|Home | About | Journals | Submit | Contact Us | Français|
Dorsal root (DR) axons regenerate in the PNS but turn around or stop at the dorsal root entry zone (DREZ), the entrance into the CNS. Earlier studies that relied on conventional tracing techniques or postmortem analyses attributed the regeneration failure to growth inhibitors and lack of intrinsic growth potential. Here, we report the first in vivo imaging study of DR regeneration. Fluorescently labeled, large-diameter DR axons in thy1-YFPH mice elongated through a DR crush site, but not a transection site, and grew along the root at > 1.5 mm per day with little variability. Surprisingly, they rarely turned around at the DREZ upon encountering astrocytes, but penetrated deeper into the CNS territory, where they rapidly stalled and then remained completely immobile or stable, even after conditioning lesions that enhanced growth along the root. Stalled axon tips and adjacent shafts were intensely immunolabeled with synapse markers. Ultrastructural analysis targeted to the DREZ enriched with recently arrived axons additionally revealed abundant axonal profiles exhibiting presynaptic features such as synaptic vesicles aggregated at active zones, but not postsynaptic features. These data suggest that axons are neither repelled nor continuously inhibited at the DREZ by growth inhibitory molecules but are rapidly stabilized as they invade the CNS territory of the DREZ, forming presynaptic terminal endings on non-neuronal cells. Our work introduces a new experimental paradigm to the investigation of DR regeneration and may help to induce significant regeneration after spinal root injuries.
Over a century ago, Ramon y Cajal labeled a subset of dorsal root ganglion (DRG) axons with Golgi staining and showed that regenerating dorsal root (DR) axons were redirected peripherally or terminated at the DREZ, the transitional zone between the CNS and PNS (Ramón y Cajal, 1928). This regeneration failure remains an important practical issue because common spinal root injuries, such as brachial plexus or cauda equina injuries, lack effective therapy (Hannila and Filbin, 2008; Havton and Carlstedt, 2009). The molecular and cellular events that repel or arrest axons at the DREZ remain poorly understood, but this regeneration failure is generally attributed to glia-associated growth-inhibitory molecules and lack of intrinsic growth potential of DRG neurons (Ramer et al., 2001a).
Although spinal root injury evokes changes similar to those induced by direct CNS injury, it does not cause an impassable glial scar. Nevertheless, the axotomized DREZ prevents regeneration efficiently: Whereas peripheral conditioning lesions, which enhance the growth potential of DRG neurons, promote intraspinal regeneration of their central axons in the dorsal columns (Neumann and Woolf, 1999; Cao et al., 2006), the same axons fail to regenerate through the DREZ (Chong et al., 1999; Golding et al., 1999; Zhang et al., 2007). Notably, repellent cues, including oligodendrocyte-associated inhibitors (Nogo, MAG, OMgp), and astrocyte-associated chondroitin sulfate proteoglycans (CSPGs), cause only brief growth cone collapse or retraction (Snow et al., 1991; Li et al., 1996; Yiu and He, 2006). Moreover, DRG axons grow despite growth cone collapse (Marsh and Letourneau, 1984; Jones et al., 2006; Jin et al., 2009). These growth inhibitory molecules therefore seem to account for the turning but not the arrest of DR axons at the DREZ. These considerations led us to suspect that a novel mechanism plays a more decisive role in preventing regeneration across the DREZ.
Previous studies relied heavily on conventional tracing techniques and postmortem analyses. The spatial and temporal resolutions of these studies were limited: One could only deduce dynamic events associated with DR regeneration by comparing static images. Fundamental questions therefore remained unanswered, including whether axons stop abruptly or attempt to turn around at the DREZ, or whether they remain immobile or regain mobility over time. Axon turning would suggest that the DREZ functions as a passive barrier that axons avoid. Brief collapse without permanent immobilization would highlight the importance of repulsive growth inhibitors that continuously collapse axons at the DREZ. On the other hand, rapid, permanent immobilization would suggest a unique mechanism that arrests axons by paralyzing or stabilizing them, as was speculated many years ago but virtually forgotten (Carlstedt, 1985; Liuzzi and Lasek, 1987).
To address these questions, we applied in vivo imaging techniques using fluorescent transgenic mice with a vital Golgi-like stain, and monitored the regeneration of identified DR axons repeatedly over weeks to months directly in living animals with and without conditioning lesions. Combined with a targeted ultrastructural analysis, we found that most axons were not repelled, but were immobilized rapidly and chronically at the DREZ, forming presynaptic terminal endings in its CNS territory.
We used adult mice (2–4 months of age) of either sex from transgenic strains thy1-YFP-H and thy1-YFP16, which express yellow fluorescent protein (YFP) under the control of the neuron-specific Thy-1 promoter (Feng et al., 2000). The original breeding pairs were purchased from The Jackson Laboratory (Bar Harbor, ME); subsequent stocks of mice used in these experiments were reared in the animal facilities at Drexel University College of Medicine. All experiments were performed in accordance with DUCOM’s Institutional Animal Care and Use Committee and National Institutes of Health guidelines.
Thy1-YFPH mice were anesthetized with an intraperitoneal injection of xylazine (8 mg/kg) and ketamine (120 mg/kg). Supplements were given during the procedure as needed. A 2- to 3-cm long incision was made in the skin of the back; the spinal musculature was reflected; and the L3-S1 spinal cord segments were exposed by hemi-laminectomies. The cavity made by the laminectomies was perfused with warm sterile Ringer’s solution or artificial cerebrospinal fluid. A small incision was made in the dura overlying the L5 dorsal root near the L3 DRG; a fine forceps (Dumont #5) was introduced subdurally and the L5 dorsal root was crushed for 10 seconds. After images were collected (see below), we attempted to minimize scar formation by tightly applying a piece of thin synthetic matrix membrane (Biobrane, Bertek Pharmaceuticals, Morgantown, WV) over the exposed cord and dura, so that scarring accumulated on the membrane rather than on the dura surface. The matrix membrane was removed and replaced at each imaging session. This membrane was stabilized with a layer of much thicker artificial dura (Gore Preclude MVP Dura Substitute, W.L. Gore and Associates, Flagstaff, AZ) that covered the laminectomy site. The musculature was then closed with sterile 5-0 sutures, and the skin with wound clips. Animals were given subcutaneous injections of lactated Ringer's solution to prevent dehydration and kept on a heating pad until fully recovered from anesthesia. Buprenorphine was given as postoperative analgesia (0.05 mg/kg subcutaneously every 12 h for 2 days). For each imaging session, we reanesthetized and surgically re-exposed the area of interest and repeated the procedures. For conditioning lesions, the sciatic nerve was crushed in the lateral thigh of the ipsilateral hind leg 10 days before the root was crushed. Animals were anesthetized as described above; the skin and superficial muscle layer of the mid thigh were opened; and the sciatic nerve was crushed for 10 s with fine forceps (Dumont #5). The muscle and skin were then closed in layers and the animals were allowed to recover on a heating pad until fully awake.
We used a Leica MZ16 fluorescent stereomicroscope or an Olympus BX51 microscope equipped with a fast shutter and a highly sensitive cooled CCD camera (ORCA-Rx2, Hamamatsu, Bridgewater, NJ) controlled by Metamorph software (Molecular Devices, Sunnyvale, CA). Body temperature was maintained by placing the animal on a thermostatically controlled heating pad. Warmed lactated Ringer’s solution was used to superfuse the exposed portion of spinal cord. Images were acquired either as single snapshots or as multiple streams of 10 to 20 frames acquired within 30- to 40-ms exposure time. In-focus images were then selected, and an overview montage was created using Photoshop (Adobe Systems, San Jose, CA). High-resolution confocal images were obtained with a Leica TCS 4D confocal microscope (Heidelberg, Germany). Z stacks were obtained at 0.3-µm step size for 20- to 40-µm depths. Leica TCS-NT acquisition software and Imaris image software (Bitplane AG, Zurich, Switzerland) were used to reconstruct z-series images into maximum intensity projections.
Following in vivo imaging, we harvested tissues and processed them in whole mounts to immunolabel astrocytes, oligodendrocytes, or Schwann cells to locate the CNS/PNS interface. The immunostaining procedure was standard (Wright et al., 2009), except for the permeabilization steps in which chilled MeOH and 1% sodium borohydride were also used. Mice were perfused transcardially with 0.9% heparinized saline solution followed by 4% paraformaldehyde in phosphate buffered saline (PBS). After 3 hours in situ postfixation at 4°C, the spinal cord segment (L3–L6) with attached dorsal roots was removed and rinsed in PBS. The tissue was then washed for 30 minutes in a blocking solution containing 0.1 M glycine and 2% bovine serum albumin (BSA) in PBS and treated in cold MeOH for 10 minutes and then 1% sodium borohydride for 5 to 10 minutes. After thorough and extensive rinsing in PBS, the spinal cord was further permeabilized with 0.2% Triton X-100 with 2% BSA in PBS (TBP) for 1 hour and then incubated with primary antibody diluted in TBP overnight. The next day the spinal cord was rinsed thoroughly in TBP and then incubated with appropriate fluorescently conjugated secondary antibodies diluted in the TBP for 1 hour at room temperature. The tissue was then rinsed in PBS, and a thin sheet of dorsal spinal cord was prepared from the DREZ and rootlet, mounted in Vectashield (Vector Laboratories, Burlingame, CA), and stored at 4°C.
To immunolabel axons at the axotomized DREZ with synaptic vesicle markers, we used the transgenic strain, thy1-YFP16, in which the entire population of large-diameter axons expresses YFP (data not shown). To analyze more axons than superficially located ones, we prepared cryostat sections, rather than whole mounts, of the DREZ after crushing dorsal roots of cervical spinal cord. Using the surgical procedures described earlier, C3–C5 roots were crushed, and the animals were allowed to recover. At 20 days post injury, the C3–C5 spinal cord and roots were harvested, postfixed overnight at 4°C, cryoprotected in 30% sucrose in PBS, and rapidly frozen in Shandon M1 embedding matrix (Thermo Electron Corporation, Pittsburgh, PA). Serial transverse sections were cut on a cryostat at 10 µm (CM3000, Leica) and collected on Superfrost Plus slides (Fisher Scientific, Pittsburgh, PA). For immunostaining, sections were postfixed in 4% paraformaldehyde in PBS for 20 min, rinsed in PBS, and blocked for 1 hour in TBP. The sections were then incubated overnight at 4°C in a cocktail of primary antibodies diluted in TBP. Sections were then rinsed in PBS and incubated with secondary antibodies in TBP for 1hour at room temperature and processed as described above.
L5 DRGs were dissected from unoperated thy1-YFPH mice and processed to obtain serial cryostat sections using the methods described above. Selected sections were stained with a fluorescent Nissl stain (Neurotrace 530/615 red fluorescent Nissl stain; Invitrogen) according to the manufacturer's instructions, washed extensively with PBS, and coverslipped using Vectashield (Vector Laboratories, Burlingame, CA). Neuron counts were made on 5 L5 DRGs from 3 animals. For each DRG at least three randomly selected sections were analyzed, taking care not to use sections that were poorly mounted or stained. Each section selected was at least 30 µm away (three sections) from either of the other selected sections. Using a Retiga EXi (Qimaging) digital camera, the entire section was photographed in segments using the 20x objective on a Leica DMRBE fluorescent microscope. The same section was photographed using both red (Nissl stained cells) and green (YFP+ cells) fluorescent filter cubes to identify neurons containing a nucleus with a visible nucleolus and to determine whether such neurons were YFP-positive. Image segments were collected at 200 magnification and combined to form a montage. Using ImageJ software (National Institutes of Health, Maryland), the cell area of all neurons containing a nucleus with a visible nucleolus from each chosen section was measured. We counted a minimum of 200 neurons per ganglion. If this number was not reached in the three sections chosen, a fourth section was counted, also in its entirety. However, because identification and counting continued even after the minimum of 200 neurons were obtained, we always counted more than 200 Nissl-stained neurons per DRG (mean 218±6 Nissl cells measured/DRG, 4.3% of the measured were YFP+). Histograms representing the cross-sectional area of all Nissl- and YFP-labeled DRG neurons measured were compiled in order to compare the distribution of the YFP-labeled cells with the total cell populations.
The primary, cell-type specific antibodies included anti glial fibrillary acidic protein (GFAP, mouse monoclonal, 1:1000, Chemicon, Millipore, Billerica, MA) to label astrocytes, anti-myelin oligodendrocyte glycoprotein (MOG, goat polyclonal, 1:200, R&D Systems, Minneapolis, MN) for labeling oligodendrocytes and anti-SC/2E (mouse monoclonal, 1:1000, Cosmo Bio USA, Carlsbad, CA) or laminin-1 (rat monoclonal, 1:200, Abcam, Cambridge, MA) to label Schwann cells. Mouse monoclonal antibodies to a synaptic vesicle protein, SV2 (1:10, Developmental Studies Hybridoma Bank, Iowa City, Iowa), or to synaptotagmin 2 (znp-1, 1:2000, Zebrafish International Resource Center, Eugene, OR) were used to label synaptic vesicles. To learn more about the phenotype of YFP+ DRG neurons, selected sections from L5 DRGs were labeled with one or more of the following methods: Neurons containing phosphorylated epitopes of high-molecular-weight neurofilament were identified using the SMI 312 antibody (mouse monoclonal antibody, 1:1000 dilution, Covance Inc., Princeton, NJ). The population of small primary afferent neurons that expresses the trkA neurotrophin receptor was labeled using an antibody to calcitonin gene-related polypeptide (CGRP, rabbit polyclonal antibody to rat CGRP, 1:2000, Bachem Americas, Torrance, CA). The population of small DRG neurons that does not express the trkA neurotrophin receptor was labeled using Griffonia simplicifolia IB4 lectin (biotin conjugate, 5µg/ml, Sigma-Aldrich, St. Louis, MO). Secondary antibodies used were Alexa 647-conjugated donkey anti-mouse 1:200, Invitrogen, Eugene, OR), Alexa-Fluor 568-conjugated goat anti-mouse IgG1 (1:200, Invitrogen, Eugene, OR), Alexa-Fluor 647-conjugated donkey anti-rabbit IgG (1:200, Invitrogen, Carlsbad, CA) and rhodamine-red conjugated rabbit anti-goat IgG (Jackson ImmunoResearch Laboratories, West Grove, PA).
The mice were perfused transcardially (with heparinized Tyrode’s solution followed by 2% paraformaldehyde and 2.5% glutaraldehyde in 0.1 M Na-cacodylate buffer. The spinal cord segments L3–L6 were then removed as one piece and rinsed in 0.1M Na-cacodylate buffer, mounted on an agarose support, and placed in the vibratome well containing chilled buffer. The most superficial longitudinal slice containing the DREZ (0< 250µm thickness) was cut and further processed for electron microscopy. To target our electron microscopic analysis to the area where axons had stalled, we applied fiducial markers to the surface of the spinal cord slice. The spinal cord sections were flattened with insect pins in Sylgard silicone elastomer-lined 35-mm petri dishes. A 1.0% solution of 1, 1'-dioctadecyl-3,3,3',3'-tetramethylindodicarbocyanine-5,5'-disulfonic acid (DiI, Invitrogen, Carlsbad, CA) was dissolved in dichloromethylene and loaded into a micropipette (resistance of 5–10 MΩ). Crystals of DiI were iontophoretically applied to the surface of the spinal cord slice in an area of the DREZ with bulb-tipped axons (e.g., see Fig. 11A). To render the DiI crystal electron dense, we excited the DiI crystals near their excitation wavelength in the presence of 3, 3’-diaminobenzidine (DAB, 5.0 mg/mL, Sigma-Aldrich) until the DiI crystal was replaced with a dark red/brown DAB precipitate (~20 minutes). After photoconversion, spinal cord slices were trimmed to contain the area of interest using the electron-dense fiducial markers as reference points. Tissue blocks were stained with 1.0% osmium tetroxide reduced in 1.5% potassium ferrocyanide for 45 minutes, then dehydrated in an ascending ethanol series, infiltrated with Araldite 502 Embed 812 resin, and polymerized at 60°C for 48 hours. Polymerized tissue blocks were sectioned (0.5 µm) with a glass knife on a Leica Ultracut R microtome (Leica, Wetzlar, Germany) until the fiducial markers were located. Serial ultrathin sections (60–70 nm) were cut and mounted on pioloform-coated slot grids. Sections were counterstained with 2.0% aqueous uranyl acetate and Reynold’s lead citrate. Sections were viewed at 75 kV on a Hitachi H-600 transmission electron microscope. Serial electron micrographs were captured at 6000× and scanned at a resolution of 1000 dpi.
To monitor DR regeneration directly in living mice, we used the H line of thy1-YFP mice (thy1-YFPH), which expresses high levels of yellow fluorescent protein (YFP) in subsets of neurons, including sensory neurons in dorsal root ganglia (DRG; Feng et al., 2000). Because DRG neurons are heterogeneous, we first characterized YFP-labeled neurons in DRG of thy1-YFPH mice. YFP positive (+) neurons were large, with an average size of 1244.7 µm2 (Fig. 1A; n = 66, compared to the mean size of 594 µm2 of 1523 Nissl-stained neurons). As expected for large DRG neurons (Ruscheweyh et al., 2007), YFP+ neurons were neurofilament-positive (+) but CGRP- and IB4- negative (−) (Fig. 1C). Additional analysis of superficial layers of the dorsal horn in thy1-YFPH mice revealed little YFP fluorescence in lamina II, where CGRP+ and IB4+ innervation is abundant (Fig. 1B). We also observed that CGRP+ axons were not labeled in the line 16 thy1-YFP (thy1-YFP16) mice, which were thought to express YFPs in nearly all neurons (data not shown, cf., Fig. 10). These results are consistent with an earlier characterization of the M line thy1-YFP mice (thy1-YFPM), in which fewer neurons are labeled than in the H line (Kerschensteiner et al., 2005). Large, neurofilament+ neurons are myelinated and their central axon processes are thought to regenerate more poorly than those of small-diameter, nonmyelinated neurons (Tessler et al., 1988; Guseva and Chelyshev, 2006). Thus, thy1-YFPH mice provide a unique opportunity to study in vivo the axon regeneration of sensory neurons whose regeneration potential may be particularly weak.
While optimizing the in vivo imaging techniques for DR regeneration, we learned that regeneration of individual DR axons is better imaged in lumbar than in cervical spinal cord of thy1-YFPH mice (Skuba et al., unpublished observations). In our typical preparation, we performed a right-sided laminectomy (Fig. 2A; T12-L5) to partially expose the L5 root and the DREZ where these processes enter the spinal cord (Fig. 2B, C, highlighted box in Fig. 2C approximates the DREZ). Superficially positioned YFP+ axons in the medial portion of the L5 root (3–10 axons) run parallel to the midline dorsal vein, curve perpendicularly to enter the DREZ, and then bifurcate within the spinal cord, with one branch that enters the dorsal column (DC) and another that enters the gray matter (Fig. 2C). This stereotypical trajectory helped us to follow regeneration of several identified axons simultaneously and repeatedly over time in vivo. It also permitted us to identify reliably the location of the DREZ in living spinal cord.
We first monitored DR axons after complete transection. The medial portion of the L5 root was cut near the L3 DRG, ~3 mm from the DREZ, using a fine-spring scissors (Fig. 2D2). After lifting the proximal stump to confirm that the DR had been completely transected, we closely apposed the cut ends of the proximal and distal stumps and then examined axons on both sides of the transection 3 and 7 days after injury. Even 7 days after the cut, no YFP+ axons from the proximal stump extended across the transection site (Fig. 2D3, n = 4 mice). The two ends were separated by approximately 500 µm of collagenous scar tissue, which seemed to prevent proximal axons from penetrating this region. After the mice were killed, we used confocal imaging to confirm that YFP+ axons located deeper in the root than those that we imaged in vivo also failed to regenerate across the injury site (Fig. 2D4). Importantly, however, few if any neurites formed bulblike endings or terminated abruptly. Instead, they turned around at the transection site and extended back along the proximal axons (e.g., arrows in Fig. 2D4; n > 95 axons, 3 mice).
Next, we monitored YFP+ axons every 24 to 48 hours for 7 days after a crush injury (Fig. 3). We crushed the medial portion of the L5 root with a fine forceps (Fig. 3A1). The next day we observed dying-back degeneration of proximal stump axons (data not shown, e.g., Fig. 4A2) and fragmentation/degeneration of the same axons distal to the crush (e.g., yellow arrowheads in Fig. 4A3; cf., Kerschensteiner et al., 2005), which confirmed that axons had been appropriately damaged. We applied several additional criteria for unambiguously distinguishing regenerating axons from axons that had been spared or recovered from the injury: These include; 1) Expansion of the non-fluorescent portion of the YFP+ axon at the crush site due to proximal and distal degeneration (in contrast to narrowing of the unlabeled gap due to fluorescent cytoplasm refilling the crush site if axons survived the injury), 2) Regenerating axons were much thinner, less brightly fluorescent and more undulating than axons that survived the injury. 3) Regenerating neurites were thinner and more dimly fluorescent than the degenerating fluorescent fragments of axons through which they extended, 4) In contrast to surviving or spared axons, regenerating axons stopped at the DREZ and 5) did not exhibit nodes of Ranvier.
In contrast to the transection injury, almost all crushed YFP+ axons extended a single neurite that grew across the site of injury by 3 days (Fig. 3A2; n > 25 axons, 5 mice) without forming additional branches (Fig. 3A3, A4). Confocal analysis 7 days after injury confirmed that almost all YFP+ axons located deep in the DR also grew through the crush site (Fig. 3A5).
To exclude the possibility that these observations were due to imaging-associated artifacts, we examined mice 4 days after crush injury that we had not previously imaged in vivo (Fig. 3B). Consistent with in vivo imaging observations, almost all YFP+ axons extended a single neurite through the crush site and grew until reaching the DREZ (Fig. 3B1, n > 85 axons, 2 mice). Notably, by 4 days, almost all of the axons had already arrived at the DREZ, about 3 mm beyond the injury site, and terminated at a similar location (Fig. 3B2, arrows). We observed few axons turning around, further evidence of the consistency of the axon response to crush injury. Thus, most, and perhaps all, YFP+ sensory axons, which have previously been thought to have a relatively weak regenerative capacity, can nevertheless regenerate across a crush site, extend within the PNS, and arrive at the DREZ.
We next investigated in vivo how quickly YFP+ DR axons regenerated along the root (Fig. 4). This process was difficult to accomplish, however, because fragments of degenerating YFP+ axons (e.g., yellow arrowheads in Fig. 4A3) retained fluorescence for 7 to 10 days and obscured the leading tips of regenerating YFP+ axons. To circumvent the problem, we minimized the number of damaged YFP+ axons by crushing only the most medial portion of the L5 root (Fig. 4A1). This strategy enabled us to crush only one or two YFP+ axons and follow their regeneration with little residual YFP fluorescence. The crush was made at the usual location, ~3 mm away from the DREZ, and the crushed axons were imaged daily. Typically, the proximal tips of the crushed axons degenerated for 2 days after injury (Fig. 4A2; e.g., an axon marked by green arrow) and then extended a short neurite, which, by Day 4, had grown about 3 mm along the fluorescent fragments of degenerating axons (i.e., endoneurial tubes marked by yellow arrowheads; Fig. 4A3) and arrived at the DREZ (Fig. 4A3). We also observed short neurites extending from some of the proximal axons imaged 2 days after injury (Fig. 4B; green arrows). Thus, YFP+ axons elongated at a rate greater than 1.5 mm per day along their original endoneurial tube trajectory. These observations are consistent with confocal analysis of mice that we did not image in vivo (cf., Fig. 3B), and demonstrate that large-diameter axons grow well along the dorsal root, with little variability.
To determine the behavior of axons at the PNS/CNS border, we continued imaging axons beyond 4 days after crush, when they arrive at the DREZ. Because more degenerating fluorescent axon fragments (i.e., residual YFP fluorescence) disappeared over the next few days, we resumed imaging 6 or 7 days after injury when the growth tips of regenerating axons were fairly easy to recognize. In several mice with L5 root crush, we were able to identify a number of axon tips arriving at the DREZ on day 6 or 7 (n => 40 axons, 6 mice) and relocate them again in subsequent imaging sessions over 2 weeks after crush (Fig. 5). Unexpectedly, the leading tips of these axons did not continue to grow forward, retract, or turn around but were completely immobile; they remained in the same location in subsequent imaging sessions (i.e., as identified by the relative location with respect to adjacent axons or landmarks such as blood vessels and fluorescent debris). Notably, their appearance also did not change except that swellings formed on the tips or shafts of some axons (Fig. 5; white arrows).
We occasionally observed slowly growing and retracting axons (Fig. 6). Close observation, however, indicated that this growth was short neuritic sprouting extending from (Fig. 6A5), and then being reabsorbed by (Fig. 6A6) axon tips that remained stationary and developed swollen endings over time (Fig. 6A4–6, pink arrows). On the last day of imaging, we sacrificed the mouse, analyzed the same axon endings with high-resolution confocal microscopy, and confirmed that no growth tip structures had been missed due to poor resolution of our live imaging set-up (data not shown, cf. Fig. 8; day 20). The apparent mobility of these axons at the DREZ was therefore due to fruitless sprouting of immobilized axon tips. Collectively, considering that axons arrive at the DREZ as early as 4 days after crush injury (cf. Fig. 4), these observations demonstrate that axons are immobilized surprisingly quickly after arriving at the DREZ. It is also notable that, in contrast to axons at the site of a transection injury (Fig. 2D), axons arriving at the DREZ after crush injury (Fig. 4) rarely turned around along the dorsal root even though this pathway contains growth-promoting Schwann cells.
We also observed the long-term response of DR axons stopped at the DREZ in mice whose L5 root had been crushed 4 months previously (Fig. 7, n = 15 axons, 2 mice). When we initiated imaging at 128 days after the crush, we observed several superficially positioned YFP+ axons and their tips at the DREZ (Fig. 7A, colored arrows). In subsequent imaging sessions at days 131 and 140 after crush, these axons were not motile but remained in the same place and looked unchanged, demonstrating chronic immobilization or stability. We also analyzed a mouse whose L5 root had been crushed 4 months previously but had not been studied with in vivo imaging (Fig. 7B). Confocal analysis revealed axon profiles at the DREZ that appeared similar to those observed within the first week after crush (cf. Fig. 3B2): almost none of the YFP+ axons turned around but terminated as single processes at a similar location at the DREZ. Axon swellings, similar in appearance to synaptic varicosities, were also frequently observed on axon shafts (Fig. 7B, yellow arrowheads). These observations show that axons quickly immobilized on entering the DREZ remained completely immobile and stable for long periods despite the absence of target innervation.
Next, we asked if axons became motile at the DREZ following a conditioning lesion of peripheral processes that enhances the growth potential of DRG neurons (Chong et al., 1999). To this end, we crushed sciatic nerves in the ipsilateral leg 10 days before crushing the L5 root at the usual location (Fig. 8, n = 6 mice). We then imaged these mice in vivo every 2 or 3 days over 3 weeks, and monitored regeneration of identified axons at the crush site, along the root, and at the DREZ. We found that most YFP+ axons had already extended through the crush site 2 days after crush (Fig. 8; day 2, 95%, n = 15), a day faster than nonconditioned axons. In addition, axons frequently extended more than one neurite, further illustrating the enhanced growth within the root due to a conditioning lesion. However, we observed no axons that had regenerated through the DREZ at 4 days after crush (Fig. 8; day 4, n > 85 axon tips located at the DREZ). Instead, like nonconditioned axons at 4 days after crush (cf. Fig. 3B, 4), these conditioned axons did not turn around upon arriving at the DREZ but terminated as single processes at a similar location at the DREZ. In subsequent imaging sessions over the next 3 weeks, they did not grow forward or retract but remained immobile (Fig. 8). The only noticeable change was the swelling of the tips and shafts of some axons (Fig. 8, yellow arrowheads). Thus, even conditioned axons with enhanced peripheral growth were quickly immobilized or stabilized at the DREZ.
At the DREZ, both astrocytes and oligodendrocytes are juxtaposed to PNS Schwann cells. Following DR injury, astrocytes proliferate and hypertrophy, occupy the DREZ, and are the first cells encountered by axons regenerating from the periphery (Bignami et al., 1984; Fraher et al., 2002). Stalled axons at the DREZ were observed to contact astrocytes (Carlstedt, 1985; Fraher, 2000; Dockery et al., 2002), and reactive astrocytes were thought to form a primary regenerative barrier at the DREZ. We were intrigued by the unexpectedly rapid and long-lasting immobilization or stabilization of axons arriving at the DREZ and wondered if axons become immobilized as they encounter astrocytes. To this end, we developed a unique immunostaining protocol that permitted antibodies to penetrate deep into the surface of spinal cords prepared in whole mount (see Methods). This method allowed us not only to identify axons that we imaged in vivo but also to label simultaneously oligodendrocytes, astrocytes, and Schwann cells that were near or directly associated with axons at DREZ (Fig. 9A,B).
We crushed L5 roots (n = 2 mice), monitored them for 2 weeks in vivo, and confirmed that most axons in these mice terminated in the usual location at the DREZ 4 days after crush, and then remained immobile (data not shown). Subsequently, oligodendrocytes (Fig. 9B) and astrocytes (data not shown) were immunolabeled in whole mounts and their relationships with the axons imaged in vivo were analyzed with high-resolution confocal microscopy. As expected (Fraher, 1999; Fraher et al., 2002), astrocytic processes extended further into the periphery than the oligodendrocytes, even in the intact, noninjured DREZ (Fig. 9A–A”) and invaded the PNS even more extensively after injury (data not shown, cf. Fig. 10A”, B”). We found that axons did not stop when they encountered astrocytic processes at the astrocyte-PNS interface but extended along them and terminated deeper in the CNS territory containing degenerating oligodendrocytes (Fig. 9B”, > 95%, n = 46 axon, 2 mice).
The rapid immobilization and subsequent swellings often formed on axon shafts and tips resembled the synaptogenic process and prompted us to test whether DR axons form synapses as they enter the CNS territory of the DREZ. We first immunolabeled cryostat sections of the DREZ where many stalled axons were observed (Fig. 10B–B’’’’), with synapse markers such as SV2 or synaptotagmin, together with a GFAP antibody to mark CNS territory. Whereas no SV2 or synaptotagmin immunoreactivity existed at the DREZ of uninjured mice (Fig. 10A’), numerous, intense immunopositive profiles were observed in the CNS territory of the DREZ of injured mice (Fig. 10B”, asterisk), and they colocalized with tips or adjacent shafts of stalled YFP+ axons (Fig. 10B’’’, inset).
We next performed an ultrastructural analysis of axons that had stopped at the DREZ (Fig. 11). To target our analysis to the CNS territory of the DREZ where axons had stopped (Fig. 11B; yellow arrowheads), we placed DiI crystals (Fig. 11A; white arrows) at the completion of in vivo imaging at 13 days after the L5 root crush. DiI crystals were then photoconverted and used as landmarks to relocate the area with electron microscopy. This strategy allowed us to observe abundant axonal profiles at the DREZ, embedded in non-neuronal profiles such as Schwann cells and astrocytes (Fig. 11C). These axonal profiles were filled with mitochondria and ~40-nm vesicles but lacked the vacuoles and disorganized microtubules that are typical of dystrophic endings (Fig. 11D, E; Erturk et al., 2007). Moreover, vesicles and mitochondria were differentially distributed within the nerve-terminal-like profiles so that vesicles were highly clustered onto one side of electron-dense membranes that resembled active zones, and mitochondria were distributed toward the other side (Fig. 11D, E). Although features of presynaptic differentiation were apparent in these axonal profiles, no indications of differentiation such as postsynaptic densities were observed postsynaptically (Fig. 11D1, E1; red-pseudocolored), excluding the possibility that postsynaptic cells are adjacent axons. Together, these observations indicate that axons became immobilized at the DREZ by forming presynaptic endings on non-neuronal cellular elements as they entered the CNS territory of the DREZ.
The generally accepted explanation for the regeneration failure at the DREZ is the combination of growth inhibitory molecules and limited intrinsic growth potential. However, growth inhibitors that transiently collapse growth cones cannot account for the unexpectedly rapid and long-lasting immobilization of YFP+ axons that we observed in vivo even after a conditioning lesion. Furthermore, they rarely turned around at the DREZ and exhibited obvious features of presynaptic terminal endings. We propose that regeneration fails at the DREZ at least in part because axons are rapidly and chronically stabilized by induction of presynaptic differentiation (Fig. 12).
The present study is the first to apply in vivo imaging to monitor DR regeneration directly in living animals. Because we were concerned about the potential confounding effects of artifacts due to phototoxicity and multiple invasive surgical/anesthetic procedures, we carried out extensive control experiments with mice that did not receive repetitive imaging or surgery. Several observations convinced us of the reliability of our techniques. First, our estimated axon elongation rate (> 1.5 mm/day) was not slower, rather it was slightly faster, than previous estimates based on conventional methods (Ramer et al., 2001a; 1mm/day). Second, consistent with an earlier study (Ylera et al., 2009), we observed that conditioning lesions caused the injured dorsal roots to begin to grow earlier and to extend more neurites than after injury alone. Third, we observed little variability among YFP+ axons in initiation of growth, elongation along the root, or behavior at the DREZ.
Our observation that YFP+ axons fail to regenerate across the DREZ following a conditioning lesion is consistent with results from previous studies in rats, in which only minor effects were reported (Chong et al., 1999; Golding et al., 1999; Zhang et al., 2007). It conflicts however with a recent report of significant regeneration in mice, which the authors attributed to species differences (Quaglia et al., 2008). It is possible that the capacity for growth varies greatly among different populations of DRG neurons in mice and that the latter study emphasized a different subpopulation, possibly small-diameter axons which were not examined in the present in vivo imaging studies. Incomplete lesions or labeling artifacts may also be responsible (cf., Steward et al., 2007). A clear advantage of in vivo imaging was that it permitted us to evaluate immediately the extent of damage and to ascertain whether the lesion was complete or incomplete.
Our observation that a conditioning lesion did not promote growth into the spinal cord agrees with the notion that the failure is likely due to growth inhibitory molecules (Golding et al., 1999; Ramer et al., 2001b), such as CSPGs synthesized by astrocytes (Rhodes and Fawcett, 2004; Silver and Miller, 2004), Nogo, MAG, and OMgp present in myelin debris (Oertle and Schwab, 2003; Yiu and He, 2006), and semaphorins expressed by fibroblasts (Fawcett, 2006). However, these inhibitors act as repulsive cues that cause brief growth cone or filopodial collapse and allow axons to turn and grow away without a significant pause or long-term immobilization (Raper and Kapfhammer, 1990; Snow et al., 1990; Drescher et al., 1995; Li et al., 1996). Moreover, DRG axons extend despite growth cone collapse (Marsh and Letourneau, 1984; Jones et al., 2006; Jin et al., 2009). Some growing axons appeared to be trapped, forming dystrophic endings when exposed to a gradient of proteoglycans, but these axon endings remained extremely motile (Tom et al., 2004). Lastly, axons entering the DREZ in vivo are accompanied by Schwann cells, which would provide an alternative growth pathway (cf., Adcock et al., 2004) or constrain migration of axons into CNS territory (Grimpe et al., 2005). We therefore anticipated that axons entering the DREZ would be mobile and dynamic, frequently turning back to the PNS.
However, we found that YFP+ DR axons rarely turned around at the DREZ but terminated abruptly and remained completely immobile. In our injury paradigm, axons arrived at the DREZ 4 days after root crush. We often resumed imaging 6 or 7 days after crush due to residual fluorescence that obscured leading axon tips. It is unlikely however that we failed to observe local motility of axons that might be particularly substantial soon after arrival (i.e., 4–7 days after crush). Conditioning lesions led to earlier clearing of residual fluorescence (Skuba et al., unpublished observation), which permitted us to monitor axon tips 4 to 7 days after injury. These axon tips remained immobile at the same location during this period, without turning, retracting, or growing forward (Fig. 8). The large-diameter DR axons that we monitored in vivo are, therefore, unexpectedly and quickly immobilized as they enter the DREZ. Consistent with our in vivo results, neurites cultured on cryosections of the DREZ rarely turned around and did not collapse or retract, but became permanently immobile within 20 minutes (Golding et al., 1999).
In vivo imaging enabled us to target our ultrastructural analysis specifically to the region of the DREZ where recently arrived axon tips were abundant. We were able to identify numerous axonal profiles within a few sections. They often looked ‘synapse-like’ (data not shown), exhibiting abundant mitochondria and membranous vesicles, and resembled the structures called ‘synaptoids’ (Carlstedt, 1985; Liuzzi and Lasek, 1987). However, these previously described ‘synapse-like’ profiles lacked the characteristic features of synaptic differentiation. It is also known that mitochondria and vesicles are abundant even in nonsynaptic, dystrophic endings (Erturk et al., 2007).
By contrast, the ‘synaptic’ profiles that we encountered exhibited obvious features of presynaptic differentiation, including active zones, differential distribution of mitochondria and synaptic vesicles, and absence of microtubules and vacuoles. Moreover, they were intensely immunolabeled by synapse markers. These synaptic profiles were relatively small and did not seem to belong to the large, club-shaped bulbous endings that developed in some axons at the DREZ (e.g., Fig. 8; yellow arrowheads). They more likely represent those axon endings that did not increase in size or varicosities that we often observed on axon shafts near the tips (e.g., Fig. 9B’).
Earlier ultrastructural studies were performed relatively long after dorsal root injury: 6 to 9 months (Carlstedt, 1985) and 3 weeks to 3 months (Liuzzi and Lasek, 1987). These studies therefore could not determine whether axon endings became ‘synapse-like’ as they entered the DREZ or whether they occasionally exhibited a synapse-like appearance after chronic remodeling. Thus, our studies provide novel and compelling evidence that axons become differentiated into presynaptic terminal endings immediately after entering the DREZ.
Synaptoids were speculated to be triggered by a physiological stop signal associated with reactive astrocytes (Liuzzi and Lasek, 1987). We also observed that portions of the axons exhibiting presynaptic profiles were in contact with filament-rich astrocytes (cf., Figure 11D and E). Importantly, however, we did not observe abundant intermediate filaments in the postsynaptic cells that were in direct apposition to presynaptic profiles. This observation raises the possibility that non-astrocytic cell types, such as NG2+ cells, may provide the stabilizing activity or induce the presynaptic differentiation of regenerating axons at the DREZ. In support of this notion, 1) The postsynaptic cells did not exhibit postsynaptic densities, excluding that these profiles are axon-axon synapses (Bernstein and Bernstein, 1971); 2) Axons did not stop when they encountered astrocytes at the CNS-PNS border but terminated deeper in the CNS territory; and 3) Presynaptic active zones were thinner than at neuron-neuron synapses and resembled those reported at neuron-NG2+ cell synapses (Bergles et al., 2000; Lin and Bergles, 2004). Importantly, NG2+ cells are present at the DREZ (Zhang et al., 2001; Beggah et al., 2005) and stabilize dorsal column axons that macrophages cause to retract at the lesion site (Busch et al., 2010).
The role of NG2+ cells is speculative, however and it is possible that reactive astrocytes are involved in the process. Recent studies revealed that astrocytes are capable of inducing and promoting synapse formation, presumably by releasing thrombospondins (Ullian et al., 2004; Wang and Bordey, 2008; Allen and Barres, 2009; Eroglu et al., 2009). An intriguing possibility, which we are currently testing, is that astrocytes prevent regeneration at the DREZ by mediating presynaptic differentiation between axons and NG2+ cells. Heparin sulfate proteoglycans may also be involved, since they trigger presynaptic assembly in the absence of specific target recognition (Lucido et al., 2009).
Efforts to overcome regeneration failure at the DREZ have included enhancing the regeneration capacity of sensory axons with neurotrophic factors and neutralizing growth inhibitors (Ramer et al., 2002; Steinmetz et al., 2005; Cafferty et al., 2007; Wang et al., 2008; Andrews et al., 2009; Harvey et al., 2009; Cafferty et al., 2010; Harvey et al., 2010; Ma et al., 2010). The effectiveness of these efforts may have been limited because they did not treat the stabilizing activity at the DREZ. Conversely, even if axons are prevented from being stabilized at the DREZ, growth inhibitors might nevertheless inhibit regeneration into and across the DREZ. It is therefore reasonable to expect the best outcome from combinatorial strategies targeting both stabilizing and inhibitory activities. It is worth however pointing out that axons can regenerate along degenerating white matter (Davies et al., 1999; Kerschensteiner et al., 2005) and that simultaneous elimination of multiple inhibitory molecules alone did not promote intraspinal regeneration (Lee et al., 2010a; Lee et al., 2010b, but see Cafferty et al., 2010). It is tempting to speculate therefore that preventing presynaptic differentiation alone, particularly during the early stage of injury before full deposition of non-permissive cues (Ramer et al., 2001b), might lead to significant regeneration after spinal root injury. It will also be important to determine whether such stabilizing activity restricts regeneration and/or anatomical plasticity elsewhere in the injured CNS.
We thank Rita Balice-Gordon, Wenbiao Gan and Joshua Trachtenberg for advice on in vivo imaging; Theresa Connors and Amy Kim for technical assistance; Marion Murray, Mickey Selzer and Veronica Tom for comments on the manuscript. This work was supported by NS062320 (Y-J.S.) and World Class University program (R31-2008-000-100069-0) through the National Research Foundation of Korea funded by the Ministry of Education,Science and Technology (Y-J.S, J.K.H).