Homeotic Transformation of Rhombomere Identity After Localized Hoxb1 Misexpression

See allHide authors and affiliations

Science  25 Jun 1999:
Vol. 284, Issue 5423, pp. 2168-2171
DOI: 10.1126/science.284.5423.2168


Segmentation of the hindbrain and branchial region is a conserved feature of head development, involving the nested expression of Hox genes. Although it is presumed that vertebrateHox genes function as segment identifiers, responsible for mediating registration between elements of diverse embryonic origin, this assumption has remained untested. To assess this, retroviral misexpression was combined with orthotopic grafting in chick embryos to generate a mismatch in Hox coding between a specific rhombomere and its corresponding branchial arch. Rhombomere-restricted misexpression of a single gene, Hoxb1, resulted in the homeotic transformation of the rhombomere, revealed by reorganization of motor axon projections.

Since the identification of homologs of Drosophila homeotic genes in vertebrates, a consensus model for their role during head development has emerged. The hindbrain is subdivided into rhombomeres, whereas adjacent tissues are subdivided into a series of branchial arches. Regional expression ofHox genes in the hindbrain is thought to confer identity to rhombomeres (1), whereas equivalent expression in the neural crest–derived branchial arches may provide a positional match between the two systems. Motor axons arising within a given pair of rhombomeres project to a single corresponding branchial arch (2). Targeted mutation (3–5), overexpression (6, 7), or manipulation of Hoxgenes by exogenous retinoids (8) all support aHox code model (9), but none of these approaches can specifically test the assumption that Hox genes act as segment identifiers mediating registration between structures along the anteroposterior axis of the head.

One way to address this issue is to generate a mismatch ofHox gene expression between the hindbrain and periphery by confining overexpression to a specific domain. The expectation is that a homeotic transformation of the domain would be manifested by an orderly respecification of motor axon projections. Accordingly, we have compared the effects of globally overexpressing a single Hoxgene, Hoxb1, with the effects of restricting its overexpression to a single rhombomere or branchial arch.

Hoxb1 has a rostral limit of expression at the boundary between rhombomere 3 (r3) and r4 (Fig. 1A) and is expressed at high levels in r4 throughout the period of overt segmentation (Fig. 1B) (10). Hoxb1 expression thus characterizes both facial motor neuron precursors in basal r4 (4) and the neural crest cells that migrate from dorsal r4 into facial motor neuron target regions of the second branchial arch (11). The gene is never expressed rostral to r4 or in the first branchial arch. To overexpressHoxb1 in chick embryos, we cloned the mouse gene into RCAS and RCAN retroviral vectors and used these to infect embryos (12) at Hamburger and Hamilton (13) stages 3 through 8. We analyzed embryos at stages 18 through 23, monitoring mouse Hoxb1 expression independently of endogenousHoxb1 (Fig. 1C).

Figure 1

Endogenous expression of Hoxb1 at stages 9 (A) and 19 (B), and ectopic Hoxb1 expression (brown) in a stage 23 hindbrain after infection with RCAS/Hoxb1 at stage 3 (C).

In normal and control (RCAN/Hoxb1-infected) embryos, r4 can be distinguished from r2 both by molecular markers and by motor neuron migration behavior (Fig. 2, A to D). First, expression of the immunoglobulin (Ig) superfamily cell surface glycoprotein BEN (14) is substantially stronger in the (facial) motor neurons of r4 than in the (trigeminal) motor neurons of r2 (15) (Fig. 2A). Second, the transcriptional control gene GATA2, a putative downstream target of Hoxb1in motor neurons (16), is expressed transiently at higher levels in r4 than in r2 (Fig. 2B) (17). Third, an r4-specific population of contralaterally migrating (vestibuloacoustic) efferent (CVA) neurons (18) can be revealed by retrograde tracing (Fig. 2C) from the cranial nerve exit point in r4 (19). Fourth, retrograde labeling of motor neurons from the branchial arches reveals that r2 motor axons innervate the first arch and r4 motor axons innervate the second arch (2); this registration was observed in all uninfected embryos (n = 39) (20) and controls (n = 23) (Fig. 2D).

Figure 2

Comparison of controls (A to E) and RCAS/Hoxb1-infected embryos (F toJ). (A and F) BEN expression. (B and G) GATA2expression. (C and H) CVA neurons labeled with DiI (arrows; floor plate shown by asterisk). (D and I) Motor axon projections, revealed by retrograde tracing from first (green) and second (red) branchial arches. (E and J) BEN-stained hindbrains in transverse section, showing ectopic nerve (arrow).

After global overexpression of Hoxb1, both BEN (Fig. 2F) andGATA2 (Fig. 2G) are up-regulated in r2, to levels normally seen in r4 (Fig. 2, A and B) (10). Retrograde labeling reveals extensive contralateral neuronal migration in r2 (n = 21) (Fig. 2H), suggesting the ectopic differentiation of CVA neurons. In the large majority of embryos (n = 32/35), the registration between motor nerves and their peripheral targets was unaltered, consistent with previous studies (4, 7) in which Hoxgene expression was changed globally. However, in a few RCAS/Hoxb1-infected embryos (n = 3/35), we noted a prominent reordering of motor axon trajectories. In these, r2 motor neurons were labeled not from their normal target, the first branchial arch, but from the second branchial arch (Fig. 2I), which they reached by way of an ectopic axon fascicle coursing longitudinally outside the hindbrain (Fig. 2, compare E and J). Thus, global overexpression of Hoxb1 can result in the transformation of r2 to r4, on the basis of molecular markers and motor neuronal migration, but rerouting of the ectopic “r4”-type motor neurons to the second arch is rare. The acquisition of a “normal” pattern of peripheral innervation may result from the transformed neurons still recognizing the first arch as an appropriate target because, with global overexpression, it too would have been transformed to a more posterior identity. The occasional rerouting of “r4”-type motor axons is consistent with this possibility, if it is assumed that the mosaic nature of infection happened to spare the first arch. This, in turn, suggests that rerouting would be a predictable consequence of confining overexpression to the basal plate of the neural tube.

Thus, we explanted the r2 basal plate of stage 10 through 11 embryos that had been globally infected at stage 3 with RCAS(N)/Hoxb1 and grafted these orthotopically into stage-matched, virus-resistant hosts (n = 22). Orthotopic grafting was as described (21), and embryos were harvested at stage 19 through 20. As expected for controls, motor neurons arising in r2 grafts that were infected with the RCAN/Hoxb1 virus (n = 5), or were uninfected (n = 14), always sent their axons exclusively to the first branchial arch (Fig. 3A, upper panel). By contrast, where the grafted basal r2 showed high levels of Hoxb1 expression, we found that motor neurons projected aberrantly to the second branchial arch at a high frequency (n = 12/17) (Fig. 3, B to D). Both carbocyanine dye tracing (n = 7/9) (Fig. 3B) and whole antibody to neurofilament staining (n = 5/8) (Fig. 3C) revealed a prominent ectopic nerve between the exit point in r2 and the root of the second arch. In all cases, there was no Hoxb1expression outside the graft (Fig. 3, A and D, lower panels). Neighboring motor neurons within RCAS/Hoxb1-infected basal r2 grafts sometimes projected to different branchial arches (Fig. 3D, upper panel). This would be consistent with a cell-autonomous effect of Hoxb1 on the choice of cell identity, if it could be shown to correlate with mosaicism of infection.

Figure 3

Comparison of control (A) and RCAS/Hoxb1-infected (B to D) r2 basal plate orthotopic grafts. (A and D) Motor neurons traced from first (green) and second (red) arches (upper panels) and viral infection in the same grafts (lower panels). (B and C) Abnormal longitudinal fascicle of axons between r2 and the second arch, revealed by retrograde tracing (B) and in antibody to neurofilament (C) (arrow). (E to G) Failure of trigeminal motor neurons correctly to innervate first branchial arch territory ectopically expressing Hoxb1. (E) Unattached trigeminal ganglion, lacking motor nerve (arrow). (F) Fused trigeminal and facial ganglia (white arrow), truncated mandibular nerve (black arrow). (G) Uninfected graft (green) within Hoxb1-positive tissue (red). (H to J) Unilateral orthotopic graft of RCAS/Hoxb1-infected r2 alar plate into an uninfected embryo. Compare the normal appearance of the trigeminal ganglion and mandibular nerve (arrow) contralateral to the graft (H), with dispersion of the ganglion and truncation of the nerve (arrow) on the operated side (I), where first arch neural crest cells express Hoxb1 (J). V: trigeminal ganglion; VII: facial ganglion.

To test the extent to which Hoxb1-infected basal r2 replicates the normal r4 phenotype, we also grafted uninfected r4 basal plates heterotopically into the r2 basal plate position. Motor neurons from the graft (n = 7/8) (20) behaved identically to r2 motor neurons expressing Hoxb1 (Fig. 3B).

Finally, we performed a complementary series of grafts to examine the effects of Hox gene misexpression in the target, independent of the basal plate of the neural tube, where motor neurons are born. Two grafting strategies were used, with motor axon outgrowth assessed by neurofilament staining. First, grafts of quail basal r2 were transplanted, unilaterally and orthotopically, into chick embryos that were heavily infected with RCAS/Hoxb1(n = 7). In some embryos, no motor axons could be observed exiting the hindbrain (n = 3) (Fig. 3E), whereas in others, first arch innervation was extensively disrupted and trigeminal and facial ganglia were fused (n = 4) (Fig. 3F). No viral spread from host to donor tissue was detected (Fig. 3G), and effects were confined to the operated side of the embryo. Second, in order to target gene manipulation to neural crest derivatives in the first arch, r2 alar plates from RCAS/Hoxb1-expressing embryos (infected at stage 3) were transplanted, unilaterally and orthotopically, into virus-resistant host embryos at stage 8. By stages 18 through 21, crest derivatives expressing Hoxb1 colocalized with disruptions in trigeminal ganglion morphology (n = 4/5). Although development was normal on the control side of the embryo (contralateral to the graft) (Fig. 3H), the mandibular (motor component of the trigeminal) nerve was truncated (Fig. 3I) in the region of ectopic Hoxb1 expression (Fig. 3J).

By surgically restricting ectopic Hoxb1 expression to the basal plate of a rhombomere that lies anterior to the limit of endogenous expression, we have demonstrated a reassignment of motor neuron identity (Fig. 4B), as assessed by axonal trajectory. Using complementary grafting strategies, we have also shown that normal axon projections into the periphery are truncated when the target selectively expresses an inappropriate Hox gene (Fig. 4C).

Figure 4

Summary diagrams showing the registration between hindbrain (r2 and r4) and branchial arches (1st and 2nd) in normal embryos (A), and when Hoxb1 expression is targeted to the basal plate of r2 (B) or first arch tissues (C).

These observations have implications for the interpretation of a number of previous studies where Hox gene levels were altered throughout the embryo. They also define precisely the previously uncertain role of Hox genes in coordinating segmental aspects of head development (22). First, the contrast between the effects of global and targeted misexpression explains the lack of systematic axonal respecification after global, targeted mutation of Hox genes in mice (4,7). Second, Hox genes appear to mediate appropriate connectivity between neural tube and peripheral targets by regulating a system of recognition cues shared by both tissues. By targeting gene misexpression to one rhombomere, the registration between neural tube and branchial arches can be systematically disrupted, overriding the normal segregation of branchiomeric segments. These results demonstrate that a classical homeotic transformation (23), in terms of gene expression, neuronal migration, and axon targeting, can be induced in a defined neuronal population by the ectopic expression of a single Hox gene. Thus, in vertebrates, as in flies, the role of Hox genes is to confer positional identity on individual elements of a meristic series.

  • * To whom correspondence should be addressed. E-mail: andrew.lumsden{at}


View Abstract

Stay Connected to Science

Navigate This Article