High-resolution intravital imaging reveals that blood derived macrophages but not resident microglia facilitate secondary axonal dieback in traumatic spinal cord injury

Transgenic mice

Single, double and triple transgenic mice from the Thy-1^YFP H (Feng et al., 2000), Thy-1^CFP (Feng et al., 2000), heterozygous Cx3cr1^+/GFP (Jung et al., 2000) and Ccr2^+/RFP (Saederup et al., 2010) lines were used. All animals are on a C57BL6 background. Multiple transgenic lines included the double transgenic lines Thy-1^+/YFP H × Cx3cr1^+/GFP and Thy-1^+/YFP H × Cx3cr1^GFP/GFP and the triple transgenic line Thy-1^+/YFP H × Cx3cr1^+/GFP × Ccr2^+/RFP and Thy-1^+/YFP H × Cx3cr1^+/GFP × Ccr2^RFP/RFP. We used all adult mice from 8–20 weeks of age, with equal numbers of male and female mice. Mice were housed and bred in the animal resource center facilities at the Case Western Reserve University according to standard husbandry practices. All experiments and animal handling were performed in accordance with protocols approved by the Institutional Animal Care and Use Committee.

Chimeras

Bone marrow transplants were performed when host animals were 10–14 weeks of age. Animals were irradiated at 980 rads, and allowed to rest for 4–5 hours. Hosts received 3 million donor cells isolated from the femur and tibia of donor animals by tail vein injection. Chimera efficiency was verified in the blood 10 weeks post-transplant. Chimera efficiency was established using flow cytometry of blood collected from these animals at 8–12 weeks after transplant. Details of flow cytometry experiments are found below in “Flow cytometry”. Chimerism was determined by comparing the number of F4/80 and CX3CR1^GFP cells. Animals in which the host and the donor were the same genotype were used as a comparison. In animals with the same host and donor, an average of 40.36% CX3CR1GFP cells were found in the blood, with a low of 36.0% and a maximum of 43.6%. In animals with a CX3CR1^GFP host and B6 donor, the average percentage CX3CR1 positive cells found in the blood was 0.58% and ranged from 1.4% to 0.1%. In animals with a B6 host and CX3CR1^GFP donor, the average number of CX3CR1 positive cells found in the blood was 38.73% and ranged from 30.8% to 54.2%. The overall average efficiency of transplantation was 95.95%. Using a two-tailed t-test assuming two samples of unequal variance, the P value comparing transplanted animals with different host and donor to those with the same genotype host and donor was * P<0.00001.

Dorsal Column Crush (DCC) spinal injury model

Mice were anesthetized with inhaled isoflurane gas (2%) for all surgical procedures. A modified version of DCC spinal cord injury described previously (Shen et al., 2009) was performed. Briefly, a laminectomy was performed to expose the T10 spinal cord segment. Small holes were made in the dura 0.5 mm lateral to the midline with a 30 gauge needle and a DCC lesion was made by inserting Dumont # 4 jeweler’s forceps approximately 1 mm into the dorsal spinal cord and squeezing, holding pressure for 10 seconds and repeating two additional times. The laminectomy site was covered with saline-soaked gelfoam (Pfizer) and the muscle and skin closed. The animals received Marcaine (1.0 mg/kg) subcutaneously along the incision site as well as buprenorphine (0.1 mg/kg) intramuscularly. For sham surgeries, the animals received a laminectomy and the intact dura was covered with gelfoam.

In vivo imaging of the DCC lesion with two-photon microscopy

Intravital imaging was performed 0, 2, 4, 5 and 8 days after lesion using a Leica SP5 confocal microscope fitted with a 16W IR laser (Coherent, Inc.). To minimize artifacts from breathing and heartbeat, the spine was stabilized using a STS-A small animal spinal cord holder (Narishige International USA, Inc.) on a customized base. A well for immersion fluid (Dulbecco’s aCSF) was then formed around the spinal clamps with OrthoJet dental acrylic (Lang Dental Manufacturing Co., Inc.). Vessels were labeled by tail vein injection of TRITC-dextran (150,000 MW; Sigma). High resolution 3D (xyz) images were taken with a z-step size of 3um were taken of the top approximately 150 µm of the spinal column. For each animal, mosaic images of the entire opened portion of the spinal column were taken. Additionally, 1–6 hour-long high-resolution 4D (xyzt) movies were acquired with a frame rate of 60 seconds/frame at the caudal side of the lesion where retraction bulbs are found. If any evidence of thermal damage was observed, samples were not used. This included bubbling, changes in general structure of the sample or swelling of the sample.

Flow cytometry

Five days after a DCC injury, Cx3cr1^+/GFP animals were euthanized by CO₂ inhalation. Blood was collected by cardiac puncture and the spleen and spinal cord were harvested. Each group consisted of 3 male mice. 20ul of blood was incubated with heparin, the entire spleen and the section of spinal cord including the DCC injury and 1mm rostral and caudal to the lesion were treated with collagenase and placed on ice for 1 hour. Single cell suspensions of the spleens and spinal cords were generated by pressing tissue through a 40um cell strainer (Corning) using the plunger from a 1 ml syringe (BD) and suspended in PBS with 5% serum (FACS buffer). Blood and tissue homogenates were red cell lysed with ACK lyse (Life Technologies) 1–3 times or until the pellet was no longer red in color, counted on a hemocytometer, and stained with PE conjugated F4/80 antibody (eBioscience) at a concentration of 1/100. Genetically expressed CX3CR1^GFP was used to identify CX3CR1+ cells. CD11b PE and CX3CR1^GFP served as positive controls for color compensation and an isotype control was also used for F4/80 (eBioscience). Samples were analyzed on an Accuri flow cytometer (BD Biosciences) using default flow parameters. Analysis was performed using CFlow Plus (BD Biosciences).

Sciatic nerve PhRodo injection

Animals received an injection of PhRodo-conjugated dextran 10,000 MW (Invitrogen) unilaterally into the sciatic nerve 6 days after a DCC injury. Briefly, the sciatic nerve of the right hind limb was exposed and crushed with #4 forceps. A 4-0 prolene suture was placed around the nerve distal to the crush site, and a hole formed in the epineurium distal to the suture with a 30-gauge needle through which a Hamilton syringe was inserted and 1 ul of PhRodo-dextran (10% in water) was injected into the crush site. Only animals where the crush site was successfully filled with dye were used for experiments. The suture was tightened enough to prevent immediate backflow of dye through the hole, but was not secured tightly around the nerve to avoid ligation of the nerve and the animal was sutured closed. Animals received sub cutaneous Marcaine (1.0mg/kg) and intramuscular bupernex (0.1 mg/kg) (Butler Schein). Animals were imaged as described previously two days after sciatic nerve injection of the PhRodo dye.

Fixed spinal cord tissue analysis by immunofluorescence histology

Double transgenic Thy-1^+/YFP × Cx3cr1^+/GFP or triple transgenic Thy-1^+/YFP × Cx3cr1^+/GFP × Ccr2^+/RFP and Thy-1^+/YFP × Cx3cr1^+/GFP × Ccr2^RFP/RFP mice were perfused with PBS followed by 4% PFA in PBS. Efficient perfusion was verified by blanching of the tissues. Tissue was then post-fixed in 4% PFA in PBS at 4 degrees C, embedded in OCT compound and cut to 20um thick slices. Endogenous RFP, GFP and YFP were used for imaging, and tissue was not counterstained with anti-fluorescent protein antibodies. Slides were mounted in Vectashield anti-fade solution (Vectorlabs) and imaged with a Leica SP5 confocal microscope (Leica).

Image processing and statistical analysis

Image processing was completed using Imaris (Bitplane) to identify axonal processes and track movement of cells. Mosaic images were compiled using XuvStitch (www.xuvtools.org/). Cellular infiltration and PhRodo inclusions were quantified via Imaris using the spots function (Bitplane) to automatically identify cells based on quality (greater than 1.2) and size of spots (greater than 6.7 um in diameter). These numbers were the optimum suggested values from Imaris. Axonal dieback was quantified in 2D projections of mosaic images in ImageJ by measuring from the leading edge of the proximal stump of injured axons to the center of the lesion as identified by a line drawn between the center of the two insertion sites for the two forceps tines. “Destructive” events at the point of macrophage contact with the axon were identified when the width of the axon measured at the initial time of either cell contact or the initial point in our imaging session decreased or increased in width by more than 50% over the imaging session. Points of contact were identified by eye and measured only after cell contact was observed. Data analysis was performed blinded when possible, although data collection was not performed blinded for practical reasons. Movies were compiled using AfterEffects (Adobe). Statistical analysis was completed for the differences between final axonal positions, number of infiltrating cells, number of PhRodo inclusions seen within Cx3cr1^+/GFP cells and migration speeds of monocytes and macrophages. In all cases, all tissue or images from a single animal was considered an N of 1. For example, if 3 sections from the same animal were used, this was considered an N of 1. All error bars are standard error of the mean. Paired Student’s T-tests were used for all analysis in Excel (Microsoft). These tests are either single or double tailed as noted and assumed two samples of unequal variance.

The influx of microglia and monocytes into the spinal cord lesion is closely correlated with axonal dieback

First, we sought to examine the DCC lesion using intravital 2-photon laser scanning microscopy. For these experiments, DCC was used since this injury has minimal functional impact on the animal and still enables us to visualize and measure axonal retraction and immune cell infiltration. This model consisted of a laminectomy followed by a 1mm × 1mm DCC injury to the dorsal columns at the T10 level (Figure 1a). The injury severs ascending sensory axons while leaving larger blood vessels intact and minimizing fibroblast invasion (Figure 1b). To visualize axons, monocytes and microglia at the site of injury in vivo without staining procedures, we bred Thy-1^+/YFP H mice (Feng et al., 2000) and Cx3cr1^+/GFP mice. To image these animals for all following experiments, we re-opened the spinal column at the site of injury. Re-opening surgeries up to 8 days after injury to gain access for imaging proved to be more reliable and introduced less inflammatory cell infiltration at the surgical site than chronic window implantation (data not shown).

This injury model reliably produced an easily quantifiable front of axons that die away from the center of the crush (Figure 1c), replicated axonal dieback and inflammatory cell infiltration (Busch et al., 2009), and produced the same end ball morphology as shown by Ramon y Cajal (Cajal, 1907) (Figure 1d). Axonal dieback was measured from the leading edge of the proximal stump of the damaged axon to the center of the crush site. The center of the site was identified as the line created between the center of the two tines of the forceps insertion points. The average difference in axonal position between 1 hour after injury and 5 days after injury was 231.51 um (Figure 1e), concurrent with an infiltration of Cx3cr1^+/GFP cells. Cx3cr1^+/GFP cells increased in number at the edges and centers of lesions and were in close proximity to the ends of severed axons (Figure 1c, f), roughly doubling in the first two days, increasing by 6 fold at day 5 and 9 fold at day 8. Results were not gender dependent as male and female mice were used. Next, we verified that our time-lapse imaging procedures were not damaging to the tissue. We did not observe any adverse effects or changes in Cx3cr1^+/GFP cell morphology due to the imaging process over several hours. Both axons and the ramified morphology of microglia remained intact with no bubbling or increase in fluorescence as typically seen with thermal damage (Supplemental movie 1). Immediately after injury, time-lapse imaging revealed that Cx3cr1^+/GFP cells began to extravasate from blood vessels and enter the lesion in the first hour after injury (data not shown).

CX3CR1 and CD11b double-positive cells increase in number in spinal cord tissue after injury

To further verify the identity of Cx3cr1^+/GFP cells within the lesion, we performed flow cytometric analysis of dissociated cells isolated from an area 1mm rostral and 1mm caudal to the lesion of injured and sham surgery spinal cords 5 days after injury. We verified the identity of the Cx3cr1^+/GFP cells by staining for CD11b/MAC1, a marker for microglia, monocytes, macrophages and granulocytes in the CNS (Babcock et al., 2003) (Figure 1g). In the spinal cord, there was a 6-fold increase in number of double positive CD11b and GFP cells after injury (Figure 1h). Cell numbers were not statistically different in the blood of DCC injured mice and sham animals (Figure 1h).

Axonal dieback occurs by axotomy in the presence of Cx3cr1^+/GFP phagocytic cells

We next investigated whether we could directly visualize the cellular process by which axons die back away from the lesion site and whether they lose parts of their retraction bulbs or retract into themselves during secondary dieback as they do in vitro. Again, we used double transgenic Thy-1^+/YFP H × Cx3cr1^+/GFP animals to identify axons, macrophages, and microglia, and imaged mice 2, 5 and 8 days after injury, during the period of secondary dieback. We observed large numbers of cell contacts between Thy-1^+/YFP axons and Cx3cr1^+/GFP cells. Surprisingly, some axon-macrophage contacts resulted in significant axonal morphologic changes and destruction during the 60 minutes following initial interaction. An example is shown in Figure 2a and Supplementary movie 2. We defined a destructive contact as one that causes a greater than 50% change in the width of the axon at the point of contact. We also observed destructive axonal contacts at 22 days after injury (Figure 2b), indicating an ongoing process over a prolonged time period. There were an average of 3.14±1.04, 3.28±0.79 and 3.47±1.41 Cx3cr1^+/GFP cell contacts per 200µm of axon per hour on days 2, 5 and 8 after injury, respectively (Figure 2c). In total, we observed 305 individual contacts, of which 11 were destructive during one to two hour imaging sessions. Six of the destructive interactions resulted in axonal thinning and five resulted in a complete severing of the axon during the imaging session. These observations indicate that secondary dieback in vivo takes place by axotomy via contact with a Cx3cr1^+/GFP cell. While there were many remnants of axons scattered about, the small percentage of observed contacts that ended in destruction of axons may be explained by the relatively short periods of observation that were possible in these animals. Many of the contacts we saw during the imaging period may have potentially resulted in a destructive interaction at later time points. Also, we have not observed any destructive events in axons without contact by a CX3CR1 positive cell, supporting the hypothesis that these two events are related. In support of an axonal shredding mechanism, bits of Thy-1^YFP axons were seen along the path of axonal dieback and within the lesion. An example at 5 days after injury is shown (Figure 2d).

Cx3cr1^+/GFP cells phagocytize parts of once intact axons

Next, we investigated whether initially intact portions of the damaged axon can be eventually phagocytized by Cx3cr1^+/GFP cells between days 6 and 8 after DCC injury, at the time when a large portion of dieback takes place. We employed the pH-sensitive PhRodo-dextran conjugated dye (Invitrogen), which becomes fluorescent only in a low pH environment such as that found within lysosomes to identify phagocytized axonal structures. Use of the pH sensitive red dye allowed us to separate inclusions within the GFP positive cells that were derived from phagocytosis of once intact dystrophic axonal ends from phagocytosis of debris generated during the formation of the lesion or debris from the portion of the axon undergoing Wallerian degeneration or from accidental detection of fluorescence in the yellow channel from very bright GFP emission. PhRodo-dextran (10 kD MW) was injected into the sciatic nerve unilaterally on day 6 after DCC injury (Figure 3a). Two days later we observed puncta of fluorescent PhRodo dye within Cx3cr1^+/GFP cells located on the same side of the spinal cord as the sciatic nerve injection (Figure 3b and c). The average number of puncta on the side of injection was 43±2, compared to 17±10 puncta on the contralateral side (P<0.05). Contralateral puncta may be in cells that migrated across the midline. Red puncta found outside the parenchyma of the spinal cord were excluded. Puncta within the spinal cord outside of CX3CR1^+/GFP cells were not observed. We also observed Thy-1⁺ inclusions within the cytoplasm of Cx3cr1^+/GFP cells, indicating that phagocytic uptake of axonal remnants by Cx3cr1^+/GFP cells is not solely restricted to those containing PhRodo-dextran (Figure 3d, e).

Blood derived monocytes comprise the majority of mobile Cx3cr1^+/GFP cells within the lesion

Since microglia are resistant to irradiation, and numbers of Cx3cr1^+/GFP cells that can enter the CNS after irradiation are proportionately small (Ajami et al., 2007), we used a bone marrow chimera animal approach designed to distinguish the involvement of spinal cord resident microglia or bone marrow-derived cells to observe the differences between these two cell types in vivo (Figure 4a). Two variations of irradiation chimera animals were used. In the first, the host animal expressed Thy-1^YFP and the donor animal was heterozygous for Cx3cr1^GFP, allowing us to identify GFP⁺ macrophages. In the second, the host animal expressed both Thy-1^YFP and was heterozygous for Cx3cr1^GFP and the donor was a C57BL6 animal expressing no fluorescent proteins, allowing us to identify resident microglia. We also used a Thy-1^+/YFP host with a double transgenic Cx3cr1^+/GFP and Ccr2^+/RFP donor to identify two lineages of blood derived monocytes/ macrophages (Geissmann et al., 2003). We also checked for chimera efficiency at replacing monocytes in the blood of transplanted mice, confirming an average efficiency of replacement of 96% of the host Cx3cr1^+/GFP blood cells with those from the donor (See methods for details). Spinal cord injury was performed 8–12 weeks after bone marrow transplantation to allow the immune system time to fully reconstitute. Using this approach, we did not observe any differences in the extent of axonal dieback or in the morphology of microglia, monocytes or axons within the lesion (Figure 4b, c). At 2 hours, 2 days, 5 days and 8 days after injury, we observed a large increase in blood-derived Cx3cr1^+/GFP cells with very little increase in the Cx3cr1^+/GFP resident cell population (Figure 4d and 4e). We found that resident microglia appeared larger and more complex than monocytes and had simple extended processes, whereas monocytes were round (Figure 4d). In the first hours after injury, we saw an insignificant difference between resident and infiltrating cells from that in the uninjured animal. Blood derived cells increased by 3 fold from day 0 to 2 and 7 fold from day 0 to 8. There was no significant difference in the amount of resident cells at any time point after initial injury (Figure 4e). These findings are in concordance with previous studies in contusion models in rat chimeras and others (Popovich and Hickey, 2001; Popovich et al., 2003; Mawhinney et al., 2012). Although irradiation of the spinal cord may attenuate proliferation of resident cells, these data revealed that the major increase in cellularity in the spinal lesion was attributed to infiltrating Cx3cr1^+/GFP cells.

Monocyte-derived macrophages, not resident microglia, are responsible for secondary axotomy

Next, we looked to see if any behavioral properties varied between resident and infiltrating Cx3cr1^+/GFP cells. Migratory capabilities of either monocytes or macrophages within traumatic spinal cord lesions have not been described previously. Using the same chimeric animals described in Figure 4A, we observed differences in the speed and displacement of migration between blood-derived monocytes and resident microglia. The main population of Cx3cr1^+/GFP cells within the lesion was composed initially of resident microglia that showed morphological activation. Within the first hour following injury, a small number of blood-derived Cx3cr1^+/GFP cells began to extravasate from the blood vessels into the lesion, while resident microglia exhibited very little movement, as shown by tracks (Figure 5a–c, Supplementary movie 3). A higher magnification of the movements of two blood derived cells is shown in Figure 5B. From the dynamic images, we measured speed and displacement. In the first hour and 8 days after injury the average speed of cord-resident microglia was 0.030 ± 0.0014 µm/second and 0.037 ± 0.0017, respectively, while these same measurements for blood-derived Cx3cr1^+/GFP cells were 0.06 ± 0.0008µm and 0.061 ± 0.0008 (Figure 5d). In the first hour and 8 days after injury the average displacement of cord-resident microglia was 4.47 ± 0.22 µm and 5.51 ± 0.40, respectively, while these same measurements for blood-derived Cx3cr1^+/GFP cells were 15.4 ± 0.32µm and 14.35 ± 0.40 (Figure 5e). In the images shown in a and c, differences in directionality of blood derived cell movement can be observed through the time-coding of cell position shown by the changing color of the tracks. Using this coding, the earliest times in imaging are identified as blue and the latest time points are identified as white. These colors can be seen in the tracking line for each cell. In the first hour after injury, cells were primarily localized within the vessel lumen at the beginning of the imaging period and were seen to migrate from the blood vessel and into the lesion by the end of the 60-minute imaging period (Figure 5a). No directional migration of cells towards either the proximal or distal portion of the lesion was noticed. This migration from the dorsal vein in the spinal cord was only seen immediately after injury. At later time points, there was a decrease in cell migration within the perivascular region and an increase in migration within the center of the lesion. No directionality or focus of movement of cells within the center of the lesion at these later time points was noted, or expected due to the amount of tissue damage seen throughout the lesion (Figure 5c). Directionality of movement was also difficult to assess due to movement of cells into and out of the field of view and large populations of stationary or near stationary CX3CR1^+/GFP cells.

Next, we observed the cell contacts between these blood derived and resident Cx3cr1^+/GFP cells within the lesion in the chimeric animals to see if a subset of these was responsible for destructive contacts. In one case, we observed a quickly moving Cx3cr1^+/GFP cell at the lesion edge contacting an axon in two places, in each case with a destructive effect (Figure 6a and Supplementary movie 4). In another example, the axon was completely severed (Figure 6b). We did not observe any destructive changes where microglia contacted axons (Figure 6c, e). We observed a total of 232 contacts and 22 destructive contacts in 10 Cx3cr1^+/GFP→WT chimeric mice (Figure 6a, b and d). We observed a total of 52 contacts and no destructive contacts in 11 WT→Cx3cr1^+/GFP chimeric mice (Figure 6 c and e). These data indicate that destructive contacts largely, but perhaps not entirely, occur between blood derived cells and axons.

Blood derived CX3CR1^GFP/+ cells are very rarely also CCR2^RFP/+ and axonal position after injury is not dependent on CCR2 expression

To further clarify the identity of infiltrating CX3CR1 phagocytic cells, we took advantage of differential expression of CCR2 and CX3CR1 that can identify different subsets of circulating macrophages (Mahad et al., 2006; Saederup et al., 2010). We did not observe any destructive events in axons outside of CX3CR1 positive cell contact, so we looked in fixed tissue sections to see if these CX3CR1 positive blood derived cells were also CCR2 positive. To this end, we generated chimeric animals with a Thy-1^+/YFP host and Ccr2^+/RFP × Cx3cr1^+/GFP marrow donor. In fixed sections from the transplanted animal in which we were able to observe CCR2 RFP using confocal imaging, CX3CR1 and CCR2 fluorescence was restricted to only blood-derived monocytes and macrophages but not microglia. Both Ccr2^+/RFP cells and Cx3cr1^+/GFP cells were blood derived and capable of infiltrating into the lesion following spinal trauma, while virtually no Ccr2^+/RFP cells were found at sites away from the lesion. Ccr2^+/RFP cells were round in appearance resembling monocyte morphology while Cx3cr1^+/GFP cells were larger with granular cytoplasm consistent with tissue phagocytes in both the normal and transplanted animals (Figure 7a). The infiltrating cells comprise three populations in the chimeric animals. CCR2 single positive cells comprised the largest population, followed by CX3CR1 single positive cells. There was also a very small population of double positive cells within the lesion (Figure 7b), possibly comprised of CCR2 cells that are in the process of down regulating CCR2 and upregulating CX3CR1 as macrophages mature. Since we observed destructive contacts with Cx3cr1^+/GFP positive cells (Figures 2, 6) and the double labeled infiltrating cells were a small minority (Figure 7b), we speculate that destructive contacts were initiated by the subgroups of infiltrating cells that are Cx3cr1^+/GFP positive, and not those that are Ccr2^+/RFP single positive. This was not directly tested due to the dimness of CCR2^RFP/+ cells in our live imaging set-up. Next, to verify that these CCR2 positive cells are unlikely to be involved in axonal dieback, we looked at the axonal position 5 days after DCC in triple transgenic Thy-1^YFP, CX3CR1^GFP/+ and either CCR2^RFP/+ or CCR2^RFP/RFP animals (Figure 7c, 7d). In this case, we did not see any difference in axonal position when measured from the leading edge of the proximal stump of the axon leading up to the injury (Figure 7e), indicating that CCR2 is not directly necessary for axonal dieback.

Mechanism of secondary dieback by axonal membrane loss

We have observed that axonal retraction during the secondary injury phase occurs via multiple axotomies along the distal portion of injured stalled axons which, in turn, results in the removal of the end of the axon, creating a new axonal end. In contrast to this, in vitro activated macrophages caused retraction of dystrophic axons over a long distance in a manner involving rapid dissolution of the axonal cytoskeleton, causing the distal axon to rapidly retract into itself (Horn et al., 2008). This difference in mechanism is not surprising, as neuron behavior in vitro occurred without the influence of extracellular matrix and surrounding support cells. Similar axotomy and loss of axonal membrane during dieback instead of axonal involution has been proposed to occur in the CNS parenchyma following several other models of injury where macrophages interact with injured axons. These include zymosan injection (Fitch and Silver, 2008) and traumatic brain injury (Greer et al., 2013). Furthermore, our real time observations of blebbing and loss of membrane from axonal ends is consistent with the suggested pattern of axonal dieback in the cortex after an ischemic stroke, where axonal debris is found at the location where the uninjured axon was present before the axon retracted towards its cell body and away from the site of injury (Hawthorne et al., 2011). Axonal thinning has also been shown in models of EAE (Powers et al., 2012). Our observations provide further distinction between the initial phase of axonal retraction that is cell intrinsic and decreased in the Wallerian degeneration slow mouse (Kerschensteiner et al., 2005) and the secondary, inflammatory phase of retraction.

Removal of membrane from axonal ends may have increased significance in the context of sprouting and attempted regrowth of axons within the core of the lesion. Axonal ends in the lesion are not stationary (Busch et al., 2009) and attempt to sprout in the first few days to weeks after injury (Kerschensteiner et al., 2005) largely in association with NG2 glia whose surface is preferred by dystrophic axons and upon which the dystrophic axon tips eventually stabilize and form synaptic –like connections (Busch et al., 2010; Filous et al., 2013). It is possible that activated macrophages, in addition to causing axonal dieback at early stages, may also be involved in curtailing the ability of axons to sprout. It is also possible that macrophages move from cell to cell long term, removing membrane continually in a fashion similar to the process of synaptic pruning that occurs during development (Schafer et al., 2012; Schafer et al., 2013).

Since infiltrating cells appear to be damaging, we looked at the CCR2 and CX3CR1 markers found on these cells in fixed tissue to differentiate monocytes expressing these two chemokine receptors (Saederup et al., 2010). The CCR2^hi/ CX3CR1^lo population of monocytes differentiates into macrophages or dendritic cells, while the CCR2^lo/ CX3CR1^hi population may aid in wound healing and other functions (Geissmann et al., 2003). Although speculative, it appears that the CCR2^+/RFP cell population is less likely to be involved in dieback as we have not seen axonal destruction events that are not associated with contact by a Cx3Cr1^+/GFP cell. Therefore, since few Cx3Cr1^+/GFP cells are also CCR2^+/RFP positive, we can infer that these cells are less likely to be involved in destructive events. This supports the idea that there are subtypes of macrophages that may be more or less harmful during trauma, and increasing the population that does not cause destruction may be therapeutic (Kigerl et al., 2009). The distinction between these subtypes of macrophages and microglia is especially interesting, since it supports the idea that microglia are involved primarily in maintaining CNS homeostasis, and continue to perform such functions even after a traumatic injury (Davalos et al., 2005).

Separating the two main Cx3cr1^+/GFP monocyte and microglial cell lineages in the spinal cord presents a technical challenge. Most models to-date involve manipulation of the animals, which could cause potential permanent changes in the CNS cellular architecture. These manipulations included an in utero transplantation chimera model (Nijagal et al., 2011), a head-shielding radiation model (Shechter et al., 2009), busulfan chimeras (Kierdorf et al., 2013a), parabiosis (Ajami et al., 2011), macrophage depletion by clodronate liposomes (Horn et al., 2008), and labeling of monocytes with CCR2^RFP (Ma et al., 2002; Saederup et al., 2010). None of these approaches, including ours, are without caveats. Many of them are technically challenging and may not have high enough efficiency to distinguish the two cell types. Irradiation has also been shown to reduce microglial activation for the first week in the retina (Bosco et al., 2012), which may persist. Busulfan is a chemotherapy drug that can be used to ablate host bone marrow in place of irradiation. Busulfan chimeras have reduced infiltration into the CNS and blood brain barrier effects than irradiation chimeras, although peripheral cell recruitment to the facial nerve after axotomy is drastically reduced (Kierdorf et al., 2013a), making the study of infiltrating cells more difficult. Clodronate liposomes subject the animal to inflammation associated with death of the cells (Van Rooijen and Sanders, 1994), preventing simple observation in an otherwise normal animal. In our hands we did not observe any irreparable differences between control animals and bone-marrow chimeras in terms of the degree of axonal dieback or changes in cell morphology, suggesting that our techniques are as minimally invasive as possible.

Further investigating knock out of chemokine receptors

In Ccr2^RFP/RFP animals, a decrease in macrophage infiltration into tissue has been observed in many models including corneal injury (Oshima et al., 2006), EAE (Gaupp et al., 2003), toxoplasma gondii infection (Dunay et al., 2008) and neuropathic pain (Abbadie et al., 2003). In cases where there is still some macrophage infiltration, there are also other chemokines that may be able to compensate for loss of CCR2 (Gaupp et al., 2003). Again, we observed an infiltration of Cx3cr1^+/GFP and CCR2^+/RFP cells into the lesion in CCR2 deficient animals, and no change in axonal position. In this case, it is likely that we do not see a change in axonal position simply because the CX3CR1^hi cells involved in axonal dieback are still able to migrate into the lesion independent of CCR2.

A second chemokine that will need to be investigated in the future is CX3CR1 which is involved in monocyte motility and is the only known receptor for Fractalkine, a chemokine that is expressed on neurons and peripheral endothelial cells (Yona et al., 2013),. Varying degrees of differences in disease progression and severity have been documented between Cx3cr1^GFP/GFP and Cx3cr1^+/GFP animals in a variety of mouse models of disease and injury. These range from no difference in mouse models of facial axotomy (Jung et al., 2000), to larger changes in Alzheimer’s disease (Bhaskar et al., 2010; Lee et al., 2010), spinal cord injury (Donnelly et al., 2011), multiple sclerosis (Huang et al., 2006), amyotrophic lateral sclerosis (ALS) and Parkinson’s disease (Cardona et al., 2006). Although not investigated here, comparison of axonal dieback between CX3CR1^GFP/GFP and CX3CR1^+/+ animals will be instructive to the understanding of these cellular interactions. Future experiments in another model of dieback, such as in the rat, where greater retraction distances occur may be helpful (Horn et al., 2008). Perhaps inhibiting more than one chemokine receptor at the same time will add specificity to the well-established data suggesting that global macrophage depletion can inhibit secondary dieback.

Further investigating destructive cell types within the lesion

Even though we have identified Cx3cr1^hi macrophages as the cell population responsible for secondary axonal dieback, we have yet to define the molecular signals involved in this process. We have also yet to determine the significance of the Cx3cr1^lo and Ccr2^hi cells in the lesion. Furthermore, the activation state of the macrophage may also be important, as Cx3cr1^+/GFP macrophages early on within the DCC lesion may be activated in a way that promotes phagocytosis and removal of damaged tissues and potential pathogens or later on in a way that promotes healing and scar tissue formation (Kigerl et al., 2009).

Many molecular candidates involved in the macrophage-dystrophic axon interaction have been proposed, including receptors for phosphatidylserine (Titsworth et al., 2008) such as milk-fat globule epidermal growth factor-like factor 8 (MFG-E8) (Li et al., 2012). Signaling may also occur via changes in membrane permeability through perforin or calcium-mediated processes (Coleman, 2005). Blockade of proteases such as MMP9 has been used to prevent axonal retraction in vitro, arguing that dieback may require protease action between macrophages and dystrophic axons (Busch et al., 2009). Chemotaxis may also be an important contributor in recruiting relevant cells to the correct sites. Although we have determined that, in this model, CCR2 is not necessary for this secondary axonal dieback phenotype, other chemoattractants are likely involved. Further investigation into what these signaling pathways are in traumatic lesion is warranted, as well as the mechanism of interaction between the macrophage and axon. Separating these two subpopulations of cells and elucidating their roles in secondary damage will play a large role when approaching more specific interventions for treatment of spinal cord injury.