Specification of distinct motor neuron identities by the singular activities of individual Hox genes

The motor neurons of the hindbrain are segmentally organised and they display subtype variation along the anteroposterior (AP) axis. They represent an attractive system in which to study patterning along the anteroposterior neural axis (Lumsden and Krumlauf, 1996). Initially each of the hindbrain segments, or rhombomeres, contains a similar basic composition of neural precursors (Clarke and Lumsden, 1993), but as development progresses, segment-specific neuronal diversification becomes evident. It is thought that this is achieved by rhombomere-specific activities of candidate genes such as the Hox genes (Lumsden and Krumlauf, 1996). In order to test for Hox gene functions, gain-of-function and loss-of-function studies have been performed in mouse, chick and zebrafish embryos. These studies have implicated Hox genes in both the initial establishment of rhombomeres as well as the maintenance of specific rhombomeric identities (Lufkin et al., 1991; Chisaka et al., 1992; Carpenter et al., 1993; Dollé et al., 1993; Mark et al., 1993; Zhang et al., 1994; Alexandre et al.,1996; Goddard et al., 1996; Studer et al., 1996; E. Bell, R. Wingate and A. Lumsden, unpublished). It remains unclear, however, how Hox genes confer positional values on neural tissue. Their overlapping expression patterns suggest that positional identity could be conferred by a cooperative action of different Hox genes (‘The Hox code’), or by the sum of individual singular Hox gene activities present at a given axial level, or by a combination of both (see McGinnis and Krumlauf, 1992). However, with the exception of the targeted disruption of Hoxa2 (Gavalas et al., 1997), which is the only Hox gene expressed in rhombomere (r)2, functional studies have focussed on areas of the developing neural tube that express other Hox genes in addition to the particular Hox gene under investigation, thereby complicating the distinction between a combinatorial and singular mode of Hox gene action. Hoxb1, for example, is initially expressed up to the r3/r4 boundary and later is maintained at high levels in r4 and, at lower levels, caudal to r6. It is coexpressed with Hoxa1, Hoxa2 and Hoxb2 in r4, which gives rise to, amongst other cells, branchiomotor neurons of the facial motor nucleus (Lumsden and Krumlauf, 1996). Gene targeting studies have demonstrated that despite the absence of Hoxb1, motor neurons differentiate in r4, suggesting that Hoxb1 might not be required to initiate motor neuron generation (Goddard et al., 1996; Studer et al., 1996). Alternatively, loss of Hoxb1 activity might be partially compensated by the remaining Hox genes, thereby giving rise to the differentiation of motor neurons in r4 with an atypical migration behaviour as described for Hoxb1−/− embryos (Studer et al., 1996; Goddard et al., 1996; Gavalas et al., 1998). However, motor neurons are still generated in the r2 of Hoxa2−/− mutants despite the fact that this is the only Hox gene expressed in that rhombomere (Gavalas et al., 1997). Although the particular subtype of motor neurons generated in the absence of Hoxa2 has not been determined, this result does suggest that Hox genes are not required for motor neuron induction. Furthermore, at more anterior levels of the axis, somatomotor neurons of the trochlear (isthmus) and oculomotor (midbrain) nerves form in the normal absence of Hox gene expression.

We were interested in finding out whether the singular activities of individual Hox genes might be sufficient to impart distinct developmental fates to neuronal precursor cells. Using a direct gain-of-function approach, and avoiding possible complications resulting from the presence of other Hox genes, we analysed the effects of misexpressing Hoxb1 and Hoxa2 in vivo in r1, i.e. anterior to the normal Hox expression domain. Our results show that ectopic Hoxb1 and Hoxa2 activity leads to the generation, respectively, of ectopic facial and trigeminal branchiomotor neurons in an area of the neural tube never normally giving rise to any motor neurons, demonstrating that Hox genes can operate through their individual singular activities to specify the development of distinct neuronal subtypes.

The full coding region of chick Hoxa2 (Prince and Lumsden, 1994) was cloned into the RCAS(BP)B retroviral vector (Hughes et al., 1987) using the adaptor plasmid Slax12 (Morgan and Fekete, 1996). The adaptor plasmid sequence was altered from the ATG of the NcoI site to add the start of the Hoxa2 coding sequence up to the SfuI site by PCR. An sfuI/AatII Hoxa2 fragment was subcloned into Slax12 digested with SfuI/HindIII. Sequences were verified to ensure no PCR errors had occurred. The Slax12/Hoxa2 was cut with ClaI and cloned into the RCAS vector cut with ClaI. The construction of the RCAS/Hoxb1 and RCAN/Hoxb1 vectors will be described elsewhere (E. Bell, R. Wingate and A. Lumsden, unpublished).

Collection of virus and infections into the neural tube or on top of the primitive streak were done according to established protocols (Morgan and Fekete, 1996). Virus titers were between 10⁸ and 10⁹ particles/ml. For each probe or antibody between 5 and 30 infected embryos were stained and analysed.

Whole-mount in situ hybridisation with digoxigenin-and fluorescein isothiocyanate-labelled riboprobes was performed as described (Wilkinson, 1992). For whole-mount antibody stains, embryos were left after blocking in primary antibodies for 3 days (in PBS/10% serum/1% Triton X-100), 4°C, followed after washing by an overnight incubation in secondary antibodies (in PBS/5% serum/1% Triton X-100), 4°C. After washing, embryos were fixed (for fluorescence-coupled secondary antibodies) or developed in DAB (for peroxidase-coupled secondary antibodies). Probes and antibodies were: chick Hoxa2 (Prince and Lumsden, 1994); chick Hoxb1 (V. Prince); mouse Hoxb1 (R. Krumlauf); chick Hoxb2 (A. Kuroiwa); chick Isl2 (Tsuchida et al., 1994); chick BEN (O. Pourquie); anti-ISL1/2 (Thor et al., 1991); anti-BEN (Pourquie et al., 1990; Developmental Studies Hybridoma Bank); anti-mouseHoxb1 (N. Manley and M. Capecchi); anti-neurofilament antibody clone RMO-270 (Zymed). The hindbrains of stained embryos were dissected out, flat-mounted and analysed by bright field and Nomarski microscopy.

Anterograde and retrograde nerve tracing was performed by injecting DiI (6 mg/ml in dimethyl formamide; Molecular Probes) into the ventral neural tube or into the branchial arches of formaldehyde-fixed embryos. After 3-7 days in fix, embryos were analysed by confocal microscopy.

To test whether the activity of a single Hox gene is sufficient to induce the generation of ectopic motor neurons, chick embryos were infected between stages HH5 and HH8 (according to Hamburger and Hamilton, 1951) with an RCAS(BP)B retrovirus expressing mouse (m) Hoxb1. Subsequently, r1 was analysed for the presence of ectopic motor neurons using an anti-ISL-1/2 antiserum. r1 never normally expresses any Hox genes and the only endogenous motor neurons present in r1 are those of the trochlear nucleus, which lie in the most anterior part of r1, in the isthmus (Fig. 1A). Misexpression of Hoxb1 led to the presence of ectopic motor neurons in infected areas of r1 when analysed 2 days after infection (HH17-19), which at these stages were found only in the ventral aspect of r1, immediately lateral to the floor plate (Fig. 1B). 4 days after infection (approximately HH25), the ectopic motor neurons had migrated from their place of birth in ventral r1 to a more lateral position within r1 (Fig. 1C,D), thereby manifesting a migration behaviour characteristic for hindbrain branchiomotor neurons (Fig. 1C) (Heaton and Moody, 1980; Covell and Noden, 1989; Simon et al., 1994) (see below). Ectopic motor neurons in r1 were only generated in infected areas of the neural tube as assessed by simultaneous detection of ISL-1/2 and mHoxb1 mRNA or mHoxb1 protein (see Fig. 2B and not shown). To control for possible non-specific viral effects, infection was performed with a control retrovirus RCAN/Hoxb1, which generates unspliced viral transcripts preventing translation of mHoxb1. Following injection of RCAN/Hoxb1, ectopic motor neurons did not appear and no abnormalities of development were detected (not shown).

To study the axonal projections of the ectopic motor neurons in r1, infected embryos were collected at HH17-19 and stained with an anti-BEN antibody (Pourquié et al., 1990). With the exception of trigeminal motor neurons, which show only a very low level of BEN expression at both the mRNA and protein levels, BEN is strongly expressed on the cell surface of young motor neurons (Guthrie and Lumsden, 1992; Simon et al., 1994). Ectopic motor neurons generated after misexpression of Hoxb1 show a high level of BEN expression (compare left and right r1 in Fig. 2A,B; see below). Three main axonal projection pathways were taken by the ectopic motor neurons located in r1. (1) Axons of ectopic motor neurons grow laterally within r1 and then join the axons of endogenous trochlear motor neurons. The latter are located immediately anterior to the ectopic motor neurons in the ventral isthmus, project dorsally around the circumference of the isthmus and, after decussation, exit from the dorsal aspect of the brain (Fig. 2C,D). (2) Ectopic motor axons join the pathway of trigeminal motor neurons. Trigeminal cell bodies are situated caudal to the ectopic motor neurons, in r2 and r3, and exit the hindbrain via an exit point in lateral r2 (Fig. 2A,C,D). (3) Ectopic projections of fasciculated motor axons project dorsally within r1 without joining either of the flanking endogenous pathways, sometimes forming ectopic nerve exit points in the dorsolateral aspect of r1 (Fig. 2D).

To analyse further the projection patterns of Hoxb1-induced motor neurons, we performed anterograde nerve tracing. Infected embryos were collected at approximately HH25 and the pathways of ectopic neurons present in the ventral aspect of r1 analysed by DiI labelling in the central region of ventral r1. In addition to joining trochlear (Fig. 2E) and trigeminal motor axon pathways (Fig. 2G), as already seen after detection of BEN, ectopic neurons in r1 were found that projected their axons posteriorly as far as r4 and entered the second branchial arch in conjunction with facial axons (Fig. 2F). Ectopic motor neurons in r1 exiting via the trigeminal exit point in r2 were also detected after retrograde tracing from the first branchial arch (Fig. 2H). Moreover, ectopic motor neurons in r1 were found to exit the hindbrain via dorsolateral exit points never normally present in r1 (Fig. 2G). These results therefore demonstrate that in addition to the induction of motor neurons exiting the hindbrain together with other cranial nerves, Hoxb1 activity is also sufficient to instruct the generation of additional nerve exit points.

Depending on the target tissue they innervate, motor neurons in the hindbrain can be classified into three different subtypes: visceromotor, branchiomotor and somatomotor. We sought to determine whether misexpression of Hoxb1 leads to the induction of a particular subtype of motor neuron or rather to the induction of a generic type of motor neuron. The migratory behaviour displayed by the ectopic motor neurons in r1 (Fig. 1D) already indicated that they might be of the branchiomotor subtype. To further test this at the molecular level, we made use of the observation that in the hindbrain only somatomotor neurons but not branchiomotor neurons express the LIM-homeobox gene Isl2 (Varela-Echavarria et al., 1996). When analysed at HH25, somatic trochlear and abducens motor neurons were found to coexpress Isl1 and Isl2, whereas branchiomotor neurons expressed Isl1 but not Isl2, as previously reported (Fig. 3A,B). Ectopic motor neurons in r1 were never found to express Isl2 in addition to Isl1 (Fig. 3A,B), consistent with their being of the branchiomotor subtype. This result demonstrated, therefore, that the activity of Hoxb1 was sufficient to selectively induce the generation of a distinct specified subtype of motor neurons.

Evidence exists for positive cross-regulatory interactions among Hox genes. The effects of misexpressing Hoxb1 might thus be due to the induction of other Hox genes. In the case of Hoxb1, both Hoxb2 and Hoxb1 itself are known targets (Popperl et al., 1995; Maconochie et al., 1997). However, we never observed any induction of Hoxb2 in Hoxb1-infected embryos when assessed 2 days after infection (Fig. 4A,B). Moreover, using a chick-specific probe, we were also unable to detect any induction of chick Hoxb1 by the retrovirally expressed mHoxb1 (Fig. 4C,D). In addition, the expression of Hoxa2 (which is not a target of Hoxb1) was unaffected by the misexpression of Hoxb1 except for a few scattered cells in the lateral aspect of r1 observed in a few embryos which, however, never colocalised with the ectopic motor neurons (Fig. 4E,F). In addition, we did not observe any transient induction of Hoxa2, Hoxb1 or Hoxb2, as assessed 24 hours after infection (data not shown). Therefore, the ectopic branchiomotor neurons found in r1 appear to be generated in response to the singular activity of Hoxb1 itself rather than to the singular or combinatorial activities of other Hox genes.

To find out whether Hox genes other than Hoxb1 might be sufficient to induce the generation of motor neurons in r1, we constructed a recombinant RCAS(BP)B retrovirus driving the expression of chick Hoxa2, which is the Hox gene expressed most anteriorly in the neural tube up to a rostral limit at the r1/r2 boundary. After infection of HH5-8 embryos and collection of the embryos 2 days later (HH17-19), we examined r1 for the presence of ectopic motor neurons by staining for ISL1/2. As in the case of Hoxb1 misexpression, Hoxa2 expression was also sufficient to confer on precursor cells in r1 the competence to form motor neurons, which were initially found in ventral r1 lateral to the floor plate (Fig. 5A).

Embryos infected with RCAS/Hoxa2 were harvested 48 and 72 hours after infection and axonal pathways analysed by BEN immunostaining. The majority of ectopic motor neurons in r1 appeared to project dorsally within r1. In a few cases, we found ectopic motor axons joining the trochlear nerve, as observed after misexpression of Hoxb1. We were unable, however, to detect ectopic motor axons joining any of the other cranial nerves (Fig. 5B).

The projections of ectopic motor neurons generated after misexpression of Hoxa2 were further investigated by retrograde and anterograde tracing from the first branchial arch and r1, respectively. We were unable to find any ectopic motor neurons in r1 exiting the hindbrain in conjunction with trigeminal motor neurons after retrograde labeling (not shown). Moreover, the anterograde labeling did not reveal any axonal projections from r1 exiting the hindbrain along with other cranial nerves nor via ectopic exit points. Occasionally, a few ectopic axons appeared to join the trochlear nerve (not shown).

To determine the subclass of motor neurons generated, we analysed the expression profiles of Isl1 and Isl2 and found that the ectopic motor neurons generated following Hoxa2 misexpression expressed Isl1 but not Isl2. (Fig. 5C,D). Therefore, as with Hoxb1, Hoxa2 misexpression also resulted in the generation of motor neurons of branchiomotor subtype identity, as shown by their lateral migration behaviour and by their Isl gene expression patterns.

In order to see whether misexpression of Hoxa2 would lead to the induction of other Hox genes, expression patterns of Hoxb1 and Hoxb2 were analysed 48 hours after infection. We did not find any changes in the expression patterns of the aforementioned genes (not shown).

Hox genes have been suggested to exert their principal influence within the anterior regions of their expression domains (see McGinnis and Krumlauf, 1992). In the case of Hoxa2, this embraces the area of the hindbrain (r2) giving rise to trigeminal branchiomotor neurons, whereas the anterior domain of Hoxb1 expression encompasses the region of the hindbrain (r4) generating facial branchiomotor neurons (see Lumsden and Krumlauf, 1996). In order to find out whether Hoxa2 and Hoxb1 activities induced the generation of trigeminal-like or facial-like motor neurons, we analysed r1 of infected embryos for the expression of BEN mRNA. BEN is only very weakly expressed by trigeminal motor neurons, in contrast to the strong expression by other motor neurons (Guthrie and Lumsden, 1992; Simon et al., 1994; see also Fig. 6E). Misexpression of Hoxa2 resulted in the generation of ectopic motor neurons with no or a very weak induction of BEN (Fig. 6A,B) (10/10 embryos), whereas the misexpression of Hoxb1 led to a strong induction of BEN expression in ectopic motor neurons in r1 (Fig. 6C,D) (10/10 embryos). Infection with the control virus RCAN/Hoxb1 did not result in any change in BEN expression (Fig. 6E). This result, therefore, suggests that at least to a certain extent, the individual activities of Hoxa2 and Hoxb1 are sufficient to impart on the neural tissue the competence to generate rhombomere-specific branchiomotor neurons of trigeminal or facial type, respectively. This is consistent with the idea of anterior domains of expression as main sites of Hox gene activity and suggests a direct role for Hox genes in determining the specific AP character of hindbrain motor neurons.

By misexpressing individual Hox genes, Hoxb1 and Hoxa2, outside (i.e. anterior to) the endogenous Hox expression domain in vivo, we have demonstrated that Hox genes are capable of conferring on neural precursor cells the competence to generate distinct cell types that are normally formed in the anterior expression domains of specific Hox genes, i.e. trigeminal motor neurons in the case of Hoxa2, and facial motor neurons in the case of Hoxb1. This shows that, at least to a certain extent, the activity of a single Hox gene is sufficient to determine the positional value of neural tissue and demonstrates that a single Hox gene is sufficient to specify the generation of neurons with a distinct subtype. Our data do not, however, rule out the possibility that for cell types other than motor neurons, or under different circumstances, a combinatorial mode of Hox gene action, for example, might be operative (see Krumlauf, 1993). Moreover, our data are consistent with the concept of posterior dominance (Duboule, 1991; Duboule and Morata, 1994; see McGinnis and Krumlauf, 1992, for a discussion), whereby the major site of influence of a specific Hox gene is in the anterior reaches of its expression domain, implying that the identities of cells at a particular axial level are determined by the activity of the most 5′ Hox gene expressed at that axial level.

Following the misexpression of Hoxa2 and Hoxb1, ectopic motor neurons could be detected throughout the entire extent of ventral r1 (Figs 1B, 5A). Cell types in the ventral neural tube have been shown to be induced by the signaling molecule sonic hedgehog, which is initially expressed by the notochord and acts on the overlying neural plate (Tanabe and Jessell, 1996). The particular subtype of ventral neural-tube derived neurons generated in response to sonic hedgehog (e.g. dopaminergic and motor neurons in the midbrain, motor neurons caudal to r1 but not in r1) is dependent on the particular anteroposterior level of the neural axis, whose specification in turn is determined by the patterning genes (e.g. Hox genes, Pax genes) expressed at a given level (Simon et al., 1995; Lumsden and Krumlauf, 1996). Thus, our data demonstrate that single Hox genes are sufficient to confer on precursor cells in r1 the competence to respond to ventralising signals with the generation of specific motor neuron subtypes not normally found in this region. Although ectopic on the anteroposterior axis, motor neuron progenitors are induced at the correct position on the dorsoventral axis of the neural tube, immediately lateral to the floor plate (Tanabe and Jessell, 1996). Following their birth, the ectopic motor neurons then migrate dorsally within the neural tube (Figs 1D, 5D), a behaviour typical for branchiomotor but not for somatomotor neurons (Heaton and Moody, 1980; Covell and Noden, 1989; Simon et al., 1994). The branchiomotor specificity of the ectopic motor neurons in r1 could further be demonstrated by assessing the expression of the LIM genes Isl-1 and Isl-2 (Figs 3, 5C,D) whose expression profiles enable the distinction to be made between the different types of motor neurons found in the hindbrain (Varela-Echavarria et al., 1996). In terms of their axonal projections, the ectopic motor neurons generated after misexpression of Hoxb1 not only exited the neural tube in conjunction with flanking cranial nerves, but also via novel exit points generated in r1. In the case of Hoxa2 misexpression, the majority of the ectopic motor neurons appeared not to exit the neural tube.

Next, we tried to determine whether misexpression of Hoxa2 and Hoxb1 might be sufficient to induce the generation of that particular type of branchiomotor neuron normally found in their most anterior expression domain, trigeminal and facial motor neurons, respectively. Although there are no markers known unequivocally to identify different branchiomotor nuclei, the observed gene expression profiles of Isl-1+/Isl-2− /BENweak following Hoxa2 misexpression (Figs 5, 6) and Isl-1+/Isl-2−/BENstrong following Hoxb1 misexpression (Figs 3, 6) mirror those, respectively, of endogenous trigeminal and facial motor neurons (Guthrie and Lumsden, 1992; Simon et al., 1994; Varela-Echavarria et al., 1996). This therefore suggests that misexpression of Hoxa2 indeed leads to the generation of trigeminal-like branchiomotor neurons and that of Hoxb1 to the generation of facial-like branchiomotor neurons. The activities of Hoxa2 and Hoxb1, therefore, are sufficient to assign AP position-specific neuronal subtypes. In the case of misexpressing Hoxa2, the axonal projection behaviour displayed by the ectopic motor neurons suggests that additional signals might be required for their further differentiation.

Analysis of Hoxb1−/− embryos has shown that even in the absence of Hoxb1, motor neurons are generated in r4. These, however, fail to migrate posteriorly to their proper position (within r6) and later die (Goddard et al., 1996; Studer et al., 1996). Using r2-specific reporter lines and gene markers it has been shown that in Hoxb1−/− embryos r4 appears to adopt an r2-like identity (Studer et al., 1996). Moreover, misexpression of Hoxb1 in r2 (normally only expressing Hoxa2) has been shown to reassign axonal projections of r2 motor neurons from trigeminal (first arch) specificity to facial (second arch) specificity (E. Bell, R. Wingate and A. Lumsden, unpublished). We now show that the singular activity of Hoxb1 rather than a cooperative action of several Hox genes is sufficient and responsible for the induction of facial-like branchiomotor neurons. Considering the remaining Hox genes being expressed in r4 in the absence of Hoxb1, Hoxa1, Hoxb2 (at a reduced expression level; Maconochie et al., 1997) and Hoxa2, it is most likely that the motor neurons generated in r4 in Hoxb1−/− embryos represent trigeminal-like motor neurons rather than facial motor neurons; this would be consistent with the absence of the posterior migration behaviour characteristic for facial branchiomotor neurons (Goddard et al., 1996; Studer et al., 1996; McKay et al., 1997) (see below). As for endogenous trigeminal motor neurons in r2, the motor neurons found in Hoxb1−/− embryos in r4 simply migrate laterally within their rhombomere of origin to condense in proximity to their dorsolateral exit point. The subsequent early death of r4 motor neurons in Hoxb1−/− embryos may result from the failure of these ectopic trigeminal-like motor neurons to encounter and innervate appropriate target tissues in the second branchial arch rather than first arch derivatives appropriate for trigeminal motor neurons, perhaps due to absence or insufficiency of correct target-derived survival factors.

In the case of Hoxa2, it has been reported that despite the absence of Hoxa2, trigeminal motor neurons are generated in r2 (no other Hox gene present) and r3 (expressing Hoxb2), which display defects in pathfinding behaviour, partly exiting the hindbrain from inappropriate exit points as assessed by nerve tracing (Gavalas et al., 1997). These findings might therefore suggest that Hoxa2 is not required to instruct motor neuron generation. Our data, however, demonstrate that Hoxa2 is sufficient to induce the generation of trigeminal-like motor neurons.

In conclusion, the specification of trigeminal and facial motor neurons by Hoxa2 and Hoxb1 activities, respectively, demonstrates that at least to a certain extent and for certain cell types the singular activities of individual Hox genes are sufficient to instruct the generation of distinct cell types.

We thank the following for kindly providing plasmids or antibodies: V. Prince (chick Hoxb1); R. Krumlauf (mouse Hoxb1); A. Kuroiwa (chick Hoxb2); T. Jessell (chick Isl2); O. Pourquie (chick BEN); T. Edlund (anti-ISL1/2); N. Manley and M. Capecchi (anti-mouse-Hoxb1); C. Cepko and D. Fekete (Slax12); S. Hughes (RCAS(BP)B); J. Gilthorpe (RCAN(BP)B). A. Graham read an early version of this paper. S. Jungbluth was supported by an EC fellowship. The Wellcome Trust and the MRC supported the work.