Recent evidence suggests that in vertebrates the formation of distinct neuronal cell types is controlled by specific families of homeodomain transcription factors. Furthermore, the expression domains of a number of these genes correlates with functionally integrated neuronal populations. We have isolated two members of the divergent T-cell leukemia translocation (HOX11/Tlx) homeobox gene family from chick, Tlx-1 and Tlx-3, and show that they are expressed in differentiating neurons of both the peripheral and central nervous systems. In the peripheral nervous system, Tlx-1 and Tlx-3 are expressed in overlapping domains within the placodally derived components of a number of cranial sensory ganglia. Tlx-3, unlikeTlx-1, is also expressed in neural crest-derived dorsal root and sympathetic ganglia. In the CNS, both genes are expressed in longitudinal columns of neurons at specific dorsoventral levels of the hindbrain. Each column has distinct anterior and/or posterior limits that respect inter-rhombomeric boundaries. Tlx-3 is also expressed in D2 and D3 neurons of the spinal cord. Tlx-1and Tlx-3 expression patterns within the peripheral and central nervous systems suggest that Tlx proteins may be involved not only in the differentiation and/or survival of specific neuronal populations but also in the establishment of neuronal circuitry. Furthermore, by analogy with the LIM genes, Tlx family members potentially define sensory columns early within the developing hindbrain in a combinatorial manner.
Subdivisions of a given neural system arise by diversification of its component cell types and pathways. During development, precursors must be differentially specified so that each functionally distinct population can form connections with its correct afferent and efferent targets. A major challenge in understanding the ontogeny of neural systems is determining how this correlated specification of neurons and their connections is achieved.
In the developing chick cranial sensory system, axons from the cranial ganglia enter the hindbrain and then turn sharply to form a series of ascending and descending tracts in the lateral portion of the pons and medulla (Kuhlenbeck 1973). These incoming afferents connect to different pools of second-order sensory neurons arranged in longitudinal columns. The emergent nuclear structures form longitudinal arrays that transect the early segmental organization of the hindbrain (Lumsden and Keynes, 1989; Marı́n and Puelles, 1995). What factors organize this transformation from a segmental arrangement and, in particular, how do populations of hindbrain sensory neurons become specified to receive inputs from specific cranial ganglia?
Recent interest has focused on the role of families of transcription factors in defining distinct subsets of neurons. Differential expression of closely related POU-domain transcription factors (Turner et al., 1994) define different classes of retinal ganglion cells and somatosensory neurons (Xiang et al., 1995; Erkman et al., 1996). In the spinal cord (Tsuchida et al., 1994) and hindbrain (Varela-Echavarrı́a et al., 1996), the differential expression of LIM-domain genes confers a “code” for specific subsets of motor neurons (Tanabe and Jessell, 1996). Intriguingly, transcription factors such as DRG-11 (Saito et al., 1995) and Phox2 (Tiveron et al., 1996;Pattyn et al., 1997) may also define synaptic pathways within a functional neuronal circuit. This raises the possibility that these genes function upstream of cell surface molecules such as the cadherins (Suzuki et al., 1997), LAMP (Pimenta et al., 1995) and BEN (Chédotal et al., 1996), which potentially mediate neuronal recognition during the development of specific connections.
A candidate transcription factor family that may play a role in the development of cranial sensory pathways is the divergent homeobox-containing gene family Tlx (T-cell leukemia translocation, also known as Hox11). Tlx-1 was originally identified in humans because of a chromosomal translocation (Dube et al., 1991; Hatano et al., 1991; Kennedy et al., 1991; Lu et al., 1991). The corresponding mouse homolog has a very complex pattern of expression within the developing embryo (Raju et al., 1993; Dear et al., 1995; Roberts et al., 1995) and has been shown to play a critical role in the normal development of the spleen (Roberts et al., 1994;Dear et al., 1995). Two additional family members have been isolated,Hox11L1/Tlx-2 and Hox11L2/Tlx-3 (Cheng and Mak, 1993; Dear et al., 1993). All Tlx family members are expressed in the central and/or peripheral nervous system (Cheng and Mak, 1993; Dear et al., 1993, 1995; Raju et al., 1993; Roberts et al., 1995; Hatano et al., 1997).
In this study, we have isolated two members of the Tlx/Hox11gene family from chick, Tlx-1 and Tlx-3, and characterized their expression in neural tissue during early development. Tlx family members not only appear to identify synaptic pathways (analogous to DRG-11 and Phox-2) within the cranial sensory system but may also represent a code (as with LIM- and POU-domain factors) for the identity of specific subsets of sensory ganglion neurons and their associated central nuclei.
MATERIALS AND METHODS
Isolation and characterization of cDNA clones. A 630 bp HindII–ApaI human HOX11 cDNA fragment (Hatano et al., 1991) containing the homeobox was used to screen a chick stage 12–15 λZAPII cDNA library (kindly provided by Dr. Angela Nieto and Dr. David Wilkinson, National Institute for Medical Research, Mill Hill, London, UK). Approximately 2 × 106 recombinant cDNA clones were plated, transferred to a Hybond-N nylon membrane (Amersham International, Buckinghamshire, UK), and hybridized for 16 hr under high-stringency conditions [50% formamide, 5× saline–sodium phosphate–EDTA (SSPE), 2× Denhardt’s solution, 10% dextran sulfate, 150 μg/ml salmon sperm DNA, and 1% glycine]. Final washes were done using 0.5× SSPE and 0.1% SDS at 60°C. Positive clones were plaque-purified, and pSK plasmids were excised using Exassist helper phage (Stratagene, La Jolla, CA) according to the manufacturer’s instructions. Clones were restriction-mapped, and relevant fragments were subcloned into pBluescript SK (Stratagene) or M13 mp18/mp19 for sequencing. Single- and/or double-stranded DNA was sequenced by the dideoxy chain termination method using the Sequenase DNA sequencing kit (United States Biochemicals, Cleveland, OH) or using the Thermo Sequenase Dye Terminator cycle sequencing premix kit (Amersham International) with either commercially available or appropriate synthetic oligonucleotide primers. Ambiguous sequences were analyzed using the guanidine analog ITP. Sequences were determined for both strands, and all sequences were analyzed using the Geneworks sequence analysis program package (IntelliGenetics, Mountain View, CA).
In situ hybridization, immunohistochemistry, and sectioning.Whole-mount in situ hybridization of intact embryos, dissected hindbrains, or dissociated cells was performed using nonradioactive digoxygenin-labeled RNA probes essentially as described by Wilkinson (1992). Embryos were staged according to the method ofHamburger and Hamilton (1951).
For Tlx-1 (GenBank accession number AF071874), an ∼800 bp probe entirely contained within the 3′ untranslated region was made by cutting the cDNA internally with PstI and transcribing with T7. For Tlx-3 (GenBank number AF071875), an ∼500 bpSstII fragment containing the homeobox was subcloned into pBluescript SK, the resulting plasmid digested with BglII and transcribed using T7. A second, more divergent 300 bpSmaI fragment located 5′ of the homeobox was used to confirm specificity of the Tlx-3 probe. Specimens were refixed using 3.5% paraformaldehyde in PBS and/or in methanol/DMSO (4:1) before immunohistochemistry, sectioning, and/or storing. In addition, for BrdU labeling, dissected hindbrains were treated with 2N hydrochloric acid for 1 hr at 37°C.
Isl-1/2 protein was detected by whole-mount immunohistochemistry using a 1:200 dilution of the rabbit polyclonal αIsl-1/2 antibody (kindly provided by Dr. T. Edlund, Department of Microbiology, University of Umea, Umea, Sweden) as per the method described by Davis et al. (1991). Disassociated cells and BrdU-labeled, dissected hindbrains were immunostained as described previously by Lumsden and Keynes (1989). BrdU incorporation was revealed using a specific monoclonal antibody (Bio-Science Products AG, Emmenbrücke, Switzerland) at a 1:10 dilution. The mouse monoclonal anti-β-tubulin antibody (Sigma, St. Louis, MO), the anti-160 kDa neurofilament antibody RMO-270 (Zymed, San Francisco, CA), and the anti r5-vimentin antibody (kindly provided by Dr. Ursula Dräger, E. K. Shriver Center, Waltham, MA) were used at 1:200, 1:2000, and 1:100 dilutions, respectively. HRP- or fluorescein-conjugated secondary antibodies to rat or mouse IgG and IgM (Jackson ImmunoResearch, West Grove, PA) were used where appropriate, at a 1:200 dilution.
After whole-mount in situ hybridization and/or immunohistochemistry, selected embryos were embedded in a mixture of gelatin and albumin, and 25–50 μm transverse sections were cut using a vibratome. Sections were then cleared using 90% glycerol in PBS and mounted under coverslips for photography. BrdU-labeled hindbrains were similarly embedded and sectioned either coronally or parasagitally, and fluorescent cell nuclei were visualized by confocal microscopy (MRC 600; Bio-Rad, Welwyn Garden City, UK).
Labeling of S-phase nuclei with BrdU. Hens’ eggs were incubated at 37°C in a humidified incubator. To prevent blood vessels forming over the roof of the shell, eggs were windowed at embryonic day 3 (E3) using a sharp pair of scissors and immediately resealed with heavy insulation tape (“tesa”; Beiersdorf AG, Hamburg, Germany). At E5, the window was opened again, and 6 μl of bromodeoxyuridine (BrdU; 15 mg/ml in Howard’s Ringer’s solution) was injected via a glass micropipette into a blood vessel of the chorioallantoic membrane. Eggs were resealed and incubated at 37°C for either 30 min or 24 hr. After fixation in 3.5% paraformaldehyde in PBS, embryos were processed sequentially for in situ hybridization and immunohistochemistry.
Dissociated hindbrain cell cultures and intact hindbrain explants. Embryos were removed from the egg at 3–5 d of incubation, dissected free from surrounding membranes, and treated for 20–30 min with 1 mg/ml dispase in L-15 media. After rinsing several times with Howard’s Ringer’s solution, hindbrains [from rhombomeres 2 (r2) to r7 inclusive] were dissected free from surrounding tissue and washed two or three times with Ca2+- and Mg2+-free HBSS containing 0.02% w/v EDTA (dissociation buffer). Dissected tissue was then incubated at 37°C for 15–20 min in fresh dissociation buffer containing between 5 and 50 μg/ml trypsin. After incubation, the tissue was mechanically disassociated using a fire-polished Pasteur pipette coated with serum and pelleted for 5 min at low speed, and cells were resuspended in F-12 and SATO media (Bottenstein and Sato, 1979). Disassociated cells were rinsed twice with F-12 and SATO, plated onto poly-l-lysine- and laminin-coated plates, and placed in a 5% CO2incubator at 37°C in F-12, SATO, and 10% heat-inactivated fetal calf serum for 20–24 hr. Cells were then fixed in 3.5% paraformaldehyde and processed sequentially for in situ hybridization and immunohistochemistry.
Hindbrains (from r1 to r7 inclusive) were similarly dissected from stage 16–17 (E3) chick embryos and embedded in collagen and DMEM. In some animals (n = 4), the fifth and seventh/eighth ganglia were removed. Hindbrain explants were bathed in modified F-12 and DMEM supplemented with chick embryo extract and antibiotics per the method of Book and Morest (1990) and incubated for 2 d in a 5% CO2 incubator at 37°C. Explants were then fixed in 3.5% paraformaldehyde and processed for in situhybridization.
Axon tracing. The fifth and/or seventh/eighth cranial sensory nerves were anterogradely labeled in live embryos using lysinated rhodamine–dextran (Molecular Probes, Eugene, OR). Briefly, hindbrains from stage 23 (E4) embryos were rapidly dissected in Howard’s Ringer’s solution, and the ganglion was cut at its root. Dextran (made up as a thick paste) was applied directly to the stump of the cranial nerve, and the isolated hindbrain (n = 7) was incubated at room temperature for 3 hr in Howard’s Ringer’s solution before fixation in 3.5% paraformaldehyde and subsequent processing for in situ hybridization. Alternatively, DiI (Molecular Probes) dissolved in dimethylformamide (6 mg/ml) was used to similarly label sensory axons in paraformaldehyde (3.5%)-fixed embryos. For labeling of the lateral longitudinal tract, DiI was injected into the roots of the fifth and/or seventh/eighth nerves of intact stage 29 (E6) embryos or into the caudal optic tectum of embryos at stage 19. Embryos were stored for up to 14 d at room temperature in 3.5% paraformaldehyde before dissection of the hindbrains for analysis. Dextran and DiI fluorescence was visualized by confocal microscopy (MRC 600; Bio-Rad).
Isolation and characterization of chickHox11/Tlx homologs
A total of five cDNA clones ranging in length from 1.7 to 2.2 kb were isolated by screening a chick stage 12–15 λZAPII cDNA library at high stringency using the human HOX11homeobox fragment (Hatano et al., 1991) as a probe. Restriction mapping and subsequent sequence analysis revealed that these clones, which we designate Tlx-1 and Tlx-3, corresponded to two different members of the previously identified Hox11/Tlxgene family (Cheng and Mak, 1993; Dear et al., 1993). Four of the five clones overlapped and were most closely related to Hox11(also known as Tlx-1), whereas the fifth was most closely related to Hox11L2 (also known as Tlx-3). A comparison of the predicted amino acid sequences for the two chick genes and their previously identified human (Hatano et al., 1991) and/or mouse (Dear et al., 1993) homologs is shown in Figure1. Overall, the chick Tlx-1 homolog was found to share greater amino acid sequence identity with the human and mouse Hox11/Tlx-1 proteins (81% for each, respectively) than it does with the mouse Hox11L1 (63%) or Hox11L2 (64%) proteins, not including deletions and insertions. Similarly, the chick Tlx-3 homolog was 85% identical to Hox11L2 overall, although it shares only 64–70% identity with the human and/or mouse Hox11/Tlx-1 and Hox11L1/Tlx-2 proteins, respectively. Furthermore, the 60 amino acid homeodomain of chick Tlx-1 and Tlx-3 are 100% identical to that found in their respective mouse and/or human homologs. Notably, the two chick homeodomains, like other members of this gene family, both contain a threonine at position 47. In both Tlx-1 and Tlx-3 proteins, there also exists considerable sequence similarity across species in the regions immediately surrounding the homeodomain.
Previous studies have shown that members of the Hox11/Tlxgene family share four other highly conserved motifs designated TH1–TH4 (Cheng and Mak, 1993) in addition to the highly conserved homeodomain. Two of the four motifs, TH1 and TH2, lie N-terminal of the homeodomain and are 100% identical or have a single conservative amino acid substitution, respectively, in the chick Tlx proteins. Similarly, the TH3 and TH4 motifs, which lie C-terminal of the homeodomain, are 100% identical.
Embryonic expression of Tlx-1and Tlx-3
We have examined the spatial and temporal pattern ofTlx-1 and Tlx-3 mRNA expression during early chick embryogenesis by in situ hybridization. With some notable exceptions, especially involving the nervous system, theTlx expression data presented here are in agreement with those described previously in mouse (Cheng and Mak, 1993; Raju et al., 1993; Roberts et al., 1994, 1995; Dear et al., 1995). In particular,Tlx-1 is expressed early in the developing branchial arches, subsequently in the sensory components of the fifth and eighth cranial nerves and later in longitudinal stripes in the hindbrain. In addition, expression is observed in the thoracic region, in the presumptive pharynx, and in a portion of the heart (data not shown). In the gut region, there is prominent expression from stage 20 onward in mesenchymal cells of the dorsal mesentery that corresponds to the splenic primordium (data not shown).
In contrast, Tlx-3 expression is restricted to neuronal populations within the peripheral and central nervous systems. In the peripheral nervous system, Tlx-3 is expressed in the placode-derived components of several cranial sensory ganglia and in the neural crest-derived dorsal root and symphathetic ganglia. In the CNS, Tlx-3 is expressed in two longitudinal columns of neurons at distinct dorsoventral levels of the hindbrain, as well as in D2 and D3 neurons within the spinal cord.
Tlx-1 is expressed early in the developing branchial arches
Tlx-1 transcripts are first detected at the three-somite stage (stage 8−) in the definitive pharnygeal endoderm. Slightly later, beginning at 5 somites (stage 8+), endoderm of the anterior intestinal portal also begins to express Tlx-1 (Fig.2 A,D). At stage 11 (13 somites), both the endoderm as well as mesenchyme expressTlx-1 (Fig. 2 E). As the branchial arches develop, Tlx-1 expression becomes restricted primarily to the central mesenchyme and ventral medial endoderm and ectoderm (Fig.2 F,G). Expression persists within the developing branchial arches until at least stage 25, the latest stage examined. However, by this stage expression is barely detectable within the mandibular component of the first branchial arch. Expression was not detected at any stage in the maxillary component of the first branchial arch.
Tlx-1 and Tlx-3 are expressed in overlapping domains within the sensory components of the fifth and eighth cranial nerves
At stage 15, the ganglia of the fifth (trigeminal) and eighth (vestibuloacoustic) cranial nerves begin to express bothTlx-1 and Tlx-3 (Fig.3 A; data not shown). AlthoughTlx-3 expression is initially restricted to the ophthalmic lobe of the trigeminal ganglion (Fig. 3 A), it soon extends into the maxillomandibular lobe where it overlaps with that ofTlx-1. By stage 17 (Figs. 2 C, 3 B), both genes are strongly expressed in overlapping domains within the ganglia of the vestibuloacoustic nerve (eighth). Overlapping expression of Tlx-1 and Tlx-3 is also seen within the ventrolateral portion of the maxillomandibular lobe of the trigeminal ganglion. At this stage, weak Tlx-1 expression also becomes apparent and overlaps with that of Tlx-3 within the ophthalmic lobe. Tlx-3, unlike Tlx-1 is also expressed in the distal, placode-derived geniculate, petrosal, and nodose ganglia of the 7th, 9th, and 10th cranial nerves, respectively (Fig. 3 B,C; see Fig. 7). The rostrocaudal order of onset ofTlx-3 expression within these ganglia is slightly delayed with respect to the fifth and eighth ganglia.
To assess the neuronal phenotype of Tlx-1- andTlx-3-expressing cells within cranial sensory ganglia, stage 21–25 embryos were processed sequentially viain situ hybridization using Tlx-1- orTlx-3-specific mRNA probes and immunohistochemically using the αIsl-1/2 antibody; a marker that labels neurons within cranial sensory ganglia. Tlx-1 and Tlx-3 transcripts were found to colocalize with Isl-1/2 protein (see Fig. 6 Fand data not shown). Furthermore, at later developmental stages (Fig.2 H; see Fig. 6 F), their expression within the maxillomandibular lobe of the trigeminal ganglia is clearly restricted to the larger, more distally located, placodally derived neurons.
By stage 25, Tlx-1 expression within the trigeminal ganglion is greatly reduced. However, expression within the vestibuloacoustic ganglia remains strong until at least stage 28 (data not shown).Tlx-3 expression is significantly reduced in the seventh, eighth, and ninth ganglia by stage 25.
Tlx-1 is expressed in parallel, bilateral stripes running rostrocaudally at distinct dorsoventral levels within the developing hindbrain
Tlx-1 is expressed in the hindbrain from stage 20 (E3.5) until at least stage 35 (E9), the latest stage examined. Figure4 A–D shows the dorsal view of dissected and flat-mounted stage 20–29 hindbrains in which the roof plate of the fourth ventricle has been removed. Initially,Tlx-1 expression is for the most part restricted dorsoventrally and anteroposteriorly to a single, bilateral stripe that runs caudally from the boundary between r1 and r2 into the cervical spinal cord. The stripe itself reflects the underlying segmental organization of the hindbrain (for review, see Lumsden, 1990). It is staggered at each successive rhombomere boundary, and there are transient rhombomere-specific elaborations of the Tlx-1expression pattern. For example, in Figure 4, a comparison betweenA (stage 20) and B (stage 23) reveals that the onset of Tlx-1 is delayed in r5, whereas in r6 there is initially a second parallel and more dorsal stripe of Tlx-1expression. Similarly, the stripe of expression within r3 begins to widen dorsally at stage 23 (Fig. 4 B), becomes more pronounced by stage 25 (Fig. 4 C), and by stage 29 (Fig.4 D, see Fig. 7) has resolved into a second stripe—a pattern evocative of medial to lateral cell migration. DiffuseTlx-1 staining extending from the caudal end of the hindbrain through the tail region is also seen in the developing spinal cord (data not shown).
Coronal sections at stages 21, 23, and 25 (Figs.4 E–G) show that Tlx-1-positive cells are located at the distal (pial) edge of the ventricular zone at the interface between nuclei of dividing cells and the postmitotic cells of the mantle. At these stages, however, the mantle is relatively thin. As an increasing number of cell bodies are deposited, the stripe ofTlx-1 expression segregates into parallel pial and ventricular components. By stage 29–31, Tlx-1 expression is clearly organized into complementary stripes at different depths within the hindbrain (Fig. 4 D,H). The more discrete ventricular stripe retains a segmental organization, despite the fact that the overtly segmented structure of the hindbrain disappears after stage 24–25. The pial stripe of expression extends anteroposteriorly from approximately the r2–r3 boundary to the r6–r7 boundary (see Fig.7). Interestingly, from stages 25 to 28, a curtain of label stretches between the ventricular and pial stripes, again highly evocative of cell migration. Within the pial stripe of expression, two foci of concentrated label are apparent at the level of r3 and r6.
By stage 34, both the ventricular and pial stripes of Tlx-1expression are significantly decreased. However, the two foci of label within the pial stripe remain distinct, and by stage 35, faintTlx-1 expression can be found only in these more pial foci (data not shown).
Tlx-1-expressing cells within the hindbrain are postmitotic neurons
To assess the phenotype of Tlx-1-positive cells within the hindbrain, short-term cultures of dissociated hindbrain cells from stage 21–25 embryos were processed sequentially via in situhybridization using a Tlx-1-specific mRNA probe and immunohistochemically using specific neuronal or glial markers. As shown in Figure 5 A, allTlx-1-positive cells coexpressed β-tubulin protein, an early neuronal marker. Similarly, most Tlx-1-expressing cells were positive for a second neuronal-specific antigen, RMO-270 (Fig. 5 B), which identifies the 160 kDa neurofilament subunit of differentiated neurons. In contrast, no cells were double-labeled in cultures similarly processed using the anti-vimentin antibody R5 (data not shown), a marker of radial glial cells (Dräger et al., 1984; Heyman et al., 1995).
Although the later pial stripe of Tlx-1 expression at stage 29 undoubtedly consists of postmitotic cells, it was unclear whether the ventricular stripe of Tlx-1 expression is located within the germinal zone or whether it lies just outside the layer of mitotic cells. To assess whether Tlx-1-expressing cells within the hindbrain are mitotically active, we used BrdU to label cells in S-phase. Thirty minutes after BrdU treatment, it is clear thatTlx-1-positive cells do not incorporate BrdU but lie just outside the proliferative, ventricular zone (Fig. 5 C). By comparison, Figure 5 D shows the pattern of double labeling in a parasagittal section of a Tlx-1-expressing ventricular stripe in a stage 29 embryo 24 hr after a similar injection of BrdU. During this period some BrdU-labeled cells will have become postmitotic and migrated out of the ventricular zone. A number of postmitoticTlx-1-positive cells have incorporated BrdU, indicating that they have divided within the last 24 hr. This suggests thatTlx-1 expression within the hindbrain can be initiated in postmitotic neurons soon after they have undergone their final division.
Tlx-1-positive neurons are closely associated with the lateral longitudinal axonal tract and do not colocalize with Isl-1/2 protein
Primary afferent axons derived from progressively more caudal cranial sensory ganglia course in successively more lateral positions within the lateral longitudinal tract of the hindbrain (Clarke and Lumsden, 1993). To examine the spatial relationship between theTlx-1-positive stripe of expression within the hindbrain and that of incoming axons, fluorescent rhodamine–dextran was used to label cranial nerve roots before in situ hybridization using a Tlx-1-specific probe. Figure 5 E shows a flat-mounted stage 24 hindbrain revealing the close spatial relationship between Tlx-1-positive cells (blue) and sensory axons of the seventh/eighth cranial nerve (red). Incoming sensory axons from the fifth cranial nerve would fill the gap between the seventh/eighth axon tract and the stripe ofTlx-1 expression. Hence, the column ofTlx-1-positive neurons lies just medial to the lateral longitudinal tract (Fig. 5 F). Interestingly, DiI labeling from the fifth cranial nerve also reveals that the incoming afferents project numerous collaterals medially toward the position of the Tlx-1 positive stripe (Fig.5 G).
Previous expression studies in mouse have suggested thatTlx-1 marks branchial motor nuclei within the developing hindbrain (Roberts et al., 1994, 1995). To directly determine whether chick Tlx-1 is expressed in motor nuclei, hindbrains from stage 29–31 embryos were processed sequentially via in situhybridization using a Tlx-1-specific probe and immunohistochemically using an αIsl-1/2 antigen that specifically labels all cranial motor nuclei within the developing hindbrain (Varela-Echavarrı́a et al., 1996). As shown in Figure6 A–D, Isl-1/2 protein does not colocalize with Tlx-1 mRNA transcripts.
Tlx-1 expression within the hindbrain is cell-autonomous
Studies in both invertebrates (Macagno, 1979; Baptista et al., 1990; Selleck and Steller, 1991) and, more recently, vertebrates (Gong and Shipley, 1995) have shown that peripheral axons may be involved in the specification of neurons within the developing CNS. Both the onset of Tlx-1 expression (initially peripheral and then progressively more central) and the position ofTlx-1-positive hindbrain neurons just medial to the lateral longitudinal axonal tract suggest a correlation with the timing and trajectory of sensory axon innervation (Moody et al., 1989;Chédotal et al., 1995). Axons invade the hindbrain before the onset of Tlx-1 expression, and we sought to investigate whether fibers from the cranial ganglia might induce or potentiate expression of Tlx-1 within the hindbrain. We tried a variety of surgical ablation and barrier approaches in ovo before deciding that the regenerative capacity and multiple origins of axons in the tract (from both cranial ganglia and midbrain) preclude a conclusive exclusion of axon ingrowth by these means. We therefore tested the ability of both cultured hindbrains deprived of their cranial ganglia and dissociated hindbrain cells to autonomously expressTlx-1. Hindbrains explanted before the onset ofTlx-1 expression with their cranial ganglia excised as well as dissociated hindbrain cells from similarly staged embryos; cultured for 1–2 d, both expressed Tlx-1 mRNA (data not shown). These findings suggest that the onset of Tlx-1 expression in the hindbrain is independent of afferent innervation and is cell-autonomous. Furthermore, dissociated cell cultures from stage 21–25 hindbrains (after the onset of Tlx-1 expression) also expressed Tlx-1 (Fig. 5 A,B), suggesting that peripheral nerve innervation is not required for the maintenance ofTlx-1 expression in central neurons.
Tlx-3 is also expressed in parallel, bilateral stripes running rostrocaudally at distinct dorsoventral levels within the developing hindbrain
Tlx-3 transcripts are first detected in the hindbrain at stage 15. As shown in Figure 4 I–L, expression for the most part is contained within two longitudinal columns of cells at specific dorsoventral levels. The more dorsally located column appears first and has a distinct rostral limit at the r3–r4 boundary (Figs.4 I, 7). The second, more medial column overlaps both spatially and temporally with that of Tlx-1 (Fig.4 J,K; data not shown). However, rostrally it is much narrower and unlike Tlx-1 continues into r1 where it expands dorsally. This expanded rostral expression of Tlx-3correlates precisely with a pronounced dorsal shift in the trajectory of axons and broadening of the lateral longitudinal tract within r1 (Fig. 5 H). Caudally, both columns appear to continue throughout the entire length of the spinal cord (summarized in Fig.7).
As for Tlx-1, there are transient rhombomere-specific elaborations of the Tlx-3 expression pattern. For example, a comparison between Figures 3 B and 4 Ireveals that the onset of Tlx-3 expression within the dorsal-most column is delayed in odd-numbered rhombomeres. In addition, this column is initially expanded ventrally in both r4 and r6 (Fig.4 I,J). By stage 23, however, this ventral expansion is no longer visible (Fig. 4 K; data not shown); a pattern that suggests medial to lateral cell migration similar to that noted for Tlx-1 (see above).
Coronal sections at stages 18, 21, and 25 (Figs. 4 M–O), show that Tlx-3-positive cells within each column are located at the distal (pial) edge of the ventricular zone; a position similar to that found for Tlx-1, suggesting thatTlx-3 like Tlx-1 is expressed in postmitotic neurons. At later developmental stages, the Tlx-3 expression pattern becomes much more complex. Expression of the medially located cell column is still apparent within the subventricular zone at stage 30, but the more dorsally located cell column appears to have diffused both ventrally and mediolaterally within the expanding mantle layer and now at least in part has a distinct caudal limit within the posterior hindbrain (Fig. 4 L). At certain anteroposterior levels, a curtain of label can also be seen stretching from the subventricular layer through the expanding mantle layer at the same dorsoventral level as the more medially located cell column (Fig.4 P).
Tlx-3 is also expressed in D2 and D3 neurons in the spinal cord
Tlx-3 expression was also found to extend throughout the length of the spinal cord in two bilateral columns of dorsally located cells (Fig. 3 G), which appear, at least initially, to be continuous with the Tlx-3-positive cell columns within the hindbrain (Fig. 4 J,K). Interestingly, Liem et al. (1997) have recently defined similarly positioned columns of neuronal cells based on their expression of specific LIM homeodomain proteins. In particular, D1 neurons, which occupy the most dorsal position, express LH2 protein, whereas D2 and D3 neurons, located in progressively more ventral positions, express Isl-1 or LIM 1/2 protein, respectively. Colocalization of Tlx-3 transcripts with Isl-1/2 protein (Fig. 6 E) confirmed that the more dorsally located column of Tlx-3-expressing cells marks D2 neurons, while the more ventrally located Tlx-3-positive, Isl-1/2-negative column of cells potentially marks D3 neurons, which are known to express LIM1/2 but not Isl-1 protein (Liem et al., 1997).
Tlx-3 is also expressed throughout the mesencephalon and posterior diencephalon in scattered cells immediately adjacent to the dorsal midline (Fig. 3 H).
We have isolated and characterized two members of the divergent homeobox-containing HOX11/Tlx gene family from chick,Tlx-1 and Tlx-3, and have shown that they are expressed in neurons in both the peripheral and central nervous systems. In the peripheral nervous system, Tlx-1 and/orTlx-3 are expressed in overlapping domains in the placodally derived components of the 5th, 7th, 8th, 9th, and 10th cranial sensory ganglia. Tlx-3, unlike Tlx-1, is also expressed in neural crest-derived dorsal root and sympathetic ganglia. In the developing hindbrain, both genes are expressed early in bilateral columns of neurons at distinct dorsoventral levels and may delineate second order sensory nuclei. Tlx-3 is also expressed throughout the spinal cord in dorsal D2 and D3 neurons and along the midline of the dorsal mesencephalon and posterior diencephalon. The expression patterns of Tlx-1 and Tlx-3 in the peripheral and central nervous systems suggests that these genes might be involved in the differentiation and/or survival of distinct neuronal cell types. Furthermore, Tlx family members define developing cranial ganglia and hindbrain sensory columns in a combinatorial manner that suggests a role in establishing appropriate connectivity.
In the peripheral nervous system, both Tlx-1 andTlx-3 are transiently expressed by neurons of the developing cranial sensory ganglia. The localization of in situlabeling together with the timing of its onset and downregulation strongly suggest that only placode-derived (as opposed to neural crest-derived) neurons express Tlx-1 and/orTlx-3. Fate-mapping studies have shown that placode-derived neurons associated with the 5th, 7th, 9th, and 10th cranial nerves are located distal to crest-derived neurons and, with the exception of the fifth cranial nerve, their cell bodies lie within separate ganglia (D’Amico-Martel and Noden, 1983; Noden, 1992). Neurons associated with the eighth cranial nerve are almost entirely derived from placodal ectoderm. Correspondingly, Tlx-positive neurons are found in the geniculate, petrosal, and nodose ganglia of the 7th, 9th, and 10th cranial nerves, respectively, in the ventrolateral portion of the trigeminal (fifth) ganglion, and throughout the whole of the vestibuloacoustic (eighth) ganglion. Furthermore, birth-dating studies have shown that cranial neural crest-derived neurons develop later than placodal neurons (D’Amico-Martel, 1982; Covell and Noden, 1989) and are still being born after Tlx genes have been downregulated at E4–E5. Interestingly, in the eighth ganglion, which contains the only placodal neurons to form later than day 5, Tlx-1expression is protracted. Overall, the spatial and temporal onset ofTlx-1 and Tlx-3 expression correlates well with the predominantly rostrocaudal gradient of maturation of placodal derivatives within cranial ganglia (D’Amico-Martel, 1982).
Placodal sensory neurons are uniquely placed to play a role in pathway specification. Because they develop axons before crest derivatives, it has been suggested that they may pioneer sensory innervation of the hindbrain (Moody et al., 1989; Noden, 1992). Therefore the expression of Tlx family homeobox genes in subsets of placodal neurons has considerable significance. In this context, the complex expression patterns of Tlx-1 and Tlx-3 in the hindbrain show a remarkable correlation with the predicted synaptic targets of the corresponding sensory neurons in the periphery (summarized in Fig. 7). In the following discussion, cranial ganglia and their appropriate recipient nuclei are considered in terms of Tlx gene expression. In defining hindbrain sensory columns, subventricular gene expression has been interpreted as identifying recently born, immature neurons (Fig. 5 D), which have yet to undergo their characteristic radial migration (Hemond and Glover, 1993; Wingate and Lumsden, 1996). The corresponding mantle expression was viewed as characterizing emergent sensory nuclei that were identified on the basis of their segmental origin (Marı́n and Puelles, 1995), mediolateral position, and distance from the ventricular surface.
Afferent axons from each cranial sensory ganglion form discrete parallel axon bundles in the lateral longitudinal tract (Clarke and Lumsden, 1993). Their axons subsequently develop collaterals (Fig.5 F), which synapse on sensory nuclei lying within three principal columns: the cochlear, vestibular, and trigeminal nuclear complexes. Most laterally, the cochlear nuclei (nucleus angularis, nucleus laminaris, and nucleus magnocellularis) receive input from the acoustic portion of the eighth nerve. These second-order sensory neurons are born at the rhombic lip of the fourth ventricle from E5.5 to E6.5 (Harkmark, 1954) and only subsequently migrate into the mantle. They do not appear to be labeled by either Tlx-1or Tlx-3. More medially, the vestibular nucleus receives input from the eighth ganglion (which colabels with bothTlx-1 and Tlx-3). Although we were not able to unequivocally locate gene expression to the various cytoarchitecturally distinct subcomponents of this complex (Wold, 1976; Glover and Petursdottir, 1988; Glover, 1993), Tlx-1 andTlx-3 have a complementary distribution that might reflect different functional populations (summarized in Fig. 7).Tlx-1 labels a rostral column restricted to r3, which may represent a caudal subdivision of cell group A (Wold, 1975;Marı́n and Puelles, 1995). By contrast, a nonoverlapping domain of Tlx-3 expression labels cells within the remainder of the vestibular column caudally from the r3–r4 boundary into the spinal cord where it merges with the column of D2 spinal neurons. At stage 20, there is a transient broadening of Tlx-3 expression in r4 and r6 (Fig. 4 J) and a transient expression of a lateral stripe of Tlx-1 in r6 (Fig. 4 A). Segment-specific developmental program may correlate with the organization of the vestibular neurons into spatially segregated clusters with different axon trajectories (Glover, 1993).
The most medial sensory recipient column is composed of the trigeminal sensory nucleus, which receives its principal input from the fifth ganglion (both Tlx-1- and Tlx-3-positive), and the solitary nucleus, which has contributions from the 7th, 9th, and 10th ganglia (all Tlx-3-positive). Beginning at stage 25, the emergent descending trigeminal spinal nucleus labels withTlx-1 within a precisely defined region stretching from the r2–r3 to the r6–r7 boundaries. Segmental swellings (Fig.6 B) are consistent with origins and development of this nucleus as described in its fate map (Marı́n and Puelles, 1995) and may also reflect its well characterized topographical divisions into at least three subnuclei. There are prominent discontinuities in the domain of Tlx-1 expression at the level of the r3–r4, and r5–r6 boundaries (Fig. 6 B). These appear to match the caudal limits of the trigeminal nucleus oralis and nucleus interpolaris respectively (Dubbeldam and Karten, 1978). Interestingly, both Tlx-1 and Tlx-3 are expressed in overlapping domains within the subventricular layer (Fig.4 D,L). Intermingling of Tlx-1 andTlx-3 label within the trigeminal spinal nucleus itself might be expected given that the latter also receives innervation from more caudal ganglia (Dubbeldam and Karten, 1978; Dubbeldam et al., 1979). By contrast, only Tlx-3-positive neurons are found rostrally (Fig. 4 L) in the principal trigeminal nucleus that lies within r1 (Marı́n and Puelles, 1995). Thus, whereas Tlx-1 primarily labels cells within the descending trigeminal spinal nucleus, Tlx-3 labels cells within the ascending principal trigeminal nucleus. Therefore, within the ophthalmic lobe (primarily Tlx-3-positive) and maxillomandibular lobe (both Tlx-1- andTlx-3-positive) of the fifth ganglion, the pattern ofTlx labeling might identify primary sensory neurons with different (ascending vs descending) axonal projections. Such a hypothesis predicts that within the colabeled maxillomandibular lobe, individual sensory neurons that express only Tlx-1contribute only descending (small-caliber) axons to the nucleus, whereas sensory cells with bifurcating, ascending, and descending axons (Windle, 1926) should express Tlx-3 andTlx-1.
The solitary nucleus overlies and intermingles with the descending trigeminal nucleus and has a distinct adaptive structure in birds (Kappers et al., 1960). The nucleus is cytoarchitecturally distinct caudal of the r6–r7 boundary (Marı́n and Puelles, 1995) and extends only a thin elongation into rostral hindbrain (Dubbeldam et al., 1979). This reflects the poorly developed gustatory system in birds and the diminished input from the seventh nerve (Kappers et al., 1960). The pattern of Tlx-3 labeling in caudal hindbrain seems to correspond to the structure and location of the solitary nucleus. Accordingly, the ganglia (7, 9, and 10) that project to this nucleus express Tlx-3 but not Tlx-1 (Fig. 7). In the spinal cord, Tlx-3 also labels D3 neurons suggesting a molecular homology between the trigeminal and solitary nuclear complex and the functionally equivalent cutaneous sensory cells in the dorsal horn.
These patterns of expression point to a functional role for Tlx proteins in establishing sensory innervation of the hindbrain. If cranial sensory innervation is orchestrated by the coordinated expression of different Tlx genes, both their regulation and function become issues of great developmental significance. In particular, how appropriate placodal expression is matched with the generation of recipient sensory neurons within the hindbrain is a central question. Interestingly, cell type-specific enhancer elements directing the expression of Tlx-1 within the peripheral nervous system may already have been located. In disrupting theTlx-1 gene, Dear et al. (1995) replaced part of the first exon and most of the first intron with an MC1–neomycin cassette and the β-galactosidase (β-gal) gene. The resulting β-gal fusion protein is expressed in developing branchial arches, CNS, and spleen but, significantly, not within the developing cranial sensory ganglia. Therefore, tissue-specific enhancers for the expression ofTlx-1 in placodally derived neurons probably lie within the first intron. Clues to the regulation of Tlx gene expression within the CNS come from the pattern of expression of Tlx-3in the developing spinal cord. Indeed, the homology between the vestibular and trigeminal sensory columns and D2 and D3 spinal neurons revealed by Tlx-3 expression raises the possibility that their development is influenced by the same factors. Recent studies of dorsal spinal cord development have shown that members of the TGF-β superfamily induce D2 and D3 interneuron differentiation (Liem et al., 1997). Therefore, the onset of Tlx-3 and possibly otherTlx family members in the hindbrain might also be regulated by the same set of roof plate-derived dorsalizing factors.
In terms of function, all Tlx proteins contain a divergent homeodomain that at least for Tlx-1 has been shown to mediate transcriptional activity (Dear et al., 1993). In addition to the homeodomain, Tlx proteins contain four other highly conserved motifs designated TH1–TH4 (Cheng and Mak, 1993). The TH1 domain is also found in Engrailed, msh, Nk1,Nk2, and goosecoid and has been shown to actively mediate transcriptional repression in vivo (Smith and Jaynes, 1996), raising the possibility that Tlx proteins function as transcriptional repressors. Recent in vitrostudies have shown that Tlx-1 expression overcomes G2 cell cycle arrest, increasing cell proliferation (Kawabe et al., 1997). However, this seems unlikely to be its function during normal neural development, given that chick Tlx-1 is only expressed in postmitotic cells (Fig. 5 C). Indeed, the coordinated expression patterns described here for Tlx-1 andTlx-3 in the peripheral and central nervous systems suggest a role for Tlx genes in the differentiation and/or survival of distinct neuronal cell types and in specifying connectivity. In the hindbrain, despite the fact that Tlx genes are only expressed after the lateral longitudinal tract is formed (Moody et al., 1989; Chédotal et al., 1995), the onset of expression in sensory nuclei occurs while afferent axon collaterals are still tipped with growth cones (Fig. 6 F). Hence Tlx genes, which are expressed both peripherally and centrally, are well placed to play a role in coordinating synaptogenesis. One model might be thatTlx genes directly or indirectly regulate sets of cell–cell recognition molecules, such as cadherins (Suzuki et al., 1997), whose presence on both presynaptic and postsynaptic elements of specific pathways might facilitate the development of appropriate connectivity. Loss-of-function mutations in mice have been made to directly test the function of one member of the Tlx gene family. Targeted disruption of the mouse Tlx-1 gene results in complete agenesis of the spleen (Roberts et al., 1994; Dear et al., 1995). Although not as yet reported, our results suggest that a more subtle neuronal phenotype might also be expected in these mice. The overlapping expression patterns of different Tlx family members (Cheng and Mak, 1993; Dear et al., 1993, 1995; Raju et al., 1993; Roberts et al., 1995; Hatano et al., 1997) may, however, complicate the analysis of such loss-of-function mutations.
Conclusions: a Tlx code for cranial sensory innervation?
Two members of the Tlx homeobox gene family,Tlx-1 and Tlx-3, are closely associated with development of cranial sensory innervation from peripheral ganglia. We propose that both subsets of hindbrain sensory neurons and their afferent inputs are developmentally specified by the combinatorial expression of members of the Tlx family of transcription factors. This potential code is defined by two important criteria: (1) where one homolog defines a single hindbrain structure, for example theTlx-3-positive solitary nucleus, the ganglia that contribute axons (7, 9, and 10) to this nucleus will also express only this single homolog (Tlx-3); and (2) where domains of Tlxexpression overlap in the periphery, as for example in the maxillomandibular lobe of the fifth ganglion, the corresponding hindbrain nucleus will express both genes, but in nonoverlapping territories (the Tlx-3-positive principal trigeminal nucleus vs the Tlx-1-positive descending trigeminal nucleus).Tlx genes are the first markers that distinguish subsets of hindbrain sensory nuclei. They are also among the first family of transcription factors to be described that define specific subsets of both afferent and target neurons within a general neural pathway. We expect that characterization of further family members will reveal more of the detailed sensory apparatus of the hindbrain and confirm or refute the predictions of a Tlx code. In addition, the continuum of Tlx-3 expression between columns of sensory nuclei in the hindbrain and identified D2 and D3 neurons in the spinal cord suggest they are specified by a common underlying mechanism.
This work was supported by the Wellcome Trust, the United Kingdom Medical Research Council, and the Howard Hughes Medical Institute, of which A.L. is an International Research Scholar. C.L. was supported by the Medical Research Council of Canada and the International Human Frontier Science Program Organization. R.J.T.W. was a Medical Research Council of the United Kingdom Training Fellow. We thank Iano Campbell for excellent technical assistance, Anthony Graham and Gilles Fortin for helpful suggestions and discussions, Jon Gilthorpe for help with computing, and Nathalie Choussat for her help with preliminary analyses. The antibodies for r5-vimentin, Islet-1/2, and the probe for human Hox-11 were generous gifts from Ursula Dräger, Thomas Edlund, and Stanley Korsmeyer, respectively.
Correspondence should be addressed to Andrew Lumsden, Department of Developmental Neurobiology, United Medical and Dental Schools, Guy’s Hospital, London SE1 9RT, UK.
Dr. Logan’s present address: Department of Anatomy and Neuroscience Research Group, Faculty of Medicine, University of Calgary, 3330 Hospital Drive Northwest, Calgary, Alberta T2N 4N1, Canada.
Dr. McKay’s present address: Department of Periodontology, St. Bartholomew’s and the Royal London School of Medicine and Dentistry, Turner Street, London E1 2AD, UK.