- Email: [email protected]
Correlated patterns of neuron differentiation and Hox gene expression in the hindbrain: a comparative analysis
Brain Research Bulletin, Vol. 55, No. 6, pp. 683– 693, 2001 Copyright © 2001 Elsevier Science Inc. Printed in the USA. All rights reserved 0361-9230/01/$–see front matter
PII S0361-9230(01)00562-7
Evolution of the Nervous System
Correlated patterns of neuron differentiation and Hox gene expression in the hindbrain: A comparative analysis Joel C. Glover* Department of Physiology, University of Oslo, Oslo, Norway ABSTRACT: Hindbrain neurons are organized into coherent subpopulations with characteristic projection patterns and functions. Many of these serve vital functions that have been conserved throughout the vertebrate radiation, but diversification to modified or highly specialized functions has also occurred. The differentiation of identifiable neuron groups in specific spatial domains must involve the regional expression of determinants within the hindbrain neuroepithelium. The Hox genes are involved in longitudinal regionalization of the neural tube, and their expression patterns in the hindbrain are closely related to the rhombomeres which partition the hindbrain into morphogenetic units. Hox gene expression also exhibits conserved patterning as well as phylogenetic variation. One plausible mechanism that may have contributed to evolutionary diversification in hindbrain neuron populations is therefore the emergence of species-specific differences in Hox gene expression. This article presents a comparative overview of the regional patterning of selected Hox genes and hindbrain neuron populations in several embryologically important species. Although tantalizing correlations exist, the relationship between Hox genes and neuronal patterning is complex, and complicated by dynamic features in each. Much more comparative and developmental data must be obtained before the link between Hox gene expression and hindbrain neuron patterning can be elucidated satisfactorily in an evolutionary context. © 2001 Elsevier Science Inc.
neuromeres, called rhombomeres, which provide a morphological grid onto which gene expression and neuron group domains can be plotted (Fig. 1). Direct and indirect comparison of neuron groups with relevant genes as well as manipulations of gene expression can therefore be carried out with relatively high spatial precision. This article presents examples of phylogenetic variation in selected hindbrain neuron populations and important developmental regulatory genes, focusing on key species commonly used in neuroembryological studies. The link between the relevant genes and the specification of these neuronal phenotypes is still poorly understood. Nevertheless, enough evidence has accumulated in favor of such a link that an underlying hypothesis can be entertained here: namely that phylogenetic constancy and variation in the expression patterns of genes such as these contribute substantially to the observed conservation and divergence of hindbrain neuron populations. SPECIFIC NEURON TYPES EXPRESS SPECIFIC COMBINATIONS OF TRANSCRIPTION FACTORS That neuronal identity might be specified by the regional expression of particular classes of transcription factors first became clear in studies of insect neurogenesis, where genetic approaches to understanding neuronal differentiation were pioneered (see e.g, [8,36]). In insects, each of the neuroblasts (neuronal progenitors) within a segmental ganglion is individually identifiable by virtue of its position and pattern of gene expression, and gives rise to a specific subset of the neurons in the ganglion. Region-specific gene expression as an important determinant of neuronal identity is also a prominent feature in the central nervous system of vertebrates. Studies of spinal motoneuron and interneuron development have provided exceptionally illustrative examples [21]. Specification of these neurons occurs in a step-wise fashion. Graded signals from midline structures induce the differentiation of progenitor cells in localized serial dorsoventral zones. Reciprocal repressive interactions then sharpen these zones into highly restricted populations of progenitors characterized by the expression of specific transcription factors. These then generate polyclones of neuronal progeny which are also distinguished by
KEY WORDS: Segmentation, Zebrafish, Chicken, Mouse, Cranial motoneurons, Reticulospinal, Vestibulospinal, Vestibuloocular, Dextranamines.
INTRODUCTION How have specific neuron populations in the vertebrate brain evolved? To answer this question it is necessary to identify which developmental regulatory genes are responsible for specifying different neuron types, and to assess how the relationship between genes and phenotypes varies among species. The hindbrain is one region of the brain that is especially well suited for such an analysis. During the period when most of its constituent neuron groups are generated, the hindbrain is transiently segmented into
* Address for correspondence: Joel C. Glover, Department of Physiology, P.B. 1103 Blindern, 0317 Oslo, Norway. E-mail: [email protected]
683
684
GLOVER
FIG. 1. Side view of a chicken embryo at Hamburger and Hamilton stage 14 (HH14). The brain is outlined (dotted line), showing rhombomeres 1 through 7 (rh1–7) and their relationship to cranial nerves (n). Other structures indicated in the hindbrain region are the otic vesicle (Otv) and the notochord (hatched). Note also the indication of neuromeres in midbrain and forebrain (see Puelles, this issue). Drawing from [45].
FIG. 2. Schematic overview of transcription factor gene expression by different motoneuron (MN) and ventral interneuron (IN) classes in a transverse section of the spinal cord (left). Ventral IN classes (V0 –V3) have been named according to differential transcription factor expression and their relative positions as they emerge from the proliferative ventricular zone (vz). These IN classes and various MN classes are distinguished moreover by their axon trajectories or peripheral targets Abbreviations: aCIN, ascending commissural IN; aIIN, ascending ipsilateral IN; adIIN, ascending and descending ipsilateral IN; symp, sympathetic ganglia; v. axial and d. axial, ventral and dorsal axial muscles, respectively; d. limb and v. limb, dorsal and ventral limb muscles, respectively. Specific projection phenotypes are thus related to specific patterns of expression of the transcription factors listed along the top of the table. Data from [30,35,44].
HINDBRAIN NEURONS AND GENE EXPRESSION
685
FIG. 3. Number of reticulospinal neurons projecting to one side of the spinal cord (half of the total number due to bilateral symmetry), in the zebrafish larva [29], the ranid frog larva [39], the 7-day chicken embryo [17], and the E15 rat embryo [2].
specific transcription factors that are believed to steer the readout of cell-type specific proteins involved in selective migration, axon outgrowth, target cell recognition, and molecular phenotype. A number of transcription factors have been identified whose combinatorial expression patterns define specific functional classes of spinal neurons (Fig. 2), including motoneurons differentiated according to peripheral axon projection pathway and stretch reflex connectivity pattern, and interneurons differentiated according to central axon projection pathways [21]. Many of the same transcription factors are expressed in the hindbrain, where similar relationships to neuron type exist [46]. Comprehensive mapping of the expression of these and similar genes will undoubtedly unveil much of the underlying genetic blueprint that specifies the diverse neuron groups in the hindbrain. This endeavour has hardly begun, however, making any attempt at a comparative analysis of the expression of these genes and neuronal phenotype grossly premature. On the other hand, transcription factors that imbue the different rhombomeres with unique identities have been studied more extensively in a comparative context. One class of these regionally specific genes, the Hox genes, will be featured in this article. Experimental manipulation of the expression of Hox genes leads to changes in the way neurons differentiate, indicating that they play an instructive role in neuron type determination [4,18,22,23]. A comparative review of the expression patterns of selected Hox genes will be presented in a later section following a phylogenetic overview of hindbrain neuron patterning. HINDBRAIN NEURON POPULATIONS HAVE EXPANDED DURING PHYLOGENESIS In general, the size of hindbrain neuron populations has increased as the vertebrate lineage evolved and branched. Documentation has only been obtained for a few neuron populations in a few species, and is furthermore complicated by counts being performed in different species at different stages during the life cycle (embryo, larva, adult), and by identification of neuron populations by different labeling paradigms. Nevertheless, it is quite clear that amniotes (reptiles, birds, mammals) have more highly populated
hindbrains than anamniotes. Among amniotes, mammals evidently have more neurons than birds and reptiles. For example, a graph of reticulospinal neuron number against vertebrate class shows a virtually logarithmic increase in the number of combined pontine and medullary reticulospinal neurons on comparison of fish to frog to chicken to rat at developmental stages of similar hindbrain maturity (Fig. 3). Similar numerical differences exist for vestibular projection neurons (vestibulospinal and vestibulo-ocular neurons, for example), although quantitation has not yet been performed to an extent that can be summarized graphically [7,17,24,32,39,42]. Regarding the data for reticulospinal neurons in Fig. 3, it should be pointed out that with increasing development in each species, the number of reticulospinal neurons is likely (or in some cases is known) to increase. This increase deserves special comment. In many fish, and perhaps other anamniotes, certain neuron populations increase in number well into adult life [9]. This protracted proliferation evidently is a special feature not seen in amniotes. It is presumably an adaptation to a continual increase in body size that occurs during the lifetime of many anamniotes, with commensurate increases in the sizes of muscles and of sensory organs. An adult fish has substantially more reticulospinal neurons, for example, than does a larval fish [27,28,33,42]. The phylogenetic variation in the population size of specific hindbrain neuron groups, on the other hand, is not related to body size. A diminutive rat embryo has far more reticulospinal neurons than a beefy adult catfish. Rather, the variation is likely to be related to evolutionary increases in the complexity of motor system regulation (especially with increasing cortical control) and to elaboration of the autonomic nervous system and its supraspinal regulation. Such factors should be borne in mind when making species comparisons. For example, excellent quantitative data for reticulospinal neurons in the lamprey [43] have not been included in Fig. 3 because it is difficult to relate the developmental stage of the lamprey specimens to the species shown. Swain et al. [43] studied larval lamprey that were 5 years old (10 –15 cm in length),
686 and counted on the order of 700 reticulospinal neurons per set (half of the bilaterally paired total). This elevates the lamprey past fish, frog, and chicken in Fig. 3, and is therefore most likely a reflection of maturity rather than evolutionary status or behavioral repertoire: simply by dint of living longer, a 5-year-old lamprey has had greater opportunity to add reticulospinal neurons to its hindbrain through ongoing proliferation. SPECIALIZED HINDBRAIN NEURON POPULATIONS EXIST IN SOME SPECIES Another important point to bear in mind is that certain species may possess special neuron populations related to special functions. This is most obvious with respect to specialized sensory systems, which are typically associated with characteristic central nuclei, for example the octavolateral nuclei in aquatic animals with a lateral line, or the pacemaker nucleus of electric fish [3]. Specialized neuron groups may be unique modifications of “generic” populations present in all species, or may be “extra” groups added through the proliferation, division, and elaboration of pre-existing populations. Thus, although elaboration seems to be a hallmark of the general evolutionary increase in the size and complexity of many hindbrain neuron populations (see below), it may also generate isolated phylogenetic features commensurate with parallel specializations elsewhere in the body. Along the same lines, some “generic” nuclei are completely missing in certain species. For example, lampreys, salmon, and electric fish lack raphe nuclei [3,31,43] and lampreys lack a specific oculomotor motoneuron pool [12]. Superficial comparisons may therefore be misleading— neuron groups must be compared with great attention to phenotypic detail and embryonic history (lineage, migratory patterns) to avoid false conclusions about evolutionary relationships. SOME, BUT NOT ALL HINDBRAIN NEURON POPULATIONS BEAR A CONSERVED RELATIONSHIP TO SPECIFIC RHOMBOMERIC DOMAINS Despite the tremendous expansion of hindbrain neuron populations during vertebrate evolution, in all vertebrate classes specific neuron populations occupy coherent spatial domains that can be related, directly or indirectly, to the overt morphological architecture of the hindbrain as delineated by the rhombomeres. A point of difficulty in assessing this regional relationship is the transient nature of the rhombomere boundaries, which itself is speciesspecific. In some species like the chicken and the frog, rhombomere boundaries are visible for an extended period during which most hindbrain neuron groups have been generated and can be mapped. In other species like the mouse, the boundaries are erased before it is possible to map many groups, and a regional analysis must depend on fate-mapping techniques or indirect cues such as lasting rhombomere-related landmarks or gene expression patterns. This is also important to remember when reviewing the existing literature, which is not without its fallacies. The rhombomere-related domain of a given neuron group is in some cases very similar across species, whereas in other cases it differs considerably. Some aspects of hindbrain regionalization are therefore highly conserved, whereas others have diverged. Cranial Motoneurons and Efferents to Sensory Organs These are the hindbrain neuron populations that have been most studied, because they are relatively easy to map unambiguously. Retrograde tracers such as DiI [19] or dextran-amines [14,16] can be applied to specific cranial nerve roots, thereby labeling a defined population of efferent neurons. Greater care must be taken when assessing the spatial relationships of subpopulations within a
GLOVER cranial motor nucleus, however. In this case, tracer applications must be targeted to specific nerve branches or peripheral targets (see, e.g., [1]). A number of studies have mapped cranial motoneuron populations in embryos and larvae of a variety of species. These studies have been reviewed recently [11], and are summarized in Fig. 4. The first thing to note is that in general the relative locations of the different cranial efferent populations are similar in different species, and can be related roughly to the same rhombomeres. The second thing to note is that despite this similarity, there are obvious differences, some striking, some subtle. In all species examined, for example, trigeminal motoneurons are found in r2 and r3. Yet the domain occupied may extend into adjacent r1 (chicken, mouse) or r4 (lamprey, mouse), or may not span the full length of r3 (elasmobranch, chicken). The trigeminal motoneurons may even split into two widely separate domains (elasmobranch, r2-3 and r6). Similarly, in nearly all species examined, abducens motoneurons are found in r5, but they may also be found in r6 (lamprey, teleost, reptile, chicken), extend into r4 (lamprey), or be localized to r6 instead of r5 (elasmobranch). The differences are even more complex for the facial motor nucleus and its subdivisions. This brings up a crucial issue when considering the relationships between neuronal domains and gene expression: it is not uncommon for neurons generated in one position to migrate en masse to another position. Migration of motoneurons has been well documented for several hindbrain motoneuron populations, including the oculomotor, trigeminal, abducens, and facial motoneurons. The migratory patterns themselves may exhibit species-related conservation and divergence (see [11]). The upshot is that using neuron position alone for assessing evolutionary relationships is risky— one must assess where a neuron population originates as well, because the expression of genes in the region of origin may be the principal determinant. Assessing origin is not always easy. It may require retrograde labeling at the very earliest stages of axon outgrowth (not equally feasible for all organisms). If motoneurons have already migrated prior to extending axons, it may additionally require fate-mapping techniques. Because of these extra difficulties, the rhombomeric patterning described in many species is still deficient in this respect. Nevertheless, several studies have managed to pinpoint the origins of some of the cranial motor nuclei in a few species. The facial nucleus is perhaps the best-studied case. Figure 5 shows where facial branchial motoneurons are believed to originate in four different species. Again, there are clear differences. Facial branchial motoneurons originate predominantly from r4 in frog, chicken, and mouse, but there are minor contributions from neighboring rhombomeres in chicken and mouse. In lamprey the predominant origin is r5, with a smaller contribution from r4. It would appear that both the genesis and final location of cranial motoneurons is linked to different domains of the hindbrain in different species. Reticulospinal Neurons After the cranial motoneurons, reticulospinal neurons are probably the most thoroughly investigated hindbrain neuron population. These make up a composite population comprising many functionally distinct neuron groups, which are often lumped together because they are easy to label simultaneously (tracers are applied to the hindbrain/spinal cord junction). A number of studies have mapped the positions of reticulospinal neurons with specific axon trajectories (unilateral tracer applications discriminate between ipsilaterally and contralaterally projecting neurons) as they relate to the rhombomeric divisions. In all species examined, each rhombomere contains reticulospinal neurons, and many of these share certain characteristics such as mediolateral location and
HINDBRAIN NEURONS AND GENE EXPRESSION
687
FIG. 4. Comparison of the rhombomeric domains of selected cranial motor nuclei (trigeminal, abducens, facial) in different species. Each image depicts schematically a frontal view of the hindbrain with rhombomeres 2– 6 indicated on each side. The two sides are separated by the midline floor plate. The trigeminal and facial nuclei are shown on the left side, and the abducens nucleus is shown on the right side. Mediolateral locations are approximate. The subdivisions of the facial nerve have not been individually characterized in all species. In such cases the domain of the facial nerve nucleus is divided diagonally to show which subpopulations are present, but not their longitudinal limits. Adapted from [11].
dendritic morphology. A common misconception in the literature, however, is that the reticulospinal neurons represent a segmentally iterated population (implying that each rhombomere contains the same set of reticulospinal phenotypes). The actual picture is one of segmental variation, which has been suggested to arise through diversification of ancestral segmentally iterated homologues [29]. Careful observation (and reading) demonstrates this fact even for the zebrafish (see [29]), which has been erroneously promulgated as an archetype of segmental iteration. That segmental variation is
FIG. 5. Domains of origin of facial branchial motoneurons in four different species.
a key developmental and evolutionary feature is particularly evident when comparing anamniotes to amniotes. In the latter, the reticulospinal population has become more elaborate (in parallel with the dramatic increase in neuron number—Fig. 3), with the evolutionary addition of extra groups and migratory patterns in specific subsets of rhombomeres [2,17]. Figure 6 illustrates the domains of ipsilaterally and contralaterally projecting reticulospinal neurons in zebrafish, frog, chicken, and rodent (rat and mouse). As a clear illustration of segmental as well as phylogenetic variation, note that in each species there is a specific rhombomeric domain that lacks contralaterally projecting reticulospinal neurons, but that this domain differs among the species (r3 in zebrafish, r6 in frog, r5 in chicken, mouse, and rat). Note also that spatial elaboration of the reticulospinal population is especially evident in the mouse and rat, where major clusters spawn subclusters that become displaced differentially to more lateral positions, thereby creating satellite groups corresponding to specific functionally identifiable reticular nuclei [2]. As is the case for the cranial motoneurons, the reticulospinal neurons occupy different rhombomeric domains in different species. Moreover, the domain boundaries of individual clusters do
688
GLOVER
FIG. 6. Domains of reticulospinal neurons in four different species. Contralaterally projecting and ipsilaterally projecting reticulospinal neurons are shown on the left and right side, respectively, of each hindbrain (the arrow indicates the side on which axons descend to the spinal cord). Note that the number of reticulospinal neurons varies greatly (see Fig. 3), such that the neuron density per domain increases substantially from zebrafish up to mouse.
not always relate precisely to rhombomere boundaries, but may extend variable distances into one rhombomere from another. Whether this non-congruence results from longitudinal migration or spatial determination out of register with the rhombomeres is not yet known. Vestibular Projection Neurons Neurons in the vestibular nuclei and associated nuclei project to several major targets, including the spinal cord, the cerebellum, the motor nuclei that steer the extraocular muscles (oculomotor, trochlear, abducens), and the contralateral vestibular nuclei. Of these subpopulations, the vestibulospinal and vestibulo-ocular have been characterized best in the context of spatial patterning within the hindbrain neuroepithelium. Studies in the chicken embryo have shown that vestibulospinal and vestibulo-ocular neurons are organized in clusters, each of which projects axons along a specific trajectory to innervate a specific target or targets (reviewed in [15]). This feature has been termed the “hodological mosaic,” because neuron groups projecting to different targets are to a great extent segregated [13]. Later studies in anamniotes have demonstrated that this patterning feature is conserved in the vertebrate lineage, but have also documented divergence in the pattern [39, 42]. Figure 7 shows the domains of vestibulospinal and vestibuloocular neurons as they relate to the rhombomeres in those species where this has been characterized. Again, certain elements of the pattern are similar in representatives of different vertebrate classes. Fish, frog, and chicken each have, within a domain roughly corresponding to r4, an ipsilaterally projecting vestibulospinal group with axons coursing in a lateral tract (the same is true in mouse and rat; [2]). They each have a contralaterally projecting vestibulospinal group roughly localized to r5, and 2 contralaterally projecting vestibulo-ocular groups in separate domains (one roughly in r2, and the other within r4 –r7). There are clear differences, however, in the longitudinal extent of several groups, and some groups are absent in zebrafish at the larval stages studied (namely, the caudal groups of ispilaterally projecting vestibulospinal and vestibuloocular neurons). There are also species differences in the degree to which vestibulospinal and vestibulo-ocular groups overlap along the longitudinal axis (see especially the ipsilaterally projecting groups).
As is the case for cranial motor nuclei and reticulospinal neurons, the relative roles of positional determination versus longitudinal migration in establishing these vestibular neuron domains is poorly characterized in all species. Non-congruence between rhombomeres and neuron domains in the vestibular population probably results from both of these mechanisms (see [7]). HOX GENES HAVE SIMILAR EXPRESSION PATTERNS IN DIFFERENT SPECIES, BUT DIFFERENCES EXIST Hox genes are evolutionary homologues of transcription factor genes that were originally identified in Drosophila, within its homeotic gene complex [25]. Hox genes have since been identified in all vertebrates. Their number increased in the vertebrate lineage as it diverged from the invertebrate, with duplications of the original homeotic complex eventually generating four different Hox gene complexes in mammals (A, B, C, and D), and further expansion within the complexes generating additional genes (Fig. 8). The peculiar feature of Hox gene expression is that the rostrocaudal and temporal order of expression within the body follow the order of the genes in the chromosome, a property termed colinearity. Hox gene nomenclature emphasizes this feature: within each complex, the genes are numbered in order of rostral to caudal expression domain, maintaining rank alignments according to sequence similarity (and thereby to presumed evolutionary relation). The rostrocaudal ordering has a few exceptions (e.g., hoxb2 is expressed rostral to hoxb1), but is surprisingly well conserved. Because the different complexes are the result of duplications, Hox genes with the same rank in different complexes are evolutionary relatives, and are therefore called paralogues. The first four paralogue groups have expression domains within the hindbrain. Figures 9 and 10 show the expression domains of four Hox genes from the first three paralogue groups (those expressed within the segmented region of the hindbrain). Dark and light domains represent areas of high and low expression, respectively. The rostral limit of expression is the most conserved feature for each gene (although this is not an absolute rule for Hox genes, especially when low expression domains are considered; see hoxd3). Certain high expression domains are also conserved. There are, nevertheless, clear species differences evident in these patterns.
HINDBRAIN NEURONS AND GENE EXPRESSION
689
FIG. 7. Domains of vestibulo-ocular (top) and vestibulospinal (bottom) neuron groups in three different species. Axon trajectories are indicated by thin lines. Ipsilaterally and contralaterally projecting groups are indicated on right and left sides, respectively, with large arrows showing the side of projection. Data from [15,39,42].
These include the caudal limits of expression and the specific domains of high versus low expression. The former feature is generally considered to be functionally irrelevant, because more caudally expressed paralogues are considered to dominate transcriptional regulation in more caudal regions. The second feature, especially as it relates to the rostral limits, may have more impact on the patterning of differentiating neurons. HOW RELATED ARE HOX GENE EXPRESSION AND HINDBRAIN NEURON GROUP DOMAINS? Hox genes are known to play important roles in specifying regional differentiation in several tissues in both vertebrates and invertebrates. Within the hindbrain, manipulation of Hox gene expression alters the regional pattern of neuronal differentiation. Ectopic expression of hoxb1, for example, leads to an apparent trans-differentiation of affected motoneurons to the facial motoneuron phenotype, consistent with a link between the normal r4-related expression of hoxb1 and the r4-related origin of facial motoneurons [4,22]. Given such effects, one might expect Hox
genes to be generally important in specifying hindbrain neurons, and therefore for their expression to be strongly related to many neuron domains. Comparison of Figs. 9 and 10 to Figs. 4 –7 is rather disappointing in this regard. Certain neuron groups do bear a relatively strong relationship to specific Hox genes. For example, hoxb1 is expressed in r4 in all species and is therefore related to the facial motoneurons (especially frog and mouse), the more rostral ipsilaterally projecting vestibulospinal neuron group (zebrafish, frog, chicken, mouse, rat), a dearth of ipsilaterally projecting vestibulo-ocular neurons (frog and chicken), and a specific lateral subgroup of contralaterally projecting reticulospinal neurons (mouse). Hoxb3 is strongly expressed in r5 in mouse and r5-6 in zebrafish and is therefore related to the abducens motoneurons in each species. The strong expression in r5 in mouse is also related to the contralaterally projecting vestibulospinal group and a dearth of contralaterally projecting reticulospinal neurons (chicken and mouse). Yet for each striking parallel, there is an equally striking discrepancy. In the chicken and lamprey (and to a lesser extent the mouse), the origin of facial motoneurons is not accurately related to
690
GLOVER
FIG. 8. Genomic organization of Hox gene clusters in mammals. Adapted from [25]. Relative chromosome distances are not intended to be correct.
the r4 hoxb1 domain. In contrast to the other species, in the zebrafish hoxb1 is not related to a dearth of ipsilaterally projecting vestibuloocular neurons, nor is hoxb3 related to a dearth of contralaterally projecting reticulospinal neurons. Similarly, hoxb3 is more related to the caudal contralaterally projecting vestibulo-ocular group than to the contralaterally projecting vestibulospinal group in zebrafish, whereas the opposite seems to be the case in other species. Most of the groups shown in Figs. 4 –7 in fact do not relate well to any of these 4 Hox genes individually. This brings up the idea of combinatorics. Just as combinations of transcription factors are required to specify individual neuron types in the transverse plane (see Fig. 2), perhaps it is the combination of two or more Hox genes which specifies the longitudinal domains of hindbrain neuron groups. The rostral boundary of a group could be delineated by the rostral boundary of expression of one Hox gene, the caudal boundary by a more caudally expressed Hox gene. But here again there are species differences that are not explained by the existing data. Take for example the ipsilaterally projecting vestibulo-ocular
group in zebrafish, which lies in r2-4 (Fig. 7, top left). Its rostral limit could be specified by the rostral limit of expression of hoxa2, and its caudal limit likewise by hoxb3. The same group in chicken has a different caudal limit, namely at r3/4, yet chicken hoxb3 also has a rostral limit of expression at r4/5 [34]. Another important consideration regarding non-congruence is that some neuron groups may comprise subgroups that occupy more restricted domains. These might relate more closely to rhombomere domains, and thereby increase the coincidence of neuron group domains and Hox gene expression domains (see [7]). More detailed and systematic analyses of neuron groups, bringing in additional phenotypic characters, is needed to resolve this issue. HERE BE DRAGONS! THE ROAD AHEAD CANNOT ESCAPE THE ISSUE OF DYNAMIC COMPLEXITY As noted previously, some neuron groups behave dynamically by migrating from sites of origin to final destinations. Similarly,
FIG. 9. Expression domains of hoxb1 and hoxa2 as they relate to the rhombomeres in zebrafish, chicken, and mouse. The right-most figure shows the expression pattern elements that are common to all three species.
HINDBRAIN NEURONS AND GENE EXPRESSION
691
FIG. 10. Expression domains of hoxb3 and hoxd3 as they relate to the rhombomeres in zebrafish and mouse. The right-most figure shows the expression pattern elements that are common to both.
Hox gene expression patterns are in many cases highly dynamic. Attempts at correlating gene expression to neuron groups are therefore subject to several important caveats. First of all, many published reports of Hox gene expression only provide information from a limited time window. Hoxb1, for example, is initially expressed in the chicken embryo from the presumptive r3/4 boundary continuously to the spinal cord, and is then gradually down-regulated in r5 and r6 and up-regulated in r4 [41]. Weak domains of Hox gene expression may thus represent areas in which down-regulation or upregulation is in progress. Indeed, at least some Hox genes are expressed in more restricted domains at later stages, evidently in a different functional context, related to neuron groups at specific dorsoventral locations rather than longitudinal domains [6]. Secondly, most documentation of Hox gene expression is based on in situ hybridization, which maps only the location of transcripts as opposed to protein. Post-transcriptional regulation could modify the expression domains seen. Thirdly, most studies of Hox gene expression have focused on the rhombomeres as landmarks, and there has been a strong tendency to tie expression domains to these. Slight overlap into adjacent rhombomeres may be interpreted as “biological noise,” yet may in fact correlate well with the domains of neuron groups that do not fall exactly into register with rhombomere boundaries (see Fig. 4 –7). Given that both neuron groups and gene expression are dynamic, it is perhaps not surprising that superficial examination does not reveal immediate correlations. Figure 11 illustrates two quite different ways by which dynamic phenomena could reduce the congruence of gene expression and neuronal differentiation. One is based on neuronal migration, the other on transient gene expression. The take-home message is that correlative studies must be detailed and systematic if they are to contribute the information necessary to understand how genes are linked to the formation of identifiable neuron groups. The link is undoubtedly there, but a great deal of analysis remains in the years ahead. The use of cell-type specific transcription factor expression as a tool for following the development of post-mitotic neurons at early as well as late stages of their differentiation is an important methodological advance (see [30]).
HOW DOES PHYLOGENETIC VARIATION IN HOX GENE EXPRESSION ARISE? The absence of reams of examples of closely related neuron and Hox gene expression domains notwithstanding, the dramatic transforming effects of Hox gene manipulation in the hindbrain [4,22] provide a strong argument that Hox genes are in fact influential players when it comes to the specification of neuronal phenotype. Obviously other genes may be involved, and several other gene classes exhibit rhombomere-related expression patterns. But if we accept for the time being that Hox genes play an important role, then it is instructive to examine briefly how evolutionary differences in Hox gene expression might arise. The regulation of Hox gene expression is quite complex, and therefore provides rich opportunities for evolutionary diversification. Most if not all Hox genes are controlled by multiple regulatory elements that make different contributions to the overall expression pattern. As described earlier, Hox genes are expressed rostrocaudally in the same sequence they maintain physically within the chromosome. This sequential readout is related to their transcriptional regulation by retinoic acid: the higher the paralogue number, the higher is the retinoid concentration required to turn on the gene [38]. This retinoid-dependent transcriptional regulation is mediated by retinoid receptors which bind to specific retinoid receptor response elements in Hox gene regulatory regions [40]. Thus, one way in which Hox gene expression could change during evolution is by modifying the spatiotemporal pattern of retinoid signaling. This could be accomplished by changing the pattern in which retinoids impinge on the hindbrain or by changing the spatiotemporal pattern of retinoid receptor expression. Species differences certainly exist for the latter (see [20]), and much research is currently focusing on the availability of endogenous retinoids in the hindbrain region in different species. Other regulatory elements control other aspects of Hox gene expression including the restriction to specific rhombomeres [10, 26,37]. This is a complex phenomenon that is beyond the scope of this review, but which represents an exciting field of research that will undoubtedly have profound impact on our understanding of hindbrain development and evolution. We can point out that for some Hox genes different elements of the expression pattern are
692
GLOVER
FIG. 11. Two alternative mechanisms that generate non-congruence between expression of a developmental regulatory gene (gene X) and a group of neurons which it determines. In the first mechanism (top), a population of naive cells spans 3 rhombomeres. Gene X is expressed in the middle rhombomere and from inception of its expression exercises the inductive effect that determines the cells to acquire a particular phenotype. Subsequently, some of the committed neurons migrate into the next caudal rhombomere, leaving the gene X expression domain. In the second mechanism (bottom), gene X is initially expressed in the two caudal rhombomeres, but has not yet exerted its effect. This occurs as the expression domain gradually recedes to the middle rhombomere. The cells in the caudal rhombomere that have been determined by gene X to acquire a particular phenotype are left behind as the gene X expression domain recedes.
regulated by independent regulatory regions, such that the overall pattern arises through the concerted activation of multiple control elements. Thus, it is not hard to imagine that mutations in regulatory regions could alter the interactions among such multiple control pathways and thereby change the spatial and temporal details of expression. Indeed, it is more likely that subtle effects on Hox gene expression exerted through modification of regulatory regions would survive in an evolutionary context than would mutations in the coding sequence [5]. The latter would more easily lead to dramatic effects like homeotic transformations of entire rhombomeres; more likely than not the resultant disruption of vital neural circuitry in the hindbrain would be deleterious to the function of the organism. CONCLUDING REMARKS We are still far from understanding how gene expression regulates the construction of neural networks in the brain. Both descriptive and experimental evidence, however, has implicated several gene families in regionalization of the hindbrain neural tube. Among these are the Hox genes, which are clearly involved in demarcating regions along the longitudinal axis. Regionalization must somehow be related to the establishment of the rostrocaudal and mediolateral domains of specific neuron populations, and it is likely that conservation and divergence in the regionalization process underlay phylogenetic constancy and diversity in the differentiation of hindbrain neuron populations and the specification of their connections. Careful and systematic comparison of gene expression patterns and neuronal domains in different vertebrate species should provide clues about this relationship, and can help
guide experimental manipulations designed to test it. Comparison is complicated by the dynamic nature of Hox gene expression and by migratory movements of neurons. The challenge now is to obtain better and more comprehensive data to obtain a clearer picture of how gene expression relates to neuron groups from the earliest stages of their development.
REFERENCES 1. Auclair, F.; Valde´ s, N.; Marchand, R. Rhombomere-specific origin of branchial and visceral motoneurons of the facial nerve of the rat embryo. J. Comp. Neurol. 369:451– 461; 1996. 2. Auclair, F.; Marchand, R.; Glover, J. C. Regional patterning of reticulospinal and vestibulospinal neurons in the hindbrain of rat and mouse embryos. J. Comp. Neurol. 411:288 –300; 1999. 3. Behrend, K.; Donicht, M. Descending connections from the brainstem to the spinal cord in the electric fish Eigenmannia. Quantitative description based on retrograde horseradish and fluorescent-dye transport. Brain Behav. Evol. 35:227–239; 1990. 4. Bell, E.; Wingate, R. J.; Lumsden, A. Homeotic transformation of rhombomere identity after localized Hoxb1 misexpression. Science 284:2168 –2171; 1999. 5. Carroll, S. Endless forms: The evolution of gene regulation and morphological diversity. Cell 101:577–580; 2000. 6. Davenne, M.; Maconochie, M. K.; Neun, R.; Pattyn, A.; Chambon, P.; Krumlauf, R.; Rijli, F. M. Hoxa2 and Hoxb2 control dorsoventral patterns of neuronal development in the rostral hindbrain. Neuron 22:677– 691; 1999. 7. Dı´az, C.; Puelles, L.; Marı´n, F.; Glover, J. C. The relationship between rhombomeres and vestibular neuron populations as assessed in quailchicken chimeras. Dev. Biol. 202:14 –28; 1998.
HINDBRAIN NEURONS AND GENE EXPRESSION 8. Doe, C. Q.; Technau, G. M. Identification and cell lineage of individual neural precursors in the Drosophila CNS. Trends Neurosci. 16: 510 –514; 1993. 9. Easter, S. S.; Hitchcock, P. F. Stem cells and regeneration in the retina: What fish have taught us about neurogenesis. Neuroscientist 6:454 – 464; 2000. 10. Ferretti, E.; Marshall, H.; Po¨ pperl, H.; Maconochie, M.; Krumlauf, R.; Blasi, F. Segmental expression of Hoxb2 in r4 requires two separate sites that integrate cooperative interactions between Prep1, Pbx and Hox proteins. Development 127:155–166; 2000. 11. Fritzsch, B. Of mice and genes: Evolution of vertebrate brain development. Brain Behav. Evol. 52:207–217; 1998. 12. Fritzsch, B.; Sonntag, R.; Dubuc, R.; Ohta, Y.; Grillner, S. Organization of the six motor nuclei innervating the ocular muscles in lamprey. J. Comp. Neurol. 294:491–506; 1990. 13. Glover, J. C. The organization of vestibulo-ocular and vestibulospinal projections in the chicken embryo. Eur. J. Morphol. 32:193–200; 1994. 14. Glover, J. C. Retrograde and anterograde axonal tracing with fluorescent dextrans in the embryonic nervous system. Neurosci. Prot. 30:1– 13; 1995. 15. Glover, J. C. Neuroepithelial “compartments” and the specification of vestibular projections. Prog. Brain Res. 124:3–21; 2000. 16. Glover, J. C.; Petursdottir, G.; Jansen, J. K. S. Fluorescent dextranamines used as axonal tracers in the nervous system of the chicken embryo. J. Neurosci. Methods 18:243–254; 1986. 17. Glover, J. C.; Petursdottir, G. Regional specificity of developing reticulospinal, vestibulospinal and vestibulo-ocular projections in the chicken embryo. J. Neurobiol. 22:353–376; 1991. 18. Grapin-Botton, A.; Bonnin, M. A.; McNaughton, L. A.; Krumlauf, R.; Le Douarin, N. M. Plasticity of transposed rhombomeres: Hox gene induction is correlated with phenotypic modifications. Development 121:2707–2721; 1995. 19. Honig, M. G.; Hume, R. I. Carbocyanine dyes. Novel markers for labeling neurons. Trends Neurosci. 12:336 –338; 1989. 20. Hoover, F.; Kielland, A.; Glover, J. C. RXRg gene is expressed by discrete cell columns within the alar plate of the brainstem of the chicken embryo. J. Comp. Neurol. 416:417– 428; 2000. 21. Jessell, T. M. Neuronal specification in the spinal cord: Inductive signals and transcriptional codes. Nat. Rev. Genet. 1:20 –29; 2000. 22. Jungbluth, S.; Bell, E.; Lumsden, A. Specification of distinct motor neuron identities by the singular activities of individual Hox genes. Development 126:2751–2758; 1999. 23. Kessel, M. Reversal of axonal pathways from rhombomere 3 correlates with extra Hox expression domains. Neuron 10:379 –393; 1993. 24. Kimmel, C. B.; Powell, S. L.; Metcalfe, W. K. Brain neurons which project to the spinal cord in young larvae of the zebrafish. J. Comp. Neurol. 205:112–127; 1982. 25. Krumlauf, R. Hox genes and pattern formation in the branchial region of the vertebrate head. TIGS 9:106 –112; 1993. 26. Kwan, C. T.; Tsang, S. L.; Krumlauf, R.; Sham, M. H. Regulatory analysis of the mouse hoxb3 gene: Multiple elements work in concert to direct temporal and spatial patterns of expression. Dev. Biol. 232: 176 –190; 2001. 27. Lee, R. K.; Eaton, R. C. Identifiable reticulospinal neurons of the adult zebrafish, Brachydanio rerio. J. Comp. Neurol. 304:34 –52; 1991. 28. Lee, R. K.; Eaton, R. C.; Zottoli, S. J. Segmental arrangement of
693
29. 30. 31. 32. 33. 34. 35. 36. 37.
38.
39. 40. 41. 42. 43. 44.
45. 46.
reticulospinal neurons in the goldfish hindbrain. J. Comp. Neurol. 329:539 –556; 1993. Metcalfe, W. K.; Mendelson, B.; Kimmel, C. B. Segmental homologies among reticulospinal neurons in the hindbrain of the zebrafish larva. J. Comp. Neurol. 251:147–159; 1986. Moran-Rivard, L.; Kagawa, T.; Saueressig, H.; Gross, M. K.; Burrill, J.; Goulding, M. Evx1 is a postmitotic determinant of V0 interneuron identity in the spinal cord. Neuron 29:385–399; 2001. Oka, Y.; Satoui, M.; Ueda, K. Descending pathways to the spinal cord in the hime´ salmon (landlocked red salmon, Oncorhynchus nerka). J. Comp. Neurol. 2554:91–103; 1986. Petursdottir, G. Vestibulo-ocular projections in the 11-day chicken embryo: Pathway specificity. J. Comp. Neurol. 297:283–297; 1990. Prasado Rao, P.; Jadhao, A.; Sharma, S. Descending projection neurons to the spinal cord of the goldfish, Carassius auratus. J. Comp. Neurol. 265:96 –108; 1987. Rex, M.; Scotting, P. J. Chick HoxB3: Deduced amino-acid sequence and embryonic gene expression. Gene 149:381–382; 1994. Saueressig, H.; Burrill, J.; Goulding, M. Engrailed-1 and Netrin-1 regulate axon pathfinding by association interneurons that project to motor neurons. Development 126:4201– 4212; 1999. Schmid, A.; Chiba, A.; Doe, C. Q. Clonal analysis of Drosophila embryonic neuroblasts: Neural cell types, axon projections and muscle targets. Development 126:4653– 4689; 1999. Sham, M. H.; Vesque, C.; Nonchev, S.; Marshall, H.; Frain, M.; Gupta, R. D.; Whiting, J.; Wilkinson, D.; Charnay, P.; Krumlauf, R. The zinc finger gene Krox20 regulates HoxB2 (Hox2.8) during hindbrain segmentation. Cell 72:183–196; 1993. Simeone, A.; Acampora, D.; Arcioni, L.; Andrews, P. W.; Boncinelli, E.; Mavilio, F. Sequential activation of Hox 2 homeobox genes by retinoic acid in human embryonal carcinoma cells. Nature 346:763– 766; 1990. Straka, H.; Baker, R.; Gilland, E. Rhombomeric organization of vestibular pathways in larval frogs. J. Comp. Neurol. 437:42–55; 2001. Studer, M.; Po¨ pperl, H.; Marshall, H.; Kuroiwa, A.; Krumlauf, R. Role of a conserved retinoic acid response element in rhombomere restriction of Hoxb-1. Science 265:1728 –1732; 1994. Sundin, O.; Eichele, G. An early marker of axial pattern in the chick embryo and its respecification by retinoic acid. Development 114:841– 852; 1992. Suwa, H.; Gilland, E.; Baker, R. Segmental organization of vestibular and reticular projections to spinal and oculomotor nuclei in the zebrafish and goldfish. Biol. Bull. 191:257–259; 1996. Swain, G. P.; Snedeker, J. A.; Ayers, J.; Selzer, M. E. Cytoarchitecture of spinal-projecting neurons in the brain of the larval sea lamprey. J. Comp. Neurol. 336:194 –210; 1993. Tsuchida, T.; Ensini, M.; Morton, S. B.; Baldassare, T.; Edlund, T.; Jessell, T. M.; Pfaff, S. L. Topographic organization of embryonic motor neurons defined by expression of LIM homeobox genes. Cell 79:957–970; 1994. Vaage, S. The segmentation of the primitive neural tube in chick embryos (Gallus domesticus). Adv. Anat. Embryol. Cell Biol. 41:1– 88; 1969. Varela-Echavarria, A.; Pfaff, S. L.; Guthrie, S. Differential expression of LIM homeobox genes among motor neuron subpopulations in the developing chick brain stem. Mol. Cell. Neurosci. 8:242–257; 1996.
PII S0361-9230(01)00562-7
Evolution of the Nervous System
Correlated patterns of neuron differentiation and Hox gene expression in the hindbrain: A comparative analysis Joel C. Glover* Department of Physiology, University of Oslo, Oslo, Norway ABSTRACT: Hindbrain neurons are organized into coherent subpopulations with characteristic projection patterns and functions. Many of these serve vital functions that have been conserved throughout the vertebrate radiation, but diversification to modified or highly specialized functions has also occurred. The differentiation of identifiable neuron groups in specific spatial domains must involve the regional expression of determinants within the hindbrain neuroepithelium. The Hox genes are involved in longitudinal regionalization of the neural tube, and their expression patterns in the hindbrain are closely related to the rhombomeres which partition the hindbrain into morphogenetic units. Hox gene expression also exhibits conserved patterning as well as phylogenetic variation. One plausible mechanism that may have contributed to evolutionary diversification in hindbrain neuron populations is therefore the emergence of species-specific differences in Hox gene expression. This article presents a comparative overview of the regional patterning of selected Hox genes and hindbrain neuron populations in several embryologically important species. Although tantalizing correlations exist, the relationship between Hox genes and neuronal patterning is complex, and complicated by dynamic features in each. Much more comparative and developmental data must be obtained before the link between Hox gene expression and hindbrain neuron patterning can be elucidated satisfactorily in an evolutionary context. © 2001 Elsevier Science Inc.
neuromeres, called rhombomeres, which provide a morphological grid onto which gene expression and neuron group domains can be plotted (Fig. 1). Direct and indirect comparison of neuron groups with relevant genes as well as manipulations of gene expression can therefore be carried out with relatively high spatial precision. This article presents examples of phylogenetic variation in selected hindbrain neuron populations and important developmental regulatory genes, focusing on key species commonly used in neuroembryological studies. The link between the relevant genes and the specification of these neuronal phenotypes is still poorly understood. Nevertheless, enough evidence has accumulated in favor of such a link that an underlying hypothesis can be entertained here: namely that phylogenetic constancy and variation in the expression patterns of genes such as these contribute substantially to the observed conservation and divergence of hindbrain neuron populations. SPECIFIC NEURON TYPES EXPRESS SPECIFIC COMBINATIONS OF TRANSCRIPTION FACTORS That neuronal identity might be specified by the regional expression of particular classes of transcription factors first became clear in studies of insect neurogenesis, where genetic approaches to understanding neuronal differentiation were pioneered (see e.g, [8,36]). In insects, each of the neuroblasts (neuronal progenitors) within a segmental ganglion is individually identifiable by virtue of its position and pattern of gene expression, and gives rise to a specific subset of the neurons in the ganglion. Region-specific gene expression as an important determinant of neuronal identity is also a prominent feature in the central nervous system of vertebrates. Studies of spinal motoneuron and interneuron development have provided exceptionally illustrative examples [21]. Specification of these neurons occurs in a step-wise fashion. Graded signals from midline structures induce the differentiation of progenitor cells in localized serial dorsoventral zones. Reciprocal repressive interactions then sharpen these zones into highly restricted populations of progenitors characterized by the expression of specific transcription factors. These then generate polyclones of neuronal progeny which are also distinguished by
KEY WORDS: Segmentation, Zebrafish, Chicken, Mouse, Cranial motoneurons, Reticulospinal, Vestibulospinal, Vestibuloocular, Dextranamines.
INTRODUCTION How have specific neuron populations in the vertebrate brain evolved? To answer this question it is necessary to identify which developmental regulatory genes are responsible for specifying different neuron types, and to assess how the relationship between genes and phenotypes varies among species. The hindbrain is one region of the brain that is especially well suited for such an analysis. During the period when most of its constituent neuron groups are generated, the hindbrain is transiently segmented into
* Address for correspondence: Joel C. Glover, Department of Physiology, P.B. 1103 Blindern, 0317 Oslo, Norway. E-mail: [email protected]
683
684
GLOVER
FIG. 1. Side view of a chicken embryo at Hamburger and Hamilton stage 14 (HH14). The brain is outlined (dotted line), showing rhombomeres 1 through 7 (rh1–7) and their relationship to cranial nerves (n). Other structures indicated in the hindbrain region are the otic vesicle (Otv) and the notochord (hatched). Note also the indication of neuromeres in midbrain and forebrain (see Puelles, this issue). Drawing from [45].
FIG. 2. Schematic overview of transcription factor gene expression by different motoneuron (MN) and ventral interneuron (IN) classes in a transverse section of the spinal cord (left). Ventral IN classes (V0 –V3) have been named according to differential transcription factor expression and their relative positions as they emerge from the proliferative ventricular zone (vz). These IN classes and various MN classes are distinguished moreover by their axon trajectories or peripheral targets Abbreviations: aCIN, ascending commissural IN; aIIN, ascending ipsilateral IN; adIIN, ascending and descending ipsilateral IN; symp, sympathetic ganglia; v. axial and d. axial, ventral and dorsal axial muscles, respectively; d. limb and v. limb, dorsal and ventral limb muscles, respectively. Specific projection phenotypes are thus related to specific patterns of expression of the transcription factors listed along the top of the table. Data from [30,35,44].
HINDBRAIN NEURONS AND GENE EXPRESSION
685
FIG. 3. Number of reticulospinal neurons projecting to one side of the spinal cord (half of the total number due to bilateral symmetry), in the zebrafish larva [29], the ranid frog larva [39], the 7-day chicken embryo [17], and the E15 rat embryo [2].
specific transcription factors that are believed to steer the readout of cell-type specific proteins involved in selective migration, axon outgrowth, target cell recognition, and molecular phenotype. A number of transcription factors have been identified whose combinatorial expression patterns define specific functional classes of spinal neurons (Fig. 2), including motoneurons differentiated according to peripheral axon projection pathway and stretch reflex connectivity pattern, and interneurons differentiated according to central axon projection pathways [21]. Many of the same transcription factors are expressed in the hindbrain, where similar relationships to neuron type exist [46]. Comprehensive mapping of the expression of these and similar genes will undoubtedly unveil much of the underlying genetic blueprint that specifies the diverse neuron groups in the hindbrain. This endeavour has hardly begun, however, making any attempt at a comparative analysis of the expression of these genes and neuronal phenotype grossly premature. On the other hand, transcription factors that imbue the different rhombomeres with unique identities have been studied more extensively in a comparative context. One class of these regionally specific genes, the Hox genes, will be featured in this article. Experimental manipulation of the expression of Hox genes leads to changes in the way neurons differentiate, indicating that they play an instructive role in neuron type determination [4,18,22,23]. A comparative review of the expression patterns of selected Hox genes will be presented in a later section following a phylogenetic overview of hindbrain neuron patterning. HINDBRAIN NEURON POPULATIONS HAVE EXPANDED DURING PHYLOGENESIS In general, the size of hindbrain neuron populations has increased as the vertebrate lineage evolved and branched. Documentation has only been obtained for a few neuron populations in a few species, and is furthermore complicated by counts being performed in different species at different stages during the life cycle (embryo, larva, adult), and by identification of neuron populations by different labeling paradigms. Nevertheless, it is quite clear that amniotes (reptiles, birds, mammals) have more highly populated
hindbrains than anamniotes. Among amniotes, mammals evidently have more neurons than birds and reptiles. For example, a graph of reticulospinal neuron number against vertebrate class shows a virtually logarithmic increase in the number of combined pontine and medullary reticulospinal neurons on comparison of fish to frog to chicken to rat at developmental stages of similar hindbrain maturity (Fig. 3). Similar numerical differences exist for vestibular projection neurons (vestibulospinal and vestibulo-ocular neurons, for example), although quantitation has not yet been performed to an extent that can be summarized graphically [7,17,24,32,39,42]. Regarding the data for reticulospinal neurons in Fig. 3, it should be pointed out that with increasing development in each species, the number of reticulospinal neurons is likely (or in some cases is known) to increase. This increase deserves special comment. In many fish, and perhaps other anamniotes, certain neuron populations increase in number well into adult life [9]. This protracted proliferation evidently is a special feature not seen in amniotes. It is presumably an adaptation to a continual increase in body size that occurs during the lifetime of many anamniotes, with commensurate increases in the sizes of muscles and of sensory organs. An adult fish has substantially more reticulospinal neurons, for example, than does a larval fish [27,28,33,42]. The phylogenetic variation in the population size of specific hindbrain neuron groups, on the other hand, is not related to body size. A diminutive rat embryo has far more reticulospinal neurons than a beefy adult catfish. Rather, the variation is likely to be related to evolutionary increases in the complexity of motor system regulation (especially with increasing cortical control) and to elaboration of the autonomic nervous system and its supraspinal regulation. Such factors should be borne in mind when making species comparisons. For example, excellent quantitative data for reticulospinal neurons in the lamprey [43] have not been included in Fig. 3 because it is difficult to relate the developmental stage of the lamprey specimens to the species shown. Swain et al. [43] studied larval lamprey that were 5 years old (10 –15 cm in length),
686 and counted on the order of 700 reticulospinal neurons per set (half of the bilaterally paired total). This elevates the lamprey past fish, frog, and chicken in Fig. 3, and is therefore most likely a reflection of maturity rather than evolutionary status or behavioral repertoire: simply by dint of living longer, a 5-year-old lamprey has had greater opportunity to add reticulospinal neurons to its hindbrain through ongoing proliferation. SPECIALIZED HINDBRAIN NEURON POPULATIONS EXIST IN SOME SPECIES Another important point to bear in mind is that certain species may possess special neuron populations related to special functions. This is most obvious with respect to specialized sensory systems, which are typically associated with characteristic central nuclei, for example the octavolateral nuclei in aquatic animals with a lateral line, or the pacemaker nucleus of electric fish [3]. Specialized neuron groups may be unique modifications of “generic” populations present in all species, or may be “extra” groups added through the proliferation, division, and elaboration of pre-existing populations. Thus, although elaboration seems to be a hallmark of the general evolutionary increase in the size and complexity of many hindbrain neuron populations (see below), it may also generate isolated phylogenetic features commensurate with parallel specializations elsewhere in the body. Along the same lines, some “generic” nuclei are completely missing in certain species. For example, lampreys, salmon, and electric fish lack raphe nuclei [3,31,43] and lampreys lack a specific oculomotor motoneuron pool [12]. Superficial comparisons may therefore be misleading— neuron groups must be compared with great attention to phenotypic detail and embryonic history (lineage, migratory patterns) to avoid false conclusions about evolutionary relationships. SOME, BUT NOT ALL HINDBRAIN NEURON POPULATIONS BEAR A CONSERVED RELATIONSHIP TO SPECIFIC RHOMBOMERIC DOMAINS Despite the tremendous expansion of hindbrain neuron populations during vertebrate evolution, in all vertebrate classes specific neuron populations occupy coherent spatial domains that can be related, directly or indirectly, to the overt morphological architecture of the hindbrain as delineated by the rhombomeres. A point of difficulty in assessing this regional relationship is the transient nature of the rhombomere boundaries, which itself is speciesspecific. In some species like the chicken and the frog, rhombomere boundaries are visible for an extended period during which most hindbrain neuron groups have been generated and can be mapped. In other species like the mouse, the boundaries are erased before it is possible to map many groups, and a regional analysis must depend on fate-mapping techniques or indirect cues such as lasting rhombomere-related landmarks or gene expression patterns. This is also important to remember when reviewing the existing literature, which is not without its fallacies. The rhombomere-related domain of a given neuron group is in some cases very similar across species, whereas in other cases it differs considerably. Some aspects of hindbrain regionalization are therefore highly conserved, whereas others have diverged. Cranial Motoneurons and Efferents to Sensory Organs These are the hindbrain neuron populations that have been most studied, because they are relatively easy to map unambiguously. Retrograde tracers such as DiI [19] or dextran-amines [14,16] can be applied to specific cranial nerve roots, thereby labeling a defined population of efferent neurons. Greater care must be taken when assessing the spatial relationships of subpopulations within a
GLOVER cranial motor nucleus, however. In this case, tracer applications must be targeted to specific nerve branches or peripheral targets (see, e.g., [1]). A number of studies have mapped cranial motoneuron populations in embryos and larvae of a variety of species. These studies have been reviewed recently [11], and are summarized in Fig. 4. The first thing to note is that in general the relative locations of the different cranial efferent populations are similar in different species, and can be related roughly to the same rhombomeres. The second thing to note is that despite this similarity, there are obvious differences, some striking, some subtle. In all species examined, for example, trigeminal motoneurons are found in r2 and r3. Yet the domain occupied may extend into adjacent r1 (chicken, mouse) or r4 (lamprey, mouse), or may not span the full length of r3 (elasmobranch, chicken). The trigeminal motoneurons may even split into two widely separate domains (elasmobranch, r2-3 and r6). Similarly, in nearly all species examined, abducens motoneurons are found in r5, but they may also be found in r6 (lamprey, teleost, reptile, chicken), extend into r4 (lamprey), or be localized to r6 instead of r5 (elasmobranch). The differences are even more complex for the facial motor nucleus and its subdivisions. This brings up a crucial issue when considering the relationships between neuronal domains and gene expression: it is not uncommon for neurons generated in one position to migrate en masse to another position. Migration of motoneurons has been well documented for several hindbrain motoneuron populations, including the oculomotor, trigeminal, abducens, and facial motoneurons. The migratory patterns themselves may exhibit species-related conservation and divergence (see [11]). The upshot is that using neuron position alone for assessing evolutionary relationships is risky— one must assess where a neuron population originates as well, because the expression of genes in the region of origin may be the principal determinant. Assessing origin is not always easy. It may require retrograde labeling at the very earliest stages of axon outgrowth (not equally feasible for all organisms). If motoneurons have already migrated prior to extending axons, it may additionally require fate-mapping techniques. Because of these extra difficulties, the rhombomeric patterning described in many species is still deficient in this respect. Nevertheless, several studies have managed to pinpoint the origins of some of the cranial motor nuclei in a few species. The facial nucleus is perhaps the best-studied case. Figure 5 shows where facial branchial motoneurons are believed to originate in four different species. Again, there are clear differences. Facial branchial motoneurons originate predominantly from r4 in frog, chicken, and mouse, but there are minor contributions from neighboring rhombomeres in chicken and mouse. In lamprey the predominant origin is r5, with a smaller contribution from r4. It would appear that both the genesis and final location of cranial motoneurons is linked to different domains of the hindbrain in different species. Reticulospinal Neurons After the cranial motoneurons, reticulospinal neurons are probably the most thoroughly investigated hindbrain neuron population. These make up a composite population comprising many functionally distinct neuron groups, which are often lumped together because they are easy to label simultaneously (tracers are applied to the hindbrain/spinal cord junction). A number of studies have mapped the positions of reticulospinal neurons with specific axon trajectories (unilateral tracer applications discriminate between ipsilaterally and contralaterally projecting neurons) as they relate to the rhombomeric divisions. In all species examined, each rhombomere contains reticulospinal neurons, and many of these share certain characteristics such as mediolateral location and
HINDBRAIN NEURONS AND GENE EXPRESSION
687
FIG. 4. Comparison of the rhombomeric domains of selected cranial motor nuclei (trigeminal, abducens, facial) in different species. Each image depicts schematically a frontal view of the hindbrain with rhombomeres 2– 6 indicated on each side. The two sides are separated by the midline floor plate. The trigeminal and facial nuclei are shown on the left side, and the abducens nucleus is shown on the right side. Mediolateral locations are approximate. The subdivisions of the facial nerve have not been individually characterized in all species. In such cases the domain of the facial nerve nucleus is divided diagonally to show which subpopulations are present, but not their longitudinal limits. Adapted from [11].
dendritic morphology. A common misconception in the literature, however, is that the reticulospinal neurons represent a segmentally iterated population (implying that each rhombomere contains the same set of reticulospinal phenotypes). The actual picture is one of segmental variation, which has been suggested to arise through diversification of ancestral segmentally iterated homologues [29]. Careful observation (and reading) demonstrates this fact even for the zebrafish (see [29]), which has been erroneously promulgated as an archetype of segmental iteration. That segmental variation is
FIG. 5. Domains of origin of facial branchial motoneurons in four different species.
a key developmental and evolutionary feature is particularly evident when comparing anamniotes to amniotes. In the latter, the reticulospinal population has become more elaborate (in parallel with the dramatic increase in neuron number—Fig. 3), with the evolutionary addition of extra groups and migratory patterns in specific subsets of rhombomeres [2,17]. Figure 6 illustrates the domains of ipsilaterally and contralaterally projecting reticulospinal neurons in zebrafish, frog, chicken, and rodent (rat and mouse). As a clear illustration of segmental as well as phylogenetic variation, note that in each species there is a specific rhombomeric domain that lacks contralaterally projecting reticulospinal neurons, but that this domain differs among the species (r3 in zebrafish, r6 in frog, r5 in chicken, mouse, and rat). Note also that spatial elaboration of the reticulospinal population is especially evident in the mouse and rat, where major clusters spawn subclusters that become displaced differentially to more lateral positions, thereby creating satellite groups corresponding to specific functionally identifiable reticular nuclei [2]. As is the case for the cranial motoneurons, the reticulospinal neurons occupy different rhombomeric domains in different species. Moreover, the domain boundaries of individual clusters do
688
GLOVER
FIG. 6. Domains of reticulospinal neurons in four different species. Contralaterally projecting and ipsilaterally projecting reticulospinal neurons are shown on the left and right side, respectively, of each hindbrain (the arrow indicates the side on which axons descend to the spinal cord). Note that the number of reticulospinal neurons varies greatly (see Fig. 3), such that the neuron density per domain increases substantially from zebrafish up to mouse.
not always relate precisely to rhombomere boundaries, but may extend variable distances into one rhombomere from another. Whether this non-congruence results from longitudinal migration or spatial determination out of register with the rhombomeres is not yet known. Vestibular Projection Neurons Neurons in the vestibular nuclei and associated nuclei project to several major targets, including the spinal cord, the cerebellum, the motor nuclei that steer the extraocular muscles (oculomotor, trochlear, abducens), and the contralateral vestibular nuclei. Of these subpopulations, the vestibulospinal and vestibulo-ocular have been characterized best in the context of spatial patterning within the hindbrain neuroepithelium. Studies in the chicken embryo have shown that vestibulospinal and vestibulo-ocular neurons are organized in clusters, each of which projects axons along a specific trajectory to innervate a specific target or targets (reviewed in [15]). This feature has been termed the “hodological mosaic,” because neuron groups projecting to different targets are to a great extent segregated [13]. Later studies in anamniotes have demonstrated that this patterning feature is conserved in the vertebrate lineage, but have also documented divergence in the pattern [39, 42]. Figure 7 shows the domains of vestibulospinal and vestibuloocular neurons as they relate to the rhombomeres in those species where this has been characterized. Again, certain elements of the pattern are similar in representatives of different vertebrate classes. Fish, frog, and chicken each have, within a domain roughly corresponding to r4, an ipsilaterally projecting vestibulospinal group with axons coursing in a lateral tract (the same is true in mouse and rat; [2]). They each have a contralaterally projecting vestibulospinal group roughly localized to r5, and 2 contralaterally projecting vestibulo-ocular groups in separate domains (one roughly in r2, and the other within r4 –r7). There are clear differences, however, in the longitudinal extent of several groups, and some groups are absent in zebrafish at the larval stages studied (namely, the caudal groups of ispilaterally projecting vestibulospinal and vestibuloocular neurons). There are also species differences in the degree to which vestibulospinal and vestibulo-ocular groups overlap along the longitudinal axis (see especially the ipsilaterally projecting groups).
As is the case for cranial motor nuclei and reticulospinal neurons, the relative roles of positional determination versus longitudinal migration in establishing these vestibular neuron domains is poorly characterized in all species. Non-congruence between rhombomeres and neuron domains in the vestibular population probably results from both of these mechanisms (see [7]). HOX GENES HAVE SIMILAR EXPRESSION PATTERNS IN DIFFERENT SPECIES, BUT DIFFERENCES EXIST Hox genes are evolutionary homologues of transcription factor genes that were originally identified in Drosophila, within its homeotic gene complex [25]. Hox genes have since been identified in all vertebrates. Their number increased in the vertebrate lineage as it diverged from the invertebrate, with duplications of the original homeotic complex eventually generating four different Hox gene complexes in mammals (A, B, C, and D), and further expansion within the complexes generating additional genes (Fig. 8). The peculiar feature of Hox gene expression is that the rostrocaudal and temporal order of expression within the body follow the order of the genes in the chromosome, a property termed colinearity. Hox gene nomenclature emphasizes this feature: within each complex, the genes are numbered in order of rostral to caudal expression domain, maintaining rank alignments according to sequence similarity (and thereby to presumed evolutionary relation). The rostrocaudal ordering has a few exceptions (e.g., hoxb2 is expressed rostral to hoxb1), but is surprisingly well conserved. Because the different complexes are the result of duplications, Hox genes with the same rank in different complexes are evolutionary relatives, and are therefore called paralogues. The first four paralogue groups have expression domains within the hindbrain. Figures 9 and 10 show the expression domains of four Hox genes from the first three paralogue groups (those expressed within the segmented region of the hindbrain). Dark and light domains represent areas of high and low expression, respectively. The rostral limit of expression is the most conserved feature for each gene (although this is not an absolute rule for Hox genes, especially when low expression domains are considered; see hoxd3). Certain high expression domains are also conserved. There are, nevertheless, clear species differences evident in these patterns.
HINDBRAIN NEURONS AND GENE EXPRESSION
689
FIG. 7. Domains of vestibulo-ocular (top) and vestibulospinal (bottom) neuron groups in three different species. Axon trajectories are indicated by thin lines. Ipsilaterally and contralaterally projecting groups are indicated on right and left sides, respectively, with large arrows showing the side of projection. Data from [15,39,42].
These include the caudal limits of expression and the specific domains of high versus low expression. The former feature is generally considered to be functionally irrelevant, because more caudally expressed paralogues are considered to dominate transcriptional regulation in more caudal regions. The second feature, especially as it relates to the rostral limits, may have more impact on the patterning of differentiating neurons. HOW RELATED ARE HOX GENE EXPRESSION AND HINDBRAIN NEURON GROUP DOMAINS? Hox genes are known to play important roles in specifying regional differentiation in several tissues in both vertebrates and invertebrates. Within the hindbrain, manipulation of Hox gene expression alters the regional pattern of neuronal differentiation. Ectopic expression of hoxb1, for example, leads to an apparent trans-differentiation of affected motoneurons to the facial motoneuron phenotype, consistent with a link between the normal r4-related expression of hoxb1 and the r4-related origin of facial motoneurons [4,22]. Given such effects, one might expect Hox
genes to be generally important in specifying hindbrain neurons, and therefore for their expression to be strongly related to many neuron domains. Comparison of Figs. 9 and 10 to Figs. 4 –7 is rather disappointing in this regard. Certain neuron groups do bear a relatively strong relationship to specific Hox genes. For example, hoxb1 is expressed in r4 in all species and is therefore related to the facial motoneurons (especially frog and mouse), the more rostral ipsilaterally projecting vestibulospinal neuron group (zebrafish, frog, chicken, mouse, rat), a dearth of ipsilaterally projecting vestibulo-ocular neurons (frog and chicken), and a specific lateral subgroup of contralaterally projecting reticulospinal neurons (mouse). Hoxb3 is strongly expressed in r5 in mouse and r5-6 in zebrafish and is therefore related to the abducens motoneurons in each species. The strong expression in r5 in mouse is also related to the contralaterally projecting vestibulospinal group and a dearth of contralaterally projecting reticulospinal neurons (chicken and mouse). Yet for each striking parallel, there is an equally striking discrepancy. In the chicken and lamprey (and to a lesser extent the mouse), the origin of facial motoneurons is not accurately related to
690
GLOVER
FIG. 8. Genomic organization of Hox gene clusters in mammals. Adapted from [25]. Relative chromosome distances are not intended to be correct.
the r4 hoxb1 domain. In contrast to the other species, in the zebrafish hoxb1 is not related to a dearth of ipsilaterally projecting vestibuloocular neurons, nor is hoxb3 related to a dearth of contralaterally projecting reticulospinal neurons. Similarly, hoxb3 is more related to the caudal contralaterally projecting vestibulo-ocular group than to the contralaterally projecting vestibulospinal group in zebrafish, whereas the opposite seems to be the case in other species. Most of the groups shown in Figs. 4 –7 in fact do not relate well to any of these 4 Hox genes individually. This brings up the idea of combinatorics. Just as combinations of transcription factors are required to specify individual neuron types in the transverse plane (see Fig. 2), perhaps it is the combination of two or more Hox genes which specifies the longitudinal domains of hindbrain neuron groups. The rostral boundary of a group could be delineated by the rostral boundary of expression of one Hox gene, the caudal boundary by a more caudally expressed Hox gene. But here again there are species differences that are not explained by the existing data. Take for example the ipsilaterally projecting vestibulo-ocular
group in zebrafish, which lies in r2-4 (Fig. 7, top left). Its rostral limit could be specified by the rostral limit of expression of hoxa2, and its caudal limit likewise by hoxb3. The same group in chicken has a different caudal limit, namely at r3/4, yet chicken hoxb3 also has a rostral limit of expression at r4/5 [34]. Another important consideration regarding non-congruence is that some neuron groups may comprise subgroups that occupy more restricted domains. These might relate more closely to rhombomere domains, and thereby increase the coincidence of neuron group domains and Hox gene expression domains (see [7]). More detailed and systematic analyses of neuron groups, bringing in additional phenotypic characters, is needed to resolve this issue. HERE BE DRAGONS! THE ROAD AHEAD CANNOT ESCAPE THE ISSUE OF DYNAMIC COMPLEXITY As noted previously, some neuron groups behave dynamically by migrating from sites of origin to final destinations. Similarly,
FIG. 9. Expression domains of hoxb1 and hoxa2 as they relate to the rhombomeres in zebrafish, chicken, and mouse. The right-most figure shows the expression pattern elements that are common to all three species.
HINDBRAIN NEURONS AND GENE EXPRESSION
691
FIG. 10. Expression domains of hoxb3 and hoxd3 as they relate to the rhombomeres in zebrafish and mouse. The right-most figure shows the expression pattern elements that are common to both.
Hox gene expression patterns are in many cases highly dynamic. Attempts at correlating gene expression to neuron groups are therefore subject to several important caveats. First of all, many published reports of Hox gene expression only provide information from a limited time window. Hoxb1, for example, is initially expressed in the chicken embryo from the presumptive r3/4 boundary continuously to the spinal cord, and is then gradually down-regulated in r5 and r6 and up-regulated in r4 [41]. Weak domains of Hox gene expression may thus represent areas in which down-regulation or upregulation is in progress. Indeed, at least some Hox genes are expressed in more restricted domains at later stages, evidently in a different functional context, related to neuron groups at specific dorsoventral locations rather than longitudinal domains [6]. Secondly, most documentation of Hox gene expression is based on in situ hybridization, which maps only the location of transcripts as opposed to protein. Post-transcriptional regulation could modify the expression domains seen. Thirdly, most studies of Hox gene expression have focused on the rhombomeres as landmarks, and there has been a strong tendency to tie expression domains to these. Slight overlap into adjacent rhombomeres may be interpreted as “biological noise,” yet may in fact correlate well with the domains of neuron groups that do not fall exactly into register with rhombomere boundaries (see Fig. 4 –7). Given that both neuron groups and gene expression are dynamic, it is perhaps not surprising that superficial examination does not reveal immediate correlations. Figure 11 illustrates two quite different ways by which dynamic phenomena could reduce the congruence of gene expression and neuronal differentiation. One is based on neuronal migration, the other on transient gene expression. The take-home message is that correlative studies must be detailed and systematic if they are to contribute the information necessary to understand how genes are linked to the formation of identifiable neuron groups. The link is undoubtedly there, but a great deal of analysis remains in the years ahead. The use of cell-type specific transcription factor expression as a tool for following the development of post-mitotic neurons at early as well as late stages of their differentiation is an important methodological advance (see [30]).
HOW DOES PHYLOGENETIC VARIATION IN HOX GENE EXPRESSION ARISE? The absence of reams of examples of closely related neuron and Hox gene expression domains notwithstanding, the dramatic transforming effects of Hox gene manipulation in the hindbrain [4,22] provide a strong argument that Hox genes are in fact influential players when it comes to the specification of neuronal phenotype. Obviously other genes may be involved, and several other gene classes exhibit rhombomere-related expression patterns. But if we accept for the time being that Hox genes play an important role, then it is instructive to examine briefly how evolutionary differences in Hox gene expression might arise. The regulation of Hox gene expression is quite complex, and therefore provides rich opportunities for evolutionary diversification. Most if not all Hox genes are controlled by multiple regulatory elements that make different contributions to the overall expression pattern. As described earlier, Hox genes are expressed rostrocaudally in the same sequence they maintain physically within the chromosome. This sequential readout is related to their transcriptional regulation by retinoic acid: the higher the paralogue number, the higher is the retinoid concentration required to turn on the gene [38]. This retinoid-dependent transcriptional regulation is mediated by retinoid receptors which bind to specific retinoid receptor response elements in Hox gene regulatory regions [40]. Thus, one way in which Hox gene expression could change during evolution is by modifying the spatiotemporal pattern of retinoid signaling. This could be accomplished by changing the pattern in which retinoids impinge on the hindbrain or by changing the spatiotemporal pattern of retinoid receptor expression. Species differences certainly exist for the latter (see [20]), and much research is currently focusing on the availability of endogenous retinoids in the hindbrain region in different species. Other regulatory elements control other aspects of Hox gene expression including the restriction to specific rhombomeres [10, 26,37]. This is a complex phenomenon that is beyond the scope of this review, but which represents an exciting field of research that will undoubtedly have profound impact on our understanding of hindbrain development and evolution. We can point out that for some Hox genes different elements of the expression pattern are
692
GLOVER
FIG. 11. Two alternative mechanisms that generate non-congruence between expression of a developmental regulatory gene (gene X) and a group of neurons which it determines. In the first mechanism (top), a population of naive cells spans 3 rhombomeres. Gene X is expressed in the middle rhombomere and from inception of its expression exercises the inductive effect that determines the cells to acquire a particular phenotype. Subsequently, some of the committed neurons migrate into the next caudal rhombomere, leaving the gene X expression domain. In the second mechanism (bottom), gene X is initially expressed in the two caudal rhombomeres, but has not yet exerted its effect. This occurs as the expression domain gradually recedes to the middle rhombomere. The cells in the caudal rhombomere that have been determined by gene X to acquire a particular phenotype are left behind as the gene X expression domain recedes.
regulated by independent regulatory regions, such that the overall pattern arises through the concerted activation of multiple control elements. Thus, it is not hard to imagine that mutations in regulatory regions could alter the interactions among such multiple control pathways and thereby change the spatial and temporal details of expression. Indeed, it is more likely that subtle effects on Hox gene expression exerted through modification of regulatory regions would survive in an evolutionary context than would mutations in the coding sequence [5]. The latter would more easily lead to dramatic effects like homeotic transformations of entire rhombomeres; more likely than not the resultant disruption of vital neural circuitry in the hindbrain would be deleterious to the function of the organism. CONCLUDING REMARKS We are still far from understanding how gene expression regulates the construction of neural networks in the brain. Both descriptive and experimental evidence, however, has implicated several gene families in regionalization of the hindbrain neural tube. Among these are the Hox genes, which are clearly involved in demarcating regions along the longitudinal axis. Regionalization must somehow be related to the establishment of the rostrocaudal and mediolateral domains of specific neuron populations, and it is likely that conservation and divergence in the regionalization process underlay phylogenetic constancy and diversity in the differentiation of hindbrain neuron populations and the specification of their connections. Careful and systematic comparison of gene expression patterns and neuronal domains in different vertebrate species should provide clues about this relationship, and can help
guide experimental manipulations designed to test it. Comparison is complicated by the dynamic nature of Hox gene expression and by migratory movements of neurons. The challenge now is to obtain better and more comprehensive data to obtain a clearer picture of how gene expression relates to neuron groups from the earliest stages of their development.
REFERENCES 1. Auclair, F.; Valde´ s, N.; Marchand, R. Rhombomere-specific origin of branchial and visceral motoneurons of the facial nerve of the rat embryo. J. Comp. Neurol. 369:451– 461; 1996. 2. Auclair, F.; Marchand, R.; Glover, J. C. Regional patterning of reticulospinal and vestibulospinal neurons in the hindbrain of rat and mouse embryos. J. Comp. Neurol. 411:288 –300; 1999. 3. Behrend, K.; Donicht, M. Descending connections from the brainstem to the spinal cord in the electric fish Eigenmannia. Quantitative description based on retrograde horseradish and fluorescent-dye transport. Brain Behav. Evol. 35:227–239; 1990. 4. Bell, E.; Wingate, R. J.; Lumsden, A. Homeotic transformation of rhombomere identity after localized Hoxb1 misexpression. Science 284:2168 –2171; 1999. 5. Carroll, S. Endless forms: The evolution of gene regulation and morphological diversity. Cell 101:577–580; 2000. 6. Davenne, M.; Maconochie, M. K.; Neun, R.; Pattyn, A.; Chambon, P.; Krumlauf, R.; Rijli, F. M. Hoxa2 and Hoxb2 control dorsoventral patterns of neuronal development in the rostral hindbrain. Neuron 22:677– 691; 1999. 7. Dı´az, C.; Puelles, L.; Marı´n, F.; Glover, J. C. The relationship between rhombomeres and vestibular neuron populations as assessed in quailchicken chimeras. Dev. Biol. 202:14 –28; 1998.
HINDBRAIN NEURONS AND GENE EXPRESSION 8. Doe, C. Q.; Technau, G. M. Identification and cell lineage of individual neural precursors in the Drosophila CNS. Trends Neurosci. 16: 510 –514; 1993. 9. Easter, S. S.; Hitchcock, P. F. Stem cells and regeneration in the retina: What fish have taught us about neurogenesis. Neuroscientist 6:454 – 464; 2000. 10. Ferretti, E.; Marshall, H.; Po¨ pperl, H.; Maconochie, M.; Krumlauf, R.; Blasi, F. Segmental expression of Hoxb2 in r4 requires two separate sites that integrate cooperative interactions between Prep1, Pbx and Hox proteins. Development 127:155–166; 2000. 11. Fritzsch, B. Of mice and genes: Evolution of vertebrate brain development. Brain Behav. Evol. 52:207–217; 1998. 12. Fritzsch, B.; Sonntag, R.; Dubuc, R.; Ohta, Y.; Grillner, S. Organization of the six motor nuclei innervating the ocular muscles in lamprey. J. Comp. Neurol. 294:491–506; 1990. 13. Glover, J. C. The organization of vestibulo-ocular and vestibulospinal projections in the chicken embryo. Eur. J. Morphol. 32:193–200; 1994. 14. Glover, J. C. Retrograde and anterograde axonal tracing with fluorescent dextrans in the embryonic nervous system. Neurosci. Prot. 30:1– 13; 1995. 15. Glover, J. C. Neuroepithelial “compartments” and the specification of vestibular projections. Prog. Brain Res. 124:3–21; 2000. 16. Glover, J. C.; Petursdottir, G.; Jansen, J. K. S. Fluorescent dextranamines used as axonal tracers in the nervous system of the chicken embryo. J. Neurosci. Methods 18:243–254; 1986. 17. Glover, J. C.; Petursdottir, G. Regional specificity of developing reticulospinal, vestibulospinal and vestibulo-ocular projections in the chicken embryo. J. Neurobiol. 22:353–376; 1991. 18. Grapin-Botton, A.; Bonnin, M. A.; McNaughton, L. A.; Krumlauf, R.; Le Douarin, N. M. Plasticity of transposed rhombomeres: Hox gene induction is correlated with phenotypic modifications. Development 121:2707–2721; 1995. 19. Honig, M. G.; Hume, R. I. Carbocyanine dyes. Novel markers for labeling neurons. Trends Neurosci. 12:336 –338; 1989. 20. Hoover, F.; Kielland, A.; Glover, J. C. RXRg gene is expressed by discrete cell columns within the alar plate of the brainstem of the chicken embryo. J. Comp. Neurol. 416:417– 428; 2000. 21. Jessell, T. M. Neuronal specification in the spinal cord: Inductive signals and transcriptional codes. Nat. Rev. Genet. 1:20 –29; 2000. 22. Jungbluth, S.; Bell, E.; Lumsden, A. Specification of distinct motor neuron identities by the singular activities of individual Hox genes. Development 126:2751–2758; 1999. 23. Kessel, M. Reversal of axonal pathways from rhombomere 3 correlates with extra Hox expression domains. Neuron 10:379 –393; 1993. 24. Kimmel, C. B.; Powell, S. L.; Metcalfe, W. K. Brain neurons which project to the spinal cord in young larvae of the zebrafish. J. Comp. Neurol. 205:112–127; 1982. 25. Krumlauf, R. Hox genes and pattern formation in the branchial region of the vertebrate head. TIGS 9:106 –112; 1993. 26. Kwan, C. T.; Tsang, S. L.; Krumlauf, R.; Sham, M. H. Regulatory analysis of the mouse hoxb3 gene: Multiple elements work in concert to direct temporal and spatial patterns of expression. Dev. Biol. 232: 176 –190; 2001. 27. Lee, R. K.; Eaton, R. C. Identifiable reticulospinal neurons of the adult zebrafish, Brachydanio rerio. J. Comp. Neurol. 304:34 –52; 1991. 28. Lee, R. K.; Eaton, R. C.; Zottoli, S. J. Segmental arrangement of
693
29. 30. 31. 32. 33. 34. 35. 36. 37.
38.
39. 40. 41. 42. 43. 44.
45. 46.
reticulospinal neurons in the goldfish hindbrain. J. Comp. Neurol. 329:539 –556; 1993. Metcalfe, W. K.; Mendelson, B.; Kimmel, C. B. Segmental homologies among reticulospinal neurons in the hindbrain of the zebrafish larva. J. Comp. Neurol. 251:147–159; 1986. Moran-Rivard, L.; Kagawa, T.; Saueressig, H.; Gross, M. K.; Burrill, J.; Goulding, M. Evx1 is a postmitotic determinant of V0 interneuron identity in the spinal cord. Neuron 29:385–399; 2001. Oka, Y.; Satoui, M.; Ueda, K. Descending pathways to the spinal cord in the hime´ salmon (landlocked red salmon, Oncorhynchus nerka). J. Comp. Neurol. 2554:91–103; 1986. Petursdottir, G. Vestibulo-ocular projections in the 11-day chicken embryo: Pathway specificity. J. Comp. Neurol. 297:283–297; 1990. Prasado Rao, P.; Jadhao, A.; Sharma, S. Descending projection neurons to the spinal cord of the goldfish, Carassius auratus. J. Comp. Neurol. 265:96 –108; 1987. Rex, M.; Scotting, P. J. Chick HoxB3: Deduced amino-acid sequence and embryonic gene expression. Gene 149:381–382; 1994. Saueressig, H.; Burrill, J.; Goulding, M. Engrailed-1 and Netrin-1 regulate axon pathfinding by association interneurons that project to motor neurons. Development 126:4201– 4212; 1999. Schmid, A.; Chiba, A.; Doe, C. Q. Clonal analysis of Drosophila embryonic neuroblasts: Neural cell types, axon projections and muscle targets. Development 126:4653– 4689; 1999. Sham, M. H.; Vesque, C.; Nonchev, S.; Marshall, H.; Frain, M.; Gupta, R. D.; Whiting, J.; Wilkinson, D.; Charnay, P.; Krumlauf, R. The zinc finger gene Krox20 regulates HoxB2 (Hox2.8) during hindbrain segmentation. Cell 72:183–196; 1993. Simeone, A.; Acampora, D.; Arcioni, L.; Andrews, P. W.; Boncinelli, E.; Mavilio, F. Sequential activation of Hox 2 homeobox genes by retinoic acid in human embryonal carcinoma cells. Nature 346:763– 766; 1990. Straka, H.; Baker, R.; Gilland, E. Rhombomeric organization of vestibular pathways in larval frogs. J. Comp. Neurol. 437:42–55; 2001. Studer, M.; Po¨ pperl, H.; Marshall, H.; Kuroiwa, A.; Krumlauf, R. Role of a conserved retinoic acid response element in rhombomere restriction of Hoxb-1. Science 265:1728 –1732; 1994. Sundin, O.; Eichele, G. An early marker of axial pattern in the chick embryo and its respecification by retinoic acid. Development 114:841– 852; 1992. Suwa, H.; Gilland, E.; Baker, R. Segmental organization of vestibular and reticular projections to spinal and oculomotor nuclei in the zebrafish and goldfish. Biol. Bull. 191:257–259; 1996. Swain, G. P.; Snedeker, J. A.; Ayers, J.; Selzer, M. E. Cytoarchitecture of spinal-projecting neurons in the brain of the larval sea lamprey. J. Comp. Neurol. 336:194 –210; 1993. Tsuchida, T.; Ensini, M.; Morton, S. B.; Baldassare, T.; Edlund, T.; Jessell, T. M.; Pfaff, S. L. Topographic organization of embryonic motor neurons defined by expression of LIM homeobox genes. Cell 79:957–970; 1994. Vaage, S. The segmentation of the primitive neural tube in chick embryos (Gallus domesticus). Adv. Anat. Embryol. Cell Biol. 41:1– 88; 1969. Varela-Echavarria, A.; Pfaff, S. L.; Guthrie, S. Differential expression of LIM homeobox genes among motor neuron subpopulations in the developing chick brain stem. Mol. Cell. Neurosci. 8:242–257; 1996.