Ectopic expression of Hoxd10 in thoracic spinal segments induces motoneurons with a lumbosacral molecular profile and axon projections to the limb


  • Veeral Shah,

    1. Department of Neurobiology and Center for Neuroscience, University of Pittsburgh, Pittsburgh, Pennsylvania
    Search for more papers by this author
  • Emily Drill,

    1. Department of Neurobiology and Center for Neuroscience, University of Pittsburgh, Pittsburgh, Pennsylvania
    Search for more papers by this author
  • Cynthia Lance-Jones

    Corresponding author
    1. Department of Neurobiology and Center for Neuroscience, University of Pittsburgh, Pittsburgh, Pennsylvania
    • Department of Neurobiology, University of Pittsburgh School of Medicine, Biomedical Science Tower W1442, 3500 Terrace Street, Pittsburgh, PA 15261
    Search for more papers by this author


Hox genes encode anterior–posterior identity during central nervous system development. Few studies have examined Hox gene function at lumbosacral (LS) levels of the spinal cord, where there is extensive information on normal development. Hoxd10 is expressed at high levels in the embryonic LS cord but not the thoracic cord. To test the hypothesis that restricted expression of Hoxd10 contributes to the attainment of an LS identity, and specifically an LS motoneuron identity, Hoxd10 was ectopically expressed in thoracic segments in chick embryos by means of in ovo electroporation. Regional motoneuron identity was assessed after the normal period of motoneuron differentiation. Subsets of motoneurons in transfected thoracic segments developed a molecular profile normally shown by LS motoneurons, including Lim 1 and RALDH2 expression. In addition, motoneurons in posterior thoracic segments showed novel axon projections to two muscles in the anterodorsal limb, the sartorius and anterior iliotibialis muscles. At thoracic levels, we also found a decrease in motoneuron numbers and a reduction in gonad size. These last findings suggest that early and high levels of Hox expression impeded motoneuron development and neural–mesodermal interactions. Despite these adverse effects, our data indicate that Hoxd10 expression is sufficient to induce LS motoneuron identity and axon trajectories characteristic of motoneurons in the LS region. Developmental Dynamics 231:43–56, 2004. © 2004 Wiley-Liss, Inc.


Motoneuron organization in limb-innervating (brachial/lumbosacral) segments of the vertebrate spinal cord is strikingly different from that in the intervening thoracic segments. Understanding how these differences are acquired is an important step in elucidating the development of motor patterns. In brachial and lumbosacral segments, motoneurons are grouped into two spatially discrete columns, a medial motor column (the MMC) containing motoneurons projecting to axial muscles and a lateral motor column (the LMC) containing motoneurons projecting to forelimb or hindlimb muscles. The LMC is further divided into lateral and medial subcolumns (the LMCl and LMCm) of motoneurons that project to dorsal and ventral limb regions, respectively (Landmesser, 1978; Hollyday and Jacobson, 1990). In contrast, thoracic segments lack an LMC, and contain two additional subtypes of motoneurons; sympathetic preganglionic motoneurons of the column of Terni (CT) and MMC motoneurons that project to body wall muscles or axial muscles (Gutman et al., 1993).

Brachial/lumbosacral segments can be distinguished from thoracic segments early because motoneuron subsets develop a characteristic molecular profile shortly after they are born. Proteins expressed in unique patterns include members of the LIM family of transcription factors (Tsuchida et al., 1994), members of the Eph receptor family of axon guidance molecules (Kilpatrick et al., 1996; Iwamasa et al., 1999; Eberhart et al., 2000), and the enzyme RALDH2 (Sockanathan and Jessell, 1998; Berggren et al., 1999). These patterns are critical for the development of appropriate motoneuron organization. For example, brachial/lumbosacral segments can be distinguished from thoracic segments because only the former contain RALDH2+ LMC motoneurons (Sockanathan and Jessell, 1998; Berggren et al., 1999). The expression of RALDH2, a key enzyme in retinoid synthesis, is critical for the development of the LMCl. Studies by Sockanathan and Jessell (1998) indicate that retinoids produced by early born LMCm motoneurons induce the expression of the LIM protein Lim1 in later born LMCl motoneurons. In turn, analyses of axonal projection patterns after ectopic Lim1 expression suggest that Lim1 specifically encodes a dorsal axon trajectory in the limb (Kania et al., 2000). Lim1 might accomplish this by regulating the distribution of Eph receptors family on motoneurons (Helmbacher et al., 1998; Eberhart et al., 2002; Kania et al., 2003).

Experiments in which embryonic spinal segments have been exchanged between brachial and thoracic (Ensini et al., 1998) or lumbosacral and thoracic regions (O'Brien and Oppenheim, 1990; Fukushima et al., 1996; Lance-Jones et al., 2001) indicate that morphological features of motoneuron organization as well as LIM and Eph receptor profiles are determined by early neural tube stages, before the onset of motoneuron differentiation. Even differences between individual segments in a single region, including projection patterns from different lumbosacral (LS) segments (Matise and Lance-Jones, 1996) and local differences in LIM profiles within brachial regions (Ensini et al., 1998), are determined before the onset of motoneuron differentiation.

As evidence of early spinal determination accrued, studies of hindbrain development indicated that Hox genes were central players in the early encoding of anterior–posterior (AP) identity. The vertebrate Hox genes encode transcription factors that are homologues of the Drosophila homeotic gene products, products that direct the development of segment phenotypes (Akam, 1987). There are four families of Hox genes (Hoxa–d) located on separate chromosomes. In each family, there are 13 possible members (i.e., Hoxa1-a13) with specific axial domains along the AP axis (Scott, 1993). Hox paralogues (i.e., Hoxa1, Hoxb1, Hoxd1) occupy parallel chromosome positions and are expressed at similar axial levels (reviewed in McGinnis and Krumlauf, 1992; Krumlauf, 1994; Carpenter, 2002). At hindbrain levels, loss of function studies implicate Hox genes in the initial formation of rhombomere segments, the differentiation of branchiomotor and somatic motor neurons in individual rhombomeres, and the maintenance of unique segmental identities (reviewed in Lumsden and Krumlauf, 1996; Capecchi, 1997). Ectopic expression of single Hox genes (Hoxb1, Bell et al., 1999; Jungbluth et al., 1999; Hoxa2, Jungbluth et al., 1999; and Hoxa3, Guidato et al., 2003) is sufficient to induce the generation of novel motoneuron subtypes and can lead to the development of axon trajectories that match the new Hox identity (Bell et al., 1999, Guidato et al., 2003).

Recent studies in the chick embryo indicate that members of the Hoxc family can program regional characteristics in brachial and thoracic spinal segments (Liu et al., 2001; Dasen et al., 2003). An early combination of signals originating from Hensen's node and paraxial mesoderm induce distinctive patterns of Hoxc expression within motoneuron populations located at different spinal levels (Liu et al., 2001). Hoxc6 expression characterizes motoneurons in brachial segments, while Hoxc9 expression is restricted to motoneurons in thoracic segments. The ectopic expression of Hoxc6 in the thoracic neural tube induces the development of a brachial molecular profile, while ectopic expression of Hoxc9 in the brachial neural tube yields a thoracic profile.

These studies indicate clear parallels between Hox function in assigning rhombomere and brachial/thoracic identities. Analyses in neonatal mice with targeted disruptions of Hox10 paralogues suggest that similar mechanisms operate to distinguish thoracic from lumbosacral regions. Hoxd10 and Hoxa10 are expressed in the lumbosacral region with anterior boundaries of expression near the thoracic/lumbosacral border (see Carpenter, 2002). In single- and double-knockout mice, the LMC is shifted posteriorly, suggesting that the absence of Hoxd10 or Hoxa10 has changed the identity of anterior lumbosacral segments to that of thoracic segments (Riijli et al., 1995; Carpenter et al., 1997; Lin and Carpenter, 2003). However, no descriptions of the molecular profile of motoneurons in mutants have been reported. It is also not known if altered expression of Hox genes affects the AP distribution of motor axons originating from posterior spinal segments. In Hoxa10 mutants, spinal nerves indeed appear to adopt the morphology of the next most rostral nerve (Rijli et al., 1995). In both single and double Hoxd10/Hoxa10 mutants, there are also truncations of sciatic nerve trunks to posterior and distal limb regions (Carpenter et al., 1997; Wahba et al., 2001), but no assessments of the AP distribution of motor axons have been made. In fact, there is little information on AP axon patterning after perturbation of Hox gene expression at any spinal level. Specific motor axon trajectories have only been examined after loss of function of Hoxc8, a gene normally expressed in brachial regions. Hoxc8 inactivation is accompanied by a loss of forelimb-innervating motoneurons and ectopic projections to forelimb muscles from anterior segments (Tiret et al., 1998). No analyses of axon patterns have been made after ectopic expression of Hox genes in the spinal neural tube alone.

The studies described here were undertaken to examine the role of one Hox gene in the development of motoneuron molecular profiles and AP axon patterns at posterior spinal levels. The central hypothesis that we set out to test was that the restricted expression of Hoxd10 plays a role in the attainment of LS spinal identity and specifically LS motoneuron identity. Hoxd10 was chosen because high expression is limited to the lumbosacral region in the chick embryo (Lance-Jones et al., 2001) and because Hoxd10 knockout mice show clear abnormalities in the anatomical development of the LMC and limb nerves (Carpenter et al., 1997; Wahba et al., 2001). We examined the development of motoneuron organization and axon projections at thoracic spinal levels after ectopic expression of Hoxd10 in the thoracic neural tube of the chick embryo. We made two principal observations. Motoneurons in thoracic segments transfected with Hoxd10 develop molecular features characteristic of the normal lumbosacral region and make novel projections to anterodorsal limb muscles.


We used in ovo electroporation to misexpress Hoxd10 at thoracic (T) levels of the chick spinal cord. All electroporations were carried out at stages 12–15 and involved DNA injection into the newly closed T neural tube. Two DNA constructs were used, one that led to the coexpression of Hoxd10 and enhanced green fluorescent protein (Hoxd10/EGFP construct), and a second control construct that led to the expression of EGFP alone. Most embryos were killed at stages 28–29, when distinctive motor column characteristics and peripheral axon patterns can be readily identified but before the normal peak period of motoneuron cell death (Hamburger and Oppenheim, 1982).

Transfected Thoracic Motoneurons Develop LS-Like Patterns of LIM and RALDH2 Expression but Do Not Show Changes in the Expression of Hoxc8

Electroporation of the Hoxd10/EGFP construct resulted in unilateral expression of Hoxd10 mRNA in T spinal segments. In electroporated embryos probed for Hoxd10 mRNA (n = 8), expression levels in lateral regions of T segments were similar to those found in normal LS segments (Fig. 1). In normal chick embryos, unique patterns of LIM and RALDH2 expression as well as Hox expression distinguish T and LS regions. To assess the regional identity of motoneurons in transfected T segments, we examined the patterns of selected LIM proteins, RALDH2, and one Hox gene (Hoxc8) in spinal cord sections from 24 embryos electroporated with Hoxd10/EGFP DNA and 7 embryos electroporated with EGFP DNA alone. Embryos used for analyses showed transfections of 20% or more postmitotic cells in ventral T spinal cord regions.

Figure 1.

Hoxd10 expression in normal embryos (A,B) and embryos electroporated with Hoxd10/enhanced green fluorescent protein (EGFP; C,D). A,C: Stage 24–25 embryos processed by means of whole-mount in situ hybridization. B,D: Transverse sections through the normal stage 29 lumbosacral cord (B) and through the thoracic cord from a stage 28 embryo electroporated at stage 14 (D). In situ hybridization was carried out on sections. Asterisks in A and C indicate approximate levels of sections in B and D. Scale bars = 200 μm.

In the LS region of normal stage 27–29 embryos, Lim 1 expression distinguishes LMCl motoneurons from all other motoneuron subtypes. In T segments, only interneurons are Lim1+ (Tsuchida et al., 1994). We first stained adjacent sections from stage 27–29 experimental embryos individually with antibodies to EGFP and Lim1/2 (Fig. 2A–D). All sections were taken from posterior T levels (T4–T6). On the transfected sides of Hoxd10/EGFP electroporated embryos, Lim1/2+ cells extended ventrally into the motor column region (Fig. 2B, arrow). These Lim1/2+ cells were typically located in a lateral position, like LMCl motoneurons in normal LS segments (compare Fig. 2B,E). Lim1/2+ ventral extensions were not found on nontransfected sides or in T segments electroporated with EGFP alone (Fig. 2C,D). To determine whether ventral Lim1/2+ cells were motoneurons or displaced interneurons, we double-labeled individual sections with antibody combinations for EGFP, Lim1/2, and the pan-motoneuron marker, Isl1/2 (Fig. 2F,G). Numerous Isl1/2+ cells were transfected in somatic motor regions of the spinal cord (Fig. 2F) and ventrally extending Lim1/2+ cells were Isl1/2+ (Fig. 2G, arrows).

Figure 2.

Enhanced green fluorescent protein (EGFP) and LIM patterns in stage 28–29 thoracic (T) segments after electroporation of Hoxd10/EGFP or EGFP at stages 13–14. A,B: Transverse sections through the T cord after electroporation of Hoxd10/EGFP. These adjacent sections were stained for EGFP (A) and Lim1/2 (B). On the transfected side of T segments, Lim1/2+ cells extend ventrally (arrow in B), like Lim1/2+ cells in a normal anterior lumbosacral (LS) section. E: No ventral extension is evident on the nontransfected side. C,D: Transverse sections through the T cord after electroporation with the EGFP construct, stained for EGFP (C) and Lim1/2 (D). No ventral extension of Lim1/2+ cells is present on the transfected side. F,G: Transverse sections from Hoxd10/EGFP electroporated embryos double-labeled with EGFP and LIM antibodies. F: Low magnification view of T cord shows extensive EGFP expression in the somatic motor region. G: High magnification view of somatic motor region (boxed area in F) in an adjacent section from the same embryo. Double labeling with Lim1/2 and Isl1/2 antibodies indicates that Lim1/2+ cells are motoneurons (arrows). H: High magnification view of the somatic motor region from a separate Hoxd10/EGFP electroporated embryo. Double labeling with EGFP and Lim1/2+ antibodies shows an isolated chain of EGFP+/Lim1/2+ cells (arrow). Scale bars = 200 μm in A (applies to A–F), 50 μm in G,H.

To estimate the numbers of Lim1/2+ motoneurons in T segments, we made counts of the total number of Lim1/2+/Isl1/2+ cells in somatic motor regions. In each experimental embryo (n = 10), three sections were chosen for counts based on their position within a single posterior T segment (anterior border, middle, and posterior border of the segment). The number of Lim1/2+ cells in T4–T6 ranged from 4 to 26 (mean = 12.8) per section. This number is considerably less than that found in normal LS sections (mean in LS1 = 67.1; n = 3; see Fig. 2E); however, counts of Isl1/2+/EGFP+ indicated that only approximately half the motoneurons present in somatic motor regions typically were transfected with Hoxd10/EGFP (mean % of Isl1/2+ also EGFP+ = 57%; n = 12). Importantly, ventral extensions of Lim1/2+ cells were found in embryos in which electroporations of Hoxd10/EGFP had been carried out at stage 12 (n = 3), stage 13 (n = 9), stage 14 (n = 7), and stage 15 (n = 5). This observation suggests that the LIM profile of T motoneurons can be changed by Hoxd10 misexpression both during (stage 12) and after (stage 15) the normal period when regional identity is programmed (Matise and Lance-Jones, 1996; Encini et al., 1998). Finally, Lim1/2 expression appeared to occur in a cell-autonomous manner. In sections through T segments where transfections were sparse, isolated clusters of EGFP+ cells in somatic motor regions often expressed Lim1/2 (Fig. 2H). Because the presence of Lim1/2+ cells is a distinguishing feature of LMCl motoneurons, these observations indicate that the LIM profile of some motoneurons in T segments has been changed to a profile resembling that of lateral LS motoneurons.

In normal stage 29 chick embryos, LS LMC motoneurons express RALDH-2, but few, if any, T motoneurons express this enzyme (Sockanathan and Jessell, 1998; Berggren et al., 1999). All Hoxd10/EGFP elec troporated embryos examined (n = 15; stage 13–15 electroporations, stage 28–29 sacrifice) showed RALDH-2 expression on the transfected side of T segments but not on the nontransfected side (Fig. 3A). The examination of sections double-labeled for RALDH2 and Isl1/2 indicated that most RALDH2+ cells were Isl1/2+ motoneurons located at the dorsolateral edge of the ventral spinal cord (Fig. 3B, arrow). Some RALDH2+/Isl1/2− cells were located outside motor column regions (Fig. 3B, arrowhead). As RALDH2 is normally expressed by some nonmotoneurons after stage 29 (Berggren et al., 1999), ectopic Hoxd10 may have initiated premature RALDH2 expression in these cells. Examination of sections double-labeled with RALDH2 and Lim1/2 indicated that most RALDH2+ cells in motor column regions were also Lim1/2+ (Fig. 3C, arrows). These observations suggest that most RALDH2+ cells in transfected T segments show an LMCl phenotype (Lim1/2+) at stage 28–29.

Figure 3.

RALDH2 patterns in stage 29 thoracic (T) segments after electroporation of Hoxd10/enhanced green fluorescent protein (EGFP) at stage 13. A: RALDH2 is expressed in the somatic motor region on the transfected side but not in the corresponding region on the nontransfected side. As in a normal embryo, expression is also present in the roof plate and the meninges. B: High-magnification view of the somatic motor region in section double-labeled with RALDH2 and Isl1/2 antibodies. Many RALDH2+ cells are Isl1/2+ motoneurons (arrow). RALDH2+ cells are also present outside the motor column region (arrowheads in B,C). C: Higher-magnification view of part of the section shown in A (boxed region). In somatic motor regions, most RALDH2+ cells are also Lim1/2+, suggesting a preponderance of LMCl cells. Scale bars = 200 μm in A, 50 μm in B,C.

We showed previously that transplantation of stage 13 tail bud tissue (the remnant of Hensen's node and the primitive streak) adjacent to the thoracic neural tube leads to an induction of Hoxd10 and an associated decrease in Hoxc8 expression (Omelchenko and Lance-Jones, 2003). We examined Hoxc8 expression in T segments electroporated with Hoxd10/EGFP to determine whether ectopic expression of Hoxd10 alone repressed the expression of Hoxc8. As Hoxc8 is normally expressed in brachial and anterior T segments (Ensini et al., 1998), we examined a subset of Hoxd10/EGFP electroporated embryos in which EGFP expression extended to these levels. In no case (n = 15), was a marked reduction of Hoxc8 evident (Fig. 4). This finding is compatible with the conclusions of Dasen et al. (2003) in suggesting that a repressive effect of 5′ Hox genes (posteriorly expressed Hox genes) on more 3′ Hox genes is limited to specific Hox subsets.

Figure 4.

Neural Hoxc8 expression in a stage 29 embryo after electroporation of Hoxd10/enhanced green fluorescent protein (EGFP) at stage 13. A: EGFP expression in an anterior thoracic segment. B:Hoxc8 expression in an adjacent section. Hoxc8 expression is similar on transfected and nontransfected sides of the cord. Scale bar = 200 μm in A (applies to A,B).

Motoneurons Within Hoxd10 Electroporated T Segments Project to Dorsal Limb Musculature

We next asked if ectopic expression of Hoxd10 in T segments affected the target choices of T motoneuron axons. We used retrograde horseradish peroxidase (HRP) labeling in stage 29 Hoxd10/EGFP electroporated embryos to determine whether T motoneuron axons projected to muscles in the hindlimb. In all of these embryos, EGFP labeling was evident in somatic motor regions of posterior T segments (see Fig. 5A,D). In 15 embryos, the sartorius muscle in each limb was injected with HRP. The sartorius is an anterodorsal thigh muscle that is normally innervated by LMCl motoneurons in LS1–2 with a few cells in T7 (Landmesser, 1978). In 11 of 15 cases, HRP+ cells were found outside the normal domain of the sartorius pool on the transfected side. As can be seen in Figure 5B,C, HRP-labeled cells were found in T5–7, in more anterior positions than normal sartorius pools. On the control (nontransfected) side, a few HRP+ cells were found in T7 but not in T5–6. Both control and transfected sides showed HRP+ cells in LS1–2 as well. (These cells are not visible on the transfected side in the section shown in Figure 5B but were evident in an adjacent section.) At higher magnification, we found a conspicuous overlap between the positions of EGFP+ cells and HRP+ cells (see Fig. 5C). Few if any cells appear to be HRP+/EGFP−. HRP+ cells were typically found in a lateral position in T segments, suggesting that these cells correspond to the Lim1/2+ cells described above (see Fig. 6A). We found anterior extensions of the sartorius pool in embryos with electroporations at stages 12–13 (n = 7) and stages 14–15 (n = 4), indicating that axon projection patterns were affected by electroporation before and after the normal period of regional determination (Matise and Lance-Jones, 1996; Ensini et al., 1998).

Figure 5.

The anterior–posterior position of motor pools to hindlimb muscles in stage 29 embryos with ectopic Hoxd10 expression. All photographs are of horizontal sections through the posterior thoracic (T) and anterior lumbosacral (LS) spinal cord. Anterior is at left. A,D: Enhanced green fluorescent protein (EGFP) expression. B,E: Horseradish peroxidase (HRP) staining in the same sections. C,F: Merged higher magnification images. A–C: Sartorius motor pools in an embryo with high EGFP expression in posterior T segments (A). HRP+ cells are present as far anteriorly as T5 on the transfected side of the cord (arrow in B). On the nontransfected side, the sartorius pool begins at T7. HRP+ cells are also visible in LS segments on the nontransfected side. They are not visible in LS segments on the transfected side in this section but were present in adjacent sections. C: A higher-magnification merged image from a section adjacent to that shown in A,B. The boxed area in A shows the approximate position of the region shown in C. D–F: Anterior iliotibialis motor pools in an embryo with high EGFP expression in T (and LS) segments (D). HRP+ cells are present as far anteriorly as T6 on the transfected side (arrow in E). F: A high-magnification merged view of the same section shown in D and E (see boxed region in D). Scale bars = 200μm in A (applies to A,B,D,E), 50 μm in C (applies to C,F).

Figure 6.

Motor pool position in the transverse plane and anatomical nerve patterns in stage 29 embryos with ectopic Hoxd10 expression. A: Horseradish peroxidase–labeled cells (arrow) are located laterally in a thoracic (T) 6 section after injection of the sartorius muscle. B: Enhanced green fluorescent protein (EGFP) expression in peripheral nerves visualized at the time of death. T spinal nerves as well as T7 and lumbosacral (LS) 1 are connected by small proximal branches (arrowhead), providing a possible route for axons from T segments to the femoral nerve (n.) and dorsal hindlimb musculature. Scale bars = 50 μm in A, 500 μm in B.

Similar results were obtained with retrograde labeling of the anterior iliotibialis muscle, a dorsal thigh muscle normally innervated by LMCl motoneurons in LS1–3. In 2/3 cases, the anterior iliotibialis pool on the transfected side included small clusters of cells in posterior T segments as well as cells in anterior LS segments (Fig. 5D–F). As with the sartorius pools, there was a close correspondence between HRP and EGFP staining patterns in transfected segments(see Fig. 5F, arrows). In summary, retrograde labeling data on both sartorius and anterior iliotibialis innervation indicate that motoneurons within Hoxd10 electroporated segments have changed their axon trajectories to project to limb targets.

In normal chick embryos and on the transfected side of embryos electroporated with Hoxd10/EGFP, small proximal nerve branches extend between T spinal nerves and between T7 and LS1 (Fig. 6B). We think that these nerve branches are the most likely pathways for motor axons from posterior T segments to the limb. We found no obvious aberrant nerves in experimental embryos examined in whole-mount for EGFP expression (Fig. 6B) or by means of neurofilament staining. Gross anatomical nerve patterns were also similar on transfected sides of embryos electroporated with the control EGFP construct (data not shown).

Transfection Patterns and Motoneuron Numbers in Hoxd10/EGFP and EGFP Electroporated Embryos

Hoxd10/EGFP electroporated T segments demonstrated two additional features distinguishing them from T segments electroporated with EGFP alone. EGFP expression was largely limited to lateral spinal cord regions in embryos electroporated with Hoxd10/EGFP (Fig. 7A–C). In contrast, EGFP expression was typically widespread in embryos electroporated with EGFP alone (Fig. 7D–F). In Hoxd10/EGFP electroporated embryos, the spinal cord was also smaller on the transfected side than on the nontransfected side. In EGFP electroporated embryos, the two sides were equivalent in size. To assess the size of motoneuron populations in experimental embryos, we first made counts of the numbers of Isl1/2+ cells in somatic motor (SM) regions (Fig. 7G). SM numbers on transfected and nontransfected sides of EGFP electroporated T segments were not significantly different (mean number of SM cells per section on transfected = 102.6% of that on nontransfected side, n = 4). These data indicate that electroporation alone did not affect cell numbers. SM motoneuron numbers were then compared on transfected sides of Hoxd10/EGFP and EGFP electroporated embryos. SM motoneuron numbers were 32.6% lower on transfected sides of Hoxd10/EGFP embryos (n = 12) than on transfected sides of EGFP embryos (n = 7; Fig. 7H). A similar reduction was evident when CT Isl1/2+ populations were compared in Hoxd10/EGFP and EGFP electroporated embryos (% reduction = 31.9%; Fig. 7H). Transfection levels (% of Isl1/2+ cells also EGFP+) were slightly lower in Hoxd10/EGFP electroporated CT regions than in CT regions of EGFP electroporated embryos or in SM regions (Fig. 7I). However, differences between transfection levels were not significant. These finding suggests that Hoxd10 expression did not specifically repress the development of CT motoneurons, a unique feature of T rather than LS segments.

Figure 7.

Enhanced green fluorescent protein (EGFP) extent and motoneuron numbers in stage 28–29 electroporated embryos. A–F: Sections through the thoracic (T) cord from embryos electroporated with Hoxd10/EGFP (A–C) and with EGFP alone (D–F). Adjacent or near-adjacent sections from each embryo were stained for EGFP/ Isl1/2 (A,C,D,F) or Isl1/2 alone (B,E). C and F are higher magnification views of the same sections shown in A and D (see boxes). In the Hoxd10/EGFP electroporated embryo, EGFP expression is lateral (A), Isl1/2+ cells are reduced on the transfected side of the cord (B), and EGFP+/Isl1/2+ cells are positioned laterally in somatic motor regions (C, arrow). In the EGFP electroporated embryo, EGFP expression is widespread (D), similar numbers of Isl1/2+ cells are present on both sides of the cord (E), and EGFP+/Isl1/2+ appear to be evenly distributed in somatic motor regions (F, arrows). G: Isl1/2-stained transverse T cord section showing method used for defining somatic motor (SM) and column of Terni (CT) regions. H: Histogram of mean motoneuron numbers per T6 section in Hoxd10/EGFP and EGFP electroporated embryos. In both SM and CT regions, numbers in the Hoxd10/EGFP group were significantly lower than numbers in the EGFP group (SM region, P = 0.0025, CT region, P = 0.05, unpaired t-tests). I: Histogram of mean % of motoneurons transfected in embryos electroporated with Hoxd10/EGFP and EGFP. No differences are significant. Bars on histograms show standard error (n = 12 Hoxd10/EGFP electroporated embryos, n = 7 EGFP electroporated embryos). Scale bars = 200 μm in A (applies to A,B,D,E,G); 50 μm in F (applies to C,F).

In Hoxd10/EGFP electroporated embryos, low motoneuron numbers and a lateral bias in the position of EGFP+ cells may have resulted because high Hoxd10 impeded the development of early born cells. T and LS motoneurons normally begin to withdraw from the cell cycle at stage 17–18, approximately 6–12 hr after electroporation, and those born earliest settle more medially than those born later (Hollyday and Hamburger, 1977; Prasad and Hollyday, 1991). The survival of more medial motoneurons or their progenitors may have been most severely compromised because these cells went through the fewest cell cycles after electroporation and maintained the highest levels of Hoxd10/EGFP. Nontransfected members of this population would thus predominate in medial ventral horn regions. Given this scenario, one might expect different results in embryos electroporated at early vs. later neural tube stages. We found no consistent differences in the spatial distribution of EGFP+ cells or Isl1/2+ cell numbers in embryos electroporated with Hoxd10/EGFP at stages 12–13 (n = 6) and stages 14–15 (n = 6). It is possible, thus, that additional factors such as abnormalities in settling patterns also affected motoneuron survival.

Ectopic Expression of Hoxd10 in the T Neural Tube Leads to Changes in the Size of the Gonads at Posterior Thoracic Levels

In the course of this study, we noted that the gonads were reduced on the transfected sides of stage 26–29 Hoxd10/EGFP electroporated embryos (Fig. 8A). Gonad size in chick embryos varies with sex and side, but sex differences are not normally apparent at stages 28–29 (Rodemer et al., 1986) and left–right asymmetry does not begin to develop until approximately stage 26 (Carlon et al., 1983). To determine whether gonadal size differences were consistent and specifically associated with ectopic Hoxd10 expression, we measured the cross-sectional area of gonads on right and left sides of a selected group of stage 28–29 Hoxd10/EGFP and EGFP electroporated embryos. All embryos showed transfections on the right side. In each case, area measurements were made on a single section taken at the midpoint of T6. In sections taken from embryos electroporated with EGFP alone, gonadal sizes were similar on right (transfected) and left (nontransfected) sides (mean gonadal area on transfected side/nontransfected side = 0.99 ± 0.9; n = 4). This finding indicates that left–right asymmetry is not evident at the T6 level at stages 28–29. In contrast, gonads in embryos electroporated with Hoxd10/EGFP were significantly reduced on the transfected side (transfected side/nontransfected side = 0.32 ± 0.12; P = 0.0074 unpaired t-test; n = 11). Similar measurements of the area occupied by the mesonephros on transfected vs. nontransfected sides indicated a small but insignificant size difference (corresponding ratio for EGFP electroporated embryos = 0.99 ± 0.6, for Hoxd10/EGFP electroporated embryos = 0.91 ± 0.12). Because we found no evidence of mesodermal transfection (see Fig. 8B), these findings suggest a disruption of interactions between T neuroepithelial and mesodermal tissues.

Figure 8.

A reduction in gonad size is found in Hoxd10/enhanced green fluorescent protein (EGFP) electroporated embryos. A: Transverse section through the trunk, stained with a neurofilament antibody and cresyl violet. In this stage 29 embryo, the gonad is absent on the transfected side at the T6 level (g, gonad). B: High-magnification view of the gonad region (boxed area in A) from an adjacent section stained with an EGFP antibody. Neural crest cells (arrow) but not mesodermal tissues show evidence of transfection. Scale bars = 500 μm in A, 250 μm in B.


We used in ovo electroporation to test the hypothesis that Hoxd10 imparts an LS identity to spinal somatic motoneurons. We made two principal findings. First, the initiation of Hoxd10 expression in T spinal segments at early neural tube stages leads to the development of motoneurons with an LMCl molecular profile normally characteristic of LS segments. Second, motoneuron axons originating in transfected posterior T segments project to anterodorsal limb muscles, in accord with induced molecular identity.

Hox Expression Patterns and the Formation of Motoneuron Subsets in the CNS

In normal embryos, Hoxd10 is expressed at high levels in LS but not T spinal segments. We show that T segments transfected with Hoxd10 contain motoneurons that express Lim1 and RALDH2, two proteins that are normally expressed by LS LMC motoneurons, but not T motoneurons, at the stages examined. These observations suggest that the expression of this single 5′ Hox gene is sufficient to induce motoneurons characteristic of the LS region. These findings add to a large body of data implicating Hox proteins in the encoding of regional identity along the AP axis of the CNS. Most prior evidence comes from analyses of the specification of hindbrain segments and branchiomotor neurons after global disruption of Hox function (see Lumsden and Krumlauf, 1996). Because Hox genes are expressed by many tissues outside the nervous system, several investigators have used techniques that permit selective alteration of Hox expression in the neural tube. Our results are compatible with those obtained at hindbrain levels where ectopic neural expression of Hoxb1 (Bell et al., 1999), Hoxa2 and Hoxb1 (Jungbluth et al., 1999), and Hoxa3 (Guidato et al., 2003) leads to an AP-matched induction of specific branchiomotor or somatic motor populations.

Our observations most clearly parallel data on the roles of Hoxc family members in specifying brachial and thoracic regional identity (Dasen et al., 2003). Hoxc6 is normally expressed in brachial segments of the spinal cord but not thoracic segments. Dasen et al. (2003) show that the ectopic expression of Hoxc6 in the thoracic neural tube leads to an induction of RALDH2+ and Lim1+ motoneurons. Our results extend these observations by demonstrating that Hoxd10 can perform a similar function at posterior spinal levels. Dasen et al. (2003) also report a pronounced and specific decrease in preganglionic motoneurons (CT motoneurons) after ectopic expression of Hoxc6 in thoracic segments. While we found a decrease in CT motoneuron numbers in Hoxd10-electroporated segments, the decrease was not proportionately different from that in somatic motoneuron populations. These observations suggest that Hoxd and Hoxc family members have unique as well as redundant functions in the differentiation of motoneuron subsets in the spinal cord. Our results also complement the findings of Carpenter et al. (1997) after targeted disruptions of Hoxd10 in mice. In these mutants, the LS LMC is shifted posteriorly. Because our results can be interpreted as a complementary anterior shift in the LMC, the phenotype in Hoxd10 mutants may be due primarily to a loss of neural expression rather than to a secondary effect of a loss of expression in mesodermal tissues.

In our experiments, motoneurons with an LMCl phenotype (RALDH2+/Lim1/2+) were induced in T segments ectopically expressing Hoxd10. Whether motoneurons with an LMCm phenotype (RALDH2+/Lim1/2-/Isl1/2+) were also induced is not known. Because we found a reduction in cell numbers on transfected sides of the spinal cord, it is possible that LMCm motoneurons were induced early but did not survive until stage 28–29 (see section below on cell numbers). Nevertheless, the results of two prior studies suggest that Hoxd10 expression leads primarily to LMCl motoneuron development. In vivo and in vitro analyses suggest that the expression of RALDH2 by early born LMCm motoneurons and the subsequent availability of RA induce Lim1 expression in later born LMCl motoneurons (Sockanathan and Jessell, 1998). These investigators hypothesize that normal LMCm motoneurons are not induced to express Lim1 because, by the time sufficient RA is available, LMCm motoneurons are no longer sensitive to RA. In our experimental embryos, Hoxd10 may have induced mainly LMCl motoneurons because it led to a premature induction of RALDH2. A second piece of evidence comes from analyses of peripheral nerve patterns in Hoxd10 knockout animals (Carpenter et al., 1997). The most obvious peripheral nerve defect in these mice is a truncation or depletion of the peroneal nerve, the major nerve trunk innervating dorsal and posterior limb musculature. Because the motor component of the peroneal nerve consists of LMCl motoneurons, these data are compatible with the suggestion that the loss of Hoxd10 leads to a specific depletion of the LMCl.

Axon Pathfinding on the AP Axis

We found that motoneurons in Hoxd10-expressing T segments changed their axon projection identity to match their induced LMCl identity. Retrograde HRP labeling of motoneurons projecting to a major anterodorsal muscle in the thigh, the sartorius muscle, showed the presence of EGFP+/HRP+ in posterior T segments (T5–6) that never normally project to this muscle. Similar results were obtained when the pool to the anterior iliotibialis muscle of the limb was mapped. These data provide direct evidence that the ectopic expression of a single Hox gene can induce a corresponding change in the registration of motoneurons and their target regions on the AP axis. Our findings amplify those of the only prior study that assessed the specificity of axon projections on the AP axis after changing neural Hox expression alone. By using a combination of transplantation and retroviral infection, Bell et al. in 1999 demonstrate that misexpression of Hoxb1 in ventral rhombomere 2 (r2) leads to a posteriorization of r2 motor projections such that they now project to branchial arch 2 (ba 2) rather than ba1.

Because our electroporations frequently resulted in the transfection of neural crest progenitors, it could be suggested that local crest cells were responsible for the novel projections made by T motoneurons. We think this unlikely. Prior studies have not implicated neural crest cells in pathway choices made by spinal motoneurons (Carpenter and Hollyday, 1992). We also found a close correspondence between the positions of EGFP+ and HRP+ cells in the transfected T cord. This finding would not be expected if neural crest cells were the primary determinants of motor axon pathway choice. Finally, in experimental embryos, where transfections were limited to intermediate and dorsal spinal cord regions (n = 3; data not shown), motor pool positions were normal. Thus, our data are most compatible with the hypothesis that Hoxd10 induced a change in motoneuron phenotype directly, rather than by means of an effect on other cell populations.

One question raised by our data is whether Hox genes play a role in the establishment or regulation of guidance cue systems in the periphery. The 5′ Hox genes are expressed in paraxial mesoderm (see Burke et al., 1995; Gaunt, 2000) and lateral (limb bud) mesoderm (see Nelson et al., 1996) where they play well-recognized roles in vertebra patterning (McGinnis and Krumlauf, 1992) and limb skeleton development (Johnson and Tabin, 1997; Zakany and Duboule, 1999). Homeotic transformation of LS vertebra and minor defects in the limb skeleton are found in mice lacking Hoxa10 or Hoxd10 (Rijli et al., 1995; Carpenter et al., 1997; de la Cruz et al., 1999). Most striking is the finding that mice with triple knock-outs of Hox10 genes (Hoxa, d, and c) show a conversion of all LS vertebrae to a T identity and a severe truncation of the proximal limb skeleton (Wellik and Capecchi, 2003). Does the pattern of peripheral Hox expression affect axon pathfinding? Data from studies where peripheral Hox expression has been targeted to a more anterior axial level than normal, suggest that this is the case. When Hox b1 is targeted to the first branchial arch (Bell et al., 1999) and Hoxc6 to cervical mesoderm (Burke and Tabin, 1996), local cranial or spinal nerves are truncated. An intriguing possibility is that axially restricted Hox patterns regulate both the development of motoneuron subtype identity and the distribution of specific guidance molecules like ephrins (see Feng et al., 2000). Alternatively, the effect of peripheral Hox expression on axon pathfinding may be less direct. For example, Hox genes have the capacity to alter Meis expression in the limb. Meis genes have been implicated in the differentiation of proximal limb regions where axon sorting and initial trajectory choices occur (see Capdevila et al., 1999; Mercader et al., 1999).

A second question raised by our findings is whether ectopic expression of Hoxd10 has imparted a general LS identity to T motoneurons or an identity specific for anterior LS segments and anterior limb muscles. The development of limb innervation depends on factors that guide or attract all LMC motoneurons into the limb as well as local cues that govern pathway choice to specific limb regions on the AP axis (see Tosney, 1992; Ebens et al., 1996). We cannot answer this question because we limited our retrograde tracing studies to muscles located in anterior and proximal limb regions. While the mapping of projections to other muscles might be undertaken, the ability of T motoneurons to detect these cues is likely to be adversely affected by distance (see Lance-Jones and Landmesser, 1981). A more informative approach might be to assess motor projections after the ectopic expression of Hoxd11, a Hox gene that is normally expressed only by posterior LS segments (C. L-J., unpublished observations).

Evaluating General Effects of Hox Electroporation and Specific Hoxd10 Effects

We found a marked decrease in motoneuron numbers in spinal segments transfected with Hoxd10. Although we used a DNA concentration similar or lower than that used by others (see Itasaki et al., 1999; Swartz et al., 2001), Hoxd10/EGFP transfected T segments contained approximately 32% fewer Isl1/2+ cells at stage 28–29 than T segments transfected with EGFP alone. Analyses of mice with a targeted disruption of Hoxb13 (Economides et al., 2003) and studies of AbdA gene function in Drosophila postembryonic neuroblasts (see Bello et al., 2003) indicate that 5′ Hox genes and invertebrate homologues can have a negative effect on neuronal cell proliferation and/or activate apoptotic pathways. However, the vast majority of data on Hox gene function indicate a positive effect on the growth of neuronal tissues. For example, overexpression of Hoxa1 leads to the appearance of an expanded hindbrain (Alexandre et al., 1996), while loss of Hoxa1 or other 3′ Hox genes leads to rhombomere deletions and the loss of hindbrain motor nuclei (see for review, Barrow et al., 2000; Gaufo et al., 2003). Similarly, Hoxc8 (Tiret et al., 1998) and Hoxd10/Hoxa10 (Lin and Carpenter, 2003) loss-of-function mutants show decreases in spinal motoneuron numbers. Given these findings, we think it is unlikely that our observations reveal a previously unrecognized function of Hoxd10. Rather, cell survival may have been adversely affected by an abnormally early and high level of Hox expression. In the chick embryo, Hoxd10 is normally expressed at low and diffuse levels at neural tube stages (stages 13–14, see Lance-Jones et al., 2001). It is not until several hours later, when cells begin to leave the cell cycle, that high expression is normally evident.

Transfected cells also may have died because they were located in an abnormal microenvironment. Guidato et al. (2003) induced abducens motoneurons in the anterior hindbrain by means of in ovo electroporation of Hoxa3 but reported a decrease in induced motoneuron numbers between stage 18 and stage 27. Since abducens motoneurons were found only in ventral regions adjacent to the floor plate at stage 27, these investigators hypothesized that induced cells located outside this region lacked adequate factors for survival. In our stage 27–29 Hoxd10-electroporated embryos, most transfected cells with an LMCl profile were located laterally, in a position corresponding to a normal LMC. Are a greater number of transfected motoneurons present at earlier stages and do they show a more widespread or aberrant distribution? We have not specifically assessed motoneuron distribution at early stages; however, examinations of a few Hoxd10-electroporated embryos at stage 18 (n = 3; data not shown) suggest that the number of Isl1/2+ cells is lower on the transfected side than the control side even at this early stage. Thus, a failure to settle in appropriate motor column positions is probably not the single cause for the observed reduction in cell numbers.

We also found a reduction in gonad size in embryos with ectopic Hoxd10 expression. Gonad development may have been impaired in a nonspecific manner associated with the decrease in spinal cord size. For example, decreases in gonad size could be the consequence of deficiencies in early signals from axial structures known to be critical for the patterning of intermediate mesoderm (Mauch et al., 2000; James and Schultheiss, 2003). Alternatively, gonad development may require region-specific signals from the T neural tube or T neural crest. Neural crest-derived sympathetic neurons do not innervate urogenital structures until late in development (Saltis and Rush, 1995); however, a separate neural crest population may contribute to urogenital tissues (Sariola et al., 1989). Loss-of-function studies in mice and studies of humans with genetic defects implicate 5′ Hox genes in the normal development and patterning of the reproductive tract and kidneys (see for reviews, Goodman and Scrambler, 2001, Kobayashi and Behringer, 2003, Patterson and Potter, 2003). Thus, appropriate expression of specific Hox genes in both neuroepithelial and mesodermal tissues may be required for the normal development of urogenital organs.


Experimental Animals

Chick eggs (Charles River) were incubated in a forced-draft incubator for 2–2.5 days when they were removed for in ovo electroporation. After opening, embryos were made visible by dorsal application of a 0.5% neutral red/saline solution and staged by somite number and the Hamburger and Hamilton series (1951). Embryos ranging in age from stage 12 (15–18 somites) to stage 15 (24–27 somites) were chosen for electroporation. Thoracic and lumbosacral levels were identified at stage 14–15 by reference to the adjacent somites or somite-equivalent length of paraxial mesoderm. Definition of axial levels at stage 12–13 was based on distance from the last somite and the tail bud (see Lance-Jones et al., 2001; Omelchenko and Lance-Jones, 2003).

Constructs and In Ovo Electroporation

Experiments were carried out with a construct made by cloning a full-length chicken Hoxd10 (kindly provided by C. Tabin) into a pMES vector. pMES was made by C. Krull by placing the IRES-EGFP sequence from the pIRES2-EGFP (Clontech) into a pCAX (Kobayashi) vector, where EGFP is driven by a chicken b-actin promoter (Swartz et al., 2001; Eberhart et al., 2002). For in ovo electroporation, DNA concentrations and current parameters were chosen based on their ability to give extensive transfection of the thoracic spinal cord with good overall embryonic morphology. Hoxd10-pMES (Hoxd10/EGFP) was microinjected (1.2–2.5 μg/μl) into the lumen of the stage 12–15 neural tube at posterior embryonic levels. Tungsten or gold electrodes (0.5 mm in diameter) were positioned on either side of the thoracic neural tube and a square-pulse electroporator (BTX) used to deliver current (three pulses of 50-msec duration, charging voltage 15–17). Electroporation of pMES (EGFP alone) was used as a control.

Initial Embryonic Assessment, In Situ Hybridization, and Immunohistochemistry

Embryos were killed 2–4 days (stages 24–29) after in ovo electroporation, placed in a cold Tyrode's or Ringer's bath, quickly decapitated, and examined for EGFP fluorescence by using a Nikon inverted fluorescence microscope. Embryos showing high EGFP fluorescence in the thoracic cord were either dissected for HRP axon tracing (see below) or fixed in 4% paraformaldehyde. A few initial embryos were processed by means of whole-mount or section in situ hybridization for Hoxd10. In addition, sections from selected embryos were probed for Hoxc8 (DNA provided by C. Tabin). Digoxigenin-labeled riboprobes were synthesized according to supplier protocol (Boehringer-Mannheim). Whole-mount and section in situ hybridization was performed using standard methods (modifications of Nieto et al., 1996; Schaeren-Wiemers and Gerfin-Moser, 1993, see Lance-Jones et al., 2001; Omelchenko and Lance-Jones, 2003). Most experimental embryos were serially cryostat-sectioned at 14–16 μm on the transverse plane. Every second or third section was placed on a separate slide series for processing with antibodies to EGFP, different regional markers, and/or neurofilament. The following antibodies were obtained commercially or kindly provided by other investigators: EGFP (Molecular Probes), Islet 1/2 and Lim 1/2 (T. Jessell and Developmental Studies Hybridoma Bank, DSHB), RALDH-2 (P. McCaffery), and neurofilament (3A10, DSHB). Both HRP-conjugated secondaries and diaminobenzamidine processing as well as fluorophore-conjugated secondaries and standard immunofluorescence techniques were used (LIM proteins, Tsuchida et al., 1994; RALDH-2, Berggren et al., 1999; neurofilament, modified protocols of Helmbacher et al., 2000, Nowicki and Burke, 2000). All sectioned material was examined and photographed by using a Nikon E600 compound microscope and a Retiga digital camera.

Axon Tracing

Standard retrograde HRP labeling techniques were used (procedures of Landmesser, 1978; Lance-Jones and Landmesser, 1981). Embryos were dissected and a ventral laminectomy performed in an oxygenated Tyrodes bath. After muscle injections of HRP at stage 29, bath temperature was raised to 32° for a 4- to 6-hr incubation. Embryos were fixed in 4% paraformaldehyde and cryostat sectioned on either the horizontal or transverse plane. HRP was visualized by using a fluorescent-conjugated secondary (Jackson Laboratories).

Cell Counts and Area Measurements

Most cell counts were made from photographs of immunostained embryonic sections cut at 14 μm. I.P. Lab software (Scanalytics, Inc., Fairfax, VA) was used to merge fluorescent images of single sections labeled with two antibodies. In selected cases, DAPI counterstaining was used for visualizing cell nuclei. To estimate the number of transfected cells present at the time of death, counts were made of DAPI+ and DAPI+/EGFP+ cells in a rectangular strip of constant width (50 μm) through the ventral spinal cord (-ventricular zone) of representative stage 28–29 experimental embryos. To assess motoneuron survival after electroporation, counts were made of Isl1/2+ cells in somatic motor (SM) and preganglionic motor (CT) regions on transfected sides of experimental embryos. SM and CT regions were defined on photos of fluorescent images as shown in Figure 7G. A line was drawn from the top of the Isl1/2+ cluster in the ventral cord to the top of the Isl1/2+cluster adjacent to the ventricular zone. This line was then bisected with a line to the floor plate, thus delineating ventrolateral SM and a dorsolateral CT regions. Counts of Isl1/2+ cells on transfected and nontransfected sides of embryos electroporated with EGFP alone were made on antibody-stained sections processed for brightfield imaging. For all embryos, counts were made on three nonadjacent sections in T6 and averaged. To assess the size of the mesonephros and gonads in experimental embryos, we measured the area occupied by these tissues in selected transverse sections at the T6 level. Neuro lucida and Neuroexplorer (Microbrightfield, Inc., Williston, VT) were used to trace the perimeter of the gonad and mesonephros and determine area.


The authors thank Peter Land, Laura Lillien, and Paula Monaghan-Nichols for helpful discussions and comments on the manuscript and Emily Sours and Jennifer Phillips for technical assistance. We also thank Cathy Krull for advise on in ovo electroporation.