Mitochondrial function in spinal cord injury and regeneration

Abstract

Many people around the world suffer from some form of paralysis caused by spinal cord injury (SCI), which has an impact on quality and life expectancy. The spinal cord is part of the central nervous system (CNS), which in mammals is unable to regenerate, and to date, there is a lack of full functional recovery therapies for SCI. These injuries start with a rapid and mechanical insult, followed by a secondary phase leading progressively to greater damage. This secondary phase can be potentially modifiable through targeted therapies. The growing literature, derived from mammalian and regenerative model studies, supports a leading role for mitochondria in every cellular response after SCI: mitochondrial dysfunction is the common event of different triggers leading to cell death, cellular metabolism regulates the immune response, mitochondrial number and localization correlate with axon regenerative capacity, while mitochondrial abundance and substrate utilization regulate neural stem progenitor cells self-renewal and differentiation. Herein, we present a comprehensive review of the cellular responses during the secondary phase of SCI, the mitochondrial contribution to each of them, as well as evidence of mitochondrial involvement in spinal cord regeneration, suggesting that a more in-depth study of mitochondrial function and regulation is needed to identify potential targets for SCI therapeutic intervention.

Introduction

Although the estimation of spinal cord injury (SCI) prevalence is not completely reliable, due to methodological problems, it is suggested that there are between 250.000 and 500.000 new cases worldwide every year [221]. The spinal cord is part of the central nervous system (CNS), and it contains nerves that receive sensory information and control the motor response [316]. Unfortunately, in humans, and mammals in general, the CNS is unable to regenerate and thus SCI is not reversible. These injuries not only compromise sensitive and motor capacities below the site of injury but are also accompanied by secondary complications as renal failure, respiratory problems, heart disease, neurological problems, and infections from untreated pressure ulcers, as well as depression and suicide [221]. All these complications affect the quality and life expectancy since people with SCI are 2 to 5 times more likely to die prematurely [221].

Currently, there is no available therapy for efficient recovery from SCI, the existing ones aim to reduce inflammation, pain, and the propagation of the initial damage, but generate poor functional improvements [221, 285]. This can be explained by the complexity and dynamics of the events, the lack of a full understanding of the factors that accompany these injuries, and the mammalian limited regenerative capacity. The damage produced by SCI is composed of two main phases: the primary injury, where the initial mechanical insult damages or destroys the tissue due to the hemostatic response and acute cell death, and the secondary injury, characterized by the propagation of the biochemical changes and extension of the injured area [110, 223, 236]. The primary injury is usually unexpected and almost impossible to intervene in, and therefore, the focus needs to be settled on the secondary injury, which implicates three different cellular phases: cell death and inflammation, cell proliferation and tissue replacement, and tissue remodeling [33]. A careful revision of the literature exposes that mitochondria play a pivotal role during SCI events at different levels: (i) excitotoxicity and reactive species, the main triggers of cell death after SCI, converge on mitochondrial dysfunction before neuronal and glial death [176, 224], (ii) cellular metabolic activity regulates immune cell phenotype and response, and inflammation [1, 135, 154], (iii) changes in the number and localization of mitochondria are involved in axonal regenerative capacity [155, 230], (iv) mitochondrial abundance and substrate utilization, regulate proliferation, differentiation, and neurogenesis, of neural stem progenitor cells (NSPC) [152, 153, 324].

In contrast to mammalian CNS, the mammalian peripheral nervous system (PNS) and the CNS of some non-mammalian organisms, such as teleost fishes, urodele amphibians, and larval stages of anuran amphibians, can regenerate after severe SCI [69, 167]. This ability provides interesting model systems to study the mechanisms underlying the successful regenerative processes, and a better understanding of why it fails in the mammalian CNS. Herein, we present a comprehensive review on the functions of mitochondria in SCI focusing on cell death, immune response and inflammation, axon damage, and neurogenesis. In addition, we highlight works demonstrating or relating a possible role of mitochondrial function in spinal cord regeneration.

Cellular response to spinal cord injury

The initial mechanical insult leads to spinal cord structural and functional loss, which is achieved by direct damage to blood vessels and cell membrane, leading to release of endothelial cells glycocalyx, vascular permeability, local hemorrhage, edema, ischemia, and loss of blood spinal cord barrier [194]. In addition, it causes direct disruption of axons, and oligodendrocyte death [101, 208]. Right after the insult, the damage is detected only in the injury epicenter, although, during the secondary injury phase, the damage spreads further both rostral and caudal to it [76, 101].

The three cellular phases involved in the secondary injury, including cell death and inflammation, cell proliferation and tissue replacement, and tissue remodeling are characterized by distinct biochemical events [73, 223, 236]. First, neuronal and glial cell death is induced by multiple factors, including excitotoxicity [224] and free radicals [61, 176], which converge on mitochondrial dysfunction followed by necrosis, apoptosis, or ferroptosis [103, 224]. Inflammation plays a dual role, on one hand, immune cells are recruited to the injury site where they remove cell debris, which has a negative impact on the regeneration of survival neurons,and on the other hand, immune cells release pro-inflammatory cytokines, free radicals, and metalloproteinase, increasing tissue damage [93]. Then, in the proliferative and tissue replacement phase, oligodendrocyte progenitors, astrocytes, and ependymal cells lining the central canal, which have in vitro neural stem cell potential [196], are induced to proliferate generating mostly astrocytes, which will form part of the glial scar, and oligodendrocytes in a slighter amount [19]. Finally, the tissue remodeling phase takes place, lasting from months to years, characterized by the formation of the glial scar, which is believed to have both protective and detrimental effects [144, 265]. It impedes a widespread infiltration of inflammatory cells, protecting from further neuronal loss and tissue degeneration [102, 315], but also it has been classically proposed to be a physical and biochemical impediment to axon regeneration [66], although the evidence presented by Anderson et al. (2016, 2018), showing that axons supplemented with growth factors can grow through the astrocytic scar and that components of this scar supports spontaneous axon regrowth, lead to a re-evaluation of its function [8, 9]. The events that are characteristics of these three cellular phases are described below in more detail, with a focus on the mitochondrial functional implications. Most of the studies have been made in rodents, thus, all following descriptions occur in rodents unless it is specified otherwise.

Cell death after SCI

In SCI, the damage generated by the initial impact triggers a cascade of events leading to increased cell death, such as the release of excess glutamate, which causes excitotoxicity, and free radicals, both resulting in mitochondrial dysfunction before cell death [73, 236]. In this section, we will describe the cell death events in mammalian and regenerative animal models, followed by a deepening on excitotoxicity and free radical-induced cell death, highlighting how they influence mitochondrial dysfunction. We will also discuss how the oxidative response and cell death are regulated in regenerative models, and finally, we will discuss how mitochondrial function and metabolism can be modified to improve SCI outcomes.

Cell death in mammals and regenerative animal models

Cell death temporality, and its differential effect in each cell type, have been widely characterized after SCI in rats. As early as 30 min after injury, necrotic cells are detected, while apoptotic cells appear 4–6 h post-injury (hpi) in the epicenter, and at 24 hpi in regions further away from the injury site [59, 180]. Neurons and oligodendrocytes present an early cell death pattern, with a peak at 8 hpi, while astrocyte death onset later and peaks at 24 hpi [101, 180]. Interestingly, neurons show a pathology more consistent with necrosis, while glial death is mainly apoptotic [101]. Neural progenitor cells are affected as well, as cell debris containing neural progenitor markers are observed in the epicenter at 24 hpi [117]. In addition, at 24 hpi, apoptotic oligodendrocytes associated with axons are also identified, which result in demyelinating axons, followed by axon degeneration, which can continue for weeks [59, 180]. One week after SCI, a second wave of cell death is observed, mainly in the white matter [180], and neuronal and glial apoptotic nuclei can be seen even 3 weeks after damage [59].

In regenerative animal models, cell death is more controlled, as it occurs during a shorter time. In hydra, the first apoptotic cells can be seen at 1 hpi, with a peak at 5 hpi, while the number of apoptotic cells start to decrease 8 hpi, and is almost absent at 16 hpi [49]. A similar pattern is observed in zebrafish,apoptotic cells appear at 6 hpi, with a peak at 24 hpi, while decreasing at 3 days post-injury (dpi), and are almost absent at 7 dpi [123]. Transcriptomic and proteomic analyses performed in our lab. suggest that the same is true for Xenopus laevis regenerative stages, as they show an upregulation of anti-apoptotic genes [169] and a decrease in pro-apoptotic proteins at 24 hpi [168].

In summary, mammals and regenerative model animals show a similar initial cell death profile, starting during the first 6 hpi, but differ in the duration and late profile, as mammals show a second wave of cell death, which may last weeks or months, and which begins while cell death almost disappears in regenerative models.

Excitotoxicity

Excitotoxicity has a prominent role during SCI, and it corresponds to a form of cell death triggered by cytotoxic levels of glutamate. High levels of glutamate lead to hyperactivation of its receptors, and a subsequent excessive calcium cellular influx, generating calcium-dependent enzyme activation and mitochondrial dysfunction, which results mainly in neurons and oligodendrocytes death. SCI generates neuron depolarization [181, 278] and microglial activation [279, 293], leading to glutamate release from both cell types, to levels over the physiological [174]. The over-activation of glutamate receptors [86] induces a massive influx of calcium [52, 186] (Fig. 1A), followed by amplification and propagation of the calcium signaling by the calcium-induced calcium release from the endoplasmic reticulum (ER), through the activation of ryanodine receptors (RyRs) and the inositol trisphosphate receptors (IP3Rs) [113, 242]. Excessive levels of calcium cause an over-activation of cytoplasmic calcium-dependent enzymes involved in mitochondrial dysfunction, necrosis, and/or apoptosis [222, 329] (Fig. 1A). For example, in neuronal culture exposed to glutamate, over-activation of nitric oxide synthase (NOS) occurs, causing an increase in nitric oxide (NO) production [7]. This increase in NO impairs the electron transport chain (ETC), specifically the activity of succinate dehydrogenase (complex II), cytochrome reductase (complex III), and cytochrome c oxidase (complex IV), leading to ROS generation [7, 29] (Fig. 1A; in neuronal cell lines, calcium-activated endonucleases that cause DNA fragmentation [139], or activation of purified cysteine proteases like calpains, which have an important role during apoptosis [58], are observed.

Fig. 1.

Mechanisms leading to cell death. A Schematic representation of excitotoxicity and oxidative damage-induced cell death in mammals after SCI. Red arrows indicate the effect of SCI over the different molecules leading to apoptosis, necrosis, or ferroptosis. The initial injury triggers the release of toxic levels of glutamate, from neurons and microglia to the extracellular space, which induce a massive influx of calcium into neurons, increasing NO production, and generating mitochondrial calcium overload. NO impairs ETC complexes in the mitochondria, leading to an increase in ROS and a decrease in ATP production, resulting in detrimental effects over mitochondrial dynamics, biogenesis, and function, converging in cell death. Reactive species released by microglia, blood vessels, and mitochondrial dysfunction, in addition to the glutamate-dependent decrease in GSH production and GPX4 activity, result in uncontrolled lipid peroxidation, whose products oxidate and nitrate proteins and nucleic acids, leading to ferroptosis. B Healthy and injured axons, magnifications show mitochondrial morphology in each case. The lightning symbolizes damage. SCI spinal cord injury; NO nitric oxide; ETC electron transport chain; ROS reactive oxygen species; GSH glutathione; GPX4 glutathione peroxidase 4

Mitochondrial calcium influx is needed for the neurotoxic effect, as decreasing mitochondrial calcium influx through several experimental strategies results in neuroprotection. Specifically, using mitochondrial uncouplers [270], blocking the outer membrane voltage-dependent anion-selective channel 1 (VDAC1) oligomerization [227], silencing the inner membrane mitochondrial calcium uniporter (MCU) [235], or inhibiting the ER calcium contribution to the mitochondria, by blockade of IP3Rs or antagonist of RyRs [242], protects against neurotoxicity. Oligodendrocytes are affected by mitochondrial calcium influx as well, and reduced oligodendrocytes death is observed when blocking VDAC1 oligomerization [227], or IP3Rs and RyRs function [243]. The elevated mitochondrial calcium, in addition to ROS, induces the assembling and opening of the mitochondrial permeability transition pore (mPTP), resulting in loss of mitochondrial membrane potential and swallowing of the matrix. These changes are followed by mitochondrial outer membrane permeabilization (MOMP), which leads to the cytosolic release of pro-apoptotic proteins, like cytochrome c and resulting in apoptosis [16, 200] (Fig. 1A), affecting the length of the lesion after SCI [268]. Therefore, the spinal cord initial injury induces over-physiological glutamate extracellular levels, resulting in excessive calcium influx to the cytosol and mitochondria, which in turn induce mitochondrial dysfunction and finally, cell death.

Free radical-induced damage

An early event after mammalian SCI is the generation of high levels of ROS [175] and reactive nitrogen species (RNS) [41, 318], which are responsible for the oxidative damage [103]. Oxidative damage markers, such as protein oxidation-related protein carbonylation (PC), protein nitration (3-NT), and lipid peroxidation-derived 4-hydroxynonenal (4-HNE), accumulate in rat spinal cord tissue samples, starting in the area of damage and followed by a later accumulation in neighboring areas, with a greater presence on the caudal side [41, 318], explaining the observed temporality and location of dying cells.

Both microglia [55] and neurons [116] produce superoxide radical (O₂⁻), which is further stimulated by mitochondrial dysfunction after SCI [116, 276] (Fig. 1A). In addition, SCI cause hemorrhage, with which hemoglobin and Fe³⁺ are released, prompting the generation of hydroxyl radicals (HO⁻) [177, 297] (Fig. 1A); and activation of NOS, both in glia (iNOS) and neurons (nNOS), leading to NO generation [29, 319] (Fig. 1A). NO can react with O₂⁻ and HO⁻, to produce the highly reactive peroxynitrite (PN) [104], which triggers lipid peroxidation on cell membrane polyunsaturated fatty acids (FAA) [103] (Fig. 1A). Lipid peroxidation is a self-propagating reaction process that can be controlled by the enzyme glutathione peroxidase 4 (GPX4), which uses glutathione (GSH) as a cofactor [32]. GSH production needs cysteine as a precursor, however, the high extracellular glutamate levels after SCI, inhibits the glutamate/cysteine transporter and thus, affects GSH production and GPX4 activity, resulting in uncontrolled lipid peroxidation [70, 320, 326] (Fig. 1A). The lipid peroxidation products, malondialdehyde (MDA), lipid peroxyl radicals (lipid ROS), 4-hydroxynonenal (4-HNE), and acrolein [78], together with RNS [17], induce protein carbonylation (PC) and nitration (3-NT) [17, 63, 103], leading to protein function impairment [115, 149], while RNS nitrate nucleic acids, inducing DNA breakage [61, 79]. All of these reactions lead to a specific cell death called ferroptosis. In addition, lipid peroxidation-induced cell membrane damage, increases the permeability for calcium ions and reactive species, further contributing to excitotoxicity [188, 193]. As neurons have a high content of polyunsaturated FAA in their membranes, they are especially sensitive to lipid peroxidation [254]. Studies using Gpx4 neuronal inducible knockout mouse showed that motor neurons in the spinal cord are the most affected by GPX4 deficiency and die by ferroptosis [48, 191].

Although mitochondria are one of the major sources of reactive species production, at the same time they are targeted by oxidative damage [7, 195, 262, 276]. Of great importance is the functional impairment of mitochondrial proteins, and metabolic enzymes [289]. For example, peroxynitrite in microglia primary culture [226] and hydrogen peroxide in guinea pig brain cortex synaptosomes [288], inhibit the activity of the Krebs cycle rate-limiting enzyme α-ketoglutarate dehydrogenase (α-KGDH) [289]. Similar results are observed when using purified enzymes, peroxynitrite induces nitration of α-KGDH [253], and 4-HNE modifies the binding site of an α-KGDH and pyruvate dehydrogenase cofactor, inactivating the enzymes [125]. In rats, it is suggested that ferroptosis induced by SCI causes mitochondria morphological adaptations such as shrunken mitochondria, reduced number of cristae, and lamellar phenotype [326]. In addition, oxidative damage can lead to metabolic changes associated with energy obtention, such as a decrease in ATP production observed in rat brain synaptosomes after traumatic brain injury [274], or fall in creatine kinase activity, which participates in the transfer of high-energy phosphates generating ATP in sites of high consumption of this nucleotide [34], in rat spinal cord tissue samples after SCI [4]. Finally, lipid peroxidation caused by β-amyloid inhibits glucose transport into cultured neurons by the binding of 4-HNE to the glucose transporter 3 [187]. Thus, reactive species generated by the initial injury, produce oxidative damage, which in turn induces mitochondrial dysfunction and triggers the production of more reactive species, increasing the extent of the oxidative damage, and ending up in cell death.

Controlled oxidative response and cell death in regenerative models

Even though the general and the more intuitive view is that cell death is a negative event after tissue damage, regenerative models such as hydra, Xenopus and planarian, have shown the opposite [27]. In hydra head amputation [49], and Xenopus tadpoles tail amputation experiments [291], apoptosis is necessary for induction of cell proliferation, and regeneration. Furthermore, the timing and amount of apoptotic cells are suggested to influence the regenerative capacity as well,apoptosis occurring during the first 24 hpi after Xenopus tadpole tail amputation is crucial, as inhibition of apoptosis during this period results in impaired regeneration [291], while planarian regeneration is affected when the number of apoptotic cells is increased by the downregulation of an anti-apoptotic gene [231]. All this evidence supports the idea that a timely and controlled event of cell death is required for proper regeneration.

The oxidative response is also regulated in regenerative models; a transcriptomic study of the regenerative response after tail amputation in Xenopus tropicalis tadpoles, showed that “hydrogen peroxide metabolic process”, “superoxide metabolic process”, and “oxygen and reactive oxygen species metabolic process” were among the most upregulated processes during the first 24 h of regeneration [183]. Likewise, the early production of reactive oxygen species (ROS) is necessary for the activation of signaling pathways that regulate cellular proliferation, which is required for initiating the regenerative program in Xenopus tropicalis [182], Axolotl [5], and Zebrafish tail regeneration [91], as well as zebrafish fin regeneration [92]. While the lipid peroxidation product, 9-hydroxystearic acid, can promote the proliferation of retinal progenitor cells during development [6]. These results indicate that the oxidative response is necessary for inducing cell proliferation and for the initiation of the regeneration program.

Thus, oxidative response and cell death are important for adequate regeneration, supporting the idea that in regenerative organisms, the damage itself triggers a regenerative response. Hence, it is tempting to speculate that modulation of cell death in non-regenerative organisms may improve their regenerative capacity. In addition, as mitochondrial dysfunction is the common event of different stimuli leading to cell death, more specific studies on mitochondrial adaptations upon SCI may contribute to designing novel treatments.

Mitochondria and metabolism regulation of cell death processes

After SCI, mitochondrial function is impaired; synaptosomes [15] and isolated mitochondria [276] obtained from injured rat spinal cord, show increased oxidative damage markers and decreased mitochondrial function at early times, with a maximum decrease at 24 hpi. These correlate with observed impairment in the function of ETC complexes in isolated mitochondria from the damaged spinal cord, particularly complex I, II, and IV [129, 137, 276]. Likewise, mitochondria ultrastructural changes are observed in spinal cord sections, including less and disordered cristae, mitochondrial swallowing between 4 and 8 hpi, and increased membrane rupture and vacuolization from 16 to 24 hpi [136].

In addition, differential temporal changes in mitochondrial fusion and fission markers, as well as mitochondrial abundance and size are observed [38]. As early as 3–6 hpi, an increase in fusion and a decrease in fission mRNA [38] and proteins [38, 136] levels are observed (Fig. 1B), accompanied by a decrease in the number of mitochondria per cross-sectional area and an increase in mitochondrial size [38], while the opposite is observed at 12–24 hpi [38, 136]. Furthermore, mitochondrial biogenesis is affected as well (Fig. 1B); the transcription factor peroxisome proliferator-activated receptor-c coactivator-1a (PGC-1α) governs mitochondrial biogenesis, and its expression in the spinal cord is decreased after contusive SCI in rats [119], and mouse [247]. Finally, altered substrate availability has also been observed in a Yucatan pig model of SCI, including a decrease in glucose and pyruvate availability, as well as an increase in lactate levels [272].

RNAseq and proteomics experiments performed in our laboratory showed that mitochondrial proteins, as well as proteins and transcripts involved in metabolic processes, e.g., fatty acid and glucose metabolism, are differentially regulated in regenerative and non-regenerative stages after Xenopus laevis SCI [168, 169]. In addition, a transcriptomic study of the regenerative response after tail amputation in Xenopus tropicalis tadpoles, showed that genes associated with organic acid metabolism and NADP/H metabolic process were among the most regulated [183]. These data lead to consider mitochondrial function and substrate utilization as key aspects that suffer adaptive responses after SCI and therefore, treatments targeting them have been proposed beneficial for improving regeneration [246].

The maintenance of mitochondrial quality is greatly achieved by the regulation of fission and fusion processes [322], thus, modulating them might prevent SCI-induced mitochondrial dysfunction and reduce the damage extent. Administration of the mitochondrial division inhibitor-1 (Mdivi-1), a selective inhibitor of the mitochondrial fission protein Drp1, previous to SCI, has a positive effect by counterattacking the mitochondrial depolarization, the decrease in ATP and reduced glutathione, cellular death, and motor dysfunction [170]. In addition, in a model of glutamate-induced damage in the spinal cord, Mdivi-1 decreased oxidative damage markers, mitochondrial dysfunction, and cell death [178].

As mitochondrial function is impaired and transcription factors that regulate mitochondrial biogenesis are decreased after SCI, the effect of enhancement of mitochondrial content on SCI has recently gained attention [246, 258]. The induction of mitochondrial biogenesis, either directly by lentiviral-mediated overexpression of PGC-1α [120], or indirectly by pharmacological approaches such as the use of Tetramethylpyrazine [119], β2-adrenoreceptors agonist [247], or 5-hydroxytryptamine 1F receptor agonists [258, 259], have shown a positive effect on locomotor recovery and lesion volume in rodents. In addition, mitochondria injected into the spinal cord, after SCI, are uptaken by neurons, astrocytes, and macrophages, and result in a reduction of apoptotic cells, increase in locomotor function, and reduction in lesion volume [171].

Mitochondrial membrane polarization state is crucial for excitotoxicity and oxidative damage outcome; mitochondrial membrane hyperpolarization results in high calcium uptake [270] and excessive ROS production [108]. Nevertheless, a mild mitochondrial uncoupling has been demonstrated to have a protective role,a single dose pre-treatment with 2,4-dinitrophenol (DNP) results in maintenance of mitochondrial function, and reduced oxidative damage and white matter damaged area, in a rat SCI model [137]. Notably, the endogenous mitochondrial uncoupling protein, uncoupling protein 2 (UCP-2), in addition to the hypothetical protein MGC78829, which has 95.1% nucleotide sequence identity with the Xenopus tropicalis UCP-2 transcript, are part of the top ten transcripts differentially regulated in regenerative versus non-regenerative stages after SCI [169]. Similarly, neonatal brains are protected from excitotoxicity because they express high levels of UCP-2 that prevent the increase in ROS and mitochondrial dysfunction [275]. On the contrary, Ucp-2-/- [12] and Ucp-3-/- [302] mice, show increased ROS production by macrophages and skeletal muscle, respectively. This evidence suggests a critical role of uncoupling proteins in preventing excitotoxicity and oxidative damage.

Pyruvate dehydrogenase enzymatic activity is altered by oxidative damage [125], therefore, the usage of alternative substrate for mitochondrial respiration, which is transformed into Acetyl-CoA entering directly to the Krebs cycle, could reverse mitochondrial dysfunction and provide energy for regeneration [36, 246]. These might be possible in the nervous system, as astrocytes are capable to produce ketone bodies (KB), both from FAA [13] and leucine [28], and then provide neurons with KB in an “astrocyte-neuron metabolic cooperation” [25]. An increasing number of studies have shown that FAA and KB can be used as an alternative substrate, enhancing SCI recovery. Acetyl-L-carnitine injections in rodent SCI models prevent and revert changes in mitochondria ultrastructure, dynamics, and function, and cell death [228, 229, 327], impeding the sparse of the damage [228, 229] and improving functional recovery [229]. KB can improve mitochondrial respiration under excitotoxicity conditions in primary neuronal cultures [185], increase mitochondrial biogenesis in rat hippocampal tissue [30], increase brain-derived neurotrophic factor expression, and nuclear factor kappa-light-chain-enhancer of activated B cells (NF-kB) activation in primary neuronal cultures [190]. In addition, a ketogenic diet (low carbohydrates and high FAA) increases GSH biosynthesis in hippocampal isolated mitochondria [132], as well as the expression of antioxidants enzymes, such as manganese superoxide dismutase and catalase in spinal cord samples [313], rescues mitochondrial respiratory capacity and increases mitochondrial biogenesis [249], while enhancing motor recovery after SCI [271].

In summary, regulation of mitochondrial function after SCI, including partial mitochondrial uncoupling, regulation of mitochondrial dynamics and biogenesis, as well as increasing available mitochondria, and the usage of alternative substrate for mitochondrial respiration, may support efficient spinal cord regeneration. Thus, future studies should be focused on the modulation of mitochondrial response and availability.

Immune cell response and inflammation

After SCI, the immune response is activated, generating an inflammatory environment needed for cell debris removal. However, the inflammation should be resolved for repair or regeneration to occur, otherwise, it results in chronic inflammation [82, 138]. In the present section, we describe the general immune cell response after damage, followed by a comparison of this response between mammals and regenerative animal models after SCI, and finalizing with a revision of immune cells regulation by the cell metabolic status and mitochondria.

General immune response after damage

After an insult, damaged or dying cells release proteins to the extracellular space, the so-called damage-associated molecular-pattern (DAMP) molecules, that induce inflammation [241]. The inflammatory response starts with activation of microglia/macrophage M1 (pro-inflammatory or neurotoxic) phenotype (Fig. 2A), crucial for cell debris removal, but at the same time, they release pro-inflammatory compounds such as tumor necrosis factor-α (TNF-α), interleukin-6 (IL-6), IL-1, and NO, which induce generalized cellular damage, widespread monocytes/neutrophils infiltration and deterioration of the extracellular matrix [56]. To prevent exacerbated tissue destruction a downregulation of this response is required. This is achieved by a change from M1 to M2 phenotype [56], which could be induced in part by the phagocytosis of apoptotic cells, as observed in macrophages culture [81] and inflamed tissues outside the CNS [126] (Fig. 2A). M2 (anti-inflammatory or neuroprotective) cells produce and release anti-inflammatory cytokines such as IL-4, IL-13, IL-10 and transforming growth factor-β (TGF-β) [56], that can suppress pro-inflammatory gene expression [105], increase fibroblast proliferation and collagen synthesis [266], induce extracellular matrix proteins expression [100], and promote angiogenesis among others [133, 161], allowing tissue remodeling and repair. Microglia and macrophages polarization to either phenotype is regulated by the cellular metabolic state, increasing the glycolytic flux and decreasing oxidative phosphorylation (OXPHOS) favors the polarization toward an M1 phenotype, while a reduction in glycolytic flux and an increase in OXPHOS promote M2 [1, 89, 135] (Fig. 2A). Although when the pro-inflammatory response is exacerbated, it can result in chronic inflammation [1] (Fig. 2A).

Fig. 2.

Immune cell response and regulation after SCI. A Schematic representation of microglia/macrophage M1 and M2 metabolic signature and regulation. Damaged cells release DAMPs, which activate quiescent microglia/macrophages by a metabolic switch involving activation of HIF-1α and NF-κB transcription factors, which favor a glycolytic metabolism, and result in M1 polarization and pro-inflammatory environment. The phagocytosis of apoptotic cells induces a second metabolic switch, by inducing transcription of PPARγ and mTOR regulated genes, favoring an oxidative metabolism, and resulting in M2 polarization and an anti-inflammatory environment. M2 polarization leads to tissue remodeling and repair unless there is myelin debris presence, which transforms M2 macrophages into foam macrophages, leading to chronic inflammation. Lightning symbolizes damage; orange circles, pro-inflammatory; and green circles, anti-inflammatory cytokines. Red and pink arrows symbolize strong or attenuated change, respectively. DAMPs damage-associated molecular-pattern molecules; PPP pentose phosphate pathway; FAO fatty acid oxidation; OXPHOS oxidative phosphorylation. B, C Graphs representing the microglia/macrophage polarization to M1 (pro-inflammatory or neurotoxic) or M2 (anti-inflammatory or neuroprotective), in response to SCI, in non-regenerative (B) and regenerative (C) model animals, over time

Immune response in mammalian spinal cord

In murine models, the immune response after SCI is characterized by a predominant microglia/macrophage M1 phenotype and result in chronic inflammation (Fig. 2B). The first immune cell type to respond are microglia [40] and neutrophils [40, 281]. Microglia/macrophages become active acquiring an M1 phenotype [154], and monocytes infiltrate at the injury site and differentiate into bone marrow-derived macrophages [312]. Parallel but transiently, some macrophages/microglia acquire the M2 phenotype, which coexists with M1 cells in the injury site during the first 1–3 days after the lesion, but after 1 week the M1/M2 phenotype ratio rises considerably [154] (Fig. 2B). This is coincident with the accumulation of myelin debris and its phagocytosis by M2 macrophages, which induces the switching of phenotype from M2 to foamy macrophages [295]. M2 macrophages engulf myelin debris in a macrophage scavenger receptor 1 dependent manner [162], as myelin debris is lipid-rich, macrophages accumulate an excess of intracellular lipids [312], resulting in dysregulated lipid intracellular homeostasis and a transcriptional profile biased towards lipid catabolism [332], which is characteristic of foamy macrophages [312, 332]. The foamy macrophages present an M1-like pro-inflammatory phenotype [312], which losses their migratory and phagocytic activity, and thus, the cellular content of apoptotic cells is accumulated, subsequently favoring chronic inflammation and decreasing regenerative capacity [295]. The pro-inflammatory microenvironment and the M1 microglia also lead to astrocytes activation [102], which, along with microglia, meningeal fibroblasts, pericytes [99], and extracellular matrix [323] form a glial scar in the lesion site [217].

Immune response in regenerative models spinal cord

The immune response after SCI is differentially regulated between regenerative and non-regenerative stages in X. laevis at the transcriptomic [169] and proteomic [168] level, and between axolotl and mammals [286]. In regenerative model animals, like zebrafish, the immune response after SCI is characterized by a dual response, starting with a microglia/macrophage M1 polarization, followed by an M2 polarization, which results in regeneration [290] (Fig. 2C). After different spinal cord injury paradigms, including stab injury [290] and electroablation [11], a quick innate immune response is observed. Neutrophils first respond, reaching a peak at 2 hpi, while macrophages and microglia have a later response, peaking at 2 dpi [290]. During the first hours, pro-inflammatory cytokines are required for induction of axon regeneration, as inhibition of Tnf-α or Il-1ß, strongly impaired axon regrowth. While at 12 hpi, a reprogramming towards an anti-inflammatory phenotype is necessary, because the maintenance of elevated levels of cytokine Il-1ß inhibits axon regeneration [290]. It is suggested that macrophages coming from peripheral tissues [203], could be responsible for this dual response, since a mutant zebrafish model, where macrophages are absent, exhibit a permanent pro-inflammatory response and impaired axon regeneration [290].

Energy status and mitochondrial regulation of immune cells

Microglia and macrophages are dynamic, exhibiting M1 and M2 phenotypes characterized by distinct gene expression profiles, structure, and function, adapting their response to multiple immunological stimuli [1, 135]. These responses are highly dependent on mitochondria and metabolic state [1, 199, 305]. Under physiological conditions, immune cells are quiescent, while a pro-inflammatory environment induces a switch from an oxidative to glycolytic metabolism, achieved by activation of the hypoxia-inducible factor-1α (HIF-1α) [150, 213, 308], and NF-κB [250]. This metabolic change results in an M1 phenotype characterized by a strong upregulation of glycolysis [96, 240, 280, 306] and pentose phosphate pathway (PPP) [96, 280], interruption of the Krebs cycle [280],and downregulation of OXPHOS [240] and fatty acid oxidation (FAO) [280] (Fig. 2A, orange M1 phenotype). This metabolic reprogramming is dependent on mitochondrial fission/fusion regulation, as fission inhibition by Mdivi-1 impedes the switch to the glycolytic state [209]. This metabolic reprogramming is necessary for inflammatory molecules production. Inhibition of glycolysis by 2-deoxyglucose impedes the accumulation of HIF-1α and thus, inhibits the generation of IL-1β and TNF-α [280]. Consistent with the role of fission in the switch to glycolysis and generation of IL-1b and TNF-α, pharmacological inhibition of mitochondrial fission impairs the pro-inflammatory cytokines and ROS production [147, 209, 225]. In addition, Krebs cycle interruption results in succinate and citrate accumulation [134]. The rise of citrate can be used for the generation of prostaglandins, NO and ROS [127], and a surge in succinate increases HIF-1α expression, resulting in induction of IL-1β [280], with the subsequent increase in inflammatory molecules production [47, 277, 280] (Fig. 2A, orange M1 phenotype).

A new metabolic reprogramming allows phenotypic change to M2, which is driven by the nuclear transcription factor peroxisome proliferator-activated receptor γ (PPARγ) [64, 219] and mammalian target of rapamycin (mTOR) complexes 1 (C1) and 2 (C2) [60, 122, 142] (Fig. 2C, green M2 phenotype). However, mTORC1 participation could be context-dependent for favoring M1 [35] or M2 [57, 142] phenotype activation. M2 phenotype is characterized by upregulation of glutamine metabolism [134], OXPHOS and FAO, and increased mitochondrial biogenesis [299]. Glycolysis and PPP rates of M2 are lower compared to M1, but higher than quiescent macrophages [89, 199, 240] (Fig. 2A, green M2 phenotype). This new metabolic change is necessary to increase the production of characteristic molecules of M2, and decrease those of M1, e.g., Glutamine is used to form α-ketoglutarate (α-KG), which induces epigenetic reprogramming in M2-genes promoters while impeding nuclear translocation of NF-κB [179] and HIF-1α accumulation [280]. In addition, FAO and OXPHOS inhibitors decrease arginase I activity [299], which in normal conditions competes with NOS for arginine substrate, resulting in inhibition of NO production [202]. Finally, the PPP non-oxidative branch is needed for ribose-5P production [109], which is used for uridine diphosphate N-acetylglucosamine (UDP-GlcNAc) generation, required for protein glycosylation, a hallmark of M2 phenotype [134] (Fig. 2A, green M2 phenotype).

M1 to M2 phenotypic shift is associated with an increase in UCP2 expression [67], leading to a partial uncoupling of the electron transport chain. Recent data suggest that the mitochondria-associated endoplasmic reticulum membranes (MAMS) accumulate miRNAs that regulate inflammation,interestingly, the levels of these miRNA levels change after uncoupling treatments [309]. Therefore, UCP2 regulation could be associated with the activity of these miRNAs.

Some studies lead to think that the lack of regenerative capacity of mammals is due to a maladaptive inflammatory response. Immunosuppression during the “refractory period” in Xenopus tadpole, restores the regenerative ability [88], while Kigerl et al. [154] showed only a transient activation of M2 phenotype after SCI in mice [154]. Moreover, lipopolysaccharides (LPS) are pro-inflammatory signals, but LPS preconditioning in SCI, results in facilitation of M2 phenotype in resident microglia, accompanied by increased vascularization next to the injury site, decreased injury area, and improved functional recovery [111]. All these data suggest that potentiation of the M2 phenotype, possibly by changing the metabolism and/or mitochondrial function, may be a promising therapeutic focus for SCI treatment in mammals, and thus, further studies are necessary.

Axonal degeneration and regeneration

SCI initial impact results in severed axons and oligodendrocyte loss, which activates a series of events leading to axon degeneration [298]. After axon damage in mammals, regenerative attempts could be seen, although they ended up in axon degeneration [114, 151, 172], while in regenerative models axons can regenerate [69]. In the present section, we will describe the axonal response after SCI, in both mammals and regenerative animals, followed by the role of mitochondria on axon regeneration.

Axonal response after SCI in mammals

Axon degeneration occurs in two phases, a fast and early phase called acute axonal degeneration (AAD), which consists of asymmetrical fragmentation of the proximal and distal axonal ends [151], followed by a later phase, in which the proximal axon end swells up forming a terminal bulb that constantly retracts, “Dieback”, away from the injured site [248] (Fig. 3A´), and the distal axon end suffers Wallerian degeneration (WD), in which axonal swelling and degeneration start near the disrupted region and continue progressively distal to the damage [24] (Fig. 3A´´). The processes of axonal degeneration and regeneration are very dynamic, the axon tries to re-connect the lost connections by extending terminal and collateral sprouting through the grey matter and the injury site [114, 172], even after the glial scar is formed [173]. However, this sprouting is transient, the velocity of growth is slow compared to regenerating axons, and they seem to lack directionality [151], failing in axon regeneration [173] (Fig. 3A, blue dashed lines). This inability to regenerate is due to both intrinsic and extrinsic factors [2, 72, 143, 282]. On one hand, the capacity of rat severed spinal cord axons to grow into a permissive environment, such as PNS grafts, evidences the presence of extrinsic factors that inhibit axonal regeneration in the spinal cord [65, 239]. On the other hand, stem cells grafted into the rat injured spinal cord can bypass the environmental inhibitory factors and differentiate into neurons extending axons through the damaged area [184], showing that spinal cord neuron intrinsic properties are not optimal for regeneration.

Fig. 3.

Axonal response after SCI. A, B Schematic representation of axon damage after SCI, resulting in degeneration in non-regenerative (mammals) (A) or regrowth in regenerative animal models (B). A´ and A´´, correspond to magnifications of a damaged axon proximal and distal end, respectively, and B´ to a magnification of a damaged axon proximal end. A After SCI in mammals, the severed axon develops a bulb end, presenting damaged/depolarized mitochondria and fragmented microtubules in both proximal and distal ends, and suffering degeneration thought dieback (A´) and Wallerian degeneration (WD) (A´´). Blue dashed lines extending from the axon proximal ends show failed regenerative attempts, in part because of lack of functional mitochondria and responsiveness to guidance cues (A´). Increasing mitochondrial dynamics induce axon regenerative capacity (A, lower portion of spinal cord). B After SCI in regenerative animal models, the severed axon develops a growth cone capable of responding to guidance cues, with functional mitochondria (B´), resulting in axon regrowth, which follows the path of a “glial bridge” formed by bipolar glial cells throughout the injury site, and continue caudally throughout the gray matter, while degenerating axons and cell debris are still present in the white matter. A´ and B´ show the effect of extrinsic factors over mitochondria, CSPGs and Nogo-A generate a decrease in mitochondrial function, while BDNF has a positive effect over mitochondrial function and NGF stimulates mitochondrial transport. Syringe symbolizes external manipulation

Axonal response after SCI in regenerative animal models

Axon regeneration occurs in many animal models including, Xenopus laevis [22, 95], axolotl [54], newts [261, 333], and zebrafish [94]. Herein, we are going to temporarily describe what happens in Zebrafish, as is one of the most widely described [94]. Days after injury there is an increase in transcription levels of genes involved in axonal growth, reaching a peak between 7 and 14 dpi [23], and axons regenerate until 6 weeks post-injury, where axon growth reaches a plateau [22], and swimming capabilities are recovered [123]. The regenerating axons elongate throughout the injury site following a “glial bridge” [97], and change their pathway to regrow throughout the gray matter caudally to the injury site, while in the white matter degenerating axons and cell debris are observed [20] (Fig. 3B). In contrast to mammals, the extrinsic factors seem to be less inhibitory for the growth of the axons [264, 294]. In addition, different populations of neurons express dissimilar intrinsic factors, determining the regenerative capacity, e.g., when comparing ascending and descending axons in the spinal cord after injury, a greater number of descending neurons express proteins involved in axonal growth and, therefore, their axons regenerate more than those of ascending neurons [21].

Mitochondria function during axon regeneration

For an axon to grow or regenerate, it has to extend a growth cone, a process that is energetically demanding, as involves a rapid sealing of the membrane, re-assembly of the cytoskeleton, transport of molecules, local protein translation, and insertion of membrane and cell surface molecules at the leading edge [31]. During axon growth, mitochondria accumulate along the axonal shaft defining axonal branching points [267], and in presynaptic buttons and axon terminals for synapse assembly and synaptic transmission [252]. In addition, mitochondrial fission/fusion balance seems relevant for growth cone lamellar protrusion, response to guidance cues, and growth cone steering [269]. In the present section, we consider the evidence concerning mitochondria transport and energy supply in axon injury and regeneration, regulation of mitochondrial activity by axon growth factors, and regulation of mitochondrial supply by mitochondrial transfer.

Mammalian severed axons fail to extend a growth cone, and instead, they form a terminal bulb and degenerate. This observation is coincident with axonal mitochondrial dysfunction after damage (Fig. 3A´, A´´), evidenced by mitochondrial depolarization [44, 330], increased mitochondrial fission [44], fragmentation, and mitophagy [148], ultrastructural changes like swelling and dilation of cisterns [160], and decreased ATP production [330], which leads to a huge deficit of energy.

There are two different pools of mitochondria in axons, the stationary that comprises around 70%, and the motile that corresponds to the remaining 30% [141]. Normally, healthy mitochondria with high membrane potential are transported preferentially anterogradely and damaged mitochondria with a low potential retrogradely, towards the neuronal soma [198] (Fig. 3B´). Interestingly, not only is mitochondrial function affected after axotomy but mitochondrial transport is also compromised. In Zebrafish peripheral sensory neurons undergoing Wallerian degeneration [216], in axotomized neurons from Drosophila adult brain [14], and mice transected intercostal nerves in vivo and in explants [201], mitochondrial motility is inhibited upon axonal injury. Even more, inhibition of mitochondrial transport in axons, both in Drosophila [14] and C. elegans [107], as well as restraining mitochondria in neuronal soma and out of the axon [237], prevents axonal regeneration and result in axon degeneration. On the contrary, mitochondrial transport to the injury site has a positive contribution to axon regeneration. Increasing mitochondrial transport by overexpressing the motor adaptor Miro1 or knocking out the mitochondria-MT anchor protein syntaphilin (snph) in mice, is needed for replacing damaged mitochondria in the axon after SCI [106], and in mature cortical neurons [330], resulting in enhancement of the number of axons forming a growth cone and improving regeneration. Similarly, overexpression of Armcx1, a mitochondrial protein induced during regeneration, potentiates mitochondrial movement, neuronal survival, and axonal regeneration in retinal ganglion cells [43].

Even more, studies in axonal models with high regenerative capacity have supported mitochondrial contribution to axon regeneration. When comparing regenerating and non-regenerating axons in C. elegans, a positive correlation between mitochondrial transport and density with regeneration capacity is observed [107]. In addition, peripheral nerves present an increased mitochondrial transport both in vitro [201] and in vivo [156], and the same is observed in cortical [42] and adult retinal ganglion cells [43] with high axonal regenerative capacity, induced by co-deletion of phosphatase and tensin homolog (PTEN) and Suppressor of cytokine signaling 3 (SOCS3).

An inverse relationship between the mitochondrial size and velocity of transport has been described [212], although mitochondrial fragments displaying lengths < 0.5 mm are considered dysfunctional [118], therefore, a meticulous regulation of the mitochondrial size is needed for proper regeneration. Among other characteristics, mitochondrial size is regulated by fusion/fission processes [46]. In addition to affected mitochondrial transport after axonal injury, changes in mitochondrial fusion/fission events are also observed,in a mouse brain injury model, an increase in mitochondrial fission-related proteins and a decrease in mitochondrial fusion related proteins are observed in the axotomized neurons [44], in a rat SCI model, mitochondrial fragmentation in the vicinity of the axonal injury is detected [148], while in mouse sciatic nerve injury, although no changes in fission/fusion protein are observed, a decrease in mitochondrial size is detected [156]. Even more, an injury-induced mitochondrial fission protein Drp1-deficient motor neurons transgenic mice presented mitochondria with increased size, decreased mitochondrial transport velocity, and increased neuronal death, after injury [156]. All this evidence demonstrates the importance of mitochondrial dynamics events for proper axon regeneration.

Mitochondria are also needed for increasing energy supply in regenerating axons; administration of creatine enhances axonal regeneration after SCI in mice [106], and in mouse mature cortical neurons in culture, electroporation of ATP in snph overexpressing axons, allows regeneration to some extent, while inhibiting ATP production with oligomycin in snph KO neurons, prevents axon regeneration [330]. Therefore, mitochondrial abundance and energy provision near the site of injury, seem to determine the neuronal injury/regeneration process (Fig. 3).

It has been demonstrated that many intrinsic and extrinsic factors, that regulate axon growth, also regulate the mitochondrial activity, supporting the idea of a role of mitochondria during axon regeneration [72, 166] (Fig. 3A´, B´). In cultured chicken embryonic dorsal root ganglion neurons, chondroitin sulfate proteoglycans (CSPGs) inhibit the extension and regeneration of axons by decreasing mitochondria density and membrane potential next to the growth cone and decreasing the axonal collateral branching by depolarizing the membrane potential of axonal mitochondria [245] (Fig. 3B´). Nogo-A, a myelin protein that inhibits axonal growth, can directly interact with proteins conforming to complex III of the mitochondrial respiratory chain [121]. The repellent guidance cue semaphoring-3A generates an increase in mitochondrial membrane potential [301]. Interestingly, growth factors influence mitochondrial function as well, NGF increases mitochondrial membrane potential [301] and stimulates mitochondrial transport towards the focus of NGF stimulation [45] (Fig. 3B´); moreover, BDNF increases mitochondrial coupling in a complex I dependent manner [189]. Kruppel-like families transcription factors regulate mitochondrial size, density, and biogenesis [269]. PTEN-induced kinase 1 regulates mitochondrial membrane potential, respiratory function, and fission/fusion events [284]. As we mentioned previously, mammalian axons inability to regenerate is due to both intrinsic and extrinsic factors [2, 143, 282], and as we stated in this section, mitochondrial dynamics and function could be acting downstream of these factors, which summed to the mitochondrial dysfunction and mitochondrial transport compromised after axotomy in non-regenerative models, make the study of the mitochondrial response to intrinsic and extrinsic factors after axotomy, highly important.

Finally, recent evidence suggests that mitochondria can be transferred from one cell to another for disposal or recycling of damaged mitochondria [263, 287]. Rat hippocampal primary neurons and astrocytes can form tunneling nanotubes between them, under stress conditions, and different organelles, including mitochondria, can be transferred from one cell to the other [314]. Rat cortical astrocytes can release functional mitochondria into the extracellular space in vitro, and transfer mitochondria to cortical neurons after stroke, increasing neuronal cell survival signals in vivo [112]. While rat retinal neurons can transfer damaged mitochondria to astrocytes for mitochondrial elimination [53]. Interestingly, rat bone marrow mesenchymal stem cells can transfer, by Gap-junctions, their mitochondria to spinal cord motoneurons in vitro, promoting their survival, and also, in vivo transplantation of those mitochondria in SCI improved locomotor recovery [171]. This is a possible mechanism that could be occurring in axonal regeneration after SCI, thus it deserves further investigation.

In summary, all this evidence supports mitochondria as fundamental players promoting and steering axonal regeneration (Fig. 3); thus, special attention should be kept on mitochondrial modulators as possible treatments for SCI.

Neural stem progenitor cells and neurogenesis

Neurogenesis is a process often associated with embryonic development, but it also occurs in the adult CNS [71, 87, 238]. The capability of certain animals to regenerate their spinal cord is strongly related to the presence of NSPC, and their ability to proliferate and differentiate in response to SCI [51, 207]. In this section, we will describe the evidence for the presence of possible mammalian spinal cord NSPC, the neurogenesis process in regenerative models, and the role of mitochondria in regulating NSPC and neurogenesis, and its implications in spinal cord regeneration.

Neurogenesis in mammalian spinal cord

Even though there is not an established NSPC in the mammalian spinal cord, the ependymal cells that surround the central canal have been suggested as possible endogenous NSPCs of this tissue. This is based on their high in vitro proliferative rate [164, 205], and their multipotent phenotype, as the neurospheres formed in vitro can be induced to differentiate into neurons, astrocytes, or oligodendrocytes [317]. Importantly, the proliferative response is potentiated by SCI, suggesting an activation of these cells after lesion [19, 85, 211]. Experiments in mice showed that these cells are not able to differentiate in vivo into neurons after SCI, but instead, they differentiate mainly into astrocytes in the center of the lesion site where they contribute to the glial scar formation, and to a lesser extent into oligodendrocytes [19, 196] (Fig. 4A). Nevertheless, they can differentiate into neurons when transplanted to neurogenic-permissive environments [255].

Fig. 4.

Neurogenesis and neural stem progenitor cells regulation. A, B Schematic representation of neurogenic response after SCI in non-regenerative (mammals) (A), and in regenerative animal models (B). SCI induces proliferation of ependymal cells lining the spinal cord central canal. In non-regenerative animals, this proliferation occurs to a lesser extent and cells differentiate principally to astrocytes, which are going to form part of the glial scar, and in a minor proportion to oligodendrocytes (A). In regenerative animals, an extensive proliferation occurs, and cells differentiate to neurons, allowing neurogenesis; bipolar glial cells, that form the “glial bridge”; and oligodendrocytes, which are going to participate in remyelination of the axons (B). C Schematic representation of mitochondrial regulation of neural stem progenitor cell (NSPC) self-renewal and differentiation processes. NSPC presents a glycolytic metabolism and less mitochondrial mass than differentiated cells, after proliferation, the cell with fewer mitochondria and glycolytic metabolism continues as NSPC, and the one with an older and more mitochondrial mass shifts towards an oxidative metabolism and differentiate

Neurogenesis in regenerative model organism’s spinal cord

Previous studies in our laboratory have shown that the neuronal progenitor marker Sox2 is expressed in cells surrounding the central canal of X. laevis spinal cord, and after SCI they increase proliferation in regenerative tadpoles, while overexpression of a dominant-negative form of Sox2, or morpholinos against this gene, impair regeneration [90, 207]. In addition, it has been recently demonstrated that these Sox2 positive cells surrounding the central canal of the X. laevis spinal cord, present an NSPC identity [74]. Therefore, indicating that activation of NSPC is necessary for spinal cord regeneration. This is also supported by studies in zebrafish [124], where SCI increases Sox2⁺ cells proliferation, and in the axolotl [83], where a lack of Sox2 disrupts the regenerative capacities after SCI. In regenerative model animals, these proliferative cells can differentiate not only to glia, but to neurons as well [26, 50, 84, 94, 207] (Fig. 4A, B).

Role of mitochondria in neural stem cell biology

The response of the Sox2⁺ cells is different in regenerative versus non-regenerative models. In X. laevis regenerative stages, Sox2⁺ cells react with a high and quick proliferative response to SCI, while in non-regenerative stages there is a delayed and very limited response [207], something similar is observed when comparing juvenile with adult axolotl [51]. Other important aspects that could explain the different capacities of regeneration among species or among stages of life cycle are the different cellular composition surrounding the central canal of the spinal cord [75], and the preference of stem cells to differentiate toward a glial phenotype in mammals [19, 196], while to a neuronal progeny in a regenerative model organism [26, 50, 84, 94, 207]. Therefore, neural stem cell regulation or transplantation into the lesion site needs a more in deep study, as could be a possible treatment for SCI [18]. We showed that spinal cord cells from regenerative tadpole transplanted into non-regenerative froglets can form neural tube-like structures and differentiate into neurons [197], in agreement with the evidence of neuro-motor improvement in injured mammals transplanted with stem cells [145, 220, 251].

Stem cells are modulated by several signals that induce intracellular changes to enhance or inhibit proliferation, maintenance, or differentiation processes [10, 68, 218, 296, 304]. Mitochondria and cellular metabolism may have an important role in both self-renewal and differentiation processes, evidenced by a correlation between mitochondrial status and cellular metabolism with stemness [128, 214, 324] (Fig. 4C). Mitochondrial mass and biogenesis have an inverse correlation with stemness; comparison between NSPC derived from embryonic stem cells, induced pluripotent stem cells [328], or from rat hippocampus [310], with differentiated neurons, show that differentiation includes an increase in mitochondrial mass (Fig. 4C). Even more, differentiation of human neural stem cells to motor neurons is accompanied by an increase in transcription factors involved in mitobiogenesis, although in this work the mitochondrial mass was unchanged [215], and in mouse hippocampal neurons in culture and PC12 cells, neuronal differentiation initiation is dependent on mitobiogenesis, which is regulated by epigenetics [292]. Furthermore, it has been seen that after mitosis of human mammary stem-like cells, the daughter cell containing fewer old mitochondria retained the stemness phenotype [146] (Fig. 4C). In addition, differentiation of NSPC is inhibited by mtDNA damage accumulation [311] and modulated by the mitochondrial fusion/fission process [131, 153]. Even more, when comparing X. laevis regenerative and non-regenerative transcriptomic and proteomic response to SCI, a decrease in proteins related to mitochondria is observed in regenerative stages [168], accompanied by an upregulation in transcripts belonging to the cell cycle category [169], and the Sox2⁺ cells proliferative response to SCI [207], supporting a relationship between mitochondrial response, NSPC proliferation, and regeneration, which needs to be further explored.

Regarding the cellular metabolism, NSPC produces ATP principally by glycolysis rather than by OXPHOS [130] (Fig. 4C); this has been observed in NSPC derived from a reprogrammed human fibroblast line [328], adult mouse hippocampus [256], and embryonic Xenopus retina [3], among others, suggesting that a reduced mitochondrial function is related to the stemness phenotype. Human pluripotent stem cells have higher expression levels of UCP2, which decrease after differentiation [244], and because of this, it has been proposed that this mitochondrial uncoupling protein prevents mitochondrial glucose oxidation by a shuttling mechanism in which pyruvate is excluded from the mitochondria, increasing glycolytic flux and nucleotide synthesis by the PPP [325]. This glycolytic preference could be explained in part by the hypoxic environment of several stem cells niches that force them to survive with anaerobic metabolism, as hematopoietic stem cells in bone marrow [77, 163, 260]. Low oxygen availability inhibits stem cells differentiation in vitro [62, 80] while enhancing in vitro generation of induced pluripotent stem cells [321]. These effects could be mediated by HIF1α expression, as it is a protein related to a pluripotent state [260, 331],particularly, HIF1α pharmacological inhibition favors oligodendrocyte fate of ependymal cells in culture [204], conditional deletion of this transcription factor in developing mouse cerebral cortex accelerates neurogenesis from radial glia [165], and neurospheres obtained from HIF1α KO and KD embryonic neural stem cells present lower self-renewal potential and increased expression of neural and astroglial markers [300], while activation of HIF1α favors induction of pluripotency [233]. Other metabolic pathways rather than glycolysis and OXPHOS have been related to NSPCs behavior as well, which have been deeply reviewed elsewhere [158, 159, 257]. For example, quiescent hippocampal NSPCs have higher rates of FAO than proliferative NSPCs, which is necessary for the quiescent state maintenance [158, 159],high levels of lipogenesis are needed for neurogenesis from adult mouse brain NSPCs [157],and more recently, it has been suggested a link between lipid metabolism and stem cells polarity during brain cortex development [98]. In addition, glutaminolysis pathway, in addition to glycolysis, is important for energetic supply for stem and progenitor cells proliferation in cortical development [140, 210].

Even though there are no studies showing how changes in the different aforementioned metabolic pathways affect NSPCs response after SCI, we previously demonstrated that SCI induces early changes in transcripts associated to metabolic processes, such as fatty acid and glucose metabolism, which are temporarily related to changes in transcripts associated to cell cycle [169] and the Sox2⁺ cells proliferative response to SCI [207] in regenerative X. laevis stages. The same has been observed by others in a X. tropicalis tail amputation model, where Affymetrix genome array [183] and RNAseq analyses [232], show changes in transcripts related to metabolic processes and cell proliferation during tail and spinal cord regeneration. All these evidences suggest a relationship between changes in metabolism, cell proliferation, and spinal cord regeneration, that should be addressed by future experiments.

In addition, single-cell transcriptomic analysis of adult mouse hippocampal NSPC, showed that NSPC activation and neurogenesis induction is accompanied by a metabolic shift from glycolysis to OXPHOS [256] (Fig. 4C). Furthermore, NSPC derived from a reprogrammed human fibroblast line showed that a decrease in glycolytic enzymes expression is essential for neuronal differentiation, otherwise, neuronal death is observed [328]. The switch to oxidative phosphorylation for ATP synthesis causes an increase in mitochondrial ROS production, which has been involved in stem cell differentiation [39, 232, 234, 303, 307], although, the exact mechanism is not completely elucidated. Even more, cellular commitment and differentiation are regulated by DNA and histone modifications [152, 273], and this metabolic switch also leads to the activation of the Krebs cycle and the generation of metabolites that are co-factors of enzymes involved in epigenetic modifications [152, 192, 324]. α-KGDH influences proliferation and neuronal commitment in the hippocampal subgranular zone [37], while acetyl-coenzyme A is a donor of an acetyl group for histone acetylation by histone acetyltransferases (HACs) [206], and α-KG is a substrate for enzymes with histone and DNA demethylation function [283]. Further investigation is needed to elucidate which signals modulate the specific phenotype after differentiation, and which intracellular pathways are involved in this process.

Together, all these data show the complexity of regeneration mechanisms and the need for several processes to occur, such as a proper balance of stem cell proliferation, migration, and differentiation into specific phenotypes. Since these processes are influenced by mitochondria and metabolic change, they should be the focus of new studies to reveal the complex events of regeneration of the spinal cord.

Conclusions

Increasing evidence demonstrates that mitochondria play a crucial role in many events of the secondary injury after SCI, including cell death, immune response, axon regeneration, and neural stem cells regulation. SCI induces mitochondrial dysfunction, which results in ROS generation, loss of bioenergetics, and cell death, that together with the initial damage, induce a switch towards a glycolytic metabolism, activating the neurotoxic M1 immune response and inflammation, which further induces cell death due to the release of pro-inflammatory cytokines and reactive species. In non-regenerative models, excessive cell death is observed, M2 immune response at attempted, but only to attain a predominant M1 immune response and chronic inflammation. Mitochondria are not able to elicit an adaptive response, showing mitochondrial dysfunction, reduced mitochondrial trafficking, and therefore, loss of bioenergetics, resulting in further cell death and inability to regenerate the spinal cord (Fig. 5B). In regenerative models, the cell death and reactive species are necessary for precursor cell proliferation, while phagocytosis of dying cells induces a new metabolic switch towards OXPHOS and FAO utilization, and mitochondrial biogenesis, allowing the polarization to a neuroprotective M2 response, which is needed for resolving inflammation and for induction of tissue remodeling and repair. Likewise, mitochondrial adaptive response leads to regulation of mitochondrial mass, position, and dynamics, in addition to the cellular metabolic state, restoring cellular bioenergetics and allowing neural precursor differentiation and axon regrowth, resulting in cellular survival, spinal cord regeneration, and functional recovery (Fig. 5A).

Fig. 5.

Regenerative and non-regenerative response to SCI. A Schematic representation of the chronic non-regenerative model response to SCI. After injury, mitochondria are damaged, leading to bioenergetics decline, which precedes an inflammatory response and cell death. Over time, mitochondria fail to adapt, therefore, bioenergetics fails to restore and inflammation to resolve, resulting in further cell death, and therefore decreased survival. B Schematic representation of regenerative animal adaptive response to SCI. After injury, mitochondria are damaged, leading to a decline in bioenergetics, which precedes an inflammatory response and cell death. Over time, the mitochondria show an adaptive response, restoring mitochondria, bioenergetics and cell survival, and inflammation resolve. The Y-axis corresponds to the relative percentage of each process represented by the different colors. The X-axis corresponds to the relative time before and after SCI, which is represented by the vertical red line. The white background in each graph corresponds to the non-injured or regenerated spinal cord, while the pink background corresponds to the injured or non-functional spinal cord

Therefore, as mitochondria are a common player in the different spinal cord regenerative events, changes in their function, at the right timing, could reduce the extent and duration of cell death, favor the anti-inflammatory over pro-inflammatory response, and induce neurogenesis and axonal regeneration after SCI, resulting in an improvement of spinal cord regenerative capacities. Because of these, modulation of mitochondrial function could have a pivotal role in the design of novel therapeutical approaches to enhance spinal cord regeneration.

Author contributions

PGS manuscript conceptualization and design, manuscript writing and editing, figure design and editing, and funding acquisition. MED-R manuscript writing, figure design and generation. MV manuscript writing. VE manuscript critical reading and editing. JL manuscript critical reading and editing, and funding acquisition.

Funding

This work was funded by FONDECYT 3190820 for PGS, 1180429 for JL and 1191770 for VE.

Data availability

Not applicable.

Declarations

Conflict of interest

The authors declare no conflict of interest.

Ethical approval

All the authors agree to participate.

Consent for publication

All the authors agree for publication.