Our previous work identified LZK as a promoter of astrocytic scar formation at the injury site following spinal cord injury, and established genetic mouse models that enable tamoxifen-inducible LZK deletion (GFAP-CreER^T2; LZK^f/f) or LZK overexpression (GFAP-CreER^T2; LZK^OE) in adult astrocytes (Chen et al., 2018). Here, we sought to assess the contribution of axon degeneration-reactive astrocytes, which are non-scar forming, to CST axon sprouting. In order to do this, we first determined if LZK regulates axon degeneration-reactive astrocytes, which would then allow for their modulation via genetic manipulation of LZK expression (Chen et al., 2018).

We utilized the unilateral photothrombotic (PT) stroke model to examine the interplay between axon degeneration-reactive astrocytes and CST axon sprouting (Figure 1A). Unilateral PT to the primary motor cortex is commonly used to experimentally study intra-spinal CST axon sprouting (Lapash Daniels et al., 2009; Lindau et al., 2014; Liu et al., 2014; Wahl et al., 2014; Kaiser et al., 2019). We targeted PT to the forelimb sensorimotor cortex unilaterally, which resulted in degeneration of CST axons originating from the ablated corticospinal neurons (CS neurons) and sprouting of intact CST axons originating from contra-lesional (uninjured) CS neurons (Figure 1A). CST axon degeneration and axon sprouting were examined at the cervical enlargement of the spinal cord, where forelimb-projecting CST axons innervate spinal motor neurons to control forelimb movement (Figure 1A).

In sham-injured mice, protein kinase C isoform gamma (PKCγ)—a marker of the CST—intensely labeled the intact CST main bundle in the dorsal funiculus of the cervical spinal cord (Figure 1B, top panel), with minimally detectable astrocytes positive for the reactivity marker GFAP (Figure 1B, top panel). Following unilateral PT stroke, degeneration of CST axons originating from the ablated neurons was evident by reduced PKCγ immunoreactivity on 14 days post injury (dpi), concurrent with the appearance of stellate GFAP⁺ reactive astrocytes (Figure 1B, bottom panel, “deg CST”). Furthermore, these axon degeneration-reactive astrocytes upregulated endogenous LZK protein expression (Figure 1C, bottom panel). Compared to the intact CST, the number of LZK⁺GFAP⁺ cells in the degenerating CST increased by four-fold on 7 dpi, and sustained until at least 28 dpi (Figure 1C, graph). These findings suggest a positive role of LZK in astrocyte reactivity to CST axon degeneration, and that manipulation of LZK gene expression may modulate such reactivity.

To test LZK-dependent regulation of astrocyte reactivity to axon degeneration, we used GFAP-CreER^T2; LZK^f/f mice (herein referred to as astrocytic LZK-KO mice) and GFAP-CreER^T2; LZK^OE mice (herein referred to as astrocytic LZK-OE mice) that allow tamoxifen-inducible LZK deletion or overexpression in adult astrocytes, respectively (Chen et al., 2018). Tamoxifen treatment was given one week prior to unilateral PT. Tamoxifen-treated GFAP-CreER^T2 littermates were used as genotype controls.

We compared astrocyte reactivity to CST axon degeneration based on GFAP immunoreactivity among control, astrocytic LZK-KO, and astrocytic LZK-OE mice in a time course after unilateral PT. In genotype control mice, astrocyte reactivity localized to the degenerating CST increased by 1.5-fold on 7 dpi that further increased by 2–3 fold on 14–28 dpi respectively, compared to sham baseline (Figures 2A,B). These axon degeneration-reactive astrocytes also expressed LZK (Figures 1C, 2A). Astrocytic LZK deletion did not affect the baseline level of GFAP immunoreactivity in the CST without injury (Figure 2B). Following unilateral PT, astrocytic LZK-KO mice showed reduced LZK expression and reactive astrocyte density within the degenerating CST (Figure 2A), though exhibited only a statistically non-significant trend toward lower astrocyte reactivity to CST axon degeneration compared to the genotype control on 7 and 14 dpi (Figure 2B).

In contrast, astrocytic LZK-OE mice showed 50% increase in astrocyte reactivity to CST axon degeneration that was sustained from 14 to 28 dpi, compared to the genotype control at the same timepoints (Figures 2A,B). This was accompanied by a strong upregulation of LZK and its fluorescent reporter tdTomato in GFAP⁺ cells (Chen et al., 2018; Figure 2A). Interestingly, in spite of this robust stimulation of astrocyte reactivity to CST axon degeneration, there was no difference in baseline astrocyte reactivity within the CST between control and astrocytic LZK-OE mice in the absence of injury (Figures 2B,C). This suggests astrocytic LZK-OE alone is insufficient to induce astrocyte reactivity in the healthy CST without injury.

Following unilateral PT, astrocyte reactivity in the spinal cord was largely localized to the degenerating CST. The level of GFAP immunoreactivity in the spared hemi-CST did not change from sham baseline in the control and astrocytic LZK-KO mice throughout the time course (Figure 2C). Minimal increase was observed in astrocytic LZK-OE mice on 28 dpi (Figure 2C). Lastly, the level of CST degeneration in the cervical spinal cord was verified to be comparable across all genotypes, based on quantification of PKCγ immunoreactivity in the degenerating hemi-CST in the cervical spinal cord (Figure 2D), indicating consistency in injury severity in all animals. Taken together, these findings demonstrate that following unilateral PT stroke to the primary motor cortex, while astrocytic LZK deletion did not abrogate reactive astrocytes localized to the degenerating CST, astrocytic LZK overexpression amplified astrocyte reactivity to CST axon degeneration in the spinal cord.

Given that LZK promotes astrocyte reactivity to CST axon degeneration in the spinal cord, we next used astrocytic LZK mutant mice to determine the effects of modulating axon degeneration-reactive astrocytes on intra-spinal CST axon sprouting following unilateral PT. As before, LZK gene manipulation in adult astrocytes was achieved by tamoxifen treatment one week prior to unilateral PT. Two weeks after injury, the neuronal tracer biotinylated dextran amine (BDA) was injected into the contra-lesional forelimb sensorimotor cortex to label CST neurons that project to the cervical spinal cord. At the end of a 4 weeks survival period after PT, CST axon sprouting across the midline in the cervical enlargement of the spinal cord was analyzed by BDA immunostaining (Figure 3A).

In the absence of injury, baseline intra-spinal CST axon sprouting was similar among all genotypes (Figures 3B,C). Unilateral PT induced CST axon sprouting by approximately two-fold in genotype control and likewise in astrocytic LZK-KO mice on 28 dpi, compared to their respective no injury baseline (Figures 3B,C). The lack of effect of astrocytic LZK deletion on CST axon sprouting correlates with its inability to abrogate astrocyte reactivity to CST axon degeneration (Figures 2A,B, 3B,C). In contrast, astrocytic LZK-OE mice exhibited a four-fold increase in injury-induced CST axon sprouting compared to its own sham baseline (Figures 3B,C), which is a two-fold enhancement compared to the injured genotype control or astrocytic LZK-KO mice (Figures 3B,C). Notably, the ability of astrocytic LZK overexpression to augment CST axon sprouting requires the presence of injury (Figures 3B,C). The stimulatory effects of astrocytic LZK overexpression on CST axon sprouting positively corresponds to its ability to amplify astrocyte reactivity to CST axon degeneration after unilateral PT (Figures 2A,B, 3B,C). Taken together, these results show that LZK deletion in adult astrocytes does not allow for testing of the role of axon degeneration-reactive astrocytes in axon sprouting by loss-of-function approach; whereas amplification of astrocyte reactivity by LZK overexpression in adult astrocytes stimulates axon sprouting but only in the presence of injury.

Because CST axon sprouting can facilitate functional recovery, we next assessed motor recovery in control, astrocytic LZK-KO, and astrocytic LZK-OE mice following unilateral PT. We first utilized the rotarod test to examine general motor coordination, balance, and endurance over 8 weeks of recovery period. Unilateral PT resulted in 30–50% deficit in rotarod performance at 1 week after injury (Figure 4A). Compared to genotype control and astrocytic LZK-KO mice that reached 76–78% of their pre-injury performance by 8 weeks after injury respectively, astrocytic LZK-OE mice showed a statistically non-significant trend toward full recovery to their pre-injury level by 8 weeks (Figure 4A).

In addition to assessing gross motor function, we also examined skilled forelimb stepping by regular horizontal ladder. The ladder task detects limb-specific, subtle or chronic deficits in paw placement accuracy and compensatory adjustments (Farr et al., 2006). No difference at baseline performance was observed among control, astrocytic LZK-KO, and astrocytic LZK-OE mice by ladder assessment (Figure 4B). Unilateral PT resulted in subtle deficit and adaptive use of the affected forelimb, shown by a two-fold increase of partially placed steps taken by the affected forelimb at 8 weeks after injury compared to pre-injury baseline in the control group (Figure 4B, middle graph). Correspondingly, there was a decrease in correctly placed steps (did not reach statistical significance, Figure 4B, left graph) and also a decrease in slips/misses (statistically significant, Figure 4B, right graph) following unilateral PT. This injury-dependent adaptation to partial placement in stepping of the affected forelimb was similar across genotypes. No statistically significant change in stepping behavior was observed in the unaffected forelimb (Figure 4C). Taken together, astrocytic LZK-OE mice showed a statistically non-significant trend toward improved recovery of general motor coordination as assessed by rotarod, but did not affect skilled forelimb stepping on horizontal ladder.

We next sought to determine the molecular basis of LZK’s ability to augment CST axon sprouting. Our previous work identified LZK as an activator of signal transducer and activator of transcription 3 (STAT3) (Chen et al., 2018). One of the transcriptional targets of STAT3 is the secreted neurotrophic cytokine ciliary neurotrophic factor (CNTF) (Keasey et al., 2013), which is in turn an activating ligand of the JAK-STAT pathway (Stahl et al., 1994). Additionally, intra-spinal AAV-mediated overexpression of CNTF is sufficient to increase sprouting of CST axons in the cervical spinal cord following unilateral transection of the CST at the brainstem (Jin et al., 2015).

Based on this evidence supporting LZK as an upstream regulator of CNTF, we tested if astrocytes in LZK-OE mice produce CNTF following unilateral PT. We first examined CNTF expression in astrocytes reactive to CST axon degeneration in the cervical cord. In genotype controls, unilateral PT induced a ten-fold increase in the number of CNTF⁺GFAP⁺ cells within the degenerating CST (Figures 5A,B). Astrocytic LZK deletion reached only a six-fold increase by the same measurement, but this reduction compared to the genotype control did not reach statistical significance (Figure 5B). While astrocytic LZK overexpression increased the number of CNTF⁺GFAP⁺ cells within the CST at no injury baseline, the number of CNTF⁺GFAP⁺ cells localized to the degenerating CST was comparable to that observed in genotype control (Figures 5A,B). These results suggest that first, astrocytes reactive to axon degeneration express CNTF and therefore have the potential to support axon growth; second, CNTF-producing reactive astrocytes within the degenerating CST, which were present in similar numbers in genotype control and LZK-OE mice, likely do not account for the ability of astrocytic LZK overexpression to enhance CST axon sprouting.

Next, we examined CNTF expression in astrocytes in the spinal gray matter on the denervated side of the spinal cord, where sprouting CST axons cross the midline just below the dorsal column (Figures 3C, 5A). Within this sprouting target in the gray matter, the genotype control had minimally detectable number of CNTF⁺GFAP⁺ cells in the absence or presence of unilateral PT, likewise for the astrocytic LZK-KO group (Figures 5A,C). In contrast, astrocytic LZK overexpression significantly increased the number of CNTF⁺GFAP⁺ cells within the sprouting target gray matter, by more than six-fold compared to the genotype control at either sham baseline, or following unilateral PT (Figures 5A,C). Notably, such increase of CNTF-producing astrocytes was not limited to the sprouting target gray matter, but observed throughout the spinal gray matter of astrocytic LZK-OE mice (Figures 5A,C). The unique presence of CNTF-producing astrocytes exclusively in the spinal gray matter of astrocytic LZK-OE mice presents a potential mechanism that promotes axon sprouting in these animals. Despite widespread CNTF-producing astrocytes in the gray matter of astrocytic LZK-OE mice, CST axon sprouting only occurred in the presence of injury (Figures 3B,C). These findings together demonstrate that LZK overexpression in adult astrocytes stimulates non-scar forming reactive astrocytes that produce CNTF in the spinal gray matter, which may contribute to enhance CST axon sprouting in astrocytic LZK-OE mice in the presence of unilateral PT stroke.

Mice. All mouse husbandry and experimental procedures were performed in compliance with protocols approved by the Institutional Animal Care and Use Committee at the University of Texas Southwestern Medical and the University of Kentucky. Tamoxifen inducible and astrocyte-targeted LZK-knockout mice (GFAP-CreER^T2;LZK^f/f) and LZK-overexpression mice (GFAP-CreER^T2;LZK^OE) have been previously described (Chen et al., 2018). Both lines were bred into C57BL/6 background (N > 10 generations) for this study. Adult male and female mice at the age of 10–12 weeks were used. To delete LZK from astrocytes in adult GFAP-CreER^T2;LZK^f/f mice, 75 mg/kg of tamoxifen was given by oral gavage once daily, consecutively for five days. To induce LZK overexpression in astrocytes of adult GFAP-CreER^T2;LZK^OE mice, 75 mg/kg of tamoxifen was given by oral gavage once daily, consecutively for two days. Tamoxifen dosage was reduced in GFAP-CreER^T2;LZK^OE to prevent severe astrogliosis-associated lethality (Chen et al., 2018). One week after the last dose of tamoxifen treatment, all mice were subjected to unilateral cortical photothrombosis. All surgical procedures, animal dissection, tissue processing, and data analyses were performed blind to genotypes.

Photothrombotic stroke. Unilateral photothrombotic stroke was performed similarly as previously described (Watson et al., 1985; Poinsatte et al., 2019; Becker et al., 2020). Mice were anesthetized with isoflurane (3.5% induction, 2% maintenance in 70% nitrous oxide and 30% oxygen mixture). While fixed in a stereotactic frame (Kopf), the mouse scalp was injected with lidocaine and incised at the midline to expose the intact skull. Rose Bengal (40 mg/kg, 5 mg/ml in saline, Sigma) was injected intraperitoneally one minute before delivery of focused 561 nm laser illumination (Coherent, laser diameter of 3 mm) to the forelimb motor cortex centered at 1.7 mm lateral to bregma under a surgical microscope (Nikon SMZ800), at 45 mW for 15 min. Sham-injured mice received anesthesia, analgesia, scalp incision, Rose Bengal injection, but no laser illumination.

Anterograde labeling of CST axons. Two weeks after photothrombosis, mice used for analyses of CST axon sprouting received BDA injections to anterogradely trace forelimb targeting CST axons originating from the contra-lesional sensorimotor cortex. 0.5 μl of 10% BDA in saline (10,000 MW, Invitrogen D1956) was injected into each of three sites: (AP, ML) in mm from Bregma: (0.5, 1.25), (−0.5, 1.25), (0, 2) at the depth of 0.6 mm. Two weeks after BDA injection (equivalent to 4 weeks after photothrombosis), mice were euthanized for brain and spinal cord dissections.

Histology. Terminal anesthesia was performed by isoflurane overdose. Mice were perfused with 4% paraformaldehyde intracardially, after which brains and spinal cords were dissected and post-fixed at 4°C overnight. Tissues were cryoprotected in 30% sucrose at 4°C overnight before embedding. C4-C7 spinal cord was embedded for transverse sections in O.C.T compound (Scigen Tissue-Plus) on dry ice. Embedded tissue blocks were sectioned at 15 μm thickness on a cryostat (Leica). Free floating sections were collected in PBS with 0.01% sodium azide for further histological examination.

Immunohistochemistry. Staining procedures have been described previously (Chen et al., 2018). Free floating sections were washed twice (0.2% triton in PBS, for 10 min), then blocked and permeabilized (0.4% triton, 5% sera matching species of secondary antibody in PBS) for 1 h at room temperature. Primary antibody incubation (see antibody concentrations below, in 0.2% triton, 1% matching sera, 0.01% sodium azide, in PBS) was carried out at room temperature overnight on a rocking platform. Sections were washed three times and incubated in secondary antibody solution (all secondary antibodies used at 1:500) for 2 h at room temperature on a rocking platform. After three washes, sections were incubated with DAPI (1 μg/ml in PBS) for 10 min, then mounted onto glass slides (Fisher Scientific) and cover slipped with Fluoromount-G (Southern Biotech). BDA staining was done as previously described (Geoffroy et al., 2015). Coronal sections of the cervical spinal cord were incubated in Vectastain ABC solution (Vector Laboratories) at 4°C overnight, washed three times in PBS, then incubated with TSA Plus solution at 1:200 (Perkin Elmer, see below) according to manufacturer’s instructions for 10 min, protected from light at room temperature.

Antibodies. Commercially available primary antibodies used were: LZK (1:500, rabbit, Sigma-Aldrich HPA016497), GFAP (1:1,000, rabbit, Dako Z0334), GFAP (1:1,000, rat, Invitrogen 2.2B10), CNTF (1:300, goat, Novus AF-557-SP), PKCγ (1:1,000, mouse, Santa Cruz SC166385). Alexa Fluor-tagged secondary antibodies used were Alexa 488, Alexa 546, and Alexa 647. BDA detection was performed using TSA Plus Cyanine 5 or Cyanine 3 (Perkin Elmer).

Microscopy and quantification. Stained tissue sections were imaged by an epifluorescent slide-scanning microscope (Zeiss Axio Scan.Z1). Image analyses were done using ImageJ software. All quantifications were conducted by observers blind to genotypes. Images of transverse sections from the cervical spinal cord at levels C7 were used for the following quantifications. For exact values of n per group, see figure legends.

To quantify PKCγ immunofluorescence intensity of the degenerating hemi-CST (area labeled as “deg CST” in Figure 1B) in the cervical spinal cord, integrated density within a semi-oval region of interest covering the degenerating hemi-CST (about 25,000 μm²) was normalized to that of the intact hemi-CST (area labeled as “deg CST” in Figure 1B). Four transverse sections of C7 spinal cord per animal were averaged to generate PKCγ signal intensity per animal. To quantify GFAP immunofluorescence intensity of the degenerating hemi-CST (area labeled as “deg CST” in Figure 1B) or intact hemi-CST (area labeled as “intact CST” in Figure 1B), integrated density within a semi-oval region of interest covering the respective hemi-CST (about 25,000 μm²) was averaged over five transverse cervical spinal cord sections per animal after subtraction of background signal.

For LZK⁺GFAP⁺ cell count in the degenerating hemi-CST, the number of cells double positive for LZK and GFAP by co-immunofluorescence staining was quantified within the semi-oval region labeled as “deg CST” (Figures 1B,C), where reactive astrocytes localized. Three transverse sections of C7 spinal cord per animal were averaged to generate cell count per animal. For CNTF⁺GFAP⁺ cell count in the degenerating hemi-CST, the number of cells double positive for CNTF and GFAP by co-immunofluorescence staining was quantified within the region outlined as “deg CST” in Figure 5A (area of 25,837 μm²). Three transverse cervical spinal cord sections per animal were averaged to generate cell count per animal. Likewise, CNTF⁺GFAP⁺ cells in the sprouting target area of the denervated gray matter were quantified within the region outlined as “target GM” in Figure 5A (area of 40,000 μm²).

CST axon sprouting was quantified as previously described (Grider et al., 2006). Density of BDA-labeled axons on the denervated side of the spinal cord was quantified within a rectangular area of 0.54 mm² in transverse sections of C7 spinal cord, and normalized against the axon density on the innervated side of the spinal cord of the same sections. Axons in the main CST were excluded from analysis. Six sections of C7 spinal cord per animal were averaged to generate the axon sprouting index per animal.

Behavioral analyses. Behavioral tests were done with the observer blind to genotypes. Two motor tasks were used to assess motor function before and after cerebral cortical photothrombosis. To evaluate gross motor coordination, mice were placed on a five-lane elevated rotating rod (IITC Life Science), facing away from the observer while walking on the rod. Rotarod was set with starting speed of 4 rpm, constant acceleration of 0.2 rpm/s, and top speed of 44 rpm. Maximum test duration was 300 s. Mice were scored based on time to fall or holding onto the rod without moving. Three trials were averaged for a final score per mouse per session. Mice had pre-injury rotarod training once a day, consecutively for 5 days before injury. Post-injury testing was done at 1, 2, 4, 6, and 8 weeks after unilateral photothrombosis. Post-injury end time of each animal was normalized to its own pre-injury end time, with the pre-injury baseline performance represented as 1 on quantification. To evaluate skilled forelimb stepping, regular horizontal ladder was used. Mice were filmed (60 fps, interlaced, and rendered to 30 fps playback) walking across an 81 cm horizontal ladder with 1.5 cm regularly spaced rungs (Farr et al., 2006). Paw placement of each forelimb was video-scored and defined as follows: correct placement (score of 6); partial placement (score of 5); misses (scores of 0–4). Each animal had three trials per day of testing. Number of steps from each placement type was expressed as a percentage of total steps. Three trials were averaged per mouse per session. Mice had pre-injury training three days before injury. Post-injury testing was done at 1, 2, 4, 6, and 8 weeks after unilateral photothrombosis.

Statistical analyses. One-way or two-way ANOVA analysis was applied depending on the experimental design (see Figure Legends), using GraphPad Prism software. Post-hoc comparisons were performed only when a main effect showed statistical significance. Bonferroni correction was used to adjust p value of multiple comparisons. All graphs show mean with SEM. N represents the number of mice per genotype or treatment group. See figure legends for exact n and p values, and statistical test applied.