Results

TrkC-Positive DRG Neurons Are Rescued in Bax/NT-3 Double Knockout Mice
Mice lacking proapoptotic protein Bax allow for distinguishing survival effects of neurotrophins from other effects. Bax-deficient sensory neurons no longer require neurotrophins for survival (White et al. 1998; Patel et al. 2000), thus they can be used to examine axonal effects. We bred NT-3 heterozygote and Bax knockout (KO) mice to obtain mice with double KO of both NT-3 and Bax genes, and examined proprioceptive axonal projections. All NT-3 and double KOs died within 48 h after birth (Tessarollo et al. 1994). We performed TrkA/TrkC double immunohistochemistry (Huang et al. 1999), enabling detection of both proteins in the same sample. TrkC-positive cells (Figure 1A) and fibers (Figure 1E) were absent in NT-3 KOs at embryonic day (E) 15. Two subsets of DRG cells expressing either TrkA or TrkC were detected in double KOs, similar to wild-type (WT) or Bax KO animals. Surprisingly, at postnatal day (P) 0, a few cells expressed TrkC even in NT-3 KO animals in the Bax +/+ genetic background, and some cells co-expressed TrkA and TrkC, regardless of the genotype (Figure 1B). In order to quantify our results, we analyzed the ratio of TrkA and TrkC immunopositive cells from four different DRGs of animals of different genotypes. At all ages studied, Bax/NT-3 double null DRGs had TrkA/TrkC ratios similar to those of Bax null DRGs, and higher than those of NT-3 null DRGs (Figure 1D). TrkC-positive neurons rescued by Bax deletion, however, failed to differentiate properly, as evidenced by the lack of expression of the proprioceptive molecular marker Parvalbumin (PV) (Figure 1C).
Figure 1  TrkA/TrkC and PV Immunohistochemistry in DRG and Spinal Cord
Red represents TrkA, green represents TrkC (and PV in [C]), and yellow represents co-expression.
(A) TrkA/TrkC immunostaining in E15 DRG. TrkC-positive neurons normally eliminated in NT-3 KOs are rescued in double KOs.
(B) TrkA/TrkC immunostaining at P0 in DRG.
(C) PV immunostaining in P0 DRG. Rescued TrkC-positive cells fail to express PV.
(D) Ratio of TrkA-immunopositive cells to TrkC-immunopositive cells in E15 and P0 DRGs. Data are presented as percentage of cells with standard deviation. Double KOs always had similar ratios to Bax KOs, and NT-3 KOs had the least amount of TrkC-positive cells, if any.
(E) TrkA/TrkC immunostaining in E15 spinal cord. Arrow points to group Ia fibers. Dorsal is up.
Scale bar: 50 μm (A–C), 1 mm (E).

NT-3 Is Necessary for Proper Innervation of Motor Neurons
TrkA/TrkC-positive fibers in the spinal cord could be detected at E15 (Figure 1E). TrkA-positive fibers were restricted to and terminated in the dorsolateral spinal cord, whereas TrkC-positive fibers entered the cord dorsomedially, and descended into the ventral horns in WT (Ozaki and Snider 1997) and Bax KO embryos. There was no detectable TrkC expression in NT-3 KO spinal cord, indicating complete absence of proprioceptive fibers. In double KO spinal cord, TrkC-positive fibers entered the dorsal spinal cord and descended medially in a manner similar to that seen in WT or Bax KO cases. However, it was not possible to follow TrkC immunolabeled fibers all the way to their terminal zones in any of the cases. Next we examined the central projections of DRG axons with the lipophilic tracer DiI at P0. In WT and Bax KO pups, proprioceptive afferents entered the dorsal spinal cord and followed a medial course towards the ventral horn. They then turned laterally towards motor neurons in the lateral motor column, where they branched and terminated (Figure 2A). DiI labeling was confined to dorsal spinal cord in NT-3 KOs (Figure 2A), as reported earlier, consistent with a complete absence of proprioceptive innervation (Ernfors et al. 1994; Tessarollo et al. 1994). In double KOs, proprioceptive afferents initially followed a trajectory similar to that of WT counterparts, but most of them failed to project all the way to the ventral cord and into the lateral motor column. Instead, they arborized near the ventral midline; some crossed the midline and extended into the contralateral ventral cord (Figure 2A and 2C). In order to distinguish between a role for NT-3 in initiation of motor neuron innervation and a role for maintenance, we repeated the DiI labeling at E17. Innervation patterns of E17 spinal cords (Figure 3) were similar to those at P0 (Figure 2). Dorsal horns of all genotypes were filled with DiI-labeled fibers corresponding to nerve growth factor–dependent nociceptive axons. In WT and Bax KO embryos, proprioceptive fibers extended towards ventral horn motor neurons (Figure 3A and 3B), whereas the ventral horns of the NT-3 KO embryos were devoid of innervation (Figure 3C). In the Bax/NT-3 double KOs, DiI-labeled fibers entered the ventral spinal cord, but extended towards the midline instead of the ventral horn (Figure 3D), in a pattern similar to that observed at P0. Our data point to a complete absence of proprioceptive innervation of the ventral horn of the Bax/NT-3 null spinal cord throughout the developmental stages investigated. As the sensory axons never reach motor neuron dendrites in the ventral horn (Figure S1), the stretch reflex arc circuit is not established. The failure to initiate contact between sensory axons and motor neurons in the absence of NT-3 suggests a requirement for NT-3 for proper axon targeting in addition to a role in sensory axon maintenance.
Figure 2  Axonal Projections in the Spinal Cord after DiI Labeling of DRG at P0
(A) Rescued DRG proprioceptive neurons fail to properly innervate motor neurons in double KOs. Instead, some axons are directed towards the ventral midline; they cross the midline and branch.
(B) Schematic drawing of the monosynaptic reflex arc as it normally develops. Small black dots represent NT-3 released centrally by the motor neurons and peripherally by the muscle spindles.
(C) High-power magnification of the inset in (A). Arrow points to the midline, and arrowheads point to synaptic bouton-like structures.
Scale bar: 1 mm (A), 400 μm (C).
Figure 3  Sensory Axons Labeled with DiI through the DRG at E17
(A) DiI-labeled fibers in WT spinal cord. Notice proprioceptive axons extending towards the motor neurons located in the ventral horn of the spinal cord in cross section.
(B) Bax null spinal cord.
(C) NT-3 null spinal cord. Stained fibers are restricted to the nociceptive axons in the dorsal horn, as evidenced by the complete absence of labeling in the ventral spinal cord.
(D) Bax/NT-3 double null spinal cord. Although fibers extend into the ventral spinal cord, they never grow towards the motor neurons, but are directed towards the midline instead.
Scale bar: 100 μm.

NT-3 Is Required for Proper Peripheral Innervation
In order to study peripheral innervation in double KO animals, we investigated spindle development in the gastrocnemius muscle. Muscle spindles could be identified easily in P0 WT and Bax KO animals by their characteristic morphology (Figure 4A and 4B). Proprioceptive fibers labeled with neurofilament-M (NF-M) antibody formed ring-like spiral endings wrapped around intrafusal bag fibers labeled with S46 antibody, specific for slow developmental myosin heavy chain protein (Miller et al. 1985). As reported earlier, there were no muscle spindles in NT-3 KO animals (Ernfors et al. 1994). On the other hand, Bax KO animals had more spindles than WT, and spindles were in clusters similar to animals over-expressing NT-3 in the muscle (Wright et al. 1997). Although NT-3-dependent cells were rescued, no muscle spindles were detected in the limbs of double KOs (Figure 4A and 4B). Muscle spindles could be observed with TrkC antibody in WT and Bax KO animals, but not in NT-3 or double KOs (Figure 4E). In both NT-3 and double KO animals, NF-M-labeled fibers could be detected in muscles, thus the muscles of these animals were not completely devoid of nerve fibers. Since there were no TrkC-positive fibers in these muscles, we think that these NF-M-labeled fibers correspond to motor axons. Absence of muscle spindles might be due to a failure in initiation of differentiation by proprioceptive axons, or a failure of maturation and maintenance in the absence of NT-3. To distinguish between these two possibilities, we investigated muscle spindle development at E15 and E17 with S46/NF-M immunostaining as well as DiI labeling. No structures with the characteristic muscle spindle shape could be detected in any of the genotypes at E15 (Figure S2). However, numerous developing spindles could be identified at E17 in WT and Bax KOs, but not in NT-3 KO embryos (Figure 4C). Bax/NT-3 double null muscles were devoid of spindles (Figure 4C), except for one spindle-like structure observed in one leg of an embryo (Figure 4C, inset, denoted by the asterisk). DiI labeling through the DRGs at E17 yielded similar results, with muscle spindles identified in parallel sections in WT and Bax null muscles (Figure 4D), but not in NT-3 or double KO animals. Although DiI-labeled fibers could be detected in double null muscles, they never formed ring-like structures characteristic of muscle spindles. We also examined TrkA/TrkC expression at P0 in the tibial nerve, which carries sensory fibers to the gastrocnemius muscle as well as the skin of the lower leg. In NT-3 KOs, TrkA-positive axons could be seen in the tibial nerve but there were no TrkC-positive axons; in contrast, TrkA- and TrkC-labeled axons are both present in WT, Bax KO, and double KO animals (Figure 4F). These results suggest that although proprioceptive axons follow proper trajectories in distal peripheral nerves, they fail to innervate their target muscles in the absence of NT-3.
Figure 4  Muscle Spindles in Gastrocnemius Muscle and TrkA/TrkC Staining in the Tibial Nerve at P0
(A) NF-M (red) and S46 (green) immunostaining in cross section of gastrocnemius muscle at P0. There are no muscle spindles detected in double KOs.
(B) NF-M (red) and S46 (green) immunostaining in parallel sections of gastrocnemius muscle at P0.
(C) NF-M (red) and S46 (green) immunostaining in parallel sections of gastrocnemius muscle at E17. Double null muscles are mostly devoid of muscle spindles, except for one spindle-like structure detected (shown in inset, denoted by the asterisk).
(D) Muscle spindles detected by DiI labeling through the DRG. Gastrocnemius muscle is sectioned at 40 μm thickness in parallel plane to the muscle fibers.
(E) Muscle spindles detected by TrkC staining in cross section of gastrocnemius muscle at P0.
(F) TrkA (red) and TrkC (green) immunostaining in the tibial nerve cross section at P0. TrkC-positive fibers are missing in NT-3 KOs. Red-green overlap (yellow) is due to the thickness of the section and overlapping of red- and green-labeled (TrkA and TrkC) fibers present at different focal depths, rather than co-localization.
Arrows indicate muscle spindles.
Scale bar: 50 μm (A, B, D), 25 μm (C, E), 75 μm (F).

NT-3 Is a Chemoattractant for DRG Axons In Vitro
To test the hypothesis that NT-3 acts as a chemoattractant for sensory axons, we performed a series of in vitro assays. Proprioceptive axons in mice enter the gray matter in the spinal cord and advance ventrally parallel to the midline by E13 and reach the motor neurons by E15 (Ozaki and Snider 1994). We co-cultured collagen-embedded E13 WT DRG explants with NT-3-soaked sepharose beads (n = 26). Control cultures were set up using bovine serum albumin (BSA)– or phosphate buffer saline (PBS)–soaked beads (n = 12). DRG axons began extending towards the localized NT-3 source by the end of the first day and consistently displayed a strong chemoattraction by 3 d in vitro, whereas they did not show such preference for BSA-loaded control beads (Figure 5A and 5B). This attraction was not due to survival support of NT-3 because Bax null ganglia displayed the same chemoattraction (Figure 5C; n = 16). NT-3 may act through either TrkC, or the p75NTR. We repeated the co-culture experiments with DRG explants derived from p75NTR KO mice (n = 18). Axons of these ganglia also showed strong chemoattraction towards the NT-3 beads (Figure 5D). Finally, we used diffusible TrkC receptors conjugated to IgG constant regions (TrkC-Fc) added to the medium (n = 6) to deplete soluble NT-3 from the collagen gels (Figure 5E). In the presence of TrkC-Fc the chemoattraction was completely blocked, demonstrating that the effect we see is specific for NT-3. In order to investigate the active range of our beads, we have repeated the cultures with WT E13 DRGs by placing the beads at increasing distances from the ganglia (n = 4 each) (Figure 5F). There was still preferred growth towards the bead at 1,200μm, the longest distance studied, although the number of axons and the extent of growth were not as robust. Next, we set up E13 DRG spinal cord explant cultures using NT-3-loaded beads placed at the midline at mid-spinal cord level as an ectopic NT-3 source (n = 15). Control cultures were set using PBS-loaded beads (n = 6). DiI labeling through the DRGs revealed numerous fibers entering the spinal cord at ectopic regions and growing towards the NT-3 beads (Figure 6A), surrounding the beads and forming bundles around them (Figure 6C–6E). In control cultures, all labeled fibers were directed towards the dorsal spinal cord and terminated there, where they normally enter the gray matter at E13 (Figure 6B). No axons were observed around the PBS-loaded control beads (Figure 6F). Axons that normally enter the gray matter through dorsal spinal cord grow towards the midline when presented with a localized NT-3 source, and new axon growth towards the NT-3 bead is initiated from DRGs, entering the spinal cord at ectopic loci at lateral mid-spinal cord (Figure 6G). NT-3 is therefore capable of acting as a chemoattractant for DRG axons.
Figure 5  Chemoattraction of E13 DRG Axons to Local NT-3 Observed by In Vitro Co-Culture Assays
(A) WT DRG with NT-3-loaded bead.
(B) WT DRG with BSA-loaded bead.
(C) Bax null DRG with NT-3-loaded bead.
(D) p75 null DRG with NT-3-loaded bead.
(E) WT DRG with NT-3-loaded bead and TrkC-Fc in the medium.
(F) WT DRG with NT-3 loaded beads placed at increasing distances away from the ganglia (range, 500–1,200 μm).
Scale bar: 150 μm (A–E), 350 μm (F).
Figure 6  Chemoattraction Towards NT-3 Beads Placed in E13 Spinal Cord DRG Explant Co-Cultures
(A) NT-3 bead placed in the midline of E13 WT spinal cord. Notice axons labeled through the DRGs (circled with black dashed lines) growing towards the bead (circled with white dashed lines) enter the spinal cord at ectopic loci instead of dorsal spinal cord.
(B) PBS-loaded bead in E13 spinal cord. All labeled axons extend along the dorsal spinal cord, where they terminate.
(C) High-power image of the bead in (A). Notice labeled axons surrounding the bead.
(D) High-power image of an NT-3-loaded bead. Notice axons bundled around the bead.
(E) High-power image of an NT-3-loaded bead. Notice the axons approaching the bead via the dorsal spinal cord.
(F) High-power image of a PBS-loaded bead. No labeled fibers were observed around control beads.
(G) Summary of our observations from E13 spinal cord DRG organotypic cultures. In control cultures fibers extend along the dorsal spinal cord, where they normally enter the gray matter at E13. In the presence of an ectopic NT-3 source localized at the midline, these axons grow towards the NT-3 bead. NT-3 also initiates axon growth from the DRGs, entering the spinal cord at ectopic lateral loci, growing towards the bead, surrounding the bead, forming nerve bundles, and branching around it.
Scale bar, 175 μm (A and B), 100 μm (C–F).