The irreversibility of injuries to the mature CNS has been known for a long time. Already 3,500 years ago, Egyptian physicians accurately described human para- and tetraplegia and the serious prognosis associated with such conditions (20). This was shown at the cellular level at the beginning of the 20th century as Santiago Ramon y Cajal, a pioneer within neuroscience, discovered that CNS axons failed to regenerate (21). Cajal demonstrated that axons from PNS, in contrast to those of the CNS, readily grew out and could re-innervate their targets following axotomy. He attributed the insufficient regenerative capabilities of CNS axons to their surroundings and hypothesized that the axons would grow if they were influenced by the same environment as peripheral nerves. To confirm this, he and his pupil Tello performed an experiment where they anastomosed a transected sciatic nerve graft to the optic nerve in a rabbit (22). They were able to demonstrate that some of the normally growth-resistant optic axons had penetrated and extended into the graft of the sciatic nerve. Cajal and Tello concluded that the environment surrounding the CNS axons at least were partially responsible for the poor regenerative capacity. However, the validity of these experiments were later questioned, as it was pointed out that the apparent extending CNS axons could represent autonomic fibers from blood vessels (23). It was not until 1981 that the field advanced, when the experiment by Cajal and Tello was repeated in a modified version by Aguayo and David (24). Using a mouse model, they grafted an autologous sciatic nerve between the medulla oblongata and lower parts of the spinal cord. After 22–30 weeks, horseradish peroxidase staining methods showed that axons had crossed from the CNS through the PNS graft penetrating the CNS tissue at the other end, where further growth stopped. This provided evidence that the mature CNS environment indeed inhibit axon regeneration, in contrast to the growth permissive PNS.

The results sparked a search for biomolecules within the CNS responsible for preventing growth of injured axons. If one or more inhibiting factors could be identified, blocking their function should theoretically promote axonal growth. Focus was turned to myelinating cells, as it was shown that mature oligodendrocytes repelled axons and neurites in co-cultures in vitro (25). This observation led to the development of a monoclonal antibody against IN-1 on the oligodendrocyte surface, later known as Nogo-A. The blocking of Nogo-A allowed in vitro ingrowth of neurites into adult optic nerve explants (26). Another study demonstrated that the neutralization of Nogo-A promoted axon sprouting at the lesion site of the corticospinal tract transection in young rats (27).

Subsequent studies identified other oligodendroglial proteins inhibiting axonal growth, including myelin-associated glycoprotein (MAG) (28) and oligodendrocyte myelin glycoprotein (OMgp) (29). These proteins were shown to impair axonal regeneration via activation of the Ras homolog member A (RhoA) and subsequently Rho-associated protein kinase (ROCK) (Figure 1) (30). ROCK mediates axon growth inhibition through phosphorylation of central molecules related to formation of the cytoskeleton (31). The importance of the myelin components in inhibiting axonal growth in the CNS was, however, later questioned when deletion of NOGO, MAG and OMgp failed to enhance axonal regrowth after SCI in knock-out mice (32). Although single deletion of each factor led to increased sprouting of uninjured axons, there was no associated behavioural improvement and no synergistic effect of deleting all three, suggesting additional contributing mechanisms responsible for the failing regeneration of the mature CNS.

In concert with these findings, other inhibiting factors for axonal regeneration were identified, such as the glial scar (33). Following injuries to the CNS, reactive astrocytes form a gliotic scar at the lesion site. This process serves to isolate the inflammation and protect the surrounding tissue but comes at the cost of axonal regrowth. Proteoglycans, including chondroitin sulphate proteoglycans (CSPGs), are necessary for the formation of the scar. Similar to Nogo-A, MAG and OMgp, CSPGs prevent axonal growth via the same Rho/ROCK pathway (34). The inhibitive effect of CSPG was supported by studies demonstrating that the CSPG-digesting enzyme chondroitinase ABC leads to axon regrowth and functional improvement following CNS injury in animal models (35, 36). Although the glial scar forms a barrier to axonal regeneration, preventing its formation may also lead to increased lesion size and worse outcomes (37). Accordingly, it was shown that astrocytes can support axonal regeneration by the production of specific growth supportive types of CSPG (CSPG4 and CSPG5). These opposing findings highlight the heterogeneity of astrocytes and their impact on axonal regeneration.

Although it was clear that the environment of the CNS prevented axonal regrowth, there had to be additional unknown factors obstructing the process of regeneration. This was observed following the transplantation of fetal neurons into rodent models of SCI. After successful intralesional implantation, the cells survived up to 16 months in both neonatal and adult rats (38). Axons from the implanted fetal neurons crossed the entire length of the lesion and formed synapses with host neurons. The functionality of these connections were, however, not assessed, as the study was strictly histological and did not include neurological deficits as outcomes. Nevertheless, the findings suggested that immature neurons could overcome the inhibitory extrinsic environment, pointing at an additional intrinsic mechanism within mature neurons that, unlike fetal neurons, prevented axonal growth within the injured CNS. An intraneural switch seemed to exist that was turned off during development for the benefit of other functions than growth, such as synaptic activity. The intrinsic brakes of regeneration were illustrated by another sentinel experiment, which included an inflicted injury to the peripheral branch of dorsal root ganglion (DRG) neurons. DRG neurons are unique in the sense that they share the features of both PNS and CNS as the distal branch belongs to the PNS and the central branch to the CNS. While the peripheral axon of the DRG neuron under optimal circumstances regrow and re-innervate its target, the central branch fails to regenerate following axotomy. If, however, the distal branch is injured one week before the central branch, the central branch will also show signs of axonal regeneration. This phenomenon was named “the conditioning effect” and indicated that the intracellular modifications following peripheral nerve injury was sufficient to prime the CNS for axon regrowth (39, 40).

Subsequent studies showed that axonal lesions in the PNS typically induce a broad and coordinated cascade of gene expression changes promoting regeneration, a cascade that is absent in injuries to the CNS (Figure 1) (41–43). These genes were named “regeneration-associated genes” (RAGs). Due to the broad apparatus necessary to initiate axon regrowth and elongation, RAGs code for several different types of proteins, including metabolic enzymes, cytoskeletal proteins and adhesion molecules. The specific function of each RAG has been determined using knock-out models or pharmacological inhibition in cell cultures as well as animal models (44, 45). These experiments have also demonstrated that not all genes expressed after PNS injury promote regeneration. Some may also inhibit regeneration, such as SOCS3. This protein is a suppressor of cytokine signalling and its overexpression leads to decreased axon growth (46). The fact that not all genes promote regeneration, illustrate the broad spectrum of gene networks that are activated in response to PNS axonal injury (47). An important question is whether the complete genetic response is necessary to sustain axonal regeneration within CNS neurons or not.

Some transcription factors (TFs) are more important than others as they seem to regulate the activity of multiple RAGs. The deletion of these “master” TFs leads to a more substantial decrease in the regenerative response as compared to knockdown of individual RAGs (47). Importantly, the regenerative program orchestrated by the master TFs is only initiated following PNS injury, but not CNS injury (48). A study identified a total of 39 TFs as master regulators for around 400 signalling pathways in the regenerative response to peripheral sensory axonal injury (49). In later experiments, 23 additional master TFs were identified in rodent models of PNS injury (50). The master TFs are highly interconnected through multiple and parallel pathways, which contribute to the robustness of the response. Master TFs and their networks may thus represent potential targets for therapeutic modulation of axonal regeneration. Important master regulators include The Krüppel-like-factor (KLF) family, C-Jun, STAT3 and AKT. The use of multi-omics analyses has also brought new insights to uncover mediators of neural regeneration. Recently, a systems genomic approach showed that RE1-Silencing Transcription Factor (REST), also known as Neuron-Restrictive Silencer Factor (NRSF), is an important upstream suppressor of the pro-regenerative program in the CNS (51). In a mature mouse model of SCI and optic nerve crush, the deletion of REST led to upregulation of RAGs with improved axonal regeneration in both models. Another recent study applied multi-omics screening to identify the four TFs, ATF3, ATF3, C/EBPγ and SHOP/Ddit3 as central regulators of in vivo injury response following optic nerve crush in mice (52). Advances in sequencing technologies will likely continue to facilitate new discoveries to unveil the complex regulation of neural regeneration.

TFs initiate transcription of DNA by binding to their target sites. Epigenetic modifications of histones and DNA play an essential role ensuring chromatin accessibility for the TFs during axonal injury (53). Histone acetylation is the most studied mechanism during regeneration. Histones are acetylated by histone acetyltransferases (HATs) and deacetylated by histone deacetylases (HDACs). Generally, histone acetylation neutralizes the positive charge of lysine, which leads to increased accessibility, whereas deacetylation results in chromatin compaction and gene silencing. Both HATs and HDACs are involved in the neuroregenerative response. The HAT p300/CBP-associated factor (PCAF) is activated after PNS injury and modifies histones at key RAGs, increasing their accessibility (Figure 1) (54). In the absence of PCAF, there was no regeneration of CNS axons following a conditioning lesion. PCAF overexpression, on the other hand, promoted regeneration of sensory fibers in a mouse model of SCI.

With HDAC5, an opposite mechanism seems to promote regeneration. Following peripheral axotomy, the influx of calcium triggers exportation of HDAC5 out of the nucleus into the cytoplasm, which results in histone H3 acetylation and activation of a regenerative transcriptional program (55). This injury-induced nuclear export is essential for axon regeneration and fails to be activated in CNS injury.

Epigenetic modifications by DNA methylation also play a role in the regenerative response after axonal injury. DNA methylation is usually associated with suppressed gene expression, whereas hypomethylated regions become more accessible for transcription (56). Methylation occurs by DNA methyltransferases (DNMTs) and demethylation by ten-eleven translocation (TET) enzymes. PNS injury leads to upregulation of TET3 via calcium signalling and is necessary for RAG expression. After PTEN knock-out, axonal regeneration of mature retinal ganglion neurons, require activation of TET1, but not TET3. This may indicate that TET signalling is required for experimental induction of axonal regeneration in CNS, yet by implicating different DNA methylation pathways than the natural regeneration in the PNS (57).

An assessment of epigenetic patterns following axonal damage, showed distinct histone acetylation signatures correlating with gene expression associated with axonal regeneration (58). The correlation was only evident in the PNS, but not in the CNS. This points to an “epigenetic barrier” in the CNS neurons, which may, at least partly, explain its poor regenerative capacity. A similar barrier exists in PNS neurons, but is overcome following axonal injury due to epigenetic modifications (Figure 1). The barrier in the CNS appears during maturation of neurons as the axons reach their targets and form synapses (59, 60).

In addition to RAGs, master TFs and epigenetic modifications, post-transcriptional regulation of gene expression also plays an important role in the intrinsic regenerative program of the PNS. Micro RNAs (miRNAs) are a class of small (~22 nucleotides) noncoding RNAs that regulate post-transcriptional gene expression by preventing translation and/or promoting mRNA degradation (61). As each miRNA typically has several targets, multiple genes may be modulated simultaneously. Based on bioinformatic predictions, miRNAs may regulate >30% of all mammalian protein coding genes (62). These post-transcriptional modifications also appear to be a central player in axonal regeneration, as deletion of Dicer, an endoribonuclease essential for the assembly of miRNAs, leads to impaired axonal regeneration both in dissociated DRG cultures and in rodent models of PNS injury (63).

Specific miRNAs have also been shown to be highly involved in the regenerative response. In a rodent DRG model, mIR-21 was upregulated by 7-fold seven days after peripheral nerve branch axotomy (64). Overexpressed mIR-21 resulted in increased neurite outgrowth in dissociated adult rat DRG neurons, suggesting that mIR-21 is an axotomy-induced miRNA that promotes peripheral axon growth. In another study, mIR-26a mediated axon regeneration of sensory neurons in a sciatic injury model in mature mice via increased expression of GSK3β (65). In addition, miR-26a also supported neurite outgrowth in rat cortical neurons by suppression of PTEN (66). Furthermore, miR-133b promoted neurite outgrowth and functional recovery in a mouse SCI model by targeting the inhibitory RhoA (67). The different miRNAs involved in post-transcriptional regulation of RAGs and regeneration-associated cellular pathways may represent potential therapeutic targets in the promotion of neural regeneration.

Several intracellular pathways are highly involved in the regenerative response to axonal injury. These include, but are not limited to, the pathways GSK3β/CRMP2 (68), JAK/STAT (69), DLK/JNK/cJUN (70), and cAMP/PKA/CREB (71). Describing each specific pathway would be beyond the scope of this review. Here, we briefly focus on the role of the PI3K/PTEN/mTOR and RhoA pathway as examples.

The PI3K/PTEN/mTOR pathway is activated by binding of growth factors, such as brain-derived neurotrophic factor (BDNF) or neurotrophin-3 (Figure 2). This activates PI3K, which in turn converts PIP2 to PIP3, thereby phosphorylating AKT. AKT disinhibits mTOR via blocking of the negative regulators TSC1 and TSC2. mTOR regulates cell growth, protein synthesis and cell proliferation, and also promotes axon regeneration. PTEN is an inhibitor of the PI3K/AKT/mTOR pathway and leads to decreased levels of mTOR. In a pioneering experiment, deletion of PTEN promoted axon regeneration following optic nerve crush in a rat model (72). This effect was shown to be dependent of mTOR, as rapamycin, an inhibitor of mTOR, cancelled the effect of PTEN knock-out. The isolated activation of mTOR by TSC2 led, however, to less regeneration, suggesting that mTOR is not the only effector in the regenerative response of PTEN deletion. In a subsequent study, the same effect was shown in a rodent SCI model as PTEN deletion enabled injured corticospinal axons to regenerate past a spinal cord lesion (73). Ultrastructual analysis and staining with vGlut1, a presynaptic marker for excitatory synapses, suggested formation of synapses past the lesion. Regrowth of corticospinal tract lesions could also be induced by PTEN deletion one year after SCI in a mouse model (74). Another study showed restored locomotor abilities in both acute and chronic SCI in mice following PTEN knockout (75).

The small GTPase RhoA is a central binding element between extracellular factors inhibiting axon regeneration and the neuron-intrinsic response within the CNS. The myelin proteins MAG, NOGO, OMgp and the astroglial CSPG all activate RhoA, thereby preventing neural regrowth (34, 76). The consequences of RhoA activation are different in neurons and astrocytes (77). In neurons, RhoA restricts axonal growth by preventing microtubuli protrusion, whereas RhoA restricts injury-induced astrogliosis and CSPG production in astrocytes. Thus, blocking RhoA in neurons may promote regeneration, but cause opposite effects in astrocytes. This illustrates the complexity in tailoring regenerative therapies for axonal regeneration.

In SCI, the goal of stem cell transplantation is to promote axon growth and remyelination, thereby restoring neural circuits and improving functional outcomes. Trials have been performed applying both NSCs and MSCs for these purposes. In a Swiss-Canadian trial, human fetal NSCs were administered to 12 patients with chronic thoracic SCI via intramedullary microinjections in an open neurosurgical procedure (142). All patients had a motoric-complete, sensoric-incomplete injury. After six years follow-up, the procedure was found to be safe. Although some segmental sensory improvements were noted in five patients, there were no significant motoric effects. Another open label study included 29 patients with subacute, cervical- or thoracic SCI in a dose-escalation study of intramedullary injected human fetal NSCs (143). A total of 11 patients had motoric and sensoric complete injury while 18 had motoric complete and sensoric incomplete injury. No safety concerns were related to cells or procedure and efficacy measures were not reported. In a subsequent trial applying the same cell product, a total of 12 patients with chronic SCI were included (144). In cohort one 15–40 million NSCs were administered intraoperatively to six patients in a dose-escalation design. In cohort two, additional six patients received 40 million NSCs, and efficacy was compared to four untreated controls. Although the treatment was shown to be feasible and safe, there were no significant clinical effects. The trial was prematurely terminated by the sponsor due to the findings in a pre-determined futility analysis.

The administration of MSCs have also been assessed in SCI, either via intrathecal or intraoperative injections (145). Similar to the trials assessing NSCs, most of these were open-label trials designed to show feasibility and safety without a control group. An exception was a randomized, controlled, double-blinded study, which investigated effects of intrathecally injected MSCs derived from umbilical cord in ten patients with chronic complete thoracic SCI in a cross-over design (146). At six months, the treatment was switched so that patients initially receiving MSCs now received placebo and vice versa. Apart from a significant increased pin-prick sensation in dermatomes below the site of injury, no improvements were noted in motor functions or neurophysiological parameters. Other types of cells have also been tested in patients with SCI, including a small phase I trials of Schwann cells (147) and case reports of olfactory ensheathing cells (148). No clear therapeutic benefits were reported.

Unlike in SCI and IS, lesions are multifocal in MS and undergo a chronic inflammatory and progressive neurodegenerative course. In theory, this may fit to the migratory capacity of NCSs and MSCs following intrathecal injection. Once the cells reach the lesions, their paracrine abilities can promote modulation of activated microglia and accelerate remyelination via stimulation of oligodendrocyte maturation. With intravenous administration, on the other hand, the cells are unable to reach the CNS as the majority get trapped in the lungs (149). Nevertheless, MSCs have been shown to provide systemic modulatory effects to the adaptive immune system, which hypothetically could have a beneficial effect in MS patients. Recently, the results of the largest study so far using MSCs in MS were reported (150). In this trial, 144 patients were randomized to receive either autologous bone marrow derived MSCs or placebo intravenously in a cross-over design. At 6 months, there were no difference in number of gadolinium-enhanced lesions, which was the primary outcome of the trial. Studies applying intrathecal injections have also been performed. Quite recently, the results of a “first in man,” phase I trial using fetal NSCs in 12 patients with progressive MS was published (151). This open-label study showed that one single intrathecal injection of NSCs was feasible and safe after 2 years follow up. All patients received immune suppression with tacrolimus during the study period to prevent unwanted immune responses and were randomized into four groups with increasing numbers of injected cells. Despite the small study population, exploratory analyses showed a dose-dependent decrease in brain atrophy on MRI and increased levels of neurotrophic factors in CSF, including GDNF, VEGF-C and stem cell factor (SCF), suggesting a beneficial effect. On the other side, there was a nearly significant worsening of EDSS during follow-up and 50% of the patients developed new brain lesions. The causality between these, both positive and negative, observations remain elusive as we do not know the fate of the cells once injected. In 2020, the results from the largest study so far using intrathecal administration of MSCs in MS were published (152). A total of 48 patients with active progressive MS were randomized to receive autologous BM-derived MSCs or placebo either intrathecally or intravenously in a cross-over design. There were no serious adverse events (SAEs) related to the cells or the procedure, and significantly fewer patients in both treatment groups (intrathecal and intravenous MSCs) experienced disease activity compared to those receiving placebo. Patients receiving MSCs also showed significant improvements in other tests, such as EDSS, 25-foot timed walking test, 9-hole peg test and OCT. Especially patients receiving intrathecal administrations had favourable outcomes. However, the study did not assess the underlying mechanism, i.e., how the MSC treatment led to clinical improvement. Consequently, it is not known whether the cells migrated to the lesions or how long they survived.

Clinical trials have also been performed utilizing stem cell therapy aiming to improve outcomes following IS. In an open-label trial, 18 patients with chronic stroke received stereotactic implantation of modified allogeneic BM-derived MSCs (153). Although there were no SAEs related to the cells, six SAEs were caused by the invasive procedure. After one year of follow-up, modest improvements were noted in different clinical scales, such as the National Institutes of Health Stroke Scale (NIHSS) and the Fugl-Meyer total score. However, efficacy was difficult to interpret in the absence of a control group.

Allogeneic adult multipotent progenitor cells from BM have also been tested in IS. In a placebo-controlled, double-blinded phase II trial, 67 IS patients received these cells intravenously within 24–48 h after stroke onset (154). The rationale of the treatment was to provide peripheral immunomodulation, thus promoting neuroprotection by alleviation of the acute neuroinflammatory response. Although the treatment was safe and showed reduced serum levels of cytokines and regulatory T cells in patients receiving the cell product, no beneficial clinical effects were noted when compared to placebo.

Another trial assessed safety and efficacy of an immortalised human neural stem cell line in a dose escalation design with up to 20 million cells administered via stereotactic injection into the ipsilateral putamen (155). A total of 11 patients with chronic stroke and stable symptoms were included. After 2 years of follow-up, four SAEs were considered related to the procedure, but none directly to the cell therapy. Disability remained unchanged in seven patients, worsened in one and improved in three. A follow-up study with 23 patients applying the same principle of treatment showed similar results and reported improvements in upper limb function in patients with residual upper limb movements at baseline, thus identifying a possible target population for a larger study (156). A phase IIb trial was initiated, but reported terminated in 2021 (NCT03629275).

Recently, another two randomized, controlled trials including 39 and 17 stroke patients failed to show improved clinical outcomes after receiving autologous MSCs, as compared to placebo (157) or those receiving standard care (158).

An example of targeting inhibitory mechanisms preventing neural regeneration is VX-210. This is a derivate of the bacterial enzyme C3 transferase, which inhibits Rho activity through covalent modification. As previously described, Rho activation stalls neural regrowth by promoting the collapse of the axonal growth cone. In mouse models of SCI, VX-210 (in the study referred to as “BA-210”) was applied to the dura mater and diffused into the spinal cord (159). Rho was deactivated in a dose-dependent manner and locomotion was improved when VX-210 was administered at the time of injury or 24 h post-injury. A similar approach was tested in a phase I/II dose-escalation trial, which included 48 patients with complete SCI at the cervical or thoracic level (160). Here, VX-210 was topically applied in a fibrin sealant to the dura mater following decompressive surgery ≤7 days after injury. Results confirmed safety and suggested improved motor strength as compared to historical controls in patients with cervical, but not thoracic SCI. A larger trial commenced, but was pre-maturely terminated due to fulfilment of predefined futility criteria after the inclusion of 70 patients (161). Heterogeneity of the patient population and suboptimal administration mode were among the highlighted factors potentially explaining the negative results. Nevertheless, this was the first randomized trial to assess an agent designated to block an inhibitory factor of neural regeneration.

In SCI, another clinical trial was performed using ATI355, a recombinant human antibody directed towards the human Nogo-A protein (162). This notion was based on the observation that intrathecally delivered Nogo-A antibodies promoted axonal growth and functional recovery in rodent and primate models of SCI (163, 164). A total of 52 patients with acute and subacute complete SCI received intrathecally administrated ATI355, either via continuous infusion or repeated spinal injections over four weeks in different doses. The Nogo-A antibody was generally well tolerated, although one case of bacterial meningitis was reported as a SAE related to administration method. The concentration of the ATI355 in CSF was mostly around 0.1 μg/mL, which should have been sufficient for therapeutic effect based on pre-clinical data. Efficacy parameters did not show any clinical improvements as compared to retrospective longitudinal data. The study was, however, not primarily designed to assess clinical efficacy.

In SCI, the loss of tissue and cystic cavity represent a substantial hurdle for regeneration and functional recovery. Polymer-based biomaterial scaffolds have been associated with prevention of cystic cavitation, less glial scarring and axonal sprouting through the scaffold in rodent and non-human primate SCI models (165, 166). The implantation of such a device within the cavity of SCI has also been assessed in a clinical study. An open-label trial treated 19 thoracal SCI patients requiring open spine surgery with a bioresorbable polymer device (Neuro-Spinal Scaffold) within 96 h postinjury (167). Patients were followed for two years, and the procedure was safe and well tolerated. Although the trial did not have a control group, comparison to historical controls indicated clinical benefit with 32% of patients converting to motor incomplete injury as opposed to ≤17% in historical data. Nevertheless, a placebo-controlled trial is required to further evaluate efficacy.

In MS, clinical trials have aimed for remyelination, thus promoting neural regeneration indirectly. One therapeutic option has been to target the Lingo-1 receptor (Figure 1). This glycoprotein is selectively expressed on CNS neurons and oligodendrocytes and inhibits oligodendrocyte differentiation, myelination and axonal regeneration (168). Opicinumab, a human monoclonal antibody against Lingo-1, has been shown to promote remyelination in rodent demyelination models (169). A randomized, phase II trial of 82 patients with a first episode of optic neuritis suggested improved optic nerve conduction of opicinumab, although not statistically significant. The findings led to a larger phase II trial, which included >400 patients with MS (170). Patients were treated with intravenously administrated opicinumab or placebo every 4 week over 72 weeks in a dose-escalation design. Results failed to show a dose linear improvement of disability, which was the primary endpoint. The highest dose of 100 mg/kg, being applied in the previous trial of optic neuritis, did not prove efficient as compared to placebo. However, the doses 10 mg/kg and 30 mg/kg showed some effect in favour of opicinumab. The authors discussed the possibility that a too high dose of opicinumab could have differentiated the oligodendrocytes too early, making them unable to migrate into the demyelinated lesions.

Retinoic acid receptor RXR-gamma agonists have also been shown to promote remyelination in models of demyelination in vitro and in vivo (171). The retinoic X receptor is a positive regulator of OPC differentiation and may thus mitigate the impaired maturation of OPCs in chronic MS lesions (172). Bexarotene, a retinoid and an antineoplastic agent used to treat cutaneous T cell lymphoma, is an agonist to the gamma retinoid X receptors. A randomized placebo-controlled phase II trial was performed to assess the remyelinating potential of bexarotene in MS (173). Results after inclusion of 52 patients showed no differences in remyelination as assessed by mean lesional magnetisation transfer ratio on MRI at 6 months. The treatment was not well tolerated as there were significantly more adverse events in the treatment group than in the placebo group.

Finally, clemastine has also been tested for remyelination in a clinical setting. Clemastine is a first generation antihistamine available over the counter in most countries and has been shown to induce OPC differentiation and myelination due to off-target antimuscarine effects in pre-clinical demyelination models (174, 175). In a phase II trial, 50 patients with relapsing remitting MS and chronic demyelinating optic neuropathy were randomized to clemastine fumarate for 90 days and placebo for 60 days in a cross-over design (176). Results showed a significant decrease in latency delay in visual evoked potentials, suggesting a remyelinating effect. There were no differences in secondary endpoints, such as MRI parameters or clinical measures, including EDSS, T25W and 6MWT. Fatigue was a common adverse event, possibly related to the antimuscarin blockade. So far, no confirmatory phase III trials have been performed. Recently, the clemastine arm in an ongoing interventional trial was terminated because three patients with progressive MS experienced increased disability during treatment (177). CSF proteomic profiling showed that clemastine may have caused innate inflammation via the purinergic P2RX7 receptor in microglia and oligodendrocytes. This discrepancy highlights important differences in treatment effects between progressive and relapsing-remitting MS.

In IS, treatment with cerebrolysin has been translated into the stage of clinical trials. Cerebrolysin is a mixture of enzymatically treated peptides derived from porcine brain tissue with proposed neurotrophic properties (178). The therapeutic effect has been suggested to occur through enhanced neurogenesis involving the sonic hedgehog pathway leading to improved outcomes in rodent stroke models (179). Other mechanisms include reduced levels of free radicals dampening excitotoxicity. In an exploratory, placebo-controlled clinical trial including 208 stroke patients, cerebrolysin was intravenously administrated within 24–72 h after stroke onset and continued for 21 days (180). The cerebrolysin group showed improved upper limb function at day 90 with good tolerability. A follow-up study did, however, not show any significant clinical effects (181). A meta-analysis including 1,417 trial participants, could also not demonstrate clinical benefits of cerebrolysin in the treatment of acute IS (182).

Granulocyte-colony stimulating factor (G-CSF) is another agent that may promote neurogenesis after stroke, thereby reducing stroke volume and inducing functional neurological improvement (183). G-CSF increases the release of myeloid cells and hematopoetic stem cells from the bone marrow and is frequently applied following chemotherapy-induced neutropenia and stem cell transplantation. The potential neuroregenerative effect in stroke is thought to occur via G-CSF receptors on neural progenitor cells, stimulating their differentiation into neurons (184). Although an initial study showed safety and a promising dose-dependent effect in acute IS (185), a subsequent phase IIb placebo-controlled trial including 328 patients failed to show efficacy both to clinical outcomes and imaging biomarkers (186). In this trial, IS patients received G-CSF intravenously over 72 h within 9 h after stroke onset. A following meta-analysis did not show improved outcomes in stroke patients receiving G-CSF, but a trend towards more frequent SAEs (187).