Therapeutic repair for spinal cord injury: combinatory approaches to address a multifaceted problem

Abstract The recent years saw the advent of promising preclinical strategies that combat the devastating effects of a spinal cord injury (SCI) that are progressing towards clinical trials. However, individually, these treatments produce only modest levels of recovery in animal models of SCI that could hamper their implementation into therapeutic strategies in spinal cord injured humans. Combinational strategies have demonstrated greater beneficial outcomes than their individual components alone by addressing multiple aspects of SCI pathology. Clinical trial designs in the future will eventually also need to align with this notion. The scenario will become increasingly complex as this happens and conversations between basic researchers and clinicians are required to ensure accurate study designs and functional readouts.


Introduction
Through silver staining neurons, Ramon y Cajal discovered that peripheral nervous system (PNS) neurons regenerate after injury, contrasting the minor regenerative response of central spinal cord neurons (Ramon y Cajal, 1928). This then prompted the question: Do central nervous system (CNS) neurons lack the intrinsic capability of regeneration or are there extrinsic factors that influence this dichotomous observation? In actuality, both of these factors come into play when spinal cord neurons are tasked with regenerating after axotomy. Several families of molecules present in the extracellular matrix (ECM) prevent axon growth including chondroitin sulphate proteoglycans (CSPGs), myelin-associated molecules, ephrins and semaphorins (Miranda et al, 1999;Chen et al, 2000;Willson et al, 2002;Silver & Miller, 2004;Geoffroy & Zheng, 2014;Worzfeld & Offermanns, 2014). Yet even when provided with a growth-permissive environment, central neurons regenerate feebly compared to their peripheral or immature CNS counterparts, indicating that they also have intrinsic growth limiting factors (Hilton & Bradke, 2017). On the brighter side and nearly 100 years on from Cajal's statement that "in adult centres, the nerve paths are something fixed, ended, immutable; everything may die, nothing may be regenerated" (Ramon-Cueto et al, 1998), we now know this statement to be not entirely true. Many research groups have reported axon regeneration and functional recovery after experimental spinal cord injury (SCI) following a variety of treatments (Thuret et al, 2006). While this is certainly a feat, many of the experimental approaches are problematic for clinical translation. For those that are, it is also becoming increasingly apparent that addressing singular aspects of the problem won't facilitate successful and functional regeneration after SCI in humans. Conversely, it will likely require the combination of various treatment strategies that address the variety of problems that result after SCI. Many attempts have been made, to varying degrees of success, that combine tissue replacement, removal of inhibitory molecules, supplying neurotrophic factors, manipulation of pro-regenerative neuronal signalling pathways and neurorehabilitation. This review will provide an update to the many therapeutic interventions after SCI and then will focus on the attempted combinatory approaches, how they could be improved and the road to their clinical translation. that can last months to years after the injury (Buss et al, 2004;Norenberg et al, 2004;Donnelly & Popovich, 2008). Some degree of spontaneous recovery is observed in experimental animal models and to a lesser extent in humans (Curt et al, 2008;Hilton et al, 2016). However, endogenous repair mechanisms are minor and recovery remains incomplete (Fawcett et al, 2007;Courtine et al, 2008). Based on the pathologies that result from SCI, researchers have identified several targets for the development of potential therapeutic interventions. These can be referred to as the "7 R's" (Fig 1). 1 Reduction of secondary damage (neuroprotection). 2 Replacement of cells lost to primary and secondary damage. 3 Removal of inhibitory molecules. 4 Regeneration to enhance the spontaneous reparative and regenerative responses. 5 Resupply of neurotrophic support to improve neuronal survival and direct axonal growth.

6
Remyelination of regenerated, replaced or spared (demyelinated) axons. 7 Rehabilitation strategies to induce neuroplasticity and/or to shape neuronal connections.
We will discuss these targets and their respective interventions in the following sections.
Reduction: neuroprotection Substantial efforts have been made to limit secondary damage. Pharmacological agents that suppress the immune system or inhibit key signalling pathways involved in inflammation were the first major strategies applied to patients. These include the non-steroidal anti-inflammatory drugs (NSAIDs), minocycline, cyclosporine A and the corticosteroid methylprednisolone (Badhiwala et al, 2018). The use of methylprednisolone remains controversial since clinical trials revealed that it may have no effect or even lead to complications (Bowers et al, 2016). Preclinically, indomethacin (an NSAID) led to Glossary Allodynia A type of neuropathic pain that is the result of increased sensitivity to otherwise innocuous stimuli. Anterograde tracing This is a research method which is used to trace axonal projections in the direction of their cell bodies towards their point of termination. The complimentary technique is known as retrograde tracing.

Autologous (transplant)
A transplant is autologous if the recipient also serves as the donor. Basso, Beattie and Bresnahan (BBB) locomotor rating scale A 21-point behavioural analysis scale that is used in spinal cord injury research to assess limb motor skills over a wide range of injury severities. Chondroitin sulphate proteoglycans (CSPGs) Components of the extracellular matrix and are composed of a protein core covalently linked to chondroitin sulphate glycosaminoglycan sidechains. CSPGs are known to inhibit axon regeneration. Chondroitinase ABC (ChABC) Chondroitinase ABC is a bacterial enzyme that degrades chondroitin sulphate.

Conditioning lesion
These involve lesioning the peripheral branch of adult sensory dorsal root ganglion neurons prior to a central lesion. This increases the regenerative capacity of the central branch of DRG neurons. Contusion An injury caused by blunt trauma to the spinal cord. Corticospinal tract (CST) One of the most important descending tracts. It controls primary motor activity from the motor cortex. Dorsal root ganglion (DRG) A cluster of neurons in dorsal spinal root of a spinal nerve. DRG neurons are pseudo-unipolar and have a central and peripheral branch. Excitotoxicity A pathological process by which neurons are damaged or killed through overstimulation of excitatory neurotransmitter receptors. Functional electrical stimulation (FES) FES involves electrophysiological stimulation of spinal cord or peripheral nerves or muscle. Growth cone Is the large actin supported extension at the tip of a developing or regenerating neurite/axon.

Induced pluripotent stem cells (iPSCs)
Are pluripotent stem cells that are generated directly from adult cells through the expression of specific transcription factors.

Kinematic analysis
The process of measuring the movement (kinematic) quantities used to describe motion.

Myelin-associated inhibitors
Components or CNS myelin that are known to be inhibitors of axon growth and regeneration. The three classic myelin-associated inhibitors are Nogo-A, myelin-associated glycoprotein and oligodendrocyte myelin glycoprotein. Neuroplasticity Refers to the adaptive changes in neurological function. This can be the result of anatomical or physiological changes. Neuroprotection Refers to the relative preservation of neural tissue during ongoing secondary damage. Neurorehabilitation Physical therapy that aims to aid recovery of nervous system injury and/or improve compensatory functions. Neurotrophins A family of regulatory factors that mediate the survival, differentiation and growth of neurons. Phase I clinical trial A clinical trial that includes a small number of human participants to determine the safety of a new drug or invasive medical device; allows the determination of drug dosage or toxicity limits. Phase II clinical trial A clinical trial that includes a larger number of human participants than a Phase I trial and is intended to evaluate the efficacy of a treatment; side effects are also monitored.

Propriospinal neuron
These are neurons contained entirely within the spinal cord that interconnect various levels of the cord.

Reticulospinal tract
A descending tract that descends from the reticular formation and is primarily responsible for locomotion and postural control. Retraction bulb Dystrophic axonal structures which are a hallmark of a failed growth response. Transection A surgical injury created by a fine transverse cut of the spinal cord.
tissue sparing and slight functional recovery (Simpson et al, 1991), while minocycline treatment reduced oligodendrocyte as well as neuronal death that improved outcome after cervical spinal cord injury in rats (Stirling et al, 2004). However, preclinical reassessment of indomethacin suggested treatment may be harmful in human SCI, and a phase II clinical trial of minocycline for acute SCI did not report beneficial effects (Guven et al, 1999;Casha et al, 2012). Similar conflicting results have been reported for the immune suppressant cyclosporine A . Since neuroinflammation has both beneficial and detrimental effects, broadspectrum suppression of inflammation may not be efficacious. Neuroprotection can also be achieved through preventing glutamate excitotoxicity by blockade of NMDA receptors by magnesium (Ditor et al, 2007) or gacyclidine (Feldblum et al, 2000), blockade of tetrodotoxin-sensitive sodium channels using riluzole (Satkunendrarajah et al, 2016), preventing apoptosis using erythropoietin (Baptiste & Fehlings, 2006), inhibition of connexin hemichannels using a mimetic peptide (Mao et al, 2017), mild to moderate hypothermia (Dietrich et al, 2009) and many more strategies (Baptiste & Fehlings, 2006;Thuret et al, 2006). These more discrete manipulations of secondary processes may prove to have greater therapeutic benefit than the broad approaches. Regardless, neuroprotective agents have failed to translate to the clinic. Stroke has a similar neuropathology to SCI, and huge efforts, particularly in the 1990s, were made to develop neuroprotective treatments for stroke. In such attempts, over 1,000 potential therapeutic agents, leading to nearly 200 clinical trials, have resulted in no successful treatments (Minnerup et al, 2012). Worryingly, clinical trials assessing neuroprotective agents for SCI may also reflect this trend. Overall, the preclinical and clinical progression of neuroprotective strategies has often stalled, and the mechanisms by which neuroprotection is conferred is not well understood. Improvements in our understanding as to the mechanisms which underpin the benefits of neuroprotective strategies could better inform their application.

Replacement: cellular transplantation
Cell transplantation for the aim of replacing lost cells has multiple historic origins stretching back to experiments conducted in the laboratory of Ramon y Cajal (Ramon-Cueto et al, 1998). The seminal works of Anders Björklund and colleagues showed that foetal tissue can promote CNS repair by replacing lost cell types in models of Huntington's and Parkinson's disease (Bjorklund & Lindvall, 2000). This led to the discovery that transplantation of foetal spinal cord tissue into the injured spinal cord results in successful graft survival, differentiation and integration into the host tissue (Bregman et al, 1993). Likewise, experiments by Richard, David and Aguayo revealed that peripheral nerve grafts could provide a permissive conduit for regenerating CNS axons (Richardson et al, 1980David & Aguayo, 1981;Benfey & Aguayo, 1982), which has since been thoroughly studied (Cote et al, 2011a). These studies provided critical evidence that CNS axons are capable of regenerating. The grafts support a variety of neuronal types including propriospinal neurons and often neurons from the brain stem (Cote et al, 2011a). It appears that only injured axons regenerate through the grafts and not sprouts from non-injured axons (Friedman & Aguayo, 1985). Despite these promising findings, there are few studies using PNGs alone which correlate this axonal growth to behavioural or recovery of axonal conduction. The likely reason for this is the remaining challenge that CNS and PNS neurons fail to extend beyond the distal grafthost interface where they encounter the inhibitory environment of the CNS. For example, lack of axon extension and functional improvements were observed in primates after spinal hemisection and PNG transplantation (Levi et al, 2002).
Cell-based transplantations have largely superseded nerve grafts for several reasons: they can be injected into the spinal cord to fill the lesion site, they are less likely to cause further damage compared to a nerve graft, and cells can be genetically modified ex vivo to secrete specific growth factors (Assinck et al, 2017). The mechanisms behind which cell transplantation confers therapeutic benefit are often multifactorial. This includes direct replacement of damaged neural cells, neuroprotection of the host cells, promoting axon regeneration and synapse formation, and/or promoting myelination of damaged or newly formed axons (Assinck et al, 2017). A variety of cells have been used for SCI transplantation, including mesenchymal stem cells (MSCs), neural progenitor cells (NPCs), Schwann cells, olfactory ensheathing cells (OECs) and induced pluripotent stem cells (iPSCs; Assinck et al, 2017). Each has advantages and disadvantages when compared to one another.
MSCs and NPCs are stem cells, typically harvested from embryonic or foetal tissue, and can differentiate into neurons or glia in vitro (Liu et al, 2000;Billon et al, 2006). Neurons obtained from in vitro differentiation can survive and integrate into the injured rat spinal cord (Deshpande et al, 2006). Whether such differentiation occurs following in vivo transplantation has been less clear in some circumstances considering early investigations reported that majority of transplanted cells remain progenitor-like or differentiate into glial cells (Vallieres & Sawchenko, 2003;Karimi-Abdolrezaee et al, 2006). It is now generally accepted that MSCs do not form neurons in vivo (Lu et al, 2004). NPCs, on the other hand, were recently shown to extend "hundreds of thousands of axons into the spinal cord" (Rosenzweig et al, 2018), and injured motor and sensory axons regenerated into appropriate domains of NPC grafts (Dulin et al, 2018). Furthermore, these NPC grafts form host-graft synaptic network formation in patterns paralleling the normal spinal cord (preprint: Ceto et al, 2019). Aside from direct tissue replacement, both MSCs and NPCs confer permissive substrates for growth ▸ Figure 1. The seven targets for therapeutic interventions following spinal cord injury.
A horizontal plane view through a region of thoracic spinal cord injury depicting some of the features of the pathology. SCI leads to immediate and continued death of neural alongside disruption of descending and ascending fibres. Seven therapeutic targets are present which can improve functional recovery after SCI: neuroprotective strategies to limit ongoing secondary damage resulting in spared tissue; tissue and cellular transplants to replace lost cells and may provide trophic or growth-permissive environments; removal of inhibitory factors such as CSPGs to allow for enhanced axonal growth; targeting neuron-intrinsic mechanisms to enhance intrinsic regenerative response which could then be directed through the resupply of trophic support; and remyelination of demyelinated axons may improve axonal conduction. Finally, rehabilitation to function in circuit remodelling and strengthens beneficial connections.
whereby injured host axons grow into the cellular grafts (Hofstetter et al, 2002;Lu et al, 2003;Ankeny et al, 2004). Injured axons grow into an NPC transplant, mediated by the secretion of various neurotrophic factors from the transplant itself (Lu et al, 2003). Similar to that observed in PNS grafts, regenerating host axons typically terminate their growth at the border of the transplant (Ruff et al, 2012;Assinck et al, 2017). A final consideration is that NPCs may also facilitate growth via paracrine actions. For example, modulation of neuroinflammatory processes may also contribute to the beneficial effects NPCs (Kokaia et al, 2012).
Schwann cells and OECs are terminally differentiated, myelinating and regeneration promoting cells found in the PNS and the olfactory system, respectively. Like other transplantation studies, both Schwann cells and OECs confer structural and trophic support (Bunge & Pearse, 2003;. Schwann cells transplanted into the damaged spinal cord of rodents reduce cavitation and promote regeneration of both ascending and descending axons into the graft, and axons became myelinated and are evidenced to be electrophysiologically active (Xu et al, 1995b(Xu et al, , 1997Pinzon et al, 2001;Takami et al, 2002). However, similar to MSC and NPC transplants, in these studies, regenerating axons failed to leave the graft distally to reinnervate the host tissue. Recovery of limb functions was reported by some (Takami et al, 2002) but not all publications (Pearse et al, 2004a). Similar observations have been witnessed for OECs. After injection of OECs into a cervical unilateral lesion site of the corticospinal tract (CST) in rodents, anterograde tracing of lesioned hindlimb CST axons revealed extensive regeneration through the graft and integration with the CST beyond the injury site (Li et al, 1997). Later, it was shown that this intervention could improve respiratory function and climbing ability of the rats (Li et al, 2003). However, it was later challenged whether OECs promote CST regeneration (Lu et al, 2006). Other studies, including veterinary trials, have also shown the ability of OECs to promote robust regeneration of some spinal pathways even beyond the graft to enhance functional recovery (Ramon-Cueto et al, 1998;Lu et al, 2002;Li et al, 2003;Jeffery et al, 2005;Toft et al, 2007). The benefits of Schwann cells and OECs can be potentiated when delivered together. For example, co-delivery of the two cell types potentiates long-distance axonal regeneration through and around guidance channels compared to what is achieved by either the cell type alone (Ramon-Cueto et al, 1998).
iPSCs may circumvent the ethical issues associated with embryonic or foetal tissue use. NPCs generated from iPSCs have demonstrated beneficial effects after transplantation for animal models of SCI . It remains unclear whether meaningful differentiation into neurons occurs or whether they play a supportive role. As is the case with other pluripotent cells, a critical safety issue for the use of iPSCs is the risk of tumorigenicity. Though promisingly, in marmosets, transplanted iPSC-NPCs at the subacute phase predominantly differentiated into neurons around the lesion site without tumorigenicity and promoted axonal regrowth and angiogenesis, and preserved myelination area (Kobayashi et al, 2012). Furthermore, one strategy to decrease the risk of tumorigenicity is the addition of gamma-secretase inhibitors, which removes tumour-initiating cells and promotes functional recovery in the subacute and chronic phases of SCI in preclinical studies (Okubo et al, 2016(Okubo et al, , 2018.
Overall, preclinical SCI studies have demonstrated that various cellular transplants are efficacious through many mechanisms. Considering the limitations listed above, there remains much experimentation. The full potential of any cell type for transplantation will likely require the combination with other synergistic therapies.
Removal: targeting chondroitin sulphate proteoglycans Removal or disruption of CSPGs can be achieved by inhibiting their synthesis, enzymatic degradation, antibody neutralisation or pharmacological targeting of effector molecules (Bradbury & Burnside et al., 2017). For example, xylosyltransferase-1 (XT-1) is crucial for the biosynthesis of glycosaminoglycan (GAG) chains of CPSGs. Reduction of XT-1 expression through using a DNA enzyme for catalytic degradation of XT-1 mRNA strongly reduced CSPG-GAGs and allowed for axons from microtransplanted dorsal root ganglions (DRGs) to grow around a dorsal column lesion in rats (Grimpe & Silver, 2004). In a similar study, mRNA-mediated knockdown of XT-1 after dorsal column transection in rats significantly reduced proteoglycan expression and increased the length and density of ascending axons through a peripheral nerve graft (Hurtado et al, 2008). This was later replicated in a model of spinal cord contusion where increases in serotonergic innervation caudal to the injury correlated with a reduction in errors during a horizontal ladder test (Oudega et al, 2012).
Enzymatic modification of CSPGs is most commonly achieved with the enzyme chondroitinase ABC (ChABC). Derived from the bacteria Proteus vulgaris, it catalyses the degradation of the glycosidic bonds between CS-GAGs of CSPGs, liberating them from the CSPG core protein (Prabhakar et al, 2005a(Prabhakar et al, , 2005b. Intrathecal infusion of ChABC for 10 days into rats following C4 dorsal column crush lesion degraded CSPGs, increased CST-axonal regeneration and improved functional scores on several behavioural tests (Bradbury et al, 2002). Using anterograde tracing and immunohistochemical markers for neuroplasticity, ChABC treatment promoted sprouting of intact and injured spinal systems and formed relay networks after SCI (Barritt et al, 2006). The beneficial effects of ChABC delivery have been replicated by many independent laboratories in a variety of injury models (Bradbury & Carter, 2011). The positive results of using ChABC in rodents have been further confirmed in larger animals. In adult cats following thoracic hemisection, for example, ChABC enhanced functional recovery of skilled locomotion and kinematic measurements of hindlimb function (Tester & Howland, 2008). Importantly, in a recent SCI clinical trial in dogs, ChABC resulted in a seemingly modest, yet significantly improved forelimb-hindlimb coordination and early improvements to bladder compliance with no evidence of long-term adverse effects (Hu et al, 2018). While average group differences are small, 10% of the treated dogs recovered independent ambulation which may represent an estimate of the "true" population effect is in such a large and heterogeneous sample size of 60 dogs (Moon & Bradbury, 2018). Cervical hemisection in rhesus monkeys and intrathecal administration 4 weeks later resulted in increased CST axon growth and formation of synapses from CST axons caudal to the lesion (Rosenzweig et al, 2019). This correlated with improved hand function compared to vehicle-treated controls. Increasing neuroplasticity may even have long-lasting effects. A single injection of ChABC into the phrenic motor pool following complete cervical hemidiaphragm paralysis in rats resulted in robust patterned respiratory recovery for more than 1.5 years (Warren et al, 2018). As such, through unmasking of the neuroplasticity that develops after injury, ChABC treatment can ensure rapid and robust functional recovery after a near lifetime of paralysis in rats.
The continuous production of CSPGs after injury may require chronic expression of ChABC. For this reason, a mammalian-compatible ChABC gene was engineered by modifying N-glycosylation sites to allow for secretion of ChABC from eukaryotic cells (Muir et al, 2010). This also avoids tissue damage of repeated administration and addresses the possible risks for immune recognition of the bacterial enzyme. Lentiviral vector (LV) gene delivery was used to achieve long-term expression of this "mammalianised" ChABC (mChABC) gene in the contused rat spinal cord which resulted in large-scale CSPG degradation and improved behavioural scores following thoracic injury (Bartus et al, 2014). Recently, these findings were replicated in the functional restoration of upper limb function and the strategy further developed to control gene expression (James et al, 2015). An immune-evasive, doxycycline Tet-on inducible LV-mChABC vector enabled temporal control over ChABC expression (Burnside et al, 2018). Such regulatable gene therapy approach could be a viable candidate for clinical translation.
There are endogenous proteins which also may affect matrix composition that could be therapeutically harnessed. These include matrix metalloproteinases (MMPs), a disintegrin and metalloproteases (ADAMs) and a disintegrin and metalloproteinase with thrombospondin motifs (ADAMTSs; Troeberg & Nagase, 2012). MMPs, ADAMs and ADAMTSs have been implicated to play a beneficial role in CNS injury by degrading CSPGs (Burnside & Bradbury, 2014). For example, knockout of MMP-9 in mice caused worsened motor deficit after traumatic brain injury (Wang et al, 2000). ADAMs and ADAMTSs are of particular interest as they display CSPG-specific substrate recognition. ADAMTS4 infusion intrathecally in the contused rat spinal cord resulted in a significant improvement in the Basso, Beattie and Bresnahan (BBB) locomotor rating scale score that was comparable to ChABC infusion (Tauchi et al, 2012). More recently, an astrocyte-selective AAV-ADAMTS4 gene therapy resulted in CSPG degradation, decreased lesion size, increased CST axon regeneration, increased caudal serotonergic inputs and significantly improved BBB and error ladder scores following thoracic contusion in rats (Griffin et al, 2020). Given the safety profile of AAV viruses (Hudry & Vandenberghe, 2019), this gene therapy could potentially represent a safer alternative to LV-ChABC gene therapy. Ultimately, removal of CSPGs has proven to be consistently advantageous for nearly two decades and we anticipate the next steps taken towards clinical translation of CSPGtargeting therapies.
Removal: targeting myelin-associated inhibitors Independent laboratories have reported CNS axon growth and recovery of limb function following the use of anti-Nogo-A antibodies (Chen et al, 2000;GrandPre et al, 2000). In non-human primates, anti-Nogo-A antibody therapies promoted the growth of CST tract axons after unilateral dorsal hemisection and regain of fine hand control (Freund et al, 2009). Interestingly, the co-delivery of anti-Nogo-A antibodies and ChABC is more effective than single treatments at enhancing functional recovery after SCI indicating that removal of multiple inhibitory factors is advantageous (Zhao et al, 2013). Other therapeutic targeting of Nogo includes a soluble Nogo-66 receptor protein (NgR (310) ecto-FC) that complexes with Nogo-A. NgR (310) ecto-FC enhances CST and raphespinal axon growth after dorsal hemisection, and after complete transection in rats significantly enhances the number myelinated fibres in the bridge and increases the area of the bridging tissue between the cord stumps Guo et al, 2012). The Nogo receptor antagonist NEP1-40 originally promoted CST and serotonergic fibre growth following thoracic dorsal hemisection in rats with striking results (GrandPre et al, 2002). The promising results obtained from the last three decades of preclinical research investigating Nogo-A and anti-Nogo-A antibodies led to the first-in-man intrathecal application of anti-Nogo-A antibodies in a phase I human clinical trial of cute spinal cord injury (Kucher et al, 2018), and further trials regarding NOGO intervention are currently ongoing (NCT03935321, NCT03989440).

Regeneration: targeting neuronal intrinsic mechanisms
Intrinsic growth capacity of central neurons declines with age as the neuron differentiates for synaptic functions (Hilton & Bradke, 2017). This is a major reason why they mount a minor regenerative response following an injury. Through studying developing immature neurons and regenerative-competent peripheral neurons we know that for an axon to regenerate, many diverse and coordinated intracellular mechanisms are required. These include cytoskeleton dynamics, axonal transport and trafficking, signalling and transcription of regenerative programmes, and epigenetic modifications (Curcio & Bradke, 2018;Fawcett & Verhaagen, 2018).
Reactivation of an intrinsic growth programme can be accomplished through conditional lesioning which involves transecting the peripheral branch of adult sensory DRG neurons prior to a central lesion . Studies correlated cyclic adenosine monophosphate (cAMP) levels in DRG neurons to growth capacity through induction of important pro-regenerative transcription factors (Qiu et al, 2002). Local administration of cAMP to DRGs can, at least in part, mimic the conditioning lesions (Cai et al, 2001;Neumann et al, 2002;Qiu et al, 2002). Rolipram prevents the hydrolysis of cAMP and sustains elevated levels of cAMP that mimic the effects of a conditioning lesion effect to some extent (Nikulina et al, 2004;Pearse et al, 2004b). cAMP-mediated gene transcription alone only partially recapitulates the conditioning lesion effect , indicating that there are opportunities remaining to discover therapeutic interventions that do so.
Genetic manipulation experiments, including deletion of PTEN or GSK3b, or overexpression of STAT3, KLFs or SOX11, have exemplified that these manipulations to aspects of regenerative programmes can produce a modest regenerative response (Curcio & Bradke, 2018). Converting these interventions into clinically useable alternatives such as pharmacological agents or viral vector gene therapies of these could be a promising avenue (Blackmore et al, 2012;Zukor et al, 2013;Wang et al, 2015). Another strategy is to modulate intracellular signalling through modulation of GTPases. This has been achieved through RhoA-GTPase inhibitors, C3-peptides, C3-ADPribosyltransferase, siRNA and ROCK inhibitors (Wu & Xu, 2016). Blocking of Rho activation using Cethrin (BA-210; SPRING-VX-210) has shown promise in preclinical and multicentre phase I/IIa clinical trials; some patients with one dose reported improvements in an ASIA grade (Fehlings et al, 2011). A further phase 2b/3 study determining efficacy and safety was initiated for patients with acute 6 of 29 EMBO Molecular Medicine 12: e11505 | 2020 ª 2020 The Authors traumatic cervical SCI , although regrettably it was discontinued shortly after due to lack of efficacy after an interim analysis (Taylor WEBARTICLE, 2018). Similarly, inhibition of the downstream effector Rho kinase (ROCK) by Y27632 and Fasudil in mouse models of SCI resulted in similarly inefficacious effects of RhoA inhibition when delivered acutely (Borisoff et al, 2003;Chan et al, 2005;Watzlawick et al, 2014). While still an intriguing target, one key reason for the unsuccessful trial is that we still understand relatively little about the mode of action of Rho blockade, both for neuronal and non-neuronal cells. We have only recently gained a deeper understanding of the physiological role of RhoA in axon growth during development and its downstream effectors. RhoA restrains axon initiation through RhoA/myosin-II-dependent actin arcs that restrict microtubule advancement in the growth cone . Future research needs to further disentangle the physiological role of RhoA in axon regeneration and how RhoA-dependent mechanisms in glial cells might affect axon regeneration.
The cytoskeleton plays an important role in axonal growth in development and regeneration. Microtubules underlie the formation of retraction bulbs and are a major contributor to the failure of axonal regeneration; stabilisation of microtubules specifies growth cone formation and protrusion (Erturk et al, 2007;Witte et al, 2008). Stabilisation of microtubules also prevents polarisation and migration of scar-forming fibroblasts which is associated with decreased deposition on CSPGs (Ruschel et al, 2015). Independent laboratories have shown that systemic administration of low doses of the microtubule-stabilising drugs epothilone B or D after SCI results in adequate CNS penetration and distribution, reduced deposition of inhibitory proteoglycans, reduced axon die-back, induced axonal growth and improved motor functions after SCI (Ruschel et al, 2015;Sandner et al, 2018). Considering these drugs are FDA-approved, easily administered and address both intrinsic and extrinsic determinants of axon regeneration, they represent a promising therapeutic candidate. Further evidence that implicates microtubule dynamics as important regulators of axon development and regeneration comes from studies investigating collapsin response mediator protein 2 (CRMP2). This protein stabilises microtubule polymerisation, while CRMP2 phosphorylation loses its affinity for cytoskeletal proteins, leading to microtubule disorganisation inhibition of axonal growth (Nagai et al, 2016). CRMP2 inactivation can be mediated via a RhoA-ROCK-dependent pathway downstream of CSPGs, MAG and Nogo (Rozes Salvador et al, 2016;Curcio & Bradke, 2018). Genetically preventing phosphorylation of CRMP2 promotes axonal regeneration attributed to suppressing microtubule depolymerisation after optic nerve injury (Kondo et al, 2019). Pharmacologically preventing phosphorylation of CRMP2 could be another therapeutic avenue.
Less is known about the role and manipulation of the actin cytoskeleton in axon regeneration. Recently however, the ADF/ cofilin family of actin-regulatory proteins that govern actin retrograde flow and dynamic were shown to promote neurite formation (Tedeschi et al, 2019). Specifically, enhanced actin turnover by ADF/cofilin is critical for axon regeneration in the adult CNS. Thus, actin turnover is a key target for future therapeutic interventions.
We are missing key pieces of the experimental puzzle such as what triggers the switch between growth competent neurons to growth-restricted mature neurons. Knowing such mechanisms could open a door to new therapeutic interventions. Whole transcriptome sequencing and bioinformatics analysis identified the calcium voltage-gated channel auxiliary subunit a2d2 (Cacna2d2) as a one such developmental switch (Tedeschi et al, 2016). Cacna2d2 promotes synapse formation and limits axon growth and regeneration in adult mouse DRG neurons (Tedeschi et al, 2016). Systemic administration of pregabalin to antagonise Cacna2d2 in adult mice after dorsal column lesion resulted in increased axon regeneration by preventing calcium influxes through Cav2 channels (Tedeschi et al, 2016). Further, gabapentinoid administration within a month after human SCI improves motor recovery (Warner et al, 2017), while other anticonvulsant drugs appear not to show such effects (Warner et al, 2019). Many different interventions are available that influence the intrinsic growth response of axons which seem to represent only a piece of the puzzle. We are yet to discover how to revert neurons to their embryonic growth state.
Resupply: trophic support Neurotrophins facilitate neuron survival, development and function. They may be administered to the damaged spinal cord by direct infusion, biomaterials and ex vivo gene therapy (cell transplantation) or by viral vector expression. Brain-derived growth factor (BDNF), neurotrophin-3, (NT-3), nerve growth factor (NGF), fibroblast growth factor (FGF) and glial cell-derived growth factor (GDNF), for example, are some of the trophic factors that have been investigated. Different neurotrophic factors have differing effects. For example, NT-3 elicits growth of CST axons, and NT-3 and NGF both promote extensive DRG-origin sensory or axon from the reticular formation and red nucleus (Houweling et al, 1998;Jakeman et al, 1998;McTigue et al, 1998;Oudega & Hagg, 1999;Namiki et al, 2000;Tuszynski et al, 2002;Blesch & Tuszynski, 2003;Shumsky et al, 2003;Zhou & Shine, 2003;Oudega et al, 2019). BDNF, by contrast, has no such effect but rather stimulates sprouting and growth of rubrospinal, reticulospinal, vestibulospinal, raphespinal and motor axons (Nakahara et al, 1996;Tuszynski et al, 1996;Grill et al, 1997;Bradbury et al, 1998;Bregman et al, 2002). There are, however, several issues pertaining to continuous infusions of trophic factors, including damage to the tissue at the infusion site, low stability, limited diffusion and inability to cross the bloodspinal cord barrier. Furthermore, the complexity of the spinal networks conveys another limitation of the use of neurotrophic factors. For example, overexpression of NGF in the dorsal horn resulted in overshooting of targets and formed inappropriate connections, resulting in severe hyperalgesia in rats (Tang et al, 2004). Similar side effects or lack of efficacy have been observed after trophic delivery in human clinical trials of diabetic neuropathy (Apfel, 2002).
Future preclinical investigations should include non-human primates that include multiple motor and sensory parameters to show whether neurotrophins safe and effective for SCI treatment. It would be invaluable to discover the required combination and temporal administration of growth factors to elicit regrowth of axons through lesions without side effects. Recently, the essential factors required to propel propriospinal axon regeneration across a complete SCI in the adult mouse were discovered . The growth capacity of mature descending propriospinal neurons was activated with viral vector-mediated expression of osteopontin, insulin-like growth factor 1 (IGF1) and ciliary-derived neurotrophic factor (CDNF) before a SCI. A growth-permissive substratum was induced with fibroblast growth factor 2 (FGF2) and epidermal growth factor (EGF), and propriospinal axons were chemoattracted with glial-derived neurotrophic factor (GDNF) delivery via spatially and temporally controlled release from biomaterial depots placed sequentially after SCI . These three mechanisms in combination stimulated propriospinal axon regrowth through astrocyte scar borders and across lesion cores of non-neural tissue that was associated with increased electrophysiological recordings in tissue below the lesion but did not result in functional improvements. It may be possible that functional recovery in these animals may only be evident if combined with rehabilitation to elicit use-dependent plasticity.

Remyelination
The validity of targeting remyelination after SCI is contentious. The myelin sheaths of myelinated axons increase the speed of transmission of action potentials. The importance of this property is overtly evident in multiple sclerosis; however, researchers have recently questioned whether the contribution of oligodendrocyte support and remyelination in pathophysiology is relevant to functional recovery after SCI (Duncan et al, 2019). After SCI, there is widespread acute oligodendrocyte death and demyelination of axons and it is predicted that remyelination could be important for protecting axons from further degeneration and enhancing conduction (Norenberg et al, 2004;Nave, 2010aNave, , 2010bAlmad et al, 2011). Two approaches are available: transplantation of cells that may directly differentiate into oligodendrocytes and to promote the recruitment and differentiation of endogenous OPCs. Several studies from various laboratories involving transplantation of Schwann cells, OECs and NPCs have demonstrated functional recovery correlates to their ability to promote remyelination (Tetzlaff et al, 2011;Karimi-Abdolrezaee & Eftekharpour, 2012). Within these experiments, the contribution of remyelination towards functional recovery is unclear as transplanted cells have many potential mechanisms of efficacy (see 3.2. Replacement: Cellular transplantation). Failure of recruitment of OPCs to the SCI lesion site is associated with the inability of axon remyelination and sensitivity to degeneration (Irvine & Blakemore, 2008). Promoting the recruitment and differentiation of OPCs to the lesion site can is a lesser-explored avenue (Duncan et al, 2019). Various growth factors and hormones can stimulate the differentiation of OPCs in vitro but their in vivo efficacy is unclear (Plemel et al, 2014). More recently however, spontaneous locomotor recovery of stepping following contusive SCI was shown to not require oligodendrocyte remyelination despite the high level of endogenous remyelination that occurs after SCI (Duncan et al, 2018). This finding raises doubts whether therapeutic targeting of oligodendrocyte remyelination is a worthwhile target for clinical translation, a debate we cannot yet conclude at this time. The contribution of Schwann cell myelination may possibly be a more meaningful target. Neuregulin-1 was found to drive trans-differentiated of OPCs into Schwann cells, and these cells were essential for Schwann cell-mediated remyelination following SCI (Bartus et al, 2016). Furthermore, preventing neuregulin-1/ ErbB signalling-controlled transformation of CNS progenitors into myelinating Schwann cells resulted in negative consequences for functional recovery following spinal cord contusion in mice (Bartus et al, 2019). Another consideration is interspecies differences. The majority of intact axons in chronically injured rats exhibit remyelination, albeit with some structural abnormalities (Powers et al, 2012). Neither was chronic demyelination observed in spared axons after SCI in mice (Lasiene et al, 2008). It is unclear whether this is also the case in humans and primates or whether remyelination contributes to the same functions as it does in rodents, owing to the need for more investigations on this topic.
Rehabilitation: neurorehabilitation, neuronal activity and electrical stimulation Recent evidence has documented that, in the right context, rehabilitation can play a critical role in regeneration and plasticity (Loy & Bareyre, 2019). Various forms of rehabilitation can enhance the recovery of motor functions after SCI in mice, rats and cats. Rehabilitation is important to prune and refine circuits that are relevant to the motor task (Ichiyama et al, 2008). Moreover, rehabilitation is also reported to stimulate the local production of neurotrophins (Ying et al, 2005;Cote et al, 2011b), modulate multiple neurotransmitter systems (Edgerton et al, 2004) and enhance sprouting of compensatory relay networks (Courtine et al, 2009;van den Brand et al, 2012). Animal models of rehabilitation can come in multiple forms of forced or voluntary physical exercise and environmental enrichment as well as on a spectrum of general to more specific approaches. Increasing general activity of rodents through environmental and social enrichment such as wheel running, multiple housing and the addition of climbing spaces, improved bedding and crawl toys all significantly increase BBB locomotor scores (Berrocal et al, 2007;Loy et al, 2018). Surprisingly, environmental enrichment prior to a SCI in mice and rats mimicked the degree of proprioceptive afferent neurons regeneration and functional recovery to a conditioning lesion (Hutson et al, 2019). The regenerative response correlated with Creb-binding protein (Cbp)-mediated histone acetylation and epigenetic changes inducing signalling pathways involved in neuronal activity, calcium modulation and the regenerative programme in proprioceptive neurons (Hutson et al, 2019). A comparative analysis of rats with incomplete SCI subjected to three forced-exercise rehabilitation paradigms (quadrupedal treadmill training, swim training and stand training) differentially improved recovery after spinal cord injury (Hutchinson et al, 2004). Quadrupedal treadmill-trained rats showed reduced allodynia and restored sensation; swim training reduced allodynia and stand training had no benefit. This indicates that stand training may not be intensive enough above spontaneous movement. Similarly, quadrupedal treadmill training was superior to hindlimb step training in terms of electrophysiological and kinematic limb analysis and the number of propriospinal labelled neurons above and below the lesion (Shah et al, 2013). Utilising task-specific rehabilitative strategies such as skilled paw retrieval, horizontal ladder training or hindlimb-specific training makes intuitive sense in that the best way to relearn a given task is to specifically train for that task. For example, hindlimbspecific treadmill training can be used after thoracic contusion to significantly improve that lost function although this negatively influenced forelimb-hindlimb inter-coordination (Griffin et al, 2020). It is apparent that emphasising a particular behaviour can negatively affect the performance of another, which suggests that reassignment of CNS resources is limited (Fawcett & Curt, 2009). For example, rats trained at skilled paw reaching improved in that task but this worsened their performance at walking on ladders, and general environmental enrichment rehabilitation can improve 8 of 29 EMBO Molecular Medicine 12: e11505 | 2020 ª 2020 The Authors locomotor function, but extinguish skilled tasks (Girgis et al, 2007;Garcia-Alias et al, 2009). Likewise, spinally transected cats could be trained in weight support or step; however, training one behaviour would extinguish the other (De Leon et al, 1998a, 1998b. We are still rudimental in our understanding of how rehabilitation influences regeneration, neuroplasticity and motor recovery following SCI. Several studies reported exercise-induced plasticity of various spinal tracts, particularly the serotonergic and reticulospinal tracts (Engesser-Cesar et al, 2007;Asboth et al, 2018;Loy et al, 2018). Conversely, an equalling number reported no such ability and proposes the role of rehabilitation in shaping newly formed connections by spontaneous mechanisms or plasticity-promoting treatments (Garcia-Alias et al, 2009;Maier et al, 2009;Wang et al, 2011a;Alluin et al, 2014). Regardless of these contrasting reports, it is clear that in general, a high-intensity programme and is key to providing robust motor function improvements in rats (Loy & Bareyre, 2019). It should be briefly mentioned that very limited recovery after anatomically complete injuries with rehabilitation is observed (Ilha et al, 2011). However, since the spinal circuitry below the lesion remains active, it can be activated through pharmacological and electrical modulation of the central pattern generator and lead to weight-bearing stepping in rats with completely transected spinal cords (Courtine et al, 2009).
We are also rudimental in our understanding of how electrical stimulation and neuronal activity affect regenerative processes and investigations are ongoing. Brief electrical stimulation accelerates axon outgrowth in models of peripheral nerve injury (Gordon, 2016), and spinal cord stimulation in human SCI patients has shown efficacy in randomised clinical trials over several decades (Sdrulla et al, 2018). Exposure to an electrical field in vitro increased neurite outgrowth of embryonic and adult neurons, and electrical stimulation of the sciatic nerve in vivo increased central DRG axon regeneration comparable to a conditioning lesion (Wood & Willits, 2006;Udina et al, 2008). Conversely, electrical stimulation has also been associated with axon growth inhibition. Electrical activity strongly inhibited axon outgrowth in cultured adult DRG neurons, an effect depended on the L-type voltage-gated Ca 2+ channel current and involved transcriptional changes (Enes et al, 2010). Knockout of the L-type voltage-gated Ca 2+ channel current in adult mice was sufficient to boost the growth ability of DRG neurons in vivo after central lesioning (Enes et al, 2010). Likewise, suppressing synaptic function through gabapentinoid-mediated blockade of the voltage-gated channel auxiliary subunit a2d2 also induces axon regeneration the adult CNS (Tedeschi et al, 2016). This counter-position hypothesises that electrical stimulation suppresses growth by triggering an increase in intracellular calcium, and conversely, a lack of electrical activity within axotomised neurons may recapitulate development and promote axon regeneration (Hilton & Bradke, 2017). It is possible that temporal regulation and the effects of conditioning neurons may account for the contrary evidences and is therefore an exciting prospect for further investigations.

Combinatory approaches
While individual therapies have shown promise in preclinical models, recovery remains incomplete and it is becoming apparent that focusing a singular barrier to repair is not going to facilitate successful and functional recovery of SCI in humans. Conversely, it will require the combination of various treatment strategies that address diverse aspects of SCI pathology. Removal of growth inhibitory molecules as well as boosting the intrinsic growth response will be required. Tissue or cell transplants may be required to replace tissue, act as a permissive bridge and/or facilitate repair. Supplying trophic support will improve survival of cells and direct axonal growth. Lastly, neurorehabilitation will function to augment the functional recovery of remodelling circuits (Tables 1-5). This section will summarise currently published studies that have combined therapies in SCI models (Fig 2).

Combining cellular transplants with neurotrophins
Early attempts at transplantation of cells into the damaged spinal cord promoted some degree of axonal regeneration into the graft. However, axons fail to grow beyond the graft, transplants typically lack appropriate orientation, and the hostile adult spinal cord environment limits transplant survival and/or differentiation. The logical addition of neurotrophins was one of the first combinatory strategies to address these problems and many studies have since been conducted over the last three decades ( Table 1). The first attempt at this strategy utilised Schwann cells in semipermeable guidance channels in combination with BDNF and NT-3 delivery via a minipump in a thoracic transection model (Xu et al, 1995a). BDNF and NT-3 infusion enhanced propriospinal axonal regeneration and, more significantly, promoted axonal regeneration of specific distant populations of brain stem neurons into grafts at the mid-thoracic level in adult rat spinal cord. In the following years, many studies reported that combining growth factor delivery with a PNG is capable of allowing axons to extend beyond the graft, although this was often not associated with improvement to behavioural measures (Cheng et al, 1996(Cheng et al, , 2004Ye & Houle, 1997;Dolbeare & Houle, 2003). This may reflect a lack of beneficial integration of the cells and a limitation of PNGs over other transplant types. Many combinatory strategies for PNGs have utilised FGF (Table 1). It appears that FGF is an important growth factor to include with cellular transplantation. Firstly, FGF tends to stimulate axonal elongation in a rectilinear rather than a branching pattern, which is critical for long-distance regrowth. Secondly, FGF tends to act on reactive astrocytes to cause them to change their shapes into a more primitive bipolar morphology, which is critical to allow axons to pass the graft-host interface (Zhou et al, 2018).
It is apparent that the transplant and growth factor delivery methods are important considerations and they need to be spatially and temporally organised. Cell and drug delivery by injection or loaded gelfoam are substandard due to washing away of cells and short-lived release of the factors, respectively. Osmotic pumps ensure long-lasting local delivery but are invasive and can cause local tissue damage. Ex vivo gene therapy delivery of cells to secrete growth factors is limited for the recurring reason that the high concentration gradient of growth factors at the transplant site does not allow axons to leave the graft (Table 1; Nakahara et al, 1996;Blesch et al, 2004;Mitsui et al, 2005). This may possibly be overcome by further expression of growth factors beyond the graft such as what was observed after lentiviral NT-3 administration beyond a bone marrow stromal cell transplant (Taylor et al, 2006 spatial expression of specific growth factors but is limited in that continued delivery after resolution is likely to be disadvantageous. The development of various natural and synthetic polymer biomaterials is advancing as a promising way to circumvent the limitations of the other delivery methods although they each also have benefits and limitations (Fuhrmann et al, 2017). Fibrin matrix-delivery of two human ESC-derived NPCs with a cocktail of ten growth factors after severe SCI allowed for extensive axonal outgrowth of the grafted cells rostral and caudal from the transplant and the combination improved functional recovery . This experiment was later repeated using human IPSCs. A similar level of extensive axonal outgrowth from the grafted cells was present, although in this case there was no improvement to motor functions which may imply an inability of IPSCs to integrate into the hose tissue to form functional connections (Table 1; Lu et al, 2014). Extensive axon growth is not always beneficial as axons may overshoot and oversupply targets, leading to spasticity, neuropathic pain and worsened motor outcome. Another issue of the combination of transplants with growth factors is that growth factors may enhance uncontrolled differentiation of NSCs into astrocytes which may lead to allodynia or pluripotent cells to overproliferate or form tumours (Hofstetter et al, 2005). For example, in a subset of animals, the combination of ESC-derived NPCs combined with platelet-derived growth factor (PDGF) and NT-3 led to tumour formation by 8 weeks after transplantation (Johnson et al, 2010a). Therefore, terminally differentiated cells such as Schwann cells or OECs may be a more suitable option. There is currently no gold standard cell type or growth factors, and combinations of the two for clinical translation and so it may be short-sighted to progress with clinical trials before this information is attained.

Combination of cellular transplants with anti-inhibitory therapies
Removing inhibitory elements through various means in conjunction with cellular transplantation has been investigated to enable regeneration in a more permissive environment (Table 2). Through such interventions, regenerating axons are capable of exiting the graft and are likely to form functional synaptic contacts. The first of these studies was in the form of administration of the IN-1 antibody (anti-Nogo-A antibody) with human Schwann cells with aFGF-fibrin glue after acute thoracic transection in rats. Inclusion of the antibody and aFGF were necessary to support sprouting and reduce axon die-back (Guest et al, 1997). Studies utilising ChABC later became a major focus in subsequent attempts that led to axonal growth beyond the grafts of Schwann cells, PNGs, OECs and NPCs (Chau et al, 2004;Houle et al, 2006;Vavrek et al, 2007;Tom & Houle, 2008;Alilain et al, 2011;Lee et al, 2013;DePaul et al, 2017). Unlike the combination of PNGs with neurotrophins, combinations with PNGs with ChABC appear to have much more pronounced    . Rather, growth factors do play an important role in the survival of the graft. Co-treatment of Nogo receptor blocking peptide (NgR) with delivery of bFGF, EGF and PDGF with adult NSCPs could overcome substantial cell loss; the growth factors increased the survival of the transplanted cells while NgR had no effect on their survival, and NgR increased axonal regeneration and myelination (Guo et al, 2012). In another study, only the combination of NPCs with ChABC and the growth factors EGF, bFGF and PDGF-AA could significantly improve neurobehavioural recovery without enhancing aberrant synaptic connectivity (Table 2; Karimi-Abdolrezaee et al, 2010 ). Yet, all of these results were achieved from infusions or microinjections of ChABC enzyme which has suboptimal stability and activity in vivo compared to gene therapy approaches (Lee et al, 2010). The combination of the dox-inducible LV-mChABC vector with other therapies is yet to be reported. It would be interesting to combine such tuning of plasticity and delivery of growth factors.

Combining anti-inhibitory therapies with neurotrophins
Combing neurotrophins with anti-inhibitory therapies was conceived to promote axonal growth in a newly permissive environment (Table 3). The first attempt used ChABC with NT-3-loaded lipid microtubes embedded in agarose gel following T10 dorsal hemisection in rats (Lee et al, 2010). Animals treated with ChABC in combination with sustained NT-3 delivery showed a significant improvement in locomotor function and enhanced growth of cholera toxin B subunit-positive sensory axons as well as sprouting of serotonergic fibres. This result was replicated by an independent group in which animals receiving combined therapy displayed the most improved body stability and inter-limb coordination (Garcia-Alias et al, 2011). Only animals with the full combination treatment recovered consistent multisynaptic responses in these motor neurons indicating formation of a detour pathway around the injury site. Similar to ChABC, anti-Nogo-A antibody in combination with NT-3 delivery by nanoparticles in an adult rat clip compression model led to increased anatomical improvements in both treatments individually but only functional improvements in the combination group (Elliott Donaghue et al, 2016). This combinatory approach may be limited in the context of complete transection injuries since increasing neuronal plasticity through anti-inhibitory therapies largely relies on spared circuitry and the formation of relay networks . For example, in a model of complete transection, co-treatment of Nogo receptor blocking peptide (NgR) with delivery of bFGF, EGF and PDGF with adult NSCPs did not have an effect on functional recovery or regeneration (Table 3; Guo et al, 2012). The growth factors, not the NgR, improved survival of the transplanted cells, while the NgR enhanced axonal regeneration and myelination. However, this is not always the case. Impressively, the combination of ChABC with neurotrophin delivery has also reported to be therapeutic in complete transection models in which ChABC typically has a very limited therapeutic benefit. ChABC and NGF delivery from an electrospun scaffold following thoracic complete transection in rats displayed a significant improvement in BBB functional recovery and that the electrospun matrix is an important factor in this observation (Colello et al, 2016). Likewise, antagonism of Nogo, EphrinB3 and sema4D receptor combined with the growth factors BDNF, NT-3 and cAMP administration after complete transection reduced the volume of cavitation, facilitated axonal regeneration and promoted neuronal generation leading to enhanced locomotion recovery (Table 3; Li et al, 2016). How does this occur? In the latter study, it was argued that new-born neurons generated at the lesion area could form the neuronal relay and enhance the locomotion recovery, but this is yet to be confirmed. There clearly are many variables between these experiments that complicate the interpretations of the outcomes, but further investigations and replications should result in more clarification.
Combination treatments involving targeting the intrinsic growth response A less explored area of research is enhancing the regenerative programme of injured neurons through targeting intracellular pathways and mechanisms as a combinatory therapy. Excitingly, this could prove to be a promising avenue for recovery after SCI (Table 4). Activation of cAMP signalling by rolipram and further cAMP injections combined with Schwann cells transplantation to the injury site following spinal cord contusion in rats further increased axonal sparing, myelination and locomotor function compared to the individual components (Pearse et al, 2004b). It was apparent that rolipram was responsible for most of the beneficial effects and the transplants alone had a minimal effect (only significant improvements to myelination). This suggested that growth factors might be required to promote axonal growth through the graft. Indeed, rolipram combined with the further addition of modified Schwann cells to secrete a bifunctional neurotrophic molecule (D15A) with both NT-3 and BDNF activity increased numbers of axons originating from the reticular formation below the lesion/ implant area and provided the greatest improvement in hindlimb function (Flora et al, 2013). Similar to the previous study, rolipram was the main contributor to the beneficial effects compared to a minor improvement observed by BDNF + NT-3. This indicates that activating the intrinsic regenerative programme is an approach that may be superior to cell-or growth factor-based therapies (Table 4). Microtubule stabilisation is fast becoming an attractive therapeutic option for promoting intrinsic regenerative capacity as well as limiting the deposition of inhibitory molecules in the injured CNS. Recently, enhancing and directing the growth of docetaxel-stabilised microtubules by the co-treatment of the growth factors BDNF and aFGF in liposome hydrogels in a rat thoracic contusion injury model was achieved . This application supported regeneration by improving neuronal survival and plasticity, and rendering a more permissive extracellular matrix environment, leading to significantly improvement of BBB score points compared to untreated controls . Co-administration of taxol with the EGFR signalling antagonist cetuximab verified therapeutic effects towards inhibition of scar deposition and promotion of neuronal differentiation, axonal outgrowth and functional recovery in a severe SCI model in rats (Table 4; Fan et al, 2018). The precise role of cetuximab in this scenario is unclear, although it was demonstrated to promote the differentiation of injury-activated endogenous-NSC differentiation into neurons. There is clearly need for more investigations to continue in this area of research. The CRMP family of proteins regulate aspects of neurite growth through stabilising microtubules and phosphorylation of CRMP2 by GSK3b reduces CRMP2's binding affinity for tubulin heterodimers leading to microtubule depolymerisation (Fukata et al, 2002). Interestingly, preventing CRMP2 phosphorylation by pharmacological inhibition of GSK3b was found to be a critical mediator of the neurotrophic response in DRG neurite outgrowth (Nagai et al, 2016). This suggests that targeting actin dynamics may also potentiate the efficacy of neurotrophic delivery which could be an attractive strategy.

Combinations of therapies with rehabilitation
Despite the many efforts made by patients, physiotherapists and trial designers, rehabilitation as a treatment for SCI achieves a modest recovery at best. Could combining rehabilitation with therapies that aim to restore intrinsic growth capacity and/or remove inhibitory factors be more effective? Plasticity-promoting therapies including ChABC can restore neuroplasticity to a level it appears in early childhood and promote compensatory sprouting of spared fibres to form new neural connections. Yet, there is frequently a mismatch between anatomical and functional recovery. For example, treatment of rat cervical spinal cord injuries with ChABC produced a modest recovery in CST function, as measured by skilled paw function despite a large observable increase in axonal regeneration and neuronal plasticity (Garcia-Alias et al, 2008). Are new neuronal connections ineffective if the host does not know how to use them? Promoting plasticity by itself may not be sufficient to promote functional recovery if it leads to random new connections. The formation of appropriate connections in the spinal cord and brain may need to be driven by appropriate rehabilitation. Indeed, subsequent studies have shown that ChABC treatment combined with rehabilitation leads to enhanced recovery through opening a window during which rehabilitation becomes more effective in the newly plastic environment (Fawcett & Curt, 2009;Garcia-Alias et al, 2009;Wang et al, 2011a;Shinozaki et al, 2016). Therefore, utilising appropriate rehabilitation regimes is emerging as a promising avenue for shaping the new circuits created by therapies into useful new connections (Table 5). However, the timing of the treatment and rehabilitation appears to be critical. In one experiment involving thoracic spinal cord contusion in rats, the animals developed an abnormal stepping pattern when rehabilitation was included in parallel with the treatment of anti-Nogo-A antibody (Maier et al, 2009). Similarly, synchronous treatment of anti-Nogo-A antibody with rehabilitation in a model of stroke resulted in "chaotic hyperinnervation" leading the researchers to hypothesise that rehabilitation at an early stage was reinforcing erroneous connections (Wahl et al, 2014). Instead, sequential rehabilitation regimes are astonishingly effective and in some cases can almost completely restore specific motor tasks (Table 5; Garcia-Alias et al, 2009;Chen et al, 2017). More recently, endogenous CSPG degradation by an astrocyte-selective AAV-ADAMTS4 gene therapy combined with hindlimb rehabilitation enhanced functional recovery following rat thoracic SCI compared to either intervention alone (Griffin et al, 2020).
ª 2020 The Authors EMBO Molecular Medicine 12: e11505 | 2020 Current reports for the role of rehabilitation with cell transplantation are limited to one preclinical study which enlisted autologous bone marrow stem cells with swim training where a difference to functional parameters between groups was observed (Carvalho et al, 2008). As we discussed above, considering swim training is less effective than high-intensity quadrupedal training, it would be  Experimental spinal cord injury research has resulted in a multitude of individual therapeutic possibilities yet recovery after injury remains incomplete. Combinatory approaches to address the seven targets will need to be a focus of translation studies. Tissue and cellular transplants will replace lost cells, among other regenerative functions. Removal of inhibitory factors such as CSPGs allows for enhanced axonal growth. Targeting neuron-intrinsic mechanisms enhance intrinsic regenerative response which can then be directed through resupply of trophic support. Remyelination of demyelinated axons may improve axonal conduction and survival. Finally, rehabilitation functions in circuit remodelling and strengthens beneficial connections.
ª 2020 The Authors EMBO Molecular Medicine 12: e11505 | 2020 of interest to test cellular transplantation with other regimes of rehabilitation. Rather, combining rehabilitation with cell transplants in human has progressed immediately to trials (Table 6). These trials have demonstrated that cell transplantation into the injury site seems to be safe and appears to only show beneficial effects for patients when combined with rehabilitation further highlighting the importance of combinatory approaches in the clinical setting (Li et al, 2015;Zhu et al, 2016;Anderson et al, 2017). Similarly, current reports for the role of rehabilitation with neurotrophic factors are limited to two preclinical studies. First, acute AAVmediated expression of NT-3 combined with sequential spinal electromagnetic stimulation rehabilitation, swim training and exercise ball training improved electrophysiology recordings, narrow beam task, error ladder task and several Catwalk gait analysis parameters (Table 5; Petrosyan et al, 2015). This result was only observed with the combination of electrical stimulation plus exercise or electrical stimulation plus exercise plus AAV-NT3, whereas any singular therapy had no effect. This indicates that electrical activity in the spinal cord may be a requirement for activity-induced neuroplasticity. Second, intrathecal administration of ChABC, PDGF, bFGF and EGF combined with daily quadrupedal treadmill training leads to enhanced neuroanatomical plasticity of CST and the serotonergicspinal pathway, although these changes failed to translate to functional improvements (Alluin et al, 2014). Considering there are many variables between these two studies, at the stage it is difficult to draw larger conclusions. Rehabilitation is limited in complete injuries, yet considering neuronal networks below the lesion site remain active, combining treadmill rehabilitation with neurotransmitter administration and electrical modulations in the lumbar spinal cord can lead to the re-establishment of walking ability in completely transected rats, a major step towards future therapy for completely paralysed patients (van den Brand et al, 2012). The use of electrical stimulation with rehabilitation in humans has been practised in the clinic for many decades. This can be achieved by functional electrical stimulation (FES; on the skin), epidural electrical stimulation (EES; stimulation of the spinal cord directly). FES has recently made large strides towards helping people with SCI. A major breakthrough showed that EES in combination with intense rehabilitation and neurotransmitter administration could help to restore walking function in three individuals with varying levels of incomplete SCIs (Wagner et al, 2018). This may be the result of the precise timing of the stimulation rather than continuous stimulation which may facilitate the naturally occurring signals. However, this study did not include a group of rehabilitation without stimulation and it is possible that the highintensity orthotic-assisted rehabilitation programme is the main therapeutic component.

Towards clinical translation: current status
Many therapeutic strategies have been discussed in terms of their potential to address the human condition since their conception, yet relatively few SCI clinical trials have been completed (Table 6). For those that have reached clinical trials, none have reached general clinical application, aside from some countries that offer medical tourism for non-validated therapies. Trials to overcome the devastating effects of SCI can be categorised into neuroprotective and regenerative approaches. Unfortunately, neuroprotection trials for human SCI have largely failed (Kim et al, 2017;Santamaria A.J., 2017). For example, trials from the neuroprotective interventions methylprednisolone, GM-1 and gacyclidine failed to show sufficient statistical evidence for efficacy to be approved for clinical use (Table 6). Neuroregeneration/repair trials have emerged more recently, yet only a small number of treatments have reached the stage of human testing. These include anti-Nogo-A antibody therapy and Cethrin, and various cell-based therapies, which have each completed phase I/IIa trials with no apparent adverse effects and some indication of functional improvements (Fehlings et al, 2011;Kucher et al, 2018). Trials of cell transplant therapies remain controversial in terms of ethical, safety and efficacy, but there is, at least, evidence that transplants may improve some sensory/motor and bladder functions (Fan et al, 2017). An exciting future prospect is for the human application of ChABC, which is yet to reach clinical trial; however, the evidence from the recent trials of SCI in dogs and rhesus monkeys supports ongoing preclinical efforts (Hu et al, 2018;Rosenzweig et al, 2019). The drive towards translating the promising results of ChABC to being clinically available is currently led by a collective team named the CHASE-IT (ChABC for Spinal Injury Therapy) consortium. A focus of the consortium is now optimising the LV-ChABC gene therapy approach to making it more clinically safe by using an AAV vector instead and with gene regulation (International Spinal Research Trust WEBARTICLE, 2019). Such approach would allow for robust, prolonged and controllable expression of mChABC in the human spinal cord which may exceed the efficacy achieved by infusion of the native enzyme. We are on the cusp of observing auspicious therapeutic candidates entering human trials; however, the road towards their clinical translation is arduous, and many factors need to be considered.

Problems for translation
One major problem towards clinical translation is the lack of robustness and reproducibility in preclinical studies (Schaffran & Bradke, 2019). With this notion, the "Facilities of Research Excellence-Spinal Cord Injury" (FORE-SCI) programme was developed to replicate key studies in the SCI field (Steward et al, 2012). Out of 12 studies selected to be replicated, it is perhaps worrying that only one study was fully replicated, six were not and the rest had mixed results or were only partially replicated. Therefore, it is imperative that a confirmatory replication of key aspects of the original study by an independent laboratory is conducted before the therapy progresses to the clinical stage. In academic research however, replication studies are rarely performed due to a lack of funding, the absence of recognition for performing such studies or problems with publishing. This is something that needs to change for us as a community to progress. This will further become more complicated when multiple strategies are combined.
Combinatorial approaches are likely to require tailored experimental design and analytical frameworks. The complexity of the combinatory treatments may project deceivingly incremental increases in efficacy. It is difficult to discern the involvement of the individual treatment entity when there are many variables that plague the reproducibility between studies and between laboratories. Facing a slew of possible therapies and combinations of such, it is a good idea to wait until we have more data before trials are initiated. What is the best cell type and biomaterial for tissue replacement? What are the ethical concerns or potential for harm when obtaining these cells? What growth factor or combination of growth factors is the best? What is the best method for removing inhibitory factors? Is the removal of multiple types of inhibitory factors required? Is the experimental model relevant to the human condition that the therapy is intended for? The experimental model should as closely represent the human condition, fine hemisection manipulations, for example, do not. Contusion and compression injuries are certainly more clinically relevant, though this does not make them the best choice for every experiment. Another consideration is that the spinal cord of humans is much thicker, has different anatomy, and drug penetration is very different. It would be preferential for therapies to be tested in multiple models, ideally including a larger animal (primate or porcine), to broaden our understanding of the limitations of the treatment. Another major problem is that outcome measures of motor improvements in humans are relatively insensitive compared to animal behavioural tests. The ASIA impairment scale is outdated, and a 5-point scale ranging from full movement to no movement is underperforming. However, FORE-SCI groups are developing additional tests of sensory and motor function that should allow more sensitive testing of outcomes after SCI in clinical trials. Another important consideration is that non-motor system directly affects quality of life and often out-ranks in importance to patients (Simpson et al, 2012). Often multiple aspects of health or motor functions are not reported in studies and so we would not see these kinds of side effects that could manifest in humans. Careful analysis of alterations in not only the locomotor system but also sensory, autonomic and possibly cognitive function should also be assessed in relevant animal models prior to entering human clinical trials. The complexity of clinical trial designs featuring combination therapies will also be an important factor to take into consideration during therapeutic translation (Fawcett et al, 2007). Ultimately, preclinical researchers and trial designers need to communicate better and more often. Trials must be large enough to prove efficacy, reflecting the parameters and outcomes of the preclinical studies.

Conclusions
SCI results in many problems that need to be addressed in order to discover a cure for the condition. Many promising individual therapies have been developed and are progressing towards clinical trials. However, many in vivo studies fail to be replicated and often fail to translate clinically which is unacceptable to individuals with SCI anxiously awaiting therapeutic options. The reproducibility of results from promising candidates must be assured before progressing to human clinical trials. Many combinational strategies have demonstrated greater beneficial outcomes than the individual components alone by addressing multiple aspects of the pathology, but these need further characterisation to ensure meaningful efficacy beyond their individual components. Clinical trials in the future will eventually need to focus on combining the best therapies that aim to address as many of the problems as possible. The scenario will become increasingly complex as this happens and conversation between basic researchers and clinicians must happen to ensure accurate study design and functional readouts.
Pending issues (i) What is the ideal combination of treatments? (ii) How to assure the reproducibility of preclinical research? (iii) What is the efficacy of combinatory treatments in larger animal models? (iv) Are these feasible, safe and efficient for clinical translation? (v) How to proceed with clinical trials involving combinatory therapies?