The detection limit of our enzyme immunoassay (EIA) system was as low as 5 pg/mL and the detection signal was dose-dependent below 1000 pg/mL, demonstrating that EIA was satisfactory for the measurement of endogenous GDNF in the injured or intact spinal cord. Furthermore, it has been confirmed that our EIA system is specific for GDNF among GDNF family proteins including artemin, neurturin, and percephin [11]. The recovery of exogenously added GDNF from the tissue homogenate was 76.7%, which confirmed the effectiveness of the measurement. Therefore, the values thus obtained by the EIA were used without correction. The intact spinal cord was dissected and cut into 9 segments of 5 mm length. The segments were numbered rostrocaudally from 1 to 9 (Figure 1A). GDNF content in each segment varied from 150 to 400 pg/g wet tissue weight (Figure 1B). The segments of the lumbar region (Nos. 7, 8, and 9) may contain much higher concentrations of GDNF than segments of other regions.

We examined GDNF content in the spinal cord after hemi-transection at T10 (Figure 2). In the transected side (left side), a marked increase in GDNF protein occurred in the rostral stumps that were adjacent or near to the injury site (segments Nos. 5 and 6) 12 h post injury (POI). The increase extended to more segments (Nos. 2–5) 1 day POI, and slightly reduced to 3 segments (Nos. 3–5) 3 days POI. The GDNF content gradually decreased 3 days POI and recovered to original levels by 7 days POI. However, no increase was observed in the caudal stumps throughout the experiments. In contrast, in the uninjured side (right), GDNF levels were not significantly high in all segments 12 h POI; however, its levels peaked in many rostral segments (segment Nos. 2–6) 1 day POI. GDNF levels returned to normal levels by 3 days POI. These results clearly demonstrated that injury-induced increase in GDNF content was conspicuous in the rostral part, but not in the caudal part, in both the injured and uninjured sides of the spinal cord.

It is likely that the non-transected side responded to the contralateral transection similarly but weakly and after a longer period of time [12,13]. It may be possible that the transection of half of the spinal cord influenced the other half by rapid gathering of macrophages/microglia to cause inflammation. The inflammation might stop axonal flow carrying GDNF protein from the rostral to caudal side, which action could result in the accumulation of GDNF at segment 6 and withdrawal of it at segment 7, thus accounting for the increase and decrease in segment 6 and 7, respectively. Since the response caused by the inflammation may be considered to be somewhat weaker and slower than the response to the transection, the increase in segment 6 of the uninjured side would be smaller than that in the corresponding segment of the injured side.

There are several possible reasons for the injury-induced increase of GDNF only in the rostral part of the spinal cord: (1) de novo synthesis of GDNF, (2) accumulation of GDNF being transported from rostral to caudal side of the spinal cord, (3) accumulation of GDNF being retrogradely transported within the peripheral nerves from the skeketal muscles to the motoneurons, (4) binding of GDNF to the injury-evoked GFRα1, coreceptor of GDNF, (5). supply of GDNF from the dorsal root ganglia (DRG).

The forelimbs may compensate for acute functional loss of the hindlimbs which may require more de novo synthesis of GDNF in the cervical enlargement. Therefore, GDNF protein might be synthesized de novo in a physical activity-dependent manner. To clarify possibilities (1), (2) and (5), we prepared an animal model with injuries of both a complete transection at T9 and a left side hemi-transection at T10 (Figure 3A). We separately measured GDNF content in the left and right side of each segment 12 h POI. If there was injury-induced enhancement of de novo synthesis in the rostral side, increase in GDNF content could occur in segment 5 and 6 of the left side and segment 5 of the right side, because they were rostral segments adjacent to the injury sites. The left and right segments 5 were similarly increased, but not the left segment 6 (Figure 3B), which could be explainable by a possibility that GDNF transporting from rostral to caudal side within axons was accumulated at segments 5 by interference with its transport. Furthermore, this result clearly confirmed that GDNF does not accumulate in the caudal stumps, suggesting that GDNF transport from caudal to rostral part does not function in the spinal cord (Figure 3B).

The possibility (5), supply of GDNF from the DRG, should be considered as a reason for the injury-induced rostral accumulation of GDNF. The DRG neurons project axons to relay neurons of specific laminae in the dorsal area of the gray matter of the spinal cord. A recent report [14] demonstrated that when afferent axons from the DRG reach the dorsal root entry zone of the spinal cord during embryonic development, they display a stereotyped pattern of T- or Y-shaped bifurcation. Therefore, DRG axons penetrate straightly into the dorsal root entry zone and form this bifurcation in the spinal cord to connect to target neurons. The anatomy of axons of DRG neurons is important to estimate the possibility of the present issue at hand. Hemitransection of the spinal cord might injure DRG neurons and consequently interfere with anterograde axonal transport of GDNF, resulting in accumulation of GDNF within these axons. However, it should be noted that axons of DRG neurons penetrate perpendicularly to the spinal cord, and orthogonally to the axons that run up and down within the cord. Therefore, only DRG axons in the vicinity of the hemitransection site would be injured and accumulate GDNF being transported within them. DRG axons apart from the injury site would function normally and show no change in GDNF distribution. If the axons of DRG neurons were injured by transection of the spinal cord, GDNF from the DRG would have accumulated around the injury sites in both rostral and caudal sides of all injury sites to which DRG axons project. Therefore, if this were the case, the two-point transection model shown in Figure 3 would have given different results. Assuming that injury to the axons of DRG neurons was responsible for GDNF accumulation in the spinal cord, then all segments both rostral and caudal to the transection sites should have contained a substantial level of GDNF. From these results and considerations, possibility (5) was excluded.

However, another explanation that the neurons located in left segment 6 which synthesized GDNF de novo would die following double transaction injury due to progressive disruption of long axon tracts and extensive tissue loss might be possible. As this possibility was based on the injury-enhanced neuronal GDNF synthesis, we evaluated GDNF mRNA expression in the all segments including left segment 6 after hemi-transection.

We evaluated GDNF mRNA expression in each segment after SCI by RT-PCR (Figure 4). In the transected left side, GDNF mRNA was evenly detected in both rostral and caudal stumps adjacent to the injury site 6 h POI onset until at least 3 days POI examination. In the non-transected right side, the expression was similarly detected in both rostral and caudal stumps; however, the expression was weaker and more transient compared with that in the left side probably because of the low severity of the injury. These results demonstrated a mismatch of GDNF mRNA and GDNF protein in their distribution after SCI. Therefore, enhanced de novo synthesis of GDNF was unlikely as a reason for rostral GDNF increment, because of the inconsistency between mRNA expression and GDNF protein levels. Therefore, it is possible that marked increase in GDNF protein in the rostral side may be responsible for the accumulation of GDNF protein that is transported within the spinal cord. We did not provide the results on mRNA expression in the sham-operated controls because it was almost the same as the results for the “control (0 h after SCI)” shown in Figure 4. Namely, GDNF mRNA expression was weak or lacking in the sham-operated uninjured spinal cord, suggesting that the substantial level of GDNF protein detected in the spinal cord (Figures 1 and 2) was not largely responsible for de novo synthesized GDNF in the spinal cord. This possibility supports constitutive transport of GDNF within the spinal cord.

The injury-induced expression of GDNF receptors [15] might cause GDNF increase in the rostral part, enhancing binding and accumulation of GDNF in this region. We examined time course of regional mRNA expression of GFRα1, GDNF binding coreceptor, after hemi-transection. GFRα1 mRNA increased rapidly from as early as 3 h POI in almost all segments of both transected and non-transected sides, and the increment continued till 3 days POI. Therefore, we noticed that the injury-induced enhancement of GFRα1 mRNA expression occurred in extensive, not limited to the rostral part of the injury site, suggesting that GFRα1 was not involved in mechanisms for rostral GDNF accumulation.

We expected at first that hemi-transection of the spinal cord would make a great difference between ipsilateral and contralateral sides. Contrary to this expectation, both sides showed a similar change, although response of the non-transected (right) side was somewhat weak and slow. However, the rapid increase of GDNF in rostral side was observed only in the transected left side. In over one day, both sides showed a similar GDNF distribution. It is likely that the transection of half of the spinal cord influences on the contralateral half by rapid gathering of macrophages/microglia to cause inflammation. Our present results demonstrated that injury-induced GFRα1 mRNA expression occurred similarly in both transected and non-transected sides as early as 3 h after the injury, suggesting that hemi-transection caused similar influence on the transected and non-transected side (Figure 5).

Distribution of GDNF-immunoreactivity (ir) in the intact spinal cord is shown in low- and high-power photographs (Figure 6). GDNF-ir was observed in the cell bodies of motoneurons and in small cell bodies of the dorsal area. The antibody preabsorbed with peptide containing the antigen sequence gave no significant signal (data not shown), demonstrating antibody specificity for GDNF. Our present results (Figure 4C) demonstrated that GDNF mRNA expression was low in the intact spinal cord when compared to the expression around the injury site, supporting the fact that GDNF mRNA was undetectable in motoneurons [16]. In contrast, GDNF generates signals through a receptor complex comprising coreceptors (GFRα1) and Ret tyrosine kinase. GFRα1 mRNA is known to be expressed in spinal neurons, including motoneurons [17], and widely distributed in the whole spinal cord. Furthermore, another molecule for signal generation, c-Ret, is also expressed in motoneurons [18,19]. Therefore, it is likely that GDNF-ir found in spinal cells, including motoneurons, was derived at least partly from GDNF synthesized elsewhere and bound to the neurons.

To obtain evidence that GDNF is transported and accumulated in the rostral side, we examined the distribution of GDNF-ir around the rostral stumps. As shown in Figure 5D, GDNF-ir did not show any obvious change in distribution and/or intensity. Therefore, we could not obtain such evidence, but exclude a possibility that GDNF being retrogradely transported within the peripheral nerves from the skeketal muscles to the motoneurons.

A question why the GDNF-ir was less clear in the immunohistochemical analysis than the EIA may arise. This may be due to differences in the principle of the detection. That is, the GDNF transporting within the axons was thought to be biologically active and diffusible form. Such moleculesm might be easily extracted from the unfixed tissues, and might be released unexpectedly from the fixed tissues. Therefore, detection efficiency of GDNF by immunohistochemical technique might be lower than it was. In this point of view, EIA is reliable to quantify active and diffusible GDNF molecules.

Another type of GDNF-ir-positive cell was detected in both the rostral and caudal rims of the transection site. Most of these cells co-expressed ED-1 antigen and GDNF-ir (Figure 7), demonstrating that they were microglia or macrophages. This GDNF-ir expression by microglia/macrophages was timely and regionally coincident with the injury-induced mRNA expression observed in both the rostral and caudal stumps adjacent to the injury site (Figure 4C). However, amount of GDNF synthesized by the immune cells was thought to be small, because the EIA for GDNF could not detect elevations in GDNF levels of the caudal stumps, except for segment 7 one day POI with significant increase (Figure 2).

Animals were handled in accordance with the Guidelines of Experimental Animal Care issued by the Office of the Prime Minister of Japan. All efforts were made to minimize the number of animals used. Female Wistar rats (7 weeks old, n = 51 animals, Nippon SLC) were anesthetized with pentobarbital (35 mg/kg i.p.), and the spinous process and the vertebral lamina were removed to expose the whole cord at the laminectomy site. Half of the spinal cord (left side) was then transected at thoracic position (T) 10 with a razor blade. After surgery, the superficial back muscles were sutured along the midline and the skin was closed with Michel wound clips. The animals were fed for various periods of time and then anesthetized again. The left side or right side of the spinal cord was separately dissected from each animal, cut into 6 pieces of 5 mm in length and stored at −80 °C until subsequent analysis.

Each sample was pulse-sonicated in 5% (w/v) 0.1 M Tris-HCl buffer, pH 7.6, containing 1 M NaCl, 0.1% bovine serum albumin (BSA), 2 mM EDTA, 80 trypsin inhibitor units/L aprotinin (Sigma-Aldrich, St. Louis, MO, USA), and 0.02% NaN₃ and centrifuged at 100,000× g for 30 min. The supernatant was mixed with an equal volume of chloroform to remove lipids and centrifuged again at 20,000× g for 30 min, following which the aqueous phase was used for the measurement of GDNF protein.

Enzyme immunoassay (EIA) was performed according to a method described previously [11]. In brief, anti-GDNF antiserum was prepared from rabbits immunized with recombinant GDNF (provided by Amgen Inc., Thousand Oaks, CA, USA), and anti-GDNF antibody was purified from the antiserum using a GDNF-linked column (Affi-gel 10; Bio-Rad, Hercules, CA, USA). Affinity-purified anti-GDNF antibody (10 mg/mL) in 0.1 M Tris-HCl buffer (pH 9.0) was coated onto the well of 96 U-bottom multiwell plates (5 mL/well) and incubated at 25 °C for 2 h. For evaluation of the background signal, control wells were treated with normal rabbit IgG. The wells were washed with 0.1 M Tris-HCl buffer (pH 7.6) containing 0.4 M NaCl, 0.1% BSA, 1 mM MgCl₂, and 0.02% NaN₃ (washing buffer), and non-occupied space was blocked by incubation with 100 mL/well of 1% (w/v) skim milk for 1 h. After washing, 30 mL of test sample or serially diluted recombinant GDNF was incubated in the wells for 2 h at 25 °C. Then, 30 mL of affinity-purified biotinylated antibody (10 ng/mL) diluted in the washing buffer was incubated for 12–18 h at 4 °C. Finally, 30 mL of β-d-galactosidase-conjugated streptavidin was added to each well. After incubation at 25 °C for 1 h, bound β-d-galactosidase activity was measured by incubation with 30 mL of 30 mM 4-methylumbelliferyl-β-d-galactoside. The intensity of fluorescence was monitored at 360 nm excitation and 448 nm emission (Model F-2000, Hitachi, Tokyo, Japan). The standard curve of GDNF was used for determination of concentrations.

Rats were anesthetized and successively perfused with cold phosphate-buffered saline (PBS, 50 mL) and cold 4% (w/v) paraformaldehyde solution prepared in 0.1 M phosphate buffer, pH 7.3 (200 mL). In the case of injured rats, 5 mm sections of the spinal cord were cut out from the transection site rostrally or caudally. These segments were postfixed for 2 h in the same fixative and frozen in embedding compounds (Miles, Elkhart, IN, USA). Coronal sections of 20-μm thickness were prepared with a cryostat (Model CM 1800, Leica, Wetzlar, Germany). For immunostaining, sections were rinsed in 0.1 M Tris-HCl buffer, pH 7.6, containing 0.3% (v/v) Triton X-100 (TT buffer) at 4 °C for 1 day to render the cell membrane permeable to antibodies and incubated with anti-GDNF rabbit antibody (0.3 μg/mL; d-20, Santa Cruz Biotechnology Inc., Santa Cruz, CA, USA) or anti-CD68 antibody (ED1) to recognize the antigen of microglia/macrophages (0.5 μg/mL; Antigenix America Inc., Huntington Station, NY, USA) in TT buffer at 4 °C for 1 day. The epitope of the antibody (D20) was mapped near the C-terminus of GDNF of human origin and recommended for detection of mouse, rat, and human GDNF. The sections were successively treated with 0.3% (v/v) H₂O₂ at 25 °C for 30 min to quench endogenous peroxidase activity, and 2% skim milk to minimize non-specific binding. They were further reacted with rhodamine-conjugated goat anti-rabbit IgG (Invitrogen, Carlsbad, CA, USA) and fluorescein isothiocyanate (FITC)-conjugated goat anti-mouse IgG (Invitrogen). Images were observed with a confocal laser microscope (LSM 510, Zeiss, Oberkochen, Germany).

RT-PCR was performed as described [3]. The specific primers used were as follows: forward primer 5′-GAGAGGAATCGGCAGGCTGCAGCTG-3′ and reverse primer 5′-CAGATACATCCACATCGTTTAGCGG-3′ for GDNF (product size: 337 bases); forward primer 5′-CCGGGCAGTCCCGTTCATA-3′ and reverse primer 5′-TCAGTCCCGAGTAGGCCAGGAG-3′ for GNRα1 (product size: 482 bases); and forward primer 5′-CGGAGTCAACGGATTTGGTCGTAT-3′ and reverse primer 5′-AGCCTTCTCCATGGTGGTGAAGAC-3′ for glyceraldehyde-3-phosphate dehydrogenase (G3PDH) (product size: 309 bases). The G3PDH gene was used as an internal control. After amplification, PCR products were subjected to 2% agarose gel electrophoresis and visualized by ethidium bromide staining. Images were captured with FLA-5100 (Fuji Film, Tokyo, Japan).

Data are presented as mean ± SE. The statistical significance of the differences between the two groups was assessed using Student’s t-test.