Research Article

Hoxa2- and Rhombomere-Dependent Development of the Mouse Facial Somatosensory Map

See allHide authors and affiliations

Science  08 Sep 2006:
Vol. 313, Issue 5792, pp. 1408-1413
DOI: 10.1126/science.1130042

Abstract

In the mouse trigeminal pathway, sensory inputs from distinct facial structures, such as whiskers or lower jaw and lip, are topographically mapped onto the somatosensory cortex through relay stations in the thalamus and hindbrain. In the developing hindbrain, the mechanisms generating such maps remain elusive. We found that in the principal sensory nucleus, the whisker-related map is contributed by rhombomere 3–derived neurons, whereas the rhombomere 2–derived progeny supply the lower jaw and lip representation. Moreover, early Hoxa2 expression in neuroepithelium prevents the trigeminal nerve from ectopically projecting to the cerebellum, whereas late expression in the principal sensory nucleus promotes selective arborization of whisker-related afferents and topographic connectivity to the thalamus. Hoxa2 inactivation further results in the absence of whisker-related maps in the postnatal brain. Thus, Hoxa2- and rhombomere 3–dependent cues determine the whisker area map and are required for the assembly of the whisker-to-barrel somatosensory circuit.

The rodent trigeminal pathway represents a suitable system to study neuronal connectivity and pattern formation (15). Somatosensory inputs from distinct facial regions are collected through the mandibular (supplying the lower jaw and lip), the maxillary (supplying the whiskers and the upper jaw and lip), and the ophthalmic branches of the trigeminal nerve by first-order neurons whose cell bodies reside in the trigeminal ganglion (Fig. 1). The central processes of trigeminal ganglion neurons enter the hindbrain and give ascending and descending branches wiring into the rostral principal (PrV) and the caudal spinal (SpV) nuclei. Sensory inputs are in turn transmitted to the somatosensory cortex through a relay station in the ventral posterior medial (VPM) nucleus of the thalamus. At all levels of the pathway, the spatial arrangements of neurons and their afferent fibers faithfully reproduce the physical distribution of peripheral sensory receptors generating somatotopic facial representations. However, little is known about the mechanisms underlying the development of somatotopy.

Fig. 1.

Rhombomere-related somatotopic organization and connectivity of PrV nucleus. (A to L) Cross-sections through hindbrain PrV [(A) to (C) and (G) to (I)] and thalamic VPM [(D) to (F) and (J) to (L)] nuclei in P4 R2::Cre;Z/AP [(A) to (F)] or Krox20::Cre;Z/AP [(G) to (L)] animals, respectively. Face maps are revealed by cytochrome oxidase (CO) histochemistry at PrV [(A) and (G)] and VPM [(D) and (J)] levels. Lower jaw– (Lj), anterior snout (As) sensory hair–, and whisker-related neuronal patterns are shown. Whisker-related neuronal modules (Br, barrelettes; Ba, barreloids) are organized into five rows, labeled A through E. The progenies of r2 (r2p) (B) or r3 and r5 (r3p and r5p) (H) are traced by alkaline phosphatase (AP) staining, and superimposed (Merge) to CO-stained adjacent sections in (A) or (G), respectively. At PrV level, r2p (C) or r3p (I) specifically contribute to the Lj or the As and Br maps, respectively. At VPM level, the Lj map is entirely targeted by AP-stained axon terminals from r2-derived neurons [(E) and (F)], whereas axon terminals from Krox20::Cre;Z/AP animals selectively target As and Ba neuronal modules, precisely matching CO-stained neuronal modules [(K) and (L)]. (M) Diagram of trigeminal circuit in mouse. (N) Summary showing the relationship between rhombomere progenies and somatotopy of PrV nucleus and its axonal connections to VPM. Md, mandibular branch of trigeminal nerve; MoV, trigeminal motor nucleus; Mx, maxillary branch; Oph, ophthalmic branch; PrV, principal sensory trigeminal; S1, somatosensory cortex; SpV, spinal sensory trigeminal column; TG, trigeminal ganglion; tl, trigeminal lemniscus; VPL, ventroposterior lateral nucleus; VPM, ventral posterior medial nucleus.

At early developmental stages, the vertebrate hindbrain is transiently segmented into rhombomeres (r) along the anteroposterior axis (6). Fate-mapping studies revealed that the progeny of individual segments form compact transverse stripes of cells running throughout the ventriculopial axis of the postnatal hindbrain (7). However, the relationship between the early segmental plan and the establishment of topographical circuitry in the developing hindbrain is still poorly understood. At the molecular level, Hox genes are crucial determinants of rhombomere identity and neuronal patterning (811). One possibility is that, at later stages of hindbrain maturation, Hox expression may set up a molecular program for somatotopic map formation. Here, we focused on Hoxa2, the most rostrally expressed Hox gene in the hindbrain, and on the development of somatotopy in the PrV nucleus, which has a fundamental role in thalamic and cortical pattern formation (12, 13). We show that cellular segregation of rhombomere progenies and maintenance of Hoxa2 expression in PrV through later stages of hindbrain development provide a set of positional labels that instructs both the wiring of peripheral trigeminal afferents and the topographic axonal mapping of PrV target neurons to VPM thalamus (summary in fig. S1).

Rhombomere-dependent somatotopy of PrV. The somatotopic organization of the rodent trigeminal system can be conveniently visualized by cytochrome oxidase (CO) histochemistry (Fig. 1, A, D, G, and J) (5, 1416). At postnatal day 4 (P4), the dorsal component of the PrV nucleus appears as a lightly CO-stained area that includes the lower jaw and lip representation (Fig. 1, A and G) and receives inputs from the mandibular branch of the trigeminal nerve. The ventral PrV component is instead organized in neuronal modules, or barrelettes, replicating the array of whiskers and sinus hairs on the snout (Fig. 1, A and G) that are innervated by the maxillary branch. Similar neuronal arrangements also exist at thalamic and cortical levels both for the lower jaw and lip and for whisker-related representations (known as barreloids and barrels, respectively) (Fig. 1, D and J). However, little is known about the cellular mechanisms underlying the establishment of interareal somatotopy during prenatal development, resulting in segregation of the lower jaw and lip and the whisker-related maps.

Here, we asked whether PrV somatotopy relates to its rhombomeric origin. To permanently label rhombomere progenies, we mated mouse lines expressing the Cre recombinase selectively in r2 (R2::Cre) (17) or in r3 and r5 (Krox20::Cre) (18) with the Z/AP reporter line (19). In this mating scheme, Cre-mediated recombination results in permanent activation of alkaline phosphatase (AP) in rhombomere progenies. AP and CO stainings on adjacent cross-sections at P4 revealed that, in the PrV, the territory including the barrelettes and sinus hair–related neuronal clusters is contributed by the r3 progeny (r3p) (Fig. 1, G to I). In contrast, the unpatterned CO-stained portion of PrV, containing the map of the lower jaw and lip, is composed entirely by the r2 progeny (r2p) and anteriorly delimited by the r1 progeny (r1p) (Fig. 1, A to C and fig. S2, A to D). Posteriorly, the barrelette field is bordered by the r4 progeny (r4p) (Fig. 1, G to I and fig. S2, E to H). Notably, the progeny of r5 (r5p) did not contribute to the PrV nucleus, nor to the more posteriorly located interpolaris (SpVi) or caudalis (SpVc) subdivisions of the SpV column (Fig. 1, G to I and fig. S2), the only other neuronal formations generating whisker-related representations (16).

Rhombomere-related topography of PrV axonal projections. Next, we mapped the AP-stained axonal projections from the brainstem onto the CO-stained representations of distinct face areas in the VPM on adjacent sections. In Krox20::Cre;Z/AP animals at P4, AP-stained axon terminals precisely matched the CO-stained barreloids and sinus hair–related neuronal modules in the dorsolateral VPM (Fig. 1, J to L). Most of these projections to VPM originated from r3-derived PrV neurons, given that the r5 progeny did not contribute to any of the nuclei generating the main stream of ascending whisker-related afferents to VPM—i.e., PrV, SpVi, or SpVc (20) (Fig. 1 and fig. S2). Notably, in R2::Cre;Z/AP animals the AP-stained projections adopted instead a complementary pattern. Axon terminals from r2-derived PrV neurons mainly segregated into the ventromedial VPM, containing the representation of the lower jaw and lip, and were excluded from the barreloid area (Fig. 1, D to F).

Rhombomere-specific patterns of trigeminal primary afferent arborization. The mapping data suggested that r3-derived PrV neurons predominantly receive inputs from the whiskers, supplied by the maxillary division of the trigeminal nerve, whereas inputs from the mandibular division are relayed through the r2-derived portion of PrV. To test this possibility, we assessed the central projection patterns of trigeminal primary afferents (Figs. 2 and 3).

Fig. 2.

Rhombomere-specific arborization patterns of trigeminal afferents and Hoxa2 expression. (A to C) Trigeminal tract somatotopy in whole-mount E14.5 wild-type brain. Maxillary (Mx) and mandibular (Md) (A), or Mx and ophthalmic (Oph) (B) branches were simultaneously retrogradely labeled by rhodamin-dextran [red in (A) and (B)] or Alexa 488-dextran [green in (A) and violet in (B)], respectively. Crosses in summary diagram (C) indicate local patterns of arborization of axons. The Md branch runs dorsally, Oph ventrally, and Mx in between. The relative width of each branch varies along the rostrocaudal axis [bars in (A) and (B)]. The star (–) indicates the trigeminal nerve entry point. Cb, cerebellum. (D to F) Arborization patterns and spatial segregation of radially oriented collaterals from retrogradely labeled Md (green) and Mx (red) branches [(D) and (E)] in coronal sections through PrV nucleus at (D) and posterior (E) to the nerve entry point [as in summary (F)]. Posteriorly, the PrV is predominantly targeted by collaterals from Mx. (G to N) Rhombomere-specific patterns of arborization of Md and Mx branches. Md (green) or Mx (red) branch dextran labeling and simultaneous detection of EGFP (blue) from r2-derived (r2p) or r3-derived (r3p) progenies in R2::Cre;Hoxa2EGFP(lox-neo-lox) [(G), (I), and (J)] or Krox20::Cre;Hoxa2EGFP(lox-neo-lox) [(L) and (M)] E14.5 fetuses, respectively. Selective arborization of Md branch in r2p [arrows, (G) and (I)], whereas no or very few collaterals are formed in r3p (L). [(J) and (M)] R3p is specifically targeted by Mx collaterals (arrows) [summary in (H), (K), and (N)]. (O to Q)[and summary in (R)] Coronal adjacent sections through PrV (outlined by dashed line) from Krox20::Cre;Z/AP E14.5 fetuses. AP staining in (O) visualizes r3p, whereas (P) and (Q) show in situ hybridization with Hoxa2 and PrV-specific Drg11 probes, respectively. High-level Hoxa2 expression in r3p, but not in r2p, of PrV target neurons correlates with maxillary branch arborization.

Fig. 3.

Spatiotemporal requirement of Hoxa2 for trigeminal afferent pathfinding and arborization. (A to C) Retrograde labeling of trigeminal nerve by whole trigeminal ganglion (TG) injection of rhodamine-conjugated dextran in E16.5 wild-type (WT) (A), Hoxa2EGFP–/– (B), and r2-specific knockout (R2::Cre;Hoxa2flox/flox) (C) fetuses. In (B) and (C), the nerve ascending branch does not stop at its normal position after entering the hindbrain, but ectopically projects to the cerebellum. (D to F) Cross-sections through E10.5 (D) and E14.5 (E) Hoxa2EGFP+/– specimen doubly labeled for rhodamine and EGFP showing the entrance of the trigeminal nerve (Vn) tract at r2 level and its point of arrest at the r1/r2 border [summary drawing in (F)]. (G to I) Cross-sections through E10.5 (G) and E14.5 (H) Hoxa2EGFP–/– specimen. In homozygous mutants, Vn ectopically projects toward the cerebellar territory [summary in (I)]. These data demonstrate an early requirement for Hoxa2 in r2 to arrest incoming ascending afferents. (J to U) Coronal sections through the hindbrain of E14.5 WT [(J) to (M)], tamoxifen (TM)–induced CMV::Cre-ERT2;Hoxa2flox/flox [(N) to (Q)], and r3/r5-specific Krox20::Cre;Hoxa2flox/flox [(R) to (U)] conditional homozygous mutant fetuses. Mandibular (Md) and maxillary (Mx) branches and their collaterals were simultaneously retrogradely labeled by Alexa 488-dextran (green) and rhodamin-dextran (red), respectively. Maxillary branch arborization is selectively inhibited both after Hoxa2 temporal inactivation during collateral formation (TM administration at E12.5 through E13.5) [arrows, (P)] and spatial inactivation in r3 [arrows, (T)] [summaries in (Q) and (U)]. Reorientation of mandibular axon collaterals is also observed [arrows in (O) and (S)]. Specimens in (N) to (P) and (R) to (T) are distinct. Cb, cerebellum; MB, midbrain.

By embryonic day 10.5 (E10.5), the trigeminal nerve has already entered the hindbrain through an entry point in r2 (10, 21), and ascending axons are arrested shortly thereafter at their final position. By fluorescent dextran tracing of trigeminal afferents and enhanced green fluorescent protein (EGFP) detection in the hindbrain of Hoxa2EGFP embryos (22), we mapped the point of arrest next to the r1/r2 border (Fig. 3D). By E14.5, somatotopy of the trigeminal tract was evident, as determined by labeling individual branches with distinct fluorescent dextrans (Fig. 2, A to C). Specifically, mandibular axons were located dorsally, ophthalmic axons ventrally, and maxillary afferents run in-between the two other divisions. Notably, the relative width of each division within the nerve varied along the rostrocaudal axis (Fig. 2, A to C), corresponding to differential patterns of arborization into the PrV target nucleus (Fig. 2, D to F). Caudally to the nerve entry point, the maxillary axons and their radially oriented collaterals contributed to most of the dorsoventral extent of the tract (in red, Fig. 2, A to C and E), with very little input from mandibular (in green, Fig. 2, A, C, E, and F) or ophthalmic (in violet, Fig. 2, B and C) divisions. In contrast, mandibular collaterals rostral to the nerve entry point were conspicuous and spatially segregated from maxillary collaterals (Fig. 2, A, D, and F); moreover, no collaterals could be identified from the ophthalmic branch (Fig. 2, B and C).

Such rostrocaudal variations suggested that the arborization patterns from mandibular or maxillary afferents may be related to the rhombomeric origin of PrV target neurons. To correlate arborization patterns with rhombomeric domains, we used the Hoxa2EGFP(lox–neo–lox) allele (22) that expresses EGFP only upon Cre-mediated deletion of the selection marker cassette, thus allowing tracing of rhombomere-specific domains of Hoxa2EGFP-expressing cells by mating with suitable Cre-expressing mouse lines. Therefore, we simultaneously visualized collaterals from individually labeled trigeminal branches and r2- or r3-derived Hoxa2-expressing territories by EGFP detection in E14.5 R2::Cre;Hoxa2EGFP(lox–neo–lox) (Fig. 2, G and K) or Krox20::Cre;Hoxa2EGFP(lox–neo–lox) (Fig. 2, L and N) fetuses, respectively. Notably, the entire dorsoventral extent of the r3-derived portion of PrV was selectively targeted by collaterals from maxillary axons (arrows, Fig. 2, J and M; summaries in Fig. 2, K and N), with very little, if any, mandibular input (Fig. 2, I, K, L, and N). In contrast, mandibular axons selectively sent collaterals dorsally within the r2-derived portion of PrV (arrows, Fig. 2G; summary in Fig. 2H). Thus, individual trigeminal divisions differentially contribute to the innervation of PrV in relation to the rhombomeric origin of target neurons, with the r3-derived PrV neurons receiving input almost uniquely from the maxillary division.

Hoxa2 differential expression in PrV. At early stages, Hoxa2 is expressed throughout r2 and r3, although at different levels (10). At E14.5, in situ hybridization in comparison with the PrV-specific marker Drg11 (23) revealed restricted expression of Hoxa2 in developing PrV neurons (Fig. 2, P and Q) but not in trigeminal ganglion cells. Comparison of transcript distribution and AP staining on adjacent sections from Krox20::Cre;Z/AP fetuses revealed that the r3-derived portion of PrV expressed high levels of Hoxa2 (compare Fig. 2, O and P). In contrast, Hoxa2 expression was much lower in r2p, specifically in the dorsal part of PrV (Fig. 2P). Such transcript distributions correlated with the r2- or r3-specific arborization patterns of mandibular or maxillary afferents, respectively, suggesting that high levels of Hoxa2 expression may be involved in selective wiring of the maxillary division.

Early Hoxa2 requirement in r2 for trigeminal nerve pathfinding. To address Hoxa2 function, we analyzed the patterns of trigeminal afferents in Hoxa2EGFP–/– homozygous mutants. A notable pathfinding defect was observed in E16.5 (Fig. 3, A and B) and E14.5 (Fig. 3, E, F, H, and I) homozygous mutants. The ascending branch of the trigeminal nerve did not stop to encapsulate the PrV nucleus, but ectopically projected to the cerebellum. Labeling of homozygous mutants at E10.5, before PrV nucleus formation (24), revealed that the pathfinding defect was already present at this early stage (arrows, Fig. 3, D and G).

To address the Hoxa2 spatial requirement, we next deleted Hoxa2 selectively in r2 by mating R2::Cre transgenic mice with our Hoxa2flox/flox allele (17). Notably, in R2::Cre;Hoxa2flox fetuses the trigeminal tract also ectopically projected to the cerebellum (Fig. 3C), similarly to the full knockout phenotype. Thus, the normal arrest of trigeminal afferents in the rostral hindbrain does not require information from PrV target neurons, but it is dependent on early expression of Hoxa2 in the r2 neuroepithelium.

Late Hoxa2 requirement for maxillary axon arborization. To investigate Hoxa2 involvement during arborization of trigeminal afferents, we used an approach based on tamoxifen (TM) inducible CMV–βactinCre-ERT2 (CMV::CreERT2)/loxP to achieve time-dependent inactivation of Hoxa2 (25). Given that arborization of trigeminal afferents into PrV neurons started at about E13 (26), CMV::Cre-ERT2;Hoxa2flox/flox fetuses and Hoxa2flox/flox controls were chronically administered with TM for three times at E12.5, E13.0, and E13.5. Fetuses were collected at E14.5 and in situ hybridization confirmed no or little residual Hoxa2 expression in the hindbrain of TM-treated CMV::Cre-ERT2;Hoxa2flox/flox mutants (25). Retrograde labeling of individual trigeminal branches revealed a selective inhibition of collateral formation from the maxillary branch (Fig. 3, J, L, M, N, P, and Q). Moreover, some collaterals from the mandibular division appeared to be abnormally oriented toward the territory normally targeted by maxillary collaterals (arrow, Fig. 3O; summary in Fig. 3Q), suggesting a degree of competitive interactions between trigeminal branches (27). In contrast, we did not observe in these animals ectopic trigeminal projections to the cerebellum, confirming that such a phenotype depended on Hoxa2 function earlier than E12.5.

To further support a Hoxa2 role for maxillary axon arborization, we generated Krox20::Cre;Hoxa2flox/flox fetuses. At E14.5, we observed a severe reduction of maxillary collaterals into the r3-derived PrV (Fig. 3, R, T, and U), with occasional ectopic reorientation of mandibular collaterals, similar to the TM-induced knockout animals (Fig. 3, O, Q, S, and U). Thus, Hoxa2 is required in PrV target neurons to induce selective arborization of maxillary axons.

Loss of EphA4 and EphA7 expression in PrV of Hoxa2 mutants. In E15.5 Krox20::Cre;Hoxa2flox/flox homozygous mutant fetuses, normal specification and organization of PrV neurons was observed, as assessed by Drg11 expression (fig. S3, A and D). However, EphA4 and EphA7 transcripts were lacking or severely reduced in the r3-derived portion of PrV, whereas ephrinA5 expression in the target thalamus was unaffected (fig. S3, B, C, E, F, J, and K). Importantly, we also observed similar spatially restricted impairments of EphA4 and EphA7 expressions in the PrV of TM-treated CMV::Cre-ERT2;Hoxa2flox/flox mutant fetuses (fig. S3, H and I). These results confirmed a late requirement of Hoxa2 and indicated that topographic wiring defects may be present in the mutants as a result of impaired EphA receptor expression.

Absence of whisker-related maps and altered topography of PrV axonal connectivity in Hoxa2 mutants. At E16.5, the analysis of AP-stained axons in Krox20::Cre;Hoxa2flox/flox;Z/AP mutants demonstrated normal pathfinding from PrV to VPM (fig. S4, A to F). However, CO-stainings at P4 revealed the absence of whisker-related neuronal patterns at both the PrV and VPM levels in these mutants (Fig. 4). Notably, analysis of AP-stained axon terminals revealed that, within the VPM, axons from PrV did not establish a normal topographic branching pattern and were mistargeted. Specifically, the majority of fibers did not project to the barreloid field, but ectopically targeted the VPM ventromedial region (Fig. 4, H, K, M, and N). Moreover, apoptotic loss of r3-derived PrV neurons was observed between P0 and P4 (Fig. 4, B and E, and fig. S5), which was not observed at prenatal stages (Fig. 4, A and D); this difference was likely the consequence of the partial deafferentation and miswiring of the circuit during the postnatal critical period of sensory experience (28). Thus, the inactivation of Hoxa2 during prenatal development altered the wiring properties of PrV neurons with regard to both peripheral afferents and thalamic VPM target neurons, resulting in topographic changes of the trigeminal circuit in mutant postnatal brain (summarized in fig. S1).

Fig. 4.

Absence of whisker-related neuronal patterns and altered topography of PrV connectivity in Hoxa2 mutants. (A to F) Cross-sections through PrV nucleus (outlined) of Krox20::Cre;Z/AP [(A) to (C)] and Krox20::Cre;Hoxa2flox/flox;Z/AP homozygous mutants [(D) to (F)] at E16.5 [(A) and (D)] and P4 [(B), (C), (E), and (F)]. Cytochrome oxidase (CO) histochemistry reveals absence of barrelettes (Br) in mutant PrV at P4 (F), and progressive loss of alkaline phosphatase (AP)-stained r3-derived (r3p) PrV neurons between E16.5(D) and P4(E). (G to L) Cross-sections through the thalamus of Krox20::Cre;Z/AP [(G) to (I)] and Krox20::Cre;Hoxa2 flox/flox;Z/AP [(J) to (L)] P4 animals. [(G) and (J)] Ventroposterior medial (VPM) nuclei are visualized by CO, whereas AP stainings on adjacent sections [(H) and (K)] map axon terminals from the PrV neurons in panels [(B) and (E)]. Lack of barreloids (Ba) is observed in Krox20::Cre;Hoxa2flox/flox;Z/AP mutant VPM [arrows, (J)], whereas AP staining demonstrates mistargeting of PrV axon terminals to ventromedial VPM [(K), and merge in (L)]. (M and N) Summary drawings.

Conclusions. We found that rhombomere-specific cellular cues are important for the spatial segregation of PrV neurons into a somatotopic pattern. Specifically, the r3 progeny contribute to the portion of the PrV that segregates into whisker-related neuronal patterns at postnatal stages. The persistence of rhombomere-specific cohesion properties of postmitotic progeny into late stages of hindbrain development may provide a cellular framework upon which to build precise neuronal connectivity. In addition, we demonstrated that Hoxa2 has multiple spatiotemporal roles in the assembly of the trigeminal circuit. First, we demonstrated an early requirement for Hoxa2 expression in the r2 neuroepithelium to prevent entering of peripheral afferents in r1. However, inner ear vestibular afferents, running just lateral to trigeminal axons, are not arrested at the r1/r2 border and normally project to the cerebellum (29). Thus, a Hoxa2-dependent molecular barrier may exist throughout r2 and/or at the r1/r2 border that is specifically involved in arresting the pathfinding of trigeminal but not vestibular axons.

Second, the onset of arborization of trigeminal axons into the PrV follows a rhombomere-specific pattern. Specifically, r3-derived neurons selectively receive collaterals from maxillary axons, whereas mandibular axons arborize predominantly into the r2-derived portion of PrV. Such a pattern correlates with differential levels and distribution of Hoxa2 transcripts in the PrV nucleus: High or low expression domains corresponded to wiring by maxillary or mandibular axons, respectively. Moreover, spatiotemporally induced inactivations established a specific requirement for Hoxa2 during collateral formation from the maxillary division (Fig. 3). Notably, trigeminal ganglion neurons, which do not express Hoxa2,displayed a normal somatotopic organization in the mutants (fig. S6), supporting the idea that the topography of trigeminal peripheral processes is independent of central influences but that target cell maturation in the brainstem is an important regulator of arborization of central afferents (26). Thus, Hoxa2 function in PrV neurons may be important to regulate the expression of molecules involved in trigeminal afferent arborization, such as neurotrophins and their receptors (30, 31), Slit proteins and Robo receptors (32), and/or semaphorins and neuropilin receptors (33).

Third, our data indicate the involvement of Hoxa2 in the topographic wiring of axonal connections to the VPM thalamus. While guidance of PrV axons to the thalamus was not affected in Hoxa2 mutants, selective changes occurred in the topographic specificity of axonal mapping within the VPM nucleus. Graded expressions of ephrins and Eph receptors have been involved in sensory mapping (3437). The finding that Hoxa2 positively regulates EphA4 and EphA7 expressions in the PrV indicates that Hoxa2-mediated control of connectivity could partly occur through Eph receptor function. By governing the distribution of molecules providing positional mapping labels, Hoxa2 could simultaneously regulate topographic wiring between brainstem (PrV) and thalamus (VPM), as well as between the periphery (incoming trigeminal afferents) and PrV. Moreover, as soluble homeobox proteins can function as guidance factors for axons (38), additional mechanisms might be at work to establish Hoxa2-mediated topographic wiring. Together with recent evidence indicating a role in motoneuron connectivity (39), these data begin to point to Hox genes as fundamental players in the building of sensorimotor circuitry in vertebrates.

Supporting Online Material

www.sciencemag.org/cgi/content/full/1130042/DC1

Materials and Methods

Figs. S1 to S6

References

References and Notes

View Abstract

Navigate This Article