A role for midbrain arcs in nucleogenesis

In the adult brain, many neurons are organized into well-delineated clusters of characteristic shape and size. The role of these clusters, known as nuclei, is to organize neuronal circuitry. Thus, for a brain neuron, the projections of its axons, the neurotransmitters it employs and the local and long-distance inputs it receives, are governed by its nuclear assignment. Despite major recent insights into the early patterning events of the neural tube (Jessell and Lumsden,1997), very little is known about the cellular and molecular patterning mechanisms that generate discrete brain nuclei of appropriate size,location and cell-type composition.

The ventral midbrain provides an elegant model system for addressing the mechanisms of brain nucleogenesis. The adult midbrain tegmentum exhibits a complex structural architecture featuring spherical, ovate and plate-shaped nuclei. During embryogenesis, this nuclear organization is preceded by a more regular organization into a reiterated series of arcuate territories called midbrain arcs (Agarwala et al.,2001; Sanders et al.,2002). Arcs are known to be molecularly distinct, differing, for example, in their expression of homeobox genes(Agarwala et al., 2001). Whether the arcs simply reflect a mechanism for regulating the numbers of distinct midbrain cell-types or serve a more specific role in allocating neurons to particular nuclear fates is unknown.

The possibility that developmental periodicities might provide a patterning substrate for nucleogenesis has been addressed in connectional and fate-mapping studies of hindbrain rhombomeres. The striking finding is that most brainstem nuclei are generated from serially adjoining sets of rhombomeres with few nuclei developing from single rhombomeres(Cramer et al., 2000;Diaz et al., 1998;Lumsden and Keynes, 1989;Marin and Puelles, 1995). Thus, the extent to which anteroposterior periodicities are crucial to hindbrain nuclear specification remains uncertain(Wingate and Lumsden, 1996). Indeed, the generation of hindbrain nuclei, many of which form longitudinal columns, may be more tightly regulated in the mediolateral dimension(Clarke et al., 1998;Marin and Puelles, 1995). This observation is of particular interest because midbrain arcs, unlike hindbrain rhombomeres, are arrayed along the mediolateral axis parallel to the ventral midline.

In these studies, we explored the relationship between arc pattern formation and the generation of midbrain nuclei by focusing on the most medial arc as a prototypical arc. Using diagnostic connectional and molecular criteria we identified the anlagen for at least two midbrain nuclei within the medial arc: the oculomotor complex (OMC) and the red nucleus (RN). Both of these nuclei are part of the motor system, but their functions and connections are so dissimilar that a shared origin in the same arc was unexpected. The OMC contains motoneurons that control eye movements and the parasympathetic regulation of accommodation and pupil contraction(Evinger, 1988). The RN, by contrast, contains no motoneurons but is a cerebellar-related nucleus mediating motor cortex and cerebellar outflow to spinal cord in the control of limb movements (Holstege and Tan,1988; Keifer and Houk,1994; ten Donkelaar,1988). The two medial arc pronuclei are also molecularly distinct,differing in their expression of and dependence on homeodomain transcription factors. In this report, we show that the prospective RN is distinguished by expression of the POU domain gene BRN3A and the homeobox geneEMX2, and that Emx2 function is required for RN development.

Within the medial arc, the RN and OMC anlagen occupy distinct spatial positions in three dimensions. We reasoned that if the formation of these midbrain nuclei is in fact governed by arc patterning, then any perturbation of the medial arc should result in strictly coordinate perturbations of its constituent pronuclei. The medial arc is flanked by two signaling centers, the Sonic Hedgehog (SHH)-rich rostral floor plate and the FGF8- and WNT1-rich midbrain-hindbrain junction (Agarwala et al., 2001; Joyner et al.,2000). Manipulations of SHH and FGF8 signals are known to result in perturbations of the midbrain arcs along the mediolateral axis (SHH)(Agarwala et al., 2001) and of midbrain generally along the anteroposterior axis (FGF8)(Crossley et al., 1996). To test the relationship between arc pattern formation and the generation of midbrain nuclei, we misexpressed SHH and FGF8 with in vivo electroporation technology. Results from these experiments show that ectopicSHH expression anywhere within the midbrain is sufficient to induce both medial arc pronuclei. Furthermore, SHH manipulations that distort the midbrain arcs elicit completely parallel distortions of the OMC and RN primordia, and FGF8 manipulations that shift the arcs forward coordinately shift both medial arc pronuclei rostrally. These results suggest that the midbrain arcs represent a patterning template for allocating midbrain progenitor cells to their correct nuclear fates.

The Fgf8 expression construct pXeX-Fgf8 was engineered by inserting a 0.9 kb mouse Fgf8b cDNA(MacArthur et al., 1995) into the Xenopus EF1α promoter expression vector pXeX(Johnson and Krieg, 1994). pXeX-SHH, which contains a cDNA encoding the 19.2 N-terminal fragment of chick SHH, was constructed as described(Agarwala et al., 2001).

Fertilized chicken eggs were incubated in a humidified forced-draft chamber until the embryos reached stage 8-15(Hamburger and Hamilton,1951). Fifty to 250 nl of endotoxin-free plasmid DNA solution (1μg/μl in 10 mM Tris, 1 mM EDTA, 0.02% Fast Green) was injected through a glass capillary needle into the midbrain vesicles of chick embryos displayed by windowing. For conventional macroelectroporations(Muramatsu et al., 1998;Sakamoto et al., 1998), two platinum electrodes ∼2 mm apart were positioned to straddle the midbrain,and a series of eight electric pulses (50 mseconds, 9V) were delivered with a square wave electroporator (BTX 820, Genetronics, CA). Microelectroporation techniques (Momose et al.,1999) were employed to elicit more restricted midbrain transfections. A 500 μm tungsten wire was etched in 10% KOH to a tip diameter of 40 μm. The tungsten electrode was inserted into the lumen of the midbrain after plasmid injection and served as the negative electrode. The positive terminal was a platinum electrode placed outside the midbrain above the eye. The electrodes were held parallel at a distant of 1 mm and three electric pulses (25 mseconds, 7V) were delivered. Electroporated embryos were returned to the incubator for 3-4 days and on harvesting were submerged in 4%paraformaldehyde in a phosphate-buffered saline (PBS) solution. Embryos that showed some sign of injury from electroporation were excluded from further study. Our analyses are based on whole-mount and section in situ hybridization of 164 embryos successfully electroporated with the SHH(n=54) or Fgf8 (n=110) constructs.

Neural connections were studied in embryonic day (E) 5 and E6 chick brains(n=94) with fluorescein-conjugated lysinated dextrans(fluoro-emerald, 10 kDa, Molecular Probes). To label the OMC, the brain neurectoderm and the third cranial nerve were dissected away from adhering mesenchymal tissues in chilled Tyrode's solution and pinned to a SYLGARD dish. Tracer resuspended in water was dried to a slurry, picked up on a pair of fine forceps and applied to the dissected end of the nerve. For RN labeling, the tracer was dried on a minuten pin and stuck into dissected brainstem. Labeled embryo tissue was incubated for 30 minutes in chilled Neurobasal Medium (Life Technologies) with 5% fetal calf serum and 30 mM glucose, and placed for 2-4 hours in a 6% CO₂ incubator at 37°C. The tissue was then immersion-fixed overnight in 4% paraformaldehyde-PBS. The fluoresceinated dextrans were detected with anti-fluorescein Fab fragments conjugated to alkaline phosphatase in modified in situ hybridization protocols.

Chick embryos were collected at stage 12-15 and dissected in cold Tyrode's saline. Ventral midbrain explants with little or no adjoining dorsal midbrain tissue were kept briefly at 4°C in L-15 medium containing 10%heat-inactivated fetal bovine serum. The explants were then placed on a Millipore filter in a drop of Neurobasal medium containing 20 mM glucose,antibiotics and 10% fetal bovine serum, and were incubated in a 6%CO₂ incubator at 37°C for 2-3 days. Cultures were fixed in 4%paraformaldehyde-PBS for analysis by in situ hybridization.

All mice were cared for according to animal protocols approved by the IACUC of the University of Chicago. Outbred CD-1 timed pregnant mice were obtained from the University of Chicago Cancer Research Center Transgenic Facility.Emx2 heterozygote mice provided by P. Gruss(Pellegrini et al., 1996) were bred and genotyped by PCR. Noon of the day of vaginal plug discovery was considered E0.5. Homozygous Emx2 mutant embryos (n=25)between E12.5-E18.5 were recovered for gene expression analyses and compared with equal or greater numbers of wild-type and heterozygote littermate mice(n=27).

Two-color whole-mount in situ hybridization was carried out on chick brain with digoxigenin- and fluorescein-labeled riboprobes synthesized from chick cDNA plasmids for ACHE, BRN3A/POU4F1, EMX2, EVX1, FGF8, FOXA2, GBX2, ISL1,OTX2, PAX3, PAX6, PHOX2A, SHH, TH and WNT1. Whole-mount and section in situ hybridization was carried out on E12.5-E18.5 mouse neurectoderm with riboprobes for mouse class III β-tubulin/Tubb6,Emx2 and Fgf8, and rat Brn3a, Isl1 and Th. Riboprobe-mRNA duplexes and fluoresceinated dextrans were detected with antibody-phosphatase conjugates (Roche Molecular Biochemicals) and demonstrated by phosphatase histochemistry employing the distinguishable tetrazolium chromagens NBT and TNBT.

Midbrain arcs are identified as acetylcholinesterase (ACHE)-rich arcuate cell-columns arrayed bilateral to the midbrain floor plate(Fig. 1A)(Ragsdale and Lumsden, 1995;Sanders et al., 2002). Two lines of evidence establish that motoneurons of the oculomotor complex (OMC)are embedded in the most medial ACHE-rich arc (arc 1). Gene expression for motoneuron-specific transcription factors PHOX2A andISL1 (Pattyn et al.,1997; Pfaff et al.,1996) is confined to the medial arc(Fig. 1A,B; data not shown),and retrograde tracer delivery to the third cranial nerve labels the medial arc PHOX2A+territory (Fig. 2A,B).

OMC neurons are not the sole constituents of the medial arc. Situated deep to the OMC, near the pial surface of the midbrain, is a second group of cells,identified as dopaminergic (DA) neurons by tyrosine hydroxylase (TH)gene expression (Fig. 1C,D). In the mouse, midbrain DA neurons have been reported to be migratory(Marchand and Poirier, 1983),raising the possibility that chick medial arc TH+ cells have migrated in from the adjoining ventral midline. This observation, combined with the late onset of TH gene expression in the chick and the lack of availability of early markers of chick DA neurons(Hynes et al., 1995;Smidt et al., 1997;Zetterstrom et al., 1997),leads to uncertainty about the origin of midbrain DA cells. We therefore confined our analyses of the medial arc to the development of its other constituents.

Overlapping the rostral end of the OMC along the anteroposterior axis and completely segregated from the OMC neurons along the ventricular-pial axis, is a third population of medial arc cells. This third territory is selectively enriched in mRNA for the homeodomain transcription factor EMX2 and for the POU domain transcription factor BRN3A (Fig. 1E-H; data not shown).

Connectional criteria establish that this EMX2+ andBRN3A+ territory of the medial arc contains neurons of the prospective red nucleus (pRN), a major brainstem nucleus involved in the control of limb movement. In tetrapod vertebrates, the only exclusively contralateral projection from midbrain tegmentum to lower brainstem and spinal cord originates in the RN (ten Donkelaar,1988). Thus, this projection is diagnostic of the RN. Small tracer deposits of fluoresceinated dextrans centered in the medial arc at E6 elicit anterograde labeling of multiple fiber-tracts descending into the hindbrain,one of which crosses the midline (Fig. 2C). In reciprocal experiments, tracer injections placed in E5 or E6 hindbrain produce retrograde labeling of a single population of exclusively contralaterally projecting cells (Fig. 2D). PHOX2A (Fig. 2E) and BRN3A (Fig. 2F) gene expression combined with retrograde tract-tracing demonstrates that these identified RN neurons lie in the BRN3A+/EMX2+territory of the medial arc.

These data establish that the medial arc contains multiple cell types distinguished by unique axonal connections and patterns of transcription factor gene expression. These cell types are not, however, spatially intermixed, but form discrete territories. Along the anteroposterior axis, the pRN, the OMC and the medial arc TH neurons appear in succession, with the OMC overlapping both the pRN and the TH neurons when viewed from the ventricular surface. Along the ventricular-pial axis, the three cell-groups are fully segregated, with OMC lying closest to the ventricular layer. Thus, at E5 when midbrain neurogenesis is still under way (T. A. Sanders, PhD thesis, University of Chicago, 2000), the mantle layer of the ventral midbrain is patterned in all three dimensions.

Previous studies have suggested that neurons of the RN, like the neurons of hindbrain precerebellar nuclei, are produced by dorsal neural tube(Cooper, 1946;Hamilton et al., 1959;Sidman and Rakic, 1982;Streeter, 1912). Indeed,modern textbooks illustrate red nucleus neurons that originate in dorsal midbrain and translocate to ventral midbrain in migratory streams(Haines, 2002;Sadler, 1995). In light of our connectional findings indicating that RN neurons are found in the medial arc of the ventral midbrain by E5, we further investigated the origin of the RN in an explant culture system. Ventral midbrain tissue was harvested for organotypic culture at stage 12-15, a time when postmitotic neurons detected in dorsal midbrain are massed in the prospective mesencephalic trigeminal nucleus adjoining the roof plate (not illustrated)(Easter et al., 1993). After 3-4 days in vitro, the ventral midbrain cultures displayed strong expression of the pRN marker BRN3A (Fig. 3). Moreover, based on PHOX2A mRNA expression, the explant pRN formed in a medial arc-like territory whose three-dimensional architecture closely matched that observed in vivo, with a BRN3A+ pRN lying anterior and pial to a PHOX2A+ OMC(Fig. 3C,D; data not shown). We detected a medial arc pRN in all explants examined, even when noPAX3+/BRN3A+ dorsal midbrain tissue remained attached to the ventral midbrain explant (Fig. 3A,C,residual tectum attached, n=54;Fig. 3B,D, no residual tectum,n=5). We conclude that BRN3A+ pRN neurons originate in ventral midbrain, and that the generation of a medial arc properly organized in three dimensions is an autonomous property of E3 ventral midbrain.

The homeobox gene EMX2 identifies the RN anlage in the medial arc. We sought genetic evidence for an EMX2 role in RN development by analyzing mice with targeted deletions of the Emx2 gene(Pellegrini et al., 1996).Emx2 gene expression establishes that, as in the chick, the mouse ventral midbrain is organized into arcuate stripes(Fig. 4A). The most medialEmx2+ arcuate stripe, when viewed in whole-mount preparations,overlaps with Isl1 and Brn3a gene expression territories,indicating the presence of both an OMC (Isl1+) and a pRN(Emx2+, Brn3a+) within the mouse medial arc(Fig. 4B-D).

In homozygous Emx2 mutant mice, Brn3a gene expression in ventromedial midbrain is retained between E12.5-E16.5, suggesting that pRN neurons are generated in the absence of Emx2 function(Fig. 5A,B). At E18.5, however,no RN can be identified with Brn3a or Class III β-tubulin gene expression (Fig. 5C-F), whereasIsl1+ and Th+ cells appear unaffected(Fig. 5G,H). Thus, within the medial arc, Emx2 is specifically expressed by the RN pronucleus and appears to be essential for RN development.

The first arc is flanked medially by the rostral floor plate, a source of the signaling molecule SHH. As illustrated inFig. 6, the relationship between the SHH source and the medial arc is dynamic between E2, the onset of arc pattern formation, and E5, when both the pRN and the OMC can be readily identified by their signature transcription factors. At E2,SHH gene expression in the midbrain is confined to the midline(Fig. 6A), but by E5 it has fanned out into the ventricular region overlying the medial arc(Fig. 6B-D). Previous work has shown that SHH can induce motoneurons and dopaminergic neurons in the ventral midbrain (Agarwala et al.,2001; Hynes et al.,1995; Wang et al.,1995; Watanabe and Nakamura,2000). Given the proximity of the BRN3A+ cells to theSHH source, we tested by in ovo electroporation whether SHHcan also elicit pRN neurons. We were particularly interested in the spatial relationship between any induced pRN and OMC neurons. As SHH is thought to specify multiple cell-types in a concentration- and position-dependent manner(Agarwala et al., 2001;Ericson et al., 1997), the observation that the pRN and the OMC are equidistant from the ventral midline SHH source at E2 would appear to predict that any induced pRN neurons would be intermixed with OMC neurons. What we found, though, was very different.

SHH misexpression in the ventral midbrain of stage 8 to 15 chick embryos produced by E5 expanded expression of SHH(Fig. 7A) and its transcriptional targets FOXA2/HNF3B and PTC1(Fig. 7B)(Agarwala et al., 2001).SHH overexpression also expanded the population of PHOX2A+oculomotor neurons and the pRN territory identified by BRN3A andEMX2 labeling (Fig. 7C-F). Remarkably, the normal spatial relationship between the OMC and the pRN territories, seen in the control (left) hemitegmentum, was maintained in the enlarged territories along all three dimensions. The expanded OMC and pRN were coextensive along the mediolateral axis(Fig. 7D,F), partially overlapping along the anteroposterior axis(Fig. 7C,E) and completely segregated along the ventricular-pial axis(Fig. 7D,F). These results demonstrate that SHH misexpression does not produce an intermixture of OMC and pRN cell-types, but rather a coordinated expansion of discrete OMC and pRN territories.

SHH misexpression in lateral and dorsal midbrain can elicit ectopic midbrain arcs (Agarwala et al.,2001). We tested whether SHH misexpression in lateral tegmentum or in tectum would also produce pRN cell-types. SHHelectroporation at the tectal-tegmental border(Fig. 7G) or in dorsal midbrain(Fig. 7H), where arcs are normally never seen, reproduced the entire pattern of midbrain arcs, including the EMX2+/BRN3A+ pRN and the PHOX2A+ OMC territories of the medial arc. The ectopic OMC and pRN occupied discrete positions within the ectopic medial arc, such that their spatial relationship always resembled that of the normal OMC and pRN in all three dimensions(Fig. 7H insert). Thus, each of a variety of SHH manipulations in midbrain development coordinately affect the size and shape of the pRN and the OMC, indicating that the patterning element regulated by SHH is the medial arc and not simply the induction of its constituent cell types.

An ectopic SHH signal in dorsal or ventral midbrain produces OMC and RN anlagen that are offset in the anteroposterior axis. This finding suggests that midbrain cells, in interpreting the SHH signal, rely on independent positional information regarding their anteroposterior position [see also Fedtsova and Turner (Fedtsova and Turner,2001)]. We have tested the role of the isthmus in supplying this positional information by shifting its location rostrally, into prospective midbrain. To shift the isthmus forward, we drew on previous findings that ectopic expression of FGF8 can transform prospective midbrain into hindbrain tissue (Irving and Mason,1999; Liu et al.,1999; Sato et al.,2001) and that the juxtaposition of hindbrain and midbrain tissue,even at ectopic sites, induces a new isthmus(Irving and Mason, 1999). This manipulation, which contracts the size of the midbrain tegmentum, allowed us to ask whether it is the position of the isthmus caudally or the distance from the forebrain rostrally that controls the anteroposterior patterning of the medial arc. If it is distance from the isthmus, then rostral midbrain cell-types should be lost as the isthmus is moved rostrally. If it is distance from the forebrain, then moving the isthmus forward should delete caudal midbrain territories. If position relative to both boundaries is important,then rostral advance of the isthmus may delete central elements.

Fgf8 delivery by electroporation to midbrain at E2 (HH stages 8-15) produced by E5 a rostral expansion of the GBX2+ hindbrain territory, a rostral regression of the OTX2+ midbrain territory and a concomitant rostral shift of the isthmus, identified by WNT1 andFGF8 gene expression (Fig. 8B,D; data not shown). In addition, the electroporated midbrain often took on the tubular appearance of rostral hindbrain(Fig. 8D). In the Fgf8electroporated hemitegmentum, the hindbrain trochlear nucleus (a rhombomere 1 derivative) was elongated and the midbrain arcs, identified by PHOX2A,PAX6 and EVX1 labeling, appeared compressed(Fig. 8A,E). Within the medial arc, PHOX2A and BRN3A gene expression territories were both shifted rostrally (Fig. 8E-H). The extent of this rostral shift was related to the anteroposterior location of the ectopic isthmus. Thus, progressively more rostral Fgf8electroporations resulted in progressively more rostrally located isthmi,which in turn produced more rostrally located OMC and RN pronuclei(Fig. 8E-H).

Remarkably, the positional relationships of the shifted PHOX2A+and BRN3A+ territories to the ectopic isthmus remained constant: the OMC territory always abutted the isthmus, the pRN territory appeared at a distance, and the rostral OMC and the caudal pRN territories overlapped. However, with very rostral Fgf8 electroporations that converted the midbrain tegmentum entirely to GBX2+/OTX2- hindbrain,BRN3A gene expression was severely reduced (n=4) or could not be detected (n=2; data not shown), with PHOX2A labeling entirely restricted to the subthalamus of caudal forebrain. These results demonstrate that manipulations that shift the medial arc rostrally produce corresponding rostral shifts in the OMC and the pRN that invariably maintain the spatial relationships between these pronuclei. In addition, our results show that the pronuclei of the medial arc are specified according to their distance from the caudally located isthmus.

We have found that the medial arc contains the anlagen of two principal nuclei of the midbrain: the OMC and the RN. The OMC was localized to the medial arc by retrograde labeling of third cranial nerve motoneurons and gene expression for motoneuron-specific transcription factors. We identified the RN anlage by the diagnostic criterion of a projection from midbrain tegmentum towards lower brainstem and spinal cord that is exclusively contralateral(Lawrence and Kuypers, 1968;Massion, 1967;ten Donkelaar, 1988). Anterograde and retrograde tracing experiments established the presence of such a crossed projection from a medial arc territory that is spatially and molecularly distinct from the OMC. A crossed rubrospinal projection has been found in all birds and mammals studied(ten Donkelaar, 1988), and is thought to originate from magnocellular RN neurons(Burman et al., 2000;ten Donkelaar, 1988). In this study, we have not distinguished between magnocellular and parvocellular subdivisions of the chick RN because in the adult bird neurons of all sizes appear throughout the RN (Wild et al.,1979).

Although many populations of neurons remain within the region of brain where they are generated, some migrate great distances to reach their final locations (Rakic, 2000). Notable examples include dorsal hindbrain neurons that migrate ventrally to populate precerebellar nuclei in the pons and medulla(Harkmark, 1954;Rodriguez and Dymecki, 2000;Wingate and Hatten, 1999). Previous studies have held that neurons of the midbrain RN, like neurons of the hindbrain cerebellar system, are initially generated in dorsal neural tube(Cooper, 1946;Haines, 2002;Hamilton et al., 1959;Sidman and Rakic, 1982;Streeter, 1912). These conclusions were primarily based on histological descriptions of streams of cells that appeared to be migrating from dorsal to ventral midbrain(Sadler, 1995). In this study,we found that SHH, a ventral midline positional signal, can induce a RN anlage and precisely regulate its size. In addition, we found that a BRN3A+pRN could develop in ventral midbrain explant cultures that had no dorsal midbrain attached. These observations lend strong support to the idea that,unlike many other cerebellar system neurons, neurons of the RN are ventral neural tube derivatives.

Two homeobox genes, BRN3A and EMX2, identify the pRN territory of the medial arc. Analysis of mutant mice indicates that both genes are needed for RN development. In Emx2-deficient mice, theBrn3a-positive neurons of the pRN are generated, but are not detected beyond E16.5. By E18.5, RN neurons are also lost in mice lackingBrn3a (McEvilly et al.,1996; Xiang et al.,1996). Thus, both Emx2 and Brn3a are essential for the proper maturation of the red nucleus, although they appear not to be required for its initial specification (this report)(McEvilly et al., 1996). ThatEmx2 and Brn3a are both expressed in the pRN suggests that the genetic requirement for these transcription factors is autonomous to pRN neurons, but it will be important to test this possibility directly in chimeric mouse experiments (Alvarez-Bolado et al., 2000).

Previous studies of Emx2 function have indicated a role for this gene in the growth and initial areal patterning of the cerebral cortex(Bishop et al., 2000;Cecchi and Boncinelli, 2000;Mallamaci et al., 2000). Our findings suggest a broader role for Emx2 in brain development, one that includes the regulation of neuronal survival and the development of nuclei. In addition, the finding that the development of the RN requires at least two homeodomain transcription factors, Emx2 and Brn3a, coupled with previous work indicating that multiple homeodomain proteins are needed for the generation of oculomotor neurons and midbrain dopamine neurons(Pattyn et al., 1997;Simon et al., 2001;Smidt et al., 2000), suggests the general hypothesis that the development of distinct nuclei within specific arcs is regulated by multiple homeodomain transcription factors.

Our in ovo electroporation experiments establish that SHHmisexpression eliciting an expansion or induction of the medial arc produces both oculomotor and pRN cell-types. With all SHH manipulations, the spatial relationships of the ectopic OMC and pRN neurons remain constant and resemble those within the unmanipulated medial arc: the OMC and pRN form distinct pronuclei that are co-extensive in the mediolateral axis and partially separated along the anteroposterior axis. We tested whether this anteroposterior segregation of the medial arc would be maintained if the medial arc were moved rostrally within the ventral midbrain. We found that as an ectopic isthmus is induced at progressively more anterior positions by ectopic FGF8 delivery, the medial arc is shifted rostrally as well. In these brains, the OMC and the pRN appear at progressively more anterior sites, until first the pRN, and eventually both the pRN and OMC territories are lost, and the entire electroporated half of the midbrain is converted to hindbrain. These results demonstrate that rostral shifts in the location of the medial arc result in corresponding shifts in the OMC and the pRN. They also show that anteroposterior patterning of medial arc pronuclei is controlled by caudal positional signals emanating from the midbrain-hindbrain junction. These signals may include FGF8 itself, as well as other isthmus-specific FGF and WNT proteins (Carl and Wittbrodt,1999; Crossley et al.,1996; Wurst and Bally-Cuif,2001). Interestingly, whatever the molecular character of these anteroposterior patterning cues, our results indicate that they are present throughout the midbrain, as SHH misexpression anywhere in midbrain results in the production of an ectopic medial arc within which the ectopic pRN and the OMC retain appropriate anteroposterior polarity.

An emerging view in studies of neural development is that a two-dimensional Cartesian coordinate system specifies cell fates within the developing neuroepithelium. According to this `grid' model, orthogonal signaling centers provide positional information for the medial-lateral and anteroposterior patterning of the nervous system (Jessell and Lumsden, 1997; Simon et al., 1995; Wolpert,1969; Ye et al.,1998). However, this model is not sufficient to account for the segregation we have found of the midbrain pronuclei along the ventricular-pial axis. What might be the mechanism for patterning midbrain nuclei along this third axis? In the rat, although the OMC and the RN are born between E12-E14,the peaks of their neurogenesis are staggered by 0.5-1.0 day(Altman and Bayer, 1981;Marchand and Poirier, 1982). Thus, cells that leave the ventricular zone could be specified successively to be OMC neurons (born E12-E13) then RN neurons (born E13-14). The mechanism underlying such a temporal switch in cell-type generation could be the length of time a progenitor cell spends in the proliferative zone. Alternatively, the switch may be mediated by specific signaling molecules(Desai and McConnell,2000).

One possibility, prompted by our observations on the changing profile of midbrain SHH gene expression, is that SHH itself could participate in ventricular-pial patterning. As Fig. 6 illustrates, although the source of SHH in the ventral midbrain is a strip at its onset, by E3 it fans out to assume a wedge-shaped profile (Ericson et al., 1995;Marti et al., 1995). Based on the rat midbrain neurogenesis studies and the onset of PHOX2A in the chick OMC by E2 and BRN3A in the pRN by E4, SHH gene expression overlies the medial arc at a time when the fates of medial arc neurons are still being specified (Altman and Bayer, 1981). That the same signaling systems can be involved sequentially in the control of patterning along different axes has been shown in studies of early Drosophila development(Gonzalez-Reyes et al.,1995).

We have examined the relationship between midbrain arc pattern formation and the specification of midbrain nuclei, using the medial arc as a prototypical arc. Our results indicate that a single arc can harbor multiple spatially segregated neuronal territories that by connectional and molecular criteria correspond to specific midbrain nuclei. These findings suggest that one role for the arcuate patterning of the ventral midbrain is the generation of nuclei.

We thank S. Assimacopoulos for the pXeX-Fgf8 construct, A. Greenlee and A. Lin for technical assistance, and E. Grove for comments on the manuscript. Nucleic acid reagents were given to us by E. Boncinelli, D. Chikaraishi, C. Goridis, M. Goulding, T. Jessell, S. Lamont, D. Latchman, A. Leutz, C. MacArthur, A. McMahon, G. Martin, C. Tabin and M. Wassef. TheEmx2 knockout mice were provided by P. Gruss with the generous assistance of K. Bishop, G. Goudreau and D. O'Leary. This work was supported by the NINDS/NIH.