Amph Nrv Sys MSWord97 Rev 2

Cross references:     



The nervous system of amphioxus: structure, development, and evolutionary significance1

HelmutWicht and ThurstonC. Lacalli

Abstract: Amphioxus neuroanatomy is important not just in its own right but also for the insights it provides regarding the evolutionary origin and basic organization of the vertebrate nervous system. This review summarizes the overall layout of the central nervous system (CNS), peripheral nerves, and nerve plexuses in amphioxus, and what is currently known of their histology and cell types, with special attention to new information on the anterior nerve cord. The inter- calated region (IR) is of special functional and evolutionary interest. It extends caudally to the end of somite 4, tradi- tionally considered the limit of the brain-like region of the amphioxus CNS, and is notable for the presence of a number of migrated cell groups. Unlike most other neurons in the cord, these migrated cells detach from the ventricu- lar lumen and move into the adjacent neuropile, much as developing neurons do in vertebrates. The larval nervous sys- tem is also considered, as there is a wealth of new data on the organization and cell types of the anterior nerve cord in young larvae, based on detailed electron microscopical analyses and nerve tracing studies, and an emerging consensus regarding how this region relates to the vertebrate brain. Much less is known about the intervening period of the life history, i.e., the period between the young larva and the adult, but a great deal of neural development must occur dur- ing this time to generate a fully mature nervous system. It is especially interesting that the vertebrate counterparts of at least some postembryonic events of amphioxus neurogenesis occur, in vertebrates, in the embryo. The implication is that the whole of the postembryonic phase of neural development in amphioxus needs to be considered when making phylogenetic comparisons. Yet this is a period about which almost nothing is known. Considering this, plus the number of new molecular and immunocytochemical techniques now available to researchers, there is no shortage of worthwhile research topics using amphioxus, of whatever stage, as a subject.

Résumé : La neuroanatomie de l’amphioxus est importante, non seulement en elle-même, mais parce qu’elle offre des perspectives sur l’origine évolutive et l’organisation de base du système nerveux des vertébrés. Notre rétrospective ré- sume la disposition générale du système nerveux central (CNS), des nerfs périphériques et des plexus nerveux chez l’amphioxus et fait le point sur les connaissances sur l’histologie et les types cellulaires, et en particulier sur les décou- vertes récentes sur la corde nerveuse antérieure. La région intercalée (RI) est d’intérêt particulier des points de vue fonctionnel et évolutif. Elle s’étend vers l’arrière jusqu’à la fin du somite 4 qui est traditionnellement considéré la li- mite de la portion ressemblant à un cerveau du CNS de l’amphioxus et elle est remarquable par la présence de plu- sieurs groupes de cellules migrantes. Contrairement à la plupart des autres neurones de la corde, ces cellules se détachent de la lumière ventriculaire et envahissent le neuropile, comme le font les neurones en développement chez les vertébrés. Nous considérons aussi le système nerveux des larves, puisqu’il y a une abondance de données nouvelles sur l’organisation et les types cellulaires de la corde nerveuse antérieure chez les jeunes larves basées sur des analyses détaillées au microscope électronique et des études de traçage des nerfs et parce qu’un consensus sur la relation entre cette région et le cerveau des vertébrés est en train de se former. La période intermédiaire du cycle biologique, c’est-à- dire entre la jeune larve et l’adulte, est beaucoup moins connue; il doit cependant s’y compléter beaucoup de dévelop- pement nerveux pour produire le système nerveux complètement mature. Il est particulièrement intéressant de noter que des phénomènes correspondant à au moins quelques événements de la neurogenèse de l’amphioxus se répètent dans l’embryon dans vertébrés. Cela implique qu’il faut tenir compte de l’ensemble de la phase post-embryonnaire du déve- loppement neural chez l’amphioxus afin de faire des comparaisons phylogénétiques. C’est néanmoins une période dont

Received 17 August 2004. Accepted 18 November 2004. Published on the NRC Research Press Web site at on 19 April 2005.

H. Wicht.2 Fachbereich Medizin der Johann Wolfgang Goethe-Universität, Dr. Senckenbergische Anatomie, Theodor-Stern-Kai 7, D-60590, Frankfurt/Main, Germany.

T.C. Lacalli. Biology Department, University of Victoria, Victoria, BC V8W 3N5, Canada.

1This review is one of a series dealing with aspects of the biology of the Protochordata. This series is one of several virtual symposia on the biology of neglected groups that will be published in the Journal from time to time.

2Corresponding author (e-mail:

Can. J. Zool. 83: 122–150 (2005) doi: 10.1139/Z04-163 © 2005 NRC Canada

on connaît à peu près rien. Dans ces conditions, le nombre de nouvelles techniques moléculaires et immunocytochimi- ques disponibles pour la recherche fait qu’il ne manque pas de sujets de recherche de valeur sur l’amphioxus à tous les stades de son cycle.

[Traduit par la Rédaction]


The common ancestor of present-day vertebrates andthe invertebrate cephalochordates haslong been extinct, butif onehad tochoose aliving species likelyto resemble itmost closely, amphioxus (Branchiostoma spp.; lancelets) wouldbe the organism ofchoice (Presley etal. 1996; Holland 2000).As such, amphioxus isof key importance to investi- gations into vertebrate origins and characteristic features of vertebrate organization. The nervous systemis ofspecial in-terest because itsbasic planis highly conserved amongver- tebrates, yetalmost nothingis known abouthow thatplan originated. Onehas little recourse exceptto conduct investi- gationsof amphioxus andother protochordates. Thiswas the rationale behindthe anatomical studiesof amphioxus carried outin thelate 19th century, principally on Branchiostoma lanceolatum (Pallas, 1774),the European species, byleading comparative zoologists ofthe day.There hasbeen something ofa hiatusin the20th century, inpart becauseof concerns thatthe apparent simplicity of amphioxus isdue toits being degenerate and derived, rather than primitive, butalso be-cause of diminishing returns from studies using conventional light microscopical methods. Muchof the original literature isin German, including major studiesby Rohde (1887), Retzius (1891), Dogiel (1903), andFranz (1923)and acom- prehensive reviewby Franz (1927). Bone (1959, 1960a, 1961) providesa comprehensive summaryof previous workfor non-German speakers, alongwith muchnew information fromhis own research. Themost recent reviewson amphi- oxusare byRuppert (1997), forthe general anatomyof theanimal asa whole,and Nieuwenhuys (1998), forthe nervous system.

Why then reviewthe nervous systemyet again?The an-swer hastwo parts. First,the nervous systemof amphioxus is surprisingly peculiar froma vertebrate perspective, soit isnot atrivial matterto geta good conceptual feelfor itsorga- nization. Justas onemight havemore thanone tourist guidewhen visiting an unfamiliar cityfor thefirst time(each pref- erably witha goodset ofmaps), thereis adistinct benefitin having morethan onereview toconsult regarding amphi- oxus neuroanatomy, preferably withgood illustrations, aseach will inevitably havea slightly different perspective. Second, andmore importantly, awhole newtool kitof mo- lecular and immunocytochemical techniques is now being employed to reexamine the structure and development of phylogenetically interesting animals, including amphioxus. New studiesof the amphioxus nervous systemnow appearon aregular basis,and gene expression studies frequently alsofocus onthe nervous system. Givenour very limited un- derstanding of amphioxus neural structure and function, thesenew results areoften difficult to interpret, typically raising more questions thanthey answer.A reexamination ofwhat wedo know,as of2004, is therefore notout ofplace. Whatis perhaps more important, however, isthe waythe

newresults highlight majorgaps inour knowledge —for example, the postembryonic events thatturn alarval nervecord intoan adultone (see below). Thus,as wellas asum- maryof facts,we tryhere toprovide insights intoareas of emerging interest and highlight issues wherewe think major advances canbe expected inthe future.We hope,by this means,to stimulate new thinking aboutthe anatomy and functional organization ofthe amphioxus nervous system, andbetter inform research intothe originof vertebrates andtheir nervous system.

Inthe account that follows, theadult nervous systemis dealtwith first, followed bya separate sectionon thelarval nervous system, specifically thatof young larvae 12–14 daysold. Separate treatment is necessary because thenature ofthe datais so different forthe two stages, being largely con- ventional anatomical descriptions basedon light microscopy and immunocytochemistry inthe caseof theadult, and cellular-level details from electron microscopical (EM)re- constructions inthe caseof thelarva. Because ourknowl- edgeof the intervening periodof development is incomplete, itis notalways clearhow structures andcell typesin the nervous systemof young larvae relateto thosein theadult, andit is consequently impossible tomake thetwo sections as seamlessly complementary asone ideally might like.

Anatomical overview

General appearance

Thecentral nervous system (CNS)of adult lancelets con-sists ofa tubular nerve cord, located directly abovethe noto- chord, that extends mostof thelength ofthe body(Fig. 1). Rostrally, the notochord extends forwardof thenerve cord,to almostthe tipof thebody, hencethe namefor this taxon: Cephalochordata. Thenerve cord itself beginsa few hundred micrometres caudalto the anteriormost coelomic cavity, which supports the rostral fin,and just forward, bya fewdozen micrometres, ofthe first myomere. Caudally, thenerve cord reachesto thetip ofthe caudalfin where, justabove thecaudal endof the notochord, itforms aterminal enlargement of variable form,the caudal ampulla (Retzius 1891; Franz 1923).

Unlikethe situation in vertebrates, the lancelet nervecord displays few externally visible landmarks exceptfor the above-mentioned ampulla anda transient anterior swelling (the cerebral vesicle) inyoung larvae. Thereare thusno neuromeres orother external indications of segmental ar- rangement otherthan the serially repeated exit pointsfor thedorsal nerves. Internally, however, pronounced cytoarchi- tectural differences canbe identified alongthe rostrocaudal axis,both inthe adultand inlarvae. Thesecan beused todefine aseries ofmajor regions alongthe nerve cord,and someof thesecan befurther subdivided. Becauseof theab- senceof external landmarks, the boundaries ofthese subdivi- sionsand their relation tothe peripheral nervous structures

Fig. 1. A highly schematic lateral view of a lancelet (Brachiostoma sp.) with 63 myomeres (m1 to m63), showing the general organization of the central and peripheral nervous systems in relation to peripheral structures. The first four and last three members of the myomere series are highlighted (light shading) to clarify the relationship between myomere number and nerve number. Regions occupied by major blocks of non-myomeric muscle are indicated by darker shading. Selected body regions are cut away, as indi- cated by the jagged lines. Abbreviations and their corresponding terms are listed in Appendix A.

are best described with reference tothe myomeres, ofwhich, on average, thereare 63pairs inB. lanceolatum (Poss and Boschung 1996). Thisposes some problems ifone istrying tocompare CNS architecture throughout development be-cause (i) thoughit islikely, itis notknown for certain whether the myomeres are permanently fixedin position in relation tothe nerve cord,and (ii) myomeres expand both lengthwise and dorsoventrally during development, which, combined withtheir chevron shape, means thereis anin- creasing zoneof overlap between themthat varies depend- ingon position alongthe dorsoventral axis. (iii) Theleft andright rowsof myomeres inthe adultare not aligned; thefirst larval somites, however, fromwhich the myomeres develop, are.In the neurula stage,the leftrow ofsomites (aswell asthe leftrow ofdorsal nervesand neuromuscular contact zones) startsto shift anteriorly (Conklin 1932).To make things evenmore complicated, this"left forward shift" isless pronounced inthe regionof thefirst five somites (myomeres). Thus, in the adult, the plane defined by the rostral tipsof thefirst myomeres isalmost perpendicular tothe longaxis ofthe body, whereas thetips of increasingly more posterior myomeres define increasingly more oblique planes. Approximately atthe levelof thesixth pairof myo- meres,the "full" oblique angleof about45° (i.e.,the typical "half-myomere staggering") isreached and maintained throughout thecaudal partof thebody. Inthe present re-view, weshall usethe transverse planes definedby thetips ofthe leftrow of myomeres as landmarks. Past accounts, however, have not always described the location of CNS landmarks ina sufficiently precise wayto overcome these problems.

The peripheral nervous system (PNS) consists ofa setof peripheral plexuses3 and a segmental seriesof intermyo- meric dorsal nerves(= dorsal rootsor "true" nerves,as op-posed tothe apparent ventral roots, whichare infact muscle processes rather than nerves). Thedorsal nervesare entirely devoidof ganglia; inother words, amphioxus hasno coun- terparts ofthe vertebrate dorsal root ganglia. Peripheral nervesin amphioxus instead consist solelyof axons, which derive fromboth central and peripheral neurons, anda few glial-like support cells.The peripheral neurons residein var-ious peripheral plexuses, whichare especially well devel- oped aroundthe gut,and insense organsand theskin. Thenerves themselves issuefrom the dorsolateral marginsof thenerve cord, their proximal parts being locatedin the myosepta, i.e.,they pass between the myomeres. Owingto the left-right asymmetry ofthe myomeres (theleft rowis displaced a half-segment forwardof theright one),the nervesare staggered leftto right.The patternof peripheral innervation fora typical nerveof the trunk is shown inFig. 2;for details seethe section (below)on thespinal cord. Innervation patterns inthe rostral andoral regionare some-what different; Fig.3 showsthis indetail, while cytological details andcell typesas seenin typical transverse sections ofthe nervecord areshown inFigs. 5and 6.

Anterior nervecord

Thetip ofthe adult nervecord rostralto myomere1 isre- ferredto asthe anterior vesicle(= cerebral vesicle (Willey

1894),Gehirn (Edinger 1906), archencephalon (von Kupffer 1906),and Stirnbläschen (Franz 1927)). The central canalis expanded inthis regionand isroughly circular incross sec-tion. Itopens onthe leftto the outside via Kölliker’s pit (Kölliker 1843),a remnantof the anterior neuropore (Willey 1894).Two pairsof nerves emerge fromthe anterior vesicleto supplythe rostral region.By convention theseare num-bered 1and 2(e.g., Dogiel 1903; Kutchin 1913; Nieuwenhuys 1998),but are referred toalso asthe rostral and anterodorsal nerves, respectively (e.g., Lacalli 2002b). They differ from remaining membersof thedorsal nervese- riesin several respects. First, becauseof the absenceof myomeres inthe rostrum, bothnerve pairs travel forwardin the connective tissue sheath surrounding thenerve cordand notochord, whereas therest ofthe dorsal nerves pass through the myosepta. Second, the various branches ofnerves 1and 2carry small bulbous clusters ofcells, thecor- pusclesof de Quatrefages (Fig.3), putative mechanosensory organs consisting ofprimary sensory cells enclosed ina cap-sule (Baatrup 1982)that occur nowhere elsein thebody. Nerve1, the rostral nerve,is unusual alsoin thatit entersthe nervecord ventrally rather than dorsally. Ithas nevertheless become customary toinclude itwith therest ofthe dorsal nerves (e.g., Dogiel 1903; Kutchin 1913; Nieuwenhuys 1998).We follow this convention here, thoughit hasthe dis- advantage thatnerve and myomere numbers thenno longer match; i.e.,a given nerven would emerge from between myomeresn –1 andn –2 (seeFig. 1).

Theregion ofthe adult nervecord extending fromthe an-terior vesicle through thefirst four myomeres is recognized as distinctive enoughto requirea special designation. Ithas been referred to variously asa deuterencephalon (von Kupfer 1906), hindbrain (Fritzsch 1996; Holland 1996),or caudal brain region (Nieuwenhuys 1998),but was regarded asa partof thespinal cord (Rückenmarksteil) byFranz (1923). Most recently, Ekhartet al.(2003) have coinedthe term intercalated region (IR)so asto avoidthe implications of homology associated withthe older terms.The IRas de-fined byEkhart etal. extends fromthe infundibular organ (Boeke 1902, 1908), which produces Reissner’s fiber (Obermüller-Wilén 1976),to thefirst giantcell ofRohde (Rohde 1887).It contains three subdivisions (anterior, inter- mediate, and posterior), described below,and is character- izedby the presence ofa numberof celltypes and groupings notseen elsewhere inthe spinal cord.The most conspicuous ofthese arethe large dorsal cellsof Joseph (Joseph 1904), whichare putative photoreceptors (seeFigs. 6,7). TheIR givesrise tonerves 3–6, which, likeall thedorsal nerves, send branches tothe corresponding component ofthe subepidermal nerve plexus in the skin. These nerves also connect withtwo peripheral nerve rings(or plexuses) associ- atedwith thebuccal region (Fig.3), namelythe buccal plexus, which innervates thebuccal cavity, cirri,and associ- ated muscles (the internal and external labial muscles, Franz 1927),and withthe velar plexus, which innervates thetenta- clesand sphincter muscleof thevelum.

Animportant non-neural structure locatedin this regionis Hatschek’s pit (Hatschek 1884), whichlies ventralto theIR atthe junction of myomeres3 and4. Thisis actuallya part

3 With respect to the common usage in the English-speaking world, we will not use the latin plural (plexus).

of the ciliated wheel organof Müller (Müller 1844; Franz 1927), locatedin theroof ofthe buccal cavity (Figs.1, 3,7C). Insome lancelet species thepit formsa direct contact withthe baseof thenerve cord(see below).

From boththe adult anatomy andgene expression datait isclear thatthe anterior vesicle andIR together forma CNSregion witha numberof features incommon withthe verte- brate brain: theyform the anterior partof the neuraxis, their nerves supplythe region associated withthe buccal cavity, and cytoarchitectural specializations occurthat arenot pres-ent inmore caudal regionsof thenerve cord. Thereis strong evidence thatthese anterior regions also contain homologs ofneural centers andsense organs thatoccur inthe brainof craniates, as discussed morefully inthe last section. Itis nevertheless difficult to determine thecaudal extentof this brain-like region withany precision anddefine itin away that applies equallyto both larvaeand adults.In adults, the

firstRohde cellis usually assumedto markthe posterior endof thebrain, butthis isbased moreon convenience (thecell isvery easyto recognize) thanon evidence, since thereare other internal landmarks and discontinuities thatcould beused justas well.On theother hand, determining theexact extentof thebrain in craniates canbe problematic, asthe boundary between spinal cordand brain,as wellas thenum- berof cranial nerves, canvary from speciesto species (Nieuwenhuys 1998).

Oneway todefine thebrain isas thatpart ofthe CNSen- closedin the cranium, butthis is obviously nota useful cri- terion when appliedto acraniates. If, instead, the vertebrate brainis definedin termsof its peripheral nervous connec- tions, which supplythe full extentof the pharyngeal region, thenthe caudalend ofthe brainof lancelets wouldcome tolie somewhere inthe middleof thebody. Inlarval lancelets, expression domainsof Otx and various Hox genes indicate

Fig. 3. A ventrolateral view of the rostrum and oral region of a lancelet, showing the first eight dorsal nerves (n1 to n8), but only those on the left side. The buccal and velar plexuses supplying the buccal chamber that extends from the base of the cirri to the velum are shaded. Also shown are the first cell of Rohde (arrow), Hatschek’s pit (white arrowhead), an anastomosis between nerves 1 and 2 (open arrowhead), and two of the neuromuscular contact zones (black arrowheads), for myomeres 5 and 6 in this instance. Note that each contact zone consists of a small dorsal and a larger ventral compartment. Abbreviations and their corresponding terms are listed in Appendix A.

that an equivalent ofthe vertebrate mid-hindbrain boundary maybe present somewhere inthe fronthalf ofsomite 2,with a hindbrain-like region extending thereafter toabout somite7 or8 (Shimeld and Holland 2005). This wouldmove the junction between brainand spinal cordto aposi- tionthree orfour myomeres caudalto thefirst Rohde cell.The nervecord inthe neurula also extends from somite1 tosomite 8, which has led to the suggestion that the entire neurulais roughly equivalent inaxial extentto the vertebrate head (Gilland andBaker 1993). Thiswould makethe nervecord inthe neurula coextensive withthe vertebrate brain, withmost ofits length being hindbrain-like in character. Whilethis maybe true,the detailsof whatit really meansto be "hindbrain-like" remainto beworked out (Jackman andKimmel 2002; Mazetand Shimeld 2002),so wemust con-clude thatthe exact caudal boundary ofthe brainin lancelets cannot currently be determined withany certainty.

Spinal cord, dorsal nerves, and associated peripheral structures

The largest regionof theCNS is commonly referred toas thespinal cord,in recognition ofits general similarity tothat structure in vertebrates. Usingthe first Rohdecell asthe

landmarkthat dividesit fromthe anterior CNSin theadult, thespinal cord stretches fromthe fifth myomereto thelast, andthus givesrise tomost ofthe dorsal nerves.Up tonerve 53,which innervates theanal sphincter muscle, thedorsal nerves connect thespinal cordto theatrial nervous system, themost extensive and complex component ofthe PNS.Fig- ure2 showshow thenerves and plexuses are arranged inthe transverse plane. Oncea typical nerve leavesthe nerve cord,it entersthe intermyotomal septumand passes laterally alongit, dividing intoa dorsaland aventral ramusas itleaves. Therami travel underthe skin lateralto themuscle block (notethat in vertebrates the corresponding nervein- steadlies medialto thetrunk musculature), dorsally inthe caseof thedorsal ramusand ventrally inthe caseof theven- tral ramus,to supply fibersto the subepidermal nerve plexus underlying theskin ofthe dorsalfin (dorsal ramus)or the subepidermal plexus beneath theskin onthe flanks (ventral ramus).At the ventral marginof the myomere, the ventral ramus divides intoa ventral cutaneous ramus, whichsup- pliesthe skinof the metapleural folds,and avisceral ramus, which turns medially towards thewall ofthe atrial cavity, whereit again divides intoan ascending anda descending branch. Theatrial epithelium andthe epithelia ofall organs

that either flankit (e.g., gonadsand pterygeal muscle)or are located withinit (pharynx and endostyle, thegut andits di- verticulum) are underlain bya massive systemof nerve cellsand fibers collectively termedthe atrial nervous system (Bone 1961).The descending branchof the visceral ramus contributes tothe gonadal and pterygeal portions ofthat sys-tem and,in addition, contributes motor fibersto thecross- striated pterygeal (= transverse) muscle. The ascending branchof the visceral ramus turns dorsally andgives offfi- bersto theplexus covering thegonads andthe lateral wallsof the atrium.It thenturns medially and reaches the denticulate ligament, which tethers the pharynx andthe gutto theroof ofthe atrium, andby thismeans itreaches thenerve plexusof the pharynx, gillbars, endostyle, andgut plus diverticulum.

On thebasis ofits internal appearance, theadult spinal cordcan be subdivided intothree major regions (thisac- count follows Franz 1923, 1927).The anterior partof thecord, atthe levelof myomeres 5–11,is characterized bythe presence oflarge Rohde cellswith descending axonsand bya high densityof pigmented, photoreceptive organsof Hesse (Hesse 1898). Peripherally, nerves connect thispart ofthe cordwith themost caudal partsof thebuccal nerve plexus, vianerve 7,while themore posterior nervesin this regionall connect withthe atrial nervous system (Bone 1961; see Figs.1, 2).

The intermediate partof thespinal cord,from myomeres 12to 38,lacks Rohde cells,and the densityof Hesse organsis reduced here. Peripherally, nerves fromthis region also connectto theatrial nervous system. Its rostrocaudal extent coincides roughly withthat ofthe gonads, thecaudal limit being approximately atthe levelof the atriopore, whose cross-striated sphincter is innervated bynerves 40and 41 (Kutchin 1913).The transition between pharynx andgut islocated approximately atthe levelof myomere26 or27. The atriocolemic funnels, enigmatic organs thatmay besensory innature, are located justabove this point. Theyare sur- roundedby the cross-striated trapezius muscles (Franz 1927;Bone 1961), whichare innervated bythe ascending visceral branches ofnerves 27–29.

The posterior partof thespinal cord extends fromthe atriopore, atthe levelof myomeres 38–40, tothe lastmyo- mere, typically number63 inB. lanceolatum (Poss and Boschung 1996).The numberof Hesse organs increases againin this regionand theRohde cells reappear. Nerves 51–53 supplythe anal sphincter muscle (Kutchin 1913). Nerves caudalto thispoint lack visceral rami (Franz 1927).The lastnerve (nerve65 inan animal with63 myomeres) leavesthe spinal cordat thecaudal borderof thelast myomere and innervates theskin adjacent tothe tailfin.

Terminal filament andcaudal ampulla

Still more caudally, thespinal cord tapers intoa thin ependymal tubule, the terminal filament (filum terminale), which connects thecord toa caudal ampulla justabove the posterior tipof the notochord. Reissner’s fiber, whichis pro-duced inthe infundibular organ, extends intothat caudal ampulla, whereit is apparently phagocytosed by specialized cellsin thewalls ofthe ampulla (Obermüller-Wilén andOlsson 1974). Thereare numerous nerve fibersof unknown origin alongthe ventral and lateral marginsof thecaudal

ampulla.These contain densecore vesicles, which implies thatthis structure may function asan endocrine organcom- parableto the urophysis of anamniotic vertebrates (Ruizand Anadón 1991a).

The adultPNS

Peripheral plexuses

Thesystem of peripheral plexuses in amphioxus isun- usualin anumber of respects. It suppliesa fine meshwork offree nerve endings thatrun through theentire epidermis (Dogiel 1903; Kutchin 1913; Franz 1923, 1927; Welsch 1968b, seeFig. 2)such thatevery epidermal cellis appar- entlyin contact withthem (Leleet al.1958). Thefree nerve endings probably arisefrom populations of intramedullary sensory neurons, ofwhich thereare twotypes, according toBone (1960a, 1961):the dorsal bipolaror Retzius bipolar celland thedorsal rootcell (seeFig. 4).The precise sourceof thefree nerve endings hasnot been determined withcer- tainty, however.

Theepithelia thatline thebuccal andatrial cavities, aswell asthe organs embedded within them,are also supplied withan extensive setof neural plexuses collectively knownas theatrial nervous system. Thoughthe systemis moreor less continuous, itis usually subdivided onthe basisof theorgans it innervates; i.e., thereare buccal, velar, gonadal, pa- rietal, pterygeal, pharyngeal, and endostylar subdivisions, andso on(Bone 1961). Whilethe various plexuses carry motor fibers innervating muscles associated withthe atrium, theyalso contain thecell bodiesand fibersof alarge numberof peripheral neurons (Figs.2, 4).It hasbeen argued (e.g., Boeke 1935)that the plexuses are homologs ofthe auto- nomic nervous systemof craniates, butthis isalmost cer- tainly false. Enteric neuronsin vertebrates develop from migratory neural crest cells,a category of embryonic cellsthat is entirely absentin amphioxus sofar ascan bedeter- mined (Bakerand Bronner-Fraser 1997). This impliesby de-fault that neuronsin the peripheral plexuses in amphioxus arise locally, insitu, though Lacalli (2004)has pointed outthat theembryo isso smallat thetime that neuronal precur- sorsare probably first deployed thatan origin closeto theneural plate, similarto thatof placodes andthe neural crest, cannotbe ruledout. Itis alsotrue, however, thatthe neuro- transmitters identified todate in amphioxus peripheral neu-rons differ fromthose released by autonomic neuronsin vertebrates. Specifically, neither acetylcholine (Flood 1974)nor catecholamines (Moretet al.2004) occurin peripheral neurons, atleast inthe atrial nervous system. FMRFamide, however, doesoccur (Boneet al.1996), andthis isa trans- mitter restricted mainlyto theCNS in vertebrates. Inaddi- tion,many ofthe peripheral neuronsin amphioxus sendaxons intothe nervecord viathe dorsal nerves (Holmes 1953;Bone 1961). This impliesa sensory function, rather thana motorone.

Basedon these unique features, Boneet al.(1996) con- cluded thatthere wereno obvious homologies between thePNS of amphioxus andthat of vertebrates. Ingeneral termsthis conclusion appears justified. Nevertheless, inrats, cen- tripetal projections areknown from peptidergic peripheral neurons (called rectospinal neurons) resident inthe wallsof theanus (Dörffler-Melly and Neuhuber 1988). Theseare ev-

idently notsimply displaced dorsal root ganglion cells,but constitute insteada novel classof vertebrate enteric neurons (Neuhuber etal. 1993). Theycould conceivably bea relicof a primitive modeof visceral innervation relatedto thatin amphioxus. Conversely, theycould bea derived featureof no phylogenetic significance.

The velarand buccal plexuses deserve special attention owingto their curious asymmetry (Dogiel 1903; Kutchin 1913; Franz 1923, 1927).As shownin Fig.3, thereare twobuccal plexuses, anouter andan innerone, anda singleve- larplexus (Bone 1961).The velar plexus derives develop- mentally fromthe oral plexus that encircles thelarval mouthas itmoves ventrally and caudally during metamorphosis (Franz 1923;Kaji etal. 2001;the term "oral" should thusbe avoided when referring toneural structures associated withthe cirri because theyare really preoral, buccal structures). Aswith other partsof theatrial nervous system, thebuccal andvelar plexuses combine sensory components (Bone 1961)with motor ones;the latter innervate thelabial andve- lar muscles. The connection tothe nervecord vianerves 1–7(1–8 according toDogiel 1903; Kutchin 1913)is highly asymmetrical. Nerves3 and4 onthe leftside are exceptional inhaving contralateral branches that connect withthe inner buccal plexuson theright side.In addition, a subsidiary branch fromthe contralateral branchof theleft nerve4 con-nects tothe rightside ofthe velar plexus, whileits leftside connects tonerve 5by meansof acaudal branch fromthat nerve. Thisis alla consequence ofthe factthat thelarval mouth develops initially onthe leftside andis innervated entirely bynerves emerging fromthe leftside ofthe nervecord (Lacalli etal. 1999).The initial connections arethen retained during subsequent development, sothe nervesare dragged alongas themouth is repositioned.

Peripheral sensory cellsand organs

Lancelets havean assortment ofsensory cellsand organs located both insidethe nervecord and outside it.The formerare dealtwith in relevant sections dealing withthe adultand larval CNS;they include the various photoreceptor systems, whichare all intramedullary, and Kölliker’s pit,for whichan olfactory function has been suggested. Hatschek’s pit, lo-cated inthe roofof thebuccal cavity (Fig.3), isnot, strictly speaking, partof the nervous system, butis considered here becauseof itsclose association withit.

Multicellular organs

The peripheral tissues arewell supplied with solitary sen-sory cellsand freenerve endings butharbor onlya few multicellular structures towhich asensory function canbe attributed. Four examples are considered here.(1) The atriocolemic funnels, first described by Lankester (1875), consistof paired conical recesses inthe dorsal surfaceof theatrial cavity that project anteriorly intothe subchordal coelom (Figs.1, 2;see Willey 1894; Franz 1927). Boththe funnels andthe surrounding striated trapezius muscleare densely innervated bya branch fromthe ascending visceral ramusof nerve27 (Holmes 1953;Bone 1961)and, to alesser degree,by neighboring dorsal nerves (Franz 1927).The fibersmay bechiefly involved inthe innervation ofthe trapezius muscle, butthe atriocoelomic funnels themselves contain manyuni- and multipolar neurons (Bone (1961) dis-

tinguishedthree types) whose axons enter visceral rami;from therethey appearto travelto thenerve cord.No func-tion hasyet been ascribed tothe funnels, however. (2)The atrial papillae ofMüller (1844) were initially thoughtto be excretory innature (asrenal papillae, Willey 1894).The papillae are locatedin thefloor ofthe atrial cavity (Fig.2) andare concentrated inthe vicinity ofthe atriopore. They consistof longitudinal stripsof tall, densely packed cells, manyof which appearto be flagellated primary sensory cells (Bone 1961;Bone etal. 1996; Ruppert 1997). Again, their function is unknown. (3)The encapsulated endingsof Fusari (1889) (seealso Bone 1960b) areformed by free nerve endings surrounded by clusters ofcell nuclei. Theyare locatedin the lateral wallsof the metapleural folds (Figs.2,

4)and maybe mechanoreceptive. (4)The corpuscles ofde Quatrefages (de Quatrefages 1845)are locatedin thecon- nective tissueof the rostrum alongthe branches ofthe firstand second nerves. Theyare typically locatedat branch points, mainlyat distal branches just proximal tothe nerves’ entryinto the subepidermal plexus (Dogiel 1903; Baatrup 1982;Fig. 3).The corpuscles consistof support cellsand peripheral neurons enclosed ina capsuleof connective tissue (Franz 1923).The neurons beara pairof ciliathat project intoa small central cavityand giverise toaxons thatenter the branches ofthe adjacent nerves (Baatrup 1982).The cor- puscles are assumedto be mechanoreceptors sensitive tothe deformation ofthe rostral connective tissue.

Solitary receptors

Solitary receptors arewidely distributed overthe entire epidermis butare most commonin theregion ofthe rostrum, buccal cirri,and tail (Dogiel 1903; Franz 1923;Bone 1960b; Stokesand Holland 1995a; Holland andYu 2002). Theyform small clusters insome instances (Sinnesknospen, Franz 1923; Schulte andRiehl 1977), especially alongthe buccal cirri.The most common receptor celltypes are referred toby convention astypes Iand II (Schulte andRiehl 1977;Bone andBest 1978). TypeI cellsare primary sensory neu-rons withan apical circletof microvilli, asingle cilium, anda basal neurite. Thereare several subtypes, butall areproba- bly mechanosensors (Baatrup 1981; Lacalli andHou 1999). Their axons projectto theCNS viathe dorsal nerves; once there, they travel alongthe cordin twofiber tracts, dorsaland subdorsal inthe terminology ofHolland andYu (2002), whichmay correspond tothe somatosensory and viscero- sensory tractsof Bone (1960a; seeFig. 4).The central axonsof typeI cells reach considerable lengths, soan axon enter-ing theCNS viathe first nervecan typically project caudally to mid-spinal levels,at leastin larvae (Holland andYu 2002). Littleis known aboutthe neurotransmitters released by peripheral neurons, butthere is evidence that at least sometype Icells are GABAergic (Anadón etal. 1998).

TypeII receptors (Fig.4) are secondary sensory cellswith synaptic terminals borneon short basal processes, usually threeper cell (Stokes and Holland 1995a; Lacalli andHou 1999). Apically, theyhave amodified nonmotile cilium sur- roundedby acollar of branched microvilli. This extensive elaboration ofthe apical surface suggestsa chemoreceptive function, but essentially nothingis knownfor certain about chemoreception in amphioxus, eitherin termsof structures or physiology (Lacalli 2004).The synaptic targetsof typeII

cells arenot known,but most likely theyare fibers belong- ingto intramedullary Retzius bipolar cells (Holland andYu 2002).

Additional sensory celltypes reported from larvaemay wellbe presentin adultsas well,but perhaps have simplynot yetbeen observed. Most notable arethe (multi)ciliary spines alongthe oral marginin larvae, each consisting ofa bundleof stiff cilia,one fromeach cellthat contributes tothe spine (Lacalli etal. 1999). These cellsare secondary sense cellsthat synapse withlocal interneurons resident inthe oral plexus, whose axonsthen travelto theCNS. It wouldbe interesting toknow whether this arrangement per-sists tothe adult.A secondway of stiffening ciliais byalter- ingtheir internal support, andthis isseen inciliary spine cells (Lacalli andHou 1999). Ciliary spine cellsare solitary sensory cells presentin small numberson thelarval rostrum, and possibly elsewhere, inwhich the ciliary axonemeis re-placed witha lamellar matrix. Again, these cellsare as-sumed tobe mechanoreceptors.

As described above, therealso are solitary sensory cellsin mostparts ofthe atrial nervous system. Bone (1961)de- scribed various typesof multi-and unipolar cellsthat send neurites intothe CNS(see Fig.4). Itis notclear, however, whether suchcells are actually primary sensory cellsor whether they participate in complicated synaptic chainsof receptors, interneurons, and projection neurons, similar to those described forthe larval oral plexus.

Hatschek’s pit

Hatschek’s pit (corresponding tothe preoral pitof thelarva) isthe central elementof thewheel organ,a systemof ciliated ridges withan accessory feeding function, locatedin theroof ofthe buccal cavity (Figs.1, 3;see Ruppert 1997for adetailed description). Asis thecase withmany other structures, Hatschek’s pitis asymmetrical: thebottom ofthe pitcomes tolie tothe rightof the notochord andpoints to-wards thebase ofthe CNS(Fig. 7C).In some speciesof lancelets (Gorbman 1999; Gorbmanet al.1999) itmay evenbe incontact withthe CNS,a situation strongly reminiscent ofthe hypothalamus–pituitary relationship in vertebrates. Consequently, Hatschek’s pithas been regarded asa good candidate for homology with Rathke’s pouchor theadeno- hyophysis of vertebrates (e.g., Tjoaand Welsch 1974; Nozakiand Gorbman 1992).As discussed elsewhere inthis issue (Sherwood etal. 2005), this hypothesis doesnot re-ceive much support froman analysis ofthe secretions by Hatschek’s pit(the evidence forthe presence oftypical adenohypohyseal hormones is somewhat inconclusive); onthe other hand, pituitary-specific transcription factors areex- pressedin Hatschek’s pitduring development (Candiani and Pestarino 1998).

There are further complications, however. First,the zoneof contact between Hatschek’s pitand theCNS inamphi- oxusis displaced abouttwo myomeres caudalto whereit shouldbe ifit werean exact homologof the vertebrate pitu- itary, since— according tothe gene expression pattensin thelarva —the amphioxus homologof the forebrain wouldbe adjacent tomyomere 1(Shimeld and Holland 2005).The junction of myomeres3 and4 lies, instead, ina regionof theCNS that expresses AmphiHox genes during early larvalde-

velopment. Thus,this regionis amore likely candidate for homology withthe vertebrate hindbrain.

Second,the development of Hatschek’s pit differs fromthat ofthe adenohypophysis of vertebrates, whichis classi- cally supposed tobe of ectodermal (placodal) origin. Hatschek’s pit,on theother hand, arises fromthe leftante- rior diverticulum ofthe endodermal embryonic foregut (see Conklin 1932; Stach 2000). This diverticulum thenopens tothe exterior byfusing witha preoral ectodermal invagin- ation, thus forming thelarval preoral pit.This pitis initially locatedat thelevel ofsomite 1,but thereis nozone ofcon- tactwith thenerve cordat this stage.The preoral pitthen shifts caudally at metamorphosis, alongwith thewhole as- semblage oforal and preoral structures, and finally develops into Hatschek’s pitat the boundary between myomeres3 and4.

Thus,despite claims thatthe adenohypophysis also devel-ops fromthe endodermal foregutin certain craniates (i.e.,in myxinoids, Gorbman and Tamarin 1985), boththe develop- mentof Hatschek’s pitand its position inadult amphioxus are sufficiently different fromthe development and position ofthe adenohypophysis intypical vertebrates towarrant some cautionin interpreting the relationship between thetwo structures. Nevertheless, itis noteworthy thatthe preoral pit differentiates veryearly in amphioxus, sothat itappears tobe functional bythe timethe larvae beginto feed.That func-tion is unknown, but feedingis themain larval activity be-sides swimming atthis early stage, which implies forthe preoral pita rolein either feedingor some related aspectof metabolic processing. Jacobsand Gates (2003) havesug- gested thatthe ancestral adenohypophysis mayhave beenan external sense organthat actedon the internal physiology ofthe animalvia some form of non-neural signaling. If the preoral pitis indeedan adenohypophyseal homolog thatacts inthis way,its involvement infeeding and metabolism in amphioxus would providea rationale forthe central roleof its vertebrate counterpart inthe controlof metabolism and growth.

Theadult CNS

General histological appearance

Mostparts ofthe nervecord are roughly triangular in transverse section, with curved sidesand aconcave basethat restson the notochord (Figs.4, 5).Only the anterior vesicle andthe caudal ampulla aremore orless circular incross sec-tion. Theshape ofthe central canal(= ventricle, ventricular system) varies from regionto region. However, withthe ex- ceptionof the anterior vesicle and the caudal ampulla, it generally hasthe formof avertical slit expanded slightly atthe topand bottom. The ventral expansion, whichruns cau-dally fromthe levelof the infundibular cells, houses Reissner’s fiber.The dorsal expansion is variable, beingpro- nouncedin some regions andabsent inothers, andmay befilled withfluid orwith cellular processes. Inthe slit-like partof the central canal(= intermediate zonein someac- counts), the opposing wallsof thenerve cordare closely apposed andthe canal itselfis almost obliterated. Trans- lumenal processes crossfrom bothsides inthis zone,and the remaining openspace contains ciliathat arisefrom both ependymal cellsand neurons.

Fig. 4. A highly schematic transverse section of amphioxus through the central nervous system and the top portion of the notochord, showing various sensory and motor cell types and the composition of the dorsal nerves. The grey spot in the ventral expansion of the central canal is Reissner’s fiber. Abbreviations and their corresponding terms are listed in Appendix A.

Fig. 5. A highly schematic transverse section of amphioxus through the anterior spinal cord, showing various types of inter- neurons and their characteristic arrangement. The grey spot in the ventral expansion of the central canal is Reissner’s fiber. Abbre- viations and their corresponding terms are listed in Appendix A.

The cell bodiesof mostof the neurons are located closeto the central canal (Figs. 4–7), andmost havean apical pro-cess that connects thecell tothe ventricular surface. Interms of vertebrate neuroanatomy, the neurons thusform adense periventricular layerand most,if notall, ofthe neuronsin thislayer areof the cerebrospinal fluid contacting type.Pop- ulations ofcells variously termed translumenal neurons (Lacalli andKelly 2003a) or commissural cells (e.g., Franz 1923;Bone 1960a) forma special subcategory of cerebro- spinal fluid contacting cells.Not onlydo they contact the ventricular surface, butparts oftheir somata— apical pro- cessesin some instances orthe wholecell bodyin others— lie transversely acrossthe ventricle. Suchcells occurin large numbersin most regionsof theCNS (Figs.4, 5,6D) andin- cludethe largest cellsin thecord, i.e.,the cellsof Rohde(see belowand Fig.5), alongwith other large neurons. Migrated neurons, i.e., those whosecell bodiesare entirely detached fromthe ventricular surface, arescarce andre- stricted tothe IRof the anterior nervecord (see below).

Theamphioxus CNSis not vascularized, andthe axonaland dendritic processes ofits neurons arenot myelinated or otherwise enveloped byglial cells (Meves 1973). Neverthe- less, various typesof glial cellsdo occurin thenerve cord(Bone 1960a; seeFig. 4). Müller’s glia (Müller 1900)con- sistsof groupsof verysmall cellswith intensely staining nu-clei thatare clustered inthe vicinity ofthe dorsal nervesand

in their peripheral branches. Some workers (e.g., Johnston 1905)have interpreted theseas dorsal root ganglion cells; however, asnoted byFranz (1927), theyalso occur between themuscle tails, i.e.,the false ventral roots (Leleet al. 1958),and theylack neurites. Bone (1960a) consequently concluded thatthey were Schwann cell analogues. Insome partsof the peripheral nerves,as wellas themuscle tails, these cellsform incomplete partial sheaths around axonsand muscle tails (Flood 1966, 1974;also seeFig. 8Bin Schulte andRiehl 1977). Thismay bea local phenomenon, however, rather thanbeing widespread throughout thePNS.

A second typeof glialcell with either short processes orno processes is Schneider’s glia (Schneider 1879;Bone 1960a), which linesthe wallsof thedorsal expansion ofthe central canal.A thirdtype ofglia, theradial glia(= ependymal glia,Bone 1960a), lines the ventricular walls. These glia send long, fiber-filled processes through thewhite matterto the connective tissue sheath surrounding thenerve cord,and therethey expandto form terminal endfeet. Thisforms es- sentially anouter limiting membrane aroundthe neural com- ponentsof thenerve cordand provides the mechanical support necessary to maintain thenerve cord’s shape. Sev-eral other glialcell typeshave been reported fromthe larval nervecord (Lacalli andKelly 2002), including one,the axial glia,that appearsto havea transient function inaxonal guid- ance. These cells arise adjacent tothe primary motoneurons, which suggests theycould berelated to vertebrate oligodendrocytes, whichalso develop froma restricted zone adjacent tothe region wherethe primary motoneurons arise.We suggest, therefore, thatthe myelination function mayhave evolved secondarily ina cellline that functioned firstin axonal guidance.

Selected celltypes

Cellsof Rohde

The cellsof Rohde (kollossale Ganglienzellen = Rohde cells,not tobe confused withthe nucleusof Rohde,a ven-tral clusterof neuronsin the anterior IR)are the largest neu-rons inthe nerve cord.As described most thoroughly byRohde (1887), theyare translumenal cellswith alarge soma acrossthe central canaland dendrites that ramifyin the white matteron bothsides ofthe nerve cord.The anterior- mostRohde cellmarks the rostral boundary ofthe spinal cord.Its cellbody isalways locatedin the vicinity ofthe leftsixth nerve (i.e.,at the boundary of myomeres4 and5; seeFigs. 1,3, 7).Its giantaxon (kollossale Faser, Rohde 1887), which arises fromthe leftside ofthe cell, turns ventrally andis themost conspicuous structure inthe ventral midlineof thespinal cord(Fig. 6).It projects caudally tothe levelof thelast myomere (Franz 1923). Muchof theinput tothe dendrites ofthe first Rohdecell appearsto bevia gapjunc- tions (Ruizand Anadón 1989),as onlya few synapses tothem haveyet been found.The giantaxon displays numer-ous en-passant synapses (Ruizand Anadón 1989); processes of somatomotor cells(see below) mightbe postsynaptic tothe giantaxon (Castroet al.2004).

The remaining Rohde cells (Fig.5) aremuch smaller thanthe firstand form separate anterior and posterior groups (Fig.1). The anterior group, excluding thefirst cell, consists ofroughly 15cells inthe regionof myomeres 5–11. Their

cellbodies arein linewith the transverse planes definedby the staggered leftand right dorsal nerves. Cellsize decreases progressively caudally, sothat themost caudalof theante- riorRohde cellsthat canbe identified withany certainty liesat thelevel ofthe right13th nerve (Franz 1923).The cellsgive riseto thick axons alternating tothe leftand right sidesin the anterior seriesof Rohde cells.The axonsturn ven- trally, crossthe midline, andthen travel caudally ina lateral fiber bundle (Fig.5).

Rohdecells reappear atthe levelof myomere 38, adjacent tonerve 39.The posterior Rohde cellsare, on average, much smaller thanthose inthe anterior group,and theyare not aligned withthe dorsal nerves. Thereare between14 (Rohde 1887)and 18(Franz 1923) posterior Rohde cells,the most caudalone beingat thelevel ofmyomere 60.The posterior cellshave mainly ascending axonsthat travelin aventro- lateral tract (Franz 1923,see Fig.5).

Asin thecase ofthe firstand largest Rohde cell, almost nothingis known aboutthe synaptic relationships andthe neurochemistry ofthe restof theRohde cell series.It has, however, recently beenshown thatall theRohde cells,in- cluding their axons,are immunopositive for progesterone (Takedaet al.2003). The significance ofthis findingis en-tirely unclear. Themost onecan currently sayis thatthe Rohde cells, though conspicuous, remain enigmatic.

Organs ofHesse

Theorgans ofHesse (=Hesse organs, dorsal ocelli) are composite photoreceptors (Hesse 1898) locatedin theven- tralpart ofthe periventricular grey (Figs.5, 7).Each organ consists ofa single rhabdomeric photoreceptor cellwhose microvilli are enveloped bya cup-shaped pigment cell (Eakinand Westfall 1962;Ruiz and Anadón 1991c; see Fig.5). The receptor cellsare primary sensory cells, i.e.,each hasan axon (Franz 1923),and these projectto the ventrolateral partof thespinal cord (Guthrie 1975).The first organto formin development isunusual in consisting ofthree cells— two photoreceptors andone pigment cell— andthe formerin thiscase areknown to innervate thedorsal compartment (DC) motoneurons (Lacalli 2002a). Itis notknown whether thisis thecase alsofor therest ofthe series, butit does suggest thatHesse organs have something todo with controlling activities that dependon slow undulations ofthe body,as opposedto thefast onesused forescape swimming. Hesse organsmay not express Pax6 during de- velopment — certainly thefirst doesnot (Glardon et al. 1998)— butthey doexpress S-antigen (arrestin), aprotein typicalof the phototransduction cascadein many animals (Mirshahi etal. 1985),and serotonin ispresent inthe recep-tor cells (Candiani etal. 2001).

Thefirst Hesse organto developis located adjacent tomyomere 5,but the anteriormost inthe adultlies atthe boundary of myomeres3 and4; itis locatedin the posterior partof theIR. Hesse organs become more numerous asone moves caudally intothe spinal cord,and theyhave aten- dencyto cluster opposite the centersof adjacent myomeres (Franz 1923)so asto forma segmental pattern. Moving fur-ther alongthe spinal cord,they first decrease innumber to-wards thecenter ofthe bodyand then increase againmore caudally. The orientation ofthe pigment cupsis not random. Instead, groupsof Hesse organsin certain regionsof thespi-

nalcord pointin particular directions. Franz (1923) under- tookthe tedious workof documenting this;for details, referto hispaper.

Motoneurons, with remarkson muscle innervation patterns

The innervation ofthe myomeric musclesin amphioxus ishighly unusual. The myomeres consistof cross-striated mus-cles arranged ina seriesof chevron-shaped blocks alongthe flank.The muscle cellseach sendlong processes (=muscle tails) towards the ventrolateral marginsof thenerve cord. These processes were formely interpreted, mistakenly, asventral nerve roots. Theirtrue naturewas first recognized byFlood (1966): themuscle processes arethe sites wherethe synapses from motoneurons are received; theaxons andsyn- aptic terminals remain entirely confined withinthe nerve cord;and transmitter release occurs acrossthe basal lamina (Figs.3, 4).The synaptic zonesare serially repeated, onefor each myomere, andare staggered leftto rightin amanner similarto thatof the myomeres. Amphioxus thuslacks any counterpart tothe ventral nerve rootsof vertebrates (Schnei- der1879; Flood 1966, 1968).

The synaptic zonesin each segment consistof two distinct domains, the ventral anddorsal synaptic compartments (Figs.3, 4).Both utilize acetylcholine asa transmitter (Flood 1974).The ventral synaptic compartments arewhere thedeep, anaerobic, fast muscle cells receive their innervation. The presynaptic motoneurons involved belongto aclass ofcells thatBone (1960a) called somatomotor (SM) cells; theymay therefore alsobe called ventral compartment moto- neurons. Theyare foundin the ventral partsof thegrey mat-ter andhave atendency tocluster opposite the synaptic contact zones,and eachhas abroad apical process connect- ingit tothe ventricular cavity. Somehave internal vacuoles, andthis character, together withsize and positional differ- ences,has beenused todefine several subtypes (Bone 1960a; onlyone suchtype, theSM1 cell,is shownin Fig.4). Theaxons ofthe SMcells project laterally intothe bundleof somatomotor fibers adjacent tothe synaptic zoneof theven- tral compartment.

Thedorsal compartment iswhere the superficial, aerobic, slow muscle cellsof the myomeres receive their innervation. TheDC motoneurons areknown from larvae (Lacalli andKelly 1999; Lacalli 2002a) buthave notyet been identified with certainty inadults. Fromthe larval data, however, itseems thatthe wholeof theDC innervation alongthe nervecord mayderive from motoneurons locatedin the anterior cordat thelevel ofsomites 2–6(see below). Thisis approxi- mately equivalent tothe zonefated tobecome theIR ofthe anterior cord, which extends from myotomes2 to4. Tracing experiments by Fritzsch (1996)and Ekhartet al.(2003) have revealed ventrally located cellswith descending projections inthe adultin this region, butit isnot clearwhat typesof cells theseare, oreven whether theyare motoneurons or interneurons. Itis also possible thatlocal commissural cells contribute tothe presynapses ofthe dorsal compartments (Fig.4). The Edinger (EC)and mid-commissural (MC)cells ofBone (1960a) sendaxons intothe bundleof somatomotor fibers adjacent tothe synaptic zoneof thedorsal compart- ments. Bone (1960a) regarded theentire DCsystem asa so-matic sensory system; hence,he classified theEC andMC

cellsas afferent cells.It isfurther possible thatthe adultdor- sal compartment is innervated bya subsetof the somatic motoneuron series, i.e.,one orthe otherof theSM celltypes (Castroet al.2004).

Unlikethe notochord ofany other chordate, the notochord of lancelets isitself amuscle, i.e.,it consists mainlyof spe- cialized striated muscle cells referred toas notochordal lamellae (Welsch 1968c; Flood 1970).The notochordal mus-cle cells contact thenerve cordby meansof processes that emerge fromthe cells dorsally, insmall bundles (notochor- dalhorns, Flood 1970),and piercethe connective tissue sheath separating the notochord andCNS (Fig.4). Thereare many thousands ofthese horns, roughly oneevery 50?m (Flood 1970), arranged intwo rowsto theleft andright ofthe midline. Their contacts withthe baseof theCNS are specialized as postsynaptic swellings thatare apposedto presynaptic terminals insidethe CNS.Thus, thereare seri-ally (butnot segmentally) repeated neurochordal synaptic contacts atthe baseof thenerve cord. Theseare thoughtto be cholinergic (Flood 1970),but their source withinthe nervecord hasnot been identified. Lacalli (2004)has madea tentative suggestion thatthey originate from sensory cells locatedin the rostrum, butthis remainsto beproven.

Theremaining musclesof thebody, besides the myomeric muscles andthe notochord, are innervated bya visceromotor (VM) systemof cellsand fibers. The neurons are distinctive in appearance and location (Bone 1960a), asthey lieimme- diately beneath the ventral expansion ofthe central canaland dorsalto themedial axonof the giant Rohde cell(Fig. 4).This places them ventralto theSM neurons (cf.Figs. 4,7), a situation that differs fromthat in vertebrates, wherethe VMneurons aredorsal toSM neurons (e.g., Nieuwenhuys 1998). Thereare bothlarge (VM1,one per segment) andsmall (VM2,many per segment) VMcells inadult amphioxus. Theyare multipolar, andaxons fromboth types leavethe CNSin thedorsal nerves; similar cellsare foundalso inlarvae (Lacalli andKelly 2002). Candiani etal. (2001) reported thatthe cell bodiesof VM2cells contain se- rotonin, butVM axonal terminals are cholinergic (Flood 1974)and arefound invarious muscles (e.g.,the pterygeal muscle) associated withthe atrium. Though referred toas visceral muscles, mostof themare cross-striated (Franz 1927; Ruppert 1997) rather than smooth. Thereis therefore some question whether theVM systemof lancelets isreally comparable tothe visceral musclesof vertebrates. Abetter comparison maybe, instead, withthe branchiomotor nervesand musclesof vertebrates, as suggested by Fritzsch and Northcutt (1993).

Intramedullary sensory neurons

Asmentioned above, lancelets donot have dorsal root ganglia. Instead, theyhave intramedullary sensory cells comparable tothe Rohon-Beard cells foundin thelarval nervecord of anamniotic vertebrates (Fritzsch and Northcutt 1993). Bone (1960a) distinguished twomajor classes, thedorsal bipolaror Retzius bipolar (RB)cell andthe dorsal root(DR) cell. Thereare two subclasses ofthe formerand threeof the latter.RB cellsare situated inthe most dorsal partof the periventricular greyand haveboth ascending and descending fibers. Thefibers area major component ofthe longitudinal somatosensory tracts, and the peripheral RB

Fig. 6. A schematic lateral view of the rostral end of the nerve cord of a lancelet, including the anterior vesicle plus the anterior and intermediate parts of the intercalated region (A), with cresyl violet-stained transverse sections (B–E) to show arrangement of cells in the transverse plane at the levels indicated in A. The diagram shows, projected onto the midsagittal plane, the various cells and (or) groups of cells that have been identified either by cytoarchitectural criteria, immunocytochemistry, or tracing experiments. Thin arrows indicate the general direction of axonal projections for selected examples; their exact targets are not shown because they are largely unknown. Abbreviations and their corresponding terms are listed in Appendix A.

Fig. 7. A schematic lateral view of the region just behind that shown in Fig. 6, to show the posterior part of the intercalated region

(A) and cresyl violet-stained transverse sections (B–D) at the levels indicated in A. Refer to Fig. 6 for details and an explanation of symbols. The black structures in the ventral part of the nerve cord are the pigment cells belonging to the Hesse organs.

processes originate as branches fromthem andenter thedor- salnerve. Thereare fewerDR cellsand mostof themhave translumenal processes (Fig.4).

As to function, theusual assumption isthat theRB andDR axonsare themost probable sourcesof receptive sensory fibersin the various peripheral plexuses andthe skin.Pre- sumably some branchand terminate asfree endings, while othersare postsynaptic to secondary receptor cells, wherever those occur. Thisis alogical supposition, andthere iscir- cumstantial evidence onthe innervation ofthe larval rostrum thattends tosupport it (Lacalli 2002b, 2004).It hasin prac- tice, however, been impossible toprove. The necessary trac-ing studies havenot yetbeen performed, for example, sowe haveno detailed information onthe peripheral connections ofeither RBor DRcells, noris itcertain thattheir processes actually travelany distance outside thecord. Thereis recent immunocytochemical evidence thatsome DRcells contain

?-aminobutyric acid (GABA) (Anadón etal. 1998),but oth- erwise nothingis known about their transmitters.


Themiddle (i.e., intermediate) zoneof the periventricular grey(Fig. 5)is occupied bya varietyof cellsthat Bone (1960a) regarded as interneurons. These include theRohde cellsand, more dorsally, translumenal dorsal commissural cellsand small dorsal cells.The latterare relatedto thesen- sory systemin thatthey sendsmall processes into somato- sensory tracts. Bone (1960a) thought thatboth types might contribute axonsto thedorsal nervesin some cases.The ECand MCcells arefound more ventrally. As mentioned above, theymay have something todo withthe DCmotor system. TheMC cellsare translumenal neuronsof anunusual type:the cellbody itself lies acrossthe central canal,and shortex- pansions project directly intothe canal itself (Figs.4, 5).

GABA, neuropeptide Y,and several other neuropeptides havebeen detected invarious cells loosely classified as interneurons (Uemuraet al.1994; Anadónet al.1998; Cas-tro etal. 2003),but this clearly represents avery diverse as- semblage of disparate celltypes about whichvery littleis known.

Three other interneuronal celltypes deserve mention. First,the cellsof Johnston (Bone 1960a) occur segmentally between thedorsal nerves, oneon eachside, andhave along process that projects tothe dorsal expansion ofthe central canaland contacts thefluid it contains (Fig.5). Their func-tion is entirely unknown. Second,a novel classof interneurons, Anadón’s cells,has been identified inthe vi-cinity ofthe ventral expansion ofthe central canal (Anadón etal. 1998;see Fig.5). Theseare verysmall GABAergic cells interspersed between thecell bodiesof SMand VMneurons (cf.Figs. 4,5). Anadónet al.(1998) have suggested thatthey mightbe comparable tothe inhibitory Renshaw cellsof vertebrates. Notably, these small cerebrospinal fluid contacting cellsdo notseem tobe identical toa third groupof interneurons, the ventral longitudinal cellsof Bone(Bone 1960a; Fig.5). Thecell bodiesof thisthird groupare alsofound between thoseof themotor neurons (cf.Figs. 4,5) butare much larger thanthose of Anadón’s cellsand arede- tached fromthe ventricle. Theyhave dorsal, ascending, and descending processes, andBone (1960a) thought thatthey also coordinate the activity ofthe motor neurons, in particu- larthat ofthe VMcells.

The anterior nervecord indetail

The region anterior tothe firstcell ofRohde has attracted less attention in cytological studies thanthe spinal cord, thoughthe two differ,as notedby Bone (1960a), interms ofboth theircell typesand their general organization. Whatis clearly lackingis a descriptive study comparable toBone’s comprehensive analysis ofthe spinal cord, citedso exten- sively above.The closestis thatof Ekhartet al. (2003), butthis isfocused moreon cell grouping, cytoarchitecture, and general morphology, while providing muchless information on individual cell types.

Anterior vesicle

The anterior vesicle (Fig.6) corresponds tothe anterior partof thelarval cerebral vesicle(= anterior CV;see belowand Fig.8). The anterior vesicle hasa central ventricular space(= central canal) but, unlike other regionsof thenerve cord,this is relatively wideand lacksthe translumenal cell processes found elsewhere inthe cord. Dorsally, onthe left-hand side, thereis aremnant ofthe anterior neuropore inthe formof aciliated pit.This is Kölliker’s pit (Kölliker 1843; Franz 1923), whichhas been ascribed an olfactory function, though without direct evidence. Infact, the neuropore re-mains anopen channel, evenin theadult (Valletet al.1985). However, itis sonarrow and clogged withcilia thatone can question how effectively itacts asa connection between the central canaland the outside. Theouter pitlacks obvious re-ceptor cells,nor arethere anynerve fibersin evidence toconnect itto the anterior vesicle (Edinger 1906;Tjoa andWelsch 1974). Thistends to reinforce the conclusion thatal- though Kolliker’s pitmay contain some specialized cell types,it isnot asense organ.

The caudalend ofthe anterior vesicleis marked dorsally bythe beginning ofthe axial columnsof Joseph cells,as wellas scattered lamellar cells,and ventrally bythe infundi- bular organ.The infundibular organ consists ofseveral rowsof columnar cellsthat secrete Reissner’s fiber.It isthough to correspond tothe flexural and subcommissural organsof vertebrates (Obermüller-Wilén 1976;see Nieuwenhuys 1998for details).

Thewalls ofthe anterior vesicle areformed byclosely packed, ciliated epithelial cells, among whichare thepig- mentcells ofthe frontal eye.While this regionhas beenex- aminedin larvaein great detailat theEM level(see below), theonly EMstudy ofthe adult anterior vesicleis thatof Meves (1973), whichis muchless complete. Mevesob- served rowsof densely stained cells (Fig.6), immediately ventral andcaudal tothe pigment cellsof the frontal eye,that may correspond tothe receptors and neuronsof thelar- val frontal eye.More caudally, inthe preinfundibular zone, i.e., between the frontal eyeand the infundibular organ,she foundtwo typesof cells,the more densely stainedof whichhad basal processes. Thesemay correspond tosome ofthe larval preinfundibular projection neurons described byLacalli andKelly (2000), including thecells ofthe balance organ. These authors found evidence forthe latter structure innewly metamorphosed juveniles, soit may persistto theadult stage.

Whetherby light microscopy (e.g., Edinger 1906; Franz 1923; Ekhartet al.2003) orEM (Meves 1973),it is difficult todiscern muchabout the neuronal andglial cellsof thean- terior vesicle, sincemost cellsare smalland rather densely stained andhave few visible distinguishing features. Franz (1923, 1927) therefore concluded thatthe entire anterior ves-icle consisted onlyof glial cells. However, GABAergic (Anadón etal. 1998)and serotoninergic neurons (Moretet al.2004) havesince been identified inthis regionin adult specimens, andsince thesame regionin thelarva iswell supplied with neurons, itmust continue to function ina neu-ral capacity throughout the animal’s life, though whatit doesin theadult isnot evident.

Thefirst nerveof thedorsal nerve series(= nerve 1, rostral nerve) entersthe anterior vesicle ventrally atits rostral margin. The nerve’s fibers then spreadout toform athin veilof white matter that coversthe lateral and ventrolateral as-pects ofthe anterior vesicle. Peripherally, thisfirst pairof nerves traverses thetop ofthe notochord, enclosed withinthe connective tissue sheath that surrounds the latter. Nearthe tipof the notochord, thenerves exitthe sheathand ram-ify underneath theskin, sothe innervation isvery muchre- stricted tojust thetip ofthe rostrum. Thisis an important point, becauseif thereare anyas yet undescribed sensory celltypes or subtypes thatoccur onlyat thetip ofthe ros-trum, itis very likely thatthey would enterthe cord bymeans ofthe first nerve, whichis exceptional anywayfor notbeing, strictly speaking, adorsal nerve,so itsCNS tar-gets may also differ from those of the rest of the dorsal nerve series. Among known fibersin thefirst nerveare ax-ons from peripheral primary sensory cells, mainly typeI mechanoreceptors, and centrally derived axonsfrom RBcells and possibly otherCNS cells.

Thesecond nerveof thedorsal series(= nerve2, anterodorsal nerve) entersthe CNS dorsally, atthe junction

betweenthe anterior vesicle andthe IR.Unlike thefirst nerve,the secondis not associated withthe notochord, butturns laterally andenters the subepidermal connective tissue. Thesecond nerve ramifies extensively inthe more caudal partsof the rostrum, and several anastomoses areformed withthe branches ofthe first nerve (Franz 1923,Fig. 3).The second nerve probably carries fiber types similarto thoseof the rostral nerves.

Theperipheral branches ofboth thefirst andthe second nerves bear numerous small swellings, the corpuscles ofde Quatrefages described above.The centripetal axons derived fromthese enterthe distal branches ofboth nerves1 and2, butit isnot known whether they actually enterthe cordin both nervesor onlyone. Someof the carbocyanine dye(DiI) tracing datahave been interpreted as indicating thatthe cor- puscles connectto theCNS mainlyvia nerve2 (Fritzsch 1996;B. Fritzsch, personal communication), whichis in- triguing but requires confirmation. The central targetsof ax-ons fromthe corpuscles are unknown.

The intercalated region (IR)

The region between thecaudal endof the anterior vesicle andthe first giant Rohdecell has sometimes been regarded asa partof thespinal cord (e.g., Franz 1927),but itclearly hasunique features that justify itsbeing regarded asa spe- cialized partof theCNS (seealso Bone 1960a; Fritzsch 1996).The most prominent ofsuch features arethe dorsal Joseph cells, present throughout mostof thiszone, anda conspicuous ventral groupof cells,the nucleusof Rohde,in the posterior partof theIR. TheIR isnot uniform alongits lengthbut instead canbe subdivided intothree partson cytoarchitectural grounds. The description ofthese thatfol- lowsis based mainlyon Ekhartet al.(2003) andsome more recent unpublished findings (H.Wicht, personal observa- tions).

Anterior IR

Theanterior partof theIR islocated adjacent tothe first myomere (Fig.6) and probably corresponds tothe posterior partof thelarval cerebral vesicle. Asin more posterior partsof theCNS, the central partof the ventricle is slit-like and traversed by cellular processes from translumenal neurons. Ventrally, asmall expansion ofthe ventricular cavity houses Reissner’s fiber. Dorsally, thereis another expansion partly filled with processes belonging to lamellar cells (Meves 1973; Ekhartet al.2003), aform ofciliary photoreceptor (Ruizand Anadón 1991b), thougha compact lamellar bodylike thatseen inthe larvais nolonger present. Thedorsal as-pect ofthe anterior partof theIR iscapped bythe rhabdomeric Joseph cells (Welsch 1968a; Ruizand Anadón 1991b, 1991c). These increase innumber ina rostrocaudal direction. Aswith theHesse organs, Joseph cellsdo notex- pressPax6 during development butdo containa rhodopsin- like protein (Watanabe and Yoshida 1986),so theyare al-most certainly photoreceptors. The ventral and lateral as-pects ofthe anterior IRare coveredby a relatively thick layerof white matter.

Neuronsand glial cellsare easily distinguished inthis re-gion. The neuronal cell bodiesare mostly locatedin the periventricular greyclose tothe central ventricular slit. Compared with neuronsin more posterior regionsof theCNS, these cellsare relatively small.In the ventral partof

the periventricular grey, however, justabove the ventral ex- pansionof the central canal, thereare individual large cellsthat resemble the somatic motoneurons (SMcells) described byBone (1960a) from spinal cord. Basedon thelarval data (e.g., Lacalli andKelly 2003b), several classesof ventral interneurons with descending axons wouldalso be expected toreside inthis area.

Immunocytochemical studiesby Holland and Holland (1993), Uemuraet al. (1994), Anadónet al. (1998), Castroet al. (2003), andMoret etal. (2004), and tracing studiesby Fritzsch (1996)and Ekhartet al. (2003), allshow thatthe periventricular greyof the anterior IR contains distinctive celltypes and groupings besides those evidentin standard histological preparations. First,in the periventricular grey, thereare numerous translumenal cells whose axons projectto thespinal cord (Ekhartet al.2003). Then,just ventralto theJoseph cellsand surrounding thedorsal expansion ofthe central canal, thereare bilateral, longitudinal bandsof neu-rons immunoreactive for urotensin and FMRFamide (Uemuraet al.1994), GABA (Anadón etal. 1998), neuropeptide Y(Castro etal. 2003),and catecholamines (the catecholamin- ergic population Iof Moretet al. (2004)). Axonsfrom someof these catecholamine-containing cells travel anteriorly intothe anterior vesicle, butfibers frommost ofthe othercell typesare directed ventrally intothe lateral and ventral neuropile, wherethey forma dense commissure orplexus beneath the central canal.

Unlikethe situation elsewhere inthe CNS,there aregroups ofneurons inthe anterior IRthat detach fromthe ependymal layerand migrate intothe white matter.In stan-dard sections ofthe anterior IRthey areseen inthe ventral midline, belowthe ventral expansion of the central canal(the anteroventral migrated (avm) group, Fig.6B), andbilat- erallyin thedorsal and lateral partsof thewhite matter (the anterolateral migrated (alm) group, Fig.6B; seeEkhart etal. 2003for details). Thereis no information onthe natureof theavm cells,but recently published data,as wellas some personal observations of immunocytochemical preparations ofB. lanceolatum, have yielded interesting detailson thealm cell group. Firstly, themore posterior almcells (slightly rostralto the junction of myomeres1 and2; black squaresin Fig.6) seemto correspond tothe anterolateral serotoninergic cellsof Holland and Holland (1993) thatwere also observed byMoret etal. (2004). Slightly more anterior (black circlesin Fig.6) isanother groupof immunocytochemically identi- fiable cells withinthe almgroup. Thisis the catechola- minergic population IIof Moretet al. (2004). Thereis some uncertainty aboutthe exact positions ofthese twocell groups, however. Moretet al.(2004) placethem adjacent tothe rostral halfof thesecond myomere. Inan independent immunocytochemical study,H. Wicht (unpublished data)lo- calized themmore anteriorly, adjacent tomyomere 1and thus withinthe confines ofthe almgroup (seeFig. 6). Wicht’s studydid confirm, however, thatboth the catecholaminergic population IIand the anterolateral sertoninergic neurons havelong descending projections tothe spinal cord.In retro- grade tracing experiments, Fritzsch (1996) found pairsof la-belled cellsin late larvae thatmay correspond tothe anterolateral serotoninergic cells, even thoughhe didnot specify their exact position, butEkhart etal. (2003),in asimilar studyin adults, didnot findsuch cells. Assuming the

latter resultis afalse negative, thecells and projections ap-pear tobe real;it isonly their exact axial position thatis amatter ofsome uncertainty.

Intermediate IR

The intermediate IR(Fig. 6)is located adjacent tomyomere 2,i.e., between dorsal nerves3 and4. Itis charac- terizedby having several layersof Joseph cells.The dorsal expansion ofthe central canalis lackingin this region, andthe canalhas instead theshape ofan inverted keyhole. The lateral and ventral migrated cell groups thatoccur inthe an-terior partof theIR arealso absent. Instead, thereis another such group (the posterodorsal migrated (pdm) group, Fig.6C) located just ventralto theJoseph cells. Cellsin thisgroup probably correspond tothe Bcells described byBone (1959, 1960a), which represent the anteriormost clusterof somatosensory RBcells found elsewhere alongthe nerve cord.A groupof small, ventral spindle-shaped cells(vsc inFig. 6)is found directly beneath the ventral expansion ofthe central canal.A numberof large cells, presumably motoneurons ofone kindor another, arefound inthe ventral partof the periventricular grey,and someof thesesend de- scending projections tothe spinal cord (Ekhartet al.2003). Therealso isa single largeVM neuronin the ventral midline just posterior tothe ventral spindle-shaped cellgroup (H.Wicht, unpublished data;see Fig.6E). Bone (1960a) claimed thatVM neurons were present onlyin thespinal cord,but theremust beat least enoughin theIR tosupply dorsal nerves 3–7, which innervate thelabial muscles andother buccal structures. In addition, thereare somevery large translumenal neuronsin this region (lac,Fig. 6D)at approximately mid-level inthe periventricular grey.They probably correspond tothe giant translumenal cellsde- scribedby Lacalli andKelly (2003b) from late-stage larvaeand the largestof the FMRFamide-positive cellsof Uemuraet al. (1994). Axonsfrom atleast someof thelarge translumenal neurons projectto thespinal cord (Ekhartet al.2003).

The periventricular greyof the intermediate IRalso con-tains neurons positive forGABA (Anadón etal. 1998)and peptides (Uemuraet al.1994), butthe pattern differs fromthat further forward. Specifically, neuronsof corresponding typeare shifted more ventrally compared withthe anterior partof theIR.

The periventricular greyin this region also contains translumenal neurons with descending projections (Ekhartet al.2003). These decrease in frequency towards themiddle andcaudal partsof the intermediate IR.From studieson lar-vae (see below),a major locomotory control center, thepri- marymotor center (PMC),is expected toreside somewhere inthis region, beginning roughlyat the anterior tipof myomere2 (Fig.6, shaded region). Thereis no obvious adult counterpart tothe PMCat this location, buteven ifthe larval neurons persistto theadult, thereare relatively fewof themand theycould easilybe overlooked.

Posterior IR

The posterior IR(Fig. 7)is located adjacent to myomeres3 and4, between dorsal nerves4 and6. Itsmost conspicu- ous featureis the nucleusof Rohde (Ekhartet al.2003, first described byRohde in1887). Thisis an agglomeration of

relativelylarge cellswith intensely staining cytoplasm (Figs.7B, 7D)that surround the ventral expansion ofthe central canal.The rostrocaudal extentof Rohde’s nucleus coincides withthat ofthe columnar epithelium ofthe wheel organand Hatschek’s pit, locatedin theroof ofthe buccal cavity. Thetip of Hatschek’s pit extendsto theside ofthe notochord and projects towards thebase ofthe CNSat the junction of myomeres3 and4 (Fig.7C). Above,we have discussed the evidence for homology between Hatschek’s pitand the vertebrate adenohypophysis. Inlight ofthis hypothe- sis,it isof course tempting to speculate that Rohde’s nu-cleus isthe equivalent ofthe neurosecretory hypothalamic cell groupsof vertebrates; however, asin thecase ofthe adenohypophyis, the evidence isso farnot very convincing. The intensely staining cytoplasm andthe large amountsof Nissl substance withinit actually point towardsan intense secretory activity; however, noneof the neuropeptides typi-cal of vertebrate hypothalamic endocrine cellshave beenob- served withinthe confines ofthe nucleusso far. Similarly, nothingis known aboutthe axonal projections ofthese cells,so itis notclear whether theyare thesource ofthe fibers that,from circumstantial evidence, appearto innervate Hatschek’s pit(Tjoa andWelsch 1974).

Theposterior partof theIR isalso wherethe rostralmost Hesse organs occur, while dorsally theJoseph cells vanish, roughlyat the boundary between myomeres3 and4. Thedorsal partof the central canal displays some isolated, bubble-shaped expansions, butthese arefilled withfluid rather thancell processes, andtheir functional significance isnot clear.

Theposterior IR contains onlya few migrated cells.The columnof pdmcells that originates farther forward (seeFig. 7) continues through this region intothe spinal cord. However, roughly coincident withthe lastof the Joseph cells,the pdm columns move medially soas to effectively mergewith the periventricular grey.

The periventricular greyof the posterior IR, flanking the slit-like partof the central canal, contains numerous trans- lumenal neurons, manyof whose axons descendto thespi- nalcord, including several especially large examples (Ekhartet al.2003). Inthe ventral partof the periventricular grey, immediately dorsalto the nucleusof Rohde (Fig.7), thereare afew large neuronsof the motoneuron series, someof whichhave descending projections (Ekhartet al.2003). Serotonin-containing neurons areabsent in this region (Moretet al.2004), buta relatively large numberof GABAergic and peptidergic cells (Uemuraet al.1994; Anadónet al.1998; Castroet al.2003) dooccur. In addition, thereare four relatively large catecholaminergic cells (popu- lationIII ofMoret etal. 2004)with translumenal processes inthe vicinity ofthe rootsof thefifth dorsal nerves.

The larval nervous system

Earlydevelopment in amphioxus resembles thatof atypi- calmarine invertebrate: theegg issmall, ca.120 ?m indi- ameter, and hatches after gastrulation asa free-swimming, but nonfeeding, ciliated larva. Thisstage undergoes neurulation and elongates toproduce amore typical chordate-type larva, roughly 1.3mm inlength, witha notochord, somites, dorsal nerve cord,and one pharyngeal

slit, thefirst ofthe series. Thelarvae thenfeed andgrow inthe plankton until, after about30 daysunder optimal condi- tions, they metamorphose to juveniles ca.5 mmlong. Excel- lent surveysof the morphological changes thatoccur during development havebeen published byHirakow andKajita (1990, 1991, 1994)and Stokesand Holland (1995a). Mostof therecent larval research hasbeen doneon different spe-cies thanthose usedfor classical anatomical studiesof theadult, notably Branchiostoma floridae (Hubbs, 1922)in North America and Branchiostoma belcheri (Gray, 1847)in Asia.The neuroanatomical differences between species ap-pear minimal, thoughthe differences observed inthe extentof theJoseph cell columns (H. Uemura, personal communi- cation) indicate thatsome cautionis required when compar- ingdata across species.

Amphioxus larvaeare wellknown fortheir asymmetry, themouth being initially onthe leftside andthe pharyngeal slitson theright. This arrangement necessitates amajor re- positioning of structures at metamorphosis, especially ofthe mouth, which shifts caudally overa distance ofseveral so-mites, andthe pharyngeal slits. Thisis adramatic process, achieved largely through differential growth, andhas at- tracteda gooddeal of attention inthe past.What isperhaps less appreciated isthe magnitude ofthe size increase that occurs duringthe larval phase,for boththe bodyas awhole andthe internal structures aswell. Thenerve cord,for exam-ple, increases froman initial diameter of 15 ?m to ca.100 ?m at metamorphosis. Bythen, thereis substantially more sensory input,so thedorsal rootsare largerand thecord hasa greater diversity of neuronal celltypes andmuch more neuropile, implying greater integrative capacity. Corre- latedwith this,the animal’s behavior becomes muchmore complex.A greatdeal ofneural development thus takes place duringthe larval phase,and the nervous system conse- quently looks quite different, evenat afairly gross anatomi- callevel, depending onthe stage examined. Reconstructing the developmental events responsible for generating achar- acteristic adult neuroanatomy thus requires thata rangeof stagesbe examined, spanninga periodof weeks,but thishas seldom been achieved in practice. Our knowledge ofneural development is therefore somewhat fragmentary, andonly provisional conclusions canbe drawnin many instances.

The first detailed description oflarval neuroanatomy wasthat ofBone (1959), whomade the prescient remark thatlar- vaeare likelyto bemore revealing about phylogenetic issues thanthe adult. Bone’s study, and much subsequent larvalwork, concentrated onrather late stages, since theseare better subjects formost staining techniques andare easierto handle thanyoung larvae. Theyoung stagesare nevertheless themost important onesin termsof understanding earlypat- ternsof neural differentiation andtract formation andfor comparison withthe rapidly increasing bodyof gene expres- siondata (Holland and Holland 1999; Shimeld and Holland 2005). Recent relevant workon young larvae includes stud-ies of peripheral innervation bymeans ofwhole mount immunostaining andDiI tracing (Yasuiet al.1998; Kajiet al.2001; Holland andYu 2002),as wellas adetailed studyof the internal microanatomy ofthe anterior nervecord atthe EMlevel, using serial sections and three-dimensional re- construction (e.g., Lacalliet al.1994; Lacalli 1996, 2002a, 2002b; Lacalli andKelly 1999, 2000, 2003a, 2003b). The

latterstudy, in particular, has revealed muchnew information of interest, fromboth a neuroanatomical andan evolutionary perspective, andis themain focusof theac- countthat follows.

Anteriornerve cord

Several distinctive subdomains canbe recognized inthe anterior nerve cord. Externally, aslightly bulbous anterior zone,the cerebral vesicle (CV),is easily distinguished fromthe restof thecord. The transition occurs slightly forwardof the boundary between somites1 and2. Internally, however, themain anatomical landmark isthe transition inthe shapeof the central canalof thecord that occursat alevel coinci- dentwith the ventral clusterof infundibular cells (Lacalli etal. 1994; Lacalli 1996;see Fig.8). Forwardof thispoint isthe anterior CV, essentially thelarval equivalent ofthe adult anterior vesicle. Herethe central canalis cylindrical andlacks afloor plate,and mostcilia project forward, someeven escaping outthe neuropore. The principal landmarks ofthe anterior CVare the frontal eye, locatedat the anterior tipof thecord, anda putative balance organ (Lacalli andKelly 2000), positioned justin frontof the infundibular cells. Mostof the neurons are ventral, columnar, and closely packed, es- pecially inthe immediate preinfundibular region, andhave caudally projecting neurites.

Theposterior partof theCV (probably corresponding tothe anterior partof theIR ofadults) beginsat the infundi- bular cells.The floor plate begins here,the central canalnar- rowsto aslit shaped likean inverted keyhole, sothat onlythe ventral portion remains open,and thecilia ofthe ependyma andfloor plate project backwards. Thedorsal partof thecord inthis zoneis occupied byan ovoidmass ofcili- ary lamellae, which together constitute the lamellar body.The corresponding ventral partof thecord istaken upfirst bya postinfundibular (= tegmental) neuropile and, behind this,the primary motor center (PMC), containing the anteriormost motoneurons andsets of interneurons withcau- dal projections thatare involved in locomotory control (Lacalli 1996; Lacalli andKelly 2003b). The transition fromthis regionto atype of organization more typicalof therest ofthe nerve cord, whatever that entails, appearsto bea grad-ual one,with noobvious landmarks (e.g.,no Rohde cells)to indicate whereit occurs. However, thecaudal limitof the lamellar body, which extends almostto the boundary be-tween somites1 and2 in late-stage larvae, andthe similar extentof Otx expression suggest thatthere is something dis- tinctive aboutthe nervecord toabout thelevel ofthe bound-ary between somites1 and2 or slightly beyond.

Despitethe usefulness ofthe infundibular cellsas anatom- ical markers, thereis noobvious transition interms of neuronal celltype atthis point. Instead, cellsof essentially anterior character arefound fromthe preinfundibular regionto the beginning ofthe PMC. "Anterior" here refersto cellswith irregular basal neurites thatform repeated varicosities containing mixed vesicle typesand few,if any, synapses. Theseare features thatare generally associated withslow transmission, often involving neuropeptides (Burnsand Au- gustine 1995). Beginning inthe PMC,most ofthe neurons have well-defined axonsand separate dendritic structures, ei-ther arborsor spines (both occur), and synaptic junctions, often with clear vesicles, predominate. This implies fast

Fig. 8. (A) Side view of the head of an amphioxus larva showing the position of the nerve cord (nc) in relation to the notochord (shaded). The head is highly asymmetric, with the mouth (heavy outline) on the left side and the pharyngeal (gill) slits (lighter out- lines) on the right. (B) Oblique dorsal view of the anterior nerve cord showing its main landmarks and selected cells. Asterisks indi- cate the third pair of large paired neurons (LPN3s), which are putative locomotory pacemaker neurons. Landmarks include the frontal eye (fe), infundibular organ (io), lamellar body (lb), and primary motor center (PMC); the zone of neuromuscular junctions is shaded. This view extends to just beyond the boundary between somites 1 and 2, roughly the extent of the cerebral vesicle as anatomically de- fined and coextensive with the zone of Otx expression. The io marks the junction between the anterior and posterior parts of the cere- bral vesicle. The two regions differ in the shape of the central canal and the direction in which most cilia project into it. Other abbreviations and their corresponding terms are listed in Appendix A. See text for further details. Modified from Lacalli (1996).

transmissionand aminergic oramino acid transmitters, whichis perhaps logical for neurons directly involved inthe locomotory control circuits.

The veryfact that distinctive regions canbe recognized withinthe anterior cord, however thisis defined anatomi- cally, raisesthe question ofhow the subdivision ofthe ante-rior cordis controlled atthe molecular level.The question is addressed indetail elsewhere inthis symposium (Shimeld and Holland 2005),but afew remarks areuseful here. Genessuch asOtx, expressed throughout mostof theCV, presum- ably definethe character ofthat zoneby some means, thoughhow isstill poorly understood. Butthis doesnot ex-plain how specific structures within eachzone are specified; for example, the frontal eyeand infundibular cells. Sig- nalling from boundaries isone possibility, andthe anterior boundary ofthe neural plate (ANB)is agood candidate, sinceit isnow recognized asan important signalling centerin vertebrates (Grove 2002).A similar rolefor theANB in amphioxus wouldhelp explain the organization ofthe fron-tal eye,in which cellsof liketype are precisely oriented inthe transverse plane, parallel tothe ANB.The infundibular regionis asecond possible candidate, notleast because cellsof similar type, including subtypes amongthe various prein- fundibular, parainfundibular, and tegmental neurons, occurat roughly equal distances inboth directions from this site. Anygene expressed earlyin apattern centered inthis regionis thenof potential interest. FoxD isone suchgene (Yuet al.2002), and further research mayturn upmore, sincewe are

still ata veryearly stagein understanding how regional sub- division of amphioxus CNSis controlled.

Frontal eye

Thefrontal eye consists ofa pigment cup, oriented soit opens dorsally, andfour rowsof neurons. Thefirst tworows consistof simple sensory neurons,6 inthe firstrow and10 inthe second, withcilia that project outthe neuropore, andbasal axons.The tworows differin termsof theextent andtype oftheir varicosities, butboth projectto the ventrolateral tractsand continue through the anterior CV,but probably notmuch further. Their close association withthe pigment cup indicates thatthese cellsare probably photoreceptors, though thishas notbeen tested experimentally. Behindthe putative photoreceptors arethe othertwo rowsof neurons. Thefirst ofthese rows(row 3) consists of6 cellswith multiple pro- cesses, typically short,by meansof whichthe cellsform multiple pointsof contact witheach other.In thefourth row,only thetwo most medial neurons showa close association withthe frontal eye.They havebasal neurites that communi- cate synaptically viatwo routes (one anterior, one at the levelof the postinfundibular neuropile) withthe dendrites ofthe thirdpair oflarge paired neurons (LPN3 cells), whichare key components ofthe locomotory control center (see below). This arrangement of photoreceptors and neurons hasbeen compared tothe vertebrate retinaby Lacalli (1996), who suggested possible homology between cellsin rows3 and4 and retinal amacrine and bipolar cells, respectively. A

further pointof similarity between the frontal eyeand thepaired eyesof vertebrates isthat both developat the anterior marginof theneural platein whatis essentially aventro- medial position. Also, projections inboth casesare tore- gions caudalto the infundibulum, toroughly midbrain level.The argument for homology isthus reasonably strong. How-ever, whilethe vertebrate retinahas a two-dimensional arrayof photoreceptors, cellsin the frontal eyeform strictly one- dimensional files,and thereis no evidence thatthis isa sec- ondarily degenerate condition, i.e.,that amphioxus everhad an image-forming eye.

Behavioral experiments showthat amphioxus larvaecan orientto light while suspended and feedingat thewater surface, probably by modulating ciliary beat on the body surface (Stokes and Holland 1995b). The frontal eyeis im- plicated inthis, thoughan appropriate neurociliary effector pathway hasyet tobe demonstrated. Asthe larva grows,the pigment spot enlarges somewhat, butthe complement of photoreceptors and neurons appearsto change very little. The function ofthe frontal eyein theadult isnot known. Further information onthis andother amphioxus photoreceptor systems canbe foundin Ruppert (1997)and Lacalli (2004).

Infundibular region

The infundibular cellsare secretory cells rather thanneu- rons,but theylie withina ventral massof about80 closely packed neurons that resemble primary sensory cells. Thereare avariety of subtypes among these.The most distinctive (14cells) have expanded, club-shaped cilia, which suggests theymay function todetect displacement, i.e.,as abalance organ (Lacalli andKelly 2000). Likeaxons ofmany ofthe surrounding cells, axonsfrom this putative balance organ projectto the postinfundibular neuropile and terminate in large varicosities. Amongthe other neuronal subtypes inthis regionare three classesof preinfundibular projection neu-rons (PPN1–3) withmixed clearand dense-core vesicles and comparatively short axons (Lacalli andKelly 2003b); four cells,the PPN2s, haveclear vesicles andaxons that travelat leastto somite7 and possibly farther (Lacalli 2002a). In general, fromthe paucityof synaptic specializations withinthe postinfundibular neuropile, itappears that paracrine re-lease isthe predominant modeof transmission, suggesting thatthis regionis mainlya modulatory center.

The lamellar bodyis thesecond major contributor tothe postinfundibular neuropile. Eachof itscells hasa single largeaxon that travels downthe sideof thecord tothe neuropile, wherea tangled massof subsidiary branches isformed (Lacalli etal. 1994).The lamellar bodyis generally accepted asa homologof the vertebrate pineal organ, which suggests that eitherthe cells themselves ortheir downstream targetsin the neuropile generatea circadian rhythm. Thereis nodirect experimental evidence forsuch rhythmsin either adultsor larvae, butthe larvae have diurnal patterns ofverti- cal migration inthe plankton undersome conditions (Wickstead andBone 1959), which implies the presence ofa circadian clock.

Primary motor center

ThePMC contains the anteriormost motoneurons inthe cordand anumber oflarge premotor interneurons. Thesecell types occur elsewhere inthe cord,but notin sucha

largecluster. The important cells, froman organizational standpoint, arethree pairsof large paired neurons (LPNs). Theseare extensively innervated bysensory inputs, bothdi- rectlyby primary sensory cellsin the periphery andby syn-apses fromthe anteriormost RBcells (=B cellsof Bone1959, aRBcells inFig. 8).The third pair,the LPN3s,are themost important andare cross-innervated ina bilaterally sym- metrical fashion, an indication thatthey maybe mutually in- hibitory andhence capableof pacemaker function (Lacalli 1996; Lacalli andKelly 2003b). Their outputis toventral compartment (VC,or fast) motoneurons via synapses andto DC(slow) motoneurons viaan unusual classof intercellular junctions (juxtareticular junctions; see Lacalli 2002a).

TheLPN3s arethus thebest candidates for neurons exert-ing adirect controlling influence overboth fastand slow swimming, which appearto havea similar neuromuscular basisin amphioxus and vertebrates (Bone 1989). Fastor es-cape swimming occursin response tosensory inputs, whichare amassive and redundant inputto theVC system. TheVC system also receives synaptic inputfrom fibersin the postinfundibular neuropile andmay besubject to additional paracrine inputas well,via fibers passing through the neuropile, allof which provides an opportunity to modulate the response tosensory stimuli. In contrast, theslow system, which drives vertical migration, isalmost devoidof synaptic input. Besides itslink via junctions tothe LPN3s, thispath- way seems to be mainly under the control of the PPN2s mentioned above,a classof preinfundibular projection neu-rons thatmake repeated junctional contacts withthe axonsof theDC motoneurons. Whatthis meansin functional termsis notclear, butthe circuitry (Fig.9) suggestsa switching deviceof somekind. Perhaps theescape response issup- pressed during migration, which might itselfbe under circa- dian control. Toassess such proposals, however, muchmore information isneeded onthe natureof the various typesof preinfundibular neurons thanis currently available.

VCmotoneurons inlarvae resemble the somatic moto- neurons (SMcells) reported fromthe adultin overall mor- phology (cf. Lacalli andKelly 1999;Bone 1959, 1960a). Theyare distributed rather irregularly inthe anterior cord,with roughly equal numberson eachof thetwo sides,but thereis nosign of bilateral pairing. They receive synapses on dendritic spinesof varying length, located allalong theaxon, which confirms the supposition thatthe thin collaterals reported fromadult motoneurons (Bone 1960a; Castroet al.2004) are dendrites. Itis usefulto notethat as the cord grows,and its neuropile expands, early dendrites wouldhave to lengthen to maintain their original connections. Spinesin theadult cordwill thusbe longer thanthose inthe larval cord,and the longest spinesare the earliest, and presumably most important, functional connections. Sincethe longest spinesin thelarva are postsynaptic toLPNs, this interpreta- tion supports the central role proposed forthese cellsin ini- tiating swimming.

Itis notknown whether thelarval motoneurons persist throughto theadult stageor whether thelarval cellsare re-placed atsome pointin development. Lacalli (2000)has ar-gued forthe former, basedon the measured lengthsof the motoneuron apices. Theseare axially elongated byan amount that roughly matches theaxial expansion ofthe so-mites during development.

Fig. 9. Schematic diagram of the main locomotory control cir- cuits in the anterior cord in young amphioxus larvae. The LPN3s (third pair of large paired neurons) are central control neurons that probably act as pacemakers. They receive external sensory inputs via several pathways and communicate with the two classes of motoneurons (DCm and VCm) by synapses or junctions as shown. There is extensive additional synaptic input (not shown) to the VC (fast) system, but almost none to the DC (slow) sys- tem except for junctions with a single class of preinfundibular neurons (type 2 preinfundibular projection neurons, PPN2s) and, more caudally (not shown), input from the dorsal ocelli. The function of the ipsilateral projection neurons (IPNs) is not clear, but they appear to provide some kind of link between the two systems. The postinfundibular (= tegmental) neuropile is a para- crine center, and the specific interactions among its components are not clear from the morphological data. Modified from Lacalli (2002a).

DCmotoneurons differ fromVC motoneurons inbeing restricted tothe anterior partof thecord, specifically somites 2–6. This restriction wasfirst inferred fromEM data, which showed thatwhile axons project both rostrally and caudally fromthe lasttwo membersof the series, locatedin somites4 and5, none travel forward frommore caudal segments (Lacalli andKelly 1999). Confirming this,the amphioxus

homologof the estrogen-related receptor gene(ERR) selectively marksthe same cells, revealing sixpairs inthe anterior somites andnone more caudally (Bardetet al.2005). Various molecular data support theidea ofa segmen- talor otherwise periodic repeatin the arrangement ofcell typesin thecord atthe levelof somites 2–7 (Jackman andKimmel 2002; Mazetand Shimeld 2002), whichis essen- tiallythe amphioxus homologof the hindbrain. TheDC motoneurons evidently forma compressed series, withmore thanone pairper segment. Thetrue natureof patterning inthis partof the anterior cordis still, therefore, notclear. Itmay bethat somecell typesshow astrictly repeating seg- mental pattern, while othersare more loosely controlled, orthere maybe several quasi-segmental patterns superimposed overone another. See Shimeld and Holland (2005)for fur-ther discussion.

Incontrast tothe detailed information now published onthe microanatomy ofthe anterior cord, nothing comparable isyet available formore caudal regions. Swimming behavior changesas thelarva grows, froma phased side-to-side bend-ing ofthe wholebody invery young stagesto whatlooks likea propagated waveof contractions (Stokes 1997).The latter impliesa locomotory signal propagated from segmentto segment, morelike the situation in vertebrates. Oneinter- pretation isthat the pacemaker circuits identified inthe ante-rior cordof young larvaeare involved in initiating locomotory contractions, butthese are probably propagated through themore caudal segments bya seriesof local pace- makers. Regardless of details, itseems clearthat the control circuits described fromthe anterior cordof young larvae cannot account fullyfor the complexity and dynamics ofbe- haviorin older larvae.

Peripheral sensory cellsand nerves

Thissection isbrief, asa recent reviewby Lacalli (2004) covers most aspectsof thelarval sensory systemand in-cludes asummary ofwhat isknown ofthe early circuitry. Asin theadult, the surface epithelium inthe larvais supplied with sensory cellsof various types.The first evident func- tional response ofthe larvais to mechanical stimulation, andthis correlates withthe early appearance ofprimary typeI sensory neuronsin the rostrum andtail. Axonsfrom these enterthe cordat eachend andtravel long distances withinit (Holland andYu 2002), usuallyin the ventrolateral tracts, wherethey make repeated synapses with ventral inter- neurons involved in locomotory control (Lacalli 2002b, 2004). Those locatedat thetip ofthe rostrum enterthe cordvia thepaired rostral nerves, whichare substantial (ca.25– 30fibers) ata timewhen thedorsal nerves consist of,at most,a few fibers.As thelarva grows, primary sensory neu-rons differentiate overmuch ofthe body surface (Stokes and Holland 1995a; Holland andYu 2002),and thedorsal roots become much largeras their fibers growinto thecord. In contrast tothe rostral andcaudal fibers, those entering viadorsal nerves passinto the expanding dorsal tract, which,by thelate larval phase,has subdivided alongmost ofthe lengthof thecord into separate dorsaland subdorsal tracts thatrun in parallel (Holland andYu 2002).

Otherneuronal celltypes identified inthe epidermal tis-sues are(i) structures inthe rostrumat the neurula stage identified asgrowth conesby Yasuiet al.(1998) becauseof

their apparently transitory nature, butwhich maybe (seeLacalli 2003a) cell bodiesof pioneering rostral neurons that differentiate early; (ii) neurons associated withthe various peripheral plexuses that synapse peripherally, including twotypes (intrinsic and extrinsic neurons) inthe oralnerve plexus (Lacalli etal. 1999); (iii) additional, more special- ized,type Isensory cells, including avariant witha modi- fied, spine-like cilium (Lacalli andHou 1999); (iv) typeII sensory neurons, putative chemoreceptors witha collarof branched microvilli andbasal synapses to peripheral nerves, which developas thelarva matures (Stokes and Holland 1995a; Lacalli andHou 1999);and (v) ventral pitcells, whichlie inrows alongthe developing metapleural folds (Stokes and Holland 1995a) andare present alsoin latelar- vae, thoughthe evidence thatthese are neuronsis equivocal. Further research islikely toreveal additional typesand de-fine subtypes among those already described. Theabove conclusion regarding theadult —that the overall organiza- tionof the peripheral systemis complex andits function ispoorly understood — applies equallyto thelarva.

Postembryonic growth: generating anadult nervecord fromthe larvalone

As the amphioxus larva grows,the various fiber tractsand regionsof neuropile increase insize, expanding the cord both ventrally and laterally. The addition ofnew neurons, however, occurs mainly through proliferation and differenti- ationof thedorsal twothirds ofthe ventricular layer,a re-gion chiefly occupied bythe intermediate zone(not tobe confused withother usesof thisterm in reference to subdivi- sionsof thenerve cordalong its longitudinal axis), which,in the transverse plane,is definedby the presence ofpopula- tionsof various typesof translumenal neurons (Fig.10; see Lacalli 2002b). Insome casesthe translumenal processes arelarge enoughto bridgeto the neuropile onthe opposite side,but more commonly theyare smalland onlyjust crossthe central canal.The largest belongto thelarval giant cells,of which thereare five forwardof thefirst Rohdecell (Bone 1959); these probably correspond tothe large translumenal neurons described abovefrom theadult IR.The larval cellshave ipsilateral axonsbut otherwise resemble Rohde cells, whichhave medialor contralateral axons. Bone (1960a) sug-gests an internuncial function forcells ofthis general type,and tracings ofsome ofthe more abundant small translumenal neuronsin larvae tendto confirm this (Lacalli andKelly 2003a). The development ofthe intermediate zone,in fact, parallels the increase in peripheral inputvia dorsal nerve roots, which itselfis a consequence ofthe in-crease inthe surface areaof thebody andthe numberof pe- ripheral sensory cells. Young larvae havevery simple sensorimotor circuits, butit is unlikely thatthese wouldcope wellwith vastly increased input, presumably excitatory innature, ifit isnot filtered or modulated insome way.This isthe problemof gain control, whichis commonto developing neural systems (e.g.,see Priebeand Ferster 2002).The addi- tional levelof processing inserted during larval development between the sensory inputand themotor outputis evidently away ofsolving this problem.

Exceptfor the addition ofradial glial cellswith fibers that bridgeto thetop ofthe notochord (e.g.,as inFig. 4),the ventral partof the ependymal region, wherethe central canal

remainsopen, changes very little throughout thelarval phase (Lacalli andKelly 2002).The progressive increase inradial glialcell numbers does appearto havea consequence, how-ever, because theapices of motoneurons inthe juvenile andthe adultopen intothe central canalmore dorsally thanin theearly larva, i.e., slightly abovethe ventral expansion rather than adjacent tothe floor plate. Unless thereis a wholesale replacement oflarval neuronsby juvenile andadult ones, which seems unlikely, thisis best explained bya passive displacement ofearly neuronsup thesides ofthe central canalas more ventrally positioned glial precursors proliferate.

Insummary, the neuronsin the amphioxus nervecord ap-pear to differentiate alongmost ofthe cord’s lengthin aven- tralto dorsal sequence. Thiscan be interpreted invarious ways.If ontogeny were,in fact,no morethan a recapitula- tionof phylogeny, itwould meanthat the ventral locomotory circuits were evolutionarily olderthan thedorsal modulatory ones. Whileit is certainly truethat some dorsal neurons arehighly specialized, andmay wellbe late-evolving cell types, thiscan betrue ofall dorsal cellsonly ifthere wasan ances- tralform thatswam butlacked anyway of modulating itslo- comotion in response to peripheral sensory input. This is highly unlikely. Abetter explanation isthat the hatching stagein amphioxus hasbeen secondarily reducedin size,to atleast some degree, during evolution. Asthe hatching larva evolved waysto utilize the resources ofthe eggmost effec- tively, the differentiation ofthe essential partsof theloco- motory circuits was accelerated atthe expenseof everything else.In this interpretation, theearly differentiation ofthe ventral circuits isa clear indication oftheir crucial impor- tanceto the hatching larva, whereas thedorsal modulatory pathways are evidently less important, suchthat their devel- opmentcan be delayed.

Thereare anumber of specialized cell groupings atthe anterior endof theadult nervecord thatare not presentin early larvae, including various typesof migrated cellsde- scribed abovefrom theadult IR. Judging fromthe timethat the anterolateral serotoninergic cells first appear (Holland and Holland 1993), thesecell groupings probably developin thelate larval phaseor during metamorphosis. Despite the proliferative activity this entails, the anteriormost region failsto thickenas muchas therest ofthe cord,so theCV progressively disappears asan externally recognizable zone.Of the late-developing cell groups, thedorsal (population I) dopamine-containing cells reported byMoret etal. (2004)are especially noteworthy. Theseare asdorsal and anterior asone canget inthe nerve cord, whichis precisely wherea telencephalic homolog wouldbe predicted toform if amphioxus hadone. Forthis andother reasons, Lacalli (2004) suggested thatthe population Icells may represent aprimi- tive versionof the olfactory bulb. Regardless ofwhether thisis eventually confirmed, thekey pointis to recognize thatthese cells, likethe Joseph cellsand other anterior migrated cell clusters, areall late-developing centers that probably actin a modulatory wayon established circuits. Itshould be possible, in principle, to determine what functions suchcells performby correlating theirtime of appearance duringde- velopment with changesin behavior. Thiscould provea use-ful experimental strategy infuture.

Theexistence of migrated cell groupsof various typesin

Fig. 10. The nerve cord in transverse section, at about the level of somite 2, in early- and late-stage amphioxus larvae. The locomotory control system in young larvae (left) depends on sets of motoneurons (mn) and caudally projecting interneurons, all of which are ven- tral, adjacent to the floor plate. Later development (right) mainly involves proliferation and expansion of the intermediate and dorsal parts of the cord, which results in increasing numbers of interneurons (tn, since most have translumenal processes) and complex cir- cuitry as increasing numbers of peripheral sensory fibers enter the cord via dorsal nerves (dn). Thus, there is a clear directional bias in the way the cord differentiates: ventral first, dorsal later. Modified from Lacalli (2004).

theanterior nervecord raisesan interpretive problem that deserves some attention. Elsewhere inthe cord,the majority ofneurons remain attached tothe ventricular surface, sotheir siteof origin with reference tothe dorsoventral axisof thenerve cordis clearby inspection. Wherethis isnot thecase, e.g.,for the anterior migrated cell groups, thefinal po-sition ofthe cellswith respectto the dorsoventral axis, whether they reside closeto the periventricular layeror deepin the neuropile, maybe secondarily altered. Thiscan hap-pen intwo ways: eitherthe cells themselves migrate dorso- ventrally orthe expanding ventricular layer leaves them behind. Thus,a migrated cellor groupof suchcells mightbe quite dorsalin termsof itspoint oforigin inthe ventricular layer,but could ultimately occupya position wellbelow the dorsalmost cellsof themature nerve cord.The time elementin development therefore needsto be explicitly considered insuch cases. Thisis an important issue, particularly incases whereone istrying to identify possible homologs of vertebrate CNS neurons, aswe doin the section below.For example, inearlier attempts todeal withthe different subcat- egoriesof RBcells (e.g., Lacalli 1996),the term "tectal cell" was appliedto the anterior group(aRB cellsin Fig.8, equiv- alentto theA cellsof Bone 1959).The termwas chosenin part,and perhaps unfortunately, becauseit seemed thatthe dorsalmost cellsin aregion judgedto be midbrain-like onthe basisof molecular criteria could,in thecourse of evolution, have generated majornew dorsal structures, including theoptic tectum. However, cellsin thedorsal partof the amphioxus ventricular layerin young larvae haveonly just

begunto proliferate and generate neurons, soit isnot obvi-ous whether thelate progenyof these cellsare moreor less dorsalin character thanearly ones.How thisall relatesto the appearance of entirely newdorsal structures andbrain regions duringthe early stagesof vertebrate evolution re-mains tobe determined.

Comparison with vertebrates

Amphioxusis now generally accepted asthe best avail- ablemodel forthe immediate invertebrate ancestor ofverte- brates (Holland 2000).As such,it isof key importance to investigations into vertebrate origins and characteristic fea-tures of vertebrate organization. ThePNS, in particular, hasbeen the subjectof much comparative analysis inthe past. Amphioxus differs from vertebrates inits reliance onperiph- erally derived sensory neurons and extensive peripheral plexuses, whichare largely replaced in vertebrates byprod- uctsof theneural crestand placodes. Thereason forthis transition andthe relation (i.e., possible homology) between the component celltypes is currently amatter ofsome inter- est,but manykey questions remain unanswered (Lacalli 2004).At the anatomical level,it isthe arrangement and innervation patterns ofthe dorsal nerves thathave received themost attention, mainly becauseof theclues these provide concerning the segmental structure ofthe vertebrate head.The keyissue hereis the relation between the serially re- peating unitsin the hindbrain (i.e., rhombomeres), pharynx (gills and gill arches), and paraxial mesoderm, and how

Fig. 11. A proposal, consistent with both anatomical and molecular data, for how the anterior nerve cord in amphioxus larvae maps to the vertebrate brain. Most of the former finds its closest counterpart, and probable homologs, in the ventral region (variously shaded) of the latter, with the exception of the lamellar body, a pineal homolog. Caudal to somite 2, the amphioxus nerve cord becomes more hindbrain-like in character, according to the molecular data, but the exact point of the transition is uncertain (indicated by the question mark), as there is no obvious corresponding anatomical transition. Abbreviations and their corresponding terms are listed in Appendix

A. Modified from Lacalli (1996).

these allrelate tothe somite series.Do they,for example, re-flect asingle underlying patternof repeats, orare theyinde- pendent patterns secondarily superimposed onone another? Todate, the contribution of amphioxus tothis debatehas proven lessthan enlightening (e.g.,see Northcutt 2001), per-haps because patterns of peripheral innervation areless con- servative thanone would ideally like.In addition, however, the peculiarities oflarval growthand metamorphosis in amphioxus, especially thecaudal shiftof themouth andoral apparatus, ensure thatthe spatial relation between thenerve cordand peripheral structures inthe headis notonly differ- entthan thatin vertebrates, butalso undergoes developmen- tal changes thathave no vertebrate counterpart.

More recently, thewealth ofdata on patterns ofgene ex- pression during development has renewed effortsto identify regional homologies between the vertebrate brain and the anterior nervecord of amphioxus. Herethe molecular and anatomical dataare largelyin agreement, anda compara- tively consistent storyis emerging. The presence ofa pineal homolog (the lamellar body)and a comparable infundibular regionhas longbeen accepted as evidence thatthe CVis ba- sicallya primitive counterpart ofthe diencephalon (Olsson 1986). Patterns ofOtx, FoxB, andHox gene expression indi-cate the presence ofregions homologous withthe forebrain +midbrain toabout thelevel ofthe boundary between somites1 and2 (Holland and Holland 1999; Shimeld and Holland 2005). This placesthe PMC,with its pacemaker neurons, atroughly midbrain level. Then, somewhere adjacent tosomite

2,a zone begins thatis hindbrain-like in character, inwhich gene expression occursin a segmental or quasi-segmental patternof repeats (Jackman andKimmel 2002; Mazetand Shimeld 2002).The microanatomy shows, however, thatex- ceptfor the lamellar body,it isonly the ventral structures thatare represented inthe larval brain of amphioxus (Fig.11). Thereare zones similarin organization tothe ven-tral diencephalon, extending fromthe preoptic areato the hypothalamus and infundibulum, thefloor ofthe midbrain, roughly equivalent tothe tegmentum, andthe anterior endof the reticulospinal system. Further, the arrangement oflongi- tudinal fiber tractsis similarin vertebrates and amphioxus larvae (cf. Hjorthand Key2002; Lacalliet al.1994), andboth developan early ventral connection between thetwo sidesof thecord inthe immediate postinfundibular region, namelythe ventral commissure in vertebrates andthe postinfundibular neuropile in amphioxus larvae.

Inaddition, whatis sofar inferred concerning thefunc- tions performed inthe anterior nervecord in amphioxus lar-vae also indicates similarities tothe vertebrate brain. Basically, avariety of modulatory inputs, including signals fromsen- sory neurons locatedin the hypothalamus orits amphioxus equivalent, converge ona ventral locomotory control center (tegmentum and reticulospinal systemin vertebrates, PMCin amphioxus), andthis initiates a locomotory response. The ancestral planone infers fromthis wouldthus havea seriesof basal centers and connecting tracts that activate swim-ming underthe controlof external sensory inputs,as wellas

Fig. 12. Differences in scale between the anterior nerve cord of amphioxus at two stages (14-day-old larva, middle diagram, and newly metamorphosed juvenile, bottom diagram) and an embryonic vertebrate brain (zebrafish at 24 h, top diagram). For the zebrafish, cell clusters involved in early tract formation are shown. Arrowheads indicate the locations of the infundibular cells in amphioxus; the small arrow points to the postinfundibular (tegmental) neuropile. Based on a previous analysis of size differences; see Lacalli (2004). See text for discussion and see Appendix A for abbreviations and their corresponding terms.

internal homeostatic signals, andwould incorporate a switching centerto coordinate escape swimming withother basic activities, e.g., migratory swimming and feeding.

In contrast, thereis no evidence in amphioxus forhomo- logsof anyof thedorsal brain structures involved inverte- brate sensory processing exceptfor thepineal organ.A second possible exception isthe olfactory bulb, which Lacalli (2004)has arguedmay be represented in rudimentary formin the advanced-stage larvaof amphioxus (and presumably alsothe adult), butthis remainsto beproven. The signifi- canceof theJoseph cells,if any,in relation todorsal visual centersin vertebrates, remains unresolved. The various ma-jor centers involved in processing inputs fromthe vertebrate organsof special sense therefore appearto beabsent in amphioxus, againwith the possible exception ofthe pineal organand olfactory bulb.This is presumably because thesense organs themselves, andthe corresponding CNSpro- cessing centers, evolvedas vertebrates evolved, after their divergence frommore basal chordate lineages.

A more detailed lookat thenerve cord reveals thatthe ar- rangement ofcell types acrossthe dorsoventral axishas anumber of recognizably vertebrate-like features: thereis afloor plate,and motoneurons are ventral, while sensory interneurons are dorsal. Thereare also enough similarities between theRB cellsin amphioxus andthe Rohon-Beard cellsof vertebrates tosuggest homology (Fritzsch and Northcutt 1993).The anatomical evidence fora relation be-tween Rohde cellsand the various typesof reticulospinal gi-ant cells foundin vertebrates is somewhat less convincing,

however.Dorsal expansion ofthe cordis a noticeable fea-ture ofCNS development duringthe late larval phase;in contrast, the ventral components ofthe cord probably change very little. This reflects theneed fora vastlyin- creased capacity for sensory integration asthe larva growsand approaches metamorphosis. Comparable eventsin verte- brate development occurin the embryo, when proliferation expands thecord dorsally and somatosensory reflex circuits complete their differentiation. These changes correlate with increased eggand embryo sizeand a prolonged periodof embryogenesis in vertebrates, which allowsfor amuch largerand better developed nervous systemat hatching (seeFig. 12), possibly in response tothe increased predatory pressures vertebrates (andtheir hatchlings) experienced dur-ing their early evolution. Thisis an important point, becauseit addsa developmental dimension to previous ideas aboutCNS evolution. Though manyof theevents of neurogenesis in vertebrate embryos have probably always been embry- onic,it islikely thatthere areothers that evolvedas addi- tionsto analready functioning CNSin anactive animal. These wouldhave been incorporated intothe embryo only secondarily, sotheir mechanism of formation andthe wayin whichthe existing circuitry accommodates newinputs should reflect thisin someway.

Byway of example, consider the development inamphi- oxusof midbrain-level dopaminergic and serotoninergic neu-rons reported byMoret etal. (2004). Neurons utilizing these transmitters are presentat aroughly comparable location in vertebrates and contribute toseveral functionally important

modulatory systems. Whydid dopaminergic neurons evolvein this particular location, andwhat canone infer about their original function? Ifthey werenot originally formedin the embryo,at whatpoint inthe life history didthey firstap- pear,and why?These questions areraised herenot becausewe have answers, butto illustrate the following general point: when constructing evolutionary scenarios, onealways needsto be thinking interms ofa sequence of changing life histories. Fromour analysis, thiswill be especially truein thecase of innovations inCNS organization and structures that develop comparatively latein embryogenesis.

In themore immediate term, future workon theamphi- oxus nervous system might usefully be directed at refining our understanding ofthe molecular differences between neuronal subclasses, interms ofboth gene expression pat-terns and neurotransmitters. Workof thistype isnow in progress ina numberof laboratories, butmuch remainsto bedone. The amphioxus CNSis well suitedfor such studies becauseit issmall and compact enough thata thorough in- ventoryof celltypes isa feasible objective inthe larvaat least,if notin theadult. A disadvantage isthat many aspectsof amphioxus neural organization areso peculiar thatthe conventional wisdomas tohow aneural system oughtto op-erate is sometimes morea hindrance thana help.On the evolutionary side, thereare questions thatcan be addressed concerning thecell typesand circuits inthe ventral partof the anterior nerve cord,and herethere maybe direct homologs inthe ventral brainstem centersof vertebrates. In contrast, amphioxus will likely provide very littlein theway ofuseful clues regarding theorigin andbasic celltypes ofdorsal structures inthe vertebrate brain,as these evidently evolved largely after lancelets and vertebrates diverged. Afi- nalpoint, especially relevant tothe subjectof this issue,is the relation to hemichordates. Here, amphioxus offersa pos-sible bridge between thehighly centralized systemof verte- bratesand the diffuse oneof hemichordates (Lacalli 2003b). Onewould hopeto find intimations ofthe latter in amphioxus sufficient to indicate whether the hemichordate systemis derivedor whetherit retainsat leastsome impor- tant ancestral features.


Anadón, R., Adrio, F., and Rodríguez-Moldes, I. 1998. Distribution of GABA immunoreactivity in the central and peripheral ner- vous system of amphioxus (Branchiostoma lanceolatum Pallas). J. Comp. Neurol. 410: 293–307.

Baatrup, E. 1981. Primary sensory cells in the skin of amphioxus (Branchiostoma lanceolatum (P)). Acta Zool. (Stockh.) 62: 147– 157.

Baatrup, E. 1982. On the structure of the corpuscles of de Quatrefages (Branchiostoma lanceolatum (P)). Acta Zool. (Stockh.) 63: 39–44.

Baker, C.V.H., and Bronner-Fraser, M. 1997. The origins of the neural crest. Part II. An evolutionary perspective. Mech. Dev. 69: 13–29.

Bardet, P.L., Schubert, M., Horard, B., Holland, L.Z., Laudet, V., Holland, N.D., and Vanacker, J.-M. 2005. Expression of estrogen-related receptors in amphioxus and zebrafish: implica- tions for the evolution of hindbrain segmentation at the invertebrate-to-vertebrate transition. Evol. Dev. In press.

Boeke, J. 1902. Über das Homologon des Infundibularorganes bei

Amphioxus lanceolatus. Anat. Anz. 21: 411–414.

Boeke, J. 1908. Das Infundibularorgan im Gehirn des Amphioxus. Anat. Anz. 32: 473–488.

Boeke, J. 1935. The autonomic (enteric) nervous system of

Amphioxus lanceolatus. Q. J. Microsc. Sci. 77: 623–658. Bone, Q. 1959. The central nervous system in larval acraniates. Q.

J. Microsc. Sci. 100: 509–527.

Bone, Q. 1960a. The central nervous system in amphioxus. J. Comp. Neurol. 115: 27–64.

Bone, Q. 1960b. A note on the innervation of the integument in amphioxus, and its bearing on the mechanism of cutaneous sen- sibility. Q. J. Microsc. Sci. 101: 371–379.

Bone, Q. 1961. The organization of the atrial nervous system of amphioxus [Branchiostoma lanceolatum (Pallas)]. Philos. Trans.

R. Soc. Lond. B Biol. Sci. No. 243: 241–269.

Bone, Q. 1989. Evolutionary patterns of axial muscle systems in some invertebrates and fish. Am. Zool. 29: 5–18.

Bone, Q., and Best, A.C.G. 1978. Ciliated sensory cells in amphi- oxus. J. Mar. Biol. Assoc. U.K. 58: 479–486.

Bone, Q., Chubb, A.D., Pulsford, A., and Ryan, K.P. 1996. FMRFamide immunoreactivity in the peripheral (atrial) nervous system of amphioxus (Branchiostoma). Isr. J. Zool. 42(Suppl.): 213–225.

Burns, M.E., and Augustine, G.J. 1995. Synaptic structure and function: dynamic organization yields architectural precision. Cell, 83: 187–194.

Candiani, S., and Pestarino, M. 1998. Expression of the tissue- specific transcription factor Pit-1 in the lancelet, Branchiostoma lanceolatum. J. Comp. Neurol. 392: 343–351.

Candiani, S., Augello., A., Oliveri, D., Passalacqua, M., Pennati, R., de Bernardi, F., and Pestarino, M. 2001. Immunocyto- chemical localization of serotonin in embryos, larvae and adults of the lancelet, Branchiostoma floridae. Histochem. J. 33: 413– 420.

Castro, A., Manso, M.J., and Anadón, R. 2003. Distribution of neuropeptide Y immunoreactivity in the central and peripheral nervous systems of amphioxus (Branchiostoma lanceolatum Pallas). J. Comp. Neurol. 461: 350–361.

Castro, A., Becerra, M., Manso, M.J., and Anadón, R. 2004. Somatomotor system of the adult amphioxus (Branchiostoma lanceolatum) revealed by an anticalretinin antiserum: an immunocytochemical study. J. Comp. Neurol. 477: 161–171.

Conklin, E.G. 1932. The embryology of amphioxus. J. Morphol.

54: 69–151.

de Quatrefages, M.A. 1845. Mémoire sur le systéme nerveux et sur l’histologie du Branchiostome ou amphioxus. Annls. Sci. Nat. 4: 197–248.

Dogiel, A.S. 1903. Das periphere Nervensystem des Amphioxus (Branchiostoma lanceolatum). Anat. Hefte, 21[= Heft 66]: 145–


Dörffler-Melly, J., and Neuhuber, W.L. 1988. Rectospinal neurons: evidence for a direct projection from the enteric to the central nervous system in the rat. Neurosci. Lett. 92: 121–125.

Eakin, R.M., and Westfall, J.A. 1962. Fine structure of photoreceptors in amphioxus. J. Ultrastruct. Res. 6: 531–539.

Edinger, L. 1906. Einiges vom "Gehirn" des Amphioxus. Anat. Anz. 28: 417–428.

Ekhart, D., Korf., H.-W., and Wicht, H. 2003. Cytoarchitecture, to- pography, and descending supraspinal projections in the anterior nervous system of Branchiostoma lanceolatum. J. Comp. Neurol. 466: 319–330.

Flood, P.R. 1966. A peculiar mode of muscular innervation in

amphioxus. Light and electron microscopic studies of the so- called ventral roots. J. Comp. Neurol. 126: 181–218.

Flood, P.R. 1968. Structure of the segmental trunk muscles in amphioxus. With notes on the course and "endings" of the so- called ventral root fibres. Z. Zellforsch. Mikrosk. Anat. 84: 389– 416.

Flood, P.R. 1970. The connection between spinal cord and noto- chord in amphioxus (Branchiostoma lanceolatum). Z. Zellforsch. Mikrosk. Anat. 103: 115–128.

Flood, P.R. 1974. Histochemistry of cholinesterase in amphioxus (Branchiostoma lanceolatum, Pallas). J. Comp. Neurol. 157: 407–438.

Franz, V. 1923. Haut, Sinnesorgane und Nervensystem der Akranier. Jena. Z. Naturw. 59: 402–526.

Franz, V. 1927. Morphologie der Akranier. Ergebn. Anat. Entwgesch.

27: 465–692.

Fritzsch, B. 1996. Similarities and differences in lancelet and craniate nervous systems. Isr. J. Zool. 42(Suppl.): 147–160.

Fritzsch, B., and Northcutt, R.G. 1993. Cranial and spinal nerve or- ganization in amphioxus and lampreys: evidence for an ancestral craniate pattern. Acta Anat. 148: 96–109.

Fusari, R. 1889. Beitrag zum Studium des peripherischen Nerven- systems von Amphioxus lanceolatus. Int. Mschr. Anat. Physiol. 6: 120–140.

Gilland, E., and Baker, R. 1993. Conservation of neuroepithelial and mesodermal segments in the embryonic vertebrate head. Acta Anat. 148: 110–123.

Glardon, S., Holland, L.Z., Gehring, W.J., and Holland, N.D. 1998. Isolation and developmental expression of the amphioxus Pax-6 gene (AmphiPax-6): insights into eye and photoreceptor evolu- tion. Development, 125: 2701–2710.

Gorbman, A. 1999. Brain – Hatschek’s pit relationships in amphioxus species. Acta Zool. 80: 301–305.

Gorbman, A., and Tamarin, A. 1985. Early development of oral, olfactory and adenohypophyseal structures of agnathans and its evolutionary implications. In Evolutionary biology of primitive fishes. Edited by R.E. Foreman, A. Gorbman, J.M. Dodd, and R. Olsson. Plenum Press, New York. pp. 165–186.

Gorbman, A., Nozaki., M., and Kubokawa, K. 1999. A brain – Hatschek’s pit connection in amphioxus. Gen. Comp. Endocrinol. 113: 251–254.

Grove, E.A. 2002. The telencephalon saved by TLC. Neuron, 35: 215–217.

Guthrie, D.M. 1975. The physiology and structure of the nervous system of amphioxus (the lancelet), Branchiostoma lanceolatum Pallas. In Protochordates. Edited by E.J.W. Barrington and R.P.S. Jefferies. Symp. Zool. Soc. Lond. No. 36: 43–80.

Hatschek, B. 1884. Mittheilungen über Amphioxus. Zool. Anz. 7: 517–520.

Hesse, R. 1898. Untersuchungen über die Organe der Lichtemp- findung bei niederen Tieren. IV. Die Sehorgane des Amphioxus. Z. Wiss. Zool. 63: 456–464.

Hirakow, R., and Kajita, N. 1990. An electron microscopic study of the development of amphioxus, Branchiostoma belcheri: cleavage. J. Morphol. 203: 331–344.

Hirakow, R., and Kajita, N. 1991. Electron microscopic study of the development of amphioxus, Branchiostoma belcheri: gastru- lation. J. Morphol. 207: 37–52.

Hirakow, R., and Kajita, N. 1994. Electron microscopic study of the development of amphioxus, Branchiostoma belcheri: the neurula and larva. Acta Anat. Nippon. 69: 1–13.

Hjorth, J., and Key, B. 2002. Development of axon pathways in the zebrafish central nervous system. Int. J. Dev. Biol. 46: 609–619. Holland, L.Z., and Holland, N.D. 1999. Chordate origins of the

vertebrate central nervous system. Curr. Opin. Neurobiol. 9: 596–602.

Holland, N.D., and Holland, L.Z. 1993. Serotonin-containing cells in the nervous system and other tissues during ontogeny of a lancelet, Branchiostoma floridae. Acta Zool. 74: 195–204.

Holland, N.D., and Yu, J.-K. 2002. Epidermal receptor develop- ment and sensory axon pathways in vitally stained amphioxus. Acta Zool. 83: 309–321.

Holland, P.W.H. 1996. Molecular biology of lancelets: insights into development and evolution. Isr. J. Zool. 42(Suppl.): 247–272.

Holland, P.W.H. 2000. Embryonic development of heads, skeletons and amphioxus: Edwin S. Goodrich revisited. Int. J. Dev. Biol. 44: 29–34.

Holmes, W. 1953. The atrial nervous system of amphioxus (Branchiostoma). Q. J. Microsc. Sci. 94: 523–535.

Jackman, W.R., and Kimmel, C.B. 2002. Coincident iterated gene expression in the amphioxus neural tube. Evol. Dev. 4: 366–374. Jacobs, D.K., and Gates, R.D. 2003. Developmental genes and the reconstruction of metazoan evolution — implications of evolu- tionary loss, limitations on inference of ancestry, and type 2 er-

rors. Integr. Comp. Biol. 43: 11–18.

Johnston, J.B. 1905. The cranial and spinal ganglia and the viscero-motor roots in amphioxus. Biol. Bull. (Woods Hole), 9: 112–127.

Joseph, H. 1904. Über eigentümliche Zellstrukturen im Zentral- nervensystem von Amphioxus. Anat. Anz. 25(Suppl.): 16–26.

Kaji, T., Aizawa, S., Uemura, M., and Yasui, K. 2001. Establish- ment of left-right asymmetric innervation in the lancelet oral re- gion. J. Comp. Neurol. 435: 394–405.

Kölliker, A. 1843. Über das Geruchsorgan von Amphioxus. (Müllers) Arch. Anat. Physiol. (Berlin), 1843: 33–55.

Kutchin, H.L. 1913. Studies on the peripheral nervous system of amphioxus. Proc. Am. Acad. Arts Sci. 49: 569–624.

Lacalli, T.C. 1996. Frontal eye circuitry, rostral sensory pathways and brain organization in amphioxus larvae: evidence from 3D reconstructions. Philos. Trans. R. Soc. Lond. B Biol. Sci. No. 351: 243–263.

Lacalli, T.C. 2000. Cell morphology in amphioxus nerve cord may reflect the time course of cell differentiation. Int. J. Dev. Biol. 44: 903–906.

Lacalli, T.C. 2002a. The dorsal compartment locomotory control system in amphioxus larvae. J. Morphol. 252: 227–237.

Lacalli, T.C. 2002b. Sensory pathways in amphioxus larvae. I. Constituent fibres of the rostral and anterodorsal nerves, their targets and evolutionary significance. Acta Zool. 83: 149–166.

Lacalli, T.C. 2003a. Ventral neurons in the anterior nerve cord of amphioxus larvae. II. Further data on the pacemaker circuit. J. Morphol. 257: 212–218.

Lacalli, T.C. 2003b. Body plans and simple brains. Nature (Lond.),

424: 263–264.

Lacalli, T.C. 2004. Sensory systems in amphioxus: a window on the ancestral chordate condition. Brain Behav. Evol. 64: 148– 162.

Lacalli, T.C., and Hou, S. 1999. A re-examination of the epithelial sensory cells of amphioxus. Acta Zool. 80: 125–134.

Lacalli, T.C., and Kelly, S.J. 1999. Somatic motoneurons in the an- terior nerve cord of amphioxus larvae: cell types, cell position and innervation patterns. Acta Zool. 80: 113–124.

Lacalli, T.C., and Kelly, S.J. 2000. The infundibular balance organ and related aspects of cerebral vesicle organization. Acta Zool. 81: 37–47.

Lacalli, T.C., and Kelly, S.J. 2002. Floor plate, glia and other sup- port cells in the anterior nerve cord of amphioxus larvae. Acta Zool. 83: 87–98.

Lacalli, T.C., and Kelly, S.J. 2003a. Sensory pathways in amphioxus larvae. II. Dorsal tracts and translumenal cells. Acta Zool. 84: 1–13.

Lacalli, T.C., and Kelly, S.J. 2003b. Ventral neurons in the anterior nerve cord of amphioxus larvae. I. An inventory of cell types and synaptic patterns. J. Morphol. 257: 190–211.

Lacalli, T.C., Holland, N.D., and West, J.E. 1994. Landmarks in the anterior central nervous system of amphioxus larvae. Philos. Trans. R. Soc. Lond. B Biol. Sci. No. 344: 165–185.

Lacalli, T.C., Gilmour, T.H.J., and Kelly, S.J. 1999. The oral nerve plexus in amphioxus larvae: function, cell types and phylogen- etic sugnificance. Proc. R. Soc. Lond. B Biol. Sci. 266: 1461– 1470.

Lankester, E.R. 1875. On some new points in the structure of amphioxus and their bearing on the morphology of Vertebrata. Q. J. Microsc. Sci. 15: 257–267.

Lele, P.P., Palmer, E., and Weddell, G. 1958. Observations on the innervation of the integument of amphioxus, Branchiostoma lanceolatum. Q. J. Microsc. Sci. 99: 421–440.

Mazet, F., and Shimeld, S.M. 2002. The evolution of chordate neu- ral segmentation. Dev. Biol. 251: 258–270.

Meves, A. 1973. Elektronenmikroskopische Untersuchungen über die Zytoarchitektur des Gehirns von Branchiostoma lanceolatum.

Z. Zellforsch. Mikrosk. Anat. 139: 511–532.

Mirshahi, M., Bouchieux, C., Collenot, G., Thillaye, B., and Faure, J.-P. 1985. Retinal S-antigen epitopes in vertebrate and inverte- brate photoreceptors. Investig. Ophthalmol. Vis. Sci. 26: 1016– 1021.

Moret, F., Guilland, J.-C., Coudouel, S., Rochette, L., and Vernier,

P. 2004. Distribution of tyrosine hydroxylase, dopamine and se- rotonin in the central nervous system of amphioxus: implica- tions for the evolution of catecholamine systems in vertebrates. J. Comp. Neurol. 468: 135–150.

Müller, E. 1900. Studien über Neuroglia. Arch. Mikrosk. Anat. Entwmech. 60: 11–62.

Müller, J. 1844. Über den Bau und die Lebenserscheinungen des Branchiostoma lubricum Costa, Amphioxus lanceolatus Yarrell. Abh. K. Preuss. Akad. Wiss. 1844: 79–116.

Neuhuber, W.L., Appelt, M., Polak, J.M., Baier-Kustermann, W., Abelli, L., and Ferri, G.-L. 1993. Rectospinal neurons: cell bod- ies, pathways, immunocytochemistry and ultrastructure. Neuro- science, 56: 367–378.

Nieuwenhuys, R. 1998. Amphioxus. In The central nervous system of vertebrates. Vol. 1. Edited by R. Nieuwenhuys, H.J. ten Donkel- aar, and C. Nicholson. Springer-Verlag, Berlin. pp. 365–396.

Northcutt, R.G. 2001. Lancelet lessons: evaluating a phylogenetic model. J. Comp. Neurol. 435: 391–393.

Nozaki, M., and Gorbman, A. 1992. The question of functional homology of Hatschek’s pit of amphioxus (Branchiostoma belcheri) and the vertebrate adenohypophysis. Zool. Sci. (To- kyo), 9: 387–395.

Obermüller-Wilén, H. 1976. The infundibular organ of Branchio- stoma lanceolatum. Acta Zool. 57: 211–216.

Obermüller-Wilén, H., and Olsson, R. 1974. The Reissner’s fiber termination in some lower chordates. Acta Zool. 55: 71–79.

Olsson, R. 1986. Basic design of the chordate brain. In Proceed- ings of the 2nd International Conference on Indo-Pacific Fishes: Indo-Pacific fish biology, Tokyo, Japan, 29 July – 3 August 1985. Edited by T. Uyeno, R. Arai, T. Taniuchi, and K. Matsuura. Ichthyological Society of Japan, Tokyo. pp. 86–93.

Poss, S.G., and Boschung, H.T. 1996. Lancelets (Cephalochordata: Branchiostomatidae): How many species are valid? Isr. J. Zool. 42(Suppl.): 13–66.

Presley, R., Horder, T.J., and Slipka, J. 1996. Lancelet development

as evidence of ancestral chordate structure. Isr. J. Zool.

42(Suppl.): 97–116.

Priebe, N.J., and Ferster, D. 2002. A new mechanism for neuronal gain control (or how the gain in brains has mainly been ex- plained). Neuron, 35: 602–604.

Retzius, G. 1891. Zur Kenntniss des Centralnervensystems von

Amphioxus lanceolatus. Biol. Unters. 2: 29–42.

Rohde, E. 1887. Histologische Untersuchungen über das Nerven- system von Amphioxus lanceolatus. Zool. Beitr. 2(1): 169–211. Ruiz, M.S., and Anadón, R. 1989. Some observations on the fine structure of the Rohde cells of the spinal cord of the amphioxus, Branchiostoma lanceolatum (Cephalochordata). J. Hirnforsch.

30: 671–677.

Ruiz, M.S., and Anadón, R. 1991a. Ultrastructural study of the filum terminale and caudal ampulla of the spinal cord of amphioxus (Branchiostoma lanceolatum Pallas). Acta Zool. 72: 63–71.

Ruiz, M.S., and Anadón, R. 1991b. The fine structure of lamellate cells in the brain of amphioxus (Branchiostoma lanceolatum, Cephalochordata). Cell Tissue Res. 263: 597–600.

Ruiz, M.S., and Anadón, R. 1991c. Some considerations on the fine structure of rhabdomeric photoreceptors in the amphioxus, Branchiostoma lanceolatum (Cephalochordata). J. Hirnforsch. 32: 159–164.

Ruppert, E.E. 1997. Cephalochordata (Acrania). In Microscopic anatomy of invertebrates. Vol. 15. Edited by F.W. Harrison and

E.E. Ruppert. Wiley, New York. pp. 349–504.

Schneider, A. 1879. Beiträge zur vergleichenden Anatomie und Entwicklungsgeschichte der Wirbelthiere. Verlag von Georg Reimer, Berlin.

Schulte, E., and Riehl, R. 1977. Elektronenmikroskopische Unter- suchungen an den Oralcirren und der Haut von Branchiostoma lanceolatum. Helgol. Wiss. Meeresunters. 29: 337–357.

Sherwood, N.M., Adams, B.A., and Tello, J.A. 2005. Endocrinol- ogy of protochordates. Can. J. Zool. 83: 225–255.

Shimeld, S.M., and Holland, N.D. 2005. Amphioxus molecular bi- ology: insights into vertebrate evolution and developmental mechanisms. Can. J. Zool. 83: 90–100.

Stach, T. 2000. Microscopic anatomy of developmental stages of Branchiostoma lanceolatum (Cephalochordata, Chordata). Bonn. Zool. Monogr. 47.

Stokes, M.D. 1997. Larval locomotion of the lancelet Branchio- stoma floridae. J. Exp. Biol. 200: 1661–1680.

Stokes, M.D., and Holland, N.D. 1995a. Embryos and larvae of a lancelet, Branchiostoma floridae, from hatching to metamorpho- sis: growth in the laboratory and external morphology. Acta Zool. 76: 105–120.

Stokes, M.D., and Holland, N.D. 1995b. Ciliary hovering in lance- lets. Biol. Bull. (Woods Hole), 188: 231–233.

Takeda, N., Kubokawa, K., and Matsumoto, G. 2003. Immuno- reactivity for progesterone in the giant Rohde cells of the amphioxus, Branchiostoma belcheri. Gen. Comp. Endocrinol. 132: 379–383.

Tjoa, L.T., and Welsch, U. 1974. Electron microscopical observa- tions on Kölliker’s and Hatschek’s pit and on the wheel organ in the head region of amphioxus (Branchiostoma lanceolatum). Cell Tissue Res. 153: 175–187.

Uemura, H., Tezuka., Y., Hasegawa, C., and Kobayashi, H. 1994. Immunohistochemical investigation of neuropeptides in the cen- tral nervous system of the amphioxus, Branchiostoma belcheri. Cell Tissue Res. 277: 279–287.

Vallet, P.G., Ody, M.G., and Huggel, H. 1985. Étude ultra- structurale du neuropore d’amphioxus adulte (Branchiostoma lanceolatum Pallas). Rev. Suisse Zool. 92: 845–849.

von Kupffer, C. 1906. Die Morphogenie des Centralnervensystems. In Handbuch der vergleichenden und experimentellen Entwickelungslehre der Wirbeltiere. Edited by O. Hertwig. Band 2, Teil 3. Verlag von Gustav Fischer, Jena. pp. 1–272.

Watanabe, T., and Yoshida, M. 1986. Morphological and histo- chemical studies on Joseph cells of amphioxus, Branchiostoma belcheri Gray. Exp. Biol. 46: 67–73.

Welsch, U. 1968a. Die Feinstruktur der Josephschen Zellen im Gehirn von Amphioxus. Z. Zellforsch. Mikrosk. Anat. 86: 252–261.

Welsch, U. 1968b. Beobachtungen über die Feinstruktur der Haut und des äu?eren Atrialepithels von Branchiostoma lanceolatum Pall. Z. Zellforsch. Mikrosk. Anat. 88: 565–575.

Welsch, U. 1968c. Über den Feinbau der Chorda dorsalis von Branchiostoma lanceolatum. Z. Zellforsch. Mikrosk. Anat. 87: 69–81.

Wickstead, J.H., and Bone, Q. 1959. Ecology of acraniate larvae. Nature (Lond.), 184: 1849–1851.

Willey, A. 1894. Amphioxus and the ancestry of the vertebrates. MacMillan, New York.

Yasui, K., Tabata, S., Ueki, T., Uemura, M., and Zhang, S.-C. 1998. Early development of the peripheral nervous system in a lancelet species. J. Comp. Neurol. 393: 415–425.

Yu, J.K., Holland, N.D., and Holland, L.Z. 2002. Am amphioxus winged helix/forkhead gene, AmphiFoxD: insights into verte- brate neural crest evolution. Dev. Dyn. 225: 289–297.

Appendix A

List of abbreviations usedin figures.

AC Anadón’s GABAergic cells acf atriocoelomic funnels

alm anterolateral migrated cell group ansm anal sphincter muscle

antRo anterior Rohde cell

antRoax anterior descending Rohde axons ap atrial papillae

aRB anterior Retzius bipolar cells atsm atriopore sphincter muscle

avm anteroventral migrated cell group bc buccal cavity

bo balance organ

cdQ corpuscles of de Quatrefages chd notochord

cJo Joseph cells cJohn Johnston cell

CNS central nervous system comv ventral commissure

cRo Rohde cell

CV (larval) cerebral vesicle DC dorsal compartment

DCC dorsal commissural cell

DCm (larval) dorsal compartment motoneuron DR dorsal root cell

e endostyle EC Edinger cell

elm external labial muscle

enc encapsulated nerve endings ep epiphysis / pineal organ

gd gut diverticulum

gon gonad

Hp Hatschek’s pit

ibp inner buccal plexus ilm inner labial muscle inf infundibulum

intf internuncial fiber bundle io infundibular organ

Köp Kölliker’s pit

lac large translumenal (= commissural) cell lc lamellar cells

m1, m2, m3… myomere 1, 2, 3…

MC mid-commissural cell

Mg Müller’s glia (Schwann cell analogues) mpf metapleural folds

mRoax median descending Rohde axon n1, n2, n3… dorsal nerve 1, 2, 3…

ncmf notochordal motor fibers/tract nRo nucleus of Rohde

NS nervous system obp outer buccal plexus oHe organ of Hesse

os optic stalk

pdm posterodorsal migrated cell group ph pharynx

pig pigment cells of frontal eye pinfn postinfundibular neuropile

pm pterygeal muscle PMC primary motor center

postRoax posterior ascending Rohde axons r1, r2, r3… rhombomere 1, 2, 3…

RB Retzius bipolar cell rdors dorsal ramus

Rf Reissner’s fiber rg radial glia

rvent ventral ramus

rventc ventral ramus, ventral cutaneous branch rvisc visceral ramus

rvisca visceral ramus, ascending branch rviscd visceral ramus, descending branch

s1, s2, s3… somite 1, 2, 3… SD small dorsal cell Sg Schneider’s glia SM somatomotor cells

SM1 somatomotor cell type 1 smf somatomotor fibers/tract ssf somatosensory fibers/tract

t telencephalon tc optic tectum

tm trapezius muscle of atriocoelomic funnel VC ventral compartment

vcav ventricular cavity

VCm (larval) ventral compartment motoneuron VL ventral longitudinal cell

VM visceromotor cell

VM1, VM2 visceromotor cell type 1, 2 vp velar plexus

vsc ventral spindle-shaped cell group vsf viscerosensory fibers/tract

vsm velar sphincter muscle

CotA AmphPDF Amph Nrv Sys MSWord97 Rev 2
140627 - 1047