Introduction

Traumatic CNS injuries often lead to irreversible impairments, such as blindness or loss of motor and sensory function, due to the inability of severed axons to regrow. This regenerative failure is mainly caused by an insufficient intrinsic growth capacity of mature CNS neurons and the inhibitory environment for axonal growth cones at the injury site [15, 45, 54].

Although there is currently no clinically applicable treatment available to human patients, various strategies exist to promote axon regeneration in experimental models of CNS injury [15, 54, 64]. For example, targeting the inhibitory environment at the lesion site by neutralizing myelin-derived factors or chondroitin sulfate proteoglycans (CSPGs) led to axonal regeneration and, in some cases, induced functional recovery after incomplete spinal cord lesions[5, 7, 34, 47]. On the other hand, it became evident that not only disinhibition but also strategies to stimulate cell-intrinsic signaling pathways are essential to promote robust axon growth. Among these, lens injury-induced expression and release of the IL-6-type cytokines ciliary neurotrophic factor (CNTF) and leukemia inhibitory factor (LIF) from retinal astrocytes transforms retinal ganglion cells (RGCs) into a regenerative state and enables moderate axon growth in the injured optic nerve [16, 38, 49, 63]. The designer cytokine hyper-IL-6 (hIL-6), a fusion protein of IL-6 and the soluble IL-6 receptor α (IL-6Rα), can enormously increase these effects. In contrast to natural cytokines, hIL-6 directly binds to the gp130 receptor and does not rely on previous interaction with cytokine-specific alpha-receptors [18, 28, 35, 60]. This enables a more potent activation of downstream signaling pathways, such as the Janus kinase/signal transducer and activator of transcription 3 (JAK/STAT3) and phosphoinositide 3-kinase/protein kinase B (PI3K/AKT) pathways [35]. As a result, AAV-mediated hIL-6 expression in RGCs elicits a more vigorous optic nerve regeneration over long distances than most other single treatments [14, 35]. Similarly, hIL-6 expression in corticospinal neurons simultaneously promotes axon regeneration of multiple descending tracts and partially restores motor function after a complete spinal cord crush [40, 60].

Another way to stimulate CNS axon regeneration is to directly activate the PI3K/AKT pathway by knockout phosphatase and tensin homolog (PTEN^-/-) [43, 53, 66]. This approach stimulated axon regeneration in the corticospinal tract even when applied one year after injury and induced functional recovery of voluntary skilled movements after incomplete spinal cord injury or pyramidotomy [12, 13, 42, 43]. PTEN^-/- or hIL-6-mediated AKT activation causes phosphorylation of several substrates, including inactivating phosphorylation of glycogen synthase kinase 3 (GSK3) [58, 59]. The specific GSK3β knockout (GSK3β^-/-) in RGCs facilitates optic nerve regeneration by reducing collapsin response mediator protein 2 (CRMP2) phosphorylation in axons, while constitutively active GSK3 (GSK3^S/A) compromises PTEN^-/- and cytokineinduced regeneration, underlining the relevance of the AKT/GSK3/CRMP2 pathway for CNS regeneration [36, 37].

Besides CRMP2, GSK3 phosphorylates microtubule-associated protein 1B (MAP1B). Phosphorylated MAP1B keeps microtubules in a dynamic state by blocking the removal of the C-terminal tyrosine from α-tubulin subunits (detyrosination) in axonal growth cones [20, 22–24, 46, 52]. In addition, inhibition of tubulin detyrosination accelerates peripheral nerve regeneration [20]. This was achieved by local or systemic application of the sesquiterpene lactone parthenolide, an inhibitor of VASH1/2, which mediates detyrosination [1, 19, 21, 51].

The current study addressed whether parthenolide affects axon regeneration in the CNS. It shows that it markedly accelerates neurite outgrowth of primary CNS neurons by reducing tubulin detyrosination in axonal tips. Moreover, intraperitoneal application of its prodrug dimethylamino-parthenolide (DMAPT) facilitates axon regeneration after optic nerve and spinal cord injuries in vivo. Furthermore, hIL-6 and PTEN^-/- were found to increase tubulin detyrosination. At the same time, parthenolide antagonized this effect, thereby further enhancing hIL-6-induced optic nerve and spinal cord regeneration and functional recovery of hindlimb locomotion after severe spinal cord injury. The effects of parthenolide on detyrosination and neurite outgrowth were reproducible in human adult RGCs, suggesting it is a promising drug candidate in patients suffering CNS injuries. In particular, parthenolide is an ideal cotreatment to accelerate axon regeneration and functional recovery in combination with other treatment strategies.

Materials and Methods

Mouse strains

Male and female C57BL/6 and 129/Ola mice were used for all experiments. All mice were housed under the same conditions for at least 10 days before the experiments and maintained on a 12 h light/dark cycle with ad libitum access to food and water. All experimental procedures were approved by the local animal care committee (LANUV Recklinghausen) and were conducted in compliance with federal and state guidelines for animal experiments in Germany.

Human eyes

The use of human eyes was approved by the ethics committee of Heinrich Heine University Düsseldorf and performed by the ethical standards as laid down in the 1964 Declaration of Helsinki and its later amendments or comparable ethical standards.

Surgical procedures

Optic nerve crush and intravitreal injections

Adult C57BL/6 mice were anesthetized by intraperitoneal injections of ketamine (120 mg/kg) and xylazine (16 mg/kg). For western blot and cell culture experiments, RGCs were transduced 3 weeks before tissue isolation by intravitreal injection of 2 µl AAV2-hIL-6. For in vivo regeneration experiments, the left optic nerve was intraorbitally crushed 1 mm behind the eyeball for 10 s using a jeweler’s forceps (Hermle), as described previously (Leibinger et al. 2009). Mice received intravitreal injections of 2 µl AAV2-hIL-6 directly after optic nerve crush. Two days before tissue isolation, regenerating axons were anterogradely labeled by intravitreal injection of 2 µl Alexa Fluor 555-conjugated cholera toxin β subunit (CTB, 0.5% in PBS; Invitrogen).

Intracortical viral injections

Adult 129/Ola mice were placed in a stereotactic frame and continuously supplied with 1.5 – 2% isoflurane for anesthesia. A midline incision was made over the skull to open the skin and reveal bregma. A micro drill with a 0.5 mm bit was used to open a 2 x 2 mm window on each side of the skull to expose the sensorimotor cortex. AAV2-hIL-6 was injected into the cortex layer V to transduce corticospinal neurons. Four injections of 500 nl each were applied into the left cortex through a glass capillary attached to a nanoliter injector (Nanoject III, Drummond) at the following coordinates: 1.5 mm lateral, 0.6 mm deep, and 0.5 mm anterior; 0.0 mm, 0.5 mm, and 1.0 mm posterior to bregma. At each injection site, five pulses of 100 nl were applied at a rate of 10 nl/s. A pause of 10 s was made after each pulse to allow the distribution of the virus solution. The needle was left in place for 1 min before moving to the next site.

Complete spinal cord crush

A complete spinal cord crush was performed as described previously [40, 61]. Adult 129/Ola mice were continuously supplied with 1.5 - 2 % isoflurane for anesthesia. A midline incision of ∼1.5 cm was made over the thoracic vertebrae. The fat and muscle tissue were cleared from thoracic vertebrae 7 and 8 (T7, T8). While holding onto T7 with forceps, a laminectomy was performed at T8 to expose the spinal cord. The complete spinal cord was then crushed for 2 s with forceps that had been filed to a width of 0.15 mm for the last 5 mm of the tips to generate a homogeneously thin lesion site [66]. After surgery, muscle layers were sutured with 5.0 resorbable sutures, and the skin was secured with wound clips.

DHT injection

To deplete serotonergic inputs to the spinal cord, mice received bilateral intracerebroventricular injections of 30 μg of the serotonin neurotoxin 5, 7-dihydroxytryptamine (DHT, Biomol). DHT was dissolved in 0.5 μl of 0.2 % ascorbic acid in saline, as described previously [40]. To this end, mice were anesthetized and fixed under a stereotactic frame described above for intracortical injections. In addition, the tip of a glass micropipette attached to a nanoject III (Drummond) was positioned at the following coordinates: 0.6 mm posterior, 1.6 mm lateral to bregma, and 2 mm deep from the cortical surface. DHT was injected at the same rate as AAV2 injections, and the pipette was left for 1 minute before the withdrawal. Thirty minutes before the DHT injection, the monoamine uptake inhibitor desipramine (Sigma) was intraperitoneally administered at 25 mg/kg to prevent DHT uptake into noradrenergic neurons.

Preparation of AAV2

AAV plasmids carrying cDNA for hIL-6 downstream of a CMV promoter were co-transfected with pAAV-RC (Stratagene) encoding the AAV genes rep and cap, and the helper plasmid (Stratagene) encoding E24, E4, and VA into AAV-293 cells (Stratagene) for recombinant AAV2 generation. In addition, purifying virus particles was performed as described previously [35, 40, 65].

Western blot assays

For protein lysate preparation, optic nerves from adult C57BL/6 mice, either untreated or 3 weeks after intraperitoneal AAV2-hIL-6 injection, were dissected and homogenized in lysis buffer (20 mM Tris/HCl pH 7.5, 10 mM KCl, 250 mM sucrose, 10 mM NaF, 1 mM DTT, 0.1 mM Na3VO4, 1% Triton X-100, 0.1% SDS) with protease inhibitors (Calbiochem, Darmstadt, Germany) and phosphatase inhibitors (Roche, Basel, Switzerland) using 5 sonication pulses at 40% power (Bandelin Sonoplus, Berlin, Germany). Lysates were cleared by centrifugation in an Eppendorf tabletop centrifuge at 4150 × g for 10 min at 4°C. Proteins were separated by sodium-dodecyl-sulfate-polyacrylamide gel electrophoresis (SDS-PAGE) using Mini TGX gels (10%, Bio-Rad, Hercules, USA) according to standard protocols and transferred to nitrocellulose membranes (0.2 μm, Bio-Rad, Hercules, USA). Blots were blocked in 5% dried milk in phosphate-buffered saline with 0.05% Tween-20 (PBS-T) (Sigma, St Louis, USA) and incubated with antibodies against S21-phosphorylated GSK3α (1:2000; Cell Signaling Technology, RRID: AB_11166651), S9-phosphorylated GSK3β (1:1000; Cell Signaling Technology, RRID:AB_2115196), total GSK3α/β (1:1000; Cell Signaling Technology, RRID: AB_10547140), T514-phosphorylated CRMP2 (1:1000; Abcam, RRID:AB_942229), total CRMP2 (1:1000; Cell Signaling Technology, RRID:AB_2094339), detyrosinated tubulin (1:1000; Millipore, RRID:AB_177350) and βIII-tubulin (1:2000; BioLegend, RRID:AB_2313773) at 4°C overnight. All antibodies were diluted in PBS-T containing 5% BSA (Sigma). A conformation-specific HRP-conjugated anti-rabbit IgG (1:4000; Cell Signaling Technology) was used as a secondary antibody. Antigen-antibody complexes were visualized using enhanced chemiluminescence substrate (Bio-Rad) on a FluorChem E detection system (ProteinSimple). Western blots were repeated at least three times to verify the results. Band intensities were quantified relative to respective loading controls by using ImageJ software.

Dissociated mouse retinal cell cultures

Adult C57BL/6 mice were used for all cell culture experiments. Some experiments included animals that had received intravitreal AAV2-hIL-6 injections 3 weeks before dissociation. Retinal cultures were prepared as described previously [25]. In brief, retinae were dissected from the eyecups and digested in Dulbecco’s Modified Eagle medium (DMEM; Thermo Fisher) containing papain (10 U/ml; Worthington) and L-cysteine (0.2 μg/ml; Sigma) at 37°C for 30 minutes. After digestion, retinae were washed with DMEM, triturated, and centrifuged in 50 ml DMEM (500 g, 7 minutes). Cell pellets were resuspended in a medium containing B27 supplement (1:50; Invitrogen) and penicillin/streptomycin (1:50; Merck) and passed through a cell strainer (40 µm; Greiner Bio-One). Cells were treated with parthenolide (Sigma; 0.25-5.0 nM in dimethyl sulfoxide), recombinant rat CNTF (Peprotech; 200 ng/ml), DMAPT (Abcam; 0.25-5.0 nM in dimethyl sulfoxide), or a combination of parthenolide or CNTF. 300 µl cell suspension was added to each well on 4-well plates (Nunc), which were coated with poly-D-lysine (0.1 mg/ml, molecular weight 70,000-150,000 Da; Sigma) and laminin (20 µg/ml; Sigma). Retinal cells were cultured at 37 °C and 5 % CO₂.

After 4 days in culture, retinal cells were fixed with 4 % paraformaldehyde (PFA; Sigma). Cells were stained with primary antibodies against βIII-tubulin (1:2000; BioLegend, RRID:AB_2313773), detyrosinated tubulin (1:2000; Millipore), and phosphorylated (Tyr705) STAT3 (1:200; Cell Signaling Technology, RRID:AB_2491009). Secondary antibodies included donkey-anti-mouse and anti-rabbit antibodies conjugated to Alexa Fluor 488 or 594 (1:1000; ThermoFisher).

For studying neurite growth, all RGC neurites were photographed under a fluorescent micro-scope (200x; Axio Observer.D1, Zeiss), and neurite length was determined using ImageJ software. To evaluate microtubule detyrosination, at least 30 axonal tips per well were photo- graphed (400x; Axio Observer.D1, Zeiss) and analyzed using ImageJ software. Axon tips were defined as the last 15 µm of βIII-tubulin positive neurites. For the quantification of STAT3 phosphorylation, at least 30 RGCs per well were evaluated (400x; Axio Observer.D1, Zeiss).

Quantification of pCRMP2 staining intensity in murine RGC cultures

To quantify pCRMP2 immunofluorescence, retinal cells were cultured in either 0.5 nM parthenolide or the vehicle. After 4 d in culture, cells were fixed and immunostained against βIII-tubulin (1:1,000; BioLegend; RRID:AB_2313773) and phospho(T514)-CRMP2 (1:2,000; Abcam; RRID:AB_942229). The staining intensity of pCRMP2 in βIII-tubulin–positive RGCs were analyzed using ImageJ software and calculated according to the following formula: Corrected total cell fluorescence = integrated density − (area of selected cell × mean fluorescence of background reading). Data represent means ± SEM of three independent experiments. In each experiment, 80 RGCs per group were analyzed in two replicate wells.

Dissociated human retinal cell cultures

Primary human retinal cultures were generated in two independent experiments with retinae isolated from two adult human patients’ enucleated eyes.

The retinae were dissected from the eyecups, and the vitreous body was removed. Afterward, the retinae were cut into 8 pieces and digested separately, each in 10 ml Dulbecco’s Modified Eagle medium (DMEM; Thermo Fisher) containing papain (10 U/ml; Worthington) and L-cysteine (0.2 μg/ml; Sigma) at 37 °C for 45 minutes. After digestion, retinae were washed with DMEM, triturated, and centrifuged in 50 ml DMEM (900 g, 5 minutes). Cell pellets were resuspended in a medium containing B27 supplement (1:50; Invitrogen) and penicillin/streptomycin (1:50; Merck) and passed through a cell strainer (40 µm; Greiner Bio-One). Retinal cultures were incubated with parthenolide (Sigma; 0.25-5.0 nM in dimethyl sulfoxide) or the vehicle. Some groups received recombinant rat CNTF (Peprotech; 200 ng/ml), either solely or in combination with parthenolide. Cells were cultured at 37 °C and 5 % CO₂ on 4-well plates (Nunc) (300 µl/well; 4 wells/group), which were coated with poly-D-lysine (0.1 mg/ml, molecular weight 70,000-150,000 Da; Sigma). After 4 days in culture, retinal cells were fixed with 4 % paraformaldehyde (PFA; Sigma) and stained with primary antibodies against βIII-tubulin (1:2000; Bio-Legend) and detyrosinated tubulin (1:2000; Millipore). Secondary antibodies included donkey-anti-mouse and anti-rabbit antibodies conjugated to Alexa Fluor 488 or 594 (1:1000; ThermoFisher). For studying neurite growth, and the percentage of detyrosinated axon tips, cultures were analyzed in the same way as described above for murine RGCs.

Intraperitoneal injections of DMAPT

Due to its high oral bioavailability of ∼70% [27] and the ability to cross the blood-brain barrier [29], DMAPT can be delivered to the CNS via a systemic route of administration. Mice were given daily intraperitoneal injections of either DMAPT (0.2, 2, or 20 µg/kg; Abcam; dissolved in DMSO and further diluted in sterile PBS) or vehicle (DMSO diluted in sterile PBS). All spinal cord regeneration experiments were performed using the 2 µg/kg dosage. Treatment started directly after the optic nerve or spinal cord crush and was continued until the day of perfusion. Intraperitoneal injections were performed under blinded conditions.

Basso Mouse Scale (BMS)

We scored mice according to the Basso Mouse Scale to assess locomotor behavior after spinal cord injury [3]. To this end, each mouse was placed separately into a round open field of 1 m in diameter and observed by two testers for 4 minutes. Scoring was based on different parameters, including ankle movement, paw placement, stepping pattern, coordination, trunk instability, and tail position, with a minimum score of 0 (no movement) to a maximum of 9 (normal locomotion). Before testing, mice were acclimated to the open field environment. BMS tests were performed before the injury or 1, 3, and 7 days after the injury, and then weekly. As quality criteria for lesion completeness, we used a BMS score of 0 on day 1 after injury and the absence of BDA or serotonin staining in transverse spinal cord sections (as described below). Mice that did not meet these criteria were excluded from the experiment. BMS tests were performed under blinded conditions.

Immunohistochemistry

Animals were anesthetized and transcardially perfused with cold PBS, followed by 4 % PFA in PBS. Brains, spinal cords, and eyes with attached optic nerves were isolated, post-fixed overnight in 4 % PFA at 4 °C, and subsequently transferred to 30 % sucrose for at least 4 h for eyes and optic nerves or 5 days for brains and spinal cords.

Tissues were embedded in the KP-cryo compound (Klinipath) and frozen at -20 °C (retina, optic nerve, spinal cord) or on dry ice (brain). Sections were cut using a CM3050 S cryostat (Leica Biosystems). Optic nerves were cut into longitudinal sections (14 µm). For spinal cords of mice subjected to T8 crush, a segment from 3 mm rostral to 8 mm caudal from the injury site was cut into sagittal sections (20 µm). Spinal cord tissue, just rostral and caudal from this segment, was used for transverse sections (20 µm) to evaluate lesion completeness by identifying potentially spared axons. Tissue from the medullary pyramids ∼1 mm above the pyramidal decussation was cut into coronal sections (20 µm). Sections were thaw-mounted onto charged glass slides (Superfrost Plus, VWR) and stored at -20 °C until further use.

The retinal, spinal cord and brain sections were thawed, washed with PBS, and permeabilized for 10 min in methanol (Sigma). Sections were stained with primary antibodies against βIII-tubulin (1:1000; BioLegend, RRID: AB_10063408), phosphorylated STAT3 (1:200; Cell Signaling Technology, RRID:AB_2491009), serotonin (1:5000; Immunostar, RRID:AB_572262, RRID:AB_572263), GFP (1:500; Novus Biologicals, RRID:AB_10128178), NeuN (1:500; Abcam, RRID:AB_2532109), GFAP (1:500; Abcam, RRID:AB_305808), or CS-56 (1:70,000; Sigma, RRID:AB_476879). Secondary antibodies included donkey-anti-rabbit, anti-goat, and anti-mouse antibodies conjugated to Alexa Fluor 488, 594 (1:1000; ThermoFisher), or 405 (1:500; Jackson ImmunoResearch).

Retinal wholemounts were prepared without prior perfusion, fixed in 4 % PFA for 30 min, and permeabilized in 2 % Triton X-100 in PBS for 1 h. RGCs were stained using a primary antibody against βIII-tubulin (1:1000; BioLegend) and an anti-mouse secondary antibody conjugated to Alexa Fluor 488 (1:1000; ThermoFisher).

Sections and wholemounts were coverslipped with Mowiol (Sigma), and pictures were taken using a slide scanner (VS120; Olympus) or a confocal laser scanning microscope (SP8; Leica). Tissue analysis was performed under blinded conditions.

Quantification of regenerated axons in the optic nerve

Pictures of 6 longitudinal sections per animal were taken using a slide scanner (VS120; Olympus). The number of CTB-labelled axons extending 0.25, 0.5, and 0.75 mm (for DMAPT experiments) or 0.5, 1, 1.5, 2, and 2.5mm beyond the lesion site (for hIL-6 + DMAPT experiments) was quantified and normalized to the width of the optic nerve at the respective measuring point. Each experimental group included 6-7 mice. Eyes with lens injury were excluded from the experiment.

Quantification of RGCs in retinal wholemounts

Retinal wholemounts were divided into four quadrants. In each quadrant, 4 non-overlapping pictures were taken using a fluorescent microscope (400x; Axio Observer.D1, Zeiss), proceeding from the center to the periphery. The number of βIII-tubulin positive RGCs was determined and normalized to an area of 1 mm². Values were averaged per retina and across all animals. Each experimental group included 5-9 retinae. Eyes with lens injury were excluded from the experiment.

Quantification of RpST axon regeneration

Sagittal and coronal spinal cord sections were stained against serotonin and imaged with a slide scanner (20x; VS120; Olympus). A grid with lines spaced every 500 µm was aligned to each image to count the number of serotonin-positive RpST axons at 500 µm intervals from 0 to 8 mm caudal from the lesion site. In addition, every other spinal cord section was analyzed, and the number of serotonin-positive axons was extrapolated to the total number of sections per spinal cord. Animals with spared serotonin-positive axons in transverse spinal cord sections > 8 mm distal from the lesion site were excluded from the experiment. Each experimental group included 7-10 animals.

Quantification of spinal cord lesion sizes and CSPG staining

To evaluate the lesion size after SCC, coronal spinal cord sections containing the T8 lesion site were stained against GFAP, and pictures were taken using a fluorescence microscope (Olympus VS120). The lesion site, defined by the GFAP-positive astrocyte scar border, was traced and measured using ImageJ to obtain the total area. Furthermore, the width at the lesion center and both lateral edges were measured and averaged to determine the lesion width. The same sections were co-stained against CS-56 to visualize CSPGs at the injury site. Using ImageJ, the total area occupied by CSPGs was measured by tracing the CS-56 staining. Lesion sizes and CSPG staining was analyzed in three sections per animal.

Statistics

Significances of intergroup differences were evaluated using Student’s t-test analysis of variance (ANOVA) followed by Tukey or Holm-Sidak post hoc tests using the Sigma STAT3.1 software (Systat Software).

Results

Parthenolide and CNTF synergistically promote neurite growth of adult RGCs

In contrast to adult RGCs, sensory neurons usually show spontaneous neurite growth [35, 39, 41, 49]. Moreover, parthenolide-induced inhibition of tubulin detyrosination in axonal growth cones markedly accelerates the regeneration of sensory neurons in culture and in vivo [20, 22]. To investigate whether parthenolide has similar effects on CNS neurons, we cultured murine RGCs in the presence of increasing compound concentrations for 4 days. As in sensory neurons [20], parthenolide significantly enhanced neurite extension concentration-dependently with maximum effects at 0.5 and 1 nM (Fig. 1 A, B). Expectedly RGC survival remained unaffected (Fig. 1 C) since RGCs do not undergo axotomy-induced cell death as early as 4 days after culturing. Moreover, cotreatment of parthenolide with CNTF showed a significant additive effect (Fig. 1 D, E), resulting in a 3.5-fold higher average neurite length than vehicle-treated controls (Fig. 1 D, E). Although parthenolide reportedly reduces JAK/STAT3 signaling at micromolar concentrations [55, 56], the nanomolar concentrations used in our experiments did not alter CNTF-induced STAT3 phosphorylation in RGCs (Fig. 1 F, G). Conversely, CNTF did not measurably affect parthenolide’s inhibitory effect on microtubule detyrosination in axon tips (Fig. 1 H, I). We also tested parthenolide on human RGCs. To this end, we prepared and cultured primary RGCs from adult human eyes. Parthenolide and CNTF significantly enhanced the average neurite length at similar concentrations as in mouse RGCs, while cotreatment enhanced neurite growth sixfold compared to controls after 4 days (Fig. 1 J, K). Parthenolide also reduced microtubule detyrosination in axonal growth cones (Fig.1 M, N) without affecting the number of RGCs per well (Fig. 1 L).

Fig. 1:

Parthenolide counteracts hIL-6 and PTEN^-/--induced microtubule detyrosination

Due to its direct binding of the signal-transducing gp130 receptor, hIL-6 promotes neurite growth of RGCs more efficaciously than CNTF and, contrary to CNTF, also measurably induces PI3K/AKT signaling in these neurons [35]. However, AKT inactivates GSK3 (Fig. 2 A), which promotes CRMP2 activity (Leibinger et al., 2017, 2019). While enhanced CRMP2 activity is favorable for axon growth [36, 44], GSK3 inhibition might also elevate tubulin detyrosination via reduced MAP1B activation (Fig. 2 A). However, as shown in the previous experiments (Fig. 1), reducing microtubule detyrosination promotes neurite growth, suggesting that this effect of AKT/GSK3 might limit optimal axonal growth. To test these possibilities, we intravitreally injected AAV2-hIL-6 into mice 3 weeks before tissue isolation. Subsequent western blot analysis of respective optic nerve lysates indeed revealed increased inhibitory phosphorylation of both GSK3 isoforms (pGSK3α, pGSK3β) by hIL-6 compared to untreated controls, without affecting total GSK3 expression (Fig. 2 B-E). Moreover, while hIL-6 reduced inhibitory phosphorylation of CRMP2 at threonine 514 (T514) without changing total CRMP2 levels (Fig. 2 B, F, G), it expectedly elevated tubulin detyrosination (Fig. 2 B, H). Since parthenolide specifically acts on tubulin detyrosination without influencing CRMP2 activity (supplementary Fig. 1 A, B), it should be suitable to counteract an hIL-6-mediated increase in axonal tubulin detyrosination. To test this hypothesis, we cultured RGCs from AAV-hIL-6-treated mice or respective controls for 4 days without or presence of different parthenolide concentrations. Confirmingly, microtubule detyrosination in axonal tips was significantly higher in hIL-6-treated cultures than in controls, and parthenolide cotreatment reversed this effect (Fig. 2 I, J). Furthermore, the increased detyrosination in hIL-6-treated neurons also affected the effective parthenolide concentration on neurite extension in these cultures. Compared to the control cultures, where parthenolide showed the most substantial increase between 0.5 and 1 nM, followed by a decrease at 5 nM (Fig. 1 A, B), neurite growth of hIL-6-stimulated RGCs was increased even at higher parthenolide concentrations (up to 5 nM) (Fig. 2 K, L). The hIL-6-induced STAT3 phosphorylation remained unaffected by any parthenolide concentration applied (supplementary Fig. 2 A, B). AKT is also activated by PTEN^-/-, another regeneration-promoting approach [4, 37, 43, 53]. To investigate microtubule detyrosination in PTEN^-/- RGC neurites, we prepared RGC cultures from PTEN-floxed mice after intravitreal AAV2-Cre injection. As expected, PTEN^-/- enhanced tubulin detyrosination in axon tips (Fig. 3 A, B, C). This effect was counteracted by parthenolide, which further improved neurite outgrowth of PTEN^-/--RGCs concentration-dependently (Fig. 3 B, C, D, E). Moreover, activation of the mTOR pathway, a significant contributor to the PTEN^-/--mediated effects on axon growth, was not compromised by parthenolide, reflected by unaltered activation of ribosomal protein S6 in these RGC cultures (supplementary Fig. 2 C, D). Thus, parthenolide counteracts AKT/GSK3-induced microtubule detyrosination and synergistically enhances hIL-6, or PTEN^-/--stimulated neurite growth.

Fig. 2:

Fig. 3:

Dimethylamino-parthenolide accelerates optic nerve regeneration

To test whether parthenolide can accelerate CNS regeneration in vivo, we used its water-soluble prodrug dimethylamino-parthenolide (DMAPT), which can cross the blood-brain barrier [27, 29, 50]. To this end, adult mice were subjected to an optic nerve crush (ONC) and received daily intraperitoneal injections of either vehicle or DMAPT (2 µg/kg) (Fig. 4 A). The number of CTB-labeled, regenerated axons was quantified in optic nerve sections 2 weeks post-injury. Compared to vehicle-treated controls, more axons reached distances of up to 0.5 mm distal to the lesion site after DMAPT application, indicating a slight but significant effect on regeneration (Fig. 4 B-E). At the same time, the RGC survival was not affected (Fig. 4 F, G). We then repeated these experiments in combination with intravitreally applied AAV2-hIL-6 (Fig. 4 H). Two weeks after injury, retinal cross-sections revealed robust phosphorylation of STAT3 in AAV2-hIL-6-treated RGCs, which remained unaffected by DMAPT treatment (Fig. 4 I). As reported previously [35], hIL-6 treatment alone enabled substantial regeneration with CTB-labeled axons reaching distances of up to 1.5 mm 2 weeks after injury (Fig. 4 J-O). Notably, additional DMAPT application at 2 µg/kg increased the number of axons at 1.5 mm fivefold, with some axons reaching distances of 2.5 mm within the same time after injury (Fig. 4 J, K, N, O). As observed in RGC cultures, DMAPT acted dose-dependently, with less pronounced effects at lower (0.2 µg/kg) or higher (20 µg/kg) dosages (Supplementary Fig. 3 A, B). Thus, the systemic application of DMAPT sufficiently enhances and accelerates the effect of hIL-6 on optic nerve regeneration.

Fig. 4:

Systemic DMAPT application accelerates raphespinal regeneration after spinal cord crush

Hyper-IL-6-mediated axonal regeneration of raphespinal fibers enables functional recovery after a severe spinal cord crush (SCC) [40]. Therefore, we investigated whether systemic DMAPT treatment also affects raphespinal regeneration or functional recovery. To this end, we performed a complete SCC at the thoracic vertebra 8 (T8) level, followed by daily intraperitoneal injections of either vehicle or DMAPT at 2 µg/kg (Fig. 5 A). Hindlimb movement was analyzed using the Basso Mouse Scale (BMS) [3] over 8 weeks after injury. One day after SCC, BMS scores dropped to 0, indicating the absence of hindlimb movement (Fig. 5 B, C). After 2 weeks, most vehicle-treated animals recovered from the injury-induced spinal shock and reestablished typical extensive ankle movement (BMS: 2) (Fig. 5 B,C; supplementary video 1). In contrast, DMAPT treatment slightly but significantly improved functional recovery. Most of these animals showed plantar hind paw placement as early as 2 weeks after injury, occasionally accompanied by bodyweight support (BMS: 3) (Fig. 5 B,C; supplementary video 2). For this reason, the spinal cord tissue of some animals was analyzed for serotonergic fiber regeneration at this early time. The absence of serotonergic fibers at ∼10 mm distal to the lesion site verified the lesion completeness in each animal included in this study to rule out the presence of spared raphespinal axons (Fig. 5 D, supplementary Fig. 4 A-H). Vehicle-treated control animals showed only typical short-distance sprouting of serotonergic axons less than 2 mm distal to the lesion site (Fig. 5 E-G, K). At the same time, DMAPT treatment increased the number and length of axons, with the longest axons reaching 4 mm (Fig. 5 E, H-K). When analyzed 8 weeks after SCC (Fig. 6 A), DMAPT further accelerated raphespinal regeneration, with the longest axons reaching >7 mm (Fig. 6 B-M). However, despite the more pronounced anatomical regeneration, the BMS score was only slightly further increased compared to 14 days after SCC (Fig. 6 N). In addition, the lesion size and CSPG expression at the glial scar remained unaffected by DMAPT treatment (supplementary Fig. 5 A-D). Thus, systemic DMAPT application accelerates raphespinal regeneration and some functional recovery after a severe spinal cord injury.

Figure 5:

Figure 6:

Raphespinal regeneration contributes to DMAPT-mediated functional recovery

We prepared additional mice to investigate whether the DMAPT-mediated effects on functional recovery depend on serotonergic fiber regeneration, as previously shown for hIL-6 [40]. To this end, we performed SCC with either vehicle or DMAPT treatment. After the DMAPT group had reached maximum functional recovery, we injected the serotonergic neurotoxin 5,7-dihydroxytryptamine (DHT), which depletes raphe neurons as early as one day after injection, as shown previously [32, 40]. Expectedly, DHT treatment reduced the number of serotonergic brainstem raphe neurons and their axons without affecting other neuronal populations (Fig. 7 A-J; supplementary Fig. 6 A, B). This did not affect BMS scores of vehicle-treated controls (Fig. 7 K, L; supplementary video 3) but immediately abolished DMAPT-induced gain of motor function (Fig. 7 K, L; supplementary video 4), suggesting an essential contribution of regenerated serotonergic axons.

Figure 7:

DMAPT accelerates hIL-6-induced functional recovery after complete spinal cord crush

Intracortical injection of AAV2-hIL-6 promotes axon regeneration and functional recovery after SCC [40]. To test for synergistic effects, as seen in the optic nerve (Fig. 4), we combined hIL-6 and DMAPT treatment. To this end, we subjected wild-type mice to SCC and injected AAV2-hIL-6 into the sensorimotor cortex. Mice received daily intraperitoneal injections of either vehicle or DMAPT for 8 weeks (Fig. 8 A). As determined by immunostaining, AAV2-hIL-6-mediated transduction of layer V cortical motorneurons induced robust phosphorylation of STAT3 (supplementary Fig. 7 A-E). Consistent with [40], intracortical AAV2-hIL-6 also transneuronally stimulated medullary raphe neurons (supplementary Fig. 7 I-M). At the same time, DMAPT treatment did not affect STAT3 phosphorylation in cortical (supplementary Fig. 7 B, F, G, H) or raphespinal (supplementary Fig. 7 J, N, O, P) neurons at applied doses used in this study. Consistent with previous results [40], AAV2-hIL-6 treatment restored plantar hind paw placement and stepping with lift-off, followed by forwarding limb advancement and re-establishment of bodyweight support, reaching a maximum BMS score of 4-5 (Fig. 8 B-C). Interestingly, combinatorial treatment with DMAPT accelerated functional recovery, allowing plantar steps in most animals already 14 d after SCC. These mice reached the maximum effect already at 21 days after injury, while animals solely treated with hIL-6 needed 42 days (Fig. 8 B-C; supplementary video 5, 6). Moreover, additional DMAPT treatment also reduced the variance, with fewer weakly responding animals, resulting in a higher average final score of 4 compared to 3.5 in the hIL-6 control group.

Figure 8:

DMAPT affects RpST fiber regeneration after spinal cord crush

Regarding fiber regeneration, additional DMAPT treatment further enhanced the hIL6-induced regeneration of serotonergic fibers 8 weeks after SCC (Fig. 8 D-K). Compared to AAV2-hIL-6 therapy, which led to robust axonal regeneration up to 7 mm distal from the lesion site (comparable to [40]) (Fig. 8, D-G, K), additional DMAPT treatment significantly increased the number of serotonergic axons at this distance, with some axons reaching even 8 mm (Fig. 8 D, H-K).

Discussion

Here we demonstrate that the systemic application of DMAPT facilitates axon regeneration in the injured optic nerve and accelerates functional recovery after severe spinal cord injury. Parthenolide’s effects on CNS regeneration in vivo were confirmed in cell culture using primary RGCs from adult mice and humans. As on sensory neurons (preprint server) [21], parthenolide reduces microtubule detyrosination in axonal tips, which increases their dynamics and promotes axon extension [20]. As for sensory neurons, this effect is likely mediated by the inhibition of VASH1/VASH2, the enzymes catalyzing the detyrosination [21]. Consistently, the effective concentration range and doses of parthenolide/DMAPT to accelerate axon regeneration of RGC, serotonergic fibers in the spinal cord, and functional recovery was similar to sensory neurons and regeneration in the injured peripheral nerve [21]. Parthenolide did not affect CRMP2 or STAT3 activity at the nanomolar concentrations used in this study, both critical for CNS axon regeneration [35, 37] and suggesting the reduction of microtubule detyrosination causes the axon growth-promoting effect.

Systemic DMAPT treatment significantly promotes axon extension in the optic nerve and regeneration of serotonergic fibers beyond the injury site. The selective elimination of serotonergic fibers by DHT [8, 11, 32] abolishes the recovered locomotion, demonstrating that the improved regeneration of serotonergic fibers is essential for recovery after SCC. These data are consistent with previously published data, showing a similar effect on AAV-hIL-6 mediated recovery [40]. Despite the fact that DMAPT-mediated functional improvement depends on the regeneration of serotonergic fibers, we cannot exclude the possibility that spinal tracts other than the raphespinal tract contributed as well. The fact that DMAPT improved axon regeneration in the optic nerve and the spinal cord also implies that sufficient amounts of the drug reached the axonal growth cones, where the drug affects the microtubule detyrosination [20]. It is likely that an oral application of DMAPT, which reportedly is >60% bioavailable [21, 27, 57], has similar effects on nerve regeneration, bolstering the potential as a drug to treat axonal CNS damage.

Another significant finding of the current study is that pro-regenerative hIL-6 and PTEN^-/- treatments induce microtubule detyrosination in RGC axons. Given that a parthenolide-mediated reduction of microtubule detyrosination promotes axon growth, these observations suggest that this effect of hIL-6 or PTEN^-/- compromises their full axon regeneration-promoting potential. hIL-6 and PTEN^-/- induced microtubule detyrosination is likely caused by inhibitory GSK3 phosphorylation via increased PI3K/AKT activity. As a result, MAP1B activity is decreased, and the axonal detyrosination rate is elevated [26, 30, 36]. However, CNTF does not affect detyrosination due to its lower AKT activation than hIL-6 [35, 48, 49]. At the same time, hIL-6 and PTEN^-/- mediated GSK3 inhibition confers disinhibitory effects toward myelin-associated molecules because less CRMP2 is inhibited [36, 37, 44]. Thus, the disinhibitory effect of GSK3 inhibition promotes regeneration, while the detyrosinating effect on microtubules rather compromises it. Therefore, by counteracting detyrosination, parthenolide/DMAPT cotreatment further increases the axon growth-promoting effect of hIL-6. Consistently, effective parthenolide concentrations reduce detyrosination levels in hIL-6-treated and PTEN^-/- RGCs, and combined treatment leads to more substantial neurite growth than each treatment alone. This notion is also supported by the fact that PTEN^-/- and hIL-6-stimulated neurons sustain parthenolide-induced additive elevation of axon growth even at higher parthenolide concentrations, which already diminish outgrowth on naïve or CNTF-stimulated RGCs. The level of detyrosinated tubulin at the maximal applied parthenolide concentration in hIL-6 treated RGCs was still similar to untreated control RGCs. The synergistic effects of DMAPT and hIL-6 were also reproduced in vivo, in the optic nerve and spinal cord. Thus, parthenolide/DMAPT neutralizes the hIL-6-mediated limiting effect of decreased microtubule dynamics and synergistically accelerates CNS regeneration.

Our findings also indicate that DMAPT mainly accelerates axon growth and has little effect on the intrinsic regenerative state of neurons. Therefore, the RGC regeneration-promoting effects of parthenolide on its own are only moderate but synergistically increase hIL-6-mediated axon extension. Thus, DMAPT enabled substantial regeneration of serotonergic RpST after complete SCC. Furthermore, the length of regenerating RpST axons after SCC further increased from 2 to 8 wpc, while the functional recovery was already almost maximal in the early phase. This observation might be based on the different possibilities of achieving active recovery. On the one hand, descending supraspinal axons might need to grow over long distances to reestablish functional connections within the lumbar central pattern generator (CPG), a network of neurons responsible for generating stereotyped hindlimb movement [10, 31, 62]. On the other hand, it is also possible that severed axons only need to overcome short distances of a few spinal segments and then connect to existing propriospinal neurons, which might serve as a relay station to the CPG [2, 9, 10]. An observation accounting for the latter possibility is increased axonal sprouting in the region of the lumbar CPG (∼6-7 mm distal to the lesion) at 8 wpc, whereas the substantial increase in functional recovery occurred already earlier (∼2-3 wpc) before the axons had reached the CPG. Thus, DMAPT-induced long-distance RpST regeneration might not be essential for its beneficial effects on functional recovery, whereas DHT experiments showed that RpST regeneration, in general, is crucial.

In conclusion, systemically applied DMAPT measurably accelerates CNS regeneration and functional recovery and overcomes the limiting effect of AKT/GSK3 on elevating microtubule detyrosination. Whether DMAPT treatment can also improve functional recovery in more clinically relevant injury models, such as contusion injuries, or even more effective in combination with other regeneration-promoting strategies [6, 17, 33] will be investigated in the future. Finally, the axon growth-promoting effects on primary adult human CNS neurons underline the potential of parthenolide/DMAPT as a possible future treatment strategy for CNS injury in humans.

Supporting information

Suppl. Video 1

Suppl. Video 2

Suppl. Video 3

Suppl. Video 4

Suppl. Video 5

Suppl. Video 6

Acknowledgements

We thank Dr. Günter Gisselmann, Anastasia Andreadaki, Christopher Brennsohn, Davina Stoesser, and Kessy Brzozowski for their technical support. The German Research Foundation supported this work.