Extensive structural remodeling of the injured spinal cord revealed by phosphorylated MAP1B in sprouting axons and degenerating neurons

Animals

Adult female Wistar rats (250 g; n = 32) were used in this study. Experimental procedures (authorization no. 75122) followed the European Communities Directive (86/609/EEC) and were approved by the Ethics Committee 3, Ile-de-France Region.

Surgery

Animals were anesthetized by i.p. injection of a solution containing xylazine (5 mg/kg; Rompun 2%, Bayer, France) and ketamine (100 mg/kg; Imalgène 500, Merieux, France). Laminectomy was performed at thoracic levels 8–9 (T8–T9) under visual guidance using an operating microscope. The skin was cut along the midline of the back, paravertebral muscles of the T8–T9 vertebrae were dissected out to expose the underlying spinal cord, the dura mater was opened and the spinal cord hemisectioned on the right side with iridectomy scissors. Muscles and skin were sutured, and the animals were left to recover, and returned to their home cages with free access to food and water. After survival times of 3–90 days post-lesion, animals were deeply anesthetized with pentobarbital, perfused transcardially with NaCl (0.9% at 37 °C), followed by 4% paraformaldehyde in 0.1 m phosphate buffer, pH 7.4 (PB). Spinal cords were dissected and cryoprotected in 30% sucrose in 0.1 m PB containing NaCl (0.9%; PBS). Serial sections (40 µm) were cut on a cryostat and maintained in PBS at 4 °C.

In some animals, the axonal tracer Dextran Biotin Amine (DBA 10000; 10%) was pressure-injected via a Hamilton syringe at 15 days post-lesion into the cortex (A:0.26, L:1.4, H:1.75/A:0.48, L:2, H:1.75/A:1, L:2.4, H:1.75) for tracing of right corticospinal tract (CST) projections. Animals were perfused 1 week thereafter.

Immunohistochemistry

Immunohistochemistry was performed as previously described (Soares et al., 2002). MAP1B and MAP1B-P immunostaining were performed using rabbit polyclonal antibody 1B-poly, and monoclonal antibody (mAb) 1B-P, respectively (hybridoma generously provided by Dr I. Fischer; Nothias et al., 1996). Neurofilaments were labeled with mAb SMI-31 (1 : 2000) and SMI-32 (1 : 4000; Interchim, France). Peptidergic primary sensory afferents were specifically labeled using a rabbit polyclonal antibody against calcitonin gene-related peptide (CGRP; 1 : 10 000, Peninsula, UK); non-peptidergic primary sensory afferents were specifically labeled using isolectin B4-FITC (IB4, Sigma, France). Macrophages that are also stained with IB4 are easily distinguished from IB4-positive axons through their morphology. Stable microtubules were detected in vitro by using monoclonal anti-acetylated tubulin antibody (1: 2000; Sigma) and beta3-tubulin by using monoclonal antibody Tuj-1 (1 : 1500; Babco, Richmond, CA, USA). Activated P-JNK was detected by a polyclonal antibody (1 : 400; Promega). Synaptophysin was detected by using a polyclonal antibody (1 : 1000, DAKO). Nuclear staining was peformed using DAPI staining. Sections were analysed on a standard light (Leica DMRB) or a confocal microscope (Leica SP1).

Relative quantification of MAP1B-P immunostaining in gray matter was performed using ImagePro-Plus software (Media Cybernetics, Silver Spring, MD, USA) on images of serial sections performed at 8, 15 and 30 days post-injury. On each section, separately for ipsi- and contralateral side to the lesion, the gray matter was divided into three areas. The number of pixels corresponding to MAP1B-P immunostaining was calculated for each area, and compared with the total surface area. For each region, the value was compared with that corresponding to the same region in a normal spinal cord and the ratio represented as histograms. Means for each area at each time point were calculated. Care was taken to ensure equal background staining intensities for each section examined.

Culture of dorsal root ganglia neurons

Adult dorsal root ganglia neurons were cultured as described (Bouquet et al., 2004). For some experiments, the following drugs were added after growing neurons for 24 h in vitro: SP600125, a specific inhibitor of activated P-JNK (Calbiochem), at 40 µm (this concentration was found to yield the maximum effect) – after 24 h, cells were fixed as described (Bouquet et al., 2004); and staurosporin (Sigma), at different concentrations (0–1 µm), used to induce apoptosis – cells were fixed 8 h after treatment. Immunocytochemistry on cell cultures were performed as described (Bouquet et al., 2004).

N2A cell experiments

N2A (neuroblastoma) cell line (generously provided by F. Propst) cultures were grown on poly lysine until confluence in DMEM−10% FCS (Invitrogen) at 37 °C and 5% CO₂. SP600125 (40 µm) was added for 24 h. For Western blotting, wells were rinsed twice with PBS, and cells solubilized in lysis buffer [SDS 4%; Tris 0.125 m, pH 6.8; EDTA 12 mm; DTT 0.3%; glycerol 20%; anti-protease cocktail (Sigma); Na₃VO₄ 1 mm]. Samples were sonicated, centrifuged and incubated for 10 min at 65 °C before loading on a 4% polyacrylamide gel (protein concentration was normalized for each sample). After transfer, nitrocellulose membranes were incubated overnight at 4 °C in BSA 5%–PBS–Tween 0.25%, then with polyclonal anti-total MAP1B in BSA 2%–PBS–Tween 0.25%, followed by goat anti-rabbit–alkaline phosphatase (1 : 3000). NBT/BCIP was used for the enzymatic detection.

MAP1B-P immunocytochemistry on N2A cells was done using the same procedure as described for primary DRG neuron cultures (see Bouquet et al., 2004).

Distribution of MAP1B-P in intact adult spinal cord

In intact adult spinal cord, MAP1B-P is never detected in neuronal somata. MAP1B-P immunostaining within gray matter is limited to fibers exhibiting the somatotopic distribution of myelinated primary sensory fibers, i.e lamina I, intermediate gray matter, and ventral horn (Fig. 1‘control’, Fig. 3c; see also Soares et al., 2002). MAP1B-P is also found in white matter, in association with most ascending and descending pathways (1, 2-9), with the notable exception of the CST. Dorsal columns are strongly labeled, as well as dorsal and ventral roots, reflecting the presence of MAP1B-P in primary sensory afferents and motor axons, as described previously (Ma et al., 2000). Finally, the ventral funiculus is also highly stained around the midline, especially on lumbar sections.

Lamina II (L-II; substantia gelatinosa) is normally devoid of MAP1B-P. CGRP, expressed by small and medium-size DRG neurons, is present in unmyelinated and thinly myelinated primary afferents terminating in laminae I and II (outer part;Fig. 5m, left) of the dorsal horn. Small-diameter non-peptidergic neuron terminals that bind Griffonia simplicifolia lectin (isolectin B4; IB4) are also present within the inner part of lamina II (Fig. 5g, left; Kitchener et al., 1993). Neither of these fiber populations co-label with MAP1B-P.

Spatio-temporal evolution of MAP1B-P distribution after lateral hemisection at T8–T9

Lateral hemisection induces marked changes in MAP1B-P staining, rostral and caudal to the lesion, both ipsilaterally and, to a lesser extent, contralaterally to the lesion.

MAP1B-P levels are increased in fibers located within gray matter

8 days post-lesion. At the lesion epicentre (Fig. 1,‘Lesion’), the dorsal horn is disrupted, and the lesion site, where a cavity has not yet formed, is filled with numerous MAP1B-P-positive fibers, intermingled with non-neuronal cells that invade the lesion (for details and references, see also von Boxberg et al., 2006). On sections close to the injury, MAP1B-P staining is strongly increased, in particular in the median to ventral part of the gray matter (Fig. 1, −1500 µm, −2400 µm). At high magnification, MAP1B-P fibers are thin and their density is high (Fig. 3a). Staining is densest proximal to the injury, particularly in dorsal and ventral roots (arrows), where individual fibers cannot be distinguished. While more prominent ipsilateral to the hemisection, an increase in MAP1B-P labeling is also observed on the contralateral, unlesioned side. Further rostrally, the distribution of MAP1B-P-positive fibers in dorsal horn is not obviously affected (Fig. 1, −6500 µm), but staining density is increased in laminae I and III, and in L–X and the median portion of L–VII around the central canal on both sides of the spinal cord, and as in L–IX, around motoneuron pools. Just caudal to the injury (Fig. 1, +950 µm), MAP1B-P staining density and distribution in dorsal horn are comparable with those in control sections, but abnormal MAP1B-P staining is detected on numerous fibers in intermediate gray matter both ipsilateral and, to a lesser extent, contralateral to the lesion. Further caudally, gray matter is still abnormally highly positive for MAP1B-P in both the ventral horn and around the central canal ipsilateral to the lesion (Fig. 1, +12 500 µm).

15 days post-lesion At the lesion epicenter, a large cavity has developed (Fig. 2,‘Lesion’) that is surrounded by MAP1B-P-positive fibers. The distribution of highly MAP1B-P-positive fibers within the gray matter has extended both rostrally and caudally (Fig. 2, −2800 µm, see also Fig. 4) compared with at 8 days (Fig. 1, −2400 µm). A higher magnification of labeled fibers in the median and ventral part of the gray matter is shown in Fig. 3b and d. The increase in MAP1B-P immunostaining and its abnormal localization in gray matter is bilateral, although always more prominent ipsilateral to the lesion.

Quantification of the density of MAP1B-P immunostaining (Fig. 4) confirmed the qualitative results presented in 1, 2.

Thus, 8 days after the lesion (even as early as 3 days, not shown) a massive increase in MAP1B-P immunoreactivity is detected in gray matter, starting around the lesion site and gradually extending rostrally and caudally. At this time-point, mean values for lesion-induced MAP1B-P labeling are generally higher on caudal than on rostral sections, and highest overall values are observed around the central canal (middle panel), close to and caudal to the lesion site.

MAP1B-P immunostaining then reaches a peak at 15 days post-injury, in all three regions examined. Around the central canal, values are now higher for rostral sections, while they appear to decline on caudal sections. Thus, sprouting revealed by MAP1B-P labeling appears to be region- and time-dependent. L-X (see bottom panel) shows highly increased MAP1B-P immnoreactivity between 8 and 15 days post-lesion both rostral and caudal to the lesion site.

By one month, the rostro-caudal extent of MAP1B-P-positive fibers within gray matter has dramatically decreased in all three areas. Labeled fibers are mostly restricted to the region around and caudal to the lesion. The lesion cavity is still present, and as at 15 days, surrounded by some stained fibers (not shown). The distribution shown at 1 month persists up to 3 months (not shown).

Hemisection induces sprouting of both large and small caliber primary afferents into the dorsal horn

Using lesion paradigms that do not directly attain the spinal cord (peripheral lesion and rhizotomy), we previously demonstrated that increased MAP1B-P levels are associated with central sprouting of myelinated, but not of CGRP-positive fibers (Soares et al., 2002). We now investigated the distribution of primary sensory afferents after spinal cord hemisection, using MAP1B-P and CGRP as specific markers for large myelinated and thin peptidergic fibers, respectively. In addition, we used IB4 as a marker for thin non-peptidergic fibers. Figure 5 illustrates the distribution of all three labels at 15 days post-injury.

In L-II of the intact dorsal horn, no MAP1B-P staining and hence no co-localization with CGRP or IB4 is noted in thin primary sensory afferents. A co-localization of CGRP with MAP1B-P is seen only in very few fibers in L-I (see also Soares et al., 2002). Following a hemisection, CGRP levels are markedly decreased rostrally near the lesion site (Fig. 5o and p), and IB4-positive fibers have completely disappeared from L-II (Fig. 5i and j). MAP1B-P staining is also decreased in dorsal horn (Fig. 5c and d; laminae I and III). This decrease in specific staining for all types of primary sensory afferents is undoubtedly due to damage of the dorsal root projecting into the hemisectioned spinal segment.

On sections further rostral or caudal to the injury, IB4-positive fibers are absent from the lateral (Fig. 5g and h) and medial (Fig. 5k) parts of L-II. Interestingly, we noted that the area of L-II devoid of IB4-positive fibers is now occupied by MAP1B-P-positive fibers (compare Fig. 5b; 5e and Fig. 5h and Fig. 5k). Careful analysis of rostral and caudal sections suggests that MAP1B-P-positive fibers from adjacent roots not damaged by the hemisection have sprouted into the region of L-II that normally receives projections of IB4-positive fibers from the damaged root. Indeed, in cases where dorsal roots were spared by the lesion, MAP1B-P staining is dramatically increased in the dorsal root, and hence also in the whole dorsal horn (see for 8 days: Fig. 1, −1500 µm).

These changes in primary sensory afferent immunostaining within L-II were observed at all time points post-injury analysed.

Partial co-localization of MAP1B-P with CGRP can be detected in ectopic sprouts at the lesion site, early after injury (8 days: Fig. 6d- f), and at later post-injury stages (1–3 months after the lesion; not shown), in fibers crossing the midline in the dorsal-most part of the median gray matter above the central canal.

Abnormal MAP1B-P accumulation in neuronal somata in gray matter close to the injury site

In intact tissue, MAP1B-P is exclusively located in axons. Following injury, we observed a strong MAP1B-P concentration in neuronal somata. In particular, on sections close to the injury site from 3 to 21 days post-lesion (Fig. 8; compare with control provided in Supplementary material, Fig. S1), Clarke's nucleus neurons are highly labeled. These cells are located caudal to the lesion (Lumbar segment 1) and project their axons rostrally in the dorsospinocerebellar tract, which is destroyed by the lateral hemisection, thus axotomizing Clarke's neuron (Fig. 8a, g and i). These cells are particulary vulnerable to spinal cord injury, and about half die by apoptosis within 2 months (Himes et al., 1994). MAP1B-P accumulation in the soma of these neurons may thus be indicative of changes in their cytoskeleton that precede their final degeneration. We also detected abnormal accumulation of other cytoskeleton components in the soma of these neurons, such as beta3-tubulin (Fig. 8b–d), and neurofilaments (Fig. 8e). Therefore, we investigated the expression of activated caspase-3 (Fig. 8i), known to be induced in neurons during early stages of apoptosis (Pettmann & Henderson, 1998). Indeed, activated caspase-3 co-localizes with MAP1B-P in neuronal somata (Fig. 8i), confirming our hypothesis that abnormal accumulation of MAP1B-P precedes death by apoptosis. TUNEL labeling performed at 1 week post-lesion revealed that MAP1B-P-positive Clarke's neurons in gray matter had not yet entered the final stage of apoptosis at this time-point, as no co-localization was observed (not shown).

To verify further our hypothesis that MAP1B phosphorylation is abnormally increased in axons and neuron cell bodies undergoing degeneration, we set up cultures of adult DRG neurons, in which apoptosis was artificially induced by addition of staurosporine (Boix et al., 1997). In control (untreated) neurons, total MAP1B is detected in all cytoplasmic compartments, cell bodies and neurites (Fig. 10d), whereas MAP1B-P is found only in neurites, distributed in a proximal to distal gradient (highest in growth cones; Fig. 10e). In staurosporine-treated cultures (Fig. 8m–q), MAP1B-P indeed accumulates in cell bodies, and in the proximal part of neurites (Fig. 8n) the proximo-distal gradient is lost. Processes that have not yet retracted or degenerated undergo swelling in their median part, with MAP1B-P appearing to spread out (Fig. 8m–o). Tubulin polymers appear also disorganized (Fig. 8q), and some neurites are seen to disintegrate (appearance of fissures along the neurites, Fig. 8m, arrow). These in vitro observations thus closely resemble the morphological state of affected neurons seen in vivo after a spinal cord lesion.

MAP1B-P shows massive structural remodeling of fibers that extends far beyond previous descriptions

We have previously shown that MAP1B-P reveals the capacity of central sprouting of primary sensory myelinated afferents in response to peripheral injury. These fibers also undergo sprouting after SC injury, as revealed by Krenz & Weaver (1998). Using MAP1B-P, we demonstrate here that sprouting of these fibers is more robust spatially, and more persistent temporally; spared dorsal roots proximal to the injury exhibited a massive increase in MAP1B-P staining detected at 3 days and up to 3 months post-injury. To confirm that these seemingly contradictory results were not due to differences in the experimental paradigms used (hemi- vs. complete transection), we also performed a complete spinal cord transection, which again yielded results similar to those described here (not shown). MAP1B-P is distributed in a proximo-distal gradient within axon shafts, and thus detected mainly at their terminals, as we have previously shown after a neurodegenerative injury in vivo (Soares et al., 1998) and here in vitro, in regenerating adult DRG neurites. This particular distribution might explain why after a hemisection, detection of MAP1B-P within corticospinal axons is limited to their projection areas within gray matter.

Interestingly, MAP1B appears also to play an important role in the maturation of neurites through establishment of the membrane skeleton during synaptogenesis in vitro (Kitamura et al., 2007). Furthermore, MAP1B-P has been used to demonstrate a correlation between functional recovery and enhanced plasticity after chondroitinase treatment (Galtrey et al., 2007). We thus suggest that remodeling of fibers after spinal cord injury, particularly massive by 2 weeks as revealed by MAP1B-P immunoreactivity, is also likely to reflect local synaptic modifications. This was also suggested previously to explain the significant electrophysiological alterations occurring after a chronic compressive injury (Nashmi & Fehlings, 2001). Accordingly, a co-localization of MAP1B-P with synaptophysin (a presynaptic marker) was observed within gray matter. The decrease in MAP1B-P immunoreactivity between 15 days and 1 month post-injury may indicate that the relatively high injury-related axonal plasticity is in fact limited in time. Bareyre et al. (2004) showed that transected CST is able to sprout and establish new contacts with short and long propriospinal neurons. Ultimately, however, only contacts with long propriospinal neurons are maintained that bridge cervical to lumbar segments, a process that can be compared with the refinement of axonal connections.

MAP1B-P staining and degenerative events

Our study reveals an association of increased MAP1B-P levels not only with axonal extension, but also with axonal degeneration in white matter, where highly MAP1B-P-positive terminal clubs are observed. These degenerative processes, occurring at progressive distances from the lesion, have been described before (Sayer et al., 2002, and references therein). Thus, spinal cord hemisection causes Wallerian degeneration of distal axon shafts, and injured axonal endings eventually retract (Hill et al., 2001; Houle & Jin, 2001). Loss of MAP1B-P-positive fibers in white matter then reflects the loss of ascending and descending axons (rostral and caudal to the lesion, respectively), a phenomenon seen to increase with time post-injury. A failure of retrograde transport might contribute to accumulation of axoplasm at the axon stump and, hence, terminal club swelling (Sayer et al., 2002). This is associated with important changes in the distribution of cytoskeletal proteins, as previously reported for phosphorylated neurofilaments (Petzold, 2005; our observations), or microtubules (for a review, see Moss & Lane, 2006).

Furthermore, we noted increased phosphorylation of MAP1B in somata of ‘suffering’ neurons that is clearly associated with injury-related apoptosis (i.e. Clarke's nucleus neurons; Himes et al., 1994), as revealed by double-labeling for activated caspase-3. As somata accumulating MAP1B-P were not detected by the TUNEL technique, we suggest that this phenomenon is indicative of an early state of degeneration, and perhaps a failure of anterograde transport. We were able to reproduce MAP1B-P accumulation in adult DRG neurons in vitro by treating them with staurosporin. This resulted in a shift from the normally proximo-distal distribution of MAP1B-P within their neurites towards an abnormal accumulation in cell bodies, accompanied by swelling and fragmentation of neurites. Abnormal MAP1B phosphorylation has also been reported for other neuronal pathologies, associated with aberrant phosphorylation of tau, such as at sites of neurofibrillary degeneration in Alzheimer's disease brains (Hasegawa et al., 1990; Ulloa et al., 1994; Good et al., 2004).

The two faces of JNK

MAP1B can be phosphorylated by two modes: while mode II phosphorylation (casein kinase II-dependent) is observed throughout developmental stages until adulthood, the developmentally regulated mode I is based on the action of proline-dependent kinases, such as GSK-3beta, cdk-5 and mitogen-activated protein kinases (MAPK; Pigino et al., 1997; Garcia-Perez et al., 1998; Lucas et al., 1998; Good et al., 2004; Kawauchi et al., 2005). Assuming that the 1B-P antibody recognizes MAP1B phosphorylated via mode I (see Discussion in Nothias et al., 1996; Soares et al., 1998; Ma et al., 2000), we further investigated a potential correlation between expression of MAP1B-P and activated c-Jun N-terminal kinases (JNKs). JNKs, of which three isoforms are known (JNK1–3), are part of the MAPK family. Various proteins in different cellular compartments are targets of JNKs, among which are microtubule-associated proteins (Chang et al., 2003; reviewed by Herdegen & Waetzig, 2001). There is increasing evidence that JNKs are potent effectors of apoptosis in the nervous system following a variety of injuries (Liu & Lin, 2005; Yin et al., 2005). On the other hand, JNKs are also involved in neuronal survival (for a review, see Liu & Lin, 2005), migration (Kawauchi et al., 2003) and axon growth (Chang et al., 2003; Xiao et al., 2006). Specific inhibition of JNK activity was shown to affect axonal regeneration from adult DRG neurons in vitro (Lindwall et al., 2004). Moreover, Chang et al. (2003) demonstrated that JNK-1-deficient mice exhibit progressive degeneration of neurites, associated with shorter microtubules and a reduced MAP1B (and MAP2) phosphorylation. The results presented here suggest that JNK signaling following spinal cord injury converges on phosphorylation of MAP1B. Thus, MAP1B-P and activated JNK co-localize in axons that undergo either sprouting or degeneration, and accumulate together in the somata of preapoptotic neurons. Furthermore, specific inhibition of JNK activity in vitro leads to a dramatic decrease in phosphorylation of the MAP1B epitope recognized by the 1B-P antibody. However, further studies are needed to elucidate the precise mechanism by which the JNK-regulated MAP1B phosphorylation state influences microtubule stability, leading to either axonal plasticity or degeneration.

In conclusion, our study shows that changes in MAP1B-P distribution after spinal cord lesion are well suited to monitoring the overall extent of the two forms of axonal remodeling, i.e. sprouting and degeneration. Furthermore, we present clues to the molecular mechanisms through which MAP1B is phosphorylated and, thus, to signaling pathways involved in structural modifications within the adult nervous system in response to injury.