PHOX2A regulation of oculomotor complex nucleogenesis

A characteristic feature of the vertebrate brain is that its neuronal cell bodies are organized into clusters. These clusters, known as nuclei, serve to organize much of the circuitry of the brain (Nieuwenhuys et al., 1998). Key attributes of nuclei include a spatial arrangement of neurons, often with multiple cell types, which have distinct inputs, axonal projections and neurotransmitter phenotypes. Only limited progress has been made in understanding how neuronal progenitor cells are allocated to different nuclear fates, and the molecular mechanisms that coordinate cell type composition, spatial arrangement and nucleus size remain largely unknown (Agarwala and Ragsdale, 2002; Diaz et al., 1998; Kawauchi et al., 2006).

The chick midbrain oculomotor complex (OMC) is an attractive model system for studying brain nucleogenesis. In vertebrates, the OMC supplies somatic motoneuron input to extraocular muscles and preganglionic autonomic input to the ciliary ganglia. The somatic and visceral midbrain motoneurons form separate pools, although the pool of visceral motoneurons in rodents is not sharply distinguished in Nissl-stained preparations or by gene expression. By contrast, the chicken shares with all birds a cleanly segregated OMC with discrete subnuclei that innervate specific autonomic and ocular muscle targets (Evinger, 1988; Gamlin et al., 1984; Heaton and Wayne, 1983). Indeed, in birds the preganglionic motoneurons arise exclusively from the Edinger-Westphal (EW) nucleus, which is the dorsal component of the OMC (Gamlin and Reiner, 1991; Reiner et al., 1991). The segregation of neurons in the developing chick OMC into discrete subnuclei occurs in a series of migrations, as documented in the classic histological studies of Levi-Montalcini (Levi-Montalcini, 1963) and Puelles (Puelles, 1978). The cellular and molecular factors that mediate these migrations are unknown (Guthrie, 2007).

Genetic studies over the past ten years have identified the PHOX2A homeobox gene as an important regulator of midbrain oculomotor development. Zebrafish embryos with a single amino acid mutation in the Phox2a homeodomain exhibit a reduction in oculomotor nucleus precursor cells (Guo et al., 1999). In Phox2a^−/− mice, midbrain oculomotor neurons are absent (Pattyn et al., 1997). In humans, PHOX2A mutations are linked to severe defects in extraocular muscle movement and in pupillary light reflexes (Bosley et al., 2006; Nakano et al., 2001), and magnetic resonance imaging has established that the oculomotor, or third, cranial nerve is missing in these patients (Bosley et al., 2006). The closely related homeobox gene Phox2b, which is required for the development of hindbrain branchiomotor and visceromotor neurons, is also expressed in midbrain oculomotor neurons but is dispensable for their production (Pattyn et al., 2000; Pattyn et al., 1997). Recent findings, however, have shown some degree of shared competency between the Phox2a/b genes in motoneuron development. Knock-in of Phox2b coding sequence into the Phox2a locus allows for the development of some midbrain oculomotor neurons (Coppola et al., 2005). Gain-of-function studies in chick spinal cord have demonstrated that exogenous Phox2a/b are able to upregulate the expression of motoneuron markers and induce ectopic axons that can contribute to existing nerve roots (Hirsch et al., 2007; Pattyn et al., 1997). Interestingly, these axons exit the cord through dorsal exit points, a feature of hindbrain branchiomotor and visceromotor neurons that distinguishes them from somatic motoneurons. This evidence establishes that Phox2a/b are required for brainstem motoneuron development and can, when misexpressed, induce motoneurons that issue axons that exit the nerve cord.

In this study, we extended the analysis of PHOX2A motoneuron induction to midbrain in order to address a very different kind of question: to what extent can a single transcription factor drive the production of a spatially organized nuclear complex that contains different cell types? First, we report on normal OMC development, showing that the visceral and somatic motoneurons of the chick OMC are generated in an ‘inside-out’ fashion, with the final position of the OMC neurons dependent on their time of origin. Second, we demonstrate that exogenous Phox2a can induce a complete OMC molecular program, including the production of both visceral and somatic motoneurons. Crucially, we document that the ectopic visceral and somatic oculomotor neurons segregate along the ventricular-pial axis, mimicking the architecture of the native OMC. Third, we demonstrate that Phox2a-induced motoneurons make ectopic cranial nerves that recruit boundary cap cells and, in some instances, target extraocular muscles directly. These results, combined with previous findings in mouse and human, establish that a single transcription factor can act as a primary developmental determinant for a brain nucleus and its constituent cell types.

Fertilized Babcock B-300 White Leghorn chicken eggs (Gallus gallus domesticus) were supplied by Phil's Fresh Eggs (Forreston, IL, USA). The first day of incubation was designated embryonic day 0 (E0). Staging of embryos followed Hamburger and Hamilton (Hamburger and Hamilton, 1951). Our chick embryo protocols were approved by the University of Chicago IACUC.

The human EF1α (EEF1A1) promoter expression construct pEFX was generated from pEF1/Myc-His C (Invitrogen) by excising a 2.2 kb PvuII fragment containing an f1 origin, neomycin resistance gene and SV40 elements. The rat Phox2a (rArix) expression construct pEFX-rArix was engineered in pEFX by insertion of a 1.6 kb cDNA containing the complete rat Phox2a coding sequence (Zellmer et al., 1995). pEFX-mPhox2b contains a 1.3 kb full-length mouse Phox2b cDNA isolated from E12.5 brain RNA by RT-PCR using the forward primer 5′-GCCATCCAGAACCTTTTCAATG-3′ and the reverse primer 5′-CCGTCTCTCCCTATCACCTACTTG-3′.

For the forced expression of Phox2a/b, we employed expression plasmid delivery by in ovo microelectroporation (Agarwala et al., 2001; Momose et al., 1999), targeting, in most experiments, the rostrolateral right ventral midbrain. Negative electrodes were fabricated from 0.025 mm tungsten wire and positive electrodes from 0.063 mm platinum wire (Goodfellow). Plasmid DNA solution (50-250 nl at 1-3 μg/μl in endotoxin-free water containing 0.02% Fast Green) was pressure-injected from a glass capillary into the midbrain vesicle of Hamburger and Hamilton stage (st) 9-24 embryos. The negative electrode was placed in the midbrain lumen and the positive electrode just outside the midbrain. Electric fields were generated with a Model 2100 Isolated Pulse Stimulator (A-M Systems) delivering three 25-millisecond pulses of 9 V, with an inter-pulse period of 1 second. Electroporated embryos were incubated for a further 1-4 days. Our analyses are based on gene expression in 704 embryos successfully electroporated with the pEFX-rArix (n=433) and pEFX-mPhox2b (n=271) constructs. Of these embryos, 316 were co-electroporated with pEFX-EGFP to assess the zone of misexpression before processing. In 23 embryos, we co-electroporated the alkaline phosphatase reporter plasmid pEFX-AP to label midbrain efferent axons. For the remaining experiments, the reproducible accuracy of plasmid delivery was confirmed through parallel electroporations of pEFX-AP, which was demonstrated by histochemistry, or of pEFX-rArix or pEFX-mPhox2b, which were detected by Phox2a/b transgene in situ hybridization.

Harvested embryos were fixed in 4% paraformaldehyde (PFA) in phosphate-buffered saline (PBS). Dissected embryonic brains were processed as whole-mounts or cryoprotected and sectioned at 40 μm on a sledge microtome. In situ hybridization was carried out with digoxigenin-, DNP- and fluorescein-labeled riboprobes synthesized from chick cDNA plasmids for CDH7, CHAT, CX36, EGR2 (KROX20), EVX1, ISL1, ISL2, NKX6.1, NRG1, PAX6, PHOX2A, PHOX2B, ROBO2, SLIT3, TBX20 and VT and from plasmids for rat Phox2a and mouse Phox2b. For light microscopy, labeled RNA duplexes were detected with antibody-alkaline phosphatase conjugates (Roche Diagnostics) and using the distinguishable formazans of NBT and TNBT (Grove et al., 1998). For fluorescence microscopy, antibody-horseradish peroxidase conjugates were substituted and peroxidase activity was demonstrated by fluorescent tyramide signal amplification using FITC-tyramide for fluorescein-labeled probes, cyanine 3-tyramide for digoxigenin-labeled probes and cyanine 5-tyramide for DNP-labeled probes (TSA Plus Fluorescence Systems Kit, PerkinElmer).

5-bromo-2-deoxyuridine (BrdU) can be very toxic to chicken embryos, particularly before E2.5 (Gould et al., 1999). Extensive scout experiments were carried out to identify a BrdU dose that elicits clear nuclear labeling while allowing embryos to survive without obvious morphological defects to E12. We found success with BrdU (Sigma) prepared as a 75 μg/ml solution in PBS containing 0.01% Fast Green and delivered as a 50 μl aliquot through a small hole made in the vitelline membrane near the heart. Twenty embryos with BrdU injection at st 8, 9 and 11-23 were harvested at E12 with PFA-PBS fixation. Serial sections (40 μm) of midbrain were first processed for TNBT gene expression histochemistry with digoxigenin-labeled ISL1 riboprobes to identify OMC structure. Sections were then treated with 2M HCl for 30 minutes to denature nuclear DNA and, after washing, incubated in a 1:50 dilution of monoclonal FITC-conjugated anti-BrdU antibody (Becton Dickinson) in 2% lamb serum. The anti-BrdU antibody was detected with the anti-fluorescein phosphatase conjugate and NBT histochemistry employed in our gene expression experiments.

Tissue was photographed with a Zeiss AxioCam digital camera attached to a Zeiss Axioskop 50 upright microscope or a Leica MZ FLIIII stereomicroscope. Images were captured through the AxioVision 4.5 software system and processed with Adobe Photoshop. Images requiring increased depth of focus were assembled with Helicon Focus 4.6 Pro software.

The avian OMC is a spatially organized nuclear complex comprising a dorsal EW nucleus and a ventral somatic motor nucleus (Evinger, 1988; Heaton and Wayne, 1983). The EW nucleus contains the midbrain visceral motoneurons, which innervate the ciliary ganglion and control pupillary constriction, accommodation and choroidal blood flow (Gamlin et al., 1982). The somatic oculomotor nucleus harbors the midbrain somatic motoneurons, which in birds are discretely organized into subnuclear pools, each innervating distinct extraocular muscle targets (Fig. 1A). In mammals, the somatic oculomotor neurons aggregate to form a single nucleus, and in many species, including mouse, the EW nucleus is challenging to identify (Weitemier et al., 2005). The anatomical clarity of the avian OMC coupled with the ease of experimental manipulation makes the chicken embryo attractive for investigations of midbrain motoneuron development.

To identify genes that reflect the connectional heterogeneity of the OMC subnuclei, we screened chick midbrain sections at E12, an age at which the OMC subnuclei are readily distinguished. Genes encoding known motoneuron markers, such as the LIM-homeodomain protein ISL1, were expressed by all oculomotor subnuclei (Fig. 1A; see Fig. S1 in the supplementary material) (Jessell, 2000). Three markers, however, identified visceral and somatic subnuclei specifically. The EW nucleus was labeled by expression of the neuropeptide gene for vasotocin (VT) (Muhlbauer et al., 1993). Somatic OMC subnuclei expressed the gap junction subunit CX36 and the T-box transcription factor TBX20 genes (Fig. 1C,D).

We constructed a timeline of OMC molecular and cellular development to relate the expression of the motoneuron subtype markers to other aspects of OMC maturation (Fig. 2). To study oculomotor nerve development, we electroporated an alkaline phosphatase-encoding expression vector into E2 midbrain and processed the embryos for phosphatase histochemistry 1-4 days later. We found that the axons are first issued from ventral midbrain at st 16 and reach the ipsilateral orbit over the next 2 days, contacting the inferior oblique (IO) muscle by st 25 (data not shown). This schedule matches that described by Chilton and Guthrie, who employed anti-neurofilament antibody staining to mark axons (Chilton and Guthrie, 2004).

We found that OMC marker onset readily fell into two groups: early onset genes that were expressed before oculomotor nerve exit and which principally encode transcription factors, and late onset genes that were expressed after nerve exit but before target innervation (Fig. 2).

The homeodomain transcription factor gene NKX6.1 was the first OMC marker that we detected in medial ventral midbrain, but it is not selective for midbrain motoneurons as it is also expressed in red nucleus progenitor domains and more lateral midbrain territories (P. M. Garfin, PhD thesis, University of Chicago, 2004).

PHOX2A, which is restricted to OMC cells, is detected at st 12+. PHOX2B is expressed in prospective OMC cells immediately thereafter, at st 13−, and the LIM homeobox genes ISL1 and ISL2 are detected by st 14+ (Fig. 2) (Varela-Echavarria et al., 1996).

The late onset genes included the motoneuron neurotransmitter enzyme choline acetyltransferase (CHAT; st 24), the axon guidance regulator ROBO2 (st 25) and the specific markers that distinguish visceral and somatic motoneuron identity. The visceral oculomotor neuron marker VT was first detected at st 20 (E3.5). The onset of the somatic oculomotor neuron markers CX36 and TBX20 occurred later, at st 25, which is when the oculomotor nerve makes initial contact with extraocular muscle (Fig. 2, black arrow). Although there might be other subtype-restricted molecules that are expressed earlier in OMC development, the differential time of onset of the visceral and somatic motoneuron markers raised the possibility that OMC visceral motoneurons are born before the OMC somatic motoneurons.

The embryonic ventral midbrain is organized into a series of arcs, with the PHOX2A-rich OMC arising in the most medial arc, arc 1 (Agarwala and Ragsdale, 2002; Sanders et al., 2002). Thus, unlike the somatic and visceral motoneurons of hindbrain (Guthrie, 2007), the midbrain somatic and visceral motoneurons appear, by gene expression, to be generated from a common progenitor pool, that is, from the ventricular zone overlying arc 1. To test for a timing difference in the neurogenesis of visceral and somatic oculomotor neurons, we delivered BrdU to embryos between E2 and E4.5 and collected them at E12, when the OMC subnuclei are easily recognized. BrdU birthdating in chick is problematic because delivered BrdU is not a true pulse as it is in mammals, but is continuously available in the egg up to 22 hours after delivery (Gould et al., 1999). Consequently, we were not able to determine precisely when neurogenesis begins, only when it ends. Heavy labeling, however, is predicted to mark cells undergoing their final divisions.

We found clear evidence for a temporal difference in the birth order of the oculomotor subtypes. Neurogenesis begins first in visceral oculomotor neurons (Fig. 3A). The neurons of the EW nucleus were most heavily labeled when BrdU was applied at st 11 and 12 (Fig. 3A; data not shown), close to the time of PHOX2A expression onset (st 12+). Most visceral oculomotor neuron progenitors have undergone their final cell division by st 18, when only a few medial EW cells still incorporated BrdU (Fig. 3C), and visceral motor neurogenesis was largely complete by st 19 (data not shown). By contrast, somatic oculomotor neuron progenitors continued to divide in large numbers through st 18 (Fig. 3B,C). BrdU labeling in the somatic subnuclei was heaviest when BrdU was delivered at st 17 (data not shown), suggesting that somatic motoneuron neurogenesis has begun by this stage and somewhat overlaps with the end of visceral motoneuron generation. By st 21, somatic oculomotor neurogenesis was largely confined to medial cells in the ventromedial subnucleus (data not shown) and was nearly complete by st 23 (Fig. 3D). Thus, there is a clear temporal difference in the birth of the oculomotor neuron subtypes in the chick embryo and this difference is part of a general gradient of OMC neurogenesis that extends from dorsolateral EW neurons to the ventromedial somatic subnuclei (Fig. 3A-D). Interestingly, the four-stage separation in neurogenesis completion for the visceral (st 19) versus somatic (st 23) motoneurons proved to be close to the five-stage separation in the onset of our visceral (VT) and somatic (CX36, TBX20) differentiation markers.

In the mature OMC, the visceral and somatic motor nuclei are segregated in the dorsal-ventral axis, which corresponds to the ventricular-pial axis of the embryonic ventral midbrain. We were interested in whether visceral and somatic motoneurons in their initial allocation to the prospective OMC prepatterned their eventual ventricular-pial segregation, or were at first fully intermixed. Our specific somatic motoneuron markers are not expressed until st 25. At st 25, TBX20-expressing somatic cells were already largely segregated from the VT-expressing EW cells, but there was some mixing of motoneuron subtypes (Fig. 4C; see Movie 1 in the supplementary material). Because of the inside-out generation of the OMC neurons, some TBX20-expressing cells in the prospective EW nucleus could be in transit, but there were also a few VT-expressing cells in the TBX20-positive territory. One stage later (st 26), nearly all VT-expressing neurons occupied a territory that was entirely separate from that of the TBX20-expressing somatic neurons, with a clear gap between the two populations (Fig. 4D).

To study the spatial allocations of OMC neurons before st 25, we employed VT (visceral) and ISL1 (visceral and somatic) gene expression. Between st 21 and 23, when motoneuron neurogenesis is still underway and somatic motoneurons must be transiting the VT territory, the prospective somatic (expressing ISL1 but not VT) and visceral (expressing both VT and ISL1) motoneurons were more intermixed along the ventricular-pial axis (Fig. 4A). By st 24, however, when motoneuron production is complete, there was a clear ISL1-rich territory that contained only a few cells expressing both VT and ISL1 (Fig. 4B). We conclude from these results that although the motoneuron subtypes are not fully segregated along the ventricular-pial axis during OMC neurogenesis and migration, by the end of motoneuron production the numbers of mislocated cells are small and these errors are rapidly corrected by remedial migration or cell death between st 24 and 26.

Previous work has shown that the signaling molecule sonic hedgehog can produce motoneurons in chick midbrain (Agarwala et al., 2001; Bayly et al., 2007; Wang et al., 1995). We were interested in the extent to which OMC transcription factors could, singly or in combination, induce OMC cell fates. PHOX2A was the clear choice for initial study: it is required for midbrain oculomotor neuron development in mouse and human (Nakano et al., 2001; Pattyn et al., 1997); in the chick embryo it is the first transcription factor we can identify that is selectively restricted to prospective OMC cells; and, most importantly, Dubreuil et al. (Dubreuil et al., 2000; Dubreuil et al., 2002) and Hirsch et al. (Hirsch et al., 2007) have shown that Phox2a/b can induce motoneuron markers, including ISL1 and CHAT, in chick spinal cord. We found that in chick midbrain, electroporation of full-length rat Phox2a (Arix) induced the early and late genes identified in Fig. 2, including motoneuron transcription factor, signaling and axon guidance genes (Fig. 5C-E; see Fig. S2 and Table S1 in the supplementary material). Induction of OMC markers was efficient, with 269/365 electroporated embryos demonstrating ectopic OMC gene expression. Similar efficiencies were seen in the induction of these markers by mouse Phox2b gene delivery (see Fig. S3 in the supplementary material).

To test for spatial restrictions to midbrain motoneuron induction, we performed two-color in situ hybridization on Phox2a-electroporated brains harvested at E6 and probed for ISL1 and the lateral midbrain markers PAX6 and EVX1 (Fig. 5E; see Fig. S4 in the supplementary material). These data establish that Phox2a induction of motoneuron markers extends across the rostral-caudal extent of the ventral midbrain, from the ventral midline at least as far lateral as the EVX1 stripe. We also found that Phox2a could induce ISL1 in the adjoining subthalamic region of forebrain (Fig. 5E; see Fig. S4 in the supplementary material). By contrast, we were unable to elicit motoneuron markers in dorsal midbrain following either Phox2a or Phox2b gene delivery (not shown). By morphology and developmental regulatory gene expression, the arcuate organization of the ventral midbrain extends into the adjoining subthalamic region of forebrain (Agarwala and Ragsdale, 2009; Sanders et al., 2002). Our Phox2a misexpression data indicate that the subthalamus also shares inductive properties with the ventral midbrain.

The neuropeptide hormone gene VT selectively labels visceral oculomotor neurons in the developing OMC. We found that Phox2a is able to induce this subset of oculomotor neurons (Fig. 5F). We were unable to detect convincing ectopic induction of the somatic oculomotor neuron marker CX36 owing to the extensive expression of endogenous CX36 laterally in the ventral midbrain. However, we were able to demonstrate that Phox2a can induce TBX20, our second marker for somatic motoneurons (Fig. 5G; see Table S1 in the supplementary material).

The mature OMC is organized with visceral motoneurons in the dorsally located EW nucleus and somatic motoneurons occupying distinct subnuclei ventrally (Fig. 1A). By E6, the visceral and somatic motoneuron markers in the native developing OMC identify two distinct populations of cells that are segregated along the ventricular-pial axis, with the VT-expressing visceral oculomotor neurons lying closer to the ventricle (Fig. 4D, Fig. 6B). We employed two-color detection to distinguish the VT-expressing and TBX20-expressing population subtypes in Phox2a-electroporated chick embryos. We identified, in the whole-mounts of four brains, 36 ectopic clusters of cells with both VT and TBX20 expression (Fig. 6A). The spatial organization of these clusters was studied in serial cross-section. Two of these clusters showed some intermixing of the VT-expressing and TBX20-expressing cells. The other 34 overlapping clusters presented full segregation of the VT-expressing and TBX20-expressing cells, with the VT-positive cell cluster invariably closer to the ventricular surface. In 26/34 clusters, the segregation featured a gap between the two populations (Fig. 6C,D).

In addition to this induction of ‘mini’-OMC clusters, we saw many isolated examples of ectopic cells (Fig. 6E,F, green arrows). In these examples, even when ectopic expression of only one of the two motoneuron class-specific markers was detected, the VT-expressing ectopic neurons appeared to lie closer to the ventricle than the TBX20-expressing ectopic neurons (Fig. 6F). To test this directly, we analyzed the arrangement of the isolated ectopic VT-expressing (n=111) and TBX20-expressing (n=43) cells along the ventricular-pial axis by measuring the distances from the base of the ventricular zone to each isolated cell or cell cluster. Because these distances were non-normally distributed (Fig. 7), we compared them with the two-sample Wilcoxon rank sum test. This analysis established that the isolated TBX20-expressing cells migrated to significantly greater tissue depths than the VT-expressing cells (mean, 87.2 versus 39.4 μm; median, 75.9 versus 32.9 μm; P<0.0001). We conclude that the spatial arrangement of the class-specific ectopic oculomotor neurons precisely follows that of the native motoneuron classes in the OMC.

The central feature of motoneurons is that their axons leave the nervous system and innervate peripheral targets. To study whether Phox2a misexpression can drive ectopic nerve production, we co-electroporated the rat Phox2a expression plasmid with an alkaline phosphatase expression vector to label cranial nerves. Phox2a overexpression generated ectopic nerve rootlets exiting from the midbrain, often several in each electroporated embryo. Table S2 in the supplementary material details the statistics on the behavior of Phox2a-induced rootlets for a sample of 19 electroporated embryos. These ectopic axons (Fig. 8) projected directly from ectopic spots within the ventral midbrain without aligning with the oculomotor nerve. Some projections stopped abruptly in the subjacent mesenchyme (n=16/49) (Fig. 8A), but many axons found their way to nearby cranial nerves III and IV, which carry oculomotor axons (III, n=14/49; IV, n=15/49) (Fig. 8B). The most striking observation was that four ectopic nerves were found to travel directly to the extraocular muscles of the orbit, bypassing the cranial nerves altogether (Fig. 8C). Finally, the timing of exit of these ectopic axons adhered closely to that of the native oculomotor nerve (st 16, n=3/7 tested). Thus, misexpressed Phox2a can produce axonal projections that exit the neuroepithelium on schedule and innervate natural targets.

Boundary cap cells (BCCs) are neural crest derivatives that accumulate at prospective nerve exit points (Vermeren et al., 2003; Yaneza et al., 2002). In the chick embryo, BCCs have been shown to express high levels of EGR2 (KROX20) and cadherin-7 (CDH7), and CDH7-positive BCCs in particular have been demonstrated at oculomotor axon exit points (Nakagawa and Takeichi, 1998; Niederlander and Lumsden, 1996). We asked whether these neural crest derivatives are also found at Phox2a-induced ectopic axon exit points. In control brains, we found EGR2 and CDH7 expression just outside the neuroepithelium, where native oculomotor axons exit (data not shown). EGR2 expression in the mesenchyme subjacent to the ventral midbrain neuroepithelium was most readily detected between E3 and E4, whereas CDH7 expression was present at high levels between E3 and E5 (Fig. 9). In Phox2a-electroporated heads, EGR2 and CDH7 expression was seen at ectopic locations, as well as at native oculomotor axon exit points (Fig. 9A,D). Cross-sections of experimental midbrains demonstrated that the CDH7-expressing cells populated the ectopic nerve exits (Fig. 9B,C). Our results establish that Phox2a induction of motor nerves constitutes a sufficient trigger for BCCs to populate axon exit points. Since ectopic exit points are unconstrained in their ventral midbrain distributions (not shown), the ectopic axons must release a BCC attractant (Bron et al., 2007; Mauti et al., 2007) or induce a BCC identity in local head crest cells.

PHOX2 transcription factors appear to be exceptionally effective in driving neuronal cell type identities. Dubreuil et al. (Dubreuil et al., 2000) have shown that exogenous Phox2b misexpression in chick spinal cord, a tissue in which PHOX2A/B are not normally expressed in motoneurons, can trigger motoneuron-specific gene expression and act as a neurogenic factor driving progenitor cells out of the cell cycle (Dubreuil et al., 2002; Pattyn et al., 2000). To explore the motoneuron inductive potential of PHOX2A/B in more detail, we studied the consequences of forced Phox2a/b expression in chick midbrain, where PHOX2A/B are natively expressed and restricted to motoneuron lineages. We found that Phox2a delivery can drive a full program of oculomotor neuron cellular and molecular fates, including the generation of a spatially organized nuclear complex. Nuclei are a central feature of vertebrate brain architecture, but little is known about the mechanisms of their production. These mechanisms will undoubtedly be multiple and vary across brain regions and cell types, but our findings with PHOX2A provide evidence for one type of mechanism that is arguably the simplest. Upstream signals drive the expression of PHOX2A in the progenitor tissue of arc 1 in ventral midbrain (Agarwala and Ragsdale, 2002; Agarwala et al., 2001). PHOX2A drives the production of both visceral and somatic motoneurons. The visceral and somatic motoneurons autonomously migrate to appropriate positions along the ventricular-pial axis, producing a spatially organized nuclear complex.

In BrdU birthdating studies, we found an inside-out pattern of neurogenesis in the chick OMC. Inside-out patterns of brain development have been most extensively studied in the formation of the layers of the mammalian cerebral cortex (Rakic, 1974). In an influential series of transplantation studies, McConnell and colleagues have demonstrated that the neocortical inside-out patterning is not a passive process, in which cells simply migrate past earlier-born cells before settling (Frantz and McConnell, 1996; McConnell, 1991). Rather, cortical progenitor cells acquire specific layer fate identity before they leave the ventricular zone of neuronal progenitors. These transplantation experiments were conducted with cortex donor tissue transplanted to cortex host tissue. Consequently, the execution of the laminar fate program was tested within native cerebral cortex, where the cues used by transplanted cells could be layer-intrinsic signals or molecules (McConnell, 1991). In our Phox2a-misexpression experiments, we have been able to drive ectopic production of isolated early-born (visceral motor) and late-born (somatic motor) neurons in lateral midbrain, that is, in neuroepithelium without a native OMC template. Thus, the ectopic singletons must employ intrinsic mechanisms, such as counting cells or measuring distances, as they migrate to their appropriate depths, or must draw on general cues that are deployed throughout ventral midbrain. A particularly attractive cueing mechanism would be a positional signaling system that supplies newborn neurons with positional information along the ventricular-pial axis. Such a system would be generally useful for arranging brain nuclei in three dimensions and, unlike a cell-counting mechanism, could accommodate brain growth. Our findings with native OMC development, particularly the (limited) intermixing of cell types seen at st 25, indicate that secondary correction mechanisms, such as cell sorting or deletion, contribute to OMC morphogenesis as well.

A striking finding in the current work is that Phox2a forced expression is able to drive the production of both visceral and somatic motoneurons broadly in ventral midbrain. The molecular mechanisms underlying the diversification of these motoneuron subtypes are unknown (Song and Pfaff, 2009). One possibility is that the ectopic cells take advantage of endogenous signaling systems, such as the NOTCH pathway, that mediate binary cell fate decisions (Mizuguchi et al., 2006; Peng et al., 2007). A second possible mechanism for somatic and visceral motoneuron diversification is the action of cell-autonomous temporal specification factors (Maurange et al., 2008; Zhu et al., 2006). A requirement for both these mechanisms is that they be resident throughout the territory competent for OMC induction or that they be triggered by Phox2a misexpression. A third possibility is that PHOX2A levels themselves directly specify visceral and somatic motoneuron fates. In vivo plasmid electroporation does not tightly control the gene dose delivered, and it is likely that different transgenic cells will express different levels of exogenous proteins. In studies of hematopoietic lineages, varying transcription factor levels have been found to regulate cell fate specification differentially (DeKoter and Singh, 2000). A similar scenario for PHOX2A in OMC cell fate determination would be that low (or high) levels of PHOX2A in the earliest progenitors drive visceral motoneuron production, and that progressive accumulation (or depletion) of PHOX2A protein leads to somatic motoneuron production. Testing this PHOX2A dose hypothesis is likely to entail the development of stem cell culture methods that permit tight regulation of PHOX2A protein delivery.

The extensive work on the specification of spinal motoneurons and hindbrain serotonin cells has emphasized the combinatorial nature of transcription factor regulation of neuronal identity (Cheng et al., 2003; Jessell, 2000; Novitch et al., 2001; Pattyn et al., 2004). The properties of PHOX2A in midbrain development, however, fit better with non-combinatorial models of cell type specification, such as those emerging from hematopoietic system studies, in which the networks of cell type control can be analyzed, at least in some cases, as hierarchies of primary and secondary regulators (Orkin, 2000; Singh, 2007). In this analysis, a regulator of a cell fate can be called primary if it (1) is required for that cell fate, (2) induces secondary regulators, such as other transcription factors, that are also required for that cell fate, (3) is activated at the time and site of the cell fate decision, and (4) is by itself sufficient to produce the cell fate. Our data, together with the extensive published work of others, establish PHOX2A is a primary regulator of oculomotor cell fates. First, genetic studies have demonstrated that PHOX2A is required for the production of midbrain motoneurons (Guo et al., 1999; Nakano et al., 2001; Pattyn et al., 1997). Second, the present study and those of others have shown that Phox2a/b induce the expression of secondary regulators of OMC cell fate, such as the transcription factor ISL1, which is required for all motoneuron development (Dubreuil et al., 2000; Pfaff et al., 1996). Third, BrdU birthdating and gene expression schedules in the chick (this study) and mouse (Pattyn et al., 1997) have established that PHOX2A is expressed in prospective oculomotor cells in the ventral midbrain just at the time when this cell fate decision is likely to be made. Fourth, this report demonstrates that forced expression of Phox2a is sufficient to produce ectopic midbrain motoneurons with the cellular and molecular features of OMC cell fates. It is possible that primary regulators of nuclear cell fate decisions will prove rare for vertebrate brain development. These results, though, establish that in at least one important example, a single transcription factor can act not only as a primary determinant of neuron identity, but also as a primary regulator of brain nucleus formation.

We thank May Tran and Melinda Drum for their help and Harinder Singh for discussions. Nucleic acid reagents were provided by V. Berthoud, M. Dixon, C. Goridis, M. Goulding, R. Ivell, T. Jessell, E. Laufer, E. Lewis, G. Martin, J. Rubenstein and M. Takeichi. This work was supported by a NEI/NIH grant to C.W.R. Deposited in PMC for release after 12 months.