We examined the spatio-temporal relationship between neurotrophic factor receptor (NTF-R) expression and motoneuron (MN) survival in the developing avian spinal cord and observed heterogeneity in the expression of NTF-Rs between, but not within, pools of MNs projecting to individual muscles. We then focused on the role of NTFs in regulating the survival of one motor pool of MNs, all of which innervate a pair of adductor muscles in the thigh and hence compete for survival during the period of programmed cell death (PCD). The complete NTF-R complement of these MNs was analyzed and found to include many, but not all, NTF-Rs. Treatment with exogenous individual NTFs rescued some, but not all, adductor MNs expressing appropriate NTF-Rs. In contrast, administration of multiple NTFs completely rescued adductor MNs from PCD. Additionally, adductor MNs were partially rescued from PCD by NTFs for which they failed to express receptors. NTF-Rs expressed by the nerve but not in the muscle target were capable of mediating survival signals to MNs in trans. Finally, the expression of some NTF-Rs by adductor MNs was not required for MN survival. These studies demonstrate the complexity in NTF regulation of a defined subset of competing MNs and suggest that properties other than NTF-R expression itself can play a role in mediating trophic responses to NTFs.
In the chick embryo, the number of postmitotic spinal motoneurons (MNs) is reduced by approximately one-half during the process of developmental programmed cell death (PCD) (Hamburger, 1975). This elimination of differentiating, synapse-forming MNs appears to reflect in large part a competition for limiting amounts of skeletal muscle-derived neurotrophic factors (NTFs) (Hollyday and Hamburger, 1976; Pittman and Oppenheim, 1979; Phelan and Hollyday, 1991; Grieshammer et al., 1998). Despite the discovery of a large number of survival-promoting NTFs expressed by muscle and other cell types during the period of MN PCD, the addition or overexpression of any one or a combination of these agents in vivo fails to rescue all spinal MNs originally destined to die (Oppenheim et al., 1993; Zurn et al. 1996; Nguyen et al. 1998). Conversely, elimination of any one NTF, or the receptor through which it acts, fails to eliminate >40% of spinal MNs originally destined to live (Henderson, 1996; Oppenheim, 1996). These findings, together with data describing spatial- and temporal-specific patterns of NTF receptors (NTF-Rs) (McKay et al., 1996; Yamamoto et al., 1997; Novak et al., 2000; Garcés et al., 2000, 2001; Homma et al., 2003), suggest that spinal and cranial MNs are heterogeneous in their trophic responsivity. Additionally, NTF-Rs expressed in cell types intimately associated with MNs during development (e.g., Schwann cells) are capable of signaling in trans to MNs, suggesting an additional level of complexity in the regulation of MN survival (Paratcha et al., 2001; Ledda et al., 2002).
The fact that not all MNs respond to increased or reduced NTF availability may simply reflect a difference in expression patterns of specific NTF-Rs. Alternatively, a simultaneously active death-promoting pathway may predominate over those initiated by even excess amounts of individual or multiple NTFs (Raoul et al., 1999, Krieglstein et al., 2000). In this model, treatment with excess NTF would not be sufficient to rescue all MNs expressing appropriate NTF-Rs. Still another alternative is that NTF-R stimulation regulates aspects of MN development other than survival, such as phenotypic identity, differentiation (e.g., target innervation), or migration of subtypes of MNs (Haase et al., 2002) and is, therefore, neither necessary nor sufficient for the survival of at least some MNs.
To determine whether MN responses to NTFs are predisposed solely by their complement of NTF-Rs, we examined expression patterns of specific NTF-Rs. Because MNs projecting to individual muscle targets are arranged into specific spatial clusters called motor pools even before the period of MN PCD (Landmesser, 1978; Hollyday, 1980), we related NTF-R expression patterns to some of these MN subpopulations. In this study, we concentrated on MNs innervating the adductor muscles in the avian thigh. MNs within this pool are homogeneous in their expression of various NTF-Rs. Therefore, we were able to determine whether NTF-R expression, in the presence of excess, physiological, or reduced amounts of NTF, was necessary or sufficient for MN survival.
Materials and Methods
Animals. White leghorn chicken embryos (Tyson, Wilkesboro, NC) were incubated until the appropriate stage (Hamburger and Hamilton, 1951). Their spinal columns were dissected, fixed overnight at 4°C in freshly made 4% paraformaldehyde in 0.1 m PBS, pH 7.4, rinsed in PBS, embedded in 30% sucrose-PBS, and snap-frozen in 3:2 30% sucrose-PBS to optimal cutting temperature (OCT) compound (Tissue-Tek; Pelco International, Redding, CA).
Immunohistochemistry. Slides were blocked in 10% FBS or 10% NGS for 1 hr at room temperature. The monoclonal antibody supernatants used were anti-TuJ1 (1:500; Boehringer Mannheim, Indianapolis, IN), anti-islet1/2 (39.4D5), anti-1E8 (avian Protein Zero, P0; a Schwann cell marker), anti-MF20 (a skeletal muscle marker; sarcomeric myosin), anti-B3/D6 (avian fibronectin), and anti-SV2 (a synaptic vesicle marker), all used at 1:100 to 1:50 and obtained from the Developmental Studies Hybridoma Bank (University of Iowa, Iowa City, IA). Mouse anti-islet1 (1:100) was kindly provided by S. Morton (Columbia University, New York, NY); rabbit anti-chicken CNTFRα (1:1000) was kindly provided by H. Rohrer (Max-Planck-Institut für Hirnforschung, Frankfurt am Main, Germany); rabbit anti-chicken tyrosine kinase B (TrkB) (1:10,000) and rabbit anti-chicken TrkC (1:1000) were kindly provided by F. Lefcort (University of Montana, Bozeman, MT); and rabbit anti-cRet (1:200) was obtained from Santa Cruz Biotechnology (Santa Cruz, CA). Sections were incubated in primary antibody overnight at 4°C in 0.03% Triton-X and 10% FBS or NGS. Slides were rinsed in 0.1 m PBS three times for 5 min and incubated in 1:250 Cy2, Cy3, or Cy5 donkey anti-mouse, donkey anti-rabbit, donkey anti-goat, or goat anti-mouse (Jackson Immunoresearch, West Grove, PA) for 45 min at room temperature, rinsed, and mounted in Gelmount (Biomeda, Foster City, CA).
Immunoblotting. Rabbit anti-neurturin (NTN) was used at 1:500 (Peprotech, Rocky Hill, NJ); rabbit anti-glial cell line-derived neurotrophic factor (GDNF) was used at 1:100 (D-20; Santa Cruz Biotechnology). Individual muscles were dissected from hindlimbs, stripped carefully of skin and connective tissue, and collected into protein lysis buffer (20 mm Tris-HCl, pH 7.5, 10 mm EDTA, 120 mm Nacl, 1% NP-40, 1 mm NaVO4, and 1× protease inhibitor mixture; Sigma, St. Louis, MO). Lysates were cleared, protein concentration was determined, and 100 μg of protein was loaded onto 12% polyacrylamide gels. After transfer to polyvinylidene difluoride membrane (NEN Life Sciences, Boston MA), proteins were stained with 0.1% India ink to ensure equal loading, rinsed in 1× TBST (0.1% Tween), blocked in 3% BSA (Fraction V, IgG free; Sigma)/TBST for 3 hr, and rotated overnight at 4°C in primary antibody/BSA. Membranes were rinsed three times for 10 min in BSA-TBST, incubated in 1:5000 peroxidase-conjugated donkey anti-rabbit, rinsed in TBST five times for 5 min, and detected with Supersignal West Pico chemiluminescent substrate (Pierce, Rockford, IL).
ELISA. For quantitation of GDNF protein by ELISA, individual chick thigh muscles at embryonic day (E) 8, entire chick thighs from E6-E15, or thighs from postnatal day (P) 0-P3 wild-type or GDNF-overexpressing mice (MYO-GDNF) (Ngyuen et al., 1998) were dissected free from connective tissue and fat and collected in 200 μl of nondenaturing lysis buffer, pH 7.4 (0.1 m Tris buffer, 0.9% NaCl, 0.5% NP-40; 10 mm NaF, 1 mm NaV, and 1× protease inhibitors; Sigma). Samples were lysed by passage through a 22 gauge syringe, vortexed several minutes, and microcentrifuged for 30 min. Protein quantity from the samples was normalized using the Bio-Rad (Hercules, CA) assay for total protein determination. GDNF protein levels were quantified by using the GDNF EMAX ImmunoAssay system (Promega, Madison, WI). One hundred microliters of 10 mg/ml protein were loaded into each well of a 96-well plate. GDNF levels for P3 wild-type and MYO-GDNF mice were comparable with those of published reports (Nguyen et al., 1998). Because the secondary detection antibody is of a chicken IgY class, control wells were incubated with capture antibody and protein but without the polyclonal GDNF antibody. Absorbances obtained from these wells represent non-specific binding of the detection antibody to chick protein extract and were thus subtracted from each absorbance obtained in the presence of the polyclonal GDNF antibody. The plate was read at 450 nm.
Retrograde labeling. Embryos at E6, E8, E10, and E12 were dissected in Tyrode's solution at room temperature, and a ventral laminectomy from the mid-thoracic to sacral spinal cord was performed rapidly. Connective tissue and parts of the pelvic girdle were removed, and the individual muscles were identified by illumination from below, which, by providing visualization of the unique orientation of the myofibers within each muscle, allows one to reliably identify individual muscles (Landmesser, 1978). Specific muscles were injected with a pulled glass filament (1 × 0.75 mm; A-M Systems, Carlsborg, WA) with 20% HRP (type VI; Sigma) in Tyrode's solution or Alexa-Fluor 488- and Alexa-Fluor 594-conjugated β-subunit of cholera toxin (used at 1 mg/ml; Molecular Probes, Eugene, OR). After muscle injection, embryos were incubated at 32°C in Tyrode's solution containing 95% oxygen and 5% CO2 for 5-7 hr and then fixed overnight in 4% paraformaldehyde (for cell counts or for immunohistochemistry) or snap-frozen in OCT compound (Tissue-Tek) for in situ hybridization. The number of retrogradely labeled MNs was quantified in embryos injected with HRP by incubating 10-μm-thick sections (cut at -30°C on a cryostat) with DAB (Vector Laboratories, Burlingame, CA) to visualize the HRP, 10 μg/ml Hoechst 33342 (Sigma) to stain nuclei, and anti-islet1/2 to label MNs. Cells were counted as retrogradely labeled MNs if they exhibited fluorescent islet1/2 immunoreactivity and their nuclear membrane was completely within the plane of a section and surrounded by DAB-positive grains. These counts were corroborated by those in which HRP was detected immunohistochemically with an antibody (1:400, goat-anti-HRP; R&D Systems, Minneapolis, MN).
In situ hybridization. Embryos were dissected at various time points in cold PBS, and spinal columns were removed and snap-frozen in OCT medium. Digoxygenin- or biotin-labeled riboprobes were hybridized at 68°C overnight according to Schaeren-Wiemers and Gerfin-Moser (1993). Biotin-labeled probes were amplified using the tyramide signal amplification kit (NEN Life Sciences). After hybridization, slides were rinsed in 0.2× SSC for 1 hr at 68°C, rinsed with 0.2× SSC for 5 min at room temperature, rinsed in 1× TBS, blocked with 2% blocking reagent (NEN Life Sciences) in 1× TBS at room temperature for 1 hr, incubated in 1:1000 streptavidin-HRP in blocking solution for 45 min at room temperature, and incubated in 1:50 FITC-Tyramide in amplification diluent (NEN Life Sciences) for 30-45 min. For visualization of both retrogradely labeled MNs and mRNA in the same section, Hoechst 33342-stained, retrogradely labeled MNs were captured before hybridization. After hybridization, sections were blocked with 10% NGS/0.3% Triton-X, incubated with 1:100 anti-islet1/2, and visualized with a fluorescently labeled secondary antibody. Sections were restained with Hoechst 33342 and mounted. Although faint signals of the original retrogradely labeled MNs were still visible after in situ hybridization, combined images were prepared by correctly realigning marked sections by superimposing old and new Hoechst-stained images in separate channels in Adobe Photoshop (Adobe Systems, San Jose, CA) before capturing the mRNA image. In some cases, islet1/2-positive MNs were identified by immunofluorescence and pasted into the third channel (blue).
Reverse transcription-PCR and primers. Semiquantitative reverse transcription (RT)-PCR was performed by determining the range of the PCR cycle number in which band intensity increased linearly. Within the central part of the curve [19 cycles for glyceraldehyde dehydrogenase (GAPDH)], serial dilutions of cDNA (all of which were reverse transcribed within the same experiment) were run for different sets of primers to ensure the validity of the selected cycle number. The cycle number was 33 for BDNF with a temperature of annealing at 57°C, 27 for hepatocyte growth factor (HGF) at 56°C, 27 for GDNF at 60°C, 25 for neurotrophin-3 (NT-3) at 56°C, and 19 for muscle-specific GAPDH at 60°C. Primers for RT-PCR were: BDNF: 5′-CCTTTTCCTTACTATGGTTA, 3′-TTTTTCTCCGCTGCTGTTAC; GDNF: 5′-ATCTTCCCCAAAAACAGGTG, 3′-GGTGCTCGTCGTCTCGGTGA; growth-promoting activity (GPA): 5′-CCCCACGCCTCAGCCTCTTC, 3′-CTCAGCATCCACCACAGCCA; GAPDH: 5′-AGCGTGACCCCAGCAACATC, 3′-CAGCCTTAGCAGCCCCAGTG; HGF: 5′-GACATCCCCCTGTGCTCTGA, 3′-AACTCTTTCCATTGCCACGA; NT-3: 5′-TCTAAGCAGGTGATGGATGT, 3′-ACGCAGGAGGTGTCTATTCT; cRet: 5′-CTACCTCAGCAAGAGCCT, 5′-ATGCTCTGTGACTGAAAT. Denaturation was at 94°C for 45 sec, annealing for 45 seconds, and elongation at 72°C for 60 sec.
Primers for constructing riboprobes for in situ hybridization were: CNTFRα: 5′-ATCTTCCCCAAAAACAGGTG, 3′-GGTGCTCGTCGTCTCGGTGA; glycoprotein130 (gp130): 5′-CCCAGAGAAGCCCAAAAACT, 3′-AAGAACATTATTAGGAGCAG; p75: 5′-AACCAGACCGTGTGTGAGCC, 3′-TAGACAGGGATGAGGTTGTC; leukemia inhibitory receptor β (LIFRβ): 5′-GATTGTCAGGGGTTTTTAGA, 3′-GTTCCAGGCATTTGTGTTTG; GDNF receptor (GFR) α4: 5′-CGGCAGTGTATCGCAGGTA, 3′-CAGGCAGTTGGGTTTCTCC. The cMet-containing plasmid was kindly provided by C. Stern (Columbia University, New York, NY), TrkB and TrkC plasmids were provided by T. Large (Millenium Pharmaceuticals, Raleigh, NC), and cRet, GFRα1, and GFRα2 plasmids were from S. Homma (Fukushima Medical University, Fukushima, Japan).
Construct preparation and in ovo electroporation. Full-length primers against the beginning and end of chick CNTFRα, including 5′-NotI and 3′-EcoRI restriction sites, were used to clone this gene from E6 spinal cord RNA into pCMV-IRES1-GFP (Clontech, Palo Alto, CA). Primary transfection of PC12 cells but not chick MNs resulted in transgene expression, so the gene was cut by NotI, blunt-ended and excised with EcoR1 and cloned into appropriate sites in the pMES vector (β-actin promoter, IRES2-GFP; kindly provided by C. Krull, University of Missouri, Columbia, MO). After primary transfections of chick MNs confirmed transgene expression, the entire CNTFRα-IRES2-GFP construct was cloned into the EcoR1 and XbaI sites of the pTRE vector (Clontech). For electroporation, eggs were windowed at stages 15-18 and injected with a DNA solution containing pTRE-CNTFRα-IRES2-GFP and pCMV-TetR (2.5 μg/μl of each plasmid) diluted in 1× DNA gel-loading dye to visualize injection. Doxycycline (1 mg/ml; Sigma) was added at daily intervals from E6-E9 to initiate transgene expression.
In vivo blocking experiments. For the experiment examining endogenous NTF inhibition, GDNF, CNTF (5 μg), a neutralizing antibody to GDNF (20 μg; R&D Systems), or the CNTFRα antagonist AADH-CNTF (100 μg; from A. Di Marco and R. Laufer, Merck, Darmstadt, Germany) (Di Marco et al., 1996) was administered in ovo daily from E6 to E9 or on E6 and E8 (anti-GDNF). Adductor muscles were retrogradely labeled with HRP and counted at E10. For the experiments summarized in Figure 7, B-E, embryos were windowed at E3, the amniotic sac was opened at E6, and the hindlimbs (thighs) were injected with a pulled pipette attached to a Hamilton syringe at E7 containing 5 μg/μl anti-GDNF or 5 μg/μl mouse monoclonal IgG (Sigma) supplemented with 0.5 μg/μl Alexa-Fluor 594-conjugated β-cholera toxin and 0.1% trypan blue. Healthy embryos were killed on E8, and spinal columns were dissected and fixed in Carnoy's solution, embedded in paraffin and cut at 8 μm, and stained with thionin. Pyknotic nuclei were counted in every 20th section throughout the lumbar motor column at 600× (Clarke and Oppenheim, 1995). For visualization of retrogradely labeled MNs after antibody treatment, the spinal columns of other treated embryos were fixed in 4% paraformaldehyde, embedded in sucrose-OCT, cut at 10 μm, and stained with anti-islet1/2 and Hoechst 33342.
For the purposes of this study, we define NTF-Rs simply as receptors with a stimulation or inhibition that has been demonstrated to result in increased or decreased MN cell death in vivo, respectively. NTFs and the receptors examined in this study include GDNF (Yan et al., 1993), comprising the receptor tyrosine kinase cRet and the GDNF-receptor 1 (GFRα1) (Jing et al., 1996; Treanor et al., 1996; Trupp et al., 1996); NTN (Kotzbauer et al., 1996), acting through c-Ret and GFRα2 (Baloh et al., 1997; Klein et al., 1997); persephin (PSP) (Milbrandt et al., 1998), requiring c-Ret and GFRα4 (Thompson et al., 1998; Masure et al., 2000); HGF/scatter factor (Stoker et al., 1987; Nakamura et al., 1989), acting through cMet (Bottaro et al., 1991; Naldini et al., 1991; Weidner et al., 1993); BDNF (Leibrock et al., 1989), acting through TrkB and/or p75 (Rodriguez-Tebar et al., 1990; Klein et al., 1991; Squinto et al., 1991); NT-3 (Hohn et al., 1990; Jones and Reichardt, 1990; Maisonpierre et al., 1990), acting through TrkC (Lamballe et al., 1991); CNTF (Varon et al., 1979), acting through CNTFRα (Ip et al., 1993); and LIFRβ/gp130 (Davis et al., 1993).
NTF-Rs are expressed in spatially heterogeneous patterns in spinal cord MNs
Whereas the existence of NTF-R-expressing subtypes within sensory and MN populations is well established (Mu et al., 1993; Horton et al., 1998; Mickleas et al., 2000; Oppenheim et al., 2000; Garcés et al., 2001), a more detailed analysis of these patterns and their function in MNs during the period of cell death is lacking (E6-E12 in the chick lumbar spinal cord). We focused on differences in NTF-R expression in the lumbar spinal cord at E8, the peak of MN cell death. An expression study of GDNF family receptors has already been performed in this region (Homma et al., 2003). The expression of cMet was restricted to the medial lumbar lateral motor column (LMC) across the entire rostrocaudal extent, suggesting that only MNs that innervate muscles derived from the ventral premuscle mass are responsive to HGF (Fig. 1A) (Landmesser et al., 1978; Hollyday, 1980). TrkB and p75 fail to be expressed by MNs in the most lateral region of the LMC in segments LS4-7. Similar to cMet, CNTFRα mRNA is found throughout the medial LMC but it is also detected in specific regions of the lateral LMC. TrkC mRNA is expressed by most, but not all, LMC MNs, whereas mRNAs for LIFRβ, gp130, IGF receptor type I, and bone morphogenetic factor receptors (BMPR-1A and 1B) are expressed by virtually all lumbar LMC MNs (Fig. 1A and data not shown). These data indicate that individual regions or subpopulations of lumbar MNs express many different NTF-Rs, and second, that these subpopulations express different combinations of NTF-R. These differences may be sufficient, therefore, to explain differences in MN responses to NTF modulation.
Hindlimb muscles homogeneously express NTFs during the period of MN PCD
Although differences in NTF-R expression may contribute to the sculpting of MN numbers during the period of cell death, target muscles to which MNs project may also express NTFs differentially, thereby regulating the survival of specific subpopulations of MNs. For example, several, but not all, forelimb muscles in the mouse express GDNF during the process of MN target innervation (Haase et al., 2002). We analyzed the expression of NTF mRNA by semiquantitative RT-PCR in various hindlimb muscles throughout the period of MN cell death. All hindlimb thigh muscles expressed similar amounts of HGF, BDNF, GDNF, and NT-3 mRNA (Fig. 1B and data not shown). GPA, a CNTF-related CNTFRα-stimulating ligand, was never detected in the hindlimb from E6 to E10 but was expressed by the sciatic nerve starting at E16 (data not shown), consistent with previous observations (Sendtner et al., 1994). We also detected qualitatively similar amounts of GDNF and NTN protein in two different hindlimb muscles at E8 by immunoblot analysis (Fig. 1C). Additionally, quantitative GDNF protein levels, measured by ELISA, were similar in four hindlimb muscles at E8 (Fig. 1C), and all exhibited increasing amounts of GDNF from E6 to E10. We cannot rule out the possibility that the access of NTF protein is differentially spatially restricted within or between specific muscles. However, a homogeneous striated pattern of GDNF immunoreactivity was present in longitudinal sections of all muscles examined (Hashino et al., 2001) (Fig. 1C and data not shown). Furthermore, GDNF immunoreactivity appeared similar in all muscle cells within a muscle, independent of their proximity to nerve branches and terminals labeled with neurofilament and synaptic vesicle antibodies (data not shown). Collectively, these data suggest that both extrafusal and intrafusal muscle fibers (or fibroblast- and Schwann cell-containing nerves within these targets) do not appear to selectively express different amounts of NTFs during the period of MN PCD.
The adductor motor pool expresses a distinct combination of NTF-Rs
Because different hindlimb muscles fail to selectively express NTFs, we next asked whether the spatial differences in NTF-R expression within the lumbar spinal cord could be related to pools of MNs innervating individual muscles. Retrogradely labeled adductor MNs were identified within the medial half of the lumbar LMC from segments LS1-3, and groups of four adjacent labeled sections were then examined for NTF-R mRNA. Four distinct patterns were identified: (1) restriction to only MNs within the adductor and not adjacent regions (CNTFRα and cMet) (see Fig. 3C); (2) expression by both adductor MNs and adjacent regions (cRet, TrkB, TrkC, p75, gp130, LIFRβ) (see Fig. 3A-C); (3) exclusion from adductor MNs but expression by adjacent regions (GFRα2 and GFRα4) (see Fig. 3A); and (4) expression by a subpopulation of adductor MNs (GFRα1) (see Fig. 3A,D). In all but the fourth group, expression of NTF-Rs did not differentiate between individual adductor MNs. Thus, with the exception of GFRα1, NTF-R expression failed to identify subtypes of adductor MNs that are differentially equipped to compete for select NTFs.
The adductor is actually composed of two distinct extensors with different fiber type (the slow, medial, or superficial adductor and the fast, lateral, or deep adductor) (Dahm and Landmesser, 1991; Schroeter and Tosney, 1991). To directly test whether GFRα1 expression is limited to the medial adductor motor pool, we analyzed GFRα1 expression from LS1-3 in the spinal cord of embryos in which the medial adductor was removed, and only the lateral adductor was injected with retrograde tracer. Fig. 2D shows that lateral adductor-innervating MNs fail to express GFRα1. These results further suggest that MNs projecting to an individual adductor muscle, and therefore competing for survival, are not different in their complement of NTF-Rs. However, because GFRα1 is present within the medial, but not lateral, adductor pool, groups of MNs projecting to different adductor muscles express different combinations of NTF-Rs.
To more rigorously examine whether individual NTF-Rs are expressed by all MNs projecting to an adductor muscle, we analyzed the expression pattern at E8 of CNTFRα protein every 150 μm along the rostrocaudal extent of the adductor (medial and lateral) motor pool. Adjacent motor pools innervating the iliochanterici, sartorius, and femerotibialis muscles are also located in this same segmental region (Landmesser, 1978). In transverse spinal cord sections throughout LS1-3, in which the adductor, femerotibialis, and sartorius motor pools were retrogradely labeled with different tracers, CNTFRα immunoreactivity is restricted to the adductor pool and is expressed by virtually every islet1/2-positive adductor MN (Fig. 3). Because some MN NTF-Rs appear temporally regulated in their expression during the cell death period (McKay et al., 1996; Yamamoto et al., 1997; Novak et al., 2000), we analyzed adductor pool-specific patterns at E6, E8, and E10 and found them to be the same for all NTF-Rs (data not shown). However, after the cell death period, cMet mRNA was downregulated in adductor MNs, whereas other trophic receptors continued to be expressed by this motor pool until at least E15 (data not shown). Finally, we examined several other lumbar motor pools and failed to detect differences in NTF-R expression between MNs in these pools (Fig. 5; data not shown). Collectively, these results suggest that MNs within individual lumbar motor pools do not exhibit differential competence to survive the period of cell death attributable to NTF-R expression patterns.
Stimulation of adductor MNs with multiple, but not individual, trophic factor ligands completely prevents their death
Although different regions of the lumbar spinal cord express different combinations of NTF-Rs, virtually all MNs within a motor pool possess the same complement. In contrast, NTFs themselves appear to be expressed by all target muscles. To determine whether MN responses to excess NTF treatment in vivo are directly related to their NTF-R expression complement, we treated embryos in ovo from E6 to E9 with various NTFs and retrogradely labeled the adductor motor pool at E10. Treatment with BDNF, CNTF, GDNF, or HGF (but not NT-3) from E6 to E9 rescued a significant percentage of adductor MNs but failed to rescue them all, suggesting that NTF-R mRNA expression is insufficient to maintain the survival of all MNs, even in the presence of excess amounts of an individual NTF (Fig. 4B). The failure to rescue lumbar MNs in vivo with NT-3 is consistent with previous studies (Kalcheim et al., 1992; Oppenheim et al., 1992).
Although the addition of single NTFs rescued only a percentage of adductor MNs from PCD, combined treatment with BDNF, GDNF, CNTF, and HGF together from E6 to E9 completely blocked adductor MN cell death (Fig. 4B,C). The addition of any three of these factors failed to completely rescue adductor MNs (data not shown). Treatment with all four factors and NGF also rescued all adductor MNs, indicating that in this situation NGF could not activate death signaling in MNs through p75 (Hughes et al., 1993; von Bartheld et al., 1994). Rescue of all adductor MNs was maintained until E15, the last day examined, but required continued treatment with these excess factors during the period from E6 to E15 (data not shown). Finally, treatment with all four factors rescued 60% of all of the MNs normally destined to undergo cell death within the entire lumbar spinal cord [18,370 (treated) vs 11,540 (control); n = 4]. These results suggest that within populations of MNs, the receptor expression profile of which is well characterized, stimulation of multiple NTFs but not individual NTFs is sufficient to completely rescue all MNs from cell death.
Although NTF-R mRNA and protein are expressed by all adductor MNs, we noticed striking differences in the subcellular pattern of NTF-R immunoreactivity in lumbar MNs, using antibodies to cRet, TrkB, TrkC, and CNTFRα (chick-specific GFRα1 and cMet antibodies for immunohistochemistry are currently unavailable). Whereas cRet immunoreactivity was detectable only in proximal somata and ventral root, TrkB immunoreactivity was observed most strongly in dendrites (process extending throughout the ventral horn and especially into the white matter) and more weakly in the ventral root. CNTFRα immunoreactivity was present in axons in the ventral root, whereas TrkC immunoreactivity was detected only in cellular processes immediately outside of and almost encircling the entire LMC (Fig. 4D). These different subcellular patterns may contribute to the ability of different NTFs to collectively stimulate greater amounts of adductor MN survival (Janiga et al., 2000).
Nerve-derived NTF-Rs regulate MN survival through signaling in trans
Whereas the above results demonstrate that individual and combined NTF treatment is sufficient to rescue some or all adductor MNs, respectively, other observations led us to question whether stimulation of specific adductor MN-derived NTF-Rs was necessary for adductor MN survival. We pursued this question in two ways: first, by testing the degree to which stimulation of NTF-Rs that are not expressed by adductor MNs themselves could nonetheless contribute to their survival, and second, by directly interfering with adductor MN-derived NTF-R expression.
We noticed that although nearly every adductor neuron expresses cRet, only rostral medial adductor MNs express GFRα1. The in ovo addition of GDNF from E6 to E9 rescued slightly more than half of the total number of adductor MNs, and these appeared equally distributed between putative medial and lateral adductor motor pools (data not shown), suggesting that excess GDNF may have activated cRet in these GFRα1-negative lateral adductor MNs by some other mechanism. We envisioned two possibilities: GDNF indirectly rescues lateral adductor MNs by activating GFRα1 and c-Ret signaling in other cells, which then produce other molecules that rescue adductor MNs; or GDNF, which can activate GFRα2 at a high concentration (Klein et al., 1997), activates cRet in trans in lateral adductor MNs through the expression of GFRα1 or GFRα2 of an adjacent cell (Paratcha et al., 2001; Ledda et al., 2002).
To distinguish between these possibilities, we first analyzed the expression of various GFRα receptors in adjacent cell types. GFRα2 and GFRα4 mRNA, but not GFRα1 or cRet mRNA, was detected in the lumbar ventral roots and peripheral nerves of E8 chick embryos (Fig. 5B; Table 1). Because receptors for NTN and PSP are expressed by cells in motor nerves but not by adductor MNs themselves, we then added NTN or PSP to embryos from E6 to E9 and retrogradely labeled these MNs on E10. Similar to the effects of GDNF in rescuing GFRα1-deficient lateral adductor MNs, both NTN and PSP rescued GFRα2- and GFRα4-deficient adductor MNs from cell death (Fig. 5A). To exclude the possibility that exogenous GDNF, NTN, or PSP exerted their MN rescue effects indirectly through other cells coexpressing cRet and GFRα, we tested whether these factors could rescue MNs that fail to express cRet (Poteryaev et al., 1999; Trupp et al., 1999). We examined the responsivity of MNs innervating the medial gastrocnemius (MG) to daily (E6-E9) treatment with exogenous GDNF. These MNs, which express GFRα1 but not cRet, fail to be rescued from cell death by excess GDNF, NTN, or PSP, suggesting that cRet expression by MNs is necessary for them to respond to excess GDNF family ligands (Fig. 5C,D). In contrast, MG MNs express CNTFRα and in ovo treatment with CNTF promotes their survival, indicating that these MNs are not inherently unresponsive to NTFs (Fig. 5C,D). In summary, these results provide the first evidence that NTFs can rescue MNs in vivo through signaling in trans from nerve-derived GFRα-expressing cells to cRet-expressing MNs.
Target-derived NTF-Rs expression fails to mediate MN survival signaling in trans
Because NTF-Rs are expressed in other cell types known to interact with MNs (Table 1), we sought to characterize their capacity to either signal indirectly or in trans to MNs. A number of NTF-Rs are expressed by non-muscle cells within the limb (Table 1), including CNTFRα, the protein of which is expressed in a distinctive lattice-like pattern in all hindlimb muscles (Fig. 6A). This pattern is similar to that observed for fibronectin expression (Fig. 6B) and other proteins associated with the muscle extracellular matrix. To determine whether CNTFRα protein was expressed by cells within the muscle or exported there by MN axons, we examined muscles such as the sartorius, the MNs of which do not express CNTFRα. The same staining pattern could be detected in sections of this muscle (data not shown), indicating that CNTFRα immunoreactivity in the limb likely originates from cells within the muscle. We confirmed this by demonstrating CNTFRα mRNA in cells within the connective tissue between muscles as well as in small non-muscle cells within the muscle (Fig. 6C). Combined with our inability to detect CNTFRα in conditioned medium extracted from E5 MN primary cultures (data not shown), these data support the idea that fibroblasts or other non-muscle cell types within the muscle express CNTFRα protein in the extracellular matrix.
Because we have shown that nerve-derived (presumably Schwann cell-derived) GFRα2 and GFRα4 are able to signal in trans to cRet-expressing MNs, we next asked whether CNTFRα expressed by fibroblasts in the muscle could act similarly to rescue MNs expressing LIFRβ and gp130 but not CNTFRα. Whereas both the adductor and sartorius muscles express CNTFRα, only MNs innervating the adductor are also CNTFRα positive. We added CNTF in ovo from E6 to E9 and retrogradely labeled both these motor pools. CNTF rescued adductor but not sartorius MNs, indicating that CNTFRα stimulation in the target is not sufficient to rescue CNTFRα-negative MNs projecting to that target (Fig. 6D). To rule out the possibility that sartorius MNs are inherently incapable of responding to CNTF because of mechanisms other than the failure to express CNTFRα (e.g., inhibitory intracellular signaling), we ectopically expressed CNTFRα in the sartorius motor pool, exclusively during the period of MN PCD (Sato et al., 2002), by in ovo electroporation (6E,F). Treatment with CNTF from E6 to E9 rescued sartorius MNs in CNTFRα-overexpressing embryos (584 vs 415; n = 3; p < 0.05, Student's t test). Finally, to determine whether exogenous CNTFRα can mediate MN survival, we added CNTF and soluble CNTFRα from E6 to E9 and counted retrogradely labeled sartorius MNs. Cotreatment with CNTF and soluble CNTFRα rescued a significant percentage of sartorius MNs from cell death (Fig. 6D). These results indicate that whereas intracellularor extracellular-derived CNTFRα can activate gp130/LIFRβ in MNs in response to exogenous CNTF, CNTFRα expressed by fibroblasts does not initiate such signaling in trans to the axons of CNTFRα-negative MNs. Together, these results indicate that glycosylphosphatidyl inositol (GPI)-linked receptors in the nerve, but not in muscle, targets activate tyrosine kinases expressed by MNs in response to NTF signals.
Perturbation of GDNF or CNTF signaling fails to increase adductor MN cell death
Because MNs are rescued by some NTFs, the receptors of which are not expressed by MNs themselves, we then asked whether MN-derived NTF-R expression was necessary at all for MN survival. Accordingly, we blocked specific NTF signaling during the period of MN death and examined the effects of this treatment on MN survival. We used a mutant CNTF protein (AADHCNTF) that binds CNTFRα but lacks the amino acids necessary for the activation of LIFRβ and thus specifically antagonizes endogenous CNTF signaling through CNTFRα (Di Marco et al., 1996). Because the amino acids mediating this interaction are conserved between mouse and chicken (Heller et al., 1995; Duong et al., 2002), we first tested the ability of this antagonist to block survival induced by either CNTF or muscle extract (MEX) on primary MN cultures derived from E5 embryos. AADH-CNTF blocked MN survival induced by CNTF but not survival induced by MEX, confirming the effectiveness of AADH-CNTF and suggesting that molecules within MEX, other than CNTF, mediate MN survival in vitro (data not shown). We next added 100 μg of AADH-CNTF daily in ovo from E6 to E9 based on previous reports that this antagonist is effective in vivo (MacLennan et al., 2000; Xu et al., 2001). The number of retrogradely labeled adductor MNs on E10 after such treatment was similar to control (Fig. 7A). To ensure the ability of exogenously applied antagonist to block CNTF signaling, we added CNTF alone (E6-E9), AADH-CNTF alone, or the two together and found that the increase in survival mediated by CNTF on adductor MNs was completely blocked by coadministration of AADH-CNTF. Whereas we and others have been unable to demonstrate the presence of chicken CNTF (GPA) in the muscle or nerve during the period of MN cell death (Sendtner et al., 1994), the expression of a relatively novel CNTFRα-stimulating heterodimeric complex, cardiotrophin-like cytokine-1 (CLC)/soluble receptor cytokine-like factor-1 (CLF) (Senaldi et al., 1999; Elson et al., 2000), has been recently demonstrated to regulate MN survival (Alexander et al., 1999: Forger et al., 2003). Activation of LIFRβ and gp130 by CLC/CLF-mediated binding is inhibited by CNTF (Elson et al., 2000), suggesting that, in our study, AADH-CNTF administration in ovo also blocked these CNTF-related cytokines from activating CNTFRα, LIFRβ, and gp130 complexes.
We also tested the requirement for endogenous GDNF signaling to maintain the survival of adductor MNs during the period of cell death. Neutralizing antibodies against human GDNF successfully inhibit avian ciliary neuron target innervation (Hashino et al., 2001) as well as lumbar spinal cord expression of large-conductance calcium-activated potassium channels (Martin-Caraballo and Dryer, 2002), when administered in ovo or into hindlimb muscles. Therefore, we delivered similar concentrations of this antibody (15 μg) in ovo at E6 and E8 and quantified adductor MNs at E10. Whereas treatment with anti-GDNF, together with GDNF, attenuated the potent increase in adductor MN survival obtained with GDNF alone, treatment with the antibody alone failed to affect the number of MNs innervating the adductor, compared with control IgG-treated embryos (Fig. 7A). Because mice deficient in the GDNF gene exhibit a profound decrease in MN number during the period of PCD (Oppenheim et al., 2000), the failure of anti-GDNF treatment to affect adductor MN number was unexpected. To ensure that the antibody gained access to MNs, we directly injected a high dose (5 μg/μl) of anti-GDNF together with a retrograde tracer (Alexa-Fluor 594-conjugated cholera toxin β-subunit) at E7 directly into the hindlimb and counted dying lumbar MNs at E8. Figure 7, B and D, shows that anti-GDNF injection markedly increases the number of dying lumbar MNs (islet1/2 staining not shown), relative to control IgG thigh-injected embryos. Although these results at first appeared at odds with those generated from in ovo anti-GDNF treatment, further analysis revealed that the increase in dying MNs observed after anti-GDNF injection was strikingly restricted to the caudal lumbar spinal cord, which contains MNs that innervate non-adductor MNs (Fig. 7E). Because retrograde tracing revealed that the adductor motor pool indeed gained access to the limb-injected anti-GDNF (Fig. 7C), these results are consistent with those obtained from systemic anti-GDNF treatment and suggest that the survival of caudally located MN pools in the lumbar spinal cord is more susceptible than the adductor motor pool to the loss of GDNF signaling.
Our results indicate for the first time that individual MNs projecting to the same target and hence competing for survival fail to differ significantly in their expression of many of the most well characterized NTF-Rs. These data also demonstrate that treatment with excess individual NTFs is incapable of rescuing all NTF-R-positive MNs within such a competition unit. However, administration of excess multiple NTFs rescues all competing MNs, suggesting that concerted NTF stimulation is sufficient to override potential death-promoting pathways. Additionally, we provide the first evidence that NTFs can rescue MNs by acting through NTF-Rs on adjacent cell types in vivo. Finally, we demonstrate that expression of some NTF-Rs is not necessary for adductor MN survival, suggesting that in the absence of these pathways there can be compensation by the activation of other pathways, or alternatively (but less likely) that these pathways are not involved in mediating MN PCD. Together, these studies provide evidence for an unexpected complexity in the regulation of MN survival by NTFs.
NTF-Rs are expressed heterogeneously in the spinal cord and homogeneously in motor pools
One important finding of the present study is the homogeneity of expression of individual NTF-Rs within a motor pool. Virtually all adductor MNs either express particular NTF-Rs by E6 (the onset of MN cell death) or fail to express them. Whether the chief purpose of these expression patterns is to mediate survival competence or other aspects of development is unclear. In particular, cMet and CNTFRα are restricted to this pool at LS1-3, raising the possibility that these molecules are part of a genetic program related to adductor target identity. However, similar to the expression of ETS transcription factors and type II cadherins (Lin et al., 1998; Price et al., 2002), both of these receptors are also expressed by other pools in more caudal segments of the lumbar cord, indicating that their putative role in MN specification is not unique to the adductor pool. Additional ongoing studies are underway to analyze the developmental function of these patterns before the period of cell death (Ebens et al., 1996).
NTFs are expressed in all hindlimb muscles during MN PCD
Every NTF examined was present within all limb muscles examined between E6 and E10. Similarly, BDNF protein levels, measured by a highly sensitive electrochemiluminescent immunosorbent assay, fail to differ between chick thigh muscles during the period of MN PCD (Vernon et al., 2004). These results, together with the fact that every neuron within a pool expresses specific NTF-Rs, suggest that differential target expression of these molecules does not likely account for the normal reduction of MNs by PCD. This is consistent with a recent finding that motor axons remain largely unaffected by ectopic sources of NTFs in the periphery (Tucker et al., 2001). Because we only examined NTF distribution and levels in individual muscles during the period of cell death (E6-E10), and after motor axons arrive at their targets (Tosney and Landmesser, 1985), we cannot exclude a more spatially restricted expression pattern of specific NTFs at earlier times that may impart a subtype-specific MN fate (Haase et al., 2002). Although we made no attempt to discriminate extrafusal from intrafusal muscle fibers, the relatively homogeneous expression of GDNF within muscles suggests that this NTF is expressed by both these fiber types.
Stimulation with multiple, but not individual, NTFs completely rescues adductor motor neurons from PCD
Although NTF treatment has previously been observed to be incapable of rescuing some populations of NTF-R-positive MNs from death (Steljes et al., 1999; Caton et al., 2000; Novak et al., 2000), this effect extended to an entire subpopulation of competing MNs. In contrast, we showed that such subpopulations themselves are heterogeneous in responding to specific NTFs. Therefore, it appears that additional factors other than NTF-R expression itself regulate trophic responses to NTFs. The identity of these factors may include extracellular inhibitors of specific NTFs/NTF-Rs, such as noggin, which antagonizes BMPRs (Zimmerman et al., 1996), intracellular regulators of NTF-R phosphorylation, such as SHP phosphatases (Tran et al., 2003), or proteins that modulate any of the numerous downstream effectors, such as TRB3, a newly discovered negative regulator of Akt (Du et al., 2003).
One potential cellular mechanism for the differential effects of individual versus combined NTF treatment on adductor MN survival may involve the strikingly different subcellular expression patterns of NTF-Rs. For example, the increased subcellular pattern of TrkB protein in dendrites and its reduced expression in ventral root axons suggest that the source of BDNF for MNs during the period of cell death, like those during MN differentiation (Jungbluth et al., 1997), may originate from afferent input and not from target muscle (Okado and Oppenheim, 1984; Furber et al., 1987; Yan et al., 1993; von Bartheld et al., 1996). Because differential sources of NTF signaling may exert specific survival and growth effects on MNs (Arce et al., 1998; Janiga et al., 2000; Peng et al., 2003), it seems possible that the complete rescue of adductor MNs by excess stimulation of various NTF-Rs may mimic or amplify multiple developmentally relevant signals.
Nerve-derived, but not target-derived, GPI-linked NTF-Rs mediate MN survival in trans
We demonstrated that GPI-anchored receptors are able to mediate trophic signals to MNs in trans (Trupp et al., 1997; Yu et al., 1998; Paratcha et al., 2001) if they are expressed by cells in the nerve (GFRα2, GFRα4) but not when expressed by cells in the target (CNTFRα). Because GFRα2 but not GFRα1 is expressed by Schwann cells, the survival effect on lateral adductor MNs mediated by GDNF signaling in trans is likely attributed to this receptor because GDNF is capable of binding to GFRα2 with low affinity (Klein et al., 1997). Alternatively, GFRα1 expressed by non-muscle cells in the limb could mediate this effect (Table 1). However, this seems less likely because of (1) the dispensability of limb-derived GRFα1 expression for achieving appropriate Ret signaling to spinal MNs in vivo (H. Enomoto, personal communication), and (2) the inability of target-derived CNTFRα to signal to CNTFRα-negative sartorius MNs. Finally, because GFRα1 has been recently shown to physically interact with the neural cell adhesion molecule and mediate GDNF signals in the absence of cRet (Paratcha et al., 2003), we examined whether GFRα1-positive, cRet-negative MNs could be rescued from PCD by exogenous GDNF. Our results showed that at least for those MNs innervating the medial gastrocnemius muscle of the ventral shank, such signaling fails to prevent MNs from undergoing PCD. Thus, cRet expression by MNs is necessary for transducing the rescue effects of GDNF family members.
There are several reasons why GPI-linked receptor expression in the target may be unable to activate receptor tyrosine kinase activation in innervating MNs during the period of cell death. First, CNTFRα expression in the nerve, but not in the extracellular matrix of muscle, colocalizes with SV2-immunoreactive nerve terminals (data not shown), suggesting that fibroblast-derived CNTFRα is membrane bound or otherwise unavailable to contact axons. Second, cRet and TrkB are expressed in dendrites, cell body, and proximal nerve and are thus more accessible to nerve-derived than limb-derived GPI-linked receptors. Finally, non-muscle cells in the limb express cRet and muscle cells express LIFRβ/gp130 mRNA during the period of MN cell death (Table 1) (Helgren et al., 1994). Thus, target-derived GPI-anchored NTF-R may be expressed in the limb for the purpose of activating tyrosine kinase-positive cells in the limb and not in MNs.
GDNF and CNTF signaling are not necessary for adductor MN survival
Interference with GDNF and CNTFR activation potentiates the death of some MN populations (caudal lumbar MNs by GDNF) but fails to increase adductor MN cell death in vivo. Similarly, although the receptors for HGF and BDNF are expressed by the oculomotor nucleus, neither excess HGF nor BDNF rescue these MNs from PCD (Steljes et al., 1999; Caton et al., 2000; Novak et al., 2000). However, inhibition of other specific NTF signaling pathways does result in the increased death of some subsets of spinal MNs during the period of PCD (Garcés et al., 2000; Novak et al., 2000; Oppenheim et al., 2000, 2001; Forger et al., 2003).
For specific subpopulations of MNs, NTF-R signaling may mediate non-survival aspects of MN development before, during or after the cell death period (Henderson, 1996). For example, mice deficient in cRet appear to exhibit deficits in neuronal number before the onset of PCD (Enomoto et al., 2001; Oppenheim et al., 2002). Similarly, subpopulations of MNs in mice lacking cMet or GDNF fail to innervate their appropriate targets (Maina et al., 2001; Haase et al., 2002). Alternatively, it is possible that the absence of GDNF and CNTF receptor signaling is compensated by the activity of other NTF-Rs in adductor MNs, because both of these molecules signal through receptor tyrosine kinases shared by other trophic factor ligands (Stahl and Yancopoulos, 1994; Airaksinen and Saarma, 2002). However, mice deficient in several different individual NTFs or NTF-Rs exhibit profound MN losses by birth, suggesting that the function of these factors in survival is not always compensated by that of others. Ongoing experiments aimed at discerning the onset and subtype specificity of MN loss in various NTF mutants will address this issue. It will also be of considerable interest to examine NTF-R signal transduction in subtypes of MNs that are affected or not affected by trophic factor addition or inhibition in vivo.
This work was supported by National Institutes of Health Grant NS20402 to R.W.O. and by a grant from the Robert Packard Center for Amyotrophic Lateral Sclerosis research at Johns Hopkins University (Baltimore, MD). We thank Lynn Landmesser, Robert Oakley, and Woong Sun for technical advice.
Correspondence should be addressed to Dr. Ronald W. Oppenheim, Department of Neurobiology and Anatomy, Wake Forest University School of Medicine, Winston-Salem, NC 27157. E-mail:.
Copyright © 2004 Society for Neuroscience 0270-6474/04/244668-15$15.00/0