac et al., 1997; Patel, Isaac & Cooke, 1999, for SnR). It appears that where transient RNA expression is appropriate, as for the wave of nodal signalling, this is achieved through an early self-activating component in the transcriptional control, followed by a damping negative control input from a co-activated or downstream-activated gene product. Thus a sequence motif in the control region of nodal suggests positive feedback (i.e. self-activation) from signalling through the nodal receptor pathway itself, while negative feedback occurs through the antagonistic activity of the closely related ligand (or anti-ligand) lefty-2/antivin on nodal signalling at the protein level (Adachi et al., 1999; Norris & Robertson, 1999; Saijoh et al., 2000). The lefty gene is expressed in a left-lateral domain that closely tracks and slightly follows the nodal one, and is probably activated directly by the nodal signal that it then goes on to quench. Prolonged and stable expression such as is required for the Pitx2 role, by contrast, is organised through initial upregulation by the upstream nodal signal, followed by self- maintaining and collateral positive maintenance inputs that include a binding site for the Nkx 2.5 transcription factor (Shiratori et al., 2001). Initial activation of lateral mesodermal SnR (chick) is bilateral, and could be considered a default state. Then with variable rapidity among individuals, it is downregulated on the left as nodal expression first appears there (Isaac et al., 1997). Experimental right-sided nodalprotein expression ablates normal SnR expression there, while antisense treatment targetting SnR expression ultimately leads to abnormal Pitx2 expression on the right (Patel et al., 1999), suggesting that SnR normally helps to suppress Pitx2 expression there. Thus, while it is clear that left nodal signalling positively activates Pitx2 transcription directly (Shiratori et al., 2001), a further, indirect double-negative control relationship may help to ensure spatial exclusivity between the key snail-related and Pitx transcription factor gene expressions normal to right and left sides. Pitx2 null mutant mice exhibit right cardiac isomerism, i.e. ‘ right-hand’ developmental character in heart precursor tissue from both sides of the midline, although the situation for more posterior viscera is less clear and may simply correspond to a lack of lateralising function, causing heterotaxia (Kitamura et al., 1999; Lin et al., 1999). Targetted ablation of msna in mice is not informative since the gene has other, prior essential developmental roles, but chick embryos after prolonged antisense disruption of SnR, having shown bilateral Pitx2 expression in heart precursors, then exhibit left-isomerism of heart-tube morphogenesis. In control is itself an intercellular signal, control can be at the protein function rather than gene activity level. Lateral Vg (left) and BMP4 (right) signals, placed at a horizontal dashed line representing the transition from early (near-midline) to ‘phylotypic ’ stages, are probably phylotypic, while the evidence that the right-lateral SnR (snail-related zinc-finger transcription factor) role is phylotypic remains incomplete. Evolutionary origin of vertebrate left/right asymmetries 381 the same experiments, antisense-mediated SnR disruption that is too brief to cause detectable right Pitx2 expression nevertheless randomises the direction of heart-tube looping. This looping direction, normally to the embryo’s right, is the first vertebrate-conserved, gross-anatomical manifestation of situs to develop. Loop reversal in these latter embryos is independent of the development of ‘ left ’ and ‘right ’ structural character in the tissue contributions to the tube inner wall, which can remain normal (Patel et al., 1999). Overall results of perturbing SnR function thus suggest a dual role in left-right development. A direct role ensures right cardiac looping and perhaps correct embryo torsion, but this is separable from the gene’s ‘cascade’ role in propagating right-hand character within tissue by ensuring repression of Pitx2. Pitx2, by contrast, appears to have a prolonged and widespread role, determining left-specificity in tissues of successively developing organs. A variety of data demonstrates that in addition to negative regulatory linkages between ‘opposing’ genes, barrier functions within the axial midline structures consti- tute a major mechanism maintaining proper laterality of expression domains normal to left and right during post- gastrular development (Meno et al., 1996; Bisgrove, Essner & Yost, 1999, 2000; Meno et al., 1999). The cellular structure of the midline appears on anatomical grounds alone to constitute a relative barrier to diffusing intercellular protein signals ; the notochord, a rod of tightly packed vacuolated cells with a tough matrix sheath, is tightly associated in vivo with the overlying floorplate, a midline strip of cells integrated into the neural plate but displaying at early stages several molecular affinities with notochord. There are also midline-associated gene expressions that appear to have a ‘barrier ’ role, notably including a member of the nodal-related lefty group (see above), whose proteins may act as an antagonist ‘ sink’ to nodal signalling by a mechanism equivalent to ‘dominant-negative’ interference with nodal ligand. A lefty is expressed in a strip at the left- hand edge of, or bilaterally in, midline structures of all vertebrates examined. Disruption or non-formation of an anatomically normal midline, whether as part of a mutant phenotype or due to microsurgical or other early embryo manipulations, predisposes embryos to heterotaxias with bilateral or absent expressions of nodal and Pitx2 (Danos & Yost, 1995; Collignon et al., 1996; Tsukui et al., 1999). Several authors have classified anomalies of left-right structure and downstream gene expressions in relation to the probable timing and axial position, within post-gastrular development, of midline barrier interruption (e.g. Bisgrove et al., 2000). Additionally, proper negative feedback onto the early, autocatalytic activation of lateral nodal expression appears necessary for its normal left restriction. Thus in a mouse lefty null mutant, heterotaxias are associated with abnormally extensive and bilaterally spreading nodal RNA expression (Meno et al., 1999; Hamada et al., 2002). As with developmental ‘master control genes ’ generally, it has thus far been hard to make progress identifying target genes whereby Pitx2 (left) and the snail-like orthologues (right), actually execute the modulations of morphogenesis and growth rate that constitute ‘ left ’ and ‘right ’ characters in relevant tissues. As regards heart loop morphogenesis, there is evidence that differential microfilament function and character of the extracellular matrix are involved (Itasaki et al., 1991; Tsuda et al., 1996). One complexity is that structures that come to be situated at right and left in the heart rudiment after its looping, were originally specified on the basis of relative antero-posterior ordering within the straight heart-tube. Thus gene expressions characterising particular parts of the formed heart, that might be considered as members of specifically right or left develop- mental gene cascades, might in fact owe their left- or right- positioning more directly to correct looping direction, for instance via right SnR expression. A role for retinoid signalling has been identified in the stable maintenance of the conserved post-gastrular left-right gene expressions, and thus the development of situs (Chazaud, Chambon & Dolle, 1999; Zile et al., 2000), though this retinoid role does not appear to extend back into the earlier, symmetry-breaking phase that will now be discussed (Chen et al., 1996). (2) Initial left-right symmetry breaking in earlier development: apparent diversification in timing and in upstream cascade steps A sector of vertebrate development beginning late in neurulation, and extending through the organisation of pharyngeal arches and a heart tube, can be regarded as a relatively conserved ‘phylotypic ’ anatomical stage, the pharyngula (although see Richardson et al., 1997). By this stage, the vertebrate-conserved left-right gene expressions described above have become established in lateral meso- derm, and they too can justifiably be referred to as ‘phylotypic ’. But vertebrate cleavage, blastula formation and early gastrulation are much more variable, and until recently, it has seemed that mechanisms of breaking symmetry to initiate the phylotypic left-right cascades on the correct sides may be as diverse, among vertebrate classes, as is this earlier developmental anatomy. Compara- tive information for this early sector is with few exceptions confined to the widely used ‘model ’ embryo types : the zebrafish (Danio rerio), the clawed frog (Xenopus laevis), the chick (Gallus gallus) and the mouse (Mus musculus), plus relevant human clinical genetic observations. Since this is essentially one embryo per taxonomic class, with no agnathans, elasmobranchs or reptiles, we cannot even be sure that the differing timings of symmetry-breaking and first lateralised gene expressions observed typify the respect- ive classes ; such arrangements might be labile on an even finer evolutionary scale. The various versions of early development seen in mammals are clearly amongthe most derived to be found in vertebrates. Observations on early events in the mouse will nevertheless be described first, since they seem to illustrate most clearly the theoretical predictions that, if the embryo is initially of bilaterally symmetrical cellular struc- ture, left-right patterning must be derived by ‘conversion’ from some form of chiral (i.e. structurally ‘handed’ in having no plane of bilateral symmetry) molecular infor- mation (Afzelius, 1976, 1985; Brown & Wolpert, 1990; Brown, McCarthy & Wolpert, 1991; Almirantis, 1995). The differing arrangements in the avian (chick) embryo will then 382 Jonathan Cooke be described. Chick not only enjoys the longest list to date of genes with left-right lateralised expressions en cascade, but was the first organism in which the recent explosive growth of molecular understanding began (Levin et al., 1995, 1997; Isaac et al., 1997; Pagan-Westphal & Tabin, 1998). This chick gene cascade extends back into earliest gastrulation or before, and is accompanied by a consistent asymmetrical anatomy at mid-gastrulation (Wetzel, 1929; Lepori, 1969; Hara, 1978; Cooke, 1995). Incomplete pictures of early events in fish and frog leave it currently unclear whether they align more closely with chick or with mouse in their earliest lateralised gene expressions. This state of affairs has been proposed to indicate a genuine evolutionary oppor- tunism or lability among vertebrate embryo types, in the initiating mechanisms that precede the conserved ‘phylo- typic ’ sector (Yost, 1999, 2001; Wagner & Yost, 2000). Such evolutionary lability of ‘upstream’ cascade steps in relation to a conserved, downstream sector in developmental mechanism is seen elsewhere, for instance in sex determi- nation, and possible reasons for its origin have been discussed by Wilkins (1995, 2002). Brown & Wolpert (1990) most clearly characterised the requirements for a molecular source of left-right symmetry- breaking information in an otherwise symmetrical embryo. This would have the formal properties of, for instance, the upper-case letter F. If molecules or molecular assemblies of this type were to become aligned in cells, with respect to two other dimensions of organisation or ‘axes ’ that the cells shared, a third dimension of alignment would necessarily be created. If the initial shared alignments were antero- posterior and apico-basal, as could be the case in epithelial structure near the midline in a gastrulating vertebrate embryo, then the molecular basis could be initiated for an organisational difference between cells to either side of the midline, or for a left-right polarised transport via gap junctions and the joint intracellular space. If an axis of beat of embryo cilia were to become fixed at an angle to an embryo-wide axial polarity shared by the epithelial cells, through a tethering according to the ‘F-molecule ’ principle of the known chiral molecular structure of ciliary basal body cross-section, then net bulk transport towards the right or the left of an anteroposterior midline axis in the epithelium could in principle result. Alternatively, a chiral (spiral) ciliary action with a particular (clockwise or anticlockwise) ‘handedness ’ could be determinedmolecularly.With certain spatial arrangements in groups of such cilia near a midline, net bulk transport to the right or left could conceivably result, although in this latter case it is important to recognise that the F-molecule principle is not involved, since individ- ual ciliary assemblies are only aligned with respect to the one, apico-basal cell polarity. The general principles of embryo symmetry-breaking by ‘conversion’ from chiral molecular structure, together with one idea for their possible realisation via ciliary activity, are illustrated in Fig. 2. It was proposed many years ago that normal ciliary activity in the embryo might somehow be translated into the correct directionality of visceral situs, based on the inherited human Kartagener’s syndrome. In these individuals, ab- normal ciliary cross-sectional structure at the electron- microscopic level, and paralysis of ciliary beat, is linked with a random incidence of visceral situs inversus and situs solitus (Afzelius, 1976, 1985). All known cilia/flagellae in the body structure, including spermatozoa tails, are affected in this particular syndrome, causing infertility and chronic airway congestion and infection as additional aspects. But structur- ally atypical cilia, the protocila, have long been known to exist in early vertebrate embryos, distributed one per cell on the apical surfaces of epithelia facing certain cavities. Their motility and functional significance had been questionable, but following a recent exciting series of reports, it is clear that such cilia are indeed motile. Gene products that are necessary for their structure and normal motility are also demonstrably required for normal symmetry-breaking in situs development, at least in mammals. A densely distributed group of protocilia on the ventral surface of the mouse embryo node (anterior tip of the primitive streak), creates in vivo a net flow toward the left across the embryo midline (Nonaka et al., 1998). Normally this must occur within the narrow extracellular space that is the homologue of the primitive vertebrate archenteron. This occurs during the middle part of gastrulation including the start of node regression or head process formation. In mouse, the midline notochord population derived by ingression at the node is at first included in the epithelial surface of the future foregut roof, and for some time protocilia remaining on these cells continue densely to populate the emerging midline anterior to the regressing node. These cilia may also function to maintain the leftward flow or, alternatively, may contribute to left-right barrier formation (see Section II.1). Ciliary activity somehow creates net flow to one side of the midline. A variety of targetted null mutant mice, lacking function of particular proteins involved in ciliary construction and activity, show either absence or paralysis of the embryonic protocilia in conjunction with gross disturbances of down- stream ‘phylotypic ’ left-right gene expression and visceral situs. One of these corresponds with the previously known iv (inversus viscerum) mutant, the gene product now being identified and characterised as left-right dynein (LRD), a relative of the ‘axonemal ’ dynein subfamily (Supp et al., 1997, 1999). In both mutant versions, where stiff, paralysed protocilia are reported though other cilia are normal, the predominant phenotype is random assignment of embryos to either normal or near-normal situs, or complete or near- complete situs inversus, with correspondingly reverse-later- alised expressions of nodal, lefty and Pitx2 genes (Okada et al., 1999). See Fig. 1 for the normal gene expression cascade. Other mutants, for proteins integral to protociliary mor- phogenesis or activity (Nonaka et al., 1998; Marszalek et al., 1999; Takeda et al., 1999) and for the forkhead transcription factor HNF4 (Chen et al., 1998), have no recognisable protociliary structures. In these, interestingly, heterotaxias may predominate, with absent or abnormal bilateral expressions of ‘phylotypic ’ left-right genes, rather than normal or reversed ones. These correlations between ciliary activity and its known molecular basis, the demonstrated extracellular ‘nodal flow’, and symmetry-breaking for visceral situs in a mam- mal, have undoubtedly impressed the community as a major insight into mechanism. They indicate that reliable net Evolutionary origin of vertebrate left/right asymmetries 383 Fig. 2. Possible conversion modes from molecular chirality to left-right developmental information. Upper two diagrams illustrate how a macromolecule or molecular assembly showing chirality (properties formallysimilar to the letter F – see Brown & Wolpert, 1990), initally present as randomly orientated copies in the cells of an epithelium having apico-basal structure and an already- developed axial or antero-posterior polarity (open arrow), could achieve alignment by means of anchoring within cell structure. If, for instance, the cells were linked by gap junctions, such a molecular assembly could mediate directional intracellular ‘morphogen’ 384 Jonathan Cooke leftward transport of an intercellular signal, initially pro- duced symmetrically around the midline, initiates the differential cascades of the lateralised gene expressions. Once started correctly, these are maintained as exclusive by their regulatory cross-relationships and by the midline barrier function (see Section II.1). Since the nodal gene is already locally active around the base of the regressing node, and indeed in the mouse this transcription normally becomes progressively enhanced on the left edge (Collignon et al., 1996), a direct working hypothesis would have nodal protein itself as the leftwards-translocated ‘morphogen’ (see Hamada et al., 2002). This idea has the added attraction that regulatory relationships between nodal and its immediately downstream and related gene lefty appear to embody the principles of local self-activation (by nodal protein) leading to longer-range lateral inhibition (by lefty protein). These have long been proposed on theoretical grounds to charac- terise symmetry-breaking morphogen systems, whose be- haviour when computer-modelled can amplify a small initial disturbance or inhomogeneity into a major, stable asym- metry in the spatial distribution of a developmental signal concentration (Turing, 1952; Gierer & Meinhardt, 1972; Gierer, 1981). Nonaka et al. (2002) now claim to have demonstrated the direct action, in mouse at least, of fluid flow across the relevant embryo surface in determining the directionality of the ‘phylotypic ’ gene expressions and of situs itself. The strikingly ingenious experiments involved culture of the embryos in a microchamber, in which medium flows of controlled rates could be imposed from either left to right or right to left in relation to the primary embryonic axis. The flow rates that reversed situs of normal embryos, or could determine that of ivx/x embryos (with inactive cilia), were grossly comparable with those caused by the normal ciliary activity. Interference was most effective in the period immediately preceding normal onset of left-lateral nodal gene expression. At the same time, the authors emphasise the incompleteness of our understanding of the physical basis of the endogenous directional flow, which appears literally to cross the field of beating cilia rather than, for instance, being whirlpool-like with a particular ‘handed- ness ’ around the node. If the beat of individual protocilia is simply spiral, the net leftward flow would need to result somehow from their dense distribution coupled with the antero-posteriorly distinctive shape of the node base. In this way chiral molecular structure with only one axis of alignment, the apico-basal epithelial axis, could be con- verted to left-right directionality at a tissue level without recourse to the ‘F-molecule ’ principle. If, on the other hand, flow caused by each individual cilium has net directionality because its beat has an off-axial in addition to a spiral component, as speculated above, the ciliary basal body structure could indeed be embodying the full ‘F ’ principle of Brown & Wolpert (1990; see also Brown et al., 1991, discussion transcript). The structure might be tethered, within each cell, in relation to some embryo-wide anatom- ical co-ordinate in addition to the apico-basal cellular one. In birds, the blastula-stage embryo is essentially a sheet of cells that are developing epithelial structure, within which a bar-like thickening due to the piling up or ‘columnarisation ’ of cells comes to define an axis, the primitive streak. As in the mouse, the definitive postgastrular midline arises as cells ingress through the shortening streak to form the new endoderm and mesoderm layers, the space beneath defini- tive endoderm ahead of the node being equivalent to primitive archenteron. These arrangements have obvious structural homology with those in the mouse, but the chick node is a less focussed structure and built on a more massive cellular scale. The mouse ‘ turret ’ structure with clear upper neurectodermal and basal endodermal epithelial layers (see Bellomo et al., 1996) is not found. As gastrulation enters the node-regression stage in chick, the notochord occupies a new middle layer from its inception, without transient integration into the midline archenteron (foregut roof). Single protocilia exist on many cells over an extensive, diffuse region near the midline, throughout gastrulation, on the apical (i.e. outer) surfaces of both endodermal and upper, neurectodermal layers. They are sparsely distrib- uted in space on endoderm due to the flattened, stretched and heterogeneous nature of these cells, and this situation persists once node regression has started (Manner, 2001; Essner et al., 2002; J. Cooke, unpublished observations). This renders implausible the production of significant directional flow as seen in the mouse, especially if ciliary beat has a spiral component. The apical neurectodermal protocilia in this central region are somewhat more prominent, but it should be recognised that any structural ‘handedness ’ within the anchored molecular assembly of cilium and basal body itself, when viewed from a particular surface of the embryo, would necessarily be reversed as between the apical cell surfaces of endoderm and neurecto- derm. To suppose anything else would be to deprive this molecular assembly of the crucial ‘conversion’ role (Brown & Wolpert, 1990) that is the whole attraction of a proposed association of ciliary action with the origin of tissue level left-right asymmetry. transport (arrow from left to right) to initiate differences across the axial midline. Lower digrams illustrate two different ways in which chiral structure inherent in cilia at the basal surface of an embryo midline region, shown facing the reader, could create net flow of extracellular medium to the embryo’s left (reader’s right), as in the observations of Nonaka et al. (1998). In the left diagram, individual cilia express chirality only as the ‘handedness ’ of their spiral mode of beating and are thus not anchored in two dimensions within cell structure according to the ‘F-molecule ’ principle. Net directional flow results in some way from co-operativity between such cilia in relation to the antero-posterior architecture of the embryo’s node region. The right diagram shows the situation where the chiral structure within individual cilia allows for anchoring at a particular angle to the embryo’s long axis, so that their intrinsically directional, whiplash-like beat produces the net leftward flow. It remains unclear, at the time of writing, which principle is most used in the real situation. Evolutionary origin of vertebrate left/right asymmetries 385 In addition to the above problems, a peri-nodal flow comparably timed with that seen in the mouse would be redundant as an initiator of lateralised gene expression in the chick. A cascade of such expressions, preceding any known for mouse, bridges between the earliest gastrula or even blastula stages and the onset of the ‘phylotypic ’ nodal/ lefty/Pitx2/SnR cascade (Levin et al., 1995, 1997; Boettger, Wittler & Kessel, 1999; Rodriguez-Esteban et al., 1999; Yokouchi et al., 1999) (see Fig. 1). Thus, while cilia are indeed diffusely present on the chick nascent archenteron surface, around the time that lateralised gene expressions are first detected in early gastrulation, these associated lateralised gene expressions are quite different from thosein mouse (e.g. Collignon et al., 1996; Lowe et al., 1996). Chick activin receptor IIA mRNA is first detected preferen- tially on the right flank of the early primitive streak, then opposite the anterior streak to the right. This occurs before the appearance of definitive node structure or regression, and thus in the absence of any structural midline barrier. Since this particular receptor is known in other circum- stances to be inducible by an appropriate ligand, a prior step in the cascade is probably the right-sided accumulation of such a ligand, perhaps by preferential gene expression. Selectivity relationships between known receptors and ligands within this superfamily are incompletely known, so despite a report of preferential right activin bB mRNA expression at the same early stages (Levin et al., 1997), it remains unclear that the ligand involved has been identified. Experimentally, activin protein perturbs downstream left- right gene expression and situs when applied to chick blastoderms during these early stages (Levin et al., 1995; Isaac et al., 1997), but experiments may not have been sufficiently controlled to ascertain that perturbation distinc- tively follows from a left-hand experimental application. Recombinant activin protein is highly diffusible, and as will be seen, further roles for other ligands of the TGFb superfamily, in addition to the well-understood one of nodal itself, have been postulated for both right and left sides. In all vertebrates, the signalling gene sonic hedgehog (Shh) has a prominent expression (and role) in the new post- gastrular mesodermal and neural midline that emerges ahead of the regressing node. Early on, in chick, this expression usually appears more prominent in the presump- tive neural ventral midline (floorplate) than in underlying chordamesoderm. Soon, it becomes significantly extended back into the node wall on the left side only. Experiments indicate that it is the preferential right-activation of the ‘activin pathway’, just described, that normally represses Shh expression to the right of the node. Chick nodal expression arises shortly afterwards in deeper-lying, meso- dermal structure immediately lateral to the Shh expression in the left node, and this clearly corresponds with the left- accentuation within the bilateral ring of nodal expression at the base of the mouse regressing node. Though this near – midline left nodal expression is maintained for some hours in both types, no cascade role for it is evident experimentally in chick. Instead, the left-lateral Shh signal at the node directly evokes the separate, broad left-side ‘phylotypic ’ nodal expression via a double-negative mechanism, involving activation of the Cerberus-Dan related extracellular signal gene caronte (Rodriguez-Esteban et al., 1999; Yokouchi et al., 1999; Zhu et al., 1999). Caronte protein is thought to counteract, on the left, the otherwise widespread bilateral bone morphogenetic protein (BMP)-4 signalling at this stage. This in turn relieves a pre-existing BMP-4 repression on nodal expression, thus involving yet another member of the TGFb superfamily in left-right determination. This early, near-midline chick cascade appears locked in, via reciprocal repressive relationships, with another main- tained just to the right of the midline. Thus the right- accentuated ‘activin pathway’, in addition to repressing Shh expression at the right in the node, activates fibroblast growth factor (FGF )-8 expression in the right anterior flank of the streak behind the node (Boettger et al., 1999; Schneider et al., 1999). FGF-8 signalling in turn can repress Shh, and activates the transcription factor gene NKX3.2 in the right streak flank as well as the right-lateral ‘phylotypic ’ SnR (see Section II.1). The potential complexity is indicated by the capacity of BMP-4 signal, whose gene is intensely activated at right of the normal node, to repress normal Shh and activate ectopic FGF-8 expression when locally applied at left (Monsoro-Burq & Le Douarin, 2001). Additionally, at around the time of these differential Shh, FGF-8 and BMP-4 expressions, the specific adhesion-mediating molecule N- cadherin is expressed in an asymmetrical pattern around the node and anterior streak (Garcia-Castro, Vielmetter & Bronner-Fraser, 2000). Role relationships are not yet clear, but interestingly, effects of blocking N-cadherin function on heart-loop direction and embryo torsion may be mediated without affecting the otherwise key downstream ‘phylo- typic ’ Pitx2 and SnR expressions. This raises the possibility of parallel independent, and not interlocked, gene expression cascades. The role of the right-preferential ‘activin-pathway’ expression, at the head of the known chick cascade to date, may be a more direct reflection of an initial symmetry- breaking process than is the subsequent cascade of later- alised expressions. Following activin protein application before stages of node regression, doubly right-sided, doubly left-sided, left-right reversed and normal versions of the downstream gene expressions are produced in approxi- mately equal numbers of embryos (Isaac et al., 1997). This suggests failure of symmetry-breaking followed by a recov- ery process, in which repressive regulatory links ensure that either ‘right ’ or ‘ left ’ cascade but not both gets activated, independently on the two sides but otherwise at random. Such an outcome is reminiscent of those resulting from the inactivations of LRD or ciliary kinesins in mouse (Collignon et al., 1996; Nonaka et al., 1998; Marszalek et al., 1999). It has been hard to establish whether embryo types represented by zebrafish and Xenopus laevis fall more into line with mouse, in proceeding directly from a symmetry- breaking mechanism to the phylotypic cascade, or with chick in utilising an intermediate cascade of lateralised but ‘near-midline ’ gene expressions. Attempts to investigate this in zebrafish through microinjection of Shh-expressing or -interfering constructs are hard to interpret, since the gene has two potential roles in left-right development. In all vertebrate embryos it supports normal function of the notochord and floorplate and thus of the midline signalling 386 Jonathan Cooke barrier, in addition to any left-specific signalling role it may have. Thus in mouse, where no functional or expression evidence for a left-specific Shh role exists, the Shh null mutant phenotype is now recognised to show left-right anomalies connected with breakthrough expression across the midline of the ‘phylotypic ’ nodal and Pitx2 expressions (Meyers & Martin, 1999; Tsukui et al., 1999). To date, no evidence suggests that the cascade just described in chick, beginning with right-hand accentuation of an ‘activin pathway’ and proceeding through left midline Shh expression via a caronte- like function to the left phylotypic cascade, is utilised by frog or fish. We have seen that the derivation of left-right organisation from chiral structure (Fig. 2), i.e. symmetry-breaking by molecular ‘conversion’ (Brown & Wolpert, 1990) occurs considerably earlier in chick development than does the ‘nodal ciliary flow’ that is proposed to initiate this in mouse (Nonaka et al., 1998). The mystery intensifies in that, by contrast, such cilia occur in Xenopus laevis (frog) and Danio rerio (teleost fish) embryos only at late gastrular or even ‘pharyngula ’ stages, and at posterior axial positions (Essner et al., 2002). Especially in the fish case, the timing and position seems hard to link with an origination of the lateralised gene cascades through archenteron-located flow. (3 ) Is there a unitary mechanism for developmental origin of vertebrate situs? Further problems of interpretation remain, for the simplest view of a vertebrate-universal, initial symmetry-breaking mechanism via leftward ciliary flow. The iv gene product LRD, despite its familyrelationships, is largely distributed cyoplasmically and in regions outside those where active protocilia occur, including an anatomically left-lateralised component, in the gastrula/neurula (Supp et al., 1999). Such a distribution would not be consistent with a role confined to initial symmetry-breaking by molecular conversion. The possibility exists that LRD, and/or some of the other gene products whose requirement for normal symmetry-breaking has been assumed to follow from their functions in cilium formation/activity, have additional functions in polarised transport machinery within the joint (gap-junction linked) intracellular space. These transport functions, directionally tethered within cells on the ‘F ’ principle (Brown & Wolpert, 1990), could also be relevant to symmetry-breaking by molecular ‘conversion’. An explanation may still be required for the apparently different consequences of LRD loss on the one hand (random situs correlated with paralysed cilia) and loss of other proteins such as kinesins on the other (heterotaxia and absent cilia), although a larger database would be desirable. Of most difficulty is a mouse mutant not yet mentioned, inv (inversion of embryo turning (Yokohama et al., 1993; Lowe et al., 1996), that almost completely reverses situs at population level rather than randomising it or causing heterotaxias. Elucidation of the nature and cell function of a protein whose mutational alteration achieves such a reversal has been eagerly awaited, during the gene’s protracted positional cloning since its initial characterisation. ‘ Inversin ’ turns out to be a large, intracellular protein of unclear relatedness, containing a domain of multiple ankyrin-type structural units believed to signify tethering to other proteins (Mochizuki et al., 1998; Morgan et al., 1998). The mutant phenotype also exhibits early lethality due to kidney malfunction, possibly signifying inversin involvement in transcellular transport mechanisms (see reference below to Ca2+ signalling), but characterisation of the cellular func- tion remains elusive. Crucially and paradoxically, inv null mutant mouse embryos do not exhibit reversed nodal ciliary flow. Instead, a possible subtle alteration in shape of the node base, and a somewhat more turbulent flow of reduced efficiency in leftward transport of experimental microparticles are re- ported (Okada et al., 1999). The authors of the nodal flow- based hypothesis of symmetry-breaking (Nonaka et al., 2002), understandably, have attempted to modify their model to incorporate and explain this finding, but at the cost of depriving the model of its original merit, which was as an explanation for de novo symmetry-breaking. The model might nevertheless be rescuable from these paradoxical observations if, for instance, it were ultimately to turn out that some intimate cellular architecture of the basal node or nearby midline was instrumental in the normal leftward net flow. The inv mutant alteration, a partial deletion within the protein structure, might conceivably alter this in a way that failed to reverse the experimentally observed flow but did reverse an in vivo functional equivalent. It must be pointed out, however, that the published report of manipulations of the flow does not include results of the formally appropriate ‘negative control ’ condition, namely a manipulated absence of net flow (Nonaka et al., 2002). Such controls are particularly desirable in view of the known tendency of culture in flowing-medium conditions to randomise situs development in mouse and indeed in other vertebrate embryos (Brown et al., 1991; Fujinaga et al., 1992; Fujinaga, Lowe & Kuehn, 2000; J. Cooke, unpublished observations on chick and blastocoel-irrigated Xenopus laevis embryos). Meanwhile, evidence has accumulated for the role of directional intracellular transport via gap junctions, in symmetry-breaking at the earliest stages of both chick and Xenopus laevis development (Levin & Mercola, 1998a, b, 1999; Levin et al., 2002). Experimental interference with gene expression for a connexin that is a principle compo- nent of embryo gap junctions, and related pharmacological channel interference, results in downstream randomisation of situs and associated gene expressions in both species. An intercommunication barrier is reported to exist, for both frog blastula and chick blastoderm, at positions that could be taken as ventral within the respective fate maps. Under these circumstances, if transport machinery had been aligned on ‘F-molecule ’ principles in relation to the embryo axis, preferential ‘morphogen’ transport could occur across a presumptive dorsal midline to build up differences between right and left. The linking of frog and chick embryo types by evidence of this sort is significant because they appear otherwise divergent, within the range of more downstream cascade steps, since the latter but not the former exhibits the near-midline sequence of local gene expressions that starts early in gastrulation and culminates in left Shh and Caronte (see Section II.2). Observations on mouse, where the claim Evolutionary origin of vertebrate left/right asymmetries 387 for causal primacy of ciliary flow seems strongest, further support a generality of the involvement of intracellular small molecule transport in the symmetry-breaking phase for situs (Pennekamp et al., 2002). A final recently forged link between apparently diverged embryo types concerns the distinctive left-right role of the veg (vg) genes, encoding yet another subgroup of the TGFb superfamily of signal proteins. Vg1, the first-identified member in Xenopus laevis, has been a strong candidate for an endogenous maternally encoded signal whereby the yolky vegetal blastomeres induce mesoderm and definitive endoderm in the equatorial zone of the blastula (Thomsen & Melton, 1993; Kessler & Melton, 1995). Orthologues and closely cognate genes are now known additionally from frog, zebrafish, chick and mouse (Seleiro, Connolly & Cooke, 1996; Sun et al., 1999; Wall et al., 2000). Their candidacy for primary axial patterning roles has perhaps been usurped by the nodal genes themselves (Varlet et al., 1997; Varlet, Collignon & Robertson, 1997; Brennan et al., 2001), even in the yolky eggs of frog and fish where maternally translated genes are prominent at early stages (Schier & Shen, 2000). A developmental patterning role for Vg genes is nevertheless suggested by their special processing requirements for release of active ligand. The enzyme system, whereby functioning C-terminal polypeptide is cleaved off from the pre-proprotein at secretion, is not the widely available one utilised by TGFb family members generally. Normal development in the frog and the chick occur without production of immunologically detectable amounts of the working Vg ligand, despite abundant bilaterally distributed mRNA and pre-proprotein. The assumption is that the endogenous role(s) of Vg genes involve localised production of the minute amounts of ligand required, due to localisation of a crucial part of the processing machinery. Experimental ectopic mis-expression of Vg genes employs chimaeric DNA or RNA constructs, in which the C- terminal distinctive Vg sequence is fused with an N-terminal sector encoding the permissive processing parts from another TGFb gene. In the frog such experimental mis- expression has been found to exert a distinctive effect on situs (Hyatt, Lohr & Yost, 1996; Hyatt & Yost, 1998; Hanafusa et al., 2000). Applied on the right, Vg is able systematically to reverse the expression of the ‘phylotypic ’ cascade and morphological situs itself, rather than randomising these or giving rise to left-isomeric development as experimental right expression of nodal or Pitx protein would tend to do. There is additional molecular evidence that the experimen- tal effects are distinctive to Vg itself, rather thandue to direct right-ectopic activation of the response pathway to nodal protein, for instance. This has led to a proposed endogenous role as a ‘ left-right co-ordinator ’ signal in normal development. The implication is that this role is normally exerted from a relatively lateral left position. The endogenous time of action is unknown, but data suggest that experimental action on situs is still fully effective in the early gastrula stage, and possibly much later than this (Toyoizumi, Mogi & Takeuchi, 2000). Either chick Vg, or chick or mouse nodal proteins have been locally mis-expressed on the right, in an extensive series of chick blastoderms in culture ( J. Cooke & S. Withington, unpublished work). Here, too, Vg exerts a distinctive effect. Whereas Nodal protein, present on the right from stages of node regression and midline formation, randomises situs and induces right ectopic expression of nodal and Pitx2 genes as would be expected, right-expressed Vg protein is able with significant frequency to reverse the expression of Pitx2 to give right expression only. Further- more, as would be expected from its downstream position in the normal cascade, right-ectopic nodal protein is unable ever to disturb the left-sided Shh expression at the node itself (or the associated gastrular structural asymmetry, see Section III.2). Comparable ectopic Vg protein expression does indeed disturb or reverse these features, when imposed from stages before they have developed. Finally, the null mutant phenotypes have been described for the mouse Vg1 orthologue Gdf-1 (Rankin et al., 2000), and for furin, a convertase component of a processing pathway that may distinctively be required for certain TGFb superfamily ligands (Constam & Robertson, 2000). These involve right isomerism and left-right inversions of particular visceral arrangements, and typically, absence of the ‘phylotypic ’ left-lateral gene expressions. These observations all suggest that Vg genes have a vital vertebrate-conserved role, situated not far downstream of the initial symmetry breaking process for situs. (4) Changing left-right co-options of gene orthologues during vertebrate diversification, and other puzzles FGF-8 in chick has a distinctive right-hand expression which is functionally integrated into that species’ early near- midline cascade (see Section II.2). The evidence in mouse, however, is that it is a left-specific player at comparable stages (Meyers & Martin, 1999). The transcription factor gene Nkx 3.2, a left-hand player apparently downstream of nodal and kept repressed by right-hand FGF-8 in chick, correspondingly appears to swap sides in mouse (Schneider et al., 1999). In chick, as also described in Section II.2, the extracellular caronte protein is crucially involved in propa- gating left information from the node into lateral mesoderm, being induced by left Shh and acting to de-repress phylotypic left-lateral nodal expression by interfering with BMP-4 signal. In mouse, where Shh has no known lateralised expression or role and where a left caronte gene expression orthologous with that of the chick has not been found, a gene cognate with caronte, Dte, is nevertheless reported to show distinctive right-hand expression at comparable stages in the node (Pearce, Penny & Rossant, 1999). This is puzzling for an additional reason; potentiation, rather than counteraction, of BMP-4 signal is believed to be important at a right-lateral position, as this antagonises Vg signal there (see above, Ramsdell & Yost, 1999; Branford, Essner & Yost, 2000). Most tested Cerberus-Dan family proteins antagonise BMP-4 signalling although dte has not explicitly been tested. In any version of a lateralising gene cascade, there is no reason why a particular gene should not have distinctive roles in different steps on opposite sides of the body, pro- vided that the cascade steps are sequential, circumscribed in 388 Jonathan Cooke time and well-regulated within the developing anatomy. BMP-4, for instance, might have more than one successive role on the right, counteracting Vg signalling from the left in all embryo types (Ramsdell & Yost, 1999; Branford et al., 2000), and restricting Shh expression at the node to the left in types utilising this expression (Monsoro-Burq & Le Douarin, 2001). Additionally, it has a later one on the left in connection with the heart in at least one embryo type (Chen et al., 1997). It is partly for this reason that negative interference or targetted mutagenesis of receptor and downstream signal transduction components can be prob- lematic in interpretation when there is a ‘ left-right ’ pheno- type. Each deleted component may have had normal roles through signalling by more than one ligand within a superfamily, and could thus have acted in ‘right ’ or in ‘ left ’ propagation, or in symmetry breaking, at different times. A recent report concerning FGF-8 in rabbit establishes that apparent ‘ side-swapping’ of gene roles between embryo types can occur even within a vertebrate class (Fischer, Viebahn & Blum, 2002), and may be more dependent upon the morphology of gastrulation than upon deeper evolution- ary change. This paper also clearly indicates the interpret- ative perils attending multiple successive roles for particular genes within one overall ‘cascade’. Despite striking apparent differences, observations cur- rently leave room for the belief that both the left-right developmental gene cascade and the underlying mechan- isms of deriving symmetry-breaking information from chiral molecular structure, might be conserved for vertebrates. Protocilia probably are present, albeit in a wider and less dense distribution, across the archenteron or equivalent surfaces of all vertebrate embryo types over a prolonged period of early development. Pending an exact understand- ing of the way in which these cilia operate, the possibility remains of their broadly conserved role involving directional extracellular transport. But there is the strong additional possibility that some early process of intracellular polarised transport is universally present, at pre-gastrula or later stages, and functionally important in symmetry-breaking for all vertebrate types. There is a need to elucidate the role of the inversin (inv) protein. Additionally, a process at or close to the head of the cascade also appears universally to in- volve the left-lateral Vg signalling, while other TGFb signal superfamily members have widespread roles. It is note- worthy that this superfamily of signal proteins, centrally involved in vertebrate axial induction and anteroposterior organisation (Smith et al., 1990; Cooke, 1991; Schier & Shen, 2000; Brennan et al., 2001), should be central also in the mechanism of situs. III. VIEWPOINTS ON THE EVOLUTIONARY ORIGINS OF SITUS (1 ) The functional ‘ just-so’ story: situs as a vertebrate (chordate) invention It can plausibly be argued that, provided a basis for its mechanism existed within cellular organisation, a reliably directional packing of the major viscera within the stream- lined body wall would have been strongly selected for during the origin of modern vertebrates. The argument runs as follows. Transition from a sedentary or crawling, detritus- or filter-feeding, to an actively swimming predatory lifestyle is likely to have accompanied the origin of the vertebrate clade. Greatly increased metabolic demands would have been involved. Additionally, a square–cube relationship applies between available surface areas for nutritional absorption and gas exchange on the one hand, and tissue mass on the other, if strictly equivalent anatomical structure increases in linear size. If, as seems likely, a progressive size increase took place, there would therefore have been strong selective pressure for proportional increase in the lengths and complexity of tubular viscera and ofpumping vascu- lature. Within the streamlined outer ‘ locomotory’ body shape that was meanwhile being selected for, such increases for rate of food absorption and of vascular exchange could only have been accomodated by the more complex looping and/or coiling of these organ systems. For gut function, the relationship is intuitively obvious, while in the case of vascular function the effect of a spiral and looped arrange- ment, in optimising the pumping output per unit mass of muscular tube wall, has recently been described for the mammalian heart (Kilner et al., 2000). Provided that the visceral packing problem is ‘solved’ in precisely the same way in all normally developing individuals, the further opportunity arises for intimate co-ordination, thus subtle co-optimisation, between the originally quite independent peristaltic actions of vascular and digestive systems. Given a bilaterally symmetrical cellular structure of the gastrula, and the disparate origins of the internal organ rudiments within the body plan, regularisation of internal packing could only occur if cascades of gene expressions, propagating separately in tissues to either side of the midline, imbued them with left and right positional identities. In this way, left-right differential growth and morphogenetic patterns could be initiated co-ordinately in the various internal rudiments. The mechanistic develop- mental problem appears as that of the reliably directional, or ‘handed’ breaking of the initial embryonic bilateral sym- metry, and its conversion into the initiation of the lateralised gene expression cascades. As detailed in Section II.2, Brown & Wolpert (1990) made an influential formulation of the problem in asserting that the only mechanism whereby such directional symmetry-breaking could occur is by ‘con- version’ from the information implicit in a chiral macro- molecule or molecular assembly. It is widely recognised that indeed some such molecular conversion process is needed if the embryo starts out bilaterally symmetrical at the cellular level of structure. Thereafter, appropriate regulatory exclus- ivities between the lateralised gene expressions that are triggered, together with a stabilising barrier function in the midline against cross-invasion of signals, will ensure geneti- cally ‘ left ’ and ‘right ’ sides with potentially separately evolvable developments. Among vertebrate embryo types examined to date, the majority indeed appear entirely bilaterally symmetrical at stages before the first features of situs develop. The bilaterian clade of animal forms, by definition, is considered to have Evolutionary origin of vertebrate left/right asymmetries 389 Fig. 3. Schematic of a proposed dexiothetic transition within vertebrate ancestry. Three body forms are shown, each in dorsal plan and in composite transverse sectional views (-axial levels indicated by dashed lines). The original, truly bilaterian ancestor on the left has paired special sense organs (black) and filter- or detritus-feeding apparatus anteriorly, paired and probably segmented mesoderm structures (stippled), and a tubular gut without left-right but possibly with dorso-ventral complexity. The central diagrams represent the body form shortly after the morphological transition, whereby an original right side became a ‘ventral ’, substratum-applied side, 390 Jonathan Cooke derived from a common ancestor having acquired an axis of true bilateral symmetry. Most recent accounts of the explosion of molecular knowledge about vertebrate left- right development have, understandably, been from mol- ecular genetic researchers whose project is a mechanistic understanding of contemporary embryonic development. They have tended to lack exposure to the disciplines of comparative anatomy and phylogenetics, with their accent on the dimension of historical contingency in evolution and their propensity to ask: ‘via which route do current biological arrangements happen to have come about? ’ It has therefore been easy to propose or assume that a symmetry-breaking mechanism, and the gene-expression cascades for situs, were ‘ invented’ entirely in response to the selective pressures attending the ancestry of vertebrates themselves, or at least that of the clade of most obviously related organisms in which they are embedded, namely the chordates. The exciting correlation of mutations that affect ciliary structure or function with disturbance of situs and of the ‘phylotypic ’ left-right gene cascades in mammals (see Section II.2), has brought the feeling that an understanding of this molecular ‘ invention’ is close at hand. The whole mechanism seems to conform with what theory would dictate as necessary, during each contemporary vertebrate embryo’s development, for directional de novo breaking of a ‘primordial ’ bilateral symmetry. (2 ) A constrasting view: situs as the co-optation of a pre-existing axis There exists a substantial alternative viewpoint to the above, however, based on comparative anatomy and phylogenetics. This would not postulate that the ur-vertebrate or even ur- chordate started out with axial symmetry of the bilaterian sort. Rather, it is seen as having evolved a secondary, derived bilaterality in the outer ‘ locomotory’ body wall and nervous system, in relation to the newly active mode of life. Enclosed within this bilateralised shell was an archaic ‘visceral ’ body that had, much earlier in the history of diversification of animal forms, abandoned the primordial symmetry ancestral to the Bilateria in a profound morpho- logical transformation. In essence, the ancestral bilaterian right-hand side is proposed to have become a new ‘ventral ’ surface applied to the substratum, with resultant loss of primary bilateral symmetry to give a sessile or slow-moving, detritus- or filter-feeding form. This proposed transform- ation, perhaps with unfortunate consequences for its sale- ability to a wider community, has been termed dexiothetism. It may have occurred with the foundation of the entire clade known as the Deuterostomia, or alterna- tively, be shared by a more restricted grouping that includes echinoderms and chordates (see below). Vertebrates would on this view retain a deeply embedded, non-symmetrical ‘visceral ’ body organisation resulting from the dexiothetic transformation, combined with an axial but only secondarily re-symmetrised ‘ locomotor ’ component. The archaic com- ponent including its regionalised developmental gene ex- pressions, laid out in a defined way with respect to this newly re-symmetrised organisation, would provide the basis for any retained, then further elaborated, internal left-right structure. Dexiothetism is illustrated in outline in Fig. 3, together with its proposed consequences for vertebrate structure. Hemichordates, echinoderms, urochordates, cephalo- chordates and vertebrates rather clearly form a clade, the now-accepted Deuterostomia, when fossil (stem) forms, anatomical structures shared by extant forms, features of embryonic development and the newer (mostly 18S ribo- somal DNA-based) molecular systematic evidence are all considered together (Gislen, 1930; Jefferies, 1975; Jefferies, 1986; Jefferies, 1991; Jefferies, Brown & Daley, 1996; Cameron, Garey & Swalla, 2000; Holland, 2000). The Hox-type homeobox gene cluster seems basal to animal axial organisation, and the possession of a relatively complete cluster of Hox gene orthologues even by echino- derms, that have no apparent axial bilaterality in their extant life-histories, justifies the assumption that the deuter- ostome clade shares a true bilaterian ancestry (Peterson, Cameron & Davidson, 2000). Current attempts at molecu- lar phylogenetics within this clade lack resolving power however, such that almost every group named above can appear to be a basal sister group to all the others if particular genomic sequences are considered
Compartilhar