The differentiation and migration of superficial dorsal horn neurons and subsequent ingrowth of cutaneous afferents are crucial events in the formation of somatosensory circuitry in the dorsal spinal cord. We report that the differentiation and migration of the superficial dorsal horn neurons are regulated by the LIM homeobox gene Lmx1b, and its downstream targets Rnx and Drg11, two transcription factors implicated in the development of dorsal horn circuitry. An analysis of Lmx1b mutants shows that Lmx1b normally acts to maintain the expression of the Ebf genes and to repress the Zic genes. Lmx1b mutants also exhibit the disruption of the cutaneous afferent ingrowth, suggesting that the dorsal horn cells might provide important cues guiding sensory axons into the dorsal spinal cord. Our results thus indicate that Lmx1b has a pivotal role in genetic cascades that control the assembly of circuitry in the superficial dorsal horn.
Introduction
Dorsal spinal cord formation begins with neuronal and glial progenitors exiting the cell cycle in the ventricular zone (VZ), and is followed by sequential migration and further differentiation of migrating neurons and glia. Once neurons have reached their appropriate region of the dorsal spinal cord, they differentiate further into distinct layers or laminae that carry out distinct physiological function(Christensen and Perl, 1970; Rexed, 1952). For example,lamina I primarily contains relay and local interneurons, whereas lamina II(substantia gelatinosa) consists of few projection neurons but contains numerous small-size interneurons, many of which make synaptic contacts with lamina I neurons (Molander and Grant,1995; Willis, 1995). The primary afferents of dorsal root ganglia(DRG) neurons project to distinct laminae of the spinal cord and establish precise neuronal connections with their targets. Small-diameter DRG neurons that express TrkA, a high-affinity neurotrophin receptor for nerve growth factor (NGF), project to laminae I-II and relay noxious and thermal sensory information, whereas large-diameter DRG neurons that express the neurotrophin receptor TrkC make connections with motoneurons in the ventral horn and relay muscle proprioceptive information(Christensen and Perl, 1970; Huang and Reichardt, 2001; Lawson and Biscoe, 1979; Scott, 1992; Snider, 1994).
During spinal cord development, the specification of dorsal interneurons is mediated by both extrinsic and intrinsic factors(Caspary and Anderson, 2003; Helms and Johnson, 2003; Lee and Jessell, 1999; Liem et al., 1997). Specifically, the intrinsic factors that control the specification and development of several classes of early-born neurons (dI1-6) have been analyzed in greater detail. In mice lacking transcription factors Lbx1 (Lbx1h - Mouse Genome Informatics) or Rnx(Tlx3 - Mouse Genome Informatics), the specification of dI5 neurons is affected, as indicated by the loss or downregulation of Lmx1b (dI5 marker) (Gross et al., 2002; Muller et al., 2002; Qian et al., 2002). These two genes are also involved in the development of late-born neurons that make up the dorsal horn of the spinal cord(Caspary and Anderson, 2003). In addition, the transcription factors Drg11 (Prxxl1 - Mouse Genome Informatics), Ebf1, Ebf3 and Zic1 have also been shown to be important for the early specification of the dorsal spinal cord neurons (Aruga et al., 1998; Aruga et al., 2002b; Chen et al., 2001; Ebert et al., 2003; Garcia-Dominguez et al., 2003; Garel et al., 1997; Wang et al., 1997). Among them, Zic1 has been shown to be a negative regulator of the differentiation of the dorsal horn neurons in mice(Aruga et al., 2002a; Aruga et al., 1998; Aruga et al., 1996a; Aruga et al., 2002b; Aruga et al., 1994; Aruga et al., 1996b). Despite these studies, our knowledge about the molecular mechanisms that govern the development of early-born neurons and the assembly of the dorsal horn circuits is still rather fragmentary.
Lmx1b is an LIM homeobox-containing gene, and was originally isolated as a mouse ortholog of the chicken Lmx1(Chen et al., 1998a; Riddle et al., 1995; Vogel et al., 1995). In the chicken, Lmx1 is involved in the specification of the dorsal cell fate in the limb, and the differentiation and morphogenesis of the isthmic organizer (Adams et al., 2000; Matsunaga et al., 2002; Riddle et al., 1995; Vogel et al., 1995). Mice lacking Lmx1b exhibit abnormal limbs and kidneys(Chen et al., 1998a). Mutation of Lmx1b results in nail patella syndrome in humans, an autosomal dominant disease characterized by abnormal skeletal patterning and renal dysplasia (Dreyer et al.,1998). In the central nervous system, Lmx1b is involved in multiple developmental processes, including the formation of eye,dopaminergic neurons, serotonergic neurons and the trajectory of the motor axons in the limb (Cheng et al.,2003; Ding et al.,2003; Kania and Jessell,2003; Kania et al.,2000; Pressman et al.,2000; Smidt et al.,2000). In the nematode Caenorhabditis elegans, the Lmx1b ortholog, Lim6, has been shown to be important for the differentiation of GABAergic neurons(Hobert et al., 1999).
In the developing dorsal spinal cord, Lmx1b is expressed in dI5 neurons and late-born neurons destined to populate the superficial layer of the dorsal spinal cord (Chen et al.,2001; Gross et al.,2002; Muller et al.,2002). In this study, we performed detailed studies of the dorsal spinal cord development in Lmx1b mutants. Our study reveals that Lmx1b is important for the development of dI5 neurons, late-born dorsal horn neurons and the projection of cutaneous afferents in the dorsal spinal cord. Our study uncovers a central role for Lmx1b in transcriptional cascades that govern the dorsal horn development.
Materials and methods
Genotyping and maintenance of animals
Lmx1b, Rnx and Drg11 mutant mice were generated and genotyped as previously described (Chen et al., 1998a; Roberts et al.,1994). The age of embryos was determined according to the plug date (the plug date is considered to be E0.5). Because the mating time among the mice may differ, the actual developmental stage of the embryos was further ascertained according to the spinal cord morphology by the use of Nissl staining, as well as the criteria described previously(Kaufman, 1998). The Lmx1b+/-, Rnx+/- and Drg11+/- mice were maintained in a mouse facility according to protocols approved by the Division of Comparative Medicine at Washington University School of Medicine.
BrdU labeling and detection by immunocytochemistry
Pregnant female mice derived from timed matings between Lmx1b or Rnx or Drg11 heterozygous mice were given a single intraperitoneal injection of BrdU (5 mg/ml solution in PBS and 60 μg/g of body weight) at 11.5 days postcoitum (dpc) and 12.5 dpc. After time periods of 2 hours, 1 day, 2 days or 3 days, embryos were removed, genotyped and sectioned as described (Chen et al.,1998a; Chen et al.,2001; Roberts et al.,1994). The sections were processed and stained sequentially with a mouse anti-BrdU antibody (Dako), biotinylated donkey anti-mouse IgG (Jackson Immunoresearch) and ABC Elite reagents (Vector). For double staining with anti-LMX1B antibody (Kania et al.,2000), the slides were first incubated with a guinea pig anti-LMX1B antibody detected enzymatically through production of a brown precipitate. For anti-BrdU antibody detection, a nickel intensification technique was used, producing a black precipitate. To quantify the distribution of BrdU-labeled neurons, we used the Photoshop (Adobe) program after dividing the dorsal horn into two parts: medial one-third and lateral two-thirds (from the midline to the lateral edge of the dorsal horn, Fig. 2). BrdU-labeled neurons in ten sections each from wild-type (n=6) and mutant (n=6)embryos were counted at the thoracic segmental level, and a comparison was performed using Student's t-test.
Nissl staining, in situ hybridization and immunocytochemistry
Nissl staining, in situ hybridization and immunocytochemistry were performed as described (Chen et al.,1998b; Chen et al.,2001). For double staining of LBX1 with BRN3A, PHOX2A with PAX2,and LMX1B and PHOX2A, Cy3-labeled donkey anti-guinea IgG (Jackson) and FITC-labeled donkey anti-rabbit or anti-mouse IgG (Jackson) were used. Antibodies included: rabbit anti-BRN3A(Fedtsova and Turner, 1995),guinea pig anti-ISL1 (Tanabe et al.,1998), guinea pig anti-LMX1B, guinea pig anti-LBX1(Muller et al., 2002), mouse anti-MAP2 (Sigma), rabbit anti-PHOX2A antibody(Tiveron et al., 1996), rabbit anti-PAX2 (Zymed Lab), rabbit anti-trkA and rabbit anti-Peripherin 57K(Chemicon). For cell counting, 10-15 consecutive sections each from wild-type(n=6) and homozygous mutant (n=6) embryos at thoracic levels were counted, and statistical analysis was performed by the use of Student's t-test.
DiI labeling
For study of the projection from the DRG to the spinal cord, a small amount of 1,1″-dioctadecyl-3,3,3″,3″-tetramethylindocarbocyanine perchlorate (DiI; Molecular Probes) crystals was placed in the DRG unilaterally. The samples were kept in the fixative at 37°C for 2-4 days,and then were sectioned transversely with a vibratome at a thickness of 50-100μm. Labeling was observed with epifluorescent or laser confocal microscopy.
In utero electroporation
For in utero electroporation, previously detailed procedures were followed(Saito and Nakatsuji, 2001). Pregnant mice at 12 dpc were anesthetized with sodium pentobarbital (40 mg/kg), followed by the exposure of the uterus and cutting of the uterine wall on both horns along the antiplacental side. pCAGGS-Lmx1b:EGFP or pCAGGS-EGFP (Niwa et al.,1991) (1-3 μl; 0.5-1.5 μg/μl) was injected into the central canal of the spinal cord using an orally controlled pipette system. After injection, square electrical pulses were delivered by the use of an Electro Square Porator (ECM830) at a rate of one pulse per second (voltage 35V, five pulses, 50 ms) to embryos by holding the utero with forceps-type electrodes. The embryos were repositioned into the abdominal cavity without sewing the uterine wall. The abdomen was filled with warmed saline and the abdominal wall and skin were sutured. Embryos were allowed to survive for 2 days. Genotyping and analysis of gene expression in the spinal cord of electroporated embryos was performed as described above.
Results
Lmx1b is expressed in dI5 and dorsal horn neurons
To further characterize dynamic expression of Lmx1b in the developing spinal cord, in situ hybridization and immunocytochemical studies were performed. At E10.5, Lmx1b was expressed in dI5 neurons(Fig. 1A). Beginning at E11.5, Lmx1b expression was expanded to the entire region lateral to the VZ of the dorsal spinal cord (Fig. 1B). At E17.5 when laminae I-II and laminae III-V are distinguishable, Lmx1b expression was prominent in laminae I-II and weak in laminae III-V (Fig. 1C). Lmx1b was not expressed in DRG neurons at any of the stages examined (Fig. 1C, data not shown). To determine whether Lmx1b is expressed in postmitotic neurons, we performed BrdU/LMX1B double labeling. After 2 hours of BrdU labeling in either E11.5 or E12.5 embryos, very few BrdU+/LMX1B+ double-labeled neurons were detected:BrdU+ neurons were mostly restricted to the VZ of the dorsal spinal cord and did not overlap with LMX1B-expressing neurons in the subventricular zone (Fig. 1D). Thus, Lmx1b expression is largely confined to postmitotic neurons.
Phox2a marks a subset of dI5 neurons and is partially lost in Lmx1b mutants
Previous studies have shown that Lmx1b expression is either absent or downregulated in dI5 neurons of Lbx1 or Rnx mutant mice(Gross et al., 2002; Muller et al., 2002; Qian et al., 2002). These observations raise the possibility that Lmx1b plays a role in the development of dI5 neurons, and that Lbx1 and/or Rnx may mediate the specification of dI5 neurons via Lmx1b. To characterize the development of dI5 neurons in the absence of Lmx1b, we examined the expression of Lbx1 (dI4-6), Brn3a (Pou4f1 -Mouse Genome Informatics) (dI1-3, dI5) and Pax2 (dI4, dI6), all markers of distinct dorsal neuronal types in Lmx1b mutants(Gross et al., 2002; Muller et al., 2002). Immunocytochemical staining revealed a similar staining pattern of LBX1, BRN3A and PAX2 between Lmx1b mutants and wild-type control at E11.0(Fig. 1E-H). Our study,however, was complicated by the observations that Lbx1 and Brn3a are also expressed in neurons located adjacent to dI5 neurons. To identify a more dI5-specific marker, we examined the co-expression of LMX1B and PHOX2A at E10.5 and E11.5 spinal cord(Tiveron et al., 1996). We found that PHOX2A expression overlaps with LMX1B in a subset of dI5 neurons,and no neurons expressing PHOX2A alone were found in the same region(Fig. 1I,J). Thus, Phox2a is a dI5 neuron-specific marker(Fig. 1I,J). In Lmx1bmutants, Phox2a expression was either lost or decreased in dI5 neurons in the mutants compared with that in wild-type control(Fig. 1K,L; data not shown). Moreover, the location of PHOX2A+ neurons was shifted more ventrally in the spinal cord of Lmx1b mutants, indicating that dI5 neurons migrated aberrantly in Lmx1b mutants(Fig. 1K,L). Cell counting confirmed that the number of PHOX2A+ neurons was indeed reduced in the mutants (Fig. 1M). Together our data suggests that the development of dI5 neurons is impaired in the absence of Lmx1b.
A migratory defect in the dorsal spinal cord of Lmx1bmutants
To examine the effect of Lmx1b mutation on the dorsal horns of Lmx1b-/- embryos, we performed Nissl staining. At E12.5,the dorsal horns of Lmx1b-/- embryos and wild-type controls were indistinguishable (Fig. 2A,B). At E15.5, in wild-type embryos, numerous laminae I-II neurons, distinguishable from laminae III-IV neurons by their smaller size,aggregate to form a separate layer, whereas laminae III-IV neurons also become distinct in being more loosely distributed(Fig. 2C). In Lmx1bmutants, these laminar boundaries were not recognizable(Fig. 2D) and the dorsal funiculus was smaller relative to wild-type controls(Fig. 2E,F).
We next used BrdU to determine the migratory behavior of the dorsal horn neurons in Lmx1b mutants. To do this, we labeled dorsal horn neurons of wild-type and Lmx1b mutant embryos with BrdU at E11.5 and E12.5,and examined their settling position at E14.5. Interestingly, neurons labeled with BrdU at E11.5 in the wild-type embryo did not migrate to the most superficial region of the dorsal horn (Fig. 2G). By contrast, neurons labeled at E12.5 migrated through earlier-born neurons and occupied the most superficial layer of the dorsal horn, consistent with previous studies(Nornes and Carry, 1978; Nornes and Das, 1974). However, unlike previous studies in which [3H]thymidine autoradiography was used, our BrdU labeling clearly revealed an inside-out migration pattern for laminae I-II neurons. We next examined the migration of BrdU-labeled neurons in Lmx1b mutants. First, we did not find any significant change in the total number of BrdU-labeled neurons between wild-type and Lmx1b mutants (data not shown). However, some neurons labeled with BrdU at E11.5 were present in the most superficial region of the dorsal horn of E14.5 Lmx1b-/- embryos(Fig. 2H), while a significantly higher number of BrdU+ neurons were accumulated near the midline region as compared with the control(Fig. 2I-K). In line with this observation, fewer neurons labeled with BrdU at E11.5 were located in the lateral region of the dorsal horn in Lmx1b-/- embryos(Fig. 2J). This altered distribution of dorsal neurons was not due to abnormal neuronal death, as the TUNEL staining pattern was similar in both Lmx1b-/- and wild-type embryos between E11.5 and E15.5 (data not shown). Together, these results indicate that there is a major defect in the migration of neurons in the dorsal horn of Lmx1b-/- embryos.
Lmx1b controls Drg11, Rnx and Ebf1, and Ebf3 expression in the dorsal horn
To further characterize the dorsal horn defects in Lmx1b mutants,we examined the expression profile of dorsal horn neuron molecular markers Drg11, Ebf1, Ebf2, Ebf3 and Rnx. In Lmx1b-/- embryos, Drg11 expression was completely abolished from the normal onset of its expression(Fig. 3A,B; see Fig. S1). At E11.5, Ebf1, Ebf2, Ebf3 and Rnx appeared to be normally expressed in Lmx1b-/- embryos (see Fig. S1). However, by E12.5, Ebf3 and Rnx expression levels were lower in Lmx1b mutants (see Fig. S1) and by E15.5, Ebf3 expression was completely absent in the Lmx1b mutants(Fig. 3C,D). At this stage, in wild-type embryos, Ebf1 was most strongly expressed in laminae III neurons and weakly in laminae III-IV neurons. By contrast, Ebf1expression was markedly reduced in Lmx1b mutant embryos(Fig. 3E,F). Similarly, Rnx was dramatically downregulated in the Lmx1b mutants(Fig. 3G,H).
The complete absence of expression of Drg11 and Ebf3expression in Lmx1b mutants promoted us to examine whether Lmx1b is able to induce the expression of these two genes in Lmx1b mutants. To address this question, Lmx1b-expression vectors were introduced into the dorsal spinal cord of Lmx1b mutant embryos by the use of in utero electroporation(Saba et al., 2003). Although the expression of Drg11 and Ebf3 was absent in the dorsal horn of Lmx1b-/- mutants electroporated with green fluorescent protein expression vectors(Fig. 3J,K; data not shown), Drg11 and Ebf3 expression was detected in the dorsal horn of Lmx1b-/- mutants electroporated with Lmx1b-expression vectors (Fig. 3L,M,O,P). The contralateral unelectroporated side of the dorsal horn showed no Drg11 and Ebf1 expression(Fig. 3L,N,O,Q). We observed that the induced expression was considerably weaker when compared with the endogenous expression, which we attribute to the inefficiency of electroporation. Because of the presence of the residual expression of Rnx and Ebf1 in the dorsal horn of Lmx1b mutants,it was difficult to determine whether Lmx1b is capable of activating these two genes in the mutants. Nevertheless, these data demonstrate that Drg11 and Ebf3 act downstream of Lmx1b in the developing dorsal horn, and Lmx1b is required either for initiating or maintaining expression of these two genes in the dorsal horn.
Lmx1b-independent dorsal horn-specific transcription factors
The expression of Lbx1, a lamina IIi-III marker which acts upstream of Lmx1b was examined in wild-type and Lmx1b mutant embryos (Fig. 4G,H). Although a few LBX1+ neurons were found in the most superficial layer of the Lmx1b-/- dorsal horn, no significant difference in the number of LBX1+ neurons was detected between the Lmx1bmutants and wild-type controls (Fig. 4I). The expression of Brn3a, a laminae III-V marker(Fig. 4J), was examined in Lmx1b mutants (Fig. 4K). Although a seemingly higher number of BRN3A+neurons were located more medially in Lmx1b mutants(Fig. 4K), neuronal counts revealed no difference in the number of BRN3A+ neurons between wild-type and Lmx1b mutants (Fig. 4L).
In the dorsal horn, Lmx1b and Pax2 expression is mutually exclusive (Gross et al.,2002). To determine whether the loss of Lmx1b affects Pax2 expression, PAX2+ neurons in the dorsal horns of the wild-type and Lmx1b-/- embryos were counted. No significant change in the number of PAX2+ neurons was detected between wild-type and Lmx1b embryos(Fig. 4M-O).
Hox6 and Hox8 are homeobox genes that are expressed in laminae I-II (Graham et al.,1991). In Lmx1b mutants, Hoxb6 and Hoxb8 expression domains appeared to have expanded(Fig. 4A,B,D,E). Although the expansion of the Hoxb6 and Hoxb8 domains could be attributed to an abnormal location of laminae I-II neurons in the deep dorsal horn, it is also possible that the expression of Hoxb6 and Hoxb8 is upregulated in laminae III-V neurons in the absence of Lmx1bfunction. If it were the latter case, the total number of Hoxb6 and Hoxb8 neurons would increase in the mutants. However, no significant difference in the number of Hoxb6 and Hoxb8 neurons was evident between wild-type and mutant animals(Fig. 4C,F). Our data thus suggest that abnormal distribution of Hox6/8 neurons is due to migration defects.
Together, these results suggest that expression of Hox6, Hox8, Lbx1,Brn3a and Pax2 is Lmx1b independent; however, the distribution pattern of neurons expressing these markers appears changed in Lmx1b mutants.
Lmx1b, Rnx and Drg11 repress Zic1 and Zic4 genes
We next asked whether there is a causal relationship between the expression of Zic genes and Lmx1b as Zic1 appears to be required for maintaining the progenitor state of undifferentiated dorsal neurons by repressing the differentiation of spinal cord progenitors(Aruga et al., 2002b; Ebert et al., 2003).
In wild-type embryos, three Zic genes (Zic1, Zic2 and Zic4) are expressed in the developing dorsal spinal cord(Fig. 5). Their expression is mainly concentrated around the midline region, suggesting that the Zic genes may play a role in modulating the differentiation of the dorsal horn neurons. Strikingly, in Lmx1b-/- mutants, while Zic2expression remained unaltered (Fig. 5E,F), expression of Zic1 and Zic4 was significantly upregulated (Fig. 5A-D). This was observed not only around the midline region, but also in the superficial layer of the dorsal horn where numerous laminae I-II neurons were present but failed to differentiate further at a later stage(Fig. 5B,D). An upregulation of Zic1 and Zic4 was also detected in Rnx mutants(Fig. 5H,J).
The maturation and differentiation of lamina I-II neurons is impaired in Lmx1b-/- mutants
The aberrant migratory behavior and altered gene expression profile of the dorsal horn neurons in Lmx1b mutants could also be viewed as the evidence of impaired maturation and differentiation of laminae I-II neurons in the absence of Lmx1b. To further examine the cellular property of laminae I-II neurons, immunocytochemical staining using anti-MAP2 antibody was performed. MAP2 is a microtubule-associated protein expressed in the dendrites, axons and somata of neurons(Lewis et al., 1986). During development, the onset of MAP2 expression concurs with the differentiation,maturation and lamination of laminae I-II neurons. In the dorsal horn of E15.5 wild-type embryos, MAP2 expression was prominent in laminae I-II neurons,especially in the dendrites (Fig. 6A,C). By contrast, MAP2 expression was much weaker in the superficial layer of the Lmx1b mutant(Fig. 6B,D). One possible explanation for the weak expression of MAP2 is that MAP2+ laminae I-II neurons could be aberrantly located in the deeper laminae of Lmx1b mutants. However, no obvious increase of MAP2 staining in these regions was found in Lmx1b mutants(Fig. 6A-D). Therefore, it is most likely that MAP2 expression is largely reduced in laminae I-II neurons of Lmx1b mutants.
Rnx and Drg11 mutants exhibit an aberrant migration and differentiation of laminae I-II neurons
The downregulation of Rnx and the loss of Drg11expression in Lmx1b mutants suggest that Lmx1b may control some aspects of the migration and differentiation via Rnx and Drg11. A previous study using BrdU labeling detected no migration deficit in the dorsal horn of Rnx mutants(Qian et al., 2002). We revisited this issue by examining the inside-out pattern of the dorsal horn neurons in Rnx and Drg11 mutants and found that neurons labeled with BrdU at E11.5 were detected in the superficial region of the dorsal horn of Rnx and Drg11 mutants(Fig. 7A,B,D,E). Quantitative analysis indicated that in Rnx and Drg11 mutants there were more BrdU+ neurons accumulating in the medial one third of the dorsal horn when compared to controls, whereas fewer BrdU+ neurons were found in the lateral two-thirds of the dorsal horn(Fig. 7C,F). Immunocytochemical staining shows that MAP2 expression was also reduced in the superficial dorsal horn of both mutants (Fig. 7G,H,I,J). Nevertheless, the abnormalities detected in Drg11 mutant are less severe than those in Rnx mutants.
Aberrant projection of TrkA+ afferent fibers into the dorsal horn of Lmx1b mutants
We reasoned that the impaired development of laminae I-II neurons in Lmx1b mutants could influence the projections of the primary cutaneous afferents which innervate the dorsal horn(Sharma and Frank, 1998). To address this possibility, we examined the projections of TrkA+fibers into laminae I-II (Huang et al.,1999). At E14.5, in wild-type mice, TrkA+ afferents had begun to ramify within laminae I-II of the dorsal horn, and immunostaining showed densely stained laminae III, an indication of innervation by primary cutaneous afferents (Ozaki and Snider,1997) (Fig. 8A). Strikingly, although Lmx1b mutants did not show apparent defects in projections of muscle afferents, as examined by anti-Peripherin antibody staining (Goldstein et al.,1991) (Fig. 8C,D),no TrkA+ afferents were detected within the gray matter of the dorsal horn of Lmx1b mutants (Fig. 8B).
We next analyzed the sensory afferent projection by DiI labeling of DRG neurons. At E14.5, in wild-type controls, projections to laminae I-IV were intensely labeled with DiI after its application to DRG cell bodies(Fig. 8E). The muscle sensory afferents that project towards the ventral horn were also labeled(Fig. 8E). In Lmx1bmutants, while the presumptive muscle sensory afferent projections appeared normal, there was a dramatic decrease in the intensity of DiI labeling in laminae I-II, although some DiI-labeled DRG afferents appeared to be present in laminae III-IV (Fig. 8F). Together, these results suggest that DRG afferent projection defects in Lmx1b mutants are specific to TrkA+ sensory afferents.
Stereotyped expression of attractive or repulsive axonal guidance molecules is crucial to guiding the precise axonal growth of neurons in the developing nervous system (Tessier-Lavigne and Goodman, 1996). To determine whether there was a change of expression of axonal guidance cues/molecules in the dorsal horn of Lmx1b mutants, we examined their expression by in situ hybridization. Sema3a is a secreted cell-surface protein that functions as a chemorepellent in the projection of cutaneous afferents in the spinal cord of the chick and the mouse (Luo et al.,1993; Messersmith et al.,1995). Its expression, however, was normal in Lmx1bmutants (data not shown). We also found that Sema3c(Feiner et al., 2001) was completely lost in the dorsal horn of Lmx1b mutants(Fig. 8G,H). By contrast,expression of the receptors for Sema3a, neuropilin 1 (Nrp1)and plexin A2 (He and Tessier-Lavigne,1997; Kolodkin et al.,1997; Takahashi et al.,1999), were markedly reduced in the mutants(Fig. 8I-L). We also examined the expression of the members of the Slit family, which have been implicated in axonal guidance and neuronal migration(Brose and Tessier-Lavigne,2000; Wong et al.,2002). Slit1 and Robo2 were most strongly expressed in laminae I-II (Fig. 8M,O). In the mutants, Slit1 and Robo2expression was dramatically reduced (Fig. 8N,P). In addition, netrin 1 expression was largely lost in the medial region of the deep dorsal horn of Lmx1b mutants(Serafini et al., 1994)(Fig. 8Q,R). Thus, the altered expression of multiple axonal guidance molecules could account for the failure of cutaneous sensory axon ingrowth in Lmx1b mutants.
Discussion
Lmx1b is required for the specification of the dI5 neuronal fate
Recent studies have begun to identify important players such as Math1,Lbx1 and Rnx in the development of early-born postmitotic dorsal neurons (Bermingham et al.,2001; Gross et al.,2002; Muller et al.,2002). In this study, we find that dI5 interneurons can be further divided into two subgroups: one that expresses Phox2a and one that does not. Phox2a is a homeobox gene and has been shown to be an important determinant of several neuronal phenotype in the nervous system(Brunet and Pattyn, 2002). The seemingly normal expression of Lbx1 and Brn3a expression in dI5 neurons of Lmx1b mutants supports the idea that these two genes may act upstream of Lmx1b, whereas the partial loss of Phox2a indicates that Phox2a lies either downstream of or in parallel to Lmx1b. It will be interesting to examine whether Phox2a plays a role in the development of dI5 neurons, and whether Lbx1 and Rnx may specify the fate of dI5 neurons through an Lmx1b-Phox2a pathway.
Lmx1b controls the ingrowth of cutaneous afferents into the dorsal horn
In Lmx1b mutants, the selective blocking of cutaneous afferent ingrowth raises the possibility that Lmx1b may coordinate the projection of TrkA+ afferents into the dorsal horn by regulating local axonal guidance cue(s). Altered ingrowth of TrkA+ afferents is also found in Lbx1 mutants and Rnx/Tlx1 double mutants(Gross et al., 2002; Muller et al., 2002; Qian et al., 2002). Nevertheless, unlike Lbx1 and Rnx, Lmx1b is not expressed in DRG neurons, therefore, the TrkA guidance defect most probably resides in the dorsal horn.
The specific block of the entry of TrkA+ afferents suggests that certain attractants may be missing in the mutants. Alternatively, the expression level of some repulsive molecules may be increased to repel the cutaneous afferents. Although these possibilities exist, no evidence for increased expression of any repellants examined in Lmx1b mutants was found. Given that multiple axonal guidance cues are expressed in the dorsal horn and their expression is lost in Lmx1b mutants, it is possible that multiple axonal guidance molecules could work synergistically to coordinate the projections of cutaneous afferents. Future analysis of mice lacking multiple axonal guidance cues may be required to reveal the identity of the cues responsible for the ingrowth of TrkA+ afferents in the dorsal spinal cord.
Lmx1b guides the differentiation and migration of the dorsal horn neurons
Our study suggests that Zic1, Zic4, Ebf1 and Ebf3function downstream of Lmx1b and Rnx. Forced expression of Zic1 in chicks represses the differentiation of Math1-expressing neurons in the dorsal neural tube(Ebert et al., 2003). Transgenic mice overexpressing Zic1 exhibit inhibited neuronal differentiation with extension of the progenitor state in the dorsal spinal cord (Aruga et al., 2002b). Loss of Zic1 in mice also leads to premature expression of theβIII tubulin in the dorsal spinal cord(Aruga et al., 2002b). In the context of these findings, the Zic gene expression upregulation is interesting, and consistent with the migration and differentiation defects in Lmx1b mutants. Zic1 could be an important component of the dorsal spinal cord neuron differentiation pathway.
The role of Zic4, Ebf1 and Ebf3 in the development of the dorsal horn is unknown. Whether dysregulation of these genes plays a causal role in aberrant development of the dorsal horn also remains to be determined. Together with previous studies, we hypothesize that the Lbx1/Lmx1b/Rnx/Drg11 pathway represents a major pathway to control the differentiation and migration of laminae I-II neurons and subsequent projection of cutaneous afferents (Fig. 9). The effects of Lmx1b and Rnx may be mediated in part through the Drg11, Ebf and Zic genes and in part via other unidentified transcription factors to control the assembly of the dorsal horn circuits.
The present study reveals the complicated relationship between Lmx1b and other transcription factors in the development of the dorsal spinal cord. For example, Lmx1b and Rnx may depend on each other for their normal expression. Although Lmx1b mutants recapitulate many aspects of the phenotype of Rnx mutants, there are noted differences in the level of regulation of downstream genes such as Zic1 and Drg11. Thus, in addition to the common sets of downstream targets, Lmx1b and Rnx may control additional distinct set of target genes.
The finding that an inside-out migration pattern and a normal differentiation program are disrupted in the dorsal horn of Lmx1bmutants raises the question of whether aberrant migratory behavior could be attributed to aberrant differentiation of the dorsal horn neurons or vice versa. Although we are not able to determine when the migration deficit first occurs in Lmx1b mutants, studies of the possible downstream targets of Lmx1b shed light onto this issue. For example, an alteration of axonal guidance cues such as netrin 1 in Lmx1b mutants might have contributed to aberrant neuronal migration(Brose and Tessier-Lavigne,2000). Moreover, some of the transcription factors downstream of Lmx1b have also been shown to be required for neuronal migration. In mice, Ebf1 is important for the migration of facial branchiomotor neurons (Garel et al., 2000). In chicks, Ebf1 and Ebf3 appear to control migration and differentiation of the dorsal neurons independently(Garcia-Dominguez et al.,2003; Garel et al.,2000). Thus, Lmx1b may have a unique role in neuronal migration in the developing spinal cord. However, this does not exclude the possibility that aberrant migration reflects some aspects of impaired differentiation of the dorsal horn neurons in the absence of Lmx1b.In fact, in addition to the dorsal horn cells, Lmx1b has also been implicated in neuronal differentiation of dopaminergic neurons and serotonergic neurons in the developing brain(Cheng et al., 2003; Ding et al., 2003; Pressman et al., 2000). Thus,it is likely that Lmx1b could play an important role in both cellular events.
Supplemental data available online
Acknowledgements
We thank D. J. Anderson, J. Brunet, E. Turner, R. Krumlauf, M. R. Capecchi,T. Jessell, J. Aruga, L. Reichardt, T. Muller, C. Birchmeier, J. Miyazaki, Y. Rao, J. Wu, M. Tessier-Lavigne, H. Fujisawa, A. Kolodkin and Q. Ma for reagents. We are particularly grateful for Dr T. Saito's advice on in utero electroporation technique. We also thank T. Jessell, J. Sanes, Y. Rao, A. L. Pearlman, I. McIntosh and M. Jacquin for comments on various versions of the manuscript; and H. Xu, S. Li, C. Xiang and Z. Zhang for technical help. This work was supported, in part, by a grant from the McDonnell Center for Cellular and Molecular Neurobiology at Washington University, and a NIH grant NS43968-01 (Z.F.C.). All animal experiments were reviewed and approved by the Animal Studies Committee (ASC) at Washington University.