Hox transcription factors influence motoneuron identity through the integrated actions of both homeodomain and non-homeodomain regions


  • Mala Misra,

    1. Department of Neurobiology, University of Pittsburgh School of Medicine, Pittsburgh, Pennsylvania
    Search for more papers by this author
  • Emily Sours,

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

    Corresponding author
    1. Department of Neurobiology, University of Pittsburgh School of Medicine, Pittsburgh, Pennsylvania
    • Department of Neurobiology, University of Pittsburgh School of Medicine, W1442 Biomedical Sciences Tower, 3500 Terrace Street, Pittsburgh, PA 15261

    Search for more papers by this author


Background: Hox transcription factors play a critical role in the specification of motoneuron subtypes within the spinal cord. Our previous work showed that two orthologous members of this family, Hoxd10 and Hoxd11, exert opposing effects on motoneuron development in the lumbosacral (LS) spinal cord of the embryonic chick: Hoxd10 promotes the development of lateral motoneuron subtypes that project to dorsal limb muscles, while Hoxd11 represses the development of lateral subtypes in favor of medial subtypes that innervate ventral limb muscles and axial muscles. The striking degree of homology between the DNA-binding homeodomains of Hoxd10 and Hoxd11 suggested that non-homeodomain regions mediate their divergent effects. In the present study, we investigate the relative contributions of homeodomain and non-homeodomain regions of Hoxd10 and Hoxd11 to motoneuron specification. Results: Using in ovo electroporation to express chimeric and mutant constructs in LS motoneurons, we find that both the homeodomain and non-homeodomain regions of Hoxd10 are necessary to specify lateral motoneurons. In contrast, non-homeodomain regions of Hoxd11 are sufficient to repress lateral motoneuron fates in favor of medial fates. Conclusions: Together, our data demonstrate that even closely related Hox orthologues rely on distinct combinations of homeodomain-dependent and -independent mechanisms to specify motoneuron identity. Developmental Dynamics 241:718–731, 2012. © 2012 Wiley Periodicals, Inc.


Members of the Hox family of transcription factors are expressed in restricted domains along the body axis and in many developing organs, where they play critical roles in the specification of cellular identity (reviewed in Maconochie et al.,1996; Pearson et al.,2005, Iimura and Pourquie,2007; Guthrie,2007; Dasen and Jessell,2009; Tumpel et al.,2009). In recent years, a wealth of data have shown that these proteins direct the patterning of the nervous system at multiple levels, from the regionalization of the spinal cord and hindbrain to the specification of individual neuronal subpopulations (Bell et al.,1999; Jungbluth et al.,1999; Ensini et al.,1998; Lance-Jones et al.,2001; Dasen et al.,2003,2005; Shah et al.,2004; Wu et al.,2008; Jung et al.,2010). A lingering question is that of how individual Hox factors deliver segment- or region-specific instructions, given that most bind similar DNA recognition sites via a highly conserved homeodomain (reviewed in Gehring,1994; Hoey and Levine,1988; Desplan et al.,1988). Such redundancy suggests that some of the specific downstream effects of Hox expression are dictated by protein:protein interactions involving non-homeodomain regions of Hox transcription factors.

Among the Hox partners known to modify Hox-DNA-binding specificity and to affect segmental patterning in both neural and non-neural tissues are two members of the TALE homeodomain family, Pbx and Meis/Prep (reviewed in Mann and Affolter,1998; Moens and Selleri,2006). Patterning defects observed in mice with mutations in Pbx1 or disruptions in Hox:Pbx interaction domains mimic Hox loss-of-function phenotypes in skeleton, cranial nerve, pharyngeal pouch, and hindbrain (Selleri et al.,2001; Cooper et al.,2003; Medina-Martínez and Ramírez-Solis,2003; Manley et al.,2004; Remacle et al.,2004; reviewed in Moens and Selleri,2006). Meis1 has been implicated in limb, spinal cord, and hindbrain patterning (Mercader et al.,1999; Dasen et al.,2005; Stedman et al.,2009) and is thought to alter Hox function through heteromeric interactions with both Hox proteins and Pbx1 (Knoepfler et al.,1997; Berthelsen et al.,1998). However, despite an abundance of in vivo phenotypic evidence, biochemical studies have raised questions as to the overall relevance of TALE cofactors in Hox DNA target selectivity. For example, Hox:Pbx interactions appear to be more important for the function of Hox factors that are expressed in anterior segments rather than posterior (LaRonde-LeBlanc and Wolberger,2003), suggesting that these posterior factors rely on alternative means to identify and bind specific DNA targets. Further, closely related Hox proteins in complex with Pbx1 all bind with highest affinity to the same DNA recognition sites in vitro (Neuteboom and Murre,1997). Thus, the exact mechanisms by which these cofactors lead to the activation or repression of unique, segment-specific downstream targets remain somewhat mysterious.

The contributions of the homeodomain itself to Hox functional specificity have proven to be equally ambiguous. In vertebrates, in vivo examinations of Hox homeodomain specificity frequently utilize an experimental paradigm in which the homeodomain-coding regions of two Hox genes are reciprocally swapped (Sreenath et al.,1996; Zhao and Potter,2001,2002; Williams et al.,2006; Yallowitz et al.,2009), but results from these swap experiments have varied greatly among experimental systems. For example, mice expressing an altered form of Hoxa11 in which its homeodomain has been replaced with that of Hoxa13 exhibit patterning phenotypes reminiscent of Hoxa13 overexpression models (Zhao and Potter,2001), suggesting that the Hoxa13 homeodomain alone carries with it the capacity to specify some aspects of segmental identity. In contrast, mice expressing a form of Hoxa4 in which the homeodomain has been replaced with that of Hoxc8 show no abnormality in vertebral development (Sreenath et al.,1996), suggesting that the two homeodomains are functionally interchangeable. Intriguingly, several studies have also suggested that direct binding of the homeodomain to DNA is in fact irrelevant to some aspects of Hox function (Caronia et al.,2003; Williams et al.,2005).

A growing body of work confirms the importance of Hox transcription factors in the patterning of motoneuron subtypes within the spinal cord (Dasen et al.,2003,2005; Shah et al.,2004; Wu et al.,2008; Jung et al.,2010). The avian spinal cord is organized into columns of motoneurons that span restricted regions along its anterior-posterior axis. Neurons that innervate axial and body wall musculature compose the medial motor column (MMC), which extends along the full length of the cord. The lateral motor column (LMC) dominates the brachial and lumbosacral (LS) regions of the cord and comprises neurons that project to fore- and hindlimb musculature. It is, in turn, subdivided into lateral (LMCl) and medial (LMCm) divisions containing dorsal-projecting and ventral-projecting neurons, respectively. During motor column development (stages 17–29 of Hamburger and Hamilton,1951), the LMC assembles in an inside-out manner, with later-born motoneurons migrating beyond their earlier-born predecessors and settling in more lateral positions (Fig. 1A). Motoneurons destined for each column can be identified during and after this migratory phase by their expression of members of the LIM-HD family of transcription factors, including Lim1, Isl1, Isl2, and Lim3 (Tsuchida et al.,1994), and the winged-helix/forkhead transcription factor Foxp1 (Dasen et al.,2008; Rousso et al.,2008).

Figure 1.

Hoxd10 and Hoxd11 have opposing effects on motoneuron subtype development but similar DNA-binding domains. A: Cartoon depicting motoneuron migration and differences in motoneuron subtype distribution in anterior and posterior LS segments. B,C: Hox mRNA expression at stage 29. Double arrow indicates LS region. D,E: Hox protein expression in the ventral cord at anterior (LS2) and posterior (LS6) axial levels. Hoxd10 expression is highest in anterior segments, where the LMCl is largest; Hoxd11 peaks in posterior segments, where the LMCl is smaller. F: Diagram of construct used for Hox misexpression. A motoneuron-specific HB9 promoter drives expression of hox-ires-egfp. G–J: Immunostained LS2 sections from stage-24 spinal cords electroporated with Hoxd10 or Hoxd11. Insets show EGFP expression. Asterisks indicate transfected sides of cords and dashed lines outline the spinal cord. G,H: Sections stained with antibodies targeting the LMCl marker Lim1 (red) and the combined pan-motoneuron markers Isl1 and Isl2 (blue). I,J: Sections stained with antibodies targeting the LMC marker Foxp1 (red) and Isl1(2) (blue). K: Alignment of homeodomain sequences from Hoxd10 and Hoxd11. Identical and similar amino acids are highlighted in light and dark gray, respectively. Scale bars = 0.1 mm.

The columnar distribution of motoneurons also varies within individual spinal regions. For example, anterior LS segments contain a sizable LMCl, but this population, as well as the LMC as a whole, diminishes in more posterior LS segments (Fig. 1A). Our prior work attributed the tuning of column size in the LS cord to the opposing actions of two Hox transcription factors, Hoxd10 and Hoxd11. Hoxd10 is expressed at highest levels in LS segments 1–4, where the LMCl is largest; Hoxd11, in contrast, is expressed only in segments posterior to LS4 (Fig. 1B–E). We have previously described the effects of altering Hoxd10 and Hoxd11 expression levels in LS motoneurons (Misra et al.,2009). Overexpression of Hoxd10 in anterior LS motoneurons using in ovo electroporation shifted motoneuron subtype proportions toward anterior lateral phenotypes, identified by their expression of the LIM-HD transcription factor Lim1 and their innervation of dorsal, anterior limb muscles, but did not affect overall LMC size as determined by expression of the LMC-specific marker Foxp1 (Fig. 1G, I). In direct opposition to Hoxd10, ectopic expression of Hoxd11 in anterior LS segments shifted subtype proportions toward Lim1-negative and Foxp1-negative medial phenotypes normally abundant in posterior LS segments (Fig. 1H,J).

The opposing effects of Hoxd10 and Hoxd11 on motoneuron subtype specification are surprising in light of the high degree of amino acid sequence similarity between their DNA-interacting domains; the two homeodomains show 68% amino acid identity and 88% similarity (Fig. 1K). Given this apparent contradiction, we were compelled to investigate the relative contributions of homeodomain versus non-homeodomain regions of these transcription factors to the altered motoneuron phenotypes produced by misexpression. To this end, we constructed hybrid forms of Hoxd10 and Hoxd11 with swapped homeodomains, as well as forms of Hoxd10 with mutations in amino acids thought to be critical for DNA binding or for Pbx interactions. Using in ovo electroporation, we misexpressed these altered Hox proteins in the LS region of the chick spinal cord and evaluated their effects on the development of multiple motoneuron subtypes. Our results suggest that the contrasting functionality of the two transcription factors is determined by the combined properties of both homeodomain and non-homeodomain regions.


Both Homeodomain and Non-Homeodomain Regions of Hoxd10 Are Required for LS LMCl Specification

To begin to assess the role of the homeodomain in motoneuron subtype specification, we first engineered a hybrid form of Hoxd10 in which the homeodomain-encoding sequence was replaced with that of Hoxd11 (Fig. 2A). The chimeric Hoxd10d11HD was then inserted into a bicistronic plasmid vector downstream of the motoneuron-specific promoter Hb9 (Arber et at.,1999; Thaler et al.,1999) and immediately upstream of a sequence encoding an internal ribosomal entry site (IRES) and EGFP (see Fig. 1F).

Figure 2.

The homeodomain of Hoxd10 is necessary for LMCl specification. A: Diagram depicting the mutant protein produced by swapping the homeodomain of Hoxd10 with that of Hoxd11. B: LS2 section from a stage-24 embryo electroporated with Hoxd10d11HD. Immunostaining with anti-EGFP (green) and anti-Isl1(2) (blue, a pan-motoneuron marker) indicates transfection extent. An asterisk marks the transfected side in this and all subsequent figures. C,D: Higher magnification images of non-transfected (C) and transfected (D) motor columns from the same section; anti-Lim1 (red), Isl1(2), and EGFP staining. Regions of Isl1(2)+ cells are outlined. E: Low-magnification image; anti-Lim1 and Isl1(2) staining. F,G: Higher magnification images as in C and D showing Lim1 expression alone. H: Graph depicting mean percentage of Lim1+ motoneurons per section on non-transfected and transfected sides of the cord of embryos electroporated with wild-type Hoxd10 or Hoxd10d11HD. N=6 embryos for each construct, 3 sections each. Error bars indicate S.E.M. in all graphs; ***P < 0.001. Counts of Isl1(2)+ cells and Lim1+, Isl1(2)+ cells were made on images like that shown in C. I: Example of the extent of neural EGFP expression along the anteroposterior axis in a stage-23 experimental embryo. Scale bars = 0.1 mm.

As in previous experiments (Misra et al.,2009), we used in ovo electroporation and an overexpression paradigm to test the effect of this hybrid DNA on motoneuron development. DNA (1.25 μg/μl) was injected into the posterior neural tube of chick embryos at stages 14–16 (Hamburger and Hamilton,1951). In this as well as the following experiments, electroporation resulted in the unilateral expression of EGFP at posterior thoracic and/or LS levels (see example in Fig. 2I). Although the anteroposterior (AP) extent of the transfection could be as much as 8 axial segments, analyses of motoneuron identity were restricted to the LS2 level because this region normally contains a substantial complement of both LMCl and LMCm motoneurons. Analyses were also confined to the early migratory phase of column formation (stages 23–25) as prior studies indicated that ectopic expression of Hoxd10 under the Hb9 promoter is downregulated during late stages of motor column formation.

Triple imunofluorescence antibody staining of tissue sections was used to define electroporation location and extent in the transverse plane (anti-EGFP, Fig. 2B,D), motoneuron identity (anti-Isl1(2) in ventral cord, Fig. 2B–E), and motoneuron LMCl subtype identity (anti-Lim1, anti-Isl1(2), Fig. 2C–E). Examination of sections from Hoxd10d11HD embryos (=embryos electroporated with Hoxd10d11HD) revealed that rather than increasing Lim1+ numbers in a manner similar to wild-type Hoxd10, expression of the chimeric protein reduced Lim1+ numbers (Fig. 2E–G). Counts of Lim1+, Isl1(2) + cells indicated mean LMCl motoneuron numbers of 39±5 on the transfected (experimental) side and 70±8 on the non-transfected (control) side (n=6 embryos, 3 sections per embryo; see Table 1).

Table 1. Quantification of Motoneuron Subtypes in Anterior LS Segments Following Hox Misexpressiona
STAGE 23–25Hoxd10bHoxd11bHoxd10d11HDHoxd11d10HDHoxd10W/QHoxd10IQN/AAA
  • a

    All numbers represent mean values ± SEM per section obtained from counts made in 4-6 embryos (n), with 3 sections counted per embryo. Lim1, Foxp1, and Lim3 values represent motoneurons as identified by coexpression of Isl1/2, with the exception of Hoxd11d10HD embryos in which Lim3+ motoneurons were identified by coexpression of Isl1 alone.

  • b

    Includes some counts previously published in Misra et al.,2009.

  • c

    nt indicates the non-transfected side of the spinal cord, t indicates the transfected side.

  • d

    P values were determined by paired t-tests comparing non-transfected and transfected sides. n.s., P > 0.05 (not significant); *P < 0.05; **P < 0.01; ***P < 0.001.

Isl1(2)-ntc156 ± 8172 ± 9190 ± 8181 ± 7161 ± 6187 ± 6
Isl1(2)-t153 ± 6152 ± 10176 ± 9168 ± 6142 ± 4177 ± 6
P valuedn.s.*********
Lim1-nt37 ± 446 ± 570 ± 841 ± 561 ± 341 ± 3
Lim1-t45 ± 318 ± 239 ± 526 ± 344 ± 351 ± 2
P value*****************
Foxp1-nt123 ± 6152 ± 8140 ± 13108 ± 4  
Foxp1-t117 ± 685 ± 5128 ± 1078 ± 4  
P valuen.s.***n.s.***  
STAGE 29      
Lim3-nt 27 ± 4 28 ± 2  
Lim3-t 35 ± 2 39 ± 2  
P value ** ***  
N 6 5  
Chx10-nt 54 ± 2 56 ± 2  
Chx10-t 70 ± 5 64 ± 3  
P value ** **  
N 4 5  

It should be noted that we found substantial variability in the size of Isl1(2)+ populations in this and subsequent experiments. For example, on the control sides of Hoxd10d11HD embryos, the mean number of Isl1(2)+ cells per embryo ranged from 148–224 (n=6 embryos). Differences in embryonic age may account for this variability, as the motor columns are forming between stages 23 and 25. Numbers also varied among sections from single embryos: the greatest mean range among 3 sections from a single embryo was 34±7 Isl/2+ cells. This finding is likely to reflect differences in precise section position within the LS2 region, as the motor columns normally expand between the beginning and end of LS2. Finally, overall motor column size was often slightly smaller on transfected versus non-transfected sides (see Fig. 2C,D and Table 1). We, therefore, represent Lim1+, Isl1(2)+ motoneurons as a proportion of total Isl1(2)+ cells on the transfected or non-transfected sides of each transverse section. As shown in Figure 2H, the proportion of Lim1+ motoneurons in LS2 of Hoxd10d11HD embryos decreased from a mean of 35 to 22% on the transfected side when compared to the non-transfected side. This represents a 37% decrease compared to the 26% increase found in wildtype Hoxd10 electroporated embryos.

These findings suggest that the native Hoxd10 homeodomain is required for the specification of LMCl fates in anterior LS segments by Hoxd10. Interestingly, a reciprocal chimeric protein containing the homeodomain of Hoxd10 flanked by the non-homeodomain regions of Hoxd11 (Fig. 3A) produced a similar effect, reducing Lim1+ motoneuron proportions from 22 to 15%, a 32% decrease (Fig. 3B–D and I). While the observed repressive effects may be due to the influence of the included portions of Hoxd11 or to other factors (see percentages from wildtype Hoxd11 embryos in 3I and Discussion section), the absence of sufficient LMCl induction suggests that both homeodomain and non-homeodomain regions of Hoxd10 are required for its role in LMCl development.

Figure 3.

Non-homeodomain regions of Hoxd10 are needed for LMCl specification. A,B: Schematic of protein and LS2 section from stage-24 embryo electroporated with Hoxd11d10HD, which encodes the reciprocal homeodomain-swapped version of Hoxd11. Immunostaining for Lim1 and Isl1(2). C,D: Lim 1+ cells within outlined motor column regions on non-transfected and transfected sides. E–H: Schematic of protein and LS2 section from a stage-24 embryo electroporated with Hoxd10W/Q that encodes a mutation likely to block Hoxd10-Pbx1 interactions. Immunostaining as in B–D. I: Graph depicting mean percentage of Lim1+ motoneurons per section on non-transfected and transfected sides of the cord in embryos electroporated with wild-type Hoxd10, Hoxd11, or mutant constructs. N=6 embryos for each construct, 3 sections each; ***P < 0.001 (see Table 1). Scale bars = 0.1 mm.

A number of prior studies have demonstrated that the Hox cofactor Pbx1 contributes to Hox-directed patterning (reviewed in Moens and Selleri,2006). Pbx1 is expressed throughout the AP axis of the cord (Dasen et al.,2005) and interacts directly with Hox paralogues 1–8 via the Hox hexapeptide motif, a conserved six–amino acid sequence situated N-terminal to the homeodomain (Chang et al.,1995; Knoepfler and Kamps,1995), though an additional linker region between the hexapeptide and homeodomain may also be important (Merabet et al.,2003). Hox paralogues 9 and 10 are also capable of interacting with Pbx1 through a single conserved tryptophan residue (Shen et al.,1997). Given the apparent requirement for non-homeodomain regions in Hoxd10 function, we next investigated the significance of Hox-Pbx1 associations in the process of LMCl specification. The relevant tryptophan residue was mutated to glutamine (Q) (as in Shen et al.,1997), and the resulting construct was electroporated into anterior LS segments. Hoxd10W/Q misexpression resulted in an 18% decrease in Lim1+ LMCl motoneuron proportions, again in direct contrast to the LMCl-inducing properties of wild-type Hoxd10 (Fig. 3E–I, see also Table1). This finding strongly suggests that Hox–Pbx1 interactions are necessary for LMCl specification.

Hoxd10-DNA Interactions Are Not Essential for LMCl Specification

Several previous studies have suggested that direct binding of the Hox homeodomain to DNA is not required for some aspects of Hox function (Caronia et al.,2003, Williams et al.,2005). To test the relevance of direct Hox–DNA interactions to LMCl motoneuron specification, we engineered a mutant Hoxd10 protein in which three amino acids—isoleucine (47), glutamine (50), and asparagine (51)—thought to be critical for interactions between the third α helix of the homeodomain and the target DNA sequence (Gehring et al.,1994) were replaced with alanines (Fig. 4A). Before testing the effects of Hoxd10IQN/AAA in vivo, we used in vitro luciferase assays to examine its affinity for the HCR, a 92-bp regulatory element derived from the Hoxd9 promoter that is capable of mediating transcriptional activation by Hoxd10 (Zappavigna et al.,1991; Caronia et al.,2003). The graph in Figure 4B depicts luminescence of lysed transfected Neuro2A cells as a ratio of the signal from firefly luciferase, an indicator of transcriptional activation via HCR binding, to the signal from renilla luciferase, an indicator of baseline transfection levels, in the presence of Hoxd10 or Hoxd10IQN/AAA. Hoxd10 increased the ratio of firefly to renilla luciferase levels several fold, confirming its ability to activate transcription by binding the HCR; Hoxd10IQN/AAA, however, produced no quantifiable effect on firefly luciferase transcription. These results confirmed that, as with previously investigated Hox family members, mutation of homeodomain amino acids 47, 50, and 51 is sufficient to block transcriptional activation by Hoxd10, suggesting that it interferes with DNA binding.

Figure 4.

Direct Hoxd10 homeodomain–DNA interactions are dispensible for LMCl specification. A: Diagrams of constructs used for luciferase assays to test binding of Hoxd10 and Hoxd10IQN/AAA, a theoretical non-DNA-binding mutant, to the HCR element (Zappavigna et al.,1991; Caronia et al.,2003). Expression of firefly luciferase (luc2) is driven by Hox binding the HCR; expression of renilla luciferase (hRluc) is constitutively active. B: Graph depicting transcriptional activation as a ratio of firefly to renilla luminescence. Co-transfection of cultured Neuro2A cells with 165 or 300 ng CMV-Hoxd10 activates luc2 transcription above baseline levels; CMV-Hoxd10IQN/AAA does not. C: Graph showing mean percentage of motoneurons expressing Lim1 on transfected versus non-transfected sides of the cord following electroporation with wild-type Hoxd10 or Hoxd10IQN/AAA. N = 6 embryos for each construct, 3 sections each; ***P < 0.001 (see also Table 1). D: Representative LS2 section from a stage-24 embryo electroporated with Hoxd10IQN/AAA and stained with antibodies targeting Lim1 and Isl1(2). E,F: Lim1 expression in non-transfected (D) and transfected (E) motor columns. Scale bars = 0.1 mm.

After confirming the molecular properties of Hoxd10IQN/AAA in vitro, we examined its in vivo effects on motoneuron subtype development. To our surprise, expression of Hoxd10IQN/AAA in anterior LS segments resulted in an increase in Lim1+ motoneurons similar in magnitude to the effect of overexpression of wild-type Hoxd10 (Fig. 4C–F, Table 1). Therefore, though the native homeodomain of Hoxd10 is required for its contribution to LMCl specification, direct DNA binding by the third α helix of the homeodomain may not be essential.

Expression of Foxp1 Is Regulated by Non-Homeodomain Regions of Hoxd11

As mentioned above, endogenous Hoxd11 levels are highest in posterior LS segments. These segments possess a smaller number of LMC motoneurons than anterior LS segments, and our previous work has suggested that the presence of Hoxd11 contributes to these diminished proportions. Ectopic expression of Hoxd11 in anterior LS segments decreases the number of motoneurons expressing the LMC marker Foxp1. Overexpression of Hoxd10 in these segments, in contrast, does not appear to affect LMC numbers. We wished to dissect the relative contributions of the Hoxd11 homeodomain and non-homeodomain regions to Foxp1 repression. To do so, we examined the effects of expressing the previously described chimeric proteins Hoxd10d11HD and Hoxd11d10HD on LMC numbers in anterior LS segments. Expression of Hoxd10d11HD did not affect Foxp1+ numbers or proportions of total motoneurons (Table 1, Fig. 5A–C and G). In contrast, the expression of Hoxd11d10HD reduced both numbers and proportions, having a similar but slightly smaller effect than ectopic wild-type Hoxd11 (Table 1, Fig. 5D–G). This finding suggests that the specific LMC-regulating functions of Hoxd11 are regulated by non-homeodomain regions of the transcription factor.

Figure 5.

Repression of LMC fate by Hoxd11 is guided by non-homeodomain regions. A–C: LS2 section from a stage-24 Hoxd10d11HD embryo immunostained with antibodies targeting the LMC marker Foxp1 and Isl1(2). Images in B and C are higher magnification views of Foxp1 staining in non-transfected (B) and transfected (C) motor columns (outlined). D–F: Equivalent images from a stage-24 Hoxd11d10HD embryo. G: Graph depicting mean percentage of motoneurons per section expressing Foxp1. N= 6 Hoxd10, 4 Hoxd11, 4 Hoxd10d10HD, and 6 Hoxd11d10HD embryos; ***P < 0.001 (see also Table 1). Scale bars = 0.1 mm.

The Homeodomains of Hoxd10 and Hoxd11 Are Functionally Redundant in the Context of Hoxd11

Given the surprising ability of Hoxd11d10HD to recapitulate the effects of ectopic wild-type Hoxd11 by repressing both Lim1 (Fig. 3B–D) and Foxp1 expression (Fig. 5D–F), we next investigated the impact of Hoxd11d10HD on the expression of other direct or indirect targets of Hoxd11 in the LS cord. Prior studies indicated that the expression of Hb9::Hoxd11 is maintained throughout column development (through stage 29) and documented changes in marker expression within specific motoneuron and interneuron populations at this stage. Thus, we focused on the assessment of the effects of ectopic Hoxd11d10HD at stage 29. Chx10 is expressed by V2a neurons, a class of excitatory glutamatergic interneurons that provide input to both motoneurons and contralateral inhibitory interneurons (Al-Mosawie et al.,2007; Crone et al.,2008). In embryos electroporated with Hb9::Hoxd11, the number of Chx10-expressing cells was increased, despite an apparent restriction of ectopic Hoxd11 to motoneuron populations. As in Hoxd11-electroporated embryos, anterior LS sections from Hoxd11d10HD embryos indeed presented increased numbers of Chx10-expressing cells at stage 29 (Fig. 6A,D; Table 1). The majority of these cells (mean=96.5%, n=5 embryos) again appeared to be EGFP-negative (Fig. 6B,C), suggesting an underlying indirect, non-cell-autonomous pathway.

Figure 6.

Positive and negative functions of Hoxd11 are guided by non-homeodomain regions. A: LS2 section from a stage-29 Hoxd11d10HD embryo, stained with the interneuron marker, anti-Chx10. Spinal cord and central canal outlined with dashed lines. B,C: Images from the transfected side of the same section showing EGFP and Chx10 expression. Box in B corresponds to image in C. Arrow marks a double-labeled cell. D: Graph showing mean numbers of Chx10+ cells on transfected versus non-transfected sides. N= 5 embryos, **P < 0.01, (see also Table 1). E: Anterior LS section from a stage-29 Hoxd11d10HD embryo, stained with anti-Lim3, a MMCm marker. F: Image from the transfected side of the same section showing overlap of EGFP and Lim3 expression in ventral cord regions. G,H: Non-transfected (G) and transfected (H) motor columns from a nearby section, stained with anti-Lim3 and anti-Isl1, a marker of LMCm and MMC motoneurons. I: Graph depicting mean numbers of Lim3+, Isl1+ motoneurons on transfected versus non-tranfected sides. N= 5 embryos; ***P < 0.001. J-M. LS2 sections from stage-29 (J,K) and stage-24 (L,M) Hoxd11d10HD embryos, stained with anti-Hoxd10 and anti-EGFP. Scale bars (C, M) = 0.05 mm; all others = 0.1 mm.

Another potential target of Hoxd11 is Lim3, a LIM-HD transcription factor expressed at stage 29 by MMC motoneurons that innervate axial muscles (MMCm motoneurons) as well as by V2a interneurons (Tsuchida et al.,1994; Sharma et al.,1998). Lim3 is known to suppress the expression of LMC markers including Foxp1 and Lim1 and, in turn, to promote the expression of MMCm characteristics (Sharma et al.,1998; Dasen et al.,2008; Rousso et al.,2008). Our previous work determined that the proportion of Lim3-expressing motoneurons was increased at stage 29 in the presence of ectopic Hoxd11. In the current studies, we used double labeling with anti-Lim3 and an anti-Isl1 antibody (different from the anti-Isl1(2) antibody used in previous experiments) to distinguish between V2a interneurons and motoneurons in anterior LS sections from stage-29 Hoxd11d10HD embryos. In normal embryos, anti-Isl1 recognizes LMCm and MMC motoneurons but not LMCl (see Fig. 6G and Tsuchida et al.,1994), allowing us to examine Lim3 expression only in non-LMCl motoneurons. Counts revealed increases in Lim3-expressing motoneurons on transfected sides, similar to those found in embryos with ectopic Hoxd11 (see Fig. 6H,I and Table 1). The proportion of Isl1+ cells per motor column region increased in parallel (44.1%, transfected side; 33.7% non-transfected side) but there was not a significant difference in the total numbers of Isl1+ cells on the two sides (89±4, transfected side; 83±4, non-transfected side; n=5 embryos) raising the possibility of a selective loss of Isl2+ LMCl motoneurons on the experimental side (see Discussion section).

In contrast to Chx10 and EGFP, we found considerable overlap between anti-EGFP and anti-Lim3 staining in more medial motor column regions where Isl1+ (LMCm and MMC) motoneurons are normally located (Fig. 6F). As with stage-29 Hoxd11-electroporated embryos (Misra et al.,2009), Lim3+ motoneurons were also more scattered on the transfected side when compared to the non-transfected side. On the control sides, Lim3+, Isl1+ (yellow/orange) and Lim3-, Isl1+ (green) cells formed conspicuous clusters, while on the experimental sides, such clusters were not obvious (Fig. G,H). These observations might reflect a change in Hox-mediated influences on adhesion/guidance molecule function (see Misra et al.,2009). It should be recognized that Lim3 is normally expressed by all motoneurons at early stages of differentiation (Shirasaki and Pfaff,2002; Briscoe and Novitch,2008). Thus, we cannot rule out the possibility that ectopic expression of Hoxd11 or Hoxd11d10HD delayed the maturation of some motoneurons.

Finally, we examined the expression of endogenous Hoxd10 protein in the presence of Hoxd11d10HD, as our prior observations suggested that wild-type Hoxd11 cell-autonomously represses Hoxd10 expression. Here, immunofluorescence staining targeting non-homeodomain regions of Hoxd10 was carried out at stage 24 as well as stage 29. Hoxd11d10HD, like Hoxd11, repressed endogenous Hoxd10 (Fig. 6J,L). Few, if any, Hoxd11d10HD-expressing cells co-expressed Hoxd10 during motoneuron columnar specification (stage 24; Fig. 6M) and after column formation was complete (stage 29; Fig. 6K). Taken together, the findings of similar changes in Chx10, Lim3, and Hoxd10 expression in Hoxd11 and Hoxd11d10HD electroporated embryos suggest a remarkable degree of functional homology between the homeodomains of Hoxd10 and Hoxd11 when placed in the context of the non-homeodomain regions of Hoxd11.


Hox proteins are thought to play a dominant role in defining spinal motoneuron subtype by inducing the expression of subtype-specific transcription factors (Dasen et al.,2003,2005; Wu et al.,2008; De Marco Garcia and Jessell,2008; Jung et al.,2010; reviewed in Carpenter,2002; Dasen and Jessell,2009). Yet, the molecular mechanisms whereby individual Hox proteins initiate and maintain unique motoneuron identities remain ambiguous, particularly in light of the high degree of homology among Hox homeodomains and their DNA recognition sequences (Gehring et al.,1994; Mann,1995).

Differing Requirements for Homeodomain and Non-Homeodomain Regions in Two Closely-Related Hox orthologs

Our previous work established that two Hox transcription factors, Hoxd10 and Hoxd11, act in opposition to determine the number and position of LMCl motoneurons in the avian LS spinal cord (Shah et al.,2004; Misra et al.,2009). We found that misexpression of Hoxd10 in the LS cord induced ectopic development of lateral, dorsal-projecting motoneurons (LMCl), while misexpression of Hoxd11 suppressed their development and instead promoted the specification of medial motoneuron subtypes that project to ventral and axial muscles (LMCm and MMC). The distinct and quantifiable effects of Hoxd10 and Hoxd11 presented a straightforward phenotypic assay to gauge the relative contributions of homeodomain-dependent and -independent interactions to Hox functional specificity in the context of spinal motoneuron development.

In the present study, we have used in ovo electroporation to express mutant forms of Hoxd10 and Hoxd11 in the LS spinal cords of embryonic chickens and then quantified motoneuron subtype proportions using immunological staining for subtype-specific markers. We first expressed chimeric forms of the two transcription factors in which the homeodomain regions were swapped. The resulting changes in motoneuron complement suggested that the ability of Hoxd10 to induce the development of LMCl-type motoneurons relies on both the specific properties of its native homeodomain and the molecular features of non-homeodomain regions, including a non-homeodomain tryptophan residue known to be essential for protein:protein interactions with the Hox cofactor Pbx1 (Shen et al.,1997). Neither the chimeric proteins Hoxd10d11HD and Hoxd11d10HD nor the tryptophan mutant Hoxd10W/Q was capable of increasing LMCl cell numbers. In contrast, the effects of Hoxd11 appear to be guided by non-homeodomain regions of the protein. Expression of Hoxd11d10HD mimicked Hoxd11 overexpression phenotypes by decreasing proportions of motoneurons expressing Lim1, Foxp1, and endogenous Hoxd10 while increasing Lim3- and Chx10-expressing cell numbers. Therefore, these closely-related Hox orthologs likely rely on distinct molecular mechanisms to drive motoneuron subtype specification.

Homeodomain-DNA Interactions During LMCl Motoneuron Specification

The observation that Hoxd10 function during LMCl specification requires its native homeodomain is consistent with several recent high-resolution analyses of homeodomain-binding preferences, which predict that even closely-related Hox proteins may prefer slightly different recognition sequences (Noyes et al.,2008; Berger et al.,2008). Many early studies of Drosophila homeotic selector genes pointed to residues in the unstructured N-terminal arm of the homeodomain as key determinants of DNA-binding specificity capable of modulating “generic” homeodomain–DNA interactions between the third α helix of the homeodomain and AT-rich sequences (Lin and McGinnis,1992; Ekker et al.,1992; Chan and Mann,1993; Furukubo-Tokunaga et al.,1993; Zeng et al.,1993; Phelan et al.,1994). More recent structural analyses have confirmed this conclusion and further suggested that the modulation of binding specificity is governed by interactions between N-terminal arm residues of the homeodomain and the DNA minor groove at specific recognition sequences (LaRonde-LeBlanc and Wolberger,2003; Joshi et al.,2007,2010).

While the Hoxd10 homeodomain appears to be necessary for LMCl specification, it is not sufficient. Our analyses also implicate a non-homeodomain element, a tryptophan residue important for interactions with Pbx1, as a crucial requirement for Hoxd10 function. This finding is consistent with structural analyses suggesting that the formation of Hox/cofactor complexes is essential for paralogue-specific interactions between the homeodomain and the minor groove (Joshi et al.,2007,2010). The same studies further suggested that the presence or absence of cofactors might determine whether Hox binding leads to activation or repression of the target gene. Interestingly, we found that misexpression of the chimeric protein Hoxd10d11HD and Hoxd10W/Q, a mutant lacking a tryptophan residue essential for Hox:Pbx interactions, both decreased the proportion of cells expressing Lim1. We hypothesize that these ectopic factors, which are capable of forming “generic” homeodomain–DNA interactions but not specific Pbx1-stabilized homeodomain–DNA interactions, may be exerting a “dominant negative” effect by competing with endogenous Hoxd10 for binding sites and subsequently repressing downstream targets (see also LaRonde-LeBlanc and Wolberger,2003).

The finding that Hoxd10IQN/AAA, a construct bearing mutations in three amino acids critical for Hox:DNA interactions, was incapable of activating transcription in vitro but able to recapitulate the effects of Hoxd10 overexpression in vivo was surprising. Results from the homeodomain-swapping experiments described above suggest that the Hoxd10 homeodomain is required for LMCl specification, but this requirement is apparently unrelated to its ability to bind DNA via the third α helix. This outcome raises an intriguing question regarding the importance of canonical homeodomain–DNA interactions for Hox function. One possible explanation is that Pbx1-stabilized interactions between the N-terminal arm of the homeodomain and the minor groove are sufficient to mediate the activation of Hoxd10′s normal transcriptional targets. Alternatively, the homeodomain may contribute to protein–protein interactions with other factors that in turn control transcription, e.g., collaborative transcription factors like Pbx1, Meis1, Pax2, and Eya1 (reviewed in Moens and Selleri,2006; Gong et al.,2007). Indeed, a number of previous studies have introduced the possibility of non-transcriptional roles for Hox proteins in segmental patterning. For example, misexpression of a mutated, non-DNA-binding form of Hoxd13 in the limb partially mimicked the effects of ectopic wild-type Hoxd13, causing a shortening of the proximal long bones (Caronia et al.,2003, Williams et al., 2005b). In addition, microarray studies have identified a number of factors whose expression is upregulated by both Hoxd13 and the non-DNA-binding form thereof (Williams et al.,2006).

In contrast to Hoxd10, Hoxd11 function appears to be governed entirely by non-homeodomain regions of the protein; substitution of its native homeodomain with that of Hoxd10 had no detectable effect on Hoxd11 function in the context of motoneuron subtype specification. The ectopic expression of Hoxd11d10HD, like that of Hoxd11, led to a decrease in Lim1+ motoneurons, an increase in Lim3+ motoneurons, and an overall decrease in Hoxd10 expression within the motor column. Indeed, the ability of Hoxd11d10HD to repress Lim1 may follow directly from its repression of Hoxd10. With both constructs, we also noted an increase in Chx10+ V2 interneurons, but these cells were located primarily above the motor columns and very few expressed EGFP, a marker of transfection, at the stage examined (stage 29). While this finding raises the possibility of non-cell-autonomous effects, an early cell-autonomous influence might also be considered; some motoneuron progenitors may have been reprogrammed to adopt a V2 interneuron identity but lost EGFP expression by stage 29. Regardless, the observed effects of Hoxd11 and Hoxd11d10HD on interneurons are consistent with prior studies showing that the misexpression of Hox in the developing hindbrain leads to changes in the expression of Irx3 (Guidato et al,2003), a protein recognized as important in the specification of V2 interneurons (see for review Briscoe and Novitch,2008).

The functional equivalence of Hoxd11 and Hoxd11d10HD in spinal neuron specification is similar to that previously observed in axial skeleton, kidney, and branchial arches following misexpression of certain Hox factors with non-native homeodomains (Sreenath et al,1996; Zhao and Potter,2002; Yallowitz et al.,2009). Hoxd11 differs from Hoxd10 in that it belongs to a sub-cluster of posterior 5′ Hox genes (Hox11–13) that do not interact with the cofactor Pbx1 (Shen et al.,1997), and lacks the critical tryptophan residue required for such interactions. Hoxd10 and Hoxd11–13 also show distinct regulatory roles in both kidney and limb morphogenesis (reviewed in Di-Poi et al.,2007; Zakany and Duboule,2007), raising the question of whether they act via similarly divergent mechanisms within these organs.

Identifying Downstream Targets of Hoxd10 and Hoxd11 During Motoneuron Specification

It is important to note that changes in motoneuron subtype proportions induced by Hox misexpression might arise either from the selective death of one subtype or from the specific conversion of one subtype to another. In embryos overexpressing Hoxd10 or Hoxd10IQN/AAA, Lim1+ motoneuron numbers increase despite small decreases in total motoneuron counts, supporting the latter hypothesis. Similarly, in Hoxd11 and Hoxd10d110HD embryos, reductions in Lim1+ motoneurons are greater than overall losses in motoneurons following electroporation, suggesting that some would-be LMCl neurons have been re-programmed to adopt characteristics of LMCm/MMC neurons. In Hoxd11d10HD and Hoxd10W/Q embryos, however, decreases in Lim1+ motoneurons approximate overall motoneuron decreases, leaving open the possibility that selective cell loss plays a role in shifting subtype proportions. The identification of direct downstream targets of Hoxd10 and Hoxd11 will help to differentiate between these two mechanisms.

Hox proteins have been found to regulate the expression of a number of critical transcription factors associated with neuronal subtype specification, including members of the Pax and Irx families (Pruitt,1994; Theokli et al.,2003). The stark effects of Hox misexpression and/or loss of function, coupled with the wide range of known or potential downstream targets, suggest that Hox proteins occupy a position at or near the top of the signaling hierarchies responsible for hindbrain and spinal cord patterning. A recent examination of Hox function by Yallowitz and colleagues (2009) identified a common enhancer site utilized by two divergent Hox factors, Hoxa2 and Hox11 paralogues, to repress or activate, respectively, the expression of the downstream transcription factor Six2 (see also Gong et al.,2007). The ability to repress or activate in this case was determined entirely by non-homeodomain regions of the two Hox factors and their specific interactions with cofactors, suggesting a high degree of functional equivalence among homeodomains. Unlike Hoxa2 and Hox11, the homeodomains of Hoxd10 and Hoxd11 are not completely functionally interchangeable. It will be important for future studies to determine precisely which targets they control and whether a subset of these targets is shared during motoneuron specification.


In Ovo Electroporation

Fertilized chick eggs (CBT Farms, Chestertown, MD, and Charles River, North Franklin, CT) were incubated in a forced-draft incubator at 37.5°C. Eggs were opened at embryonic day (E) 2.5 (stages 14–16 of Hamburger and Hamilton,1951) and stained with 0.25% neutral red in physiological saline to increase visibility and facilitate stage assessment. Neural tubes from stage-14–16 chick embryos were microinjected at future LS levels with 1.25 μg/μl DNA plasmid constructs encoding wild-type or mutant Hoxd10, Hoxd11, and EGFP. DNA was diluted in Tris-EDTA, pH 8.0, with 0.05% Fast green for visibility during injection. Following injection, embryos were bathed in sterile saline and electroporated using gold 0.5-mm electrodes. Current was delivered in 3 pulses (50-ms duration, charging voltage of 17V) by a square pulse electroporator (BTX). Electroporated embryos were incubated until E4–E6 (stages 23–29 of Hamburger and Hamilton,1951). At sacrifice, embryos were placed in cold avian saline, staged, and dissected to a trunk/limb preparation. Preliminary analyses of dissected tissue indicated unilateral EGFP expression in multiple segments of the spinal cord at posterior thoracic and/or LS levels.

In Ovo Gene Expression Constructs

Expression vectors used for in ovo electroporation were generated by insertion of wild-type or mutant Hoxd10 and Hoxd11 genes and the ires-egfp fragment from pIRES2-EGFP (Clontech, Palo Alto, CA) into a pBluescript-based vector containing a minimal CMV enhancer and an abbreviated Hb9 promoter (provided by S. Pfaff), which drives gene expression specifically in postmitotic motoneurons (Arber et al.,1999; Thaler et al.,1999). Wild-type Hoxd10 and Hoxd11 were kindly provided by C. Tabin. Homeodomain-swapped forms of these two genes were generated using megaprimer PCR (Barik,2002). The 180-bp homeodomain-encoding region of the “donor” gene was amplified via serial PCR reactions with primers adding the homeodomain-flanking sequences of the “recipient” gene to each end of the donor homeodomain. Restriction enzyme recognition sites within the flanking sequences allowed subsequent insertion of the chimeric fragment into the recipient gene. Single and multiple amino acid substitutions were generated using GeneTailor (Invitrogen, Carlsbad, CA) and QuikChange (Stratagene, La Jolla, CA) site-directed mutagenesis kits. Mutations were verified using DNA sequencing.

Luciferase Assays

Luciferase assays were carried out using the Dual-Luciferase Reporter Assay System (Promega, Madison, WI). The 92-bp HCR regulatory element from the Hoxd9 promoter region (Zappavigna et al.,1991; Caronia et al.,2003) was cloned directly from the chick genome and inserted into the multiple cloning site of pGL4.10[luc2] (Promega), from where it drives expression of a synthetic firefly luciferase in response to Hox binding (“HCR-luc2”). Wild-type and mutant Hoxd10 were inserted into the pIRES2-EGFP vector (Clontech) following the removal of the ires-egfp sequence (“CMV-Hox”). A vector encoding renilla luciferase (pGL4.74[hRluc/TK]; Promega) under a constitutively active HSV-TK promoter was used as an internal transfection control (“hRluc”). One day prior to transfection, Neuro2A cells were plated in 6-well plates at 200,000 cells/well in minimal essential medium with 10% fetal bovine serum and 1% penicillin-streptomycin. Each well was transfected with 497 ng HCR-luc2, 20 ng hRluc, and 165 or 300 ng CMV-Hox using TransIT Neural transfection reagent (Mirus Bio, Madison, WI). Cells were lysed 15–24 hr after transfection in passive lysis buffer (Promega). Luminescence was measured using a Turner Biosystems (T3, Sunnyvale, CA) luminometer. All assays were performed in triplicate or quadruplicate.


Dissected chick embryos were fixed in 4% paraformaldehyde for 1.5–2 hr, washed, embedded in a 50:50 mix of 30% sucrose and OCT, and sectioned at 14 μm. Alternating serial sections were placed on separate sets of slides for staining with different antibody combinations. Immunohistochemistry was performed according to standard protocols with the following primary antibodies: mouse anti-Isl1(2) (1:100; Developmental Studies Hybridoma Bank), rabbit anti-Isl1 (1:2,000; T. Jessell), rabbit anti-Lim1 (1:20,000; T. Jessell), rabbit anti-Foxp1 (1:32,000; T. Jessell), rabbit anti-Chx10 (1:4,000; T. Jessell), guinea pig anti-Hoxd10 (1:8,000; T. Jessell), mouse anti-Lim3 (1:100; Developmental Studies Hybridoma Bank, Iowa City, IA), and goat anti-EGFP (1:500; Rockland Immunochemicals, Gilbertsville, PA). Cy2-, Cy3-, and Cy5-conjugated secondary antibodies raised in donkey (Jackson Immunoresearch, West Grove, PA) were used for fluorescent imaging.

Cell Quantification

To quantify differences between transfected and non-transfected sides of the spinal cord following electroporation, counts of neuronal subtypes identified via immunohistochemistry were made on transverse sections through a single segment length of anterior LS cord. Most counts were made at the LS2 level. Due to section availability, some Lim3 and Chx10 counts extended into posterior LS1 or anterior LS3, however, no correlation was found between segment position and outcome. Segment number and boundaries were identified by reference to dorsal root ganglia and spinal nerves on the non-transfected side. Somatic motoneuron status was assigned to cells positive for Isl1(2) antibody staining (a pan-motoneuron marker) and located within three cell-widths of the dorsal edge of the visible somatic motor column cluster. Similar techniques were used to assign LMCm + MMC motoneuron identity to cells positive for Isl1. Images of triple-labeled sections were taken using an Olympus (Center Valley, PA) Fluoview FV1000 confocal unit fitted to an Olympus BX61 microscope; images of double-labeled sections were taken with a QImaging (Surrey, Canada) Retiga 2000R camera and a Nikon (Melville, NY) Eclipse E600 microscope. Cells of a particular molecular profile in these images were subsequently dotted using Adobe Photoshop. Dotted images were then exported to a counting program (provided by N. Roy). Counts of cells on transfected and non-transfected were made on 3 non-adjacent sections per embryo, the spacing between sections being determined by the number of available sections stained with a particular antibody combination. For each construct tested, counts were made on tissue from 4–6 embryos. Because of variability likely to be associated with slight differences in the position of sections within the anterior LS cord and embryonic age (see Results section), the numbers of a specific motoneuron subtype are usually represented as a proportion of the total number of motoneurons per motor column region. No correlation was noted between slight differences in embryonic age and outcome. Significance of numerical differences between transfected and non-transfected sides was determined using paired Student's t-tests.


We thank Laura Lillien and Jack Lee for critical reading of the manuscript. We also thank Thomas Jessell, Samuel Pfaff, Cliff Tabin, and the Developmental Studies Hybridoma Bank for reagents, and Nicholas Roy for the cell counting program. We are grateful to Paula Monaghan-Nichols for cloning advice and to Kathleen Salerno and Kathryn Albers for help with luciferase assays. This work was funded by NIH R01-HD025676 to C.L.-J.