Mechanisms of convergence and extension by cell intercalation · Convergence and extension...

Mechanisms of convergenceand extension by cell intercalation

Ray Keller1 Lance Davidson1 Anna Edlund1 Tamira Elul2 Max Ezin1David Shook1 and Paul Skoglund1

1Department of Biology Gilmer Hall University of Virginia CharlottesvilleVA 22903 USA2Department of Physiology Howard Hughes Medical Institute University of California San Francisco CA 94143- 6723 USA

The cells of many embryonic tissues actively narrow in one dimension (convergence) and lengthen in theperpendicular dimension (extension) Convergence and extension are ubiquitous and important tissuemovements in metazoan morphogenesis In vertebrates the dorsal axial and paraxial mesodermal tissuesthe notochordal and somitic mesoderm converge and extend In amphibians as well as a number of otherorganisms where these movements appear they occur by mediolateral cell intercalation the rearrange-ment of cells along the mediolateral axis to produce an array that is narrower in this axis and longer inthe anteroposterior axis In amphibians mesodermal cell intercalation is driven by bipolar mediolaterallydirected protrusive activity which appears to exert traction on adjacent cells and pulls the cells betweenone another In addition the notochordal somitic boundary functions in convergence and extension bycapturingrsquo notochordal cells as they contact the boundary thus elongating the boundary The prospectiveneural tissue also actively converges and extends parallel with the mesoderm In contrast to the meso-derm cell intercalation in the neural plate normally occurs by monopolar protrusive activity directedmedially towards the midline notoplate^poundoor-plate region In contrast the notoplate^poundoor-plate regionappears to converge and extend by adhering to and being towed by or perhaps migrating on the under-lying notochord Converging and extending mesoderm stiiexclens by a factor of three or four and exerts upto 06 m N force Therefore active force-producing convergent extension the mechanism of cell inter-calation requires a mechanism to actively pull cells between one another while maintaining a tissuestiiexclness sucurrencient to push with a substantial force Based on the evidence thus far a cell^cell tractionmodel of intercalation is described The essential elements of such a morphogenic machine appear to be(i) bipolar mediolaterally orientated or monopolar medially directed protrusive activity (ii) thisprotrusive activity results in mediolaterally orientated or medially directed traction of cells on oneanother (iii) tractive protrusions are concentned to the ends of the cells (iv) a mechanically stable cellcortex over the bulk of the cell body which serves as a movable substratum for the orientated or directedcell traction The implications of this model for cell adhesion regulation of cell motility and cell polarityand cell and tissue biomechanics are discussed

Keywords convergence extension gastrulation Xenopus morphogenesis motility

1 INTRODUCTION

(a) What are the convergence and extensionmovements

Convergence and extension are classical terms fornarrowing and lengthening respectively of a cell popu-lation during morphogenesis (reviewed in Keller et al1991ab) (centgure 1a) These movements occur ubiquitouslyin metazoan development and probably account for moretissue distortion in embryogenesis than any other singleprocess (Keller 1987 Keller et al 1991b) Examples includearchenteron elongation in echinoderms (Ettensohn 1985Hardin amp Cheng 1986) germ band extension (Irvine ampWieschaus 1994) and imaginal leg disc evagination inDrosophila (see Condic et al 1991) Hydra regeneration

(Bode amp Bode 1984) nematode body axis elongation(Priess amp Hirsh 1986 Williams-Masson et al 1997) andelongation of dorsal axial embryonic tissues of ascidians(Cloney 1964 Miyamoto amp Crowther 1985) centshes(Warga amp Kimmel 1990 Kimmel et al 1994 Concha ampAdams 1998) birds (Schoenwolf amp Alvarez 1989)mammals (Sausedo amp Schoenwolf 1994) and amphibians(Vogt 1929 Schechtman 1942 Burnside amp Jacobson 1968Keller 1984 Keller et al 1991ab) Convergence can becoupled directly to extension with conservation of tissuevolume the decrease in width accounted for by a propor-tional increase in length (centgure 1a) More often howeverconvergence produces thickening as well as lengtheningthe proportion of each depending on the specicentc case(centgure 1b) For example the notochordal and somiticmesoderm of the amphibian both converge extend andthicken but convergence results in less extension andmore thickening in the case of the somitic mesoderm (seeKeller et al 1989a Keller 2000) Thus the collective term

Phil Trans R Soc Lond B (2000) 355 897^922 897 copy 2000 The Royal Society

Author for correspondence (rek3kunixmailvirginiaedu)

doi 101098rstb20000626

`convergent extensionrsquo can be used for convenience but itcan be misleading unless one remembers that the rela-tionship between convergence extension and thickeningcan be and usually is more complex than this termsuggests

Convergence and extension movements are representa-tive of a larger more general class of `mass movementsrsquoinvolving change in tissue proportions with approximateconservation of volume Divergence and shortening(centgure 1c) the reverse of convergent extension occursduring Drosophila germ band retraction (Hartenstein ampCampos-Ortega 1985) Epiboly an increase in tissue areaby thinning (centgure 1d ) occurs in the animal region ofearly amphibian (see Keller 1978 1980) and teleost centshesembryos (Betchaku amp Trinkaus 1978) to mention onlytwo examples Convergence and extension are oftenmajor components of what are usually identicented as othertypes of morphogenic processes For example invagina-tion of a disc of cells to form a tubular gut such asprimary invagination of a echinoderm vegetal plate toform the gut is often portrayed as a simple bending insectional view (centgure 1e) In fact such an invaginationinvolves a form of convergent extension the progressivenarrowing of the circumference of the disc and extensionalong its radius (the length of the tube) (centgure 1e)

Convergence and extension have been studied most insituations where tissue volume changes due to cell growthdoes not occur at all or is insignicentcant such as in theearly development of sea urchins (Hardin amp Cheng 1986)amphibians (Burnside amp Jacobson 1968 Keller 1986)and poundies (Condic et al 1991) However these movementsalso occur in embryos that grow dramatically in earlydevelopment such as those of the mouse (Snow 1977) and

in these cases tissue volume changes may also come intoplay Analysis of the mechanism of convergent extensionin organisms that grow rapidly may reveal variations onthe themes developed here specicentcally the potentialinvolvement of directed anisotropic growth

Convergence and extension movements as well asrelated mass tissue movements may be passive responsesto forces generated elsewhere in the embryo or they maybe active force-producing processes We will focus on theactive movements here although the passive ones are alsoimportant in morphogenesis and misunderstanding themcan result in erroneous conclusions For example it is acommon assumption that if a mutation targeted to aparticular tissue in the embryo disrupts convergent exten-sion of that tissue the movement is an active oneHowever the movement could be a passive response toforce generated elsewhere and the true function of thewild-type gene could be to increase the deformability ofthe tissue allowing it to be stretched and narrowed bythe active tissue

(b) Why study convergence and extensionThese movements present a major challenge and an

opportunity to understand cell function in the morpho-genesis of integrated populationsThey represent a categoryof morphogenic processes the so-called `mass movementsrsquoin which a tissue distorts by virtue of integrated cell beha-viours within its boundaries due to interactions amongthese cells rather than interactions with external substratesLittle is known of the cellular molecular and bio-mechanical mechanisms of these types of movements animportant decentciency in the light of their dramatic role inshaping embryonic body form in so many systems

898 R Keller and others Convergence and extension by cell intercalation

PhilTrans R Soc Lond B (2000)

(a) (c) (e)

(i)

(ii)(b)

(d)

Figure 1 Diiexclerent types of mass tissue movements including (a) convergence and extension (b) convergence extension andthickening (c) divergence and shortening (d ) epiboly and (e) invagination in (i) sectional view and (ii) in three dimensions

The conceptual framework for analysis of cell inter-actions within integrated populations is poorly developedIt is dicurrencult to visualize and interpret cell motility inembryos and thus little is known of the cell behaviourdriving these movements The common paradigm foranalysis of cell function in morphogenesis is the study ofmotility of individual cells in culture at low densitycrawling on a rigid transparent substratum In contrastmost cell movements in embryos involve high densities ofcells interacting with one another or with the extra-cellular matrix between them Moreover the forces indi-vidual cells generate have largely local eiexclects on theirmovement in culture whereas these forces have bothlocal eiexclects and eiexclects that are integrated over the cellpopulation in embryos The mechanism for global inte-gration of locally generated forces and the molecularbasis of the necessary cell and tissue biomechanical prop-erties are poorly understood in both concept and experi-mental measurement Finally too often the paradigm forgenetic and molecular analysis of tissue movements hasbeen compositional rather than mechanistic Experimentshave been focused on interdicting the expression of a geneor the function of a molecule and showing that the geneor molecule it encodes is necessary for a morphogenicprocess to occur Collectively these experiments resolve

the composition of the machine rather than how themachine works Here we will discuss `morphogenicmachinesrsquo largely at the cell and cell population levelCell biology has decentned a large number of molecularmachines for generating force for generating cell polarity(Brunner amp Nurse this issue Schwab et al this issue)imparting mechanical integrity for regulating motility(Hall amp Nobes this issue) and for imparting cell ad-hesion (see Takeichi et al this issue) and contact beha-viour (see Xu et al this issue)

It is our goal here to decentne and characterize thecellular machines that connect these molecular machinesto massive movements of large coherent cell populationsin the embryo

(c) The paradox of convergence and extension mobility in the presence of stiiexclness

We will focus on what is emerging as the most commontype of convergence and extension movements those thatoccur by cell rearrangement The paradox of this type ofconvergence and extension is that the cells actively inter-calate between one another to produce a change in shapeof the tissue while forming a stiiexcl array that can distortsurrounding passive tissues The major question to beaddressed here is how can cells move with respect to one

Convergence and extension by cell intercalation R Keller and others 899

Phil Trans R Soc Lond B (2000)

BP

lateneurula

lategastrula

earlygastrula

lateralsurface

Fb

So

HM

HM

N

Hb and Sc

midsaggital

3D cut-awayof mesoderm

Figure 2 Expression of the convergence and extension movements during gastrulation and neurulation of Xenopus laevis Tissuesundergoing convergence and extension include the prospective hindbrain (Hb) and spinal cord (Sc) regions of the neural tissue(light shading) the prospective somitic mesoderm (So medium shading) and the prospective notochord (N dark shading) Alsoshown are the prospective forebrain (Fb) the prospective head mesoderm (HM) and the blastopore (BP)

another and yet form a stiiexcl tissue capable of pushing anddistorting surrounding tissues

We will discuss convergence and extension movementsin the amphibian Xenopus laevis in more detail It is one oftwo systems in which cell behaviours have been describedIt oiexclers at least two examples of convergent extensionone in the posterior mesodermal tissue and one in theposterior neural tissue consisting of what appear to bevariants of a basic mechanism Analysis of these twosystems may reveal common and perhaps essentialfeatures of the mechanism of convergent extension by cellintercalation Finally convergence and extension inamphibians is demonstrably an active force-producingprocess and it is the active forms of these movements thatwill be focused on here although those examples ofpassive responses to external forces are interesting in theirown right

2 EXPRESSION AND FUNCTION OF CONVERGENT

EXTENSION DURING FROG GASTRULATION

AND NEURULATION

During gastrulation the prospective somitic (centgure 2medium shading) and notochordal mesoderm (centgure 2dark shading) involute over the blastoporal lip and asthey do so they converge in the mediolateral orientationaround the circumference of the blastopore and extendin the anteroposterior orientation to form the elongatedbody axis (see Keller 1975 1976 Keller et al 1991b)(centgure 2) Meanwhile the posterior neural plateconsisting of the prospective spinal cord and hindbrain(centgure 2 light shading) likewise converges in the medio-lateral orientation and extends in the anteroposterior axis(centgure 2) more or less coincident with the underlyingconverging and extending mesodermal tissue (Keller et al1992a) Similar movements occur in the urodele (Vogt

1929 Schechtman 1942 Jacobson amp Gordon 1976) butthey diiexcler in their timing and the degree of tissue distor-tion

These movements play major roles in gastrulation andbody axis formation During gastrulation the forces ofconvergence form a hoop stress around the blastoporewhich squeezes the blastopore shut in the normal aniso-tropic fashion towards the ventral side of the embryo(Keller et al 1992b) These forces also contribute to involu-tion of the mesodermal and endodermal tissues (Shih ampKeller 1992a Keller et al 1992b Lane amp Keller 1997)although the vegetal endodermal rotation movementsrecently discovered by Winklbauer amp Schulaquo rfeld (1999)account for much of the early involution movementsFinally convergent extension also morphologically decentnesthe anteroposterior body axis The prospective anteropos-terior body axis of vertebrates is remarkably short andwide in the pregastrula stages The dorsal mesodermaland neural tissues converge and extend tremendouslyduring gastrulation and neurulation pushing the headaway from the tail and thereby morphologically decentningthe anteroposterior body axis that is so important ingetting through life head centrst (centgure 2)

3 AMPHIBIAN CONVERGENT EXTENSION

IS AN ACTIVE FORCE-PRODUCING PROCESS

The centrst step in analysis of an example of specicentcconvergent extension movement is determining whether itis an active force-producing process or a passive responseto forces produced elsewhere in the embryo The source ofthe forces causing these movements in amphibiansremained a mystery until explants of tissue activelyconverged and extended in culture mechanically isolatedfrom the rest of the embryo (Schechtman 1942 Holtfreter1944 Jacobson amp Gordon 1976 Keller et al 1985) In



Fb

(b)

(a)

So

No

Hb and Sc

Figure 3 Comparison of the convergence and extension movements in (a) the whole embryo and (b) an explant of the dorsalsector of the gastrula Normally the prospective notochordal (No dark shading) and somitic (So medium shading) mesoderminvolutes beneath the prospective posterior hindbrain (Hb) and spinal cord (Sc light shading) and the mesodermal and neuraltissues converge and extend posteriorly together (arrowheads in a) When the dorsal sectors of two early gastrulae are cut outand sandwiched with their inner deep cell surfaces together the explant converges extends and diiexclerentiates in culture withoutforces generated in the rest of the embryo and without traction on an external substratum (b)



Figure 4 (a) An explant between the stage and the probe of a computer-controlled biomechanical measuring machine the`Histowigglerrsquo during a compressive stress relaxation test for tissue stiiexclness (see Moore et al 1995) (b) Compression stressrelaxation tests compare stiiexclness (E in N m7 2 after 180 s compression) of the mesoderm of the early gastrula before involutionwith stiiexclness of the same tissue at the late midgastrula stage after involution (i) Before involution E(180) ˆ sup13^4 N m7 2(ii) after involution E(180) ˆ sup114^15 N m7 2 The tissues tested are shown in grey shading the test was done with compressionalong the anteroposterior axis The force of extension is measured with an isometric force measurement (c) in which a convergingand extending explant (shaded) extends against the probe of the Histowiggler with a force (dark arrow) that is balanced by anopposing force from the Histowiggler probe (light arrow) such that the length of the explant does not increase Force measure-ment 10^12 m N for a sandwich of two and 05^06 m N for a dorsal mesoderm (d ) Convergence and extension of the mesodermis independent of the blastocoel roof When deprived of its normal migratory substrate the blastocoel roof (i) dorsal notochordal(N) and somitic tissue (stained) converge and extend (arrows) without the overlying substrate (ii) The blastopore is indicated byBP or a triangle (e) The extending axial and paraxial mesoderm controls the shaping of the embryo When a blastocoel roofsubstrate is oiexclered to the extending dorsal mesoderm oiexcl-axis at the sides of the embryo (dorsal is at the bottom of the centgure) (i)the dorsal mesoderm extends up the midline of the embryo and the `substratesrsquo on both sides are pulled dorsally over theextending mesoderm (ii arrows)

explants composed of two dorsal sectors of the earlyXenopus gastrula sandwiched together the prospectivenotochordal and somitic mesoderm and the prospectiveposterior neural tissues converge and extend by internallygenerated forces independent of traction on the substrateand of forces generated elsewhere in the embryo (Keller etal 1985 Keller amp Danilchik 1988) (centgure 3)

(a) Tissues undergoing convergent extension forma stiiexcl embryonic skeleton that decentnes the shape

of the embryoThe independent convergence and extension of sand-

wich explants in culture implies that these tissues mustpush and exert enough force to overcome their owninternal resistance to distortion and to deform the neigh-bouring passive tissues in the embryo as well In fact theconverging and extending tissues stiiexclen to form anembryonic `skeletonrsquo capable of pushing forces thatelongate the embryonic axis and take a leading role indecentning the shape of the early embryo Moore andassociates (Moore 1992 Moore et al 1995) developed acomputer-controlled biomechanical measuring devicethe `Histowigglerrsquo (centgure 4a) capable of measuringmechanical properties of embryonic tissues and the forcesthey generate Uniaxial compression stress relaxationtests made with this machine showed that the dorsal meso-dermal tissue stiiexclens in the axis of extension by a factor ofthree or four during extension (centgure 4b) and isometricforce measurements show that it is capable of pushing witha force of about 05 m N (Moore 1992 Moore et al 1995)(centgure 4c) The onset of this stiiexclness is regulated such thatit occurs after involution of the mesodermal tissuesPatterning the expression of this increased stiiexclness to thepost-involution region appears to be essential for gastrula-tion When substantial pieces of the post-involution meso-derm are grafted back to the lip of the blastopore they willnot involute but will sit up on the lip of the blastopore likea canoe on the edge of a waterfall apparently too stiiexcl toturn the corner (R Keller unpublished data)

The role of the stiiexcl forcefully extending mesoderm inshaping the embryo is illustrated by experiments in whichthe roof of the blastocoel was removed (Keller amp Jansa1992) Under these conditions the axial and paraxialmesodermal tissues involute and extend into the liquidmedium without an overlying blastocoel roof to serve as asubstratum (centgure 4d ) Moreover these dorsal meso-dermal tissues rather than the blastocoel roof `substratumrsquodominate the shaping of the embryo If fragments ofblastocoel roof are oiexclered as substrata for migration inregions to the sides of the axis of extension these tissues donot change their direction of extension to accommodatethe substratum rather the substratum is pulled over thestiiexclened extending tissues (centgure 4e)

The neural region can also push although the force itgenerates has not been measured When put in serialopposition with the mesodermal component by barricadesat both ends of the extending explant the neural regionbuckles in the face of extension of the mesodermal regionarguing that the neural region is the least stiiexcl and thusprobably weakest extender of the two In the embryo themesodermal tissues are fastened to the overlying neuralplate by a strong attachment of the notochord to theregion of the neural plate overlying it the `notoplatersquo

(Jacobson amp Gordon 1976) These attached neural andmesodermal tissues extend together (A Edlund andR Keller unpublished data Keller et al 1992a) In thissituation the force generated should be at least the sumof the forces generated separately but may be greater ifcontact between them makes one the other or both moreforceful in extension

(b) Biomechanics of other systemsIt is not clear whether many examples of convergence

and extension are active or passive processes and ifactive how much force they generate The axial andparaxial tissues of centshes can extend as explants in isola-tion from the embryo (see Laale 1982) and they can pushinto the yolk in teratogenized embryos (see Bauman ampSanders 1984) implying an internal force-generatingprocess However the notochord and somitic centles in theseexplants are often kinked as if they bucked in the processof extending suggesting that the converging andextending tissues of the teleost centshes may be less stiiexcl andextend with less force than those of Xenopus The variationin design of early morphogenesis is substantial evenbetween closely related taxa of amphibians (Keller 2000)and thus mechanical variations of this type should beexpected among the vertebrates Imaginal leg discs ofDrosophila (see Condic et al 1991) converge and extendduring evagination in culture which indicates an activeprocess Experimental manipulation and centnite elementmodelling suggests that the archenteron of echinodermsconverges and extends by an internal force-producingprocess (Hardin amp Cheng 1986 Hardin 1988) and muta-tions aiexclecting germ band elongation in Drosophila suggestbut do not prove that the same may be true of thisprocess (see Irvine amp Wieschaus 1994)


OCCURS BY CELL INTERCALATION

Because convergent extension of axial structures ofamphibians occurs in absence of cell growth and theappropriate changes in cell shape Waddington (1940)suggested that these movements occur by the rearrange-ment of cells In fact morphological studies tracings ofcells with pounduorescent labels and time-lapse recordings ofcell behaviour in both mesodermal (Keller amp Tibbetts1988 Keller et al 1989a Wilson et al 1989 Wilson ampKeller 1991 Shih amp Keller 1992ab) and neural regions ofXenopus (Keller et al 1992a Elul et al 1997 Elul amp Keller2000) have shown that these tissues converge and extendby two types of cell rearrangement To understand theserearrangements it is important to know that the regionsinvolved consist of a single layer of supercentcial epithelialcells and several layers of deep mesenchymal cells (centgure5a) In the centrst half of gastrulation the deep mesench-ymal cells of the mesoderm (Wilson et al 1989 Wilson ampKeller 1991) and the posterior neural tissue (Keller et al1992a) undergo radial intercalation (Keller 1980) inwhich they intercalate along the radius of the embryonormal to its surface to produce a thinner array that isalso longer in the prospective anteroposterior axis(centgure 5a) Radial intercalation of several layers of deepcells to form one is also the cellular basis of the exten-sion of the neural and mesodermal tissues in sandwich



explants (Wilson amp Keller 1991 Keller et al 1992a)(centgure 5b) Radial intercalation also occurs in theanimal cap region of the embryo but there the increasein area is more uniform in all directions rather thanprimarily along one axis The mechanism of this diiexcler-ence between the animal cap and the neural and meso-dermal tissue is not known

The supercentcial epithelial cells do not participate inradial intercalation in Xenopus but spread and divide toaccommodate the larger area of the spreading deep cellpopulation (Keller 1978 1980) This is not the case inmany other species of amphibian particularly theembryos of the urodeles in which the deep cells intercalatebetween the supercentcial epithelial cells to form one layer

according to the published work (reviewed in Keller1986) The intercalation of deep mesenchymal cellsoutwards into the epithelium has never been observeddirectly nor is the mechanism known by which this mighthappen while maintaining tight junctions and an intactphysiological barrier

Following radial intercalation convergence and exten-sion occurs as the deep cells of the prospective mesodermaland neural tissue undergo mediolateral intercalation bymoving between one another along the mediolateral axis toproduce a narrower longer and usually somewhat thickerarray (Keller et al 1989a 1992b Shih amp Keller 1992ab)(centgure 5ab) The supercentcial epithelial cells accommodatethis narrowing and extension of the deep cell array by also



mediolateralintercalation


deep

(a)

(b)

deep

epith

epith

radialintercalation

radialintercalation

Figure 5 The expression of radial cell intercalation and mediolateral cell intercalation in (a) whole embryo and (b) explantsRadial intercalation consists of several layers of deep cells intercalating along the radius of the embryo (normal to the surface) toform fewer layers of greater area Mediolateral cell intercalation consists of multiple rows of deep cells intercalating along themediolateral axis to form a longer narrower array Radial intercalation occurs centrst followed by mediolateral intercalation inboth the mesodermal and neural tissues The overlying epithelial layer (epith) of the embryo is shown From Keller et al(1991a b)

intercalating mediolaterally as well as by dividing andspreading (see Keller1978)

Mediolateral intercalation is patterned in space andtime occurring progressively in both the mesodermaland neural tissue in the second half of gastrulation andthrough neurulation It originates in the anterior lateralregions of both tissues and spreads posteriorly andmedially (Shih amp Keller 1992a Domingo amp Keller 1995Elul amp Keller 2000) (see 9) Mediolateral intercalationof mesodermal cells occurs mostly at and beyond the blas-toporal lip in the post-involution region whereas radialintercalation of mesodermal cells occurs predominantlyin the pre-involution region (centgure 5ab) However thereare exceptions to this rule In late gastrulation and earlyneurulation the prospective somitic mesodermal cells ofthe lateral and ventral margins of the blastopore undergoradial intercalation followed by mediolateral intercala-tion in the post-involution region (Wilson et al 1989)Also the rapid spreading of the mesoderm^endoderm justafter its rotation and application to the roof of the blasto-coel appears to involve radial intercalation (Winklbaueramp Schulaquo rfeld 1999) Mediolateral intercalation appears tobe coincident with the stiiexclening of the post-involutionmesoderm described in frac12 9(a)(c)

(a) Mechanisms in other systemsMost examples of convergent extension appear to occur

by cell intercalation although other processes may beinvolved as well Cell intercalation contributes to conver-gent extension of the sea urchin archenteron (Ettensohn1985 Hardin amp Cheng 1986 Hardin 1989) the germband of Drosophila (Irvine amp Wieschaus 1994) and theaxial paraxial and neural structures of ascidians(Cloney 1964 Miyamoto amp Crowther 1985) centshes(Warga amp Kimmel 1990) birds (Schoenwolf amp Alvarez1989) and mammals (Sausedo amp Schoenwolf 1994) Incontrast the imaginal leg disc of Drosophila undergoesconvergent extension mostly by cell shape change(Condic et al 1991) although cell intercalation may beinvolved as well Likewise change in cell shape as well asintercalation plays a role in nematode morphogenesis(Priess amp Hirsh 1986 Williams-Masson et al 1997) Inaddition to cell intercalation orientated cell division is afactor in convergent extension of the neural plate of birds(Schoenwolf amp Alvarez 1989) the early morphogenesis ofthe centsh embryo (Concha amp Adams 1998) and Drosophilagerm band morphogenesis (Hartenstein amp Campos-Ortega 1985)

Cell growth could also potentially contribute toincreased length during convergence and extension insome systems such as the mouse which develop an earlyyolk sac placenta and thus have an external source ofnutrients that allows them to grow substantially duringgastrulation and neurulation (Snow 1977) Amphibianswhich have no large external supply of nutrients but relyinstead on intracellular stores of yolk do not show ageneral increase in cellular volume prior to feeding (seeTuft 1962) the exception being the osmotic swelling of thenotochordal cells in the tailbud stages (see Adams et al1990) Teleost centshes (Warga amp Kimmel 1990) and birds(see Schoenwolf amp Alvarez 1989) probably also do notgrow signicentcantly prior to the development of thecirculatory system

5 MECHANISMS OF MEDIOLATERAL CELL

INTERCALATION

To understand convergence and extension by cell inter-calation we must understand the motility that cells use tomove themselves between one another in the face ofinternal and external mechanical loads A description of aspecicentc cellular motility a cellular machinersquo is neededThe cellular basis of mediolateral cell intercalation willbe discussed exclusively here although similar cell moti-lity and mechanical principles may be involved in radialintercalation (see Keller 1980 Davidson amp Keller 1999)

What do cells do in order to intercalate Resolving cellbehaviour in a cell population presents several challengesFirst the Xenopus embryo consists of an outer epithelial layerand several layers of deep non-epithelial mesenchymal-like cells raising the question of the relative contributionof each to the forces producing convergent extension Forboth mesodermal (Shih amp Keller 1992b) and neural tissue(Elul et al 1997) the deep mesenchymal cell populationalone can produce convergent extension The supercentcialepithelial layer of the Xenopus embryo has not shownconvergence and extension by itself in explants but thiscould be an artefact of explantation and culture condi-tions Thus whether the epithelial layer can also contri-bute to force generation when attached to the deep regionin the embryo is not known For the present the deepregion is the only layer that has been demonstrated toconverge and extend by itself

To visualize deep cell behaviour the `open-facedexplantrsquo and the `deep cell explantrsquo were designed toexpose these deep cells to modern imaging methods(Keller et al 1985 Keller amp Danilchik 1988 Shih ampKeller 1992ab) The open-faced explant consists ofculturing one of the dorsal mesodermal^neural unitsmaking up the sandwich explant (centgure 3) alone withthe deep cells facing the microscope objective The deepcell explant consists of just the deep cells of the sameembryonic region isolated at the midgastrula stage (Shihamp Keller 1992b) In addition a solution mimicking thecomposition of the blastocoel pounduid (Danilchikrsquos solution)was developed in order to support normal deep cell beha-viour in culture (for the original version see Keller et al(1985) and Keller amp Danilchik (1988) for a more recentversion see Sater et al (1993))

Low-light pounduorescence microscopy and time-lapserecordings of deep cells labelled with pounduorescent markersrevealed the protrusive activity and cell behaviour under-lying mediolateral cell intercalation in the notochord(Keller et al 1989a) in the dorsal mesoderm before andafter formation of the notochordal somitic boundary(NSB) (Shih amp Keller 1992ab) and in the neural tissueunder several conditions of patterning (Elul et al 1997Elul amp Keller 2000) From these studies several majortypes of cell behaviours were found to underlie activesubstrate-independent cell intercalation (i) a bipolarmediolaterally orientated protrusive activity in themesoderm (ii) a boundary-mediated boundary-capturemechanism in the notochordal mesoderm (iii) a mono-polar medially directed protrusive activity in the neuraltissue and (iv) a bipolar mediolaterally orientated protru-sive activity in the neural tissue which occurs only underabnormal tissue interactions and probably represents a



latent pattern of cell motility that is not used in theembryo

(a) Mesodermal cell intercalation the bipolarmediolaterally orientated mode

The prospective dorsal mesodermal cells of Xenopusearly gastrula initially show multipolar protrusive activityin all directions (centgure 6a) At the midgastrula stagewhen mediolateral cell intercalation and the resultingconvergent extension begin the cells locally adopt astrongly bipolar morphology and show protrusive activityorientated strongly in the medial and lateral directions(centgure 6b) These protrusions appear to adhere directly toadjacent cells and result in traction on their surfaces As aresult the cells become elongated and their long axesbecome aligned parallel to the mediolateral axis of theembryo (Shih amp Keller 1992ab) (centgure 6b) The medio-laterally orientated traction of neighbouring cells on oneanother appears to pull them between one another thus

producing a forceful mediolateral intercalation that servesto narrow and elongate the tissue (centgure 6b)

(b) Mesodermal cell intercalation the monopolar boundary-capture mode

At the late midgastrula stage the NSB forms withinthis array of bipolar mediolaterally intercalating cells Atthis point a second boundary-associated mechanism ofmediolateral cell intercalation comes into play Tracings ofthe NSB backwards in time-lapse recordings (Shih ampKeller 1992a) show that the prospective boundary is anirregular line passing in zig-zag fashion between thefuture notochordal and somitic cells The boundary takesform as this line straightens and a smooth interface isformed between the notochordal and somitic cells(centgure 6c) Notochordal cells that initially form theboundary bleb for a short time and then cease protrusiveactivity at their boundary ends forming a poundat surface inthe plane of the boundary (centgure 6d ) Very few of thesecells leave the boundary once they are in contact with itand thus they are capturedrsquo by the boundary resulting inits elongation (Shih amp Keller 1992a) (centgure 6de) Mean-while the protrusive activity characteristic of the formerbipolar state continues at the inner (medial) ends of thesecells which is thought to exert traction on the interiornotochordal cells thus pulling them into the boundaryThe cells in the interior of the notochord continue tointercalate using the bipolar mode thus progressivelyconverging the notochord and bringing its interior cellsinto contact with the boundary where they also undergo`boundary capturersquo These processes continue until ulti-mately the notochord has extended greatly and convergedto less than two cells in width (centgure 6e) (see Keller et al1989a) Note that this mode of cell intercalation involvesmodicentcation of the bipolar orientated form to a mono-polar directional form of intercalation behaviour inwhich one end of the cell the boundary end is quiescentand the other continues traction on adjacent bipolar cellsthat have not yet contacted the boundary Boundarycapture was invoked originally by Jacobson and associatesas a boundary-mediated mechanism of elongating thenervous system (see Jacobson et al 1985 1986 Jacobson ampMoury 1995)

(c) MisunderstandingsSeveral points of confusion have emerged over the

years since these behaviours were centrst described (Shih ampKeller 1992ab) First the bipolar mode of mediolateralcell intercalation is an independent mechanism of inter-calation that occurs before the boundary-capturemechanism comes into play Its centrst expression is in thevegetal alignment zone (VAZ) (discussed in 9(a)) whichforms before the notochordal somitic mesodermalboundary Moreover the bipolar mode continues withinthe interior non-boundary regions of both the somiticand notochordal regions after formation of the NSBSecond evidence thus far suggests that the bipolar modeconsists of an orientated behaviour not a directionalbehaviour There is no evidence that the medial andlateral ends of the cells diiexcler in any way nor is thereevidence that they are undergoing a `directed migrationrsquoIn fact their activities appear to be balanced and equiva-lent in the medial and lateral directions and we make the



(a)

(b)

(c)

)(d

(e)

Figure 6 The cell behaviours of mediolateralcell intercalation(a) Cells are initially multipolar and show random rapidprotrusive activity in the early gastrula (b) At themid-gastrula stage they show bipolar medially and laterallyorientated protrusive activity extending protrusions thatappear to exert traction on adjacent cells As a result theyelongate in the mediolateral direction and appear to pullthemselves between one another along the mediolateral axisAt the late midgastrula stage the notochordal^somiticboundary (NSB) begins to form The decision as to where itwill form is made (white line c) but initially it has nomorphological identity Then the boundary takes form as itstraightens and the sides of the cells facing the boundary orentering the boundary spread on the boundary and ceaseprotrusive activity (light shading d ) Once they touch theboundary they rarely leave it again (boundary capture) Theinternal notochord cells continue to intercalate to form anarrower array and they progressively come into contactwith the boundaries (d e) As the invasive tractive ends of theinternal bipolar cells contact the boundary (dark shadingarrowhead) they spread on the boundary and become quies-cent (light areas) Redrawn from Keller et al (1992b)

argument below (6(g)) that this balance is necessary forconvergent extension to occur Thus the behaviour isorientated parallel to the mediolateral axis but it is notdirectional In contrast the monopolar mode of intercala-tion is a directional behaviour with a stabilized lateralend and the tractional protrusive activity pointed medi-ally Third the exercise of the bipolar mode within thenotochord will inevitably bring all the internal notochordcells into contact with the boundary without any bias inthe motility of these cells toward the boundary Fourth itis an attractive idea that the bipolar internal notochordalcells recognize a signal emanating from the boundary atsome point in their inevitable movement towards it andaccelerate their movement in that direction Howeverthere is no evidence supporting this notion If such anacceleration occurs it must occur very close to theboundary and will require a more detailed statisticalanalysis of rates of movement with distance from theboundary than has been done this far

(d) Neural cell intercalation pattern-specicentcmodesoumlthe bipolar and the monopolar modes

As noted above the neural tissue of Xenopus centrstextends and thins by radial intercalation and thenactively converges and extends using mediolateral cellintercalation (Keller et al 1992b) As with the mesodermthe cell behaviour driving radial intercalation has notbeen observed and remains unknown

Turning to the mediolateral intercalation Elul andassociates (Elul et al 1997 Elul amp Keller 2000) foundthat the deep neural cells are capable of two mechanismsof convergence and extension depending on degree ofvertical interaction with the underlying mesoderm Deepneural tissue explanted into culture from a late mid-gastrula shows no morphological or behavioural regional-ization (centgure 7ac) In contrast deep neural tissueexplanted with the underlying notochordal and somiticmesoderm shows the normal subdivision into a medialnotoplate which is that part of the neural plate overlyingthe notochord (see Jacobson amp Gordon 1976) and alateral neural plate (centgure 7bd )

The deep neural explant undergoes mediolateral cellintercalation and convergent extension using a bipolarform of cell behaviour similar to that seen in the meso-derm (centgure 7c) However it diiexclers in having a greaterfrequency of formation of new protrusions per hourcompared with mesodermal cells (Elul et al 1997) It isalso more episodic instead of elongating mediolaterallyand maintaining that shape as the mesodermal bipolarcells do the neural cells extend protrusions elongate andthen shorten again in more of an inchworm fashion (seeElul et al 1997) These diiexclerences could repoundect diiexclerencesin how forces are generated by the cells The mesodermalcells may exert relatively continuous traction using shortlocal extensions attachment and retraction of protru-sions with the retractive force repoundecting a local con-tractile event and short episodes of traction In contrastthe neural cells appear to extend further shorten moreand involve more of the whole cell in an active con-traction

The explants of deep neural tissue made with under-lying mesoderm show a dramatically regionalized cellbehaviour the regions corresponding to the notoplate



neural plate

midgastrulaembryo

notoplate

(a)

(b)

(c) )(d

(e)

Figure 7 Two types of explants were used to study neuralcell intercalation (a) The deep layer of the midgastrulaneural plate was isolated by removing both the underlyingmesoderm and the outer epithelial layer of the neural plate(b) The deep neural-over-mesoderm explant was made byexcising the dorsal mesoderm along with the neural plate andremoving both the outer neural epithelial layer and theendodermal epithelial layer from the mesoderm (c d ) Bothexplants expose the deep cells to time-lapse recordingsof cell behaviour (c) In deep neural explants convergenceand extension occurs as the cells intercalate using a bipolarmediolaterally orientated protrusive activity (d ) In the deepneural-over-mesoderm explants convergence and extensionoccur as deep neural cells show a monopolar mediallydirected protrusive activity in the lateral neural plate In thenotoplate the cells show multipolar randomly orientatedprotrusive activity (e) In these explants monopolar mediallydirected protrusions appear to exert traction on more medialcells and pull themselves between medial neighbours to form alonger narrower array At the neural plate^notoplateboundary (shaded bar) some of the neural plate cells turnposteriorly but none enters the domain of the notoplate FromElul amp Keller (2000)

the neural plate and the boundary between them (Elulamp Keller 2000) (centgure 7d ) The deep neural platecells located on both sides of the midline notoplate showa monopolar medially directed protrusive activity(centgure 7de) The monopolar medially directed protrusiveactivity appears to exert traction on adjacent cells andpull the cells between one another in a fashion similar tothat seen in the bipolar mode but in this case the motilityis directed towards the midline (centgure 7de) In contrastthe midline notoplate cells show a multipolar randomlyorientated protrusive activity (Elul amp Keller 2000)(centgure 7de)

The boundary between the notoplate and neural plate isnot a physically well-decentned border like that seen betweenthe notochordal and somitic mesoderm (see Shih amp Keller1992a) However cells do not cross the boundary in eitherdirection after stage 11 the midgastrula stage in explants(Elul amp Keller 2000) or in the embryo (A Edlund andR Keller unpublished data) Cells of the neural platethat approach the notoplate^neural plate boundary eitherstop or they turn posteriorly and show protrusive activityin that direction (centgure 7e) It is possible that this be-haviour results in a shearing force of the neural plateagainst the notoplate at the boundary between themtending to pull the neural plate posteriorly but there is noevidence for or against this notion

(e) The notoplate attachmentdirectional spreading and towing

The notoplate is an isolated cell population afterstage 11 the midgastrula stage when it nominallyconsists of a deep and a supercentcial layer Deep cells willnot cross the notoplate^neural plate boundary in explants(Elul amp Keller 2000) or in embryos after this stage (AEdlund and R Keller unpublished data) In dissectionsof living embryos the deep cells of the notoplate aretightly joined to the underlying notochord from stage 11to 115 (late midgastrula) onwards whereas the lateralneural plate is much more easily removed from the under-lying somitic mesoderm (R Keller unpublished data) Asthe deep layer of cells of the two-layered neural plateundergoes radial intercalation to form the single layer ofcells making up the neural tube nearly the entire poundoorplate is formed from these tightly attached deep cells ofthe notoplate (A Edlund and R Keller unpublisheddata) The notoplate cells show multipolar protrusiveactivity in all directions It is dicurrencult to see how the noto-plate would actively converge and extend using this typeof isotropic protrusive activity When challenged withnotochord and somitic tissue as substrates the deep noto-plate cells as well as the intact tissue spread on the noto-chord (R Keller and A Edlund unpublished data) Thusone possibility is that in the embryo the notoplate cellsmay spread posteriorly on the notochord as the notochordelongates in that direction and thus be towed along bythe underlying notochord The neural plate attached tothe lateral aspect of the notoplate would then in turn betowed along by the notoplate This towing would be inaddition to its endogenous convergent extension driven bythe monopolar mode of intercalation described above(5(d)) These are speculations The contribution ofpassive towing of the neural tissue by virtue of its attach-ment to the extending notochord the contribution of

active migration of the notoplate posteriorly on theextending notochord and the contribution of activeextension of the notoplate^neural plate should be deter-mined as part of a comprehensive mechanism of neuralextension

(f) Diiexclerences between the bipolarand monopolar modes of neural cell intercalationThe medially directed monopolar mode of neural cell

intercalation diiexclers in several ways from the bipolarmediolaterally orientated mode First the monopolarmode results in a conservative pattern of cell intercalationwith mostly nearest neighbours intercalating with oneanother at any one time In contrast the bipolar moderesults in a promiscuous pattern of intercalation in whichcells mix with neighbours further acenteld and more often(Elul amp Keller 2000) Second the bipolar mode seemsless ecurrencient than the monopolar mode in producingconvergent extension Extension is usually less in the deepneural explant which uses the bipolar mode than in thedeep neural tissue plus mesoderm explant which uses themonopolar mode This diiexclerence may be due to inherentproperties of the two modes of neural cell intercalationIn the bipolar mode the neighbour changes appear morechaotic and involve relatively rapid long-distance move-ments of cells centrst in one direction and then the otheralong the mediolateral axis often resulting in exchange ofplaces along the mediolateral axis without producing netconvergence and extension We think that the bipolarmode depends on balanced traction which may make itinherently a less stable and ecurrencient mechanism of inter-calation than the monopolar mode which of course doesnot require balanced traction (see 6(a)) On the otherhand the bipolar mode as it is expressed in the meso-derm appears to be more ecurrencient at producing conver-gent extension than the neural bipolar mode Oneexplantation for this diiexclerence may be that the neuralbipolar mechanism is a latent mechanism needed as anunderpinning for the decentnitive monopolar mode butnever used alone in normal development (see below) andnot capable of producing ecurrencient convergent extension byitself

Alternatively or perhaps in addition the deep neural-over-mesoderm explants may extend better than the deepneural explants because of the possible towing of the noto-plate by the underlying notochord or the possible activeposterior spreading of the notoplate referred to aboveResolving these issues depends on being able to design anexplant in which the mesoderm is allowed to induce themonopolar intercalation behaviour but then is removedso that the notochord cannot act as a substrate or as atowing device This is now possible

Finally the episodic bipolar mediolaterally orientatedmode of neural cell intercalation is probably not used innormal development of the embryo This mode of inter-calation behaviour occurs only in explants developingunder the transient planar and vertical signals from theorganizer tissue that function during the centrst half ofgastrulation (Elul et al 1997 Poznanski et al 1997 Elul ampKeller 1999) The monopolar directional protrusiveactivity emerges only with persistent vertical interactionsbetween mesoderm and the overlying neural tissue Thusin the embryo the bipolar mode is either not used or



used very briepoundy during midgastrulation before themonopolar mode comes into play

Why then is the bipolar mode of neural cell intercala-tion expressed at all under the inpounduence of transient inter-actions in the early gastrula stage If it is used little ornot at all in the embryo why has the subcellularmachinery essential for its expression persisted and whydoes it emerge as a functioning mode of cell intercalationwhen deep neural tissue develops under abnormal signal-ling Perhaps the bipolar mode is preserved becauseexpression of the monopolar mode depends on compo-nents of the bipolar machinery If so the bipolar modemay be a latent mechanism never used for its own sakebut retained only as a prerequisite for the monopolarmode Under the simplicented planar and transient signal-ling regime under which the deep cell explant developsthe bipolar neural mode of intercalation re-emerges as anindependent mechanism We view the normal mechan-isms of morphogenesis as products of much evolutionaryhistory with parts of some earlier versions of morpho-genic machines used as underpinnings for the extantmechanisms Manipulating signalling pathways and tissueinteractions in embryos may uncover these latentmachines as separate mechanisms which perhaps nolonger work all that well by themselves

(g) Local motility global displacement the ecurrenciency of cell intercalation

as a mechanism of convergent extensionConvergent extension by cell intercalation is a highly

ecurrencient mechanism of translating local cell traction intoglobal displacement of cells There is the small localdisplacement of a given cell with respect to its neighboursdue to its own motility and there is the much largerglobal displacement of the cell that occurs by virtue of thecellrsquos membership in the aggregate The integration ofsmall local movements over the tissue makes the lattermovements very large If each cell moves an average ofhalf a cell diameter with respect to its neighbours in themediolateral direction the length of the tissue will bedoubled and its width halved For this reason the centrstdirect visualization in time-lapse recordings of cell inter-calation as a mechanism of convergent extension wassubtle looking locally not much seemed to be happeningbut the displacement of the cells overall was great (Kelleret al 1989a Shih amp Keller 1992ab) Note that in theXenopus mesoderm cell shape change is counter-productive in producing convergent extension the cellsare elongated in the dimension of convergence andshortened in the dimension of extension Neverthelessmultiple rounds of cell intercalation are very eiexclective andecurrencient in elongating and narrowing the axial and para-xial tissues

Finally a cellrsquos apparent speed and direction of move-ment is determined by the reference point used in visua-lizing converging and extending tissues If in the case ofthe extending neural tissue for example the position ofthe forebrain is held constant cells progressively moreposterior in the axis will appear to move posteriorly atgreater rates although the local motility of the cells maybe exactly the same as those cells further anteriorly thatappear to be moving more slowly If the posterior end ofthe neural plate is used as the reference point cells

appear to move anteriorly and the rate of displacementin regard to anteroposterior position is reversed Theseconsiderations should be taken in to account whendescribing examples of convergent extension

6 A CELL plusmn CELL TRACTION MODEL

OF CELL INTERCALATION

Here we hope to identify the important properties ofcells and the interactions between cells that will allowthem to intercalate forcefully and yet maintain a stiiexclarray capable of generating the pushing forces describedabove (3(a)) Concepts about how the motility and theadhesion of cells are related to a force-producing conver-gent extension on the part of the tissue as a whole arepoorly developed To address this decentciency we willdescribe a model of cell intercalation specifying para-meters of cell adhesion cell traction and cell contactbehaviour that are based on and account for experi-mental observations of cell intercalation We consider themechanisms proposed to be working hypotheses to guidefurther observations and experiments

It seems wise at this point to consider both cell^celland cell^matrix-mediated intercalation mechanismsbecause both seemed to be involved Supporting the cell^cell traction model the intercalating cells appear to beexerting traction directly on the surfaces of neighbouringcells in both the mesoderm (Shih amp Keller 1992ab) andin the neural plate (Elul et al 1997 Elul amp Keller 2000)In addition there is evidence that specicentc protocadherinsin the axial (Yamamoto et al 1998) and paraxial meso-derm (Kim et al 1998) may be involved in convergentextension which in its simplest interpretation implies acell^cell-mediated process In scanning electron micro-graphs (SEMs) connecting protrusions seem to be applieddirectly to adjacent cells without intervening matrix(centgure 9 Keller et al 1992b) However there is alsoevidence that cell^matrix interactions are involved inNSB-mediated intercalation (P Skoglund unpublisheddata see 8) and perhaps in other regions of the meso-derm as well

(a) Cell^cell traction model requires regionalizedfunction tractive protrusions at the ends of the cells

and the cell body serving as movable substrataBoth the monopolar and bipolar modes of cell inter-

calation require the cell body to serve as a substratumthus providing stable anchorage for the tractive protru-sions that pull the cells between one another (centgure 8ab)Moreover each cell must restrict its tractive protrusiveactivity to the medial and lateral ends in the case of thebipolar mode (centgure 8a) and to the medial ends in thecase of the monopolar mode (centgure 8b) Thus each cellhas a dual function one or both ends of the cell exertstraction on a substratum and at the same time thesurfaces of the cell bodies especially the apposed anteriorand posterior surfaces serve as `movable substratarsquoIntercellular traction must be concentned to the end(s) ofthe cells When `tractoringrsquo is spread over the cell bodysuch as in the cortical tractor model of cell movementdiscussed in frac12 6(f ) the substrate function of the cellbody is pre-empted no stable substratum is availableand the cell^cell traction model will not work Based



on time-lapse recordings of mesodermal cell behaviour(Shih amp Keller 1992ab) the tractive force appears to begenerated locally These cells show short advances andretractions of the tractive protrusions at the medial andlateral ends tugging on and distorting neighbouring cellsin the process In the neural cell intercalation behaviourwhich is made up of larger more episodic extension andretraction cycles traction may involve contraction of alarger part of the cell body These cells show a greaterchange in aspect ratio in a cyclic fashion (see Elul et al1997) suggesting that the traction-generating machinerymay involve more of the cell body Nevertheless the

traction seems to be concentned largely to the protrusions atone or both ends rather than along the entire cell bodybased on distortions of adjacent cells in response toprotrusive activity of a given cell

(b) The cell body as a substratum an attachmentfor tractive protrusions that will resist tangential

deformation (shear with respect to the cell surface)As a substrate the cell surface must oiexcler resistance to

deformation parallel to the cell surface (resistance toshear) under the traction of adjacent cells (centgure 8c) Thestiiexclness that decentnes the substrate function of the cell



old

old

new

new

(a) (i)

(i)

(ii)

(ii)

anchor tocortex transcellular

anchor

(iv)

(iv)

(v)

(v)

(iii)

(iii)(b)

(c)

)(d

)( f(e)

Figure 8 The cell^cell traction model of mediolateral cell intercalation for (a) the bipolar and (b) the monopolar modes ofintercalation The elements of this model are (i) large tractive protrusions (light shading a b details in c) at both the medial andlateral ends of the cells in the bipolar mode of intercalation (a) or at the medial ends of the cells in the monopolar mode(b) (ii) many small `stiiexclening adhesionsrsquo across the cell body (dark shading in a b details in d ) (iii) a stiiexcl cortex of the cellbody which serves as a substrate by oiexclering adhesions to tractive protrusions that resist tangential displacement (c) or(iv) cytoskeletal elements spanning the cell between tractive adhesions (a b) and (v) an elastic or contractile element of thecytoskeleton spanning the length of the cell The large tractive protrusions should induce disassembly of the stiiexclening adhesions(e) but pass one another without interacting ( f )

body could be a permanent property of the cell cortex orthis property could be contact induced specicentcally bycontact of a tractive protrusion on a cell body Such resis-tance to shear must involve anchorage of the cell adhesionmolecules to the cytoskeleton We envisage that therelevant cytoskeleton could be a cortical cytoskeletonperhaps permanently in place or a transcellular cyto-skeleton induced by the tractive adhesions as theyadvance (centgure 8ab) We do not know how the cyto-skeleton of the mesodermal and neural cells is organizedin the three-dimensional tissue array while they are inter-calating Visualizing the cytoskeleton in these large yolkycells is dicurrencult but such studies should be done

Arguing from results in other systems the resistance todeformation on the part of the cell body could be inducedby the forces exerted on the cell body by the tractiveprotrusions The attachment of integrin matrix receptorsto the cytoskeleton is strengthened by resistance oiexclered tothe movement of a matrix-coated bead (Choquet et al1997) The identity of the receptor^ligand system involvedin the adhesion of the putative tractive protrusions to thecell body are not known and therefore it is not known ifsuch a response is likely among intercalating cellsHowever time-lapse recordings of pairs of pounduorescentlylabelled intercalating cells often show repeated attach-ments and tugging on one another with the substrate cellbody seemingly oiexclering greater resistance on repeatedtuggings (R Keller unpublished data) These observa-tions are anecdotal and should be pursued in a formalway perhaps using two-colour pounduorescent time-lapseimaging to distinguish better the deformation of the cellexerting traction and the one oiexclering its cell body assubstrate

A better way to approach the problem of both tractiveforce and resistance of the substratum to shear would beto test both using mechanical manipulation of beadscoated with the appropriate ligands However an impor-tant missing piece of information for analyses of this typeis the identity of the cell adhesion system involved in thetraction events This information is needed because theresponse of the system to mechanical stress is likely to bespecicentc to the adhesion system The analysis of mechan-ical aspects of regulating adhesion molecule or matrixreceptor location function and attachment to the cyto-skeleton promises to be complex (see Evans et al 1991 Leeet al 1993 Choquet et al 1997) and perhaps all the moreinteresting in the specialized mechanical environment ofthe intercalating cells

A common and important element in our model ofboth monopolar and bipolar modes of intercalation is thateach cell spends a high proportion of time functioning asa tension-bearing element in a chain of tension-bearingelements The tensile elements forming the cortex or span-ning the cell cytoplasm must link the traction sites ofother cells on the body of a given cell to that cellrsquos owntractive protrusions or to traction sites of another cellpulling from the opposite direction (centgure 8ab) A novelfeature of this chain is that the links (the cells) can movethus shortening the chain but the movement of the linksmust not break the chain In the case of the bipolarmode this means that the cells must show continuousbalanced traction on the medial and lateral ends If thetraction of a cell becomes substantially unbalanced it will

move in the direction of the strongest traction and thechain is essentially broken And if many cells becomeunbalanced randomly in the medial or lateral directionthe cells will simply exchange places rather than shortenthe chain In the case of the monopolar mode tractionmust be strongly and constantly biased in the medialdirection Again if cells show mixed medially directedand then laterally directed traction a medially or laterallydirected `migrationrsquo will occur again resulting in exchangeof places without extension-producing intercalation

(c) The paradox of stiiexclness and pushing in thepresence of neighbour exchange requires a special

type of adhesion the stiiexclening adhesionrsquoAs noted above (3(a)) the dorsal mesodermal tissue is

stiiexcler in the anteroposterior axis by a factor of three tofour during convergent extension (Moore et al 1995)which allows it to push with a force of about 05 m N(Moore 1992) Stiiexclness is necessary because if a tissue isto push it must resist bending under the load imposed byinternal resistance and the resistance of surroundingtissues To form a stiiexcl tissue the cells composing thetissue must resist deformation and they must resist re-arrangement under an external load However at thesame time mediolateral cell intercalation must occur inorder for the tissue to extend

How then do cells of a tissue exchange neighbourswhile the tissue remains stiiexcl enough to support acompression load The cells must be attached tightly toone another in such a way that they resist exchange ofneighbours under external loads but can exchange neigh-bours as a result of the internally generated tractiveforces proposed above This requires a second specialtype of cell^cell adhesion the stiiexclening adhesion

Stiiexclening adhesions are predicted to lie along the ante-rior and posterior surfaces of the cell bodies (centgure 8ab)where they join cells together fore and aft and resistseparation normal to the surface of the cell (centgure 8d )However these stiiexclening adhesions must oiexcler little resis-tance to tangential movements (parallel to the cellsurface) in order to allow the cells to slide by one anotherduring cell intercalation (centgure 8d ) Moreover as a celladvances between two others and exchanges neighboursit must assemble new stiiexclening adhesions with its newneighbours Thus at the ends of the cells there should bezones of regulated assembly (or disassembly) of the stif-fening adhesions Finally the stiiexclening adhesions mustalso allow the tractive protrusions to pass through themand thus they should show a contact-induced disassemblyon approach or on contact by the large advancing trac-tive protrusions that pull cells between one another(centgure 8e) It is important to emphasize that we do notexpect this disassembly to be due to mechanical forcespulling the adhesions apart but a contact-mediatedcontrolled process of disassembly

(d) Rules of contact-mediated regulationof protrusive activity and adhesion

The invasive tractive protrusions should not showmutual contact inhibition of advance or contact-inducedretraction when meeting head-on but should be able topass by one another (centgure 8f ) Likewise they should notshow contact inhibition by the cellsrsquo surfaces serving as



substrata if so they could not advance across thesesurfaces Finally the tractive protrusions should be ableto transfer their adhesion from one cell to the next as theopportunity arises (centgure 8f )

Is there evidence for these stiiexclening adhesions Deepmesodermal cells have large numbers of small centliformprotrusions attaching the long anterior and posteriorsurfaces of adjacent cells to one another as seen in SEM(Keller et al 1992b) (centgure 9ab) These intercellularcontacts may not exist as protrusions in the embryo Inthe embryo the cells appear to be tightly packed togetherand low-light pounduorescence microscopy of labelled cellsreveals few such centliform protrusions on these surfaces(Shih amp Keller 1992a) Instead these protrusions mayrepoundect the presence of small local adhesions between

tightly packed cells which are then pulled out intocentliform structures (centgure 9b) during the shrinkage asso-ciated with SEM which can be 10 or more in thissystem (see Keller amp Schoenwolf 1977) These numerouscontact points on the anterior and posterior surfaces ofthese elongated cells could serve as the stiiexclening ad-hesions between cells regardless of whether or not theyexist as centliform protrusions

The large lamelliform protrusions at the medial andlateral ends of bipolar mesodermal cells seen in low-light pounduorescence time-lapse recordings are also seen inSEM (centgure 9ab) These we believe to be the tractiveprotrusions

(e) Common criticisms of the cell traction^cellsubstrate mechanism

There are a number of objections that are commonlyraised against the cell traction^cell substrate mechanismof cell intercalation described here One of these is thatthe two halves of the substrate cell in the bipolar modelare somehow diiexclerent in polarity one end pointing oneway and the other end pointing the other way and thatsomehow the tractive protrusions will fail to producefurther movement when they pass beyond the midpoint ofan elongated cell However no such polarity is implied inthe model or observed in cell behaviour The modelproposes that the cell body is an unpolarized substratumuniform from end to end (centgure 8ab) The important



Figure 9 Scanning electron micrographs at (a) low and (b)high magnicentcation show large lamelliform tractive protrusionsat the medial and lateral ends of elongated and aligned meso-dermal cells (arrowhead a large arrowheads b) andnumerous small adhesions mediated by centlopodia at the ante-rior and posterior surfaces of these cells (small arrowheads b)From Keller et al (1989b)

(b)

(a)

Figure 10 The cortical tractor model of cell motility resultsin parallel tractoring in the same direction (top box in aright box in b) in both (a) bipolar and (b) monopolar modesof cell intercalation Opposing directions of tractoring occuronly at partial overlap of bipolar cells (lower box in a) Thearrows indicate the pattern of tractoring of the cell cortexbipolar cells are assumed to tractor from both endsto the middle

element in the model is that the cells maintain a contin-uous structural organization to transmit tension over partor most of its mediolateral length in either direction asthe tractive protrusion of another cell crawls along itssurface from either end to the other there is no reason tosuppose that this function involves polarization(centgure 8ab) This objection arises from assuming that thetraction of a bipolar cell occurs over the whole cortex ofthe cell and in opposite directions from both endstowards the middle (centgure 10a) As discussed below(6(f )) this type of cortical tractoring over large areas ofthe cell presents problems for both the bipolar and mono-polar modes of intercalation But cortical tractoring is notseen in time-lapse recordings instead the traction seemslocalized to the protrusions at the ends of the cells (Elulamp Keller 2000) (see 6(g))

A second objection is that some unusual circumstancewill arise when a tractive protrusion reaches the end of itssubstrate cell and transfers to another cell This objectionseems to arise from the assumption that there will be ormay be a critical reversal of polarity at this transferpoint But again the cell^cell traction model proposedhere does not imply such a polarity and there is noevidence that these cells are polarized in such a fashionAs a tractive protrusion reaches the end of its substratecell it has the opportunity to transfer to an adjacent celland continue exerting traction on the new substrate cell(centgure 8f ) We do not envisage the substratum function ofthe cell body as oiexclering less resistance to traction in onedirection or the other over the whole cell body Indeedreview of cell behaviour in low-light pounduorescence time-lapse centlms of neural and mesodermal cells intercalatingdoes not betray any obvious diiexclerences in protrusivebehaviour as a cell advances along another cell (R Kellerunpublished data) In terms of the model the chains oftension are not broken when a cell transfers its tractiveadhesion to another substrate cell but is remodelled suchthat a given medial segment might be hooked up with anew and diiexclerent lateral segment

A third objection is that when cells come to lie paralleland at the same mediolateral position they are using oneanother for traction and thus cannot move with respect toone another But in fact time-lapse recordings show thatany two such cells are available as substrates for cellsmedial and lateral to them and in turn can use these cellsas substrates because of the partial interdigitation of theirmedial and lateral ends Therefore the question of theirmutual traction is largely irrelevant Given the maximumdensity packing of the cells and the partial interdigitationof these fusiform cells tractive protrusions on a given cellhave many opportunities to jump over to other substratecell bodies when reaching the limits of the currentsubstrate cell And again traction appears localized tothe end or ends of the cells rather than spread over thecell body

(f) The cortical tractorrsquo model of cell motilityas a counter-example

The `cortical tractorrsquo model of cell motility provides agood counter-example to the localized tractionmechanism of cell intercalation described here because itdoes not restrict traction to localized protrusions at theends of cells The cortical tractor model proposes that

cells move forward by tractoring their cortex backwardsover much or all of the cell body from the leading edge ofthe cell (centgure 10b) (Jacobson et al 1985 1986 centg 8)Jacobson and co-workers have used the cortical tractoringmodel of cell motility to explain the rolling of the neuralplate into a trough and to explain the elongation of thenotoplate^neural plate boundary by cell intercalation andboundary capture in the urodelean amphibian (Jacobsonet al 1986) In this application the notoplate cells at theboundary with the neural plate show cortical tractoringtoward the boundary thus pulling interior notoplate cellsinto the boundary where they in turn adopt thisbehaviour (Jacobson et al 1986 centg 8)

Although successful in this application to morpho-genesis of the urodele neural plate invoking the corticaltractor across the whole centeld of Xenopus deep neural cellsas a mechanism of moving the cells towards the midlineillustrates the need for localized traction traction res-tricted to the ends of the cells Applying the corticaltractor model to the monopolar mode it appears that allthe tractoring surfaces would abut one another and attemptto use one another as a source of traction (centgure 10b)And of course they would fail to move because foreach cell the substrate is moving in the same direction asthe applied traction Another cell could not enterbetween any two other cells likewise because of paralleltractoring Applied to the bipolar mode the corticaltractor might actually work well initially in the stages ofpartial overlap of two intercalating cells because theapproaching ends of bipolar cells would be of oppositetractoring directions But tractoring would becomeprogressively less eiexclective as the cells near completeoverlap when a greater part of their adjacent surfacesdisplay parallel tractoring (centgure 10b) Again the impor-tant diiexclerence between the cortical tractor mode of cellmotility and the localized traction mode of cell motilitydescribed here is that the entire surface of the cell isserving as a tractor and as a substrate in the corticaltractor model whereas these two functions are separatedspatially on the cell in the localized traction model

(g) Cell behaviour observed during intercalationThe localized cell^cell traction mechanism centts the

observations of the behaviour of intercalating cells in boththe mesoderm and neural tissue First the observedprotrusive activity of intercalating cells is local andconcentned to the ends of the cells both ends in the case ofbipolar mode in both mesodermal (Shih amp Keller 1992a)and neural tissue (Elul et al 1997 Elul amp Keller 2000)and the medial end in the case of the monopolar neuralmode (Elul amp Keller 2000) In the latter case the cellsoften have small lateral protrusions perhaps representingretraction centbres oiexcl the trailing edge of the cells There isthus far no evidence of global cortical tractoring amongstintercalating mesodermal or neural cells in Xenopus Whenboth substrate and tractive cell are labelled and thus canbe seen in recordings the putative tractive protrusions atthe ends of the cells appear to tug on and distort the adja-cent cells implying that these protrusions are in facttractive Similar protrusions seen on various types of cellscultured on deformable rubber sheets are associated withtraction visualized by deformation of the sheets (Harris1980) Time-lapse recordings of intercalating mesodermal



cells in explants suggests that the most convergent exten-sion occurs when there is balanced bipolar protrusiveactivity When these cells appear to be unbalanced in theirtraction they `migratersquo medially or laterally as indivi-duals exchanging places and little convergent extensionis produced (R Keller unpublished data) Mediolaterallyorientated rows of cells at specicentc anteroposterior levelsare occasionally observed to slide medially and laterallyas a unit an observation consistent with the idea thatintercalating cells form mediolaterally orientated chainsunder tension that span the mediolateral axis and serve asthe organizational unit for producing convergent cellintercalation Finally the tractive protrusions do not showcontact inhibition of one another but instead pass by oneanother when meeting head-on and attach to the cellbodies behind one another The predicted contact induceddisassembly of the stiiexclening cell adhesions along the ante-rior and posterior sides of the cells has not be observeddirectly but this may be because of the close apposition ofthe cells The fact that cells seen intercalating in time-lapse recordings are found to have large numbers oflateral contacts in SEM suggests that these contacts oiexclerlittle resistance to shearing of cells past one another

(h) Cell polarity and cell intercalation behaviourare context dependent

Eiexclorts to dissociate the intercalating mesodermal cellsand culture them as individuals on decentned substrateshave failed thus far to produce very much useful informa-tion about the cell intercalation behaviour Whencultured on centbronectin these cells attach spread andbecome transiently multipolar then bipolar and backagain to multipolar displaying none of the highly orga-nized behaviour seen in the explants (R Keller un-published data) Immediate reaggregation of dissociateddorsal mesodermal cells failed to produce convergentextension (R Keller unpublished data) a result alsoobtained long ago by Holtfreter (1944) Whatever thebasis of the bipolar and monopolar intercalation beha-viours it is contextual and requires maintenance of inter-cellular communication either by contact or by signallingover longer distances It also does not appear to be self-organizing in a dissociated scrambled population of cellsThe overlying epithelial layer of the organizer appears tohave a role in setting up the organization necessary forcell intercalation in the deep region in the centrst half ofgastrulation (Shih amp Keller 1992c) but we do not knowthe mechanism of this eiexclect

The polarization of cell morphology and protrusiveactivity in the bipolar mode the monopolar mode and inthe boundary-capture mode is very strong when observedin the intact explant In the context of the explant thesecells are large volume rectangular containers of yolkplatelets in which it has been dicurrencult to resolve details ofcytoskeletal organization that underlies this dramaticpolarity of behaviour and morphology Also little isknown about the motility regulating factors (see Hall ampNobes this issue) that are expected to control the typeand distribution of protrusive activity We must developmethods to study the cytoskeleton and its regulation inthe context of the explant or resolve the contextual cuesthat will allow expression of intercalation behaviour in asimpler context where the cytoskeleton can be studied

more easily Or we must study these aspects in one ormore of the systems mentioned above that shows similarintercalation behaviour but is more amenable to visualiza-tion and analysis of the cytoskeleton

7 FUNCTION OF CELL ADHESION MOLECULES

IN CELL INTERCALATION

(a) CadherinsAre there specicentc cell^cell adhesion systems associated

with either the tractive protrusions or the putative stiiexclen-ing adhesions holding the cells together or do cell adhe-sion molecules serve only a general function of holdingthe cells of the embryo together with the specicentcity of thepolarized cell behaviour lying elsewhere A number oftypes of cadherins calcium-dependent cell adhesionmolecules are found in early Xenopus developmentincluding maternally derived cadherins (EP- or C-cadherinand XB- or U-cadherin) E-cadherin which comes on atgastrulation in the ectoderm and N-cadherin whichreplaces E-cadherin in the neural ectoderm (see Levineet al 1994 Heasman et al 1994 Herzberg et al 1991)Another cadherin Xenopus cadherin-11 (Xcadherin-11) isexpressed at the gastrula stage in mesenchymal cells(Hadeball et al 1998) One of these EP- or C-cadherin isthe major cadherin expressed in the early embryo and itis the primary cell adhesion molecule binding the cells ofblastula and early gastrula together (Heasman et al 1994Lee amp Gumbiner 1995) It appears to be downregulatedin functional activity when convergent extension isinduced experimentally in animal caps suggesting thatdecreased cadherin-mediated cell^cell adhesion is asso-ciated with increased morphogenic movements (Brieheramp Gumbiner 1994) Expression of a dominant inhibitoryform of this cadherin consisting of the deletion of thecytoplasmic tail by RNA injections blocked involutionand closure of the blastopore when expressed in the dorsalsector of the marginal zone but not when expressed in theanimal cap or ventral sector of the marginal zone (Lee ampGumbiner 1995) In this study expression of wild-typeC-cadherin rescued the eiexclect of the dominant negativeform when co-expressed and generated gastrulationdefects when expressed alone somewhat diiexclerent fromand more severe than those of the dominant negativeform A monoclonal antibody to the ectodomain ofC-cadherin AA5 is a strong activator of C-cadherin-mediated adhesion of early frog blastomeres and inhibitsthe elongation of activin-treated animal caps arguingthat a decrease in C-cadherin-mediated adhesion isnecessary for convergent extension (Zhong et al 1999)These studies argue that C-cadherin is playing a role inconvergence and extension but it is not clear specicentcallywhat that role might be It appears to be downregulatedin activity in in vitro assays but that does not necessarilymean that the same occurs in the context of the inter-calating cells of the axis in that context it may beregulated locally and it is only a default lower-adhesionstate that is expressed in disassembled axial tissues Thegeneral problem with arguing that downregulation ofadhesion is a major element in mesodermal convergentextension is that the cells must stick together in order toform a stiiexcl array that is able to push in the fashiondescribed in 3) Therefore if specicentc cadherin-mediated



adhesion is decreased cell^cell adhesion must beincreased by a diiexclerent cadherin by a diiexclerent cell adhe-sion molecule or by cell matrix adhesion in order to forma stiiexcl extension

N-cadherin is expressed in the neural tissue ofXenopus during the period of its convergent extension(Detrick et al 1990) Expression of N-cadherin along withb-galactosidase to mark the expressing cells suggestedthat N-cadherin overexpression prevents the normalamount of cell mixing Expression of N-cadherin alsoaiexclected neural tube formation ranging from its failure toclose to aiexclects on its cellular arrangement (Detrick et al1990) It is not known whether these phenotypes weredue to eiexclects on convergent extension and the type of cellintercalation described above for the neural tissue orother prior events in early development Rapid and exten-sive radial and mediolateral intercalations are involved inreorganization of the dorsal neural tube during its closure(Davidson amp Keller 1999) perhaps these intercalationswere aiexclected in these experiments Interestingly anothercell adhesion molecule the calcium-independent neuralcell adhesion molecule N-CAM which is normallyexpressed in the neural tube has no apparent eiexclect onearly neural development when overexpressed (Kintner1988) Expression of extracellularly truncated of E- andEP-cadherins in Xenopus (Broders amp Thiery 1995) andPleurodeles (Delarue et al 1998) embryos disrupted theconvergent extension movements by aiexclecting the patternsof cell intercalation RNA encoding cadherin-11 isexpressed in the animal cap and marginal zone of theXenopus gastrula and later in the sclerotome lateral platemesoderm and neural crest but not the converging andextending tissues of notochordal and somitic mesoderm(Hadeball et al 1998) Overexpression of Xenopuscadherin-11 results in failure of convergent extension inexplants abnormalities of the neural tube and notochordand head in the embryo (Hadeball et al 1998)

These experiments suggest an involvement of cadherinsat some level in convergent extension but do not charac-terize their role The challenge is to determine which celladhesion molecules have general roles in convergentextension to determine which if any have specicentc rolesin convergent extension and to characterize these roles interms of cell and tissue biomechanics

(b) The role of protocadherinsTwo protocadherins paraxial protocadherin (PAPC)

and axial protocadherin (AXPC) (Kim et al 1989 Yama-moto et al 1998) have been implicated more specicentcallyin convergent extension In Xenopus PAPC-encodingRNA is centrst expressed in all posterior mesoderm of thelate blastula and early gastrula and when the mesoderminvolutes it disappears from the notochord but persists inthe somitic mesoderm In contrast AXPC-encodingRNA is expressed in the notochord as PAPC RNA dis-appears in this tissue AXPC and PAPC show mutuallyexclusive specicentc cell adhesion properties in dissociationreaggregation and sorting out studies Expression of asecreted extracellular domain of PAPC has a dominantnegative eiexclect and disrupts convergence of paraxialmesoderm into the normal somite centles of Xenopusmimicking to some degree the defects of somitic meso-derm convergence seen in the eiexclect of the mutation

spadetail in the zebracentsh (Kimmel et al 1989) The domi-nant negative PAPC disrupts elongation of animal capexplants induced with activin but the normal PAPCpotentiates expression of convergent extension movementsand adoption of the elongate shape typical of the interca-lating mesodermal cells in animal caps treated with sub-threshold levels of activin (Kim et al 1998)

Surprisingly truncations of most of the intracellulardomain did not have a dominant inhibitory eiexclect aswould be predicted from the behaviour of classicalcadherins of this type Instead the truncated formshowed greater positive eiexclects on cell adhesion than thenormal molecules suggesting that the cytoplasmic tails ofthese molecules may function in decreasing adhesion(Kim et al 1998) These results suggest involvement inregulation of cell adhesion during convergent extensionand perhaps regulation of the expression of cell intercala-tion behaviour Kim and associates suggest that a specicentcregion of the intracellular domain of PAPC may associatewith an unidenticented protein to lower cell adhesionmaking possible the transient cell associations necessaryduring cell intercalation Could PAPC function in the`stiiexclening adhesionsrsquo postulated above by promotinghigh adhesion of the anterior and posterior surfaces of theintercalating mesodermal cells but yet disassemble theseadhesions in a regulated way as an intercalating cellleaves one neighbour and reassemble an adhesion as anew one is acquired

(c) PossibilitiesIn principle it is now possible to test what specicentc role

a cadherin a protocadherin a cytoskeletal element orregulator of protrusive activity might have in the contextof the cell^cell traction model described above (6)using time-lapse recordings of cell behaviour and bio-mechanical measurements together to evaluate theprotrusive activity the stiiexclness of the tissue and responseof the cells to mechanical manipulation In practiceapplication of these sorts of analyses requires workingwith the explants which are dicurrencult to make and visua-lizing cell behaviour with low-light time-lapse pounduores-cence microscopy which requires sophisticated equipmentand knowledge of imaging And biomechanical analysesare likewise dicurrencult Nevertheless with the appropriateapplication of these methods it will soon be possible todetermine not only whether a molecule is necessary forconvergent extension but what specicentc function it plays

8 CELL plusmnMATRIX MODEL OF CELL INTERCALATION

An alternative to or perhaps an addition to the cell^cell traction model of cell intercalation is a cell traction-on-extracellular matrix model for active cell inter-calation Formally centbrils of an extracellular matrix couldoccupy the spaces between the intercalating cells andserve as a substratum for an orientated or directed migra-tion (centgure 11a) To produce cell intercalation the matrix-mediated migration would have to result in wedging ofthe cells between one another to make a longer arrayalong the anteroposterior axis The matrix might beinterstitial as shown here or as a mat on one or bothsurfaces of the intercalating cell array Of course in eithercase the matrix would have to be deformable or capable



of being remodelled in the course of convergent extensionin order to allow the tissue to lengthen and narrow Aninterstitial matrix or a mat of matrix could serve as asubstratum for a monopolar medially directed cellcrawling of the type seen in the neural tissue but it is notobvious how such a matrix would serve as a substratumfor the bipolar mechanism of cell intercalation seen in themesoderm and in neural tissue under minimal inter-actions with mesoderm Formally the model of thebipolar mode of intercalation seems to work only as acell^cell traction model Finally an extracellular matrixcould be involved in the `boundary capturersquo described in 5(b)) for the NSB in this model cells would crawl onthe matrix perhaps directionally until reaching the

boundary where they would spread on the matrix in theboundary and then cease their motility (centgure 11b)

What is the evidence for extracellular matrix functionin convergent extension An external substratum such asthe overlying blastocoel roof is not essential for conver-gent extension of the mesodermal tissue of Xenopus (seeShih amp Keller 1992b Keller amp Jansa 1992) Likewise theneural tissue can converge and extend without the under-lying mesoderm (Keller amp Danilchik 1988 Elul amp Keller2000) Most matrix research has focused on the prominentcentbronectin-containing matrix lining the blastocoel roofand its role in mesodermal migration (see Winklbauer ampKeller 1996) Whether this matrix is needed for neuralextension is not known Because it is presumably retainedat least initially in the deep neural explants it could beused as a substratum for neural convergent extensionAlthough the mesodermal cells do not need the organizedcentbrillar matrix on the blastocoel roof to converge andextend a centbrillar system of matrix might lie between themesodermal cells and serve as a substrate for the tractionof these cells Arguing against this notion is the fact thatno such centbrillar system has been characterized in eitherthe neural or mesodermal tissue

However it is possible that extracellular matrixcomponents function in a non-centbrillar form Molecularcentbronectin as opposed to organized centbrils is foundthroughout the mesodermal tissue through gastrulationand beyond (D DeSimone personal communication)Whether this form of centbronectin functions in themesodermal mediolateral intercalation behaviour (MIB)is not known Arguing against this notion is the fact thatantibodies and peptides blocking the integrin-mediatedmigration of mesodermal cells on the blastocoel roof havemuch less or no aiexclect on convergent extension move-ments (Ramos amp DeSimone 1996 Ramos et al 1996Winklbauer amp Keller 1996) But it is not known whetherthe antibody and peptide reagents used in these experi-ments penetrate between the densely packed mesodermalcells of the extending axis

There is extracellular matrix in the NSB and thismatrix may have a function in setting up the boundaryblocking cell movements across the boundary andperhaps even organizing the boundary-mediated cellcapture described above (5(b)) (centgure 11b ) Extra-cellular matrix is abundant in the notochordal sheathand this structure arises from the notochordal somiticmesodermal boundaries that develop as morphologicallyvisible structures at midgastrula stage (Shih amp Keller1992a) Molecules such as laminin (Fey amp Hausen 1990)centbronectin (Lee et al 1984) and elastin (P Skoglundunpublished results) are found by immunopounduorescence inthe notochordal sheath at neurula stages but have notbeen detected at earlier stages of boundary formation

However the matrix molecule centbrillin is found inthe forming boundary at midgastrula stages (centgure 11cP Skoglund unpublished data) Fibrillin is a componentof microcentbrils (Sakai et al 1986) widely distributed 10^12 nm diameter extensible extracellular centbres that canserve mechanical functions for example in ligaments ofthe human eye (reviewed in Ramirez et al 1993) There-fore this molecule is a strong candidate for involvement inthe changing biomechanical properties that dorsal meso-derm exhibits as these boundaries form In addition



(b)

(c)

(a)

Figure 11 (a) An extracellular matrix might function in cellintercalation by providing a non-cellular substrate on whichcells could crawl directionally in the case of the monopolarmode of cell intercalation (b) Matrix may prevent cells fromcrossing the NSB and also control the boundary-capturemode of cell intercalation by oiexclering a substrate on whichcells could adhere and spread In both cases matrixremodelling would have to occur for large extensionsAntibody staining shows the presence of centbrillin protein in theearly NSB (pointers) at the midgastrula stage viewed from thedorsal side (c) the blastopore is at the bottom

because cells have receptors for centbrillin (Pfaiexcl et al 1996Sakamoto et al 1996) centbrillin is a candidate molecule fora role in regulating the boundary-capture mode of cellbehaviour seen there

Extracellular matrix functions in the straightening andstiiexclening of the notochord in the tailbud stages Thedense centbrillar matrix of the notochordal sheath developsin the late neurula through the tailbud stages and forms acentbre-wound tube The notochord cells then vacuolate andgenerate pressure within the matrix-wound tube whichresults in stiiexclening and straightening of the notochord inthis period (see Adams et al 1990)

9 PATTERNING OF MEDIOLATERAL INTERCALATION

BEHAVIOUR PROGRESSIVE ANTERIOR-TO-

POSTERIOR AND LATERAL-TO-MEDIAL EXPRESSION

Expression of mediolateral intercalation behaviour(MIB) is spatially and temporally regulated in both the

mesodermal and neural tissues Expression occurs fromanterior to posterior and from lateral to medial in bothtissues such that the two undergo parallel movementsthe mesodermal on the inside and the neural on theoutside of the blastoporal lip (Keller et al 1992a) Thisprogressive organization is essential for the function ofboth neural and mesodermal convergent extensionmachines in gastrulation

(a) Patterning of MIB in the mesodermExpression of MIB in the prospective mesoderm is best

visualized in explants of the entire involuting marginalzone (IMZ) of the embryo (centgure 12a) (Shih amp Keller1992a) The expression of the bipolar mode of MIB isrepresented diagrammatically as elongated fusiformshapes (centgure 12) First consider the expression of MIBin explants that are not allowed to converge and extend(centgure 11b) MIB is centrst expressed in the anterior lateralprospective somitic mesoderm and progresses medially as



(b)

(e)(i)

(ii)

(a)d

d

v v

so

v v

no

so

noM

MMP

P

NSB

early gastrula

late midgastrula

noso

stag

e 10

+

onw

ards

stag

e 11

onw

ards

stag

e 10

+ to

10

5

A AL

LL

(d)

(c)

vegetal alignment zone

no

so

Figure 12 To visualize the pattern of expression of mediolateral intercalation behaviour (MIB) (a) explants of a large part orall of the involuting marginal zone were made and cultured in two ways Some explants were cultured beneath a restrainingcover-slip such that they did not undergo convergence and extension (b) or more loosely restrained and allowed to convergeand extend (c) Under these conditions the progress of MIB was determined from time-lapse recordings and represented here byfusiform cell procentles (b c) (d ) The expression of MIB in the whole embryo at the early gastrula and late midgastrula is alsoshown The prospective anterior (A) to posterior (P) and lateral (L) to medial (M) axes as well as the dorsal (d) and ventralsides (v) in the somitic and notochordal centelds are indicated in (a) Note that the progress of MIB is from anterior to posteriorand from lateral to medial (e) Similar anterior-to-posterior and lateral-to-medial expression of MIB occurs in the neural plateand notoplate Also shown is the notochord (no) the somitic mesoderm (so) and the NSB (i) Deep neural plate (ii) deep neuralplate over the mesoderm

arcs which meet in the midline to form the VAZ(centgure 12b solid arrows stage 10+ to 105) The NSBthen forms at the midgastrula stage within the VAZ andit proceeds posteriorly (centgure 12b small arrows stage 11onwards) Then MIB is expressed from anterior toposterior in the lateral region of somitic centeld and fromanterior to posterior in the lateral boundary region ofthe notochordal mesoderm (centgure 12b solid arrows stage11 onwards) From this lateral origin MIB progressesmedially towards the medial aspect of both the somiticcenteld and the notochordal centeld (centgure 12b solid arrowsstage 11 onwards) If the explants are not allowed toconverge and extend medial posterior regions of boththe prospective somitic and notochordal mesodermalcentelds fail to express MIB and diiexclerentiate into theirrespective tissues (centgure 12b asterisks stage 11 onwards)

If the explants are allowed to converge and extend theexpression of MIB follows the same pattern from anteriorto posterior and lateral to medial in both notochordal andsomitic regions (centgure 12c) However the convergencebrings the lateral boundaries closer to the medial parts ofboth tissues and under these conditions MIB spreadsthroughout the medial parts of both the notochordal andsomitic regions This fact and the fact that MIBprogresses medially from a lateral origin suggests thatthe lateral boundaries of these tissues might be thesources of the signals organizing MIB as well as cell fateThis notion is supported by the fact that if the lateralboundaries are not allowed to converge and move closertogether the medial cells do not diiexclerentiate MIB orproper tissue fates (Shih amp Keller 1992a Domingo ampKeller 1995) If the tissue boundaries are sources ofsignals of limited range convergent extension would haveto be self-reinforcing convergence of the lateral signal-ling boundaries would bring more cells into rangeallowing more convergence and so on

These results also suggest that the signalling systemsorganizing MIB and perhaps centnal cell fate are extantduring gastrulation and that cell fate is not yet deter-mined in the early gastrula stage To test these notionslabelled cells from the notochordal somitic andepidermal regions of early to midgastrula stages weregrafted into the notochordal region of early gastrulawhole embryos and explants (Domingo amp Keller 1995)(centgure 13a) Their motility was recorded with time-lapseimaging and their cell fate determined by marker expres-sion (Domingo amp Keller 1995) (centgure 13b) These graftedcells express MIB and diiexclerentiate as notochord in thesame anteroposterior and lateral^medial order as thenative notochordal cells (centgure 13b) Thus the notochordalcells are not pre-programmed to express MIB in theobserved pattern at the outset of the behaviour butsignals operating during gastrulation probablyemanating from the boundary regions organize MIBduring gastrulation Moreover in Domingorsquos experi-ments cells of epidermal prospective fate grafted into thenotochordal region were converted to notochord as wellas induced to express MIB showing that these cells wereexposed to mesodermal cell fate-inducing signals as wellas MIB-inducing signals during gastrulation

Mediolateral cell intercalation appears to be regulatedby the Wnt signalling pathway A dominant inhibitoryform of Xenopus dishevelled which is involved in the Wnt

signalling pathway inhibits convergent extension (Sokol1996) as does a dominant-inhibitory form Xenopusfrizzled-8 another upstream component of the pathway(Deardoriexcl et al 1998) Evidence for the involvement ofthe `planar polarityrsquo non-canonical Wnt signallingpathway in controlling cell motility of convergent exten-sion is described in Smith (this issue)

(b) Are there two zones of MIB organizationIn an attempt to learn if microtubules were involved in

development of cell polarity during MIB Lane amp Keller(1997) found evidence for two zones of organization ofMIB and cell fate If microtubules are disrupted with the



(b)

(a)

Figure 13 Labelled cells from another prospectivenotochordal somitic or epidermal tissue of a gastrula stageembryo were grafted into the notochordal region of anunlabeled early gastrula (long arrow a) (b) The onset ofMIB was then recorded with time-lapse pounduorescencemicroscopy Expression of MIB in the labelled cells occurredfrom lateral to medial (solid arrows a) and from anterior toposterior (open arrows a) From Domingo amp Keller (1995)

drug nocodozole in early gastrulation (ie before theVAZ forms) the VAZ never forms convergent extensionnever occurs involution never occurs and a smallamount of notochord and somitic tissue diiexclerentiates inthe anterior (VAZ) region of the mesoderm This meso-derm diiexclerentiates in the pre-involution position and theremaining posterior axial and paraxial mesoderm doesnot express MIB or diiexclerentiate If microtubules aredisrupted after the VAZ has formed the mesodermal cellsexpress MIB converge and extend and diiexclerentiate asnotochord and somite without microtubules This includesthe posterior axial and paraxial mesoderm Absence ofmicrotubules could prevent expression of MIB duringformation of the VAZ in a number of ways and thus noclear interpretation of this result can be made Howeverthe surprising and important result is that expression ofMIB in the notochordal and somitic regions posterior tothe VAZ does not require microtubules provided the VAZhas formed Although the anterior VAZ region and theremaining posterior part of the axis appear to be identicalin the type of MIB expressed and in the diiexclerentiationnotochord and somites these regions may represent twounits of axial organization each operating on diiexclerentmechanisms at some level (see Lane amp Keller 1997)

(c) Patterning of MIB in the neural tissueThe neural versions of MIB also originate in the ante-

rior lateral region of the prospective neural plate andprogress posteriorly and medially from this origin (Elulamp Keller 2000) This progression occurs in both the deepneural explants and in the deep neural over mesodermexplants and applies to the both the bipolar mode and themonopolar mode of neural plate cell intercalation and tothe notoplate-specicentc multipolar behaviour (centgure 11e)These progressions of cell behaviours follow the samepattern as the cell divisions and neuronal diiexclerentiationin the neural plate (see Hartenstein 1989)

These centndings have implications for understandingneural patterning The convergent extension movementscan be induced by planar signals from Spemannrsquos orga-nizer (Keller et al 1992c) However if these signals areproceeding from the organizer toward the lateral bound-aries of the neural plate as would be expected andexpression of neural MIB follows the order in which thecells received these signals then the expression of MIBshould be from posterior to anterior and medial to lateralInstead the cell behaviours are organized medially fromthe lateral boundary and posteriorly from the anteriorboundary mimicking the pattern seen in the mesodermThe simplest explanation is that inducing signals from theorganizer interact with opposing signals from theepidermal region to establish a neural plate boundary (seereview by Harland amp Gerhart 1997) After the boundary isestablished MIB is organized by secondary signalsproceeding inwardly or posteriorly from this boundary

(d) The neural midline and the monopolar modeof cell intercalation

The monopolar medially directed protrusive activitysuggests that it might be stimulated by a signal emanatingfrom a midline structure the obvious one being thenotoplate early and following that the poundoor plate derivedfrom it Elul amp Keller (2000) observed a stronger

polarization of the protrusive activity of the lateral neuralplate cells as these cells approached the notoplate whichis consistent with a polarizing signal emanating from thenotoplate and therefore stronger near it than furtheraway Arguing against this notion however is the obser-vation that embryos without notochords and apparentlywithout poundoor plates appear to develop relatively normalneural tubes (Malacinski et al 1981) Whether the nervoussystems of these embryos converged and extended by thenormal monopolar mode or the bipolar mode is notknown Perhaps they used the less ecurrencient bipolar modebut its decentciencies were oiexclset by towing forces from theunderlying mesoderm

An alternative mechanism is that the notoplate^poundoorplate despite its prominent position has nothing to dowith organizing the monopolar MIB but instead theunderlying somitic mesoderm imposes a medial direction-ality on the overlying neural cells In this mechanismsomitic mesoderm rather than notoplate^poundoor platewould be necessary for medially polarized protrusiveactivity Experiments are underway to test these possi-bilities

(e) Function of progressive expression of MIBThe progressive expression of MIB is essential for

proper gastrulation The centrst expression of MIB theVAZ forms an arc across the dorsal lip of the gastrulaand as MIB shortens this arc a hoop stress is generatedacross the dorsal lip pulling it vegetally over the bottlecells and thus contributing to involution (centgure 12d earlygastrula) The endodermal rotation movements describedrecently by Winklbauer amp Schulaquo rfeld (1999) initiate invo-lution and set up the blastoporal groove below the VAZand drive much of the early involution After this timeMIB is expressed in the postinvolution region progres-sively posteriorly from the now involuted VAZ and as aresult a wave of convergence and resulting hoop stressproceeds posteriorly on the inside of the blastopore(centgure 12d mid^late gastrula) driving continued involu-tion and squeezing the blastoporal lip toward closure at apoint over the ventral region of the vegetal endodermBecause the mesoderm is expressing MIB posteriorlyfrom the anterior postinvolution region and the neuraltissue is expressing MIB posteriorly from an anteriororigin in the prospective hindbrain on the outside of thegastrula these tissues are expressing convergent extensionin the same order and more or less in spatial correspon-dence (see Keller et al 1992a)

(f) Mechanical consequences of the bipolarand monopolar modes of cell intercalation

The bipolar mode of MIB has no inherent directionalitybut only an orientation the mediolaterally orientatedprotrusive activity tends to shorten the array of cellspulling the ends together without a bias as to which endor perhaps both is to move the most in accommodatingthe shortening The anchorage points and deformabilityof the neighbouring anchoring tissues determine theseparameters In the embryo the mesodermal cells expres-sing MIB in progressively more posterior regions can bevisualized as forming arcs that are anchored at each endnear the margins of the large vegetal endodermal massand span the across the IMZ (centgure 14a) It should be



kept in mind that in the embryo the process is contin-uous we only portray the process as separated arcs forclarity of illustration As MIB progresses posteriorlythese arcs shorten in sequence and thus generate a hoopstress across the IMZ progressing posteriorly asdescribed above In this case neither end moves butrather the arcs shorten and the IMZ is pulled across theblastoporal lip (centgure 13a)

If the arcs are broken somewhere along their length onboth sides the medial segments of the arcs of MIB areunanchored and thus pull dorsally where they convergeextend and form an isolated extension and at the sametime fail to involute (centgure 14bc light shading) (Keller1981) The isolated lateral segments of the broken arcs inthese experiments have their medial ends unanchoredand therefore they pull these ends towards (convergetowards) the vegetal endoderm and extend in the per-pendicular direction parallel to the endodermal margin(centgure 13bc dark shading) Such movement results inreopening of the partially closed blastopore (centgure 13c)

This type of embryo appears often in the classical litera-ture where it is known as a `ring embryorsquo and it forms incases where the continuity of the arcs of convergence arebroken (see Keller 1986)

In the case of the monopolar mode of MIB in theneural tissue the medially directed protrusive activitytends to pull the cells to the midline thus pulling theedges of the neural plate towards the middle The lateralanchorage of the neural plate is in the epidermis whichappears to be deformable and thus it stretches to allowthe edges of the neural plate to move medially Note thatthere is no inherent reason why the bipolar mode wouldnot work for convergent extension of the neural tissue aslong as the epidermis is deformable

(g) Specicentcity of morphogenesis involvesbiomechanical integration of distributed information

It is the global context of local cell behaviour that deter-mines the specicentcity of the morphogenic eiexclect It is thegeometry and the lateral anchorage points of the arc-like



(b)

(a) (c)(i)

ant

ant

mid

mid

(ii)

(iii)post post

Figure 14 The mechanism of progressive mediolateral intercalation of mesodermal cells in bringing about involution of themarginal zone (a) The mediolateral intercalation of mesodermal cells is illustrated as a series of arcs of intercalating cells aprospective anterior (ant) one (i) a prospective middle (mid) one (ii) and a prospective posterior (post) one (iii) all anchoredat each end at appropriate points near the large vegetal endodermal mass (a left) During gastrulation these arcs are shortenedby mediolateral cell intercalation beginning with (i) and following with (ii) and (iii) As each successive arc is shortened thecorresponding region of the marginal zone is rolled over the lip by the hoop stress passing across the dorsal and lateral lips of theblastopore (a right) (b) If these arcs are cut dorsolaterally the isolated medial segments are no longer anchored laterally buthave free ends These arc segments shorten by convergence pulling each unanchored end medially and as a result they extenda proboscis-like structure that normally would have involuted out into the medium (b right c) Likewise the isolated lateralsegments now have a free medial end and an anchored lateral end (b) These arcs converge toward the vegetal endodermbringing their free medial ends towards the vegetal endoderm and they extend around the circumference of the vegetalendoderm causing the embryo to reopen its blastopore (c) A light micrograph shows an embryo of this type often called a `ringembryorsquo

expression patterns of MIB in the mesoderm thatdetermine the specicentc eiexclects of MIB Taken alone as inthe sandwich explants (centgure 3) the arc-like array MIB-expressing cells will produce simple convergence andextension In contrast if anchored at the edges of thevegetal endoderm the arc-like pattern of progressive MIBexpression will produce involution as well The specicentcityof this type of morphogenesis depends on distributedinformation information that is integrated by thebiomechanical properties and geometry of the tissues todetermine the output and signicentcance of local cellbehaviour This has major implications for genetic andmolecular analysis of morphogenesis

We oiexcler this paper in remembrance of Nigel Holder We werenot fortunate enough to know him personally but his examplehis published work and his many less-decentnable but importantcontributions to the tenor of the centeld have gained our admira-tion and deep respect

REFERENCES

Adams D Keller R E amp Koehl M A R 1990 Themechanics of notochord elongation straightening andstiiexclening in the embryo of Xenopus laevis Development 110115^130

Bauman M amp Sanders K 1984 Bipartite axiation followsincomplete epiboly in zebracentsh embryos treated withchemical teratogens J Exp Zool 230 363^376

Betchaku T amp Trinkaus J P 1978 Contact relations surfaceactivity and cortical microcentlaments of marginal cells of theenveloping layer and of the yolk syncytial and yolk cyto-plasmic layers of Fundulus before and during epiboly J ExpZool 206 381^426

Bode P amp Bode H 1984 Formation of pattern in regeneratingtissue pieces of Hydra attenuata III The shaping of the bodycolumn Devl Biol 106 315^325

Brieher W amp Gumbiner B 1994 Regulation of C-cadherinfunction during activin induced morphogenesis of Xenopusanimal caps J Cell Biol 126 519^527

Broders F amp Thiery J-P 1995 Contribution of cadherins todirectional cell migration and histogenesis in Xenopusembryos Cell Adhesion Commun 3 419^440

Burnside M B amp Jacobson A 1968 Analysis of morphogeneticmovements in the neural plate of the newt Taricha torosa DevlBiol 18 537^552

Concha M L amp Adams R J 1998 Oriented cell divisions andcellular morphogenesis in the zebracentsh gastrula and neurulaa time-lapse analysis Development 125 983^994

Choquet D Felsenfeld D P amp Sheetz M P 1997Extracellular matrix rigidity causes strengthening of integrin^cytoskeleton linkages Cell 88 39^48

Cloney R 1964 The development of the ascidian notochordActa Embryol Morphol Expo 7 111^130

Condic M Fristrom J amp Fristrom D 1991 Apical cellshape changes during Drosophila imaginal leg disc elonga-tion a novel morphogenetic mechanism Development 11123^33

Davidson L amp Keller R 1999 Neural tube closure in Xenopuslaevis involves medial migration directed protrusive activitycell intercalation and convergent extension Development 1264547^4556

Deardoriexcl M Tan C Conrad L amp Klein P 1998 Frizzled-8is expressed in the Spemann organizer and plays a role inearly morphogenesis Development 125 2687^2700

Delarue M Saez F Boucaut J-C Thiery J-P amp Broders F1998 Medial cell mixing during axial morphogenesis of theamphibian embryo requires cadherin function Devl Dynam213 248^260

Detrick R J Dickey D amp Kintner C 1990 The eiexclects ofN-cadherin misexpression on morphogenesis in Xenopusembryos Neuron 4 439^506

Domingo C amp Keller R 1995 Induction of notochord cell inter-calation behavior and diiexclerentiation by progressive signals inthe gastrula of Xenopus laevis Development121 3311^3321

Elul T amp Keller R 2000 Monopolar protrusive activity a newmorphogenic cell behaviour in the neural plate dependent onvertical interactions with the mesoderm in Xenopus Devl Biol(In the press)

Elul T Koehl M amp Keller R 1997 Cellular mechanismunderlying neural convergent extension in Xenopus laevisembryos Devl Biol 191 243^258

Ettensohn C 1985 Gastrulation in the sea urchin embryo isaccompanied by the rearrangement of the invaginatingepithelial cells Devl Biol 112 383^390

Evans E Berk D amp Leung A 1991 Detachment of agglutinin-bonded red blood cells I Forces to rupture molecular-pointattachments Biophys J 59 838^848

Fey J amp Hausen P 1990 Appearance and distribution oflaminin during development of Xenopus laevis Diiexclerentiation42 144^152

Hadeball B Borchers A amp Wedlich D 1998 Xenopuscadherin-11 (Xcadherin-11) expression requires the WgWtsignal Mech Devl 72 101^113

Hardin J D 1988 The role of secondary mesenchyme cellsduring sea urchin gastrulation studied by laser ablationDevelopment 103 317^324

Hardin J D 1989 Local shifts in position and polarized moti-lity drive cell rearrangement during sea urchin gastrulationDevl Biol 136 430^445

Hardin J D amp Cheng L Y 1986 The mechanism andmechanics of archenteron elongation during sea urchingastrulation Devl Biol 115 490^501

Harland R amp Gerhart J 1997 Formation and function ofSpemannrsquos organizer A Rev Cell Dev Biol 13 611^667

Harris A 1980 Silicone rubber substrata a new wrinkle in thestudy of cell locomotion Science 208 177^179

Hartenstein V 1989 Early neurogenesis in Xenopus the spatio^temporal pattern of proliferation and cell lineages in theembryonic spinal cord Neuron 3 399^411

Hartenstein V amp Campos-Ortega J A 1985 Fate mapping inwild-type Drosophila melanogaster Rouxrsquos Arch Devl Biol 194181^193

Heasman J Crawford A Goldstone K Garner-HamrickP Gumbiner B McCrea P Kintner C Noro C Y ampWylie C 1994 Overexpression of cadherins and under-expression of b catenin inhibit dorsal mesoderm induction inearly Xenopus embryos Cell 79 791^803

Herzberg F Wildermuth V amp Wedlich D 1991 Expression ofXbcad a novel cadherin during oogenesis and early develop-ment of Xenopus Mech Devl 35 33^42

Holtfreter J 1944 A study of the mechanics of gastrulation IIJ Exp Zool 95 171^212

Irvine K D amp Wieschaus E 1994 Cell intercalation duringDrosophila germband extension and its regulation by pair-rulesegmentation genes Development 120 827^841

Jacobson A amp Gordon R 1976 Changes in the shape of thedeveloping vertebrate nervous system analyzed experimen-tally mathematically and by computer simulation J ExpZool 197 191^246

Jacobson A amp Moury J D 1995 Tissue boundaries and cellbehavior during neurulation Devl Biol 171 98^110



Jacobson A Odell G amp Oster G 1985 The cortical tractormodel for epithelial folding application to the neural plateIn Molecular determinants of animal form University of CaliforniaLos Angeles symposia on molecular and cellular biology (new series)vol 31 (ed G M Edelman) pp143^167 NewYork Liss

Jacobson A Oster G F Odell G M amp Cheng L Y 1986Neurulation and the cortical tractor model for epithelialfolding J Embryol Exp Morphol 96 19^49

Keller R E 1975 Vital dye mapping of the gastrula and neurulaof Xenopus laevis I Prospective areas and morphogenetic move-ments of the supercentcial layer DevlBiol42 222^241

Keller R E 1976 Vital dye mapping of the gastrula andneurula of Xenopus laevis II Prospective areas andmorphogenetic movements of the deep layer Devl Biol 51118^137

Keller R E 1978 Time-lapse cinemicrographic analysis ofsupercentcial cell behavior during and prior to gastrulation inXenopus laevis J Morphol 157 223^248

Keller R E 1980 The cellular basis of epiboly an SEM studyof deep-cell rearrangement during gastrulation in Xenopuslaevis J Embryol Exp Morphol 60 201^234

Keller R E 1981 An experimental analysis of the role of bottlecells and the deep marginal zone in gastrulation of Xenopuslaevis J Exp Zool 216 81^101

Keller R E 1984 The cellular basis of gastrulation in Xenopuslaevis active postinvolution convergence and extension bymediolateral interdigitation Am Zool 24 589^603

Keller R E 1986 The cellular basis of amphibian gastrulationIn Developmental biology a comprehensive synthesis 2 The cellularbasis of morphogenesis (ed L Browder) pp 241^327 New YorkPlenum

Keller R 1987 Cell rearrangement in morphogenesis Zool Sci4 763^769

Keller R 2000 The origin and morphogenesis of amphibiansomites CurrTop Devl Biol 47 183^246

Keller R E amp Danilchik M 1988 Regional expressionpattern and timing of convergence and extension duringgastrulation of Xenopus laevis Development 103 193^209

Keller R amp Jansa S 1992 Xenopus gastrulation without a blas-tocoel roof Devl Dynam 195 162^176

Keller R E amp Schoenwolf G C 1977 An SEM study ofcellular morphology contact and arrangement as related togastrulation in Xenopus laevis Wilhelm Rouxrsquos Arch 182165^186

Keller R E amp Tibbetts P 1989 Mediolateral cell intercalationin the dorsal axial mesoderm of Xenopus laevis Devl Biol 131539^549

Keller R E Danilchik M Gimlich R amp Shih J 1985 Thefunction and mechanism of convergent extension duringgastrulation of Xenopus laevis J Embryol Exp Morphol89(Suppl) 185^209

Keller R E Cooper M Danilchik M Tibbetts P ampWilson P 1989a Cell intercalation during notochord develop-ment in Xenopus laevis J Exp Zool 251 134^154

Keller R Shih J amp Wilson P 1989b Morphological polarityof intercalating deep mesodermal cells in the organizer ofXenopus laevis gastrulae In Proceedings of the 47th Annual Meetingof the Electron Microscopy Society of America (ed G W Bailey)pp 840^841 San Francisco CA San Francisco Press

Keller R Shih J Wilson P amp Sater A 1991a Pattern andfunction of cell motility and cell interactions during conver-gence and extension in Xenopus In 49th Symposium of the Societyfor Developmental Biology cell cell interactions in early development(ed J Gerhart) pp 93^107 Wiley

Keller R Shih J amp Wilson P 1991b Cell motility control andfunction of convergence and extension during gastrulation ofXenopus In Gastrulation movements patterns and molecules (edR KellerW Clark amp F Gricurrenn) pp101^119 Plenum

Keller R Shih J amp Sater A 1992a The cellular basis of theconvergence and extension of the Xenopus neural plate DevlDynam 193 199^217

Keller R Shih J amp Domingo C 1992b The patterning andfunctioning of protrusive activity during convergence andextension of the Xenopus organizer Development 1992(Suppl)81^91

Keller R Shih J Sater A amp Moreno C 1992c Planar induc-tion of convergence and extension of the neural plate by theorganizer of Xenopus Devl Dynam 193 218^234

Kim S-H Yamamoto A Bouwmeester T Agius E ampDeRobertis E M 1998 The role of paraxial protocadherinin selective adhesion and cell movements of the mesodermduring Xenopus gastrulation Development 125 4681^4691

Kimmel C Kane D Walker C Warga R amp Rothman M1989 A mutation that changes cell movement and cell fate inthe zebracentsh embryo Nature 337 358^362

Kimmel C Warga R amp Kane D 1994 Cell cycles and clonalstrings during formation of the zebracentsh central nervoussystem Development 120 265^276

Kintner C 1988 Eiexclects of altered expression of the neural celladhesion molecule N-CAM on early neural development inXenopus embryos Neuron 1 545^555

Laale H 1982 Fish embryo culture observations on axial corddiiexclerentiation in presomitic isolates of the zebracentshBrachydanio rerio (Hamilton^Buchanan) Can J Zool 601710^1721

Lane M C amp Keller R 1997 Microtubule disruption revealsthat Spemannrsquos organizer is subdivided into two domains bythe vegetal alignment zone Development 124 895^906

Lee C H amp Gumbiner B M 1995 Disruption of gastrulationmovements in Xenopus by a dominant-negative mutant forC-cadherin Devl Biol 171 363^373

Lee G Hynes R amp Kirschner M 1984 Temporal and spatialregulation of centbronectin in early Xenopus development Cell36 729^740

Lee T L Lin Y C Mochitate K amp Grinnell F 1993 Stress-relaxation of centbroblasts in collagen matrices triggers ecto-cytosis of plasma membrane vesicles containing actinannexins II and VI and beta 1 integrin receptors J Cell Sci105 167^177

Levine E Chung H L Kintner C amp Gumbiner B 1994Selective disruption of E-cadherin function in early Xenopusembryos by a dominant negative mutant Development 120901^909

Malacinski G M amp Youn B-W 1981 Neural plate morphogen-esis and axial stretching in `notochord-defectiversquoXenopus laevisembryos Devl Biol 88 352^357

Miyamoto D amp Crowther R 1985 Formation of the noto-chord in living ascidian embryos J Embryol Exp Morphol86 1^17

Moore S W 1992 Direct measurement of dynamic biomecha-nical properties of amphibian embryonic tissues PhD thesisUniversity of California Berkeley CA USA

Moore S Keller R amp Koehl M 1995 The dorsal involutingmarginal zone stiiexclens anisotropically during its convergentextension in the gastrula of Xenopus laevis Development 1213131^3140

Pfaiexcl M Reinhardt D P Sakai L Y amp Timpl R 1996 Celladhesion and integrin binding to recombinant humancentbrillin-1 FEBS Lett 384 247^250

Poznanski A Minsuk S Stathopoulos D amp Keller R 1997Epithelial cell wedging and neural trough formation areinduced planarly in Xenopus without persistent vertical inter-actions with mesoderm Devl Biol 189 256^269

Priess J R amp Hirsh D I 1986 Caenorhabditis elegans morpho-genesis the role of the cytoskeleton in elongation of theembryo Devl Biol 117 156^173



Ramirez F Pereira L Zhang H amp Lee B 1993 Thecentbrillin^Marfan syndrome connection BioEssays 15 589^594

Ramos J W amp DeSimone D W 1996 Xenopus embryonic celladhesion to centbronectin position-specicentc activation of RGDsynergy site-dependent migratory behavior at gastrulation JCell Biol 134 1^14

Ramos J W Whittaker C A amp Desimone D W 1996Integrin-dependent adhesive activity is spatially controlled byinductive signals at gastrulation Development 122 2873^2883

Sakai L Y Keene D R amp Engvall E 1986 Fibrillin a new350-kD glycoprotein is a component of extracellular micro-centbrils J Cell Biol 103 2499^2509

Sakamoto H Broekelmann T Cheresh D A Ramirez FRosenbloom J amp Mecham R P 1996 Cell-type specicentcrecognition of RGD- and non-RGD-containing cell bindingdomains in centbrillin-1 J Biol Chem 274 4916^4922

Sater A Steinhardt R A amp Keller R 1993 Induction ofneuronal diiexclerentiation by planar signals in Xenopus laevisDevl Dynam 197 268^280

Sausedo R A amp Schoenwolf G C 1994 Quantitative analysesof cell behaviors underlying notochord formation and exten-sion in mouse embryos Anat Rec 239 103^112

Schechtman A M 1942 The mechanics of amphibian gastrula-tion I Gastrulation-producing interactions between variousregions of an anuran egg (Ityla regila) Univ Calif Publ Zool 511^39

Schoenwolf G C amp Alvarez I 1989 Roles of neuroepithelialcell rearrangement and division in shaping of the avianneural plate Development 106 427^439

Shih J amp Keller R 1992a Patterns of cell motility in theorganizer and dorsal mesoderm of Xenopus Development 116915^930

Shih J amp Keller R 1992b Cell motility driving mediolateralintercalation in explants of Xenopus laevis Development 116901^914

Shih J amp Keller R 1992c The epithelium of the dorsalmarginal zone of Xenopus has organizer propertiesDevelopment 116 887^899

Snow M 1977 Gastrulation in the mouse growth and regional-ization of the epiblast J Embryol Exp Morphol 42 293^303

Sokol S 1996 Analysis of Dishevelled signalling pathwaysduring Xenopus development Curr Biol 6 1456^1467

Tuft P 1962 The uptake and distribution of water in theembryo of Xenopus laevis (Daudin) J Exp Biol 39 1^19

Vogt W 1929 Gestaltanalyse am Amphibienkein mit olaquo rtlicherVitalfalaquo rbung II Teil Gastrulation und Mesodermbildung beiUrodelen und Anuren Wilhelm Roux Arch EntwMech Org 120384^706

Waddington C H 1940 Organizers and genes CambridgeUniversity Press

Warga R amp Kimmel C 1990 Cell movements during epibolyand gastrulation in the zebracentsh Development 108 569^580

Williams-Masson E M Malik A N amp Hardin J 1997 Anactin-mediated two-step mechanism is required for ventralenclosure of the C elegans hypodermis Development 1242889^2901

Wilson P amp Keller R 1991 Cell rearrangement during gastru-lation of Xenopus direct observation of cultured explantsDevelopment 112 289^300

Wilson P Oster G amp Keller R E 1989 Cell rearrangementand segmentation in Xenopus direct observation of culturedexplants Development 105 155^166

Winklbauer R amp Keller R E 1996 Fibronectin mesodermmigration and gastrulation in Xenopus Devl Biol 177 413^426

Winklbauer R amp Schulaquo rfeld M 1999 Vegetal rotation a newgastrulation movement involved in the internalization of themesoderm and endoderm in Xenopus Development 1263703^3713

Yamamoto A Amacher S Kim S-L Geissert DKimmel C amp DeRobertis E M 1998 Zebracentsh paraxialprotocadherin is a downstream target of spadetail involvedin morphogenesis of gastrula mesoderm Development 1253389^3397

Zhong Y Brieher W amp Gumbiner B 1999 Analysis of C-cadherin regulation during tissue morphogenesis with an acti-vating antibody J Cell Biol 144 351^359



`convergent extensionrsquo can be used for convenience but itcan be misleading unless one remembers that the rela-tionship between convergence extension and thickeningcan be and usually is more complex than this termsuggests

Convergence and extension movements are representa-tive of a larger more general class of `mass movementsrsquoinvolving change in tissue proportions with approximateconservation of volume Divergence and shortening(centgure 1c) the reverse of convergent extension occursduring Drosophila germ band retraction (Hartenstein ampCampos-Ortega 1985) Epiboly an increase in tissue areaby thinning (centgure 1d ) occurs in the animal region ofearly amphibian (see Keller 1978 1980) and teleost centshesembryos (Betchaku amp Trinkaus 1978) to mention onlytwo examples Convergence and extension are oftenmajor components of what are usually identicented as othertypes of morphogenic processes For example invagina-tion of a disc of cells to form a tubular gut such asprimary invagination of a echinoderm vegetal plate toform the gut is often portrayed as a simple bending insectional view (centgure 1e) In fact such an invaginationinvolves a form of convergent extension the progressivenarrowing of the circumference of the disc and extensionalong its radius (the length of the tube) (centgure 1e)

Convergence and extension have been studied most insituations where tissue volume changes due to cell growthdoes not occur at all or is insignicentcant such as in theearly development of sea urchins (Hardin amp Cheng 1986)amphibians (Burnside amp Jacobson 1968 Keller 1986)and poundies (Condic et al 1991) However these movementsalso occur in embryos that grow dramatically in earlydevelopment such as those of the mouse (Snow 1977) and

in these cases tissue volume changes may also come intoplay Analysis of the mechanism of convergent extensionin organisms that grow rapidly may reveal variations onthe themes developed here specicentcally the potentialinvolvement of directed anisotropic growth

Convergence and extension movements as well asrelated mass tissue movements may be passive responsesto forces generated elsewhere in the embryo or they maybe active force-producing processes We will focus on theactive movements here although the passive ones are alsoimportant in morphogenesis and misunderstanding themcan result in erroneous conclusions For example it is acommon assumption that if a mutation targeted to aparticular tissue in the embryo disrupts convergent exten-sion of that tissue the movement is an active oneHowever the movement could be a passive response toforce generated elsewhere and the true function of thewild-type gene could be to increase the deformability ofthe tissue allowing it to be stretched and narrowed bythe active tissue

(b) Why study convergence and extensionThese movements present a major challenge and an

opportunity to understand cell function in the morpho-genesis of integrated populationsThey represent a categoryof morphogenic processes the so-called `mass movementsrsquoin which a tissue distorts by virtue of integrated cell beha-viours within its boundaries due to interactions amongthese cells rather than interactions with external substratesLittle is known of the cellular molecular and bio-mechanical mechanisms of these types of movements animportant decentciency in the light of their dramatic role inshaping embryonic body form in so many systems



(a) (c) (e)

(i)

(ii)(b)

(d)

Figure 1 Diiexclerent types of mass tissue movements including (a) convergence and extension (b) convergence extension andthickening (c) divergence and shortening (d ) epiboly and (e) invagination in (i) sectional view and (ii) in three dimensions








BP

lateneurula

lategastrula

earlygastrula

lateralsurface

Fb

So

HM

HM

N

Hb and Sc

midsaggital







AND NEURULATION









Fb

(b)

(a)

So

No

Hb and Sc



























deep

(a)

(b)

deep

epith

epith

radialintercalation

radialintercalation








INTERCALATION

















(a)

(b)

(c)

)(d

(e)










neural plate

midgastrulaembryo

notoplate

(a)

(b)

(c) )(d

(e)




































old

old

new

new

(a) (i)

(i)

(ii)

(ii)


anchor

(iv)

(iv)

(v)

(v)

(iii)

(iii)(b)

(c)

)(d

)( f(e)

























(b)

(a)













































(b)

(c)

(a)













(b)

(e)(i)

(ii)

(a)d

d

v v

so

v v

no

so

noM

MMP

P

NSB

early gastrula

late midgastrula

noso

stag

e 10

+

onw

ards

stag

e 11

onw

ards

stag

e 10

+ to

10

5

A AL

LL

(d)

(c)


no

so











(b)

(a)
























(b)

(a) (c)(i)

ant

ant

mid

mid

(ii)

(iii)post post




REFERENCES


















































































































BP

lateneurula

lategastrula

earlygastrula

lateralsurface

Fb

So

HM

HM

N

Hb and Sc

midsaggital







AND NEURULATION









Fb

(b)

(a)

So

No

Hb and Sc



























deep

(a)

(b)

deep

epith

epith

radialintercalation

radialintercalation








INTERCALATION

















(a)

(b)

(c)

)(d

(e)










neural plate

midgastrulaembryo

notoplate

(a)

(b)

(c) )(d

(e)




































old

old

new

new

(a) (i)

(i)

(ii)

(ii)


anchor

(iv)

(iv)

(v)

(v)

(iii)

(iii)(b)

(c)

)(d

)( f(e)

























(b)

(a)













































(b)

(c)

(a)













(b)

(e)(i)

(ii)

(a)d

d

v v

so

v v

no

so

noM

MMP

P

NSB

early gastrula

late midgastrula

noso

stag

e 10

+

onw

ards

stag

e 11

onw

ards

stag

e 10

+ to

10

5

A AL

LL

(d)

(c)


no

so











(b)

(a)
























(b)

(a) (c)(i)

ant

ant

mid

mid

(ii)

(iii)post post




REFERENCES















































































































AND NEURULATION









Fb

(b)

(a)

So

No

Hb and Sc



























deep

(a)

(b)

deep

epith

epith

radialintercalation

radialintercalation








INTERCALATION

















(a)

(b)

(c)

)(d

(e)










neural plate

midgastrulaembryo

notoplate

(a)

(b)

(c) )(d

(e)




































old

old

new

new

(a) (i)

(i)

(ii)

(ii)


anchor

(iv)

(iv)

(v)

(v)

(iii)

(iii)(b)

(c)

)(d

)( f(e)

























(b)

(a)













































(b)

(c)

(a)













(b)

(e)(i)

(ii)

(a)d

d

v v

so

v v

no

so

noM

MMP

P

NSB

early gastrula

late midgastrula

noso

stag

e 10

+

onw

ards

stag

e 11

onw

ards

stag

e 10

+ to

10

5

A AL

LL

(d)

(c)


no

so











(b)

(a)
























(b)

(a) (c)(i)

ant

ant

mid

mid

(ii)

(iii)post post




REFERENCES




































































































































deep

(a)

(b)

deep

epith

epith

radialintercalation

radialintercalation








INTERCALATION

















(a)

(b)

(c)

)(d

(e)










neural plate

midgastrulaembryo

notoplate

(a)

(b)

(c) )(d

(e)




































old

old

new

new

(a) (i)

(i)

(ii)

(ii)


anchor

(iv)

(iv)

(v)

(v)

(iii)

(iii)(b)

(c)

)(d

)( f(e)

























(b)

(a)













































(b)

(c)

(a)













(b)

(e)(i)

(ii)

(a)d

d

v v

so

v v

no

so

noM

MMP

P

NSB

early gastrula

late midgastrula

noso

stag

e 10

+

onw

ards

stag

e 11

onw

ards

stag

e 10

+ to

10

5

A AL

LL

(d)

(c)


no

so











(b)

(a)
























(b)

(a) (c)(i)

ant

ant

mid

mid

(ii)

(iii)post post




REFERENCES

















































































































INTERCALATION

















(a)

(b)

(c)

)(d

(e)










neural plate

midgastrulaembryo

notoplate

(a)

(b)

(c) )(d

(e)




































old

old

new

new

(a) (i)

(i)

(ii)

(ii)


anchor

(iv)

(iv)

(v)

(v)

(iii)

(iii)(b)

(c)

)(d

)( f(e)

























(b)

(a)













































(b)

(c)

(a)













(b)

(e)(i)

(ii)

(a)d

d

v v

so

v v

no

so

noM

MMP

P

NSB

early gastrula

late midgastrula

noso

stag

e 10

+

onw

ards

stag

e 11

onw

ards

stag

e 10

+ to

10

5

A AL

LL

(d)

(c)


no

so











(b)

(a)
























(b)

(a) (c)(i)

ant

ant

mid

mid

(ii)

(iii)post post




REFERENCES





















































































































(a)

(b)

(c)

)(d

(e)










neural plate

midgastrulaembryo

notoplate

(a)

(b)

(c) )(d

(e)




































old

old

new

new

(a) (i)

(i)

(ii)

(ii)


anchor

(iv)

(iv)

(v)

(v)

(iii)

(iii)(b)

(c)

)(d

)( f(e)

























(b)

(a)













































(b)

(c)

(a)













(b)

(e)(i)

(ii)

(a)d

d

v v

so

v v

no

so

noM

MMP

P

NSB

early gastrula

late midgastrula

noso

stag

e 10

+

onw

ards

stag

e 11

onw

ards

stag

e 10

+ to

10

5

A AL

LL

(d)

(c)


no

so











(b)

(a)
























(b)

(a) (c)(i)

ant

ant

mid

mid

(ii)

(iii)post post




REFERENCES



















































































































neural plate

midgastrulaembryo

notoplate

(a)

(b)

(c) )(d

(e)




































old

old

new

new

(a) (i)

(i)

(ii)

(ii)


anchor

(iv)

(iv)

(v)

(v)

(iii)

(iii)(b)

(c)

)(d

)( f(e)

























(b)

(a)













































(b)

(c)

(a)













(b)

(e)(i)

(ii)

(a)d

d

v v

so

v v

no

so

noM

MMP

P

NSB

early gastrula

late midgastrula

noso

stag

e 10

+

onw

ards

stag

e 11

onw

ards

stag

e 10

+ to

10

5

A AL

LL

(d)

(c)


no

so











(b)

(a)
























(b)

(a) (c)(i)

ant

ant

mid

mid

(ii)

(iii)post post




REFERENCES













































































































































old

old

new

new

(a) (i)

(i)

(ii)

(ii)


anchor

(iv)

(iv)

(v)

(v)

(iii)

(iii)(b)

(c)

)(d

)( f(e)

























(b)

(a)













































(b)

(c)

(a)













(b)

(e)(i)

(ii)

(a)d

d

v v

so

v v

no

so

noM

MMP

P

NSB

early gastrula

late midgastrula

noso

stag

e 10

+

onw

ards

stag

e 11

onw

ards

stag

e 10

+ to

10

5

A AL

LL

(d)

(c)


no

so











(b)

(a)
























(b)

(a) (c)(i)

ant

ant

mid

mid

(ii)

(iii)post post




REFERENCES


































































































































(b)

(a)













































(b)

(c)

(a)













(b)

(e)(i)

(ii)

(a)d

d

v v

so

v v

no

so

noM

MMP

P

NSB

early gastrula

late midgastrula

noso

stag

e 10

+

onw

ards

stag

e 11

onw

ards

stag

e 10

+ to

10

5

A AL

LL

(d)

(c)


no

so











(b)

(a)
























(b)

(a) (c)(i)

ant

ant

mid

mid

(ii)

(iii)post post




REFERENCES






















































































































































(b)

(c)

(a)













(b)

(e)(i)

(ii)

(a)d

d

v v

so

v v

no

so

noM

MMP

P

NSB

early gastrula

late midgastrula

noso

stag

e 10

+

onw

ards

stag

e 11

onw

ards

stag

e 10

+ to

10

5

A AL

LL

(d)

(c)


no

so











(b)

(a)
























(b)

(a) (c)(i)

ant

ant

mid

mid

(ii)

(iii)post post




REFERENCES






















































































































(b)

(e)(i)

(ii)

(a)d

d

v v

so

v v

no

so

noM

MMP

P

NSB

early gastrula

late midgastrula

noso

stag

e 10

+

onw

ards

stag

e 11

onw

ards

stag

e 10

+ to

10

5

A AL

LL

(d)

(c)


no

so











(b)

(a)
























(b)

(a) (c)(i)

ant

ant

mid

mid

(ii)

(iii)post post




REFERENCES




















































































































(b)

(a)
























(b)

(a) (c)(i)

ant

ant

mid

mid

(ii)

(iii)post post




REFERENCES

































































































































(b)

(a) (c)(i)

ant

ant

mid

mid

(ii)

(iii)post post




REFERENCES













































































































REFERENCES











































































































Mechanisms of convergence and extension by cell intercalation · Convergence and extension...

Documents

Transcript of Mechanisms of convergence and extension by cell intercalation · Convergence and extension...