ryuichi shirasaki and samuel l. pfaff - brandeis · pdf fileryuichi shirasaki and samuel l....

22 May 2002 9:48 AR AR160-08.tex AR160-08.sgm LaTeX2e(2002/01/18)P1: ILV10.1146/annurev.neuro.25.112701.142916

Annu. Rev. Neurosci. 2002. 25:251–81doi: 10.1146/annurev.neuro.25.112701.142916

Copyright c© 2002 by Annual Reviews. All rights reserved

TRANSCRIPTIONAL CODES AND THE CONTROL OF

NEURONAL IDENTITY

Ryuichi Shirasaki and Samuel L. PfaffGene Expression Laboratory, The Salk Institute for Biological Studies, La Jolla,California 92037; email: [email protected]; [email protected]

Key Words LIM code, motor neuron subtypes, axon pathfinding, topography,spinal cord

■ Abstract The topographic assembly of neural circuits is dependent upon thegeneration of specific neuronal subtypes, each subtype displaying unique propertiesthat direct the formation of selective connections with appropriate target cells. Studiesof motor neuron development in the spinal cord have begun to elucidate the molecularmechanisms involved in controlling motor projections. In this review, we first describethe actions of transcription factors within motor neuron progenitors, which initiate acascade of transcriptional interactions that lead to motor neuron specification. We nexthighlight the contribution of the LIM homeodomain (LIM-HD) transcription factors inestablishing motor neuron subtype identity. Importantly, it has recently been shown thatthe combinatorial expression of LIM-HD transcription factors, the LIM code, confersmotor neuron subtypes with the ability to select specific axon pathways to reach theirdistinct muscle targets. Finally, the downstream targets of the LIM code are discussed,especially in the context of subtype-specific motor axon pathfinding.

INTRODUCTION

One of the fundamental issues in the field of developmental neuroscience is tounderstand the mechanisms that control the identity of distinct classes of neuronslocated at defined positions within the nervous system. This has been the goal ofconsiderable research effort in vertebrates and invertebrates alike, for the ability ofan animal to accomplish its behavioral repertoires depends critically on the gen-eration of appropriate neuronal types and the establishment of specific neuronalconnections. Importantly, in adult organisms, the mature identity of a neuron, whichunderlies the functional unit of a neural circuit, is represented by its characteristicfeatures that include the soma position, the axonal projection pattern, the elabora-tion of dendritic morphology, the expression of specific ion channels and neuro-transmitter receptors, and the production of appropriate neurotransmitter. Many ofthese characteristic traits are coordinately regulated and acquired during develop-ment (for reviews, see Edlund & Jessell 1999, Jessell 2000). Recent genetic studiesof vertebrates and invertebrates have identified a number of genes that regulate

0147-006X/02/0721-0251$14.00 251

22 May 2002 9:48 AR AR160-08.tex AR160-08.sgm LaTeX2e(2002/01/18)P1: ILV

252 SHIRASAKI ¥ PFAFF

the expression of these unique phenotypes, and further suggested that the ultimateoutcome of neuronal identity is under the control of complex regulatory networksof transcription factors acting in a sequential fashion (for recent reviews, see Jurataet al. 2000, Lee & Pfaff 2001). It should be noted, however, that many of thesetranscription factors act in the progenitor cells for neurons prior to cell cycle exit(for reviews, see Bang & Goulding 1996, Doe & Skeath 1996). Therefore, for themost part, mutation of progenitor cell transcription factors alters the expressionof the downstream transcription factors driving neural development, leading toa change in expression of the terminal differentiation genes in postmitotic neu-rons. This results in the alteration of the mature phenotype of neurons. Therefore,whereas progenitor cell transcription factors are important for establishing neu-ronal cell identity, it is unlikely that they directly control the expression of uniquecharacteristics of postmitotic neurons.

Axon pathfinding is one aspect of the neuronal differentiation program thatoccurs in postmitotic neurons. We now understand that the correct wiring of thenervous system depends upon the ability of individual axons to recognize specificguidance cues during their growth toward their final cellular targets (for reviews,see Tessier-Lavigne & Goodman 1996, Mueller 1999). In addition, the topographicorganization and elaboration of neural networks in higher organisms is dependentupon the generation of related, but distinct, neuronal subclasses (or subtypes) frombroad classes of neurons during development. In general, the axonal trajectoryof a single class of neurons is composed of several discrete pathways, some ofwhich are shared by multiple subtypes and others are unique pathways used byindividual subtypes. The classical experimental studies using avian embryos haveprovided a wealth of information on these axonal pathfinding steps (for a review,see Landmesser 1992). The emerging view is that each neuronal subtype possessesan intrinsic capacity to detect its own unique path soon after it becomes postmitotic(for a review, see Eisen 1994). This suggests that, before they reach their specificpathfinding decision points, they have already been genetically programmed to beable to sense their unique pathway-specific guidance cues. Accumulating evidencenow suggests that the intrinsic genetic programs are in most cases encoded not by asingle but rather by unique combinatorial arrays of certain classes of transcriptionfactors (for a review, see Pfaff & Kintner 1998).

This review focuses primarily on the role of postmitotically expressed trans-cription factors that ultimately control axon pathfinding during the developmentof vertebrates and invertebrates. In particular, we describe the role of transcriptionfactors in the development of a single class of neurons in the spinal cord, motorneurons, from the generation of motor neuron progenitors to the acquisition ofspecific subtype properties. In recent years, numerous reviews have covered por-tions of this work, most of which focused on inductive signals and the controlof neuronal patterning along the dorsoventral axis of the spinal cord as well asthe transcriptional mechanisms for determination of neuronal progenitor identity.Thus, rather than providing a comprehensive review on the transcriptional mech-anisms for motor neuron fate specification, we mainly discuss the role of oneclass of postmitotically expressed transcription factors, the LIM homeodomain


MOTOR NEURON SUBTYPE SPECIFICATION 253

(LIM-HD) transcription factors, in fate determination of motor neuron subtypesand in subtype-specific motor axon pathfinding, especially in light of intrinsicgenetic regulators of axon guidance.

SPECIFICATION OF PROGENITOR CELL IDENTITYBY HOMEODOMAIN TRANSCRIPTION FACTORS

We first describe the early patterning events that confer the identity of neuronalprogenitors in the spinal cord. During the early development of the vertebrate cen-tral nervous system (CNS), dividing progenitor cells in the ventricular zone acquireregionally restricted characteristics that later impose distinct identities on their neu-ronal progeny. The properties of progenitor cells are defined by secreted inductivesignals, which diffuse from their cellular source to form a concentration gradient.For example, in the dorsal spinal cord, bone morphogenetic proteins, which aresecreted from the surface ectoderm and the roof plate, control the specificationof dorsal cell types such as neural crest cells and dorsal sensory interneurons (fora review, see Lee & Jessell 1999). In contrast, in the ventral spinal cord, SonicHedgehog (Shh), a glycoprotein secreted from the notochord and the floor plate,specifies the pattern of generation of motor neurons and certain classes of ventralinterneurons in a concentration-dependent manner (for a recent review, see Briscoe& Ericson 2001). Indeed, in mice deficient in Shh function, motor neurons andventral interneurons fail to develop (Chiang et al. 1996).

Progenitor cells respond to the graded inductive signals by translating the spe-cific concentration of the signals into the patterned expression of transcriptionfactors (for a recent review, see Gurdon & Bourillot 2001). Recent studies haveidentified a set of homeodomain transcription factors expressed by ventral neuronalprogenitors that function as intermediaries in interpreting graded Shh signaling.The combinatorial expression of these homeodomain transcription factors subdi-vides the ventral spinal cord into five progenitor domains, each of which gives riseto a distinct class of postmitotic neurons (Figure 1) (Briscoe et al. 2000). Thesehomeodomain transcription factors can be categorized into class I and class II fac-tors based on their mode of regulation by Shh. The class I proteins (Pax7, Dbx1,Dbx2, Irx3, and Pax6) are repressed by Shh signaling, whereas the class II proteins(Nkx6.2, Nkx6.1, Olig2, Nkx2.2, and Nkx2.9) depend on Shh signaling for theirexpression (Briscoe et al. 2000, Mizuguchi et al. 2001, Novitch et al. 2001, Vallstedtet al. 2001). It should be noted that Olig2 is a basic helix-loop-helix (bHLH) trans-cription factor (see below), unlike the other factors which contain homeodomains.

Evidence has been provided that cross-repressive interactions between neigh-boring class I and class II factors subsequently act to refine and sharpen theboundaries of these progenitor domains (Briscoe et al. 2000), thus ensuring thatprogenitor cells within individual domains express distinct combinations of thesehomeodomain transcription factors. Importantly, it has also been shown that mostof these homeodomain transcription factors, which share a motif related to the eh1region of the Engrailed repressor domain (Smith & Jaynes 1996), act directly as



Figure 1 A combinatorial code of homeodomain transcription factors and basic helix-loop-helix transcription factors specifies the identity of neuronal progenitor cells in the ventralneural tube. Graded Sonic Hedgehog (Shh) signaling establishes distinct ventral progenitordomains by regulating the expression of transcription factors, which are subdivided intoclass I and class II factors based on whether their expression is repressed or induced byShh signaling. These transcription factors act as intermediaries in Shh-dependent neuralpatterning. Crossrepressive interactions between class I and class II factors, which abut acommon progenitor domain, contribute to establishing individual progenitor domains (p3-p0)with sharp boundaries. Each progenitor domain gives rise to a specific class of postmitoticneurons. The identity of potential class II factors (mX) counteracting the activity of thePax7 class I factor remains unknown. In this model, for example, the boundaries of themotor neuron (MN) progenitor domain are delineated ventrally by Pax6/Nkx2.2 and dorsallyby Irx3/Olig2. Nkx6.2 expression overlaps Nkx6.1 ventrally in the chick but is restricted tothe p1 domain in the mouse.

transcriptional repressors through the recruitment of Groucho-TLE corepressors(Muhr et al. 2001). Consistent with the idea that these transcription factors partic-ipate in a code for establishing neuronal identity, genetic studies have shown thatthe pattern of generation of certain ventral neuronal classes is impaired in rodentscarrying mutations in these genes such asPax6, Dbx1, Nkx2.2, Nkx6.1, andNkx6.2(Burrill et al. 1997, Ericson et al. 1997, Osumi et al. 1997, Briscoe et al. 1999,Sander et al. 2000, Pierani et al. 2001, Vallstedt et al. 2001).

Together, these findings suggest that mutual repression of class I and class IItranscriptional repressors at progenitor domain boundaries, as well as the



combinatorial expression of these factors within each progenitor domain, playsan instructive role in the specification of neuronal progenitor fate in the develop-ing spinal cord (Figure 1). Furthermore, the regulation of neuronal fate by nega-tively acting transcription factors in progenitor cells implicates a derepressionstrategy for neuronal fate determination. This model therefore predicts that cellidentity within each domain is achieved through active suppression of all the otherpossible neuronal fates directed by other transcription factors (see below).

NEURONAL SPECIFICATION AND CELL CYCLEEXIT BY bHLH TRANSCRIPTION FACTORS

Recent studies have now provided evidence that bHLH transcription factors invertebrates can also impose some aspects of neuronal identity, in addition to theirmore general roles in neurogenesis (for reviews, see Kageyama & Nakanishi 1997,Lee 1997, Guillemot 1999; see also Gowan et al. 2001). For example, in theperipheral nervous system,Mash1, a mammalian homolog ofDrosophila achaete-scute, is expressed in progenitor cells of autonomic neurons, where it is requiredfor the acquisition of a noradrenergic neurotransmitter phenotype (Hirsch et al.1998). In addition, bHLH transcription factors have the ability to direct the exitof neural progenitors from the cell cycle (Morrow et al. 1999, Farah et al. 2000,Mizuguchi et al. 2001, Novitch et al. 2001). In contrast, none of the class I/IIhomeodomain transcription factors described above have so far been shown toregulate cell cycle exit in the ventricular zone of the neural tube.

In the developing spinal cord, bHLH transcription factors are expressed in anumber of discrete domains along the dorsoventral axis of the neural tube. Someof these bHLH transcription factors, such as theDrosophila biparous/tap-relatedgenesNeurogenins(Ngns) and theatonal-related geneMath1, show expressionpatterns restricted to specific progenitor domains of the spinal cord (Akazawa et al.1995, Ben-Arie et al. 1996, Gradwohl et al. 1996, Sommer et al. 1996, Ma et al.1997, Helms & Johnson 1998, Lee et al. 1998, Gowan et al. 2001). Indeed, like theaction of homeodomain transcription factors in the ventral spinal cord describedabove, crossinhibitory regulation between Neurogenin1, Math1, and Mash1 hasbeen shown to determine discrete progenitor domains and neuronal fate in thedorsal spinal cord (Gowan et al. 2001).

Recently, another family of bHLH transcription factors has been identified andtwo of these members, Olig1 and Olig2, show a restricted pattern of expressionwithin the ventral spinal cord (Lu et al. 2000, Takebayashi et al. 2000, Zhou et al.2000). Furthermore, functional studies using ectopic expression of these geneshave suggested an instructive role in oligodendrocyte differentiation (Lu et al.2000; Zhou et al. 2000, 2001). Importantly, the onset of expression of these genesprecedes the appearance of oligodendrocyte progenitors and, in particular, they areprecisely expressed in the progenitor domain for motor neurons in the ventral spinalcord (Mizuguchi et al. 2001, Novitch et al. 2001), suggesting an additional rolefor these bHLH transcription factors in motor neuron fate specification. Indeed,



misexpression of Olig2 in the chick neural tube generated ectopic motor neurons.The ectopic cells formed at the expense of more dorsal neuronal types are correlatedwith Olig2’s ability to repress the expression of the class I homeodomain factorIrx3 (Mizuguchi et al. 2001, Novitch et al. 2001).

The involvement of Neurogenin2 (Ngn2), another member of the bHLH tran-scription factor family, in neuronal fate specification has been suggested from theregulatory interactions betweenNgn2andPax6, revealed by the analysis of micecarrying mutations in these genes (Scardigli et al. 2001). Interestingly, consis-tent with the temporal and spatial expression pattern of Olig2 and Ngn2 in theventral spinal cord, ectopic expression of Olig2 in the neural tube induced theupregulation of Ngn2 (Novitch et al. 2001) and, in some regions of the neuraltube, triggered expression of homeodomain transcription factors such as MNR2which is involved in motor neuron specification (Mizuguchi et al. 2001, Novitchet al. 2001). This therefore suggests that Olig2 regulates factors for establishingcell identity and neurogenesis in a coordinated manner. In support of this notion,mice deficient inNgn2function have reduced motor neuron numbers and alteredmolecular marker profiles (Scardigli et al. 2001). Because Ngns have been shownto regulate neurogenesis and promote cell cycle exit, the role of Olig2 appears notjust to activate specific motor neuron determinants but also to restrict the number ofcell cycles available to motor neuron progenitors and activate pan-neuronal traitsvia the activation of Ngn2.

MOLECULAR PATHWAY FOR SPECIFICATION OF MOTORNEURON IDENTITY BY COMBINATORIAL ACTIONS OFHOMEODOMAIN AND bHLH TRANSCRIPTION FACTORS

The combinatorial expression of homeodomain and bHLH transcription factorsthus delineates distinct progenitor domains in the ventricular zone of the spinalcord, each of which is responsible for the fate determination of their neuronalprogeny. Here, we describe in detail a model of the action of these transcriptionfactors on motor neuron fate specification.

The motor neuron progenitor (pMN) domain is demarcated by the V2 interneu-ron progenitor (p2) domain dorsally and by the V3 interneuron progenitor (p3)domain ventrally (Figure 1). The combinatorial actions of the class I/II factors,Pax6/Nkx2.2 and Irx3/Olig2, define the boundaries of the pMN domain. Withinthis pMN domain itself, where progenitors express Nkx6.1 and Nkx6.2, the ac-tivities of these two Nkx6 repressor proteins inhibit the expression of other celldeterminants while allowing Olig2 to become expressed for the coordinate regu-lation of factors involved in motor neuron differentiation such as Ngn2, MNR2,and HB9 (Briscoe et al. 2000, Sander et al. 2000, Vallstedt et al. 2001). Thesefindings raise the issue of whether Olig2 acts in parallel with or independently ofNkx6 and Ngn2 in this molecular pathway. Gain-of-function and loss-of-functionstudies have demonstrated that Olig2 acts downstream of Nkx6 proteins (Novitchet al. 2001).



Taken together, these findings establish a molecular hierarchy of transcriptionfactors in motor neuron fate specification, i.e., within the pMN domain, Nkx6class proteins initiate the generation of motor neuron progenitors by upregulatingthe expression of Olig2, which then promotes the expression of dedicated deter-minants such as MNR2 that impose specific motor neuron identity on cells, inparallel with the promotion of Ngn2 expression to drive motor neuron progenitorsto leave the cell cycle and acquire panneuronal traits characteristic of postmitoticneurons.

TRANSCRIPTIONAL CONTROL OF MOTOR NEURONFATE BY HOMEODOMAIN FACTORS, MNR2 AND HB9

The onset of expression of MNR2 and subsequent expression of HB9 correlateswith the progression of progenitor cells toward a motor neuron fate. In the chickembryo, MNR2 is first expressed by motor neuron progenitors during their fi-nal division cycle at the time when motor neurons attain independence from Shhsignaling (Ericson et al. 1996, Tanabe et al. 1998), and the expression persiststransiently in postmitotic motor neurons. Functional significance for the MNR2homeodomain factor is provided by the finding that ectopic expression of MNR2is sufficient to activate a program of somatic (skeletal muscle–innervating) motorneuron differentiation, characterized by the expression of motor neuron–specifichomeodomain factors [e.g., Islet-1 (Isl1), Islet-2 (Isl2), and Lhx3 (also termedLim3 or P-Lim)] and the enzyme choline acetyltransferase (ChAT) for neurotrans-mitter synthesis, and by the extension of axons that exit from the ventral spinalcord (Tanabe et al. 1998). In addition, the observation that MNR2 can positivelyregulate its own expression (Tanabe et al. 1998) points to a molecular basis forthe transition from an Shh-dependent to an Shh-independent state, thus contribut-ing to an irreversible commitment for the specification of somatic motor neurondifferentiation (Tanabe et al. 1998).

HB9 (Harrison et al. 1994), a homeobox gene whose expression in the CNS isalso restricted to motor neurons (Pfaff et al. 1996, Saha et al. 1997, Arber et al.1999, Thaler et al. 1999), possesses a homeodomain almost identical to that ofMNR2and shares with MNR2 the ability to induce motor neurons when ectopi-cally expressed in the chick neural tube (Tanabe et al. 1998). In contrast to MNR2,chick HB9 appears only in postmitotic motor neurons (Tanabe et al. 1998). Inter-estingly, mice seem to lackMNR2, but the expression pattern of HB9 in mice showsa pattern that approximates a composite of MNR2 and HB9 in chick (Arber et al.1999, Thaler et al. 1999). In mice lackingHB9 function, motor neurons are gener-ated on schedule and in normal number. However, soon after motor neurons haveleft the cell cycle, they transiently express the Chx10 homeodomain transcriptionfactor ordinarily characteristic of V2 interneurons.

Although this aberrant regulation of the gene profile in motor neurons doesnot cause a complete fate conversion of motor neurons to V2 interneurons, this



abnormal transcriptional code is accompanied by topological disorganization ofmotor columns and marked errors in the pattern of axonal projections of somaticmotor neurons in the periphery (Arber et al. 1999, Thaler et al. 1999). Thus, thesestudies show that, although HB9 is not required for motor neuron generation, un-like other factors such as the postmitotically expressed Isl1 (Pfaff et al. 1996), itis indispensable for facilitating a normal program of motor neuron differentiationand for suppressing expression of inappropriate interneuron genes. Indeed, therepressive role of HB9 deduced from the above loss-of-function studies is comple-mentary to the MNR2 gain-of-function study in which ectopic expression of MNR2suppresses the generation of V2 interneurons in the chick (Tanabe et al. 1998).

It is important to discuss why motor neurons, in the absence ofHB9 function,preferentially upregulate V2 interneuron characteristics rather than other interneu-ron genes. Cell lineage–tracing studies in the chick and the zebrafish have sug-gested that a single clonal progenitor can generate both motor neurons and ventralinterneurons (Leber et al. 1990, Kimmel et al. 1994). More recently, it has beenshown in mice that neuronal progenitors of motor neurons and V2 interneuronsexpress a related profile of homeodomain transcription factors, including Pax6 andLhx3/Lhx4 (Ericson et al. 1997, Sharma et al. 1998). These findings therefore sug-gest that motor neurons and V2 interneurons have genetically related progenitorcells. In addition, it has been shown in the chick that ectopic expression of Lhx3is sufficient to induce the V2 interneuron marker Chx10 in the absence of MNR2(Tanabe et al. 1998). Thus, the emerging view from the analysis ofHB9 knock-out mice is that motor neurons inherently possess the potential to initiate geneticprograms representative of V2 interneurons. In this context, active suppression ofV2 interneuron genetic programs within developing motor neurons is essential toconsolidate and establish the identity of postmitotic motor neurons, because Lhx3is shared by both V2 interneurons and motor neurons.

Thus, like other homeodomain transcription factors, MNR2 and HB9 contributeto the determination of neuronal fate by restricting the intrinsic cellular potentialto express conflicting genetic programs within a developing neuron (e.g., Tanabeet al. 1998, Arber et al. 1999, Siegler & Jia 1999, Thaler et al. 1999, Moran-Rivardet al. 2001, Pierani et al. 2001).

TRANSCRIPTION FACTORS EXPRESSEDIN POSTMITOTIC NEURONS: LIM-HDTRANSCRIPTION FACTORS

The initiation of axon outgrowth and subsequent pathfinding steps by a neuronis one characteristic feature of the neuronal differentiation program that occurspostmitotically. However, as described above, many of the transcription factorsthat have critical roles in neuronal fate specification act prior to or around thefinal cell division of progenitor cells, making it unlikely that they directly regulatethe onset of the specific events occurring in postmitotic neurons. In this context,



transcription factors expressed in postmitotic neurons are considered to have keyfunctions to activate axon guidance programs for specific pathway choices andtarget recognition.

Among transcription factors categorized in this way, LIM-HD transcriptionfactors are a family of genes in which all members identified to date in highervertebrates are expressed in distinct subsets of postmitotic neurons (for a recentreview, see Hobert & Westphal 2000). Apart from a DNA-binding homeodomain,one of the unique properties of LIM-HD transcription factors is their structurewith two copies of a specialized zing-finger motif called the LIM domain, afterits original discovery in twoCaenorhabditis elegansgene products, lin-11 andmec-3, and the insulin-enhancer-binding protein Isl1 (Way & Chalfie 1988, Freydet al. 1990, Karlsson et al. 1990). The LIM domain is thought to mediate protein-protein interactions, and, in particular, all LIM-HD transcription factor interactionsreported to date are dependent upon the LIM domains (for reviews, see Dawidet al. 1998, Jurata & Gill 1998), except for the interaction between the MEC-3and the POU homeodomain (POU-HD) transcription factor UNC-86 (Xue et al.1993). It has also been shown that NLI (also termed Ldb or CLIM), a widelyexpressed nuclear protein with intrinsic dimerization capacity, binds with highaffinity to the LIM domains of all LIM-HD transcription factors (Agulnick et al.1996, Jurata et al. 1996, Jurata & Gill 1997, Bach et al. 1997) and interestingly, itis mainly expressed in postmitotic cells in the developing CNS (Jurata et al. 1996).Importantly, evidence exists that the nuclear interactor NLI can assemble differentLIM-HD transcription factors into a single complex (Figure 2) (Jurata et al. 1998).

Thus, the unique combinatorial arrays of LIM-HD transcription factors in post-mitotic neurons have the potential to assemble homomeric and heteromeric com-plexes in a neuron-specific fashion. It seems likely that the formation of suchhigher-order complexes allows LIM-HD transcription factors to engage in acti-vation of terminal differentiation genes responsible for the expression of uniqueproperties of postmitotic neuronal subclasses (J. P. Thaler, S.-K. Lee, L. W. Jurata,G. N. Gill, S. L. Pfaff, submitted).

FUNCTIONS OF EXEMPLARY MEMBERS OFLIM-HD TRANSCRIPTION FACTORS

Before proceeding to the role of LIM-HD transcription factors in the determinationof motor neuron fate, we describe briefly the functions of exemplary members ofthis class of transcription factors,lin-11, mec-3, apterous, andIsl1.

lin-11

The complete genome sequence ofC. elegansreveals seven LIM-HD genes, whichare expressed in largely nonoverlapping sets of cell types (for a review, see Hobert& Westphal 2000).lin-11 was first identified on the basis of its role in regulat-ing asymmetric cell divisions in vulval cell lineages (Ferguson & Horvitz 1985,



Figure 2 The formation of NLI-mediated LIM-HD transcription factor complexes. NLIspecifically interacts with the LIM domain of LIM-HD transcription factors. The intrinsicdimerization capacity of NLI allows any given LIM-HD transcription factors to form homo-meric (A) and heteromeric (B) complexes with other LIM-HD transcription factors. Thus,combinations of LIM-HD transcription factors are thought to assemble into unique complexesfor the regulation of effector genes in postmitotic neurons responsible for the characteristicproperties of individual neuronal subclasses.

Ferguson et al. 1987, Freyd et al. 1990). Subsequent work has shown that post-mitotic AIZ interneurons, which are a major component of the thermoregulatoryneural network, expresslin-11. The gene activity is not required for the generationof these neurons per se, but has a role in specifying the function of these cells(Hobert et al. 1998).lin-11 is also expressed in ventral cord motor neurons andthe gene activity is not necessary for the axonal outgrowth of these motor neuronsbut is required for them to fasciculate correctly (Hobert et al. 1998). Recently, ithas been shown thatlin-11 is also necessary for the functional specification ofthe AWA olfactory neurons by regulating expression of the AWA-specificodr-7nuclear hormone receptor gene (Sarafi-Reinach et al. 2001), which promotes the



expression of AWA-specific characteristics and suppresses the fate of sibling AWColfactory neurons (Sengupta et al. 1994, Sagasti et al. 1999). It should be notedthat the vertebratelin-11 homologLim1 (also termedLhx1) is also expressed in asubset of motor neurons in the spinal cord (Tsuchida et al. 1994, Kania et al. 2000).

mec-3

Themec-3gene is required for the terminal differentiation of the six mechanosen-sory neurons (the touch receptor neurons) inC. elegans(Way & Chalfie 1988).In mec-3mutants, the expression ofmec-4, which encodes a putative ion channelsubunit necessary for touch sensitivity (Driscoll & Chalfie 1991), is absent and theexpression of mec-7, which encodes aβ-tubulin present in the processes of touchcells (Savage et al. 1989), is greatly reduced (Hamelin et al. 1992, Mitani et al.1993). The POU-HD transcription factorunc-86is necessary for the generation oftouch cells and continues to be expressed by postmitotic touch cells (Finney et al.1988, Finney & Ruvkun 1990). Subsequent molecular analyses have suggested thatMEC-3 and UNC-86 bind cooperatively to sites in themec-3promoter region andcan synergistically activate transcription from themec-3promoter (Xue et al. 1993,Lichtsteiner & Tjian 1995). Recently, it has been shown that MEC-3 binds poorlyto mec-4andmec-7promoters in the absence of UNC-86, and the hetero-oligomercomplexes of UNC-86 and MEC-3 directly activate two downstream genes,mec-4and mec-7, that are necessary for the specific function of touch cells (Dugganet al. 1998). Combinatorial transcription control by LIM-HD and POU-HD trans-cription factors can also be seen in development of the mammalian pituitary.For example, the LIM-HD transcription factor Lhx3 and POU-HD transcrip-tion factor Pit-1 synergistically activate several downstream target genes, suchas the prolactin and thyroid-stimulating hormoneβ-subunit genes (Bach et al.1995).

apterous

Theapterousgene is expressed in small subsets of postmitotic interneurons andmuscle precursors during embryonic development ofDrosophila(Bourgouin et al.1992, Lundgren et al. 1995). It has been shown thatapterouscontrols the abilityof these neurons to make their proper pathway choices and selectively fasciculatewith one another (Lundgren et al. 1995).apterousis also expressed in dorsal cellsof the wing imaginal disc and is required for cells of the dorsal compartment toassume a dorsal identity (Cohen et al. 1992, Diaz-Benjumea & Cohen 1993, Blairet al. 1994). Interestingly, ectopic expression ofapterousinduces ectopic ventralexpression of PS1 integrin, which is normally expressed in a dorsal-specific pattern.In contrast, loss of apterous causes the ectopic dorsal expression of PS2 integrin,a ventral-specific characteristic (Blair et al. 1994). These results suggest that celladhesion molecules are included among the downstream target genes of LIM-HDtranscription factors.



Isl1

Isl1 was originally identified as a protein that binds to enhancer elements in the ratinsulin gene (Karlsson et al. 1990). In the adult rat, Isl1 is expressed in all pancreaticislet cell types and in a variety of other polypeptide hormone-producing cells of theendocrine system (Thor et al. 1991). In addition, Isl1 is also expressed in a subsetof neurons, including motor neurons in the spinal cord and the brain stem (Thoret al. 1991). The expression pattern of Isl1 in the vertebrate CNS, together withthe findings from genetic studies inC. elegansandDrosophiladescribed above,raised the intriguing possibility that LIM-HD transcription factors may contributeto the specification of neuronal cell fate in the vertebrate nervous system. This ideagained credence when it was found in the chick spinal cord that Isl1 is expressedin all motor neurons immediately after they leave the cell cycle and before theyinitiate the expression of other differentiated properties (Ericson et al. 1992). Sim-ilar observations have been made in studies of spinal motor neurons in the mouse(Pfaff et al. 1996), of zebrafish primary motor neurons (Korzh et al. 1993, Inoueet al. 1994), and ofDrosophilamotor neurons and interneurons (Thor & Thomas1997). InIsl1 knockout mice and in chick embryo spinal cord explants cultured inthe presence ofIsl1 antisense oligonucleotides, the generation of motor neuronsdoes not occur, leading to the appearance of apoptotic cells at the stage when Isl1is first expressed (Pfaff et al. 1996). In contrast,Drosophila isletis not requiredfor the generation of motor neurons and interneurons. Instead, inislet mutants,these neurons show axon pathfinding errors and fail to exhibit their proper neuro-transmitter phenotype (Thor & Thomas 1997). Recently, the role of zebrafishIsl2,an ortholog ofDrosophila islet, in neuronal differentiation has been examined bydisrupting the heteromeric complex of Isl2 and NLI dimers and through the use ofantisense morpholino oligonucleotides againstIsl2 mRNA (Segawa et al. 2001).This functional repression of Isl2 causes abnormal soma settling within the spinalcord, elimination of ventrally projecting axons, and alteration in neurotransmitterphenotype by subsets of primary motor neurons. Interestingly, in these embryos,the peripheral projections of the Isl2-positive primary sensory neurons are neverformed.

MOTOR NEURON SUBTYPE IDENTITY INVERTEBRATES: CELLULAR ORGANIZATIONOF MOTOR NEURON SUBTYPES

Motor neurons located at different positions in the spinal cord project their ax-ons in a highly stereotyped manner to innervate distinct targets in the periphery(Figure 3). This high degree of spatial order establishes a topographic neuralmap. The topographic and functional organization of the spinal cord motor pro-jections is a consequence of the generation of subtypes of motor neurons duringdevelopment (for a review, see Pfaff & Kintner 1998). Motor neuron subtypesbecome evident when their axons select specific pathways to their muscle targets,and their cell bodies settle into longitudinally aligned columns within the spinal



cord (for a review, see Landmesser 1980). In higher vertebrates, the arrangement ofthis organization is hierarchical and composed of several layers. First, motor neu-rons that project to a single muscle are clustered together in a longitudinal columncalled a “pool.” Second, pools are grouped into larger columns on the basis of theirspecific target location in the periphery. Individual columns occupy characteristicpositions along the rostrocaudal axis of the spinal cord. Third, the limb- and axialmuscle–innervating motor columns are further subdivided into the medial and thelateral compartments corresponding to a dorsoventral subdivision in the positionof their respective targets. In addition, motor neurons are also categorized into twoclasses based on whether their axons emerge dorsally or ventrally from the neuraltube, as well as on the cell type that they innervate (i.e., somatic or visceral).

In the chick spinal cord (Figure 3), for example, somatic motor neurons thatinnervate trunk muscles form a medial motor column (MMC), which is continu-ous along the length of the spinal cord. In contrast, somatic motor neurons thatinnervate limb muscles are present only at brachial and lumbar levels, wherethey constitute a discontinuous lateral motor column (LMC). At thoracic levels,visceral motor neurons that innervate sympathetic neurons form a preganglionicmotor column of Terni (purple, Figure 3). Furthermore, the MMC and LMC canbe subdivided into two groups, respectively. A medial group of MMC, MMCm,extends along the entire rostrocaudal length of the spinal cord and projects axonsto axial muscles (red, Figure 3). A lateral group, MMCl, found only at thoraciclevels, projects axons to muscles of the ventral body wall (orange, Figure 3).Within the LMC, a medial group, LMCm, projects to ventral limb muscles (green,Figure 3), whereas a lateral group, LMCl, projects to dorsal limb muscles (blue,Figure 3).

DISTINCT INTRINSIC PROPERTIES OFMOTOR NEURON SUBTYPES

The specificity of motor projections to individual muscle targets at different pe-ripheral locations depends critically on the pathfinding decisions made by axonalgrowth cones of each subtype of motor neuron during embryonic development(for reviews, see Landmesser 1992, Eisen 1994). A remarkable feature of axonpathfinding by motor neuron subtypes is that after leaving the spinal cord, theyinitially follow a common pathway to grow ventrally, at which point they divergeto select a subtype-specific pathway leading toward their appropriate targets. Forexample, the axons of MMCm neurons break away from the common path, theventral root, and form a nerve branch (termed the dorsal ramus) to innervate the der-momyotome, the final target of MMCm axons (Tosney & Landmesser 1985a,b). Incontrast, the axons of LMCm and LMCl neurons appear to ignore the choice pointof MMCm axons and instead continue to grow ventrolaterally to reach the baseof the limb. At this point, LMCm axons project exclusively into the ventral limbmesenchyme, whereas LMCl axons project only to the dorsal limb mesenchyme(Landmesser 1978; Tosney & Landmesser 1985a,b). Importantly, motor neuronsdevelop their target specificity prior to their target innervation, as suggested by the



analyses of motor axon projection patterns after a variety of surgical manipulationsof chick and zebrafish embryos (for reviews, see Landmesser 1992, Eisen 1994;see also Matise & Lance-Jones 1996). Furthermore, in embryonic zebrafish, it hasbeen shown by transplanting identified primary motor neurons to new spinal cordlocations that the identity of motor neuron subtypes is determined just prior to theonset of axon outgrowth (Eisen 1991).

COMBINATORIAL EXPRESSION OF LIM-HDTRANSCRIPTION FACTORS MARKS MOTORNEURON SUBTYPES: A LIM CODE

Tsuchida et al. (1994) identified a family of chick LIM-HD transcription factorsand found that four members,Isl1, Isl2, Lim1, andLhx3, are expressed in mo-tor neurons in the embryonic chick spinal cord. Interestingly, the expression ofa single LIM-HD transcription factor was not sufficient to distinguish individualmotor columns, nor did it obviously delineate individual motor pools. However,the combinatorial expression of these genes (LIM code) defines subtypes of motorneurons that occupy different columns in the spinal cord, select specific axonalpathways in the periphery, and innervate distinct targets (Figure 3). Furthermore,each subtype of motor neuron expresses its own LIM code prior to the segre-gation of motor neurons into columns and before distinct axonal pathways areestablished in the periphery. Thus, a LIM code expressed in motor neurons ac-curately marks a motor neuron subtype long before they reach their own muscletargets. Subsequent studies have shown that the patterned expression of LIM-HDtranscription factors correlates with the axonal trajectory of primary motor neu-rons in embryonic zebrafish as well (Appel et al. 1995, Tokumoto et al. 1995). Inaddition, the transplantation of zebrafish primary motor neurons to new spinal cordpositions initiated a new LIM code in these motor neurons, accompanied by axonalprojection patterns appropriate for the new position (Appel et al. 1995), suggestinga tight correlation between the LIM-HD factors and the commitment to a particularmotor neuron fate represented by subtype-specific axonal projection patterns.

Together, these findings raise the possibility that the LIM code confers subtypesof motor neurons with the ability to select distinct axonal pathways and to recognizespecific muscle targets in the periphery. In this context, one function of the LIMcode may be to regulate the expression of downstream genes that determine motoraxons’ responsiveness to guidance cues specific for a subtype of motor neurons.

DYNAMIC EXPRESSION OF LIM-HD TRANSCRIPTIONFACTORS IN MOTOR NEURONS

The expression patterns of LIM-HD transcription factors change rapidly in motorneurons during the initial period in which they are generated (Tsuchida et al. 1994,Appel et al. 1995, Sharma et al. 1998), suggesting that precise temporal regulation



of LIM-HD transcription factor expression potentially expands the LIM code. Us-ing CRE-mediated lineage tracing in mice, Sharma et al. (1998) found that, for abrief period around the time when motor neurons are born,Lhx3and the redundantgeneLhx4are expressed in all motor neuron subtypes that extend axons ventrallyfrom the neural tube (Figure 4). In contrast, motor neurons whose axons emergedorsally from the neural tube never express these genes. Surprisingly, when ven-trally exiting motor neurons become topographically organized into columns andextend axons in the periphery, they rapidly downregulate the expression of Lhx3and Lhx4, except for MMCm neurons that maintain expression of these factors(Figure 4). Because all motor neurons begin to express Isl1 soon after they exitthe cell cycle irrespective of the axonal trajectory (Ericson et al. 1992), the dy-namic expression of Lhx3 and Lhx4 in combination with Isl1 is found to define atransient LIM code that is predicted to assign the axonal exit point selected by mo-tor neurons. To test whether these LIM-HD transcription factors regulate axonalexit point identity, Sharma et al. (1998) analyzed mice deficient in bothLhx3andLhx4. They found that, although motor neuron differentiation proceeds, ventrallyexiting motor neurons switch their subtype identity to become motor neurons thatextend axons dorsally from the neural tube. Conversely, they showed using chickembryo electroporation that ectopic expression of Lhx3 in dorsally exiting motorneurons is sufficient to reorient their axonal projections ventrally from the neuraltube.

Thus, at the birth of motor neurons, Isl1 and Lhx3/Lhx4 act together in a binaryfashion during a brief period to direct motor axons ventrally from the neuraltube, providing the first functional evidence that the LIM code plays a role in axonpathfinding in vertebrates. Notably, this “early” LIM code is very dynamic and cellsrapidly switch to new LIM codes as motor column identity begins to be established.

FUNCTIONAL TESTS OF THE LIM CODE FOR MOTORAXON PATHWAY SELECTION IN THE PERIPHERY

The LIM code model, as first conceived (Figure 3) (Tsuchida et al. 1994), pre-dicted that axonal projections in the periphery of the embryo were regulated by thecombinatorial activities of the LIM-HD transcription factors expressed by a givenmotor neuron subtype. For example, ifLhx3/Lhx4 were knocked out, cells fatedto become MMCm neurons were predicted to express an Isl1/Isl2 LIM code char-acteristic of MMCl and LMCm neurons, and were therefore predicted to projectaxons to the targets of these motor neuron subtypes. However, because Lhx3 andLhx4 have an extremely dynamic expression pattern and thus seem to act at severaldistinct phases of motor neuron development (Figure 4), previous studies usingLhx3- andLhx4-deficient mice were not able to reveal the role of these factorsin motor axon pathfinding in the periphery owing to the cell fate conversion toa dorsally exiting motor neuron identity (Sharma et al. 1998). To define the laterfunctions of these factors in motor axon guidance, Sharma et al. (2000) therefore



employed a mice knock-in strategy to ectopically and stably express Lhx3 in allsomatic motor neurons. This genetic alteration was sufficient to convert motor neu-ron subtypes to the MMCm identity, as defined by the cell body settling patternwithin the spinal cord and gene expression profile. Importantly, this resulted in adramatic increase, but not total conversion, in the number of axonal projectionstoward axial muscles, the normal targets of MMCm neurons (Figure 5). Thus,during the later period in which motor axons extend in the periphery, a LIM codespecific for MMCm neurons (i.e., Isl1/Isl2 and Lhx3/Lhx4) now controls motoraxon pathway selection and target recognition, presumably by regulating MMCmaxons’ responsiveness to guidance cues that promote and direct their axonal growthtoward axial muscles.

Another line of evidence for a role of the LIM code in the control of mo-tor axon pathfinding in vertebrates was provided by focusing on the selectiveexpression of Lim1 in LMCl neurons (Kania et al. 2000). A characteristic fea-ture of LMCl neurons is that, at the base of the limb, the LMCl axons projectinto dorsal limb mesenchyme. This is in marked contrast to the axons of LMCmneurons which select a ventral trajectory within the limb (Figure 6A). By intro-ducing an axonal tracer tau-LacZ into theLim1 locus, Kania et al. analyzed theaxonal trajectory of LMCl neurons within the limb. They found that, in micedeficient inLim1 function, the specification of LMCl neurons, such as the cellmigration pattern in the spinal cord and appearance of characteristic molecular at-tributes, proceeds normally, but the LMCl axons project into the dorsal and ventralhalves of the limb at equal incidence, thus randomizing the selection of a dorsal orventral pathway by these axons (Figure 6B). In addition, inLim1 mutants, LMClaxons are unable to extend into the distal part of the limb. These results suggestthat, althoughLim1 is not essential for dorsally directed growth of LMCl axons, aLIM code for LMCl neurons (i.e., Isl2 and Lim1) may control the ability of LMClaxons to respond to specific guidance cues which ensure that a correct pathfindingdecision is made at the bifurcation point.

←−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−−Figure 4 Dynamic expression of LIM-HD transcription factors during motor neurondevelopment in mice. The onset of expression of Lhx3 and Lhx4 occurs in the final celldivision of ventrally exiting motor neuron progenitors. Lhx3 and Lhx4 are not detectedin dorsally exiting motor neuron progenitors. When motor neurons exit the cell cycle,they begin to express Isl1. Although MMCm neurons maintain the expression of Lhx3and Lhx4 even when their axons exit ventrally from the neural tube to project dorsallytoward their axial muscles, other motor neuron subtypes rapidly downregulate theexpression of these factors when they begin to form specific motor columns withinthe spinal cord and extend axons ventrally from the neural tube to grow toward theirnonaxial muscle targets. LIM-HD gene expression is dynamic and produces severaltransient combinations of factors that contribute to motor neuron development in asequential fashion.



The action of the LIM code for axon pathfinding decisions seems to be con-served through evolution. InDrosophilaembryos, motor neurons that project axonsthrough the b branch of the intersegmental nerve (ISNb) express bothisletandlim3,a homolog of vertebrateLhx3, whereas motor neurons that project axons throughthe d branch (ISNd) express onlyislet (Thor et al. 1999). As in vertebrates, sub-types of motor neurons inDrosophilatherefore express their unique combinationsof LIM-HD transcription factors. Thor et al. (1999) found that, inlim3 mutants,the axons of ISNb motor neurons behave like those of ISNd motor neurons. Con-versely, ectopic expression oflim3 in ISNd motor neurons now causes the axonsof these neurons to behave as those of ISNb motor neurons in a predictable way.

Together, the findings from these genetic studies therefore establish the role ofthe LIM code in motor axon pathway selection, which represents a critical step forgenerating the topographic organization of motor projections. More generally, theLIM code expressed in postmitotic neurons is likely to participate in determiningthe fates and axonal projections of subtypes of neuronal classes throughout theCNS (for a recent review, see Hobert & Westphal 2000; see also Varela-Echavarr´ıaet al. 1996, Mar´ın et al. 2000, Nakagawa & O’Leary 2001).

DOWNSTREAM TARGETS OF THE LIM CODE FORSUBTYPE-SPECIFIC MOTOR AXON GUIDANCE

The downstream targets of the LIM code that control motor axon pathway selectionin vertebrates as well as the subtype-specific axon guidance cues remain poorlycharacterized. Here, we discuss candidate guidance cues and their motor neuronsubtype-specific receptors that may be regulated by the LIM code.

Exit Point from the Neural Tube: Ventral or Dorsal

Analysis of the fate of motor neurons in mice deficient in bothLhx3 andLhx4revealed that the combinatorial expression of Isl1 and Lhx3/Lhx4 specifies motorneuron axonal exit point identity (Sharma et al. 1998). The finding that the absenceof Lhx3 and Lhx4 causes motor axons to exit more dorsally from the neural tubesuggests that the status of these factors may control the responsiveness of motoraxons to either attractants or repellents secreted from the region outside the motorcolumn. It has been shown, for example, that the floor plate secretes diffusiblechemorepellents, such as netrin-1 and semaphorins, for axons that show dorsallydirected growth away from the ventral midline (Colamarino & Tessier-Lavigne1995, Guthrie & Pini 1995, Tamada et al. 1995, Shirasaki et al. 1996, Varela-Echavarr´ıa et al. 1997). To date, involvement of netrin-1 and semaphorins in theformation of the initial ventral trajectory seems unlikely, for in mice deficientin the netrin receptorDCC and semaphorin receptorsneuropilins, motor axonexit from the ventral neural tube is not perturbed (Fazeli et al. 1997, Kitsukawaet al. 1997, Chen et al. 2000, Giger et al. 2000). However, among the repellents



expressed in floor plate cells, Slit2 is able to repel spinal motor axons at a distancein vitro (Brose et al. 1999). Because spinal motor neurons express the Slit receptorRoboaround the time when these axons exit the neural tube (Kidd et al. 1998,Brose et al. 1999, Yuan et al. 1999), it is possible that Slit/Robo signaling in someway contributes to the guidance of motor axons from the neural tube. Alternatively,Lhx3 and Lhx4 in motor neurons might regulate the expression of receptors forchemorepellents secreted from the dorsal neural tube. If this is the case, the sourceof chemorepellents might exclude the roof plate because this tissue did not repelspinal motor axons in vitro (Augsburger et al. 1999). The release of repellentactivity from the dorsal neural tube for subsets of cranial motor axons has recentlybeen reported (see Caton et al. 2000).

Growth Away from the Neural Tube

It has been observed that motor neuron subtypes other than MMCm neurons rapidlydownregulate the expression of Lhx3 and Lhx4 after their axons exit the neuraltube (Figure 4) (Sharma et al. 1998). Interestingly, in the developing neural tube,commissural axons have been shown to rapidly change their responsiveness toguidance cues upon encountering the floor plate, an intermediate target of theseaxons (Shirasaki et al. 1998, Shirasaki & Murakami 2001, Zou et al. 2000). Thesestudies suggest that growing axons still possess an ability to change their initialnavigation program during their growth toward their final targets. Thus, the rapidchange in the status of Lhx3/Lhx4 within motor neurons might be relevant to thechange in their initial axon-navigation program, which may enable axons of motorneuron subtypes other than MMCm to extend ventrally away from the neural tube.On the other hand, the maintenance of Lhx3/Lhx4 in MMCm motor neurons asthey extend axons in the periphery raises the intriguing possibility that a relatedguidance cue is used both for directing motor axons ventrally from the neuraltube and for their selection of the dorsal ramus pathway in the periphery. Finally, itshould be noted that the involvement of HGF/c-met signaling in directed growth ofLMC axons toward the base of the limb is still controversial, for the HGF receptorc-met has not been observed to be expressed in LMC neurons when they first growtoward the limb (Ebens et al. 1996, Novak et al. 2000), and the axonal projectionstoward the limb are unaffected inHGFmutant mice (Ebens et al. 1996), suggestinganother signal may be responsible.

Directed Growth Toward Axial Muscles

It has been shown that ablation of the dermomyotome, the precursor of axial mus-cles, in the chick embryo prevents the dorsal deviation of MMCm axons (Tosney1987, 1988), suggesting the possibility that the dermomyotome secretes a chemoat-tractant for the axons of MMCm neurons. Similarly, evidence has been providedthat Xenopus somitic myoblasts can attract neurites from Xenopus neural tubeexplants in culture (McCaig 1986). Furthermore, in the zebrafish mutantspade-tail, which is deficient in trunk myotome development (Kimmel et al. 1989), the



projection patterns of motor axons toward the myotome develop abnormally (Eisen& Pike 1991). Although the hypothesis of the existence of a dermomyotome-derived chemoattractant for MMCm axons has fascinated developmental neurobi-ologists for many years, direct evidence has yet to be provided.

Tessier-Lavigne and colleagues have attempted to detect such a chemoattractiveactivity from the dermomyotome using a collagen-gel coculture assay, but wereunable to detect its activity (Ebens et al. 1996). In addition, they have also testedwhether candidate secreted molecules expressed in the dermomyotome could pro-mote the outgrowth of motor axons from ventral spinal cord explants. However,the factors that they have tested such as netrin-1, basic FGF, and TGFβ-1 failedto promote motor axon outgrowth (Ebens et al. 1996). Consistent with this, mostof the known secreted molecules expressed in the dermomyotome do not showoutgrowth-promoting activity for MMCm axons (R. Shirasaki & S. L. Pfaff, un-published data). Possible explanations for these failures are that the cultured targettissue fails to differentiate properly or motor axons lose responsiveness to thisputative chemoattractant when placed in vitro (Ebens et al. 1996).

Most importantly, however, the observation that, in mice in which Lhx3 isectopically expressed in all motor neurons, many motor axons that have first by-passed the normal choice point of MMCm axons now are dramatically redirectedto grow back toward axial muscles (Figure 5) (Sharma et al. 2000) provides furtherin vivo evidence to support the notion that the axial muscle, the target of MMCmaxons, secretes a long-range chemoattractant specific for MMCm axons. To date,the nature of the chemoattractive activity specific for MMCm axons as well as theidentity of the chemoattractant receptor remain unknown.

Choice of Axonal Entry into the Limb: Ventral or Dorsal

The cellular environment of the dorsal and ventral halves of the limb mesenchymeis molecularly distinguished by the restricted dorsal expression of LIM-HD trans-cription factor Lmx1b (Riddle et al. 1995, Vogel et al. 1995). Kania et al. (2000)found that, inLmx1bmutants, the axons of both LMCl and LMCm neurons ran-domly project into the dorsal and ventral limb, suggesting that LMCl and LMCmaxons can no longer sense the distinction of guidance cues that direct these ax-ons toward their appropriate territory. Together with the complementary resultsobtained inLim1 mutants described above (Figure 6B) (Kania et al. 2000), thesefindings suggest that one of the downstream genes regulated byLim1 in LMClneurons may encode a receptor that recognizes the guidance cues preferentiallyexpressed in the dorsal or the ventral half of the limb mesenchyme. In addition, atthe base of the limb, the axons of LMCl and LMCm neurons segregate immedi-ately to enter either the dorsal or ventral limb mesenchyme (Tosney & Landmesser1985a), suggesting that the guidance cues that control this pathfinding decisionmay operate at short range.

Recent studies have shown that the Eph family of tyrosine kinase receptorsand their membrane-bound ligands, the ephrins, mediate axon guidance by a



contact-dependent mechanism (for a recent review, see Wilkinson 2001). Indeed,some members of the Eph/ephrin signaling system are differentially and tran-siently expressed in subsets of motor neurons (Kilpatrick et al. 1996, Ohta et al.1996, Araujo et al. 1998, Iwamasa et al. 1999). Among these, it has been shownthat in chick and mouse embryos, the EphA4 protein is predominantly expressedin the axons of LMCl neurons during their pathfinding decision at the base ofthe limb (Ohta et al. 1996, Helmbacher et al. 2000). Moreover, the ventral limbmesenchyme appears to express higher levels of two EphA4 ligands, ephrin-A2and ephrin-A5 proteins, as LMC axons reach this choice point (Ohta et al. 1997,Eberhart et al. 2000). Strikingly, in mice deficient inEphA4function, the LMClaxons fail to enter the dorsal limb and instead select a ventral trajectory (Fig-ure 6C) (Helmbacher et al. 2000). Because ephrin-A2 and ephrin-A5 have aninhibitory effect on axonal outgrowth of motor neurons expressing EphA4 (Ohtaet al. 1997), the misprojection by LMCl axons inEphA4mutant mice could be dueto the failure of these axons to respond to the repellent ligands in the ventral limb.

Although the phenotype of LMCl axon trajectories inEphA4mutants does notperfectly correspond to that inLim1 mutants (compare Figure 6B with 6C), thefinding that both of these molecules are critically involved in the binary choice inaxonal trajectory at their specific decision point provides a potential link betweenthe downstream genes of a LIM code involving Lim1 and the Eph/ephrin signalingthrough EphA4 within the LMCl subtype.

CONTRIBUTION OF THE HOX CODEAND THE ETS CODE

Although the LIM code defines the columnar identity of motor neurons in thespinal cord, the currently known LIM-HD transcription factors are insufficientto account for the entire range of motor neuron subtypes. For example, the LIMcode alone does not distinguish the motor pool identity of limb-innervating LMCneurons. Lin et al. (1998) found that specific motor pools are defined by thecombinatorial expression of LIM-HD transcription factors and two closely-relatedETS transcription factors. In addition, the two ETS transcription factors, ER81 andPEA3, are also expressed in subsets of muscle sensory afferent neurons. Strikingly,they found a high degree of concordance in the profile of ETS protein expressionby sensory and motor neurons that project to the same muscle target (Lin et al.1998), thus revealing a molecular matching between functionally related sensoryand motor neurons. It is important to note that the onset of expression of ETStranscription factors occurs at a relatively late stage when LMC axons reach thebase of the limb (Lin et al. 1998). It has recently been shown that individual motorpools exhibit their characteristic features before LMC axons enter the limb (Milneret al. 1998, Milner & Landmesser 1999). Thus, it is likely that specification of motorpool identity occurs prior to the initiation of ETS gene expression. Indeed, inEr81mutant mice, the pool identity as well as the columnar identity of LMC neurons



do not appear to be perturbed, although Ia sensory afferents fail to establish theircharacteristic termination zone within the ventral spinal cord (Arber et al. 2000).

Several lines of evidence suggest that theHoxgenes, which have been implicatedin specifying positional identities along the anteroposterior axis of the caudalneural tube (for a review, see Lumsden & Krumlauf 1996), may contribute to theestablishment of the pool identity of motor neurons. Members ofHoxcandHoxdclusters have been shown to be expressed by subsets of LMC neurons (Ensini et al.1998, Tiret et al. 1998, Lance-Jones et al. 2001). In addition, the specification ofthe motor pool seems to occur prior to the onset of expression of the LIM code(Matise & Lance-Jones 1996), and the reprogramming of motor neuron subtypeproperties is associated with the reconfiguration ofHoxgene profiles (Lance-Joneset al. 2001). In support of a role forHoxgenes in motor neuron pool development,targeted disruptions ofHoxd9andHoxd10cause alterations in peripheral nerveprojections from specific motor pools (Carpenter et al. 1997, de la Cruz et al. 1999).Furthermore, inHoxc8deficient mice, the topographic organization of motor poolsthat innervate forelimb distal muscles becomes disorganized (Tiret et al. 1998).Thus, these findings support the idea that the action of the Hox code within motorneurons is a part of the mechanism that imposes on motor neurons the patterninginformation required for the fine tuning of subtype identity necessary to establishconnections between motor pools and muscle targets.

CONCLUSIONS

The ultimate outcome of neuronal fate specification is the accomplishment ofsynapse formation with appropriate target cells and the acquisition of appropri-ate physiological traits. Among a series of stages of the neuronal differentiationprogram, axon pathfinding that occurs in postmitotic neurons plays a fundamentalrole in the formation of functional neural circuits. Classical embryological studiesfocusing on topographic projections of spinal motor axons have suggested the ex-istence of intrinsic differences between motor neuron subtypes before these axonsselect their specific pathways to reach their proper target muscles. One of the keydiscoveries during the past decade is that unique combinatorial arrays of LIM-HDtranscription factors, the LIM code, define the target specificity of individual motorneuron subtypes. Recent molecular and genetic studies have provided compellingevidence to support the notion that the LIM code confers motor neuron subtypeswith the ability to select specific axon pathways to reach their targets.

However, a number of important questions remain to be addressed. For example,it is still unclear to what extent the initial intrinsic genetic programs can control thelater aspects of motor neuron development, such as positionally selective synapto-genesis by motor pools onto skeletal muscles (Laskowski & Sanes 1987, Feng et al.2000) and specific expression of ion channels on motor nerve endings. Notably,it has recently been shown that individual motor pools exhibit their characteris-tic patterned bursting activity even while motor axons are growing toward theirmuscle targets (Milner & Landmesser 1999). In addition, motor neurons within



a single pool can be further subdivided into two classes, based on whether theyinnervate fast or slow muscle fiber types. Recent studies have shown that, longbefore motor axons reach their muscle targets, fast- and slow-muscle innervatingmotor neurons are molecularly distinct in their axonal surface properties requiredfor their selective fasciculation and target recognition (Rafuse et al. 1996, Milneret al. 1998). These findings suggest that, although the LIM code explains neitherpool-specific nor subpool-specific properties, certain genetic programs involvedin regulation of their characteristic membrane surface properties on their growingaxons must be active prior to target innervation.

Finally, an important challenge is now to identify the downstream genes thatare directly regulated by axon pathfinding-related transcriptional codes. Implicitin the suggestion from the LIM code hypothesis is that the code determines thedifferential responsiveness of motor axons to guidance cues specific for a subtypeof motor neurons. In this context, it is reasonable to speculate that the receptor orreceptor-associated cofactor for sensing such guidance cues is a prime candidateto be among the downstream targets. Because the transcriptional codes seem tobe conserved through evolution, the identification of such key players will surelylead to the opening of a new and exciting chapter in this field of biology.

ACKNOWLEDGMENTS

Our research is supported by the National Institutes of Health, the ChristopherReeves Paralysis Foundation, the Human Frontier Science Program, and the Mus-cular Dystrophy Association. RS is supported by the Japan Society for the Pro-motion of Science Postdoctoral Fellowships for Research Abroad and SLP by thePew Charitable Trusts and the Mathers Foundation.

The Annual Review of Neuroscienceis online at http://neuro.annualreviews.org

LITERATURE CITED

Agulnick AD, Taira M, Breen JJ, Tanaka T,Dawid IB, Westphal H. 1996. Interactionsof the LIM-domain binding factor Ldb1with LIM homeodomain proteins.Nature384:270–72

Akazawa C, Ishibashi M, Shimizu C, Naka-nishi S, Kageyama R. 1995. A mammalianhelix-loop-helix factor structurally related tothe product of Drosophila proneural geneatonal is a positive transcriptional regulatorexpressed in the developing nervous system.J. Biol. Chem.270:8730–38

Appel B, Korzh V, Glasgow E, Thor S, EdlundT, et al. 1995. Motoneuron fate specification

revealed by patterned LIM homeobox geneexpression in embryonic zebrafish.Develop-ment121:4117–25

Araujo M, Piedra ME, Herrera MT, Ros MA,Nieto MA. 1998. The expression and regula-tion of chick EphA7suggests roles in limbpatterning and innervation.Development125:4195–204

Arber S, Han B, Mendelsohn M, Smith M,Jessell TM, Sockanathan S. 1999. Require-ment for the homeobox geneHb9 in the con-solidation of motor neuron identity.Neuron23:659–74

Arber S, Ladle DR, Lin JH, Frank E, Jessell TM.



2000. ETS geneEr81 controls the forma-tion of functional connections between groupIa sensory afferents and motor neurons.Cell101:485–98

Augsburger A, Schuchardt A, Hoskins S, DoddJ, Butler S. 1999. BMPs as mediators of roofplate repulsion of commissural neurons.Neu-ron 24:127–41

Bach I, Carriere C, Ostendorff HP, AndersenB, Rosenfeld MG. 1997. A family of LIMdomain-associated cofactors confer tran-scriptional synergism between LIM and Otxhomeodomain proteins.Genes Dev.11:1370–80

Bach I, Rhodes SJ, Pearse RV 2nd, Heinzel T,Gloss B, et al. 1995. P-Lim, a LIM home-odomain factor, is expressed during pitu-itary organ and cell commitment and syner-gizes with Pit-1.Proc. Natl. Acad. Sci. USA92:2720–24

Bang AG, Goulding MD. 1996. Regulation ofvertebrate neural cell fate by transcriptionfactors.Curr. Opin. Neurobiol.6:25–32

Ben-Arie N, McCall AE, Berkman S, EicheleG, Bellen HJ, Zoghbi HY. 1996. Evolution-ary conservation of sequence and expres-sion of the bHLH protein Atonal suggestsa conserved role in neurogenesis.Hum. Mol.Genet.5:1207–16

Blair SS, Brower DL, Thomas JB, ZavortinkM. 1994. The role ofapterousin the con-trol of dorsoventral compartmentalizationand PS integrin gene expression in the de-veloping wing ofDrosophila. Development120:1805–15

Bourgouin C, Lundgren SE, Thomas JB. 1992.apterousis a Drosophila LIM domain generequired for the development of a subset ofembryonic muscles.Neuron9:549–61

Briscoe J, Ericson J. 2001. Specification of neu-ronal fates in the ventral neural tube.Curr.Opin. Neurobiol.11:43–49

Briscoe J, Pierani A, Jessell TM, Ericson J.2000. A homeodomain protein code speci-fies progenitor cell identity and neuronal fatein the ventral neural tube.Cell 101:435–45

Briscoe J, Sussel L, Serup P, Hartigan-O’Connor D, Jessell TM, et al. 1999. Homeo-

box geneNkx2.2and specification of neu-ronal identity by graded sonic hedgehog sig-nalling.Nature398:622–27

Brose K, Bland KS, Wang KH, Arnott D, Hen-zel W, et al. 1999. Slit proteins bind Robo re-ceptors and have an evolutionarily conservedrole in repulsive axon guidance.Cell96:795–806

Burrill JD, Moran L, Goulding MD, Saueres-sig H. 1997. PAX2 is expressed in multiplespinal cord interneurons, including a pop-ulation of EN1+ interneurons that requirePAX6 for their development.Development124:4493–503

Carpenter EM, Goddard JM, Davis AP, NguyenTP, Capecchi MR. 1997. Targeted disruptionof Hoxd-10affects mouse hindlimb develop-ment.Development124:4505–14

Caton A, Hacker A, Naeem A, Livet J, Maina F,et al. 2000. The branchial arches and HGFare growth-promoting and chemoattractantfor cranial motor axons.Development127:1751–66

Chen H, Bagri A, Zupicich JA, Zou Y, StoeckliE, et al. 2000. Neuropilin-2 regulates the de-velopment of selective cranial and sensorynerves and hippocampal mossy fiber projec-tions.Neuron25:43–56

Chiang C, Litingtung Y, Lee E, Young KE, Cor-den JL, et al. 1996. Cyclopia and defectiveaxial patterning in mice lackingSonic hedge-hoggene function.Nature383:407–13

Cohen B, McGuffin ME, Pfeifle C, Segal D,Cohen SM. 1992.apterous, a gene requiredfor imaginal disc development in Drosophilaencodes a member of the LIM family of de-velopmental regulatory proteins.Genes Dev.6:715–29

Colamarino SA, Tessier-Lavigne M. 1995. Theaxonal chemoattractant netrin-1 is also achemorepellent for trochlear motor axons.Cell 81:621–29

Dawid IB, Breen JJ, Toyama R. 1998. LIM do-mains: multiple roles as adapters and func-tional modifiers in protein interactions.Trends Genet.14:156–62

de la Cruz CC, Der-Avakian A, SpyropoulosDD, Tieu DD, Carpenter EM. 1999. Targeted



disruption ofHoxd9 and Hoxd10alters lo-comotor behavior, vertebral identity, and pe-ripheral nervous system development.Dev.Biol. 216:595–610

Diaz-Benjumea FJ, Cohen SM. 1993. Interac-tion between dorsal and ventral cells in theimaginal disc directs wing development inDrosophila.Cell 75:741–52

Doe CQ, Skeath JB. 1996. Neurogenesis in theinsect central nervous system.Curr. Opin.Neurobiol.6:18–24

Driscoll M, Chalfie M. 1991. Themec-4geneis a member of a family of Caenorhabditiselegans genes that can mutate to induce neu-ronal degeneration.Nature349:588–93

Duggan A, Ma C, Chalfie M. 1998. Regula-tion of touch receptor differentiation by theCaenorhabditiselegansmec-3and unc-86genes.Development125:4107–19

Ebens A, Brose K, Leonardo ED, Hanson MGJr, Bladt F, et al. 1996. Hepatocyte growthfactor/scatter factor is an axonal chemoat-tractant and a neurotrophic factor for spinalmotor neurons.Neuron17:1157–72

Eberhart J, Swartz M, Koblar SA, PasqualeEB, Tanaka H, Krull CE. 2000. Expressionof EphA4, ephrin-A2 and ephrin-A5 duringaxon outgrowth to the hindlimb indicates po-tential roles in pathfinding.Dev. Neurosci.22:237–50

Edlund T, Jessell TM. 1999. Progression fromextrinsic to intrinsic signaling in cell fatespecification: a view from the nervous sys-tem.Cell 96:211–24

Eisen JS. 1991. Determination of primary mo-toneuron identity in developing zebrafish em-bryos.Science252:569–72

Eisen JS. 1994. Development of motoneuronalphenotype.Annu. Rev. Neurosci.17:1–30

Eisen JS, Pike SH. 1991. Thespt-1 mutationalters segmental arrangement and axonal de-velopment of identified neurons in the spinalcord of the embryonic zebrafish.Neuron6:767–76

Ensini M, Tsuchida TN, Belting HG, JessellTM. 1998. The control of rostrocaudal pat-tern in the developing spinal cord: specifi-cation of motor neuron subtype identity is

initiated by signals from paraxial mesoderm.Development125:969–82

Ericson J, Morton S, Kawakami A, Roelink H,Jessell TM. 1996. Two critical periods ofSonic Hedgehog signaling required for thespecification of motor neuron identity.Cell87:661–73

Ericson J, Rashbass P, Schedl A, Brenner-Morton S, Kawakami A, et al. 1997. Pax6controls progenitor cell identity and neuronalfate in response to graded Shh signaling.Cell90:169–80

Ericson J, Thor S, Edlund T, Jessell TM, Ya-mada T. 1992. Early stages of motor neu-ron differentiation revealed by expression ofhomeobox geneIslet-1. Science256:1555–60

Farah MH, Olson JM, Sucic HB, Hume RI, Tap-scott SJ, Turner DL. 2000. Generation of neu-rons by transient expression of neural bHLHproteins in mammalian cells.Development127:693–702

Fazeli A, Dickinson SL, Hermiston ML, TigheRV, Steen RG, et al. 1997. Phenotype of micelacking functional deleted in colorectal can-cer (Dcc) gene.Nature386:796–804

Feng G, Laskowski MB, Feldheim DA, WangH, Lewis R, et al. 2000. Roles of ephrinsin positionally selective synaptogenesis be-tween motor neurons and muscle fibers.Neu-ron 25:295–306

Ferguson EL, Horvitz HR. 1985. Identifica-tion and characterization of 22 genes that af-fect the vulval cell lineages of the nematodeCaenorhabditis elegans.Genetics110:17–72

Ferguson EL, Sternberg PW, Horvitz HR. 1987.A genetic pathway for the specification ofthe vulval cell lineages of Caenorhabditis el-egans.Nature326:259–67

Finney M, Ruvkun G. 1990. Theunc-86geneproduct couples cell lineage and cell identityin C. elegans.Cell 63:895–905

Finney M, Ruvkun G, Horvitz HR. 1988. TheC. elegans cell lineage and differentiationgeneunc-86encodes a protein with a home-odomain and extended similarity to tran-scription factors.Cell 55:757–69

Freyd G, Kim SK, Horvitz HR. 1990. Novel



cysteine-rich motif and homeodomain in theproduct of the Caenorhabditis elegans celllineage gene lin-11.Nature344:876–79

Giger RJ, Cloutier JF, Sahay A, Prinjha RK,Levengood DV, et al. 2000. Neuropilin-2 isrequired in vivo for selective axon guidanceresponses to secreted semaphorins.Neuron25:29–41

Gowan K, Helms AW, Hunsaker TL, CollissonT, Ebert PJ, et al. 2001. Crossinhibitory ac-tivities of Ngn1 and Math1 allow specifica-tion of distinct dorsal interneurons.Neuron31:219–32

Gradwohl G, Fode C, Guillemot F. 1996.Restricted expression of a novel murineatonal-related bHLH protein in undifferen-tiated neural precursors.Dev. Biol.180:227–41

Guillemot F. 1999. Vertebrate bHLH genes andthe determination of neuronal fates.Exp. CellRes.253:357–64

Gurdon JB, Bourillot P-Y. 2001. Morphogengradient interpretation.Nature413:797–803

Guthrie S, Pini A. 1995. Chemorepulsion of de-veloping motor axons by the floor plate.Neu-ron 14:1117–30

Hamelin M, Scott IM, Way JC, CulottiJG. 1992. Themec-7beta-tubulin gene ofCaenorhabditis elegans is expressed primar-ily in the touch receptor neurons.EMBO J.11:2885–93

Harrison KA, Druey KM, Deguchi Y, TuscanoJM, Kehrl JH. 1994. A novel human homeo-box gene distantly related to proboscipediais expressed in lymphoid and pancreatic tis-sues.J. Biol. Chem.269:19968–75

Helmbacher F, Schneider-Maunoury S, TopilkoP, Tiret L, Charnay P. 2000. Targeting of theEphA4 tyrosine kinase receptor affects dor-sal/ventral pathfinding of limb motor axons.Development127:3313–24

Helms AW, Johnson JE. 1998. Progenitorsof dorsal commissural interneurons are de-fined by MATH1 expression.Development125:919–28

Hirsch MR, Tiveron MC, Guillemot F, BrunetJF, Goridis C. 1998. Control of noradrener-gic differentiation and Phox2a expression by

MASH1 in the central and peripheral nervoussystem.Development125:599–608

Hobert O, D’Alberti T, Liu Y, Ruvkun G. 1998.Control of neural development and functionin a thermoregulatory network by the LIMhomeobox genelin-11. J. Neurosci. 18:2084–96

Hobert O, Westphal H. 2000. Functions of LIM-homeobox genes.Trends Genet.16:75–83

Inoue A, Takahashi M, Hatta K, Hotta Y,Okamoto H. 1994. Developmental regulationof islet-1 mRNA expression during neuronaldifferentiation in embryonic zebrafish.Dev.Dyn.199:1–11

Iwamasa H, Ohta K, Yamada T, Ushijima K,Terasaki H, Tanaka H. 1999. Expressionof Eph receptor tyrosine kinases and theirligands in chick embryonic motor neuronsand hindlimb muscles.Dev. Growth Differ.41:685–98

Jessell TM. 2000. Neuronal specification in thespinal cord: inductive signals and transcrip-tional codes.Nat. Rev. Genet.1:20–29

Jurata LW, Gill GN. 1997. Functional analysisof the nuclear LIM domain interactor NLI.Mol. Cell. Biol.17:5688–98

Jurata LW, Gill GN. 1998. Structure and func-tion of LIM domains.Curr. Top. Microbiol.Immunol.228:75–113

Jurata LW, Kenny DA, Gill GN. 1996. NuclearLIM interactor, a rhombotin and LIM home-odomain interacting protein, is expressedearly in neuronal development.Proc. Natl.Acad. Sci. USA93:11693–98

Jurata LW, Pfaff SL, Gill GN. 1998. The nuclearLIM domain interactor NLI mediates homo-and heterodimerization of LIM domain tran-scription factors.J. Biol. Chem.273:3152–57

Jurata LW, Thomas JB, Pfaff SL. 2000. Tran-scriptional mechanisms in the developmentof motor control.Curr. Opin. Neurobiol.10:72–79

Kageyama R, Nakanishi S. 1997. Helix-loop-helix factors in growth and differentiation ofthe vertebrate nervous system.Curr. Opin.Genet. Dev.7:659–65

Kania A, Johnson RL, Jessell TM. 2000. Co-ordinate roles for LIM homeobox genes in



directing the dorsoventral trajectory of motoraxons in the vertebrate limb.Cell 102:161–73

Karlsson O, Thor S, Norberg T, Ohlsson H, Ed-lund T. 1990. Insulin gene enhancer bindingprotein Isl-1 is a member of a novel classof proteins containing both a homeo- and aCys-His domain.Nature344:879–82

Kidd T, Brose K, Mitchell KJ, Fetter RD,Tessier-Lavigne M, et al. 1998. Roundaboutcontrols axon crossing of the CNS midlineand defines a novel subfamily of evolution-arily conserved guidance receptors.Cell 92:205–15

Kilpatrick TJ, Brown A, Lai C, Gassmann M,Goulding M, Lemke G. 1996. Expressionof the Tyro4/Mek4/Cek4gene specificallymarks a subset of embryonic motor neuronsand their muscle targets.Mol. Cell. Neurosci.7:62–74

Kimmel CB, Kane DA, Walker C, Warga RM,Rothman MB. 1989. A mutation that changescell movement and cell fate in the zebrafishembryo.Nature337:358–62

Kimmel CB, Warga RM, Kane DA. 1994. Cellcycles and clonal strings during formation ofthe zebrafish central nervous system.Devel-opment120:265–76

Kitsukawa T, Shimizu M, Sanbo M, HirataT, Taniguchi M, et al. 1997. Neuropilin-semaphorin III/D-mediated chemorepulsivesignals play a crucial role in peripheralnerve projection in mice.Neuron 19:995–1005

Korzh V, Edlund T, Thor S. 1993. Zebrafish pri-mary neurons initiate expression of the LIMhomeodomain protein Isl-1 at the end of gas-trulation.Development118:417–25

Lance-Jones C, Omelchenko N, Bailis A, LynchS, Sharma K. 2001.Hoxd10induction and re-gionalization in the developing lumbosacralspinal cord.Development128:2255–68

Landmesser L. 1978. The development of mo-tor projection patterns in the chick hind limb.J. Physiol.284:391–414

Landmesser LT. 1980. The generation of neu-romuscular specificity.Annu. Rev. Neurosci.3:279–302

Landmesser LT. 1992. Growth cone guidancein the avian limb: a search for cellular andmolecular mechanisms. InThe Nerve GrowthCone, ed. PC Letourneau, SB Kater, ERMacagno. pp. 373–85. New York: Raven

Laskowski MB, Sanes JR. 1987. Topographicmapping of motor pools onto skeletal mus-cles.J. Neurosci.7:252–60

Leber SM, Breedlove SM, Sanes JR. 1990. Lin-eage, arrangement, and death of clonally re-lated motoneurons in chick spinal cord.J.Neurosci.10:2451–62

Lee JE. 1997. Basic helix-loop-helix genes inneural development.Curr. Opin. Neurobiol.7:13–20

Lee KJ, Mendelsohn M, Jessell TM. 1998. Neu-ronal patterning by BMPs: a requirement forGDF7 in the generation of a discrete classof commissural interneurons in the mousespinal cord.Genes Dev.12:3394–407

Lee KJ, Jessell TM. 1999. The specification ofdorsal cell fates in the vertebrate central ner-vous system.Annu. Rev. Neurosci.22:261–94

Lee S-K, Pfaff SL. 2001. Transcriptional net-works regulating neuronal identity in thedeveloping spinal cord.Nat. Neurosci.4(Suppl.):1183–91

Lichtsteiner S, Tjian R. 1995. Synergistic acti-vation of transcription by UNC-86 and MEC-3 inCaenorhabditis elegansembryo extracts.EMBO J.14:3937–45

Li H, Witte DP, Branford WW, Aronow BJ,Weinstein M, et al. 1994.Gsh-4encodes aLIM-type homeodomain, is expressed in thedeveloping central nervous system and is re-quired for early postnatal survival.EMBO J.13:2876–85

Lin JH, Saito T, Anderson DJ, Lance-Jones C,Jessell TM, Arber S. 1998. Functionally re-lated motor neuron pool and muscle sensoryafferent subtypes defined by coordinateETSgene expression.Cell 95:393–407

Lu QR, Yuk D-I, Alberta JA, Zhu Z, Pawlitzky I,et al. 2000. Sonic hedgehog-regulated oligo-dendrocyte lineage genes encoding bHLHproteins in the mammalian central nervoussystem.Neuron25:317–29



Lumsden A, Krumlauf R. 1996. Patterningthe vertebrate neuraxis.Science274:1109–15

Lundgren SE, Callahan CA, Thor S, ThomasJB. 1995. Control of neuronal pathway se-lection by the Drosophila LIM homeodomaingene apterous.Development121:1769–73

Ma Q, Sommer L, Cserjesi P, Anderson DJ.1997. Mash1 and neurogenin1 expressionpatterns define complementary domains ofneuroepithelium in the developing CNS andare correlated with regions expressing notchligands.J. Neurosci.17:3644–52

Marin O, Anderson SA, Rubenstein JL. 2000.Origin and molecular specification of striatalinterneurons.J. Neurosci.20:6063–76

Matise MP, Lance-Jones C. 1996. A critical pe-riod for the specification of motor pools in thechick lumbosacral spinal cord.Development122:659–69

McCaig CD. 1986. Myoblasts and myoblast-conditioned medium attract the earliestspinal neurites from frog embryos.J. Physiol.375:39–54

Milner LD, Landmesser LT. 1999. Choliner-gic and GABAergic inputs drive patternedspontaneous motoneuron activity before tar-get contact.J. Neurosci.19:3007–22

Milner LD, Rafuse VF, Landmesser LT.1998. Selective fasciculation and divergentpathfinding decisions of embryonic chickmotor axons projecting to fast and slow mus-cle regions.J. Neurosci.18:3297–313

Mitani S, Du H, Hall DH, Driscoll M, ChalfieM. 1993. Combinatorial control of touch re-ceptor neuron expression inCaenorhabditiselegans. Development119:773–83

Mizuguchi R, Sugimori M, Takebayashi H,Kosako H, Nagao M, et al. 2001. Combina-torial roles of Olig2 and Neurogenin2 in thecoordinated induction of pan-neuronal andsubtype-specific properties of motoneurons.Neuron31:757–71

Moran-Rivard L, Kagawa T, Saueressig H,Gross MK, Burrill J, Goulding M. 2001.Evx1 is a postmitotic determinant of V0 in-terneuron identity in the spinal cord.Neuron29:385–99

Morrow EM, Furukawa T, Lee JE, Cepko CL.1999. NeuroD regulates multiple functionsin the developing neural retina in rodent.De-velopment126:23–36

Mueller BK. 1999. Growth cone guidance: firststeps towards a deeper understanding.Annu.Rev. Neurosci.22:351–88

Muhr J, Andersson E, Persson M, Jessell TM,Ericson J. 2001. Groucho-mediated tran-scriptional repression establishes progenitorcell pattern and neuronal fate in the ventralneural tube.Cell 104:861–73

Nakagawa Y, O’Leary DDM. 2001. Com-binatorial expression patterns of LIM-homeodomain and other regulatory genesparcellate developing thalamus.J. Neurosci.21:2711–25

Novak KD, Prevette D, Wang S, Gould TW, Op-penheim RW. 2000. Hepatocyte growth fac-tor/scatter factor is a neurotrophic survivalfactor for lumbar but not for other somaticmotoneurons in the chick embryo.J. Neu-rosci.20:326–37

Novitch BG, Chen AI, Jessell TM. 2001. Co-ordinate regulation of motor neuron subtypeidentity and pan-neuronal properties by thebHLH repressor Olig2.Neuron31:773–89

Ohta K, Iwamasa H, Drescher U, Terasaki H,Tanaka H. 1997. The inhibitory effect on neu-rite outgrowth of motoneurons exerted bythe ligands ELF-1 and RAGS.Mech. Dev.64:127–35

Ohta K, Nakamura M, Hirokawa K, Tanaka S,Iwama A, et al. 1996. The receptor tyrosinekinase, Cek8, is transiently expressed on sub-types of motoneurons in the spinal cord dur-ing development.Mech. Dev.54:59–69

Osumi N, Hirota A, Ohuchi H, Nakafuku M,Iimura T, et al. 1997. Pax-6 is involved inthe specification of hindbrain motor neuronsubtype.Development125:2961–72

Pfaff S, Kintner C. 1998. Neuronal diversifica-tion: development of motor neuron subtypes.Curr. Opin. Neurobiol.8:27–36

Pfaff SL, Mendelsohn M, Stewart CL, EdlundT, Jessell TM. 1996. Requirement for LIMhomeobox geneIsl1 in motor neuron gener-ation reveals a motor neuron-dependent step



in interneuron differentiation.Cell 84:309–20

Pierani A, Moran-Rivard L, Sunshine MJ,Littman DR, Goulding M, Jessell TM. 2001.Control of interneuron fate in the developingspinal cord by the progenitor homeodomainprotein Dbx1.Neuron29:367–84

Rafuse VF, Milner LD, Landmesser LT. 1996.Selective innervation of fast and slow mus-cle regions during early chick neuromusculardevelopment.J. Neurosci.16:6864–77

Riddle RD, Ensini M, Nelson C, Tsuchida T,Jessell TM, Tabin C. 1995. Induction of theLIM homeobox geneLmx1by WNT7a estab-lishes dorsoventral pattern in the vertebratelimb. Cell 83:631–40

Sagasti A, Hobert O, Troemel ER, Ruvkun G,Bargmann CI. 1999. Alternative olfactoryneuron fates are specified by the LIM home-obox genelim-4. Genes Dev.13:1794–806

Saha MS, Miles RR, Grainger RM. 1997.Dorsal-ventral patterning during neural in-duction in Xenopus: assessment of spinalcord regionalization with xHB9, a marker forthe motor neuron region.Dev. Biol.187:209–23

Sander M, Paydar S, Ericson J, Briscoe J,Berber E, et al. 2000. Ventral neural pattern-ing byNkxhomeobox genes:Nkx6.1controlssomatic motor neuron and ventral interneu-ron fates.Genes Dev.14:2134–39

Sarafi-Reinach TR, Melkman T, Hobert O, Sen-gupta P. 2001. Thelin-11 LIM homeoboxgene specifies olfactory and chemosensoryneuron fates inC. elegans. Development128:3269–81

Savage C, Hamelin M, Culotti JG, CoulsonA, Albertson DG, Chalfie M. 1989.mec-7is a beta-tubulin gene required for the pro-duction of 15-protofilament microtubules inCaenorhabditis elegans. Genes Dev.3:870–81

Scardigli R, Schuurmans C, Gradwohl G,Guillemot F. 2001. Crossregulation betweenNeurogenin2and pathways specifying neu-ronal identity in the spinal cord.Neuron31:203–17

Segawa H, Miyashita T, Hirate Y, Higashijima

S, Chino N, et al. 2001. Functional repres-sion of Islet-2 by disruption of complex withLdb impairs peripheral axonal outgrowth inembryonic zebrafish.Neuron30:423–36

Sengupta P, Colbert HA, Bargmann CI. 1994.The C. elegans geneodr-7 encodes anolfactory-specific member of the nuclear re-ceptor superfamily.Cell 79:971–80

Sharma K, Leonard AE, Lettieri K, Pfaff SL.2000. Genetic and epigenetic mechanismscontribute to motor neuron pathfinding.Na-ture406:515–19

Sharma K, Sheng HZ, Lettieri K, Li H, Kara-vanov A, et al. 1998. LIM homeodomain fac-tors Lhx3 and Lhx4 assign subtype identitiesfor motor neurons.Cell 95:817–28

Shirasaki R, Katsumata R, Murakami F. 1998.Change in chemoattractant responsiveness ofdeveloping axons at an intermediate target.Science279:105–7

Shirasaki R, Mirzayan C, Tessier-Lavigne M,Murakami F. 1996. Guidance of circumfer-entially growing axons by netrin-dependentand -independent floor plate chemotropismin the vertebrate brain.Neuron17:1079–88

Shirasaki R, Murakami F. 2001. Crossing thefloor plate triggers sharp turning of commis-sural axons.Dev. Biol.236:99–108

Siegler MV, Jia XX. 1999. Engrailed nega-tively regulates the expression of cell adhe-sion molecules connectin and neuroglian inembryonicDrosophilanervous system.Neu-ron 22:265–76

Smith ST, Jaynes JB. 1996. A conserved re-gion of engrailed, shared among all en-, gsc-,Nk1-, k Nk2- and msh-class homeoproteins,mediates active transcriptional repression invivo. Development122:3141–50

Sommer L, Ma Q, Anderson DJ. 1996. Neu-rogenins, a novel family of atonal-relatedbHLH transcription factors, are putativemammalian neuronal determination genesthat reveal progenitor cell heterogeneity inthe developing CNS and PNS.Mol. Cell.Neurosci.8:221–41

Takebayashi H, Yoshida S, Sugimori M,Kosako H, Kominami R, et al. 2000. Dy-namic expression of basic helix-loop-helix



Olig family members: implication of Olig2 inneuron and oligodendrocyte differentiationand identification of a new member, Olig3.Mech. Dev.99:143–48

Tamada A, Shirasaki R, Murakami F. 1995.Floor plate chemoattracts crossed axons andchemorepels uncrossed axons in the verte-brate brain.Neuron14:1083–93

Tanabe Y, William C, Jessell TM. 1998. Specifi-cation of motor neuron identity by the MNR2homeodomain protein.Cell 95:67–80

Tessier-Lavigne M, Goodman CS. 1996. Themolecular biology of axon guidance.Science274:1123–33

Thaler J, Harrison K, Sharma K, Lettieri K,Kehrl J, Pfaff SL. 1999. Active suppres-sion of interneuron programs within devel-oping motor neurons revealed by analysis ofhomeodomain factor HB9.Neuron23:675–87

Thor S, Andersson SG, Tomlinson A, ThomasJB. 1999. A LIM-homeodomain combinato-rial code for motor-neuron pathway selec-tion. Nature397:76–80

Thor S, Ericson J, Brannstrom T, Edlund T.1991. The homeodomain LIM protein Isl-1 isexpressed in subsets of neurons and endo-crine cells in the adult rat.Neuron7:881–89

Thor S, Thomas JB. 1997. The Drosophilaisletgene governs axon pathfinding and neuro-transmitter identity.Neuron18:397–409

Tiret L, Le Mouellic H, Maury M, Brulet P.1998. Increased apoptosis of motoneuronsand altered somatotopic maps in the brachialspinal cord ofHoxc-8-deficient mice.Devel-opment125:279–91

Tokumoto M, Gong Z, Tsubokawa T, HewCL, Uyemura K, et al. 1995. Molecular het-erogeneity among primary motoneurons andwithin myotomes revealed by the differentialmRNA expression of novel islet-1 homologsin embryonic zebrafish.Dev. Biol.171:578–89

Tosney KW. 1987. Proximal tissues and pat-terned neurite outgrowth at the lumbosacrallevel of the chick embryo: deletion of the der-mamyotome.Dev. Biol.122:540–58

Tosney KW. 1988. Proximal tissues and pat-

terned neurite outgrowth at the lumbosacrallevel of the chick embryo: partial and com-plete deletion of the somite.Dev. Biol.127:266–86

Tosney KW, Landmesser LT. 1985a. Devel-opment of the major pathways for neuriteoutgrowth in the chick hindlimb.Dev. Biol.109:193–214

Tosney KW, Landmesser LT. 1985b. Growthcone morphology and trajectory in the lum-bosacral region of the chick embryo.J. Neu-rosci.5:2345–58

Tsuchida T, Ensini M, Morton SB, BaldassareM, Edlund T, et al. 1994. Topographic orga-nization of embryonic motor neurons definedby expression of LIM homeobox genes.Cell79:957–70

Vallstedt A, Muhr J, Pattyn A, Pierani A,Mendelsohn M, et al. 2001. Different lev-els of repressor activity assign redundant andspecific roles toNkx6 genes in motor neu-ron and interneuron specification.Neuron31:743–55

Varela-Echavarria A, Pfaff SL, Guthrie S. 1996.Differential expression of LIM homeoboxgenes among motor neuron subpopulationsin the developing chick brain stem.Mol. Cell.Neurosci.8:242–57

Varela-Echavarria A, Tucker A, Puschel AW,Guthrie S. 1997. Motor axon subpopulationsrespond differentially to the chemorepellentsnetrin-1 and semaphorin D.Neuron18:193–207

Vogel A, Rodriguez C, Warnken W, Izpisua Bel-monte JC. 1995. Dorsal cell fate specified bychick Lmx1 during vertebrate limb develop-ment.Nature378:716–20

Way JC, Chalfie M. 1988.mec-3, a homeobox-containing gene that specifies differentiationof the touch receptor neurons in C. elegans.Cell 54:5–16

Wilkinson DG. 2001. Multiple roles of EPH re-ceptors and ephrins in neural development.Nat. Rev. Neurosci.2:155–64

Xue D, Tu Y, Chalfie M. 1993. Cooperativeinteractions between the Caenorhabditis el-egans homeoproteins UNC-86 and MEC-3.Science261:1324–28



Yuan W, Zhou L, Chen JH, Wu JY, Rao Y, OrnitzDM. 1999. The mouse SLIT family: secretedligands for ROBO expressed in patterns thatsuggest a role in morphogenesis and axonguidance.Dev. Biol.212:290–306

Zhou Q, Choi G, Anderson DJ. 2001. ThebHLH transcription factor Olig2 promotesoligodendrocyte differentiation in collabora-tion with Nkx2.2.Neuron31:791–807

Zhou Q, Wang S, Anderson DJ. 2000. Identifi-cation of a novel family of oligodendrocytelineage-specific basic helix-loop-helix tran-scription factors.Neuron25:331–43

Zou Y, Stoeckli E, Chen H, Tessier-Lavigne M.2000. Squeezing axons out of the grey mat-ter: a role for slit and semaphorin proteinsfrom midline and ventral spinal cord.Cell102:363–75

25 May 2002 11:59 AR AR160-08-COLOR.tex AR160-08-COLOR.SGM LaTeX2e(2002/01/18)P1: GDL

Figure 3 Columnar organization of motor neuron subtypes in the chick spinal cordand the target specificity of motor neuron subtypes are defined by distinct combinationsof LIM-HD transcription factors. In the spinal cord, motor neurons are organized intolongitudinal columns, as shown in the open-book whole-mount view on theleft. In thispreparation, the center is the floor plate (grey) at the ventral midline of the neural tube.The axonal pathways of motor neuron subtypes are represented in transverse sectionsat brachial and thoracic levels on theright, respectively. The expression of LIM-HDtranscription factors in individual motor neuron subtypes, i.e., the LIM code for motorneuron subtypes, is shown by color coding. bw, body wall musculature; dlb, dorsallimb bud musculature; dm, dermomyotome; sg, neurons of the sympathetic ganglia;vlb, ventral limb bud musculature. LMCl (blue), lateral half of lateral motor column;LMCm (green), median half of lateral motor column; MMCl (orange), lateral halfof medial motor column; MMCm (red), median half of medial motor column; PMC(purple), preganglionic motor column (also called the column of Terni in chick).Lhx4(also termedGsh-4) is a gene highly related toLhx3in mouse (Li et al. 1994). Becausechick and mouse LIM-HD genes typically exhibit similar expression,Lhx4 is likelyexpressed in chick MMCm neurons as well.


Fig

ure

5A

LIM

code

for

MM

Cm

neur

ons

conf

ers

axon

sof

this

subt

ype

with

the

abili

tyto

sele

cta

spec

ific

path

way

tow

ard

thei

rm

uscl

eta

rget

s,th

ede

rmom

yoto

me.

(A

)A

LIM

code

for

MM

Cm

neur

ons

isde

fined

byth

eun

ique

expr

essi

onof

Lhx3

and

redu

ndan

tfac

tor

Lhx4

inco

mbi

natio

nw

ithIs

l1/2

.(B)

Tote

stw

heth

erth

esu

btyp

e-sp

ecifi

cLI

Mco

deco

ntro

lsm

otor

axon

path

way

sele

ctio

nin

the

perip

hery

acco

rdin

gto

the

LIM

code

hypo

thes

is,

allm

otor

neur

ons

inm

ice

wer

ege

netic

ally

forc

edto

stab

lyex

pres

sLh

x3in

orde

rto

esta

blis

han

MM

Cm

-like

code

.Thi

sge

netic

alte

ratio

nis

suffi

cien

tto

conv

ertt

hece

llbo

dyse

ttlin

gpa

ttern

,gen

e-ex

pres

sion

profi

lean

dax

onal

proj

ectio

npa

ttern

sof

mot

orne

uron

sto

that

ofth

eM

MC

msu

btyp

e.H

owev

er,

owin

gto

aco

mpe

titiv

ein

tera

ctio

nin

embr

yos

with

larg

enu

mbe

rsof

MM

Cm

cells

,som

eax

ons

are

dive

rted

from

the

derm

omyo

tom

e(n

otsh

own)

.


Fig

ure

6LM

Cla

xon

path

findi

ngat

the

base

ofth

elim

b.(

A)

Lim

1m

arks

the

LMC

lsub

type

,an

dE

phA

4is

pref

eren

tially

expr

esse

din

LMC

laxo

nsw

hen

they

exte

ndin

toth

edo

rsal

limb.

At

the

bifu

rcat

ion

poin

t,LM

Cla

xons

(b

lue)

sele

cta

dors

alpa

thw

ay,

whi

leLM

Cm

axon

s(gr

ee

n)se

lect

ave

ntra

ltra

ject

ory

toen

ter

the

limb

bud.

(B

)In

Lim

1m

utan

tm

ice,

LMC

laxo

nspr

ojec

tin

toth

edo

rsal

and

vent

ral

path

way

sw

ithin

the

limb

ateq

uali

ncid

ence

.In

this

mut

ant,

the

proj

ectio

nsof

LMC

max

ons

are

nota

ffect

ed.T

hus,

the

LIM

code

forL

MC

lne

uron

sco

mpo

sed

ofLi

m1

and

Isl2

cont

ribut

esto

the

fidel

ityof

abi

nary

choi

cein

axon

altr

ajec

tory

atth

eba

seof

the

limb.

(C

)In

Ep

hA

4m

utan

tmic

e,LM

Cla

xons

fail

toin

vade

the

dors

alco

mpa

rtm

ento

fthe

limb

and

inst

ead

sele

cta

vent

ralp

athw

ayto

ente

rth

eve

ntra

llim

bm

esen

chym

e.

ryuichi shirasaki and samuel l. pfaff - brandeis · pdf fileryuichi shirasaki and samuel l....

Documents