Peroxisome proliferator-activated receptor γ (PPARγ): A master gatekeeper in CNS injury and repair

Abstract

Peroxisome proliferator-activated receptor γ (PPARγ) is a widely expressed ligand-modulated transcription factor that governs the expression of genes involved in inflammation, redox equilibrium, trophic factor production, insulin sensitivity, and the metabolism of lipids and glucose. Synthetic PPARγ agonists (e.g. thiazolidinediones) are used to treat Type II diabetes and have the potential to limit the risk of developing brain injuries, such as stroke, by mitigating the influence of comorbidities. If brain injury develops, PPARγ serves as a master gatekeeper of cytoprotective stress responses, improving the chances of cellular survival and recovery of homeostatic equilibrium. In the acute injury phase, PPARγ directly restricts tissue damage by inhibiting the NFκB pathway to mitigate inflammation and stimulating the Nrf2/ARE axis to neutralize oxidative stress. During the chronic phase of acute brain injuries, PPARγ activation in injured cells culminates in the repair of gray and white matter, preservation of the blood-brain barrier, reconstruction of the neurovascular unit, resolution of inflammation, and long-term functional recovery. Thus, PPARγ lies at the apex of cell fate decisions and exerts profound effects on the chronic progression of acute injury conditions. Here, we review the therapeutic potential of PPARγ in stroke and brain trauma and highlight the novel role of PPARγ in long-term tissue repair. We describe its structure and function and identify the genes that it targets. PPARγ regulation of inflammation, metabolism, cell fate (proliferation/differentiation/maturation/survival), and many other processes also has relevance to other neurological diseases. Therefore, PPARγ is an attractive target for therapies against a number of progressive neurological disorders.

1. Introduction

Peroxisome proliferator-activated receptor γ (PPARγ) is a member of the pleiotropic nuclear receptor 1C (NR1C) family (also known as the PPAR family) (Janani and Ranjitha Kumari, 2015). This nuclear receptor family encompasses a group of ligand-modulated transcription factors with broad tissue distributions and a wide array of target genes and functions (Grygiel-Gorniak, 2014). PPARγ is a master gatekeeper of the expression of numerous genes, including G-protein coupled receptors, growth factors, antioxidant enzymes, stem cell genes, kinases, cytokines/chemokines, pro-inflammatory factors, ion channels, and transporters. PPARγ plays a well-established role in the regulation of adipocyte differentiation and the metabolism of lipid and glucose (Janani and Ranjitha Kumari, 2015). Thus, synthetic PPARγ ligands such as thiazolidinediones (TZDs or glitazones) are currently prescibed worldwide to treat hyperglycemia and diabetes and can be readily side-tracked for use in other conditions. Aside from its role in metabolic regulation, anti-inflammatory and protective effects of PPARγ have also been widely studied. For example, PPARγ activation is known to mitigate neuroinflammation and exert direct neuronal protection after central nervous system (CNS) injuries (Gillespie et al., 2011).

Acute CNS injuries are among the leading causes of disability, mortality, and morbidity worldwide (Murray and Lopez, 1997). Tissue damage in these conditions is elicited by both primary and secondary mechanisms, and impacts almost every component of the CNS, including grey matter, white matter, and the vascular network. A growing body of evidence indicates that PPARγ is critically involved in the long-term promotion of tissue repair and rescue of brain cells. In addition, PPARγ directly or indirectly controls the manifestation of comorbid diseases such as diabetes and hypertension, which can profoundly influence the onset and outcome of CNS injuries. Given its dominant position as a master gatekeeper of gene expression, PPARγ is an attractive therapeutic target for injury conditions such as ischemic stroke, hemorrhagic stroke, traumatic brain injury (TBI), and spinal cord injury (SCI).

In this article, we briefly review the numerous functions of PPARγ and their effects on CNS injury. We begin by describing the pathophysiological mechanisms underlying CNS injury in both the acute injury phase and the chronic repair phase. We discuss the contributions of comorbidities to increased disease incidence and negative clinical outcomes and their modulation by PPARγ. The key role of PPARγ in tissue repair and regeneration is highlighted throughout this review. Finally, we will end with our perspectives on PPARγ-related translational therapies.

2. PPARγ overview

2.1 PPARγ structure and CNS distribution

PPARs, or NR1Cs, were first cloned in rodent hepatocytes in the 1990s, and found to be activated by a group of peroxisome proliferators (Issemann and Green, 1990). PPARs are ligand-activated transcription factors that influence the expression of a number of genes, among which metabolism-related genes are the best documented (Michalik and Wahli, 2006). Three isotypes of PPARs have been identified in mammals—PPARα (NR1C1), δ/β (NR1C2), and γ (NR1C3). As with all other nuclear receptors, PPARs share similar structural features and contain the following functional domains: a N-terminal ligand-independent functional domain, a DNA-binding domain that specifically recognizes peroxisome proliferator response elements (PPREs), a flexible hinge region, and a C-terminal ligand binding domain (LBD) (Zoete et al., 2007). The three PPAR isotypes are encoded by genes located on different chromosomes (Greene et al., 1995) and exhibit distinct tissue distribution patterns and biological functions. PPARα is mainly expressed in tissues with high fatty acid metabolism, including liver, kidney, and white and brown adipose tissue (Delerive et al., 2000; Lo Verme et al., 2005; Neschen et al., 2007). PPARδ/β is ubiquitously and abundantly expressed in a broad range of tissues and regulates fatty acid oxidation (Wang et al., 2003b; Stephen et al., 2004). PPARγ consists of two isoforms in human and rodents—PPARγ1, a ~54.5 kDa shorter form, and PPARγ2, a ~57.6 kDa longer form (Chen et al., 2012) (Figure 1A). PPARγ2 is restricted to adipose tissue and plays an important role in adipocyte differentiation. In contrast, PPARγ1 is expressed in various cell types, including brain cells such as neurons and glia, and bone marrow derived immune cells (Zhu et al., 1995; Elbrecht et al., 1996)

Figure 1.

PPARγ structure and post-translational modifications (PTMs). (A) Human PPARγ consists of two isoforms, PPARγ1 (477 aa) and PPARγ2 (505 aa). Like all other nuclear receptors, PPARγ has four functional domains: a ligand-independent activation domain, a DNA binding domain, a hinge domain, and a ligand-binding domain. (B) PPARγ activity is regulated by PTMs. Ser112 phosphorylation results in different transcriptional outcomes depending on the kinases involved. CDK7/9 mediated Ser112 phosphorylation leads to increased PPARγ activity whereas MAPK elicits the opposite effect. Lys107 SUMOylation is associated with increases in PPARγ activity. Ser273 phosphorylation results in decreased insulin sensitivity. SUMOylation of Lys395 mediates transrepression. Abbreviations: Mitogen-activated protein kinases (MAPKs), cyclin-dependent kinase (CDK).

Post-translational modulations (PTMs) of specific amino acids influence PPARγ activity and its functional states (Kim et al., 2013). The major types of PTMs that regulate PPARγ function are phosphorylation, SUMOylation, and ubiquitination (Figure 1B). Ser112 phosphorylation decreases PPARγ activity if induced by mitogen-activated protein kinases (Hu et al., 1996; Shao et al., 1998), but increases PPARγ activity if induced by cyclin-dependent kinase (CDK) 7 and CDK9 (Compe et al., 2005; Iankova et al., 2006). CDK5-mediated Ser273 phosphorylation of PPARγ leads to reduced insulin sensitivity (Choi et al., 2010; Choi et al., 2011). SUMOylation at Lys107 results in decreased activity, whereas Lys395 SUMOylation is strongly associated with PPARγ transrepression of nuclear factor (NF)-κB (Yamashita et al., 2004; Pascual et al., 2005; Jennewein et al., 2008; Pourcet et al., 2010). The latter function is particularly important in anti-inflammatory effects and will be descibed in Section 2.3. Ubiquitination of PPARγ tags it for proteasomal degradation, a process that is enhanced by interferon (IFN)-γ (Waite et al., 2001) and tumor necrosis factor (TNF)-α (He et al., 2008), and repressed by sirtuin 1 (Picard et al., 2004; Floyd et al., 2008).

All three subtypes of PPARs are expressed in the CNS, albeit at different levels (Elbrecht et al., 1996; Braissant and Wahli, 1998; Cullingford et al., 1998; Moreno et al., 2004). Whereas PPAR δ/β is broadly and robustly expressed in the CNS, PPARα and PPARγ exhibit a more restricted pattern of distribution (Braissant and Wahli, 1998; Cullingford et al., 1998; Moreno et al., 2004). Under physiological conditions, PPARγ is readily detectable in select areas, such as the basal ganglia, thalamus, piriform cortex, and hippocampus (Braissant and Wahli, 1998 ; Moreno et al., 2004) and this expression is mainly in neuronal cells (Ferguson et al., 2014). A fraction of astrocytes (20-40%) also express PPARγ, predominantly in processes rather than somata (Warden et al., 2016). Although PPARγ expression has been observed in microglial cultures (Bernardo et al., 2000), it is barely detectable in this cell type in vivo under physiological conditions. However, lipopolysaccharide (LPS) stiumlation significantly increases microglial PPARγ expression in the brain, indicating that microglial PPARγ expression may be dependent on inflammation status (Warden et al., 2016).

2.2 PPARγ ligands

PPARγ ligands can be classified into three major categories. Category A includes natural (endogenous) agonists, such as unsaturated fatty acids, eicosnanoids, oxidized phospholipids, and nitroalkenes. Among these agonists, an eicosanoid termed prostaglandin metabolite 15-deoxy-D12, 14 prostaglandin J2 (15d-PGJ2) was the first specific endogenous ligand of PPARγ to be identified and has been widely used in studies of PPARγ in various disease models (Forman et al., 1995). Category B includes synthetic agonists: 1) TZDs such as pioglitazone (PGZ), ciglitazone (CGZ), triglitazone (TGZ), and rosiglitazone (RGZ), 2) non-TZD agonists, 3) dual-α/γ agonists, 4) pan-α/β/γ agonists, and 5) selective PPARγ modulators. Among these, TZDs were the first to be synthesized and are widely used in diabetic patients (Lehmann et al., 1995), as well as in experimental models to assess the effects of PPARγ activation. Category C includes synthetic antagonists, such as bisphenol A diglyceryl ether, GW9662, LG100641, PD068235, T0070907, and SR-202. Among these, GW9662 is an irreversible PPARγ antagonist that has been widely applied in research on PPARγ (Leesnitzer et al., 2002; Chen et al., 2012).

It is important to note that almost all of the aforementioned PPARγ agonists not only activate PPARγ, but also activate other PPARs, as well as PPAR-independent pathways (Park et al., 2003; Bernardo et al., 2009). As a result, the effects of PPARγ agonists, including TZDs, work through both PPARγ-dependent and independent mechanisms that are not easily differentiated and can confound the interpretations of studies. Furthermore, regardless of the route of delivery (iv, ip, oral, or intracerebral), TZDs can affect PPARγ in the CNS (Culman et al., 2007), although it is not known whether all TZDs cross the blood-brain barrier (BBB). PGZ has been reported to cross the BBB (Grommes et al., 2013), but unpublished observations by GlaxoSmithKline have been used to claim that RGZ cannot cross the BBB (Pedersen et al., 2006), although this awaits verification.

2.3 PPARγ functioning patterns

As with all other nuclear receptors, PPARγ functioning patterns involve ligand-independent repression, ligand-dependent transactivation, and ligand-dependent transrepression (Glass and Ogawa, 2006) (Figure 2). As there are no known endogeous PPARγ-inhibiting ligands, all “ligands” in this review refer to PPARγ-activating ligands.

Figure 2.

Functional patterns of PPARγ. (A) Ligand-independent repression and ligand-dependent transactivation. Under resting conditions, the PPARγ/RXR heterodimer is connected to co-repressor (HDAC and NCoR), which blocks PPRE target gene transcription. After binding to the PPARy ligand, conformational changes lead to dissociation of the co-repressor and recruitment of a co-activator (e.g. TRAP220), which triggers transcription of PPRE target genes. (B) Ligand-dependent transrepression. Binding to its ligand leads to SUMOylation of PPARγ, which stabilizes the co-repressor of NF-κB, leading to blockade of NF-κB target gene expression. (C) PPARγ can also inhibit NF-κB target genes by 1) mediating NF-κB degradation, 2) facilitating NF-κB export out of the nucleus, 3) cofactor competition with NF-κB, and 4) steric inhibition of NF-κB binding. Abbreviations: retinoid X receptor (RXR), histone deacetylase (HCAC), nuclear receptor corepressor (NCoR), thyroid hormone-associated protein 220 (TRAP220), peroxisome proliferator response elements (PPREs), ubiquitin (Ub), nuclear factor-κB (NF-κB).

The classic view of PPARγ function is that it forms a heterodimer with the retinoid X receptor (RXR) (Berger and Moller, 2002). Under basal conditions, in the absence of ligands of RXR or PPARγ, the heterodimer is associated with corepressors, such as nuclear receptor corepressor (NCoR), transducin β-like protein, and histone deacetylases (HDACs). This complex binds to PPREs in the promoter region of target genes and retains the genes in a suppressed state defined as ligand-independent repression. Following ligand binding, however, the PPARγ/RXR heterodimer undergoes conformational changes, dissociates from co-repressors, and recruits co-activators such as thyroid hormone-associated protein 220, thereby upregulating PPRE target gene transcription (Dekkers et al., 2012). This effect is known as ligand-dependent transactivation (Figure 2A). Ligands of either PPARγ or RXR can both initiate ligand-dependent transactivation, whereas binding of both receptors simultaneously can lead to additive effects on gene expression (Fajas et al., 1997). Both ligand-independent repression and ligand-dependent transactivation involve the regulation of promoters located at PPREs, and PPARγ is known to maintain lipid metabolic homeostasis by these mechanisms.

Ligand-dependent transrepression is more complex and denotes PPARγ-dependent repression of genes other than PPREs, typically NF-κB and activator protein-1, in the absence of direct binding of PPARγ to DNA. The HDAC3/NCoR complex binds to NF-κB to maintain a repressed state, which is terminated upon NCoR degradation. Ligand-dependent transrepression of PPARγ occurs when ligands induce PPARγ SUMOylation, which guides PPARγ to the HDAC3/NCoR complex, inhibits NCoR degradation, and causes NF-κB target gene repression (Pascual et al., 2005; Dekkers et al., 2012) (Figure 2B). As NF-κB is part of a prototypical pro-inflammatory signaling cascade, SUMOylation-dependent transrepression is the predominant means by which PPARγ promotes anti-inflammatory effects. Other mechanisms involved in PPARγ-induced NF-κB inhibition include ubiquitin-dependent degradation of NF-κB, exportation of NF-κB out of the nucleus, cofactor competition with NF-κB, and steric inhibition of NF-κB binding to DNA (Sauer, 2015) (Figure 2C). The sophisticated, multidimensional nature of PPARγ PTMs and functioning patterns confers this protein with the flexibility demanded of a master gatekeeper.

2.4 PPARγ target genes

The observation that PPARγ knockout is embryonically lethal indicates a vital and indispensible role for this protein (Semple et al., 2006). PPARγ response networks are complex and different genes are regulated in cell-specific manners. PPARγ regulates a variety of target genes relevant to adipogenesis (e.g. adipoq, lpl, nr1h3, ucp1), fatty acid metabolism (e.g. acadl, acadm, acox1), lipid transport (e.g. angptl4, apoe, olr1), cell proliferation (e.g. clu, eln, hspd1), insulin signaling (e.g. cpt1a, dgat1, pck1), and inflammation (e.g. nfkb, mmp9). The activation of PPARγ increases metabolic activities such as lipid metabolism and storage, adipogenesis, insulin sensitization, glucose homeostasis, and sodium and fluid retension (Ahmadian et al., 2013). Nevertheless, not all of these functions involve direct transcriptional regulation. The regulated genes can be grouped into three major categories (Chen et al., 2012): 1) fatty acid/glucose metabolism, 2) inflammation/oxidative stress/apoptosis, and 3) cancer. A large variety of genes are found in the area of overlap of these three categories and positive regulation is more common than negative regulation. Most of PPARγ signal transduction cascades in CNS pathologies are directly or indirectly associated with NF-κB, and this serves as a major mechanism whereby PPARγ regulates the stress response to injuries in the brain.

In addition to NF-κB, PPARγ also regulates the prototypical redox-sensing nuclear factor erythroid 2- related factor 2 (Nrf2) pathway, acting in synergy with this pathway to mute the destructive effects of oxidative stress (Figure 3) (Ikeda et al., 2000; Shih et al., 2005; Cho et al., 2010; Polvani et al., 2012; Zhao and Aronowski, 2014). Nrf2 is a master transcription factor that binds the antioxidant response element (ARE) and activates numerous genes critical for the maintenance of redox homeostasis, such as those related to the glutathione and superoxide dismutase antioxidant systems (Zhang et al., 2013). The interactions between PPARγ-Nrf2 signaling pathways can be classified into four categories. First, in addition to PPARγ-mediation modulation of the Nrf2/ARE axis, Nrf2 also regulates PPARγ, suggestive of a bidirectional loop (Cho et al., 2010). For example, Nrf2-mediated PPARγ induction protects mice against acute oxidant damage to the lung in vivo (Cho et al., 2010). Second, some antioxidant genes, such as catalase (Guan et al., 2000; Girnun et al., 2002), glutathione S-transferase (Rushmore and Pickett, 1990; Rushmore et al., 1991; Park et al., 2004a) and superoxide dismutase (SOD) (Yoo et al., 1999; Zelko et al., 2002; Jurkunas et al., 2010) contain both a PPRE and an ARE and are regulated by both PPARγ and Nrf2 to elicit anti-oxidative effects. Third, microglial/macrophagic CD36 is under the regulation of both PPARγ (Tontonoz et al., 1998; Yamanaka et al., 2012) and Nrf2 (Ishii et al., 2004; Wang et al., 2014; Zhao et al., 2015a). CD36 is important in microglial/macrophage phagocytosis and subsequent debris clearance, which facilitates resolution of neuroinflammation (Li et al., 2015). Fourth, PPARγ and Nrf2 synergistically elicit anti-inflammatory effects by inhibition of the NF-κB pathway (Wardyn et al., 2015; Zhao et al., 2015b).

Figure 3.

Interactions between PPARγ and Nrf2 pathways. (1) A bidirectional loop exists between PPARγ and Nrf2. PPARγ can upregulate Nrf2 expression and vice versa. (2) PPARγ and Nrf2 synergistically increase the expression of some antioxidants, including catalase, SOD, and GST, which have both PPRE and ARE elements in their genes. These proteins inhibit the generation of ROS, thereby mitigating neuroinflammation and improving debris clearance. (3) Microglial CD36 is regulated by both PPARγ and Nrf2. CD36 is critical for microglial phagocytosis and subsequent debris clearance, both of which facilitate the resolution of neuroinflammation. (4) PPARγ and Nrf2 synergistically inhibit NF-κB, the major proinflammatory pathway. Abbreviations: nuclear factor erythroid 2-related factor 2 (Nrf2), glutathione S-transferase (GST), reactive oxygen species (ROS), superoxide dismutase (SOD).

PPREs have been found in target genes with potential roles in tissue preservation and repair and functional recovery after acute brain injury. Examples include tight junction protein zona occludens-3 (BBB integrity), synaptogyrin 4 (synaptic function), ubiquitin conjugating enzyme E2 S (protein quality control), G1/S-specific cyclin, p21-activated kinase 4, homeodomain-interacting protein kinase 4 (cell cycle), mitogen activated protein kinase organizer 1, transciption factor jun-D (cell survival), TOMM40, TIMM13, TIMM44, and electron transfer flavoprotein subunit beta (mitochondrial function) (Heinaniemi et al., 2007). Furthermore, a genome-wide identification of PPREs revealed functional clusters of PPAR-regulated genes related to chromatin remodeling, DNA damage response, cell differentiation, and cell growth/maintenance (Lemay and Hwang, 2006). Aside from the tissue reparative nature of these target genes, PPREs with well-established roles in lipid metabolism and inflammation might be critical for the repair of myelin and clearance of debris in the CNS. Indeed, a critical role for PPARγ in peripheral wound repair has been well established (Yessoufou and Wahli, 2010); PPARγ is activated by lipid mediators produced by necrotic tissue, and upregulates genes involved in anti-fibrotic, anti-oxidative, and anti-apoptotic effects (Michalik and Wahli, 2006). As discussed further below, the genomic studies identifying potential PPREs are consistent with the view that PPARγ lies at the nexus between perturbations of the internal milieu and a concerted effort to transcribe an abundance of genes that preserve homeostasis and repair tissue.

3. Evidence favoring a beneficial role of PPARγ in the injured CNS

CNS injuries encompass a variety of pathogenic mechanisms, including neurovasculopathy, trauma, neurodegeneration, tumor formation, and CNS autoimmunity, and may arise due to systemic diseases. In the present review, we will restrict our focus to those conditions with CNS abnormalities as the primary underlying pathology. This group of CNS injuries—ischemic stroke, hemorrhagic stroke, TBI, and SCI—share a number of pathophysiologic/repair processes in common. In the next section, we briefly describe alterations in PPARγ expression in CNS injuries and present evidence favoring its beneficial effects. We will then introduce the mechanisms underlying PPARγ-mediated protection against CNS injury, promotion of CNS repair, and regulation of comorbidities (Figure 4).

Figure 4.

Pathophysiology of CNS injuries and summary of PPARγ effects. The acute phase of CNS injuries (e.g. stroke, TBI, SCI, etc.) is characterized by neuroinflammation, neuronal death, demyelination, and BBB disruption. In the chronic recovery phase, repair processes are initiated, such as resolution of inflammation, clearance of debris and infiltrating cells, neurogenesis, remyelination, angiogenesis, and BBB repair. Comorbidities affect the risk of CNS injuries, and influence the course of the injury and self-reparatory activities. PPARγ inhibits the detrimental effects of injuries, promotes repair, and mitigates comorbidities. Abbreviations: blood brain barrier (BBB).

3.1 PPARγ expression and activity following CNS injuries

Ischemic stroke is caused by transient or permanent local reduction of cerebral blood flow, and is characterized by a number of cellular disturbances, including BBB dysfunction, vasogenic brain edema, and neuronal death (Yin et al., 2014; Krueger et al., 2015). In response to middle cerebral artery occlusion (MCAO), PPARγ mRNA levels are upregulated by 3.3-fold and 7.5-fold in the ipsilateral cerebral cortex of the rat at 6h and 24h post-MCAO, respectively (Ou et al., 2006). In situ hybridization studies further reveal that PPARγ mRNA expression is restricted to neurons within the ischemic territory (Ou et al., 2006). Similarly, Victor et al. observed PPARγ upregulation at both the mRNA and protein levels in ischemic neurons following MCAO in rats (Victor et al., 2006). Elevated expression of PPARγ protein was detected by immunohistochemistry as early as 4h after stroke, with a peak at 24h, and lasted for at least 14d after stroke (Victor et al., 2006). Despite increased PPARγ expression in neurons after ischemic insults, there is decreased PPARγ DNA binding activity in the ischemic hemisphere. In the ischemic brain, PPARγ-PPRE binding activity 8 h after the insult is only ~44% of the values in the contralesional brain, as evaluated by electrophoretic mobility shift assays and densitometry (Victor et al., 2006). The reason for the reduction in the PPRE binding activity of PPARγ after ischemic injury is still unknown. It is possible, however, that PPARγ protein levels may be elevated in the ischemic brain to compensate for the decrease in PPARγ DNA binding activity.

Hemorrhagic stroke is elicited by the rupture of a weakened blood vessel or a brain aneurysm followed by bleeding into surrounding brain regions. There are two subtypes of hemorrhagic stroke—intracerebral hemorrhage (ICH) and subarachnoid hemorrhage (SAH) (Qureshi et al., 2001; Tholance et al., 2015). Xu et al. injected autologous whole blood into the right striatum of male Sprague-Dawley rats to induce an ICH model (Xu et al., 2016). Significantly elevated expression of PPARγ and 15(S)-hydroxyeicosatetraenoic acid (15(S)-HETE) was found in the ipsilateral brain in ICH rats compaired to sham-treated controls (Xu et al., 2016). Furthermore, PPAR-γ protein levels and PPARγ-targeted gene expression were significantly increased in rats treated with the PPARγ agonist, exogenous 15(S)-HETE, in comparison to vehicle-treated ICH rats (Xu et al., 2016). Intrastriatal injections of 15d-PGJ₂ significantly increased PPARγ-DNA binding activity and the expression of catalase in the perihemorrhagic area 3h after ICH (Zhao et al., 2006). However, in an experimental SAH model, PPARγ mRNA and protein levels were significantly reduced in the cortex compared to sham controls (Chang et al., 2015a).

Following TBI, cerebral inflammation is one of the major triggers of progressive neuronal cell death (Morganti-Kossmann et al., 2001). In a TBI model induced by controlled cortical impact (CCI), PPARγ mRNA expression was not altered either following TBI or treatment with the PPARγ agonists PGZ and RGZ (Thal et al., 2008; Thal et al., 2011). However, inflammatory genes such as inducible nitric oxide synthase (iNOS), TNF-α, interleukin (IL)-1β and IL-6 were upregulated following TBI, and this response was inhibited by PGZ treatment (Thal et al., 2011). In contrast, Yi et al. reported a two-fold increase in PPARγ mRNA expression 24h after experimental TBI and this was increased four-fold by RGZ treatment (Yi et al., 2008). The methodological differences in experimental trauma models might explain these discrepancies.

It is well accepted that apoptosis, inflammation, and excitotoxicity contribute to secondary neuronal damage after SCI. One day following experimental SCI, the expression of PPARγ is significantly increased in the spinal cortex surrounding the lesion area (Zhang et al., 2010). In contrast, another study demonstrated that PPARγ protein levels were significantly reduced in spinal cord homogenates one day after SCI (Paterniti et al., 2013).

3.2 Protective effects of PPARγ in CNS injuries

A large body of seminal work in healthy control, diabetic, or obese animals revealed that PPARγ agonists decrease myocardial infarct size and increase glucose uptake and insulin sensitivity (Yue Tl et al., 2001; Khandoudi et al., 2002; Shiomi et al., 2002; Sidell et al., 2002; Wayman et al., 2002; Lee and Chou, 2003; Liu et al., 2004). Consistent with these studies, mounting evidence also demonstrates that PPARγ activation is beneficial to the injured brain (Zhao and Aronowski, 2014). For example, Zhao and colleagues demonstrated that neuron-specific PPARγ knockout mice experience greater brain damage and oxidative stress after MCAO (Zhao et al., 2009b). Furthermore, systemic or intracerebroventricular administration of PPARγ agonists, either prior to and/or after stroke, elicit potent neuroprotective effects in various animal models of ischemic stroke, as shown by reductions in infarct size and superior neurological function (Sundararajan and Landreth, 2004; Luo et al., 2006; Wang et al., 2009; Kaundal and Sharma, 2010). Even delayed intraperitoneal administration of the PPARγ agonist RGZ—beginning 24 h after cerebral ischemia—is able to confer protection in a middle cerebral artery embolization stroke model in rats (Allahtavakoli et al., 2009). These findings may be important because they suggest a wide therapeutic window for PPARγ agonists. However, the dose must be chosen with care, as PPARγ agonists may promote cell survival at low concentrations but increase cell loss at high concentrations (Wang et al., 2002b; Lin et al., 2006; Wu et al., 2009).

Aside from ischemic stroke, PPARγ agonists also exhibit neuroprotective effects in hemorrhagic stroke models (Zhao et al., 2006; Zhao et al., 2007; Zhao et al., 2009a; Wu et al., 2015). For example, in a rat model of striatal ICH, intrastriatal 15d-PGJ₂ treatment increased PPARγ DNA binding activity and subsequent catalase expression, resulting in significant decreases in neurologic deficits compared to rats receiving saline (Zhao et al., 2006). Similarly, intraperitoneal administration of RGZ at 5 min, 6 h, and 24 h after TBI increased PPARγ mRNA levels in cortical tissue surrounding the contusion area within 24 h after brain injury and significantly reduced the cortical lesion volume compared to vehicle-treated mice (Yi et al., 2008). Similarly, the survival of hippocampal CA3 neurons was increased after intraperitoneal injection of 6 mg/kg RGZ in a controlled cortical impact model of TBI (Liu et al., 2016). Nevertheless, a number of studies have shown contradictory results, particularly in TBI models. It has therefore been argued that the neuroprotective effects of PGZ in TBI—as shown by reduced contusion volume, decreased cerebral inflammation, and reduced microglial activation—are the result of PPARγ-independent mechanisms (Thal et al., 2011; Liu et al., 2016). Other possible reasons for these discrepancies include: 1) differing rodent species and strains; 2) varying degrees of injury severity, reflecting differences in the diameter of the impactor tip, impact velocity, impact duration, and displacement; 3) differences in lesion location within the brain; 4) different administration methods and timing of PPARγ agonists; and 5) different times of evaluation. In short, the interpretation of this body of work awaits further confirmation and greater reproducibility across studies.

Application of PPARγ agonists is protective in various animal models of SCI. In a study by McTigue and colleagues, rats received a moderate mid-thoracic contusion and were intraperitoneally injected with either low or high doses of PGZ or vehicle at 15 min after SCI and then every 12h for 7 days (McTigue et al., 2007). In this study, PGZ significantly protected white matter, gray matter, and motor neurons from SCI. In a compression model of SCI, vascular clips were applied to the mouse dura with a force of 24g via four-level T₅-T₈ laminectomy (Genovese et al., 2008). In this model, intraperitoneal injections of the PPARγ agonist 15d-PGJ₂ ameliorated spinal cord inflammation, tissue injury, neutrophil infiltration, NF-κB activation, and cell apoptosis after SCI. In addition, the PPARγ agonist PGZ was found to ameliorate sensory dysfunction after SCI (Iwai et al., 2008) and reduce neuropathic pain after SCI and peripheral nerve injury (Jia et al., 2013; Griggs et al., 2015). Although there remain some debates, the abovementioned findings suggest that PPARγ agonists elicit structural and functional protection of the brain in diverse injury models.

4. PPARγ protects against CNS injury

A series of pathological processes are initiated during the acute phase following CNS injuries. Following the acute period, a wave of secondary injury expansion also emerges. Necrosis of neural tissue triggers a powerful inflammatory cascade, resulting in further neuronal apoptosis and excitotoxicity, oxidative stress, and other feed-forward mechanisms. Neuroinflammation contributes to the amplification and spread of local injury, predisposing surrounding cells to secondary injuries (Kawabori and Yenari, 2015). Vulnerable brain vessels and the BBB release pro-inflammatory factors and facilitate immune cell infiltration into the injured brain (de Wit et al., 2017). In this section, we will summarize pathophysiological processes after CNS injury and the beneficial effects of PPARγ in these processes.

4.1 PPARγ protects against neuroinflammation following CNS injuries

The immune system serves as the sentinel and guardian of the organism. For many decades, the CNS was falsely believed to be immunoprivileged. The CNS is now viewed as immune competent, actively engaging in interactions with the peripheral immune system (Carson et al., 2006). Furthermore, a lymphatic vascular system has been discovered in the brain (Molina-Holgado and Molina-Holgado, 2010; Aspelund et al., 2015) and neural inflammation is thought to drive disease processes into a secondary phase of injury expansion.

4.1.1 An introduction on neuroinflammation following CNS injuries

Inflammatory responses following CNS injury are initiated by damage-associated molecular patterns (DAMPs) that are released from injured tissue (Sharma and Naidu, 2016). High mobility group box 1 is rapidly released by cells undergoing necrotic or pyroptotic cell death (Sharma and Naidu, 2016). High mobility group box 1 is detected in the serum half an hour after 90 min stroke in rodents, and is observed within the first hour after symptom onset in ischemic stroke patients (Liesz et al., 2015). In addition, heat shock protein (HSP) 72 can be detected in the serum of CNS trauma patients (Pittet et al., 2002). Release of S-100 protein from the injured CNS occurs within the first 24h in patients with TBI, ischemic stroke, and transient ischemic attacks (Elting et al., 2000). Furthermore, the early release of nuclear DNA, the IL-1 family, histones, and other intrinsic cell components from the injured CNS serves to ignite the inflammatory response after acute CNS injuries (Sharma and Naidu, 2016).

Both local and systemic inflammatory cells participate in acute CNS injury. Microglia are the resident immune cells of the brain and are among the first cells to respond to CNS injury. Injury signals are transported through the CNS lymphatic vessels and the blood vessels, possibly delivered by the drainage of damage-associated molecular patterns directly to peripheral immune organs (Louveau et al., 2015). Injury signals might also be carried by dendritic cells (DCs) (Clarkson et al., 2012). Macrophages take up residence in the CNS vessel bed and are also poised to sense and transmit information about disruptions in homeostasis (Goverman, 2009; Kivisakk et al., 2009). Activation of the peripheral immune response results in a surge of infiltrating immune cells that home in on the CNS lesion zone (Pennypacker and Offner, 2015).

In ischemic stroke, the pro-inflammatory profiles of microglia may be inactivated at early stages after ischemic stroke. However, as the immune reactions progress, pro-inflammatory microglia participate in the promotion of neural inflammation (Hu et al., 2012; Patel et al., 2013; Wang et al., 2013; Jin et al., 2017). In a transient MCAO model, brain infiltration of macrophages appeared as early as 12h after ischemia and peaked after 24h, followed by a gradual decrease by day 7 (Gelderblom et al., 2009). Similarly, infiltrating neutrophils were detected at 12 h, peaking on day 2-3 and gradually disappearing by day 7 (Gelderblom et al., 2009). Although a deleterious role has been reported for natural killer cells in ischemic stroke (Gan et al., 2014), the infiltration of this cell population is not always evident. However, T and B lymphocytes from the adaptive immune systems appear in the ischemic lesion site at 3-7 days after stroke (Gelderblom et al., 2009), even though they represent only a small portion of infiltrated immune cells. In animal models of hemorrhagic stroke, microglial activation is observed 1-4 h after hemorrhage, peaking at 3-7 days, and returning to baseline by 21-28 days (Hickenbottom et al., 1999; Gong et al., 2000; Xue and Del Bigio, 2000; Wang et al., 2003a; Wang and Tsirka, 2005). Neutrophil infiltration occurs as early as 4h after the formation of the hematoma, peaking at 2-3 days, and subsiding by 7 days.

In traumatic CNS injuries such as TBI and SCI, the BBB is directly disrupted by the primary damage, followed by the entry of blood components into the CNS (McKee and Lukens, 2016). As in peripheral injured tissue, neutrophils are the first to be mobilized to the brain after TBI. In a rat model of CCI, neutrophil infiltration was detected within hours post-trauma, and this further exacerbated the loss of BBB integrity. Microglia and macrophages are increased between 12 and 72 h in damaged cortical regions after TBI (Clark et al., 1994; Soares et al., 1995). A large number of chemokines are released into the injured CNS, inducing the migration of activated neutrophils, lymphocytes, monocytes, and macrophages (Ghirnikar et al., 1998).

Together with the resident microglia in the CNS, infiltrated peripheral immune cells promote ongoing neural inflammation with direct cytotoxicity or secretion of inflammatory factors. Microglia produce proinflammatory cytokines such as IL-1, IL-6, IL-12, IL-23, nitric oxide (NO), and TNFα, and contribute to local oxidative stress by inducing iNOS and generating reactive oxygen species (ROS) (Hu et al., 2012; Wang et al., 2013). CD8⁺ cell cytotoxic T lymphocytes (CTLs) and natural killer cells produce cytolytic granzyme-b (Chaitanya et al., 2011) and perforin (Mracsko et al., 2014) to induce direct neuronal death after ischemic stroke. Immune cell-derived cytokines, including IFN-γ, TNF-α, various interleukins, macrophage inflammatory protein 1α and 1β, and monocyte chemotactic protein (MCP)-1β not only convey immune signals but also induce apoptosis of cells in the stroke penumbra (Jin et al., 2010b; Gan et al., 2014). In addition, chemokines are produced in the inflamed zone of injury, both by infiltrating leukocytes and the injured brain tissue itself, in order to attract even more peripheral immune cells in a self-amplifying cascade (Losy et al., 2005; Jin et al., 2010b; Wang et al., 2012b; Wolinski and Glabinski, 2013). Matrix metalloproteinases (MMPs) are released mainly from neutrophils and cause damage to the BBB, facilitating the entry of additional peripheral leukocytes. In addition, activated endothelial cells in microvessels express adhesion molecules that promote the rolling and firm adhesion of leukocytes and platelets (Yilmaz and Granger, 2008). It is worth noting that the deleterious effects of the aforementioned inflammatory factors are relatively non-specific, resulting in broad injury to cells in the lesion zone and surrounding tissues, including neurons, white matter-forming oligodendrocytes, and all components of the BBB.

4.1.2 PPARγ protects against post-injury neuroinflammation

Widespread expression of PPARγ has been detected in various immune cells, including CNS glial cells (Bernardo et al., 2000) and peripheral leukocytes (Ricote et al., 1998; Clark et al., 2000; Faveeuw et al., 2000; Padilla et al., 2000; Zhang et al., 2004a). A strong suppressor of the pro-inflammatory transcriptional factor NF-κB, PPARγ fine-tunes neural inflammation by limiting the activation of immune cells, reducing the generation of pro-inflammatory cytokines, chemokines, and adhesion molecules, and mitigating oxidative stress (Kapadia et al., 2008).

Treatment with PGZ reduces microglial proliferation in stroke-prone renovascular hypertensive rats (Lan et al., 2015). In addition, RGZ decreases parenchymal accumulation of neutrophils in a mouse MCAO model (Luo et al., 2006) and 15d-PGJ2 inhibits neutrophil infiltration following ICH injury (Zhao et al., 2006). In rodent TBI models, RGZ reduces activated microglia/macrophages, and this can be reversed by GW9662 (Yi et al., 2008). In vitro experiments reveal that RGZ impedes the maturation of human monocyte-derived DCs (Gosset et al., 2001). Although it is not known if lymphocyte activation is affected by PPARγ, the tempering of DC activation suggests this to be the case.

A large body of evidence indicates that activation of PPARγ attenuates neural inflammation by reducing inflammatory mediators in ischemic stroke (Culman et al., 2007), ICH (Zhao et al., 2007; Aronowski and Zhao, 2011), TBI (Qi et al., 2010), and SCI (Zhang et al., 2010; Li et al., 2013). PPARγ stimulation inhibits the expression of a wide array of pro-inflammatory mediators in microglia and macrophages, including TNF-α, IL-1β, IL-6, iNOS, inducible cyclooxygenase (COX) 2, and the IL-12 family (Bernardo et al., 2000; Luna-Medina et al., 2005; Storer et al., 2005). Appel et al. reported that 15d-PGJ2 or TGZ treatment of magnet-sorted human DCs (stimulated with toll-like receptor ligands) significantly down-regulated the production of cytokines (IL-12, IL-6, TNF-α, etc.) and chemokines (CCL2, CCL3, CCL5) that are involved in T cell recruitment and activation (Appel et al., 2005). Furthermore, PPARγ agonists have been shown to reduce the expression of IFNγ (da Rocha Junior et al., 2013), IL-17, and IL-23 in T cells (Klotz et al., 2009). In human endothelial cells, adenovirus-mediated expression of a constitutively active, mutant version of PPARγ suppressed the expression of vascular adhesion molecules intracellular adhesion molecule-1, vascular cell adhesion molecule-1 and E-selectin (Wang et al., 2002a).

Severe neural inflammation following acute CNS injury causes apoptosis of neurons (Aktas et al., 2007), degeneration of white matter (Rosenberg, 2009), and breakdown of the BBB (Hawkins et al., 1991). PPARγ activation ameliorates these consequences of neural inflammation and oxidative stress in various acute CNS diseases (Culman et al., 2007; Qi et al., 2010; Zhang et al., 2010; Aronowski and Zhao, 2011). However, PPARγ may also be essential for microglial survival (Liu et al., 2010; Kane et al., 2011; Pan et al., 2013). Furthermore, a PPARγ agonist has been shown to restore compromised pro-inflammatory responses in diabetic mice soon after stroke, whereas inflammatory responses were suppressed at later time points in both control and diabetic mice (Kumari et al., 2010). These data demonstrate a sophisticated, bidirectional regulation of microglia by PPARγ.

4.2 Neuronal death

4.2.1 Mechanisms of neuronal death following CNS injury

CNS injury-associated neuronal death can be classified as primary or secondary. Primary neuronal death arises from energy failure in ischemic stroke, mechanical damage associated with mass effects in intracerebral hemorrhage (Qureshi et al., 2009), and mechanical damage from blunt force traumas in TBI and SCI (McTigue, 2008). The loss of neurons typically occurs immediately following injury and continues for hours to days. An important feature of primary neuronal death is that it is generally believed to be non-regulatable and non-intervenable. On the contrary, secondary neuronal death may be regulated with interventions and is therefore the focus of the following discussion.

Glutamate receptors belong to two major classes—G-protein coupled metabotropic receptors and ligand-gated ionotropic receptors, such as the N-methyl-D-aspartic acid (NMDA) receptor, α-amino-3-hydroxy-5-methylisoxazole-4-propionic acid (AMPA) receptor, and kainate receptor. Excitotoxicity is induced by excessive activation of glutamate receptors by extracellular glutamate and is a major contributor to neuronal death following CNS injury (Puyal et al., 2013). Energy failure after ischemia leads to decreased ATP production, thereby disrupting Na⁺/K⁺-ATPase, Ca²⁺/H⁺-ATPase and Na⁺/Ca²⁺-transporters and leading to depolarization of the cellular membrane. Diffusely depolarized presynaptic neurons release excessive glutamate into the synaptic cleft (Terasaki et al., 2014; Hamming et al., 2016). Another mechanism of excessive glutamate release is via connexin hemichannels in activated astrocytes and microglia (Takeuchi et al., 2006; Ouyang et al., 2014). Ultimately, over-activation of the NMDA receptor leads to the sustained excitement of neurons and an excessive influx of calcium, which causes loss of neuronal viability (Randall and Thayer, 1992; Tymianski et al., 1993; Owens et al., 1997; Lai et al., 2014).

Aside from excitotoxicity-induced neuronal cell death, oxidative stress may also elicit cell loss in CNS injury conditions. The predominant ROS are the superoxide anion (O₂^▪), hydrogen peroxide (H₂O₂), hydroxyl radical (^▪OH), peroxyl (RO₂^▪), peroxynitrite anion (ONOO^▪), and nitrogen dioxide (^▪NO2) (Dhawan, 2014; Yang et al., 2016). Excessive intracellular Ca²⁺ accumulation results in overactivation of calcium-dependent enzymes such as nicotinamide adenine dinucleotide phosphate oxidase (NOX) and xanthine oxidase, both of which are capable of producing ROS. Disruptions in the electron transport chain also elicit the leakage of ROS from mitochondria (Piantadosi and Zhang, 1996; Galeffi et al., 2016). Other mechanisms of ROS production include lipid peroxidation and increased mitochondrial ROS production stimulated by the release of Ca²⁺, Na⁺, and adenosine diphosphate (ADP) from damaged cells (Schulz et al., 2000; Viola and Hool, 2013). Furthermore, blood components, particularly hemoglobin, can undergo auto-oxidation upon erythrocyte lysis, which may serve as a major source of ROS in ICH, SAH, TBI, and SCI (Simoni et al., 1990; Asano, 1999). ROS are typically released within 1h post-injury in diverse models (Gaetani et al., 1990; Marklund et al., 2001). The cellular source of this oxidative stress might be neurons, other parenchymal cells, and vascular cells (Chen et al., 2014). ROS can directly damage neuronal components by promoting lipid peroxidation, protein breakdown, and DNA damage, and also act as executioners of cell death by activating pro-apoptotic pathways (Zhang et al., 2004b). Finally, ROS can promote neuroinflammation, BBB disruption, and the release of vasospastic mediators that indirectly harm neurons by induction of leukocyte infiltration and enhancement of inflammation, thereby exacerbating ischemic injury (Clark et al., 1993).

4.2.2 PPARγ protects against neuronal death following CNS injury

PPARγ has been shown to decrease excitotoxicity in various pathologic settings. For example, 15d-PGJ2 protects neurons against NMDA-induced toxicity both in vivo and in vitro (Zhao et al., 2006). Similarly, the PPARγ agonist RGZ promotes neuronal survival and neurite outgrowh in vitro following exposure to glutamate (Zuhayra et al., 2011). In a rodent ischemic stroke model, the PPARγ antagonist T0070907 abolished neuroprotection mediated by ischemic preconditioning, and increased glutamate release and reduced glutamate uptake in neuron-astrocyte co-cultures (Romera et al., 2007). Similar effects have also been reported in other disease models. TGZ diminishes glutamate-induced excitotoxicity to the same degree as NMDA receptor antagonists in retinal ganglion cells (Aoun et al., 2003). Furthermore, RGZ decreases glutamate release in glioblastoma multiforme, and this can be reversed by the PPARγ antagonist GW9662 (Ching et al., 2015). The underlying mechanism by which PPARγ reduces excitotoxicity may involve upregulation of the glutamate transporter (GLT)-1 (Romera et al., 2007; Ching et al., 2015). GLT-1 is found in astrocytes and is responsible for up to ~90% of glutamate uptake (Robinson, 1998; Schousboe and Waagepetersen, 2005; Arbo et al., 2016). On the other hand, Yao et al observed no changes in GLT-1 expression in response to RGZ injection in a model of TBI (Yao et al., 2015). In addition, no evidence supporting the effects of PPARγ on excitotoxicity has been reported in SCI and hemorrhagic stroke models. Nevertheless, increased astrocytic glutamate uptake following PPARγ activation would be expected to reduce overall tissue damage (Romera et al., 2007; Ching et al., 2015), as excitotoxicity is a major cause of death not only for neurons, but also for oligodendrocytes and other types of CNS cells (Huria et al., 2015).

The anti-oxidative effects of PPARγ are better established than its effects on excitotoxicity. First, PPARγ can directly inhibit ROS generation, partly by inhibiting ROS-generating enzymes. RGZ has been shown to promote neuronal survival and neurite outgrowh in vitro following exposure to the Parkinson’s oxidative toxicant 6-hydroxydopamine (Zuhayra et al., 2011). Systemic administration of RGZ or PGZ after ischemia/reperfusion reduces ROS and the expression of COX-2, an important enzyme involved in ROS generation, and attenuates neuronal loss (Collino et al., 2006). Similarly, 15d-PGJ2 inhibits iNOS expression and subsequent apoptosis in cerebellar granule cells (Heneka et al., 1999). NOX activation is another major driving force in ROS generation, and its relationship with PPARγ has been examined in primary cortical neurons subjected to oxygen-glucose deprivation (OGD) (Wu et al., 2016). NOX expression was enhanced by both GW9662 treatment and PPARγ-siRNA transfection, and was inhibited by PPARγ transfection or exposure to 15d-PGJ2. Second, PPARγ prevents oxidative stress by induction of anti-oxidant molecules. For example, both RGZ and PGZ can restore glutathione, the most ubiquitous biological antioxidant, in a mouse MCAO model (Collino et al., 2006). Furthermore, in a rodent TBI model, RGZ induces the expression of the antioxidant enzymes catalase and SOD-1 (also known as Cu/Zn SOD), as well as heme oxygenase-1 and the chaperones heat shock protein (HSP) 27 and HSP70, and these responses are abolished by GW9662 administration (Yi et al., 2008). As mentioned above, PPARγ also interacts with the Nrf2/ARE axis to mute the toxic sequelae of oxidative stress (Ikeda et al., 2000; Shih et al., 2005; Cho et al., 2010; Polvani et al., 2012; Zhao and Aronowski, 2014). Collectively, these findings reveal the molecular mechanisms underlying powerful anti-oxidative actions of PPARγ. As oxidative stress contributes to almost all CNS pathologies and impairs tissue repair, the upregulation of multiple antioxidant pathways by PPARγ is expected to accelerate the recovery process.

Another mechanism underlying PPARγ-afforded neuroprotection may be through the energetic and neurotrophic support provided by glial cells. Production of nerve growth factor (NGF) and brain-derived neurotrophic factor (BDNF) in CNS trauma models is enhanced after administration of PPARγ agonists, although the cellular source of the neurotrophins has not been identified (Meng et al., 2011; Mandrekar-Colucci et al., 2013). Astrocytes have long been known to provide nutritional and metabolic support to neurons (Wang et al., 2016), and activated astrocytes produce BDNF following injury (Zhao et al., 2004). Consistent with its established role in bioenergetics, PPARγ modulates glucose metabolism in astrocytes. For example, PPARγ agonists increase glucose uptake and glycolysis in primary astrocytes (Dello Russo et al., 2003; Izawa et al., 2009). The subsequent increase in cerebral glucose utilization and metabolism might attenuate neurodegeneration (Garcia-Bueno et al., 2007). These effects might be especially important in diseases related to energy deprivation—such as ischemic stroke—and may broaden the temporal window during which drugs might be administered, in addition to increasing cell survival. Microglia also secrete trophic factors (e.g. insulin-like growth factor (IGF)-1 and BDNF) to support neuronal survival (Choi et al., 2008; Trang et al., 2011; Wang et al., 2017). However, it is not known if PPARγ activation increases trophic factor expression in microglia.

4.3 Demyelination

4.3.1 Mechanisms of demyelination following CNS injury

White matter injury is characterized by demyelination and axonal degeneration. Demyelination is defined as loss of the axonal myelin sheath that is formed by the processes of oligodendrocytes. Myelination is essential for efficient signal transduction across different brain regions, and serves to protect the ensheathed axon by providing metabolic support and acting as a physical barrier against biochemical insults (Irvine and Blakemore, 2008; Funfschilling et al., 2012; Lee et al., 2012). Axonal degeneration can ensue as a consequence of demyelination, and these secondary effects are largely responsible for the debilitating effects of white matter injury. Thus, the consequences of demyelination include sensorimotor dysfunction and profound neurobehavioral and cognitive impairments (Desmond, 2002).

Oligodendrocytes in white matter are highly vulnerable to ischemic injury for a number of reasons. First, white matter has much lower blood flow and lack of collateral circulation in comparison with gray matter (Rosenberg, 2016). Second, cells of the oligodendrocyte lineage, such as oligodendrocyte precursor cells (OPCs), are particularly vulnerable to ischemia-induced oxidative stress and inflammation (Husain and Juurlink, 1995). In the rat MCAO model, 30 min of vascular occlusion leads to conspicuous swelling of oligodendrocytes and 3 h of occlusion results in lethal injury in large numbers of oligodendrocytes (Pantoni et al., 1996). Demyelination persists for at least 4 weeks after 60 min of transient MCAO in mice, as evidenced by decreased expression of myelin basic protein (MBP) in white matter (Jiang et al., 2016). Consistent with these in vivo observations, in vitro studies show that 30 min of OGD results in the death of 90% of oligodendrocytes within 9h after reperfusion (Tekkok and Goldberg, 2001). The pattern of demyelination in other CNS injuries, such as intracerebral hemorrhage, SCI, and TBI is similar to that observed in ischemic stroke. In other words, demyelination commences within hours to days after injuries and is sustained for a prolonged period, often up to months or even years (Wasserman and Schlichter, 2008).

The molecular mechanisms underlying CNS injury-induced demyelination and oligodendrocyte death can be summarized as follows. First, oligodendrocytes express all three types of glutamate receptors—AMPA, kainate, and NMDA (Salter and Fern, 2005; Butt, 2006), and are therefore subject to excitotoxicity. As in neurons, excess extracellular glutamate leads to over-activation of glutamate receptors on oligodendrocytes and elicits the accumulation of cytosolic calcium, which triggers oligodendrocyte toxicity (Benarroch, 2009). Second, oxidative stress is another major contributor to demyelination. Oligodendrocytes have high energy demands, abundant iron stores, and low levels of molecular antioxidants, rendering them particularly sensitive to increased ROS (Juurlink, 1997; Oyinbo, 2011; Shereen et al., 2011; Yang et al., 2017). Third, inflammatory reactions also contribute to demyelination after ischemia. Activation of resident microglia and astrocytes, as well as the infiltration of peripheral immunocytes lead to excess production of proinflammatory cytokines, key mediators of pathology in demyelination disorders. For example, high levels of IFN-γ lead to oligodendrocyte apoptosis and demyelination (Lin et al., 2005). Furthermore, TNF-α inhibits the proliferation and differentiation of OPCs, and induces oligodendrocyte apoptosis (Hovelmeyer et al., 2005; Pang et al., 2005; Shi et al., 2015).

Interactions between oligodendrocyte lineage cells and other white matter components such as microglia, astrocytes, and axonal fibers also play a critical role in injury-induced demyelination. Activated microglia can exhibit pro-inflammatory and anti-inflammatory properties, respectively (Hu et al., 2015) and may influence oligodendrocyte survival (Wang et al., 2013). Similarly, activated astrocytes also exhibit dual roles. Glutamate is taken up by astrocytes and converted into lactate, which then serves as an energy substrate for axons and oligodendrocytes following intracellular entry through monocarboxylate transporters and connexins. Astrocytes thereby transiently provide energy for oligodendrocytes and axons under stressful conditions (Matute et al., 2007). However, activated astrocytes may exacerbate oligodendrocyte cell death by increasing intracellular uptake of zinc and stimulation of Ca²⁺-permeable AMPA receptors (Johnstone et al., 2013).

Traumatic axonal injury (TAI) is localized to white matter and is initiated by mechanical force, such as impact to the head and rapid acceleration-deceleration of the head or the spinal cord (Armstrong et al., 2016b). The proliferation, migration, survival, and differentiation of oligodendrocytes depend upon axon-derived signals (Shi et al., 2015), such that loss or degeneration of axons can directly cause myelin pathologies. TAI causes axonal degeneration, which is accompanied by mitochondrial swelling, disruption of vesicular transport, cytoskeleton breakdown, and the formation of end bulbs at the sites of axotomy or disconnection. An important feature of TAI is that the injury is dispersed among the intact axons within white matter tracts; this expands the lesion and increases disease severity (Gennarelli, 1993). Mild TBI with low TAI density does not impair oligodendrocyte survival, as a single oligodendrocyte can maintain myelin sheaths around a cohort of axons, and damaged axons are a minor proportion of the total number of axons present. Nevertheless, damage to axons can dysregulate myelin maintenance signals to the oligodendrocyte, resulting in aberrant myelin synthesis and subsequent formation of redundant myelin, as evident in double-layered myelin sheaths around damaged axons (Rosenbluth, 1966; Armstrong et al., 2016a). Mild TBI may be accompanied by some degree of demyelination, as there also co-exist other pathogenic factors related to demyelination, including glutamate excitotoxicity, excessive ROS, neuroinflammation, and microglial activation (Clarner et al., 2012; Sullivan et al., 2013; Mierzwa et al., 2015). Consistent with these observations, in a closed-skull impact mouse model mimicking mild TBI, demyelination of intact axons is observed as early as 3 days, and lasts up to 6 weeks post-injury, in the absence of frank oligodendrocyte death (Mierzwa et al., 2015). In that model, chronic inflammation contributes to the long-lasting demyelination. Moderate or severe TBI with TAI is often accompanied by microhemorrhages or severe neuroinflammation (Armstrong et al., 2016a), resulting in focal lesions, and a higher proportion of axons in the lesion site are damaged, leaving the myelinating oligodendrocytes without any viable axons to wrap (Armstrong et al., 2016b). This lack of nearby axons causes myelin degradation and oligodendrocyte death. It is worth noting that TBI and SCI also cause mechanical damage in gray matter, leading directly to neuronal death (Bayly et al., 2005). Thus, axons can become disconnected from damaged neurons and undergo orthograde degeneration, ultimately leading to myelin degradation and oligodendrocyte death, as outlined above. For these and other reasons, death of oligodendrocytes in white matter is evident within 12 h in a rat model of moderate fluid percussion injury, and this increases progressively through the following week and is maintained for 2 months post-injury (Conti et al., 1998), perhaps mediated by the caspase-3 apoptotic pathway (Flygt et al., 2013).

4.3.2 PPARγ protects against demyelination following CNS injury

Most studies on PPARγ in CNS injury focus on gray matter. In a recent study, however, Han et al. investigated the effect of PGZ on white matter integrity in a 60 min transient MCAO mouse model, and reported less loss of white matter integrity at 21d post injury, partly due to increased microglial M2 polarization (Han et al., 2015). There are also protective effects of PPARγ on autoimmunity-induced demyelination, consistent with its beneficial role in stroke, TBI, and SCI. In rodent experimental autoimmune encephalomyelitis (EAE) models, TGZ decreases pathology by attenuating pro-inflammatory cytokine production (Niino et al., 2001). Similarly, PGZ reduces injury symptoms in EAE models, and this is accompanied with decreased lymphocyte infiltration, demyelination, and chemokine and cytokine expression, and increased expression of the NF-κB inhibitor, IκB, in the brain (Feinstein et al., 2002).

In vitro studies have demonstrated potential mechanisms underlying the protective effects of PPARγ against demyelination. For example, endogenous prostaglandin derivatives are natural PPARγ activators and inhibit production of nitrite and pro-inflammatory cytokines, including IL-6, IL-1β, TNF-α, and MCP-1 from LPS-stimulated microglia and astrocytes (Storer et al., 2005). In a study exploring T-cell mediated demyelination in three-dimensional aggregating rat brain cell cultures, Duvanel et al. demonstrated that PGZ rescued anti-myelin oligodendrocyte glycoprotein autoantibody-induced demyelination by reducing heat shock responses and downregulating TNF-α (Duvanel et al., 2003). Apart from its anti-inflammatory properties, PPARγ also mitigates oxidative stress in oligodendrocytes. Treatment of rat oligodendrocytes with PGZ or 15d-PGJ2 enhances their antioxidant defenses by increasing levels of catalase and Cu/Zn SOD, while maintaining the equilibrium of the glutathione redox system (Bernardo et al., 2009). Although there is little direct evidence regarding the influence of PPARγ on excitotoxicity-induced demyelination, PPARγ is known to decrease excitotoxicity in general, as discussed in section 4.2.2. However, whether PPARγ can help relieve demyelination caused by TAI-induced oligodendrocyte death is not yet known.

Collectively, the available evidence supports the therapeutic potential of PPARγ in CNS injury- induced demyelination, although evidence regarding direct effects of PPARγ on demyelination is still limited, especially in the context of acute CNS injuries. Furthermore, the underlying mechanisms are poorly understood and warrant further studies.

4.4 BBB disruption

4.4.1 Mechanisms of blood-brain barrier disruption in CNS injury

As a physical, transport, and metabolic barrier, the blood-brain barrier (BBB) controls the exchange of various molecules and nutrients between the blood and brain compartments, while keeping bacteria, viruses, and other xenobiotics outside the brain (Weiss et al., 2009). The major structural bases of the BBB consist of brain capillary endothelial cells, the tight junctions that connect them, astrocyte end feet, and pericytes (Weiss et al., 2009; Thal and Neuhaus, 2014; Attwell et al., 2016). Rapid alterations of endothelial permeability occur in many neurological conditions, including ischemic stroke and TBI (Salehi et al., 2017). The effects of CNS injuries may occur either directly on the endothelium or indirectly through other cells types in the neurovascular unit that affect the endothelium (Thal and Neuhaus, 2014). Reactive oxidants, inflammatory cytokines, and MMPs contribute to BBB permeability and cause the degradation of endothelial junctions, loss of vascular integrity, and BBB dysfunction after CNS injuries. Within 30-60 min following ischemic stroke, aberrant actin polymerization, stress fiber formation, and destruction of junctional proteins in brain endothelial cells result in an early breach of the BBB in response to activation of Rho-associated protein kinase/myosin light chain signaling and loss of inhibition by actin depolymerizing factor. This is followed by the infiltration of MMP-secreting neutrophils and macrophages into the brain, setting in motion a cascade of events that culminates in tissue damage and loss of neurological function (Shi et al., 2016a; Shi et al., 2016b). Increased BBB permeability may also facilitate edema formation and hemorrhagic transformation, thereby aggravating neurological outcomes (Dirnagl et al., 1999). All of these changes lead to primary brain damage and post-ischemic secondary injury (Hacke et al., 1996; Yin et al., 2010).

4.4.2 PPARγ protects against blood-brain barrier disruption following CNS injury

The BBB is partly composed of cerebral vascular endothelial cells and the tight junctions holding them together. A growing number of studies reveal that PPARγ is expressed in the cerebral vascular endothelium and plays important roles in the regulation of cerebrovascular structure and function (Culman et al., 2007; Beyer et al., 2008; Halabi et al., 2008; Hamblin et al., 2009). For example, PPARγ activation by TZDs can regulate endothelial migration and angiogenesis (Duan et al., 2008), consistent with a role for PPARγ in BBB repair. PPARγ promotes human umbilical vein cell migration by inducing the expression of a member of the class-3 semaphorins, Sema3g (Ming et al., 2015). Activated endothelial cells trigger immune responses through the secretion of pro-inflammatory cytokines/chemokines, such as IL-1 and TNF-α (Ross, 1993, 1999; Tesfamariam and DeFelice, 2007). Many studies have demonstrated that PPARγ agonists reduce the activation and inflammation of endothelial cells by inhibiting proinflammatory cytokines/chemokines (Jackson et al., 1999; Marx et al., 2000; Pasceri et al., 2000; Imamoto et al., 2004; Verrier et al., 2004; Sasaki et al., 2005; Chen et al., 2015). PPARγ in endothelial cells protects against IL-1β-induced endothelial dysfunction (Mukohda et al., 2016). Furthermore, Yin and colleagues found that the PPARγ agonist PGZ significantly reduces OGD-induced endothelial cell death (Yin et al., 2013). In the same study, the effects of adenovirus-mediated gain or loss of PPARγ function strongly support a cytoprotective role for PPARγ in endothelial cultures. Consistent with these in vitro findings, PPARγ agonists also significantly reduce ischemia-triggered increases in cerebrovascular/BBB permeability in mice (Yin et al., 2013). Hind and colleagues established an in vitro BBB model using human brain microvascular endothelial cells and human astrocyte co-cultures (Hind et al., 2015). They discovered that the PPARγ agonist cannabidiol decreased BBB permeability when administered either before or after OGD. Furthermore, vascular cell adhesion molecule-1 expression was decreased. Min et al. reported that activation of PPARγ by telmisartan attenuates BBB impairment in a type 2 diabetic mouse model (Min et al., 2012). In that study, significant decreases in the expression of tight junction proteins (ZO-1, occludin, claudin3, and claudin5) and increases in MMP-2 and MMP-9, oxidative stress, and pro-inflammatory cytokines (MCP-1, IL-6, TNF-α) were observed in the mouse brain after treatment with the PPARα antagonist (Min et al., 2012). Collectively, these studies indicate that PPARγ activation regulates BBB permeability and this may partly underlie its protective role against CNS injuries.

5. PPARγ promotes CNS repair

The chronic phase of CNS injury can progress for weeks to years, even after the CNS commences repair to restore homeostasis. Thus, the injury and reparatory phases are not completely separable and exhibit considerable overlap. During the chronic phase, neural inflammation is eventually controlled and resolved. Debris in and around the lesion zone is isolated and removed by local microglia and/or by infiltrating peripheral immune cells. Neurons may be replenished in limited numbers through neurogenesis. White matter is repaired by neurite outgrowth and remyelination. Angiogenesis and revascularization reseal the compromised BBB. In the next section, we will discuss repair processes, mainly in the chronic phase, and describe the involvement of PPARγ. As discussed in Section 2.4, the existence of PPREs on gene clusters that control cell survival, differentiation, chromatin remodeling, and the DNA damage response support an important role of PPARγ in tissue repair (Lemay and Hwang, 2006; Heinaniemi et al., 2007).

5.1 Resolution of inflammation

5.1.1 An introduction on the resolution of inflammation following CNS injury

Resolution of inflammation is initiated immediately after the onset of CNS challenges, at the same time as the inflammation commences. Clearance of debris is the first step toward the resolution of inflammation, and will be discussed in Section 5.2. Contraction of the immune cell population in the CNS is also essential to restore homeostasis. Neurotrophic factors from the immune system help regenerate injured tissue and prevent cell loss. Anti-inflammatory cells and factors work in concert to complete the process of inflammation resolution.

Both innate and adaptive immune systems have components that regulate inflammation. Immediately after CNS injury, the pro-resolution M2 phenotype predominates in the microglial and macrophage populations for the first 3-5 days, and this is followed by M1 polarization (Hu et al., 2012; Wang et al., 2013). However, within weeks following CNS injury, the pro-inflammatory M1 microglial phenotype gradually subsides and is replaced by the pro-resolution M2 phenotype. The M2 phenotype promotes 1) phagocytosis to clear debris, 2) secretion of anti-inflammatory cytokines, such as IL-10 and transforming growth factor (TGF) β, and 3) release of a plethora of trophic factors, including IGF-1. Concomitantly, infiltrating neutrophils and macrophages undergo apoptosis to help resolve inflammation. The N2 neutrophil population expresses pro-resolution markers, such as CD206 and Ym1, and this is associated with increased neutrophil clearance in the MCAO mouse model (Cuartero et al., 2013). In the adaptive immune system, the best-studied regulatory lymphocytes are known as CD4+CD25+FoxP3+ regulatory T cells (Tregs), which facilitate the resolution of inflammation after CNS injury (Shevach et al., 2006; Zhou et al., 2017). Treg cells can directly suppress effector T cells and myeloid cells by secreting suppressor cytokines such as IL-10, IL-35, and TGF-β, and compete for IL-2, which induces T cell apoptosis due to deprivation of activating cytokines and granzyme-dependent cytotoxicity. Furthermore, by suppressing the functions of antigen presenting cells, Treg cells indirectly modulate the responder T cells. Tregs prevent activation, maturation, and access to effector cells, impeding the immune stimulatory function of antigen presenting cells (Shevach, 2009). In addition, Tregs can express neurotrophic factors such as BDNF and the levels of circulating BDNF+ Tregs are correlated with positive disease outcomes in ischemic stroke patients (Feddersen and Van Ness, 1989).

5.1.2 PPARγ promotes resolution of inflammation

Although beneficial effects of PPARγ agonists have been widely reported in ischemic stroke, ICH, TBI, and SCI, the mechanisms whereby PPARγ promotes resolution of inflammation in acute CNS diseases are poorly understood. Our previous study reported that application of RGZ in a transient MCAO mouse model increased the expression of CD206 on microglia, a marker of the inflammation-resolving M2 phenotype (Han et al., 2015). Furthermore, treatment with RGZ enhances the phagocytic activity of microglia in the mouse distal permanent MCAO model (Ballesteros et al., 2014). Zhao and colleagues reported that RGZ administration promotes hematoma resolution by enhancing microglial phagocytosis after intrastriatal blood injection in both mice and rats (Zhao et al., 2007). Administration of RGZ also promotes infiltration of neutrophils into the ischemic lesion zone, followed by direct neutrophil polarization to the N2 phenotype and microglia/macrophage clearance of neutrophils in the MCAO mouse model (Cuartero et al., 2013). RGZ treatment in mice subjected to permanent MCAO accelerates neutrophil recruitment into the ischemic core, as shown by increased numbers of neutrophils in the ipsilesional hemisphere at 24h after ischemia. RGZ also increased the expression of Ym1 and CD206 by two-fold in these infiltrated neutrophils. Furthermore, RGZ treatment facilitated neutrophil clearance in the infarct core by microglia/macrophages. Taken together, these findings indicate that PPARγ activation might enhance the beneficial functions of microglia and neutrophils. Furthermore, PPARγ agonists prevent excessive activation of neutrophils by promoting their clearance (Cuartero et al., 2013). These collective actions of PPARγ are expected to accelerate tissue repair and functional recovery.

In contrast to the effects of PPARγ activation on microglia and neutrophils, the effects of PPARγ on Tregs in acute CNS diseases are poorly understood. PPARγ is known to be essential for the development of Tregs in highly enriched mouse adipose tissue (Cipolletta et al., 2012; Park and Pan, 2015). Administration of CGZ enhances Treg development and its suppression of Th17 cells in a graft-versus-host disease mouse model (Wohlfert et al., 2007; Park and Pan, 2015). Although the cellular source has not been established, intraperitoneal administration of RGZ after SCI increases the levels of NGF and BDNF (Meng et al., 2011; Mandrekar-Colucci et al., 2013). Furthermore, Tregs can be a source of BDNF and other trophic factors, as discussed in Section 5.1.1.

In conclusion, resolution of inflammation is not an isolated process; it is fully integrated into the chronic recovery phase. Timely clearance of necrotic debris or injured cells prevents oxidative stress and the release of pro-inflammatory factors, impeding the feed-forward cycles that characterize inflammation. Contraction of the immune cell populations after fulfillment of their beneficial functions decreases the risk of excessive immune reactions and the resulting deleterious auto-immune responses. Neurotrophic factors are also essential for the reconstruction of tissue in acute CNS diseases. Direct and indirect evidence outlined above suggests that activation of PPARγ fosters the resolution of inflammation and the recovery of function through some of these mechanisms.

5.2 Clearance of debris and infiltrating cells

5.2.1 An introduction on the clearance of debris and infiltrating cells

Severe CNS injuries cause cell death and the subsequent production of cellular debris. Necrotic or apoptotic neuron somata and axonal debris or detached myelin pieces are detrimental for several reasons. First, they activate immune responses and lead to subsequent chronic inflammation, which causes further cell death and tissue damage. Second, the debris-induced chronic inflammation impairs axonal regeneration and maturation of OPCs, which are vital for post-injury remyelination. Furthermore, infiltrating polymorphonuclear neutrophils (PMNs) and erythrocytes may also be detrimental after hemorrhage and trauma. Application of PMNs onto post-OGD organotypic hippocampal slice cultures causes a remarkable exacerbation of neuronal damage (Neumann et al., 2008), perhaps due to the release of ROS, proteases, and proinflammatory cytokines such as TNF-α (Jordan et al., 1999). In ICH, extravasated erythrocytes in the hematoma undergo lysis within hours to days after ICH, releasing cytotoxic hemoglobin, heme, and iron, thereby negatively influencing the viability of surrounding neurons and oligodendrocytes (Zhao et al., 2009a). Thus, efficient clearance of damaged cellular components by phagocytosis relieves secondary tissue damage and favors tissue repair after CNS injury.

Phagocytosis after CNS injuries is mainly carried out by resident microglia and infiltrated hematogenous macrophages. Using green fluorescent protein (GFP) transgenic bone marrow chimeric mice to distinguish microglia from peripheral macrophages, Schilling and colleagues reported that activated microglia exhibited phagocytic responses beginning on the first day after 30 min of transient MCAO (Schilling et al., 2005). Furthermore, the number of activated microglia increased gradually and peaked on day 10. Peripheral macrophage infiltration was not evident until day 4 after MCAO, peaking in the following 3 days, and then decreasing until 2 weeks after the ischemic insult. Phagocytic microglia accounted for one quarter of total activated microglia and reached maximal levels as early as 24 h post-ischemia and were maintained at this level until the infiltration of hematogenous macrophages. These findings lead to the hypothesis that microglia play a predominant role in debris clearance in the early stages, while assistance from infiltrated macrophages occurs subsequently. Similarly, in a moderate contusion injury model of SCI, Greenhalgh et al. used lysozyme M enhanced GFP-knockin mice to tag hematogenous macrophages and discovered that 1) phagocytic microglia appeared as early as 24 h post-SCI, 2) microglia were the predominant phagocytic cell type until day 3, and 3) infiltrating macrophages become the predominant phagocytic cell on day 3 and persisted for up to 42 days (Greenhalgh and David, 2014). It is noteworthy that only a small percentage of microglia contained phagocytic material at 7d (~3%), while phagocytic material was still evident in a high percentage of infiltrating macrophages at 42 d (~30%) after SCI. Phagocytosis of red blood cells by microglia has also been observed in ICH and promotes hematoma resolution (Zhao et al., 2007). In a closed skull injury-induced mouse TBI model, activated microglia were immunoreactive for galactose-specific lectin (Galectin)-3 (formerly called Mac-2), a receptor involved in the activation of myelin phagocytosis (Rotshenker, 2009). The microglia were most abundant one day following injury, followed by a decrease up to 28 days after TBI, but still were significantly elevated in TBI brains compared to sham control brains (Venkatesan et al., 2010).

The benefits of phagocytosis after acute CNS injuries have been well documented. As mentioned above, external application of invading PMNs to OGD-treated organotypic brain slices dramatically enhances ischemic neurotoxicity (Neumann et al., 2008). However, additional application of microglia counteracted this effect by rapid engulfment of PMNs, followed by neuroprotection that could be abrogated by interfering with microglial PMN engulfment (Neumann et al., 2008). Phagocytosis of blood components by resident microglia and macrophages in ICH mitigates further damage (Zhao et al., 2009a). Accordingly, phagocytosis and subsequent hematoma resolution are associated with improvements in neurological recovery (Chen et al., 2017), whereas loss of phagocytosis exacerbates brain swelling and impedes clot resolution (Ni et al., 2016). Enhanced microglial phagocytosis of myelin and cell debris is also beneficial for the survival of injured neurons and myelin regeneration in acute SCI and TBI, thereby facilitating functional recovery (Rotshenker, 2003; Boekhoff et al., 2012; Redondo-Castro et al., 2013). Molecular mechanisms involved in microglia/macrophage-afforded phagocytosis have been widely studied. In this context, priming is defined as any condition or event that facilitates the interaction between phagocytic receptors and their targets. Priming allows the phagocyte to engulf its targets by exposing ligand binding sites. This exposure endows phagocytic receptors with increased mobility in the plane of the membrane, thereby enhancing the probability of encountering target particles (Freeman and Grinstein, 2014). A large number of inflammatory stimuli promote the transition of the integrins from a bent (closed) conformation to an extended (open) conformation, resulting in enhanced exposure of binding sites (Shattil et al., 2010). Other priming factors include G protein-coupled receptor ligands (Freeman and Grinstein, 2014), Toll-like receptor (TLR) ligands (Patel and Harrison, 2008), and growth factors such as macrophage colony-stimulating factor (Kheir et al., 2005). After priming, specific membrane receptors on phagocytes and downstream signaling pathways contribute to the recognition and engulfment of target microparticles. In general, there are two functionally distinct types of receptors—one with a high binding affinity to foreign microbial pathogens, such as TLRs, and one that recognizes apoptotic cellular substances, such as triggering receptor expressed on myeloid cells 2. During post-injury debris clearance, the latter receptor is more instrumental. Other receptors involved in phagocytosis include Fc receptors (Anderson et al., 1990), complement receptors (Ross et al., 1992), scavenger receptors (Patel et al., 2004), pyrimidinergic receptor P2Y (Koizumi et al., 2007), G-protein coupled receptors (Riyahi et al., 2011; Dustin, 2016) macrophage antigen complex 2 (Rotshenker, 2009), and mannose receptors (Ezekowitz et al., 1990). Furthermore, different surface receptors on phagocytes may be involved in tethering (recognition and binding of cell corpses) and tickling (internalization and activation of downstream signaling) processes. Phosphatidylserine is a dominant “eat me” signal when cells undergo apoptosis (Fadok et al., 1998). The recognition of phosphatidylserine is mediated by at least one tethering and one internalization receptor (Henson and Hume, 2006). For example, integrins and CD36 are involved in tethering, while Gas and protein S are involved in tickling (Majai et al., 2007). Tethering and tickling lead to activation of Rac pathway, which mediates the uptake of dead cells (Majai et al., 2007).

In the context of acute CNS injuries, CD36 and Galectin-3 are the most important and widely studied phagocytosis receptors. CD36, a member of the scavenger receptors, is involved in the recognition and removal of modified lipoproteins, apoptotic cells, and pathogens (Canton et al., 2013). Upregulation of CD36 during the acute (3d) and recovery (7d) periods has been documented in mouse models of ischemia and is correlated with higher phagocytic indices in brain immune cells (Woo et al., 2016). Furthermore, CD36 deficiency compromises microglial phagocytosis of RBCs and slows hematoma absorption (Fang et al., 2014). The importance of CD36 is confirmed by the observation that addition of anti-CD36 antibodies to human macrophages reduces phagocytosis of apoptotic neutrophils by 50% (Fadok et al., 1998). In addition, CD36 is involved both in tethering and ligand internalization processes. Specifically, CD36 functions as an integrin-associated protein and cooperates with integrins to accomplish tethering (Majai et al., 2007). During ligand internalization, CD36 forms a heteromeric complex with β1 and/or β2 integrins and the tetraspanins CD9 and/or CD81, thereby linking to the adaptor Fc receptor γ, which bears an immunoreceptor tyrosine activation motif. By coupling to Fc receptor γ, CD36 is able to engage the Src-family kinases and Syk, which in turn drives the internalization of CD36 and its bound ligands (Majai et al., 2007). Finally, Galectin-3 may activate phagocytosis by upregulating and prolonging KRas-GTP-dependent phosphatidylinositol 3-kinase activity (Rotshenker, 2009).

During phagocytosis, high levels of pro-oxidative molecules and pro-inflammatory mediators are produced and may adversely affect the survival of phagocytes (Zhao et al., 2007; Yi et al., 2008). Thus, resolution of inflammation and oxidative burden facilitate phagocytosis by preventing injury to microglia/macrophages. For example, upregulation of the antioxidant enzyme catalase is known to promote phagocytosis (Donnan et al., 2008).

5.2.2 PPARγ promotes clearance of debris and infiltrating cells

The functional role of PPARγ in microglia/macrophage-afforded phagocytosis has been widely studied. PPARγ is dramatically upregulated during the maturation of human macrophages from monocytes, a process that endows macrophages with more efficient phagocytic capacities (Majai et al., 2007). PPARγ agonists can upregulate microglial phagocytosis of Staphylococcus aureus in bacterial brain abscesses (Kielian et al., 2008). In the brains of Alzheimer’s patients, PPARγ activation is associated with elimination of Aβ by microglia and a reduction in the oxidative burden in damaged tissue (Yamanaka et al., 2012). Consistent with these findings, PPARγ deficiency in macrophages delays skin wound healing in mice by impairing apoptotic cell clearance (Chen et al., 2015). Furthermore, the application of PPARγ antagonists during the differentiation process impairs macrophage-mediated phagocytosis of apoptotic cells, which may be partly attributed to downregulation of CD36 (Majai et al., 2007).

PPARγ-afforded promotion of phagocytosis may be attributed to three factors. First, PPARγ may promote inflammation-resolving M2 polarization in microglia/macrophages. Although somewhat controversial, phagocytosis is more often considered an M2 characteristic (Tang and Le, 2016). Second, PPARγ favors the resolution of inflammation and may therefore reduce proinflammatory cytokines such as IFN-γ, IL-1β, and TNF-α, which could impair the phagocytic activity of microglia (Koenigsknecht-Talboo and Landreth, 2005). Third, PPARγ reduces oxidative stress, which also serves to promote phagocytosis (Donnan et al., 2008).

A number of studies support the view that PPAR-γ activation in microglia and macrophages promotes phagocytosis after CNS injuries. In models of ICH, the activation of PPARγ increases the clearance of blood deposited in primary microglia cultures and in animals, thereby improving hematoma resolution (Heppner et al., 2015). PPARγ activation is followed by upregulation of catalase and CD36 by phagocytes (Donnan et al., 2008) and by inhibition of NF-κB in ICH, all of which serve to boost phagocytosis. Similarly, PPARγ activation with RGZ induces CD36 expression in resident microglia in a mouse model of permanent focal cerebral ischemia, increasing the ability of microglia to engulf dead neutrophils (Ballesteros et al., 2014). The effects of PPARγ activation on phagocytosis in SCI or TBI are not known.

5.3 Neurogenesis

5.3.1 An introduction on neurogenesis in the adult brain after CNS injury

For many years, it was believed that the brain is extremely limited in its capacity for self-renewal (Streilein, 1995). However, we now know that the adult CNS undergoes active self-repair and neurogenesis after CNS injuries, a finding that holds immense translational potential. Two neurogenic niches containing neural stem/progenitor cells (NSCs/NPCs) have been identified in the adult brain—the subventricular zone (SVZ) of the lateral ventricles and the subgranular zone (SGZ) of the hippocampal dentate gyrus (Lindvall and Kokaia, 2015). Neurogenesis in these two niches follows different patterns. Putative NSCs within the SVZ are known as type B cells and are intermixed with ependymal cells. They are characterized by the astrocytic marker glial fibrillary acidic protein (GFAP) and astrocyte-like processes. Once quiescent type B cells are activated, they proliferate and begin to express NPC markers such as nestin. At this stage, they are called transit amplifying precursors or type C cells. After three proliferative cycles, type C cells transform into neuroblasts (type A cells) that migrate towards the olfactory bulb along the rostral migratory stream and differentiate into mature neurons while integrating into olfactory circuits (Doetsch et al., 1997; Quinones-Hinojosa et al., 2006). Neurogenesis in the SGZ originates in another type of NSCs (type I cells) that share properties with radial glial cells and express GFAP and nestin. When activated, type I cells transform into NPCs (type IIa cells) that further develop into type IIb cells and express markers of neuroblasts such as doublecortin. Type IIb cells then generate type III neuroblasts and develop dendrites. Notably, approximately 80% of type III cells die of apoptosis within 2 weeks (Arvidsson et al., 2002), and those that survive mature into NeuN-expressing neurons (Eriksson et al., 1998; Seri et al., 2001). These newly generated neurons form synaptic connections with pyramidal cells of the hippocampus, which play a critical role in cognitive recovery after CNS injury (Seri et al., 2001; Ramirez-Amaya et al., 2006; Clelland et al., 2009; Sahay et al., 2011).

During the past two decades, novel neurogenic niches other than in the SVZ and SGZ have been reported, including in the hypothalamus (Kokoeva et al., 2005; Xu et al., 2005; Migaud et al., 2010; Pierce and Xu, 2010), striatum, neocortical areas, amygdala (Bernier et al., 2002; Dayer et al., 2005; Luzzati et al., 2006), and brainstem (Bauer et al., 2005). Neurogenesis in the spinal cord after SCI is more elusive. Spinal cord ependymal cells may act as NSCs (Duan et al., 2016), but develop into oligodendrocytes and astrocytes and not into neurons, although their NSC potential is preserved in vitro (Johansson et al., 1999; Meletis et al., 2008; Barnabe-Heider et al., 2010). One explanation is that the spinal cord provides a non-permissive microenvironment that specifically blocks neuronal lineage differentiation, unlike the brain (Lee-Liu et al., 2013). However, these ideas await confirmation.

Stroke and TBI both trigger the proliferation, differentiation, and migration of neurons in the SVZ and SGZ (Arvidsson et al., 2002; Masuda et al., 2007; Shen et al., 2008), suggesting a link between CNS injury and neurogenesis. However, whether neurogenesis is vital for CNS recovery after injury is not clear. When neurogenesis is abolished by ablation of doublecortin-expressing neuroblasts in mice subjected to MCAO, infarct sizes and behavioral deficits are increased at 24 h post-injury (Jin et al., 2010a). Furthermore, these effects last up to 12 weeks after MCAO, suggesting a crucial role for neurogenesis in long-term recovery (Wang et al., 2012a). A similar phenomenon is also evident in models of TBI with selective ablation of nestin-expressing NPCs (Blaiss et al., 2011).

5.3.2 PPARγ in neurogenesis

Only a few studies have focused on the role of PPARγ in neurogenesis, and none of them involve CNS injuries. Nevertheless, PPARγ enables normal brain development and global knockout is embryonically lethal (Wada et al., 2006). This has led to the speculation that PPARγ is indispensable in adult neurogenesis. In this section, we will discuss PPARγ-dependent neurogenesis in physiological and disease conditions, and make some inferences about its neurogenic role in CNS injuries. A role for PPARγ in neurogenesis is consistent with genomic studies confirming the presence of PPREs in gene clusters that control cell survival and differentiation (Lemay and Hwang, 2006; Heinaniemi et al., 2007).

PPARγ is essential in normal stem cell proliferation and neural differentiation in vitro. Wada et al. reported a decrease in cell growth rate in cultured PPARγ^+/− and PPARγ^−/− NSCs (Wada et al., 2006). In PPARγ^+/+ NSCs, cell proliferation was enhanced by RGZ and reduced by GW9662. PPARγ agonists can also promote neural differentiation. For example, 15d-PGJ2 promotes the differentiation of embryonic midbrain cells into dopaminergic neurons (Park et al., 2004b). Similarly, PGZ promotes neural differentiation and neurite outgrowth in SH-SY5Y neuroblastoma cells (Park et al., 2004b). Taheri et al. reported an increase in PPARγ mRNA in both the NPC stage and neural cell stage during the differentiation of cultured human embryonic stem cells (hESCs) (Taheri et al., 2015). Notably, NPC markers are decreased by GW9662, but not increased by 15d-PGJ2.

The effects of PPARγ on neurogenesis have been examined in some disease models. During an exploration of hyperglycemia-induced decreases in neurogenesis, advanced glycan end products (AGEs) were found to inhibit NSC proliferation and differentiation, and this was associated with decreased PPARγ expression (Wang et al., 2009). Exogenous RGZ administration rescued AGE-induced repression of NSC proliferation by inhibition of caspase-dependent NSC apoptosis. However, RGZ failed to prevent AGE-mediated inhibition of NSC differentiation (Wang et al., 2009). Intraperitoneal delivery of LPS (5 mg/kg) decreases hippocampal neurogenesis by 50% and this is associated with memory impairments that are reversed by oral intake of RGZ, indicating a protective role of PPARγ against inflammation-induced suppression of neurogenesis (Ormerod et al., 2013). However, conflicting results were reported during congenital cytomegalovirus (CMV) infection of the CNS (Rolland et al., 2016). In that study, CMV infection was associated with increased PPARγ expression and activity, as well as increases in 9-hydroxyoctadecadienoic acid, a known PPARγ agonist. PPARγ activation with 15d-PGJ2 led to impaired neurogenesis in uninfected NSCs, and treatment of NSCs with 9-hydroxyoctadecadienoic acid decreased NSC neural differentiation. Consistent with these findings, treatment of CMV-infected NSCs with the PPARγ inhibitor T0070907 restored normal differentiation rather dramatically. These results suggest complex effects of PPARγ on neurogenesis, and possibly distinct roles in various diseases and experimental models. Although PPARγ is significantly increased by CNS injuries such as cerebrovascular events and trauma, its effects on neurogenic processes remain unknown and need further exploration.

5.4 Remyelination

5.4.1 An introduction on remyelination during CNS repair

Remyelination refers to the progressive restoration of the entire myelin sheath around demyelinated axons, a process that reinstates saltatory conduction (Smith et al., 1979) and resolves functional deficits (Jeffery and Blakemore, 1997; Liebetanz and Merkler, 2006). Remyelination is characterized by the formation of thinner myelin with shorter internodes compared to sheaths that are not injured (Blakemore, 1974). The g ratio refers to the circumference of the axon divided by the circumference of the myelin sheath, and is used to assess the degree of remyelination (Franklin and Ffrench-Constant, 2008). Generally, remyelination consists of three phases: OPC proliferation, OPC recruitment, and OPC maturation. There is consensus that OPCs serve as the major origin of newly generated oligodendrocytes post-demyelination (Dawson et al., 2003). First, retroviral and autoradiographic tracing studies indicate that remyelinating oligodendrocytes originate from dividing cells in normal adult white matter, which are likely to be adult OPCs, although this awaits confirmation (Gensert and Goldman, 1997; Carroll et al., 1998). Second, transplanted OPCs remyelinate demyelinated axons quite effectively (Groves et al., 1993). OPCs repopulate focal demyelination sites, where both oligodendrocytes and OPCs have died, prior to the appearance of new oligodendrocytes; these temporal and spatial patterns strongly suggest that OPCs are indeed the source of remyelinating cells (Sim et al., 2002). Fourth, cells transiently express OPC as well as oligodendrocyte markers at the onset of remyelination (Fancy et al., 2004). In response to injury, microglia, astrocytes, and other inflammatory cells become activated and produce mitogenic factors and pro-migratory factors, including platelet-derived growth factor and basic fibroblast growth factor, which switch OPCs from a quiescent state to a regenerative phenotype and recruit OPCs to the demyelination zones (Franklin and Ffrench-Constant, 2008). In the final phase of remyelination, the recruited OPCs differentiate into myelinating oligodendrocytes and ensheath the demyelinated axons.

The adult brain undergoes remyelination and white matter repair after CNS injuries, but only to a very limited extent. In the closed-skull impact mouse model of TBI, newly generated oligodendrocytes appear as early as 3 days post-injury (Mierzwa et al., 2015). Remyelination is evident within 1 week post-TBI and prolonged for at least 6 weeks. Similar patterns of post-injury remyelination have been reported in animal models of ischemic stroke (Mandai et al., 1997), SCI (Harrison et al., 1975; Tripathi and McTigue, 2007; Hesp et al., 2015), and ICH (Joseph et al., 2016). Remyelination failure can result from failure of OPC recruitment or failure of differentiation. Experimental and clinical data suggest that differentiation is the most vulnerable phase of remyelination (Franklin and Ffrench-Constant, 2008).

5.4.2 PPARγ in remyelination (703)

As mature oligodendrocytic processes form the myelin sheath, promoting NSC differentiation toward the mature oligodendrocytic lineage might directly increase myelination. PPARγ agonists can augment NSC differentiation into mature oligodendrocytes (Kanakasabai et al., 2012; Wan Ibrahim et al., 2013), although the underlying mechanisms remain unknown (Figure 5A).

Figure 5.

PPARγ promotes regeneration and maturation of cells of the oligodendrocyte lineage. (A) PPARγ promotes differentiation of neural stem cells (NSCs) into oligodendrocyte precursor cells (OPCs), and enhances further maturation of OPCs into mature, myelinating oligodendrocytes. (B) Signaling pathways involved in OPC maturation: (1) PPARγ enhances the formation of myelin components (e.g., lipid plasmalogen) by increasing ADAPS expression, which in turn promotes OPC process extension. (2) Rapid lipid metabolism results in oxidative stress and lipid peroxidation, but this is counteracted by PPARγ-mediated upregulation of antioxidants such as catalase and Cu/Zn SOD. (3) PPARγ enhances mitochondrial activation by increasing activity of mitochondrial respiratory chain complex IV and upregulating ATP-induced Ca²⁺ oscillations. Abbreviations: neural stem cell (NSC), oligodendrocyte precursor cell (OPC), alkyl-dihydroxyacetone phosphate synthase (ADAPS), superoxide dismutase (SOD), reactive oxygen species (ROS), respiratory chain complex I, II, III, and IV, adenosine 5′-triphosphate (ATP), inositol 1,4,5-trisphosphate (IP3), inositol 1,4,5-trisphosphate receptor (IP₃R).

The effects of PPARγ in OPC maturation are better studied than its effects on NSCs (Figure 5B). Recent findings demonstrate that PPARγ activation positively contributes to differentiation of multiple lines of OPCs. Studies on B12 glioma-derived cells and primary rat spinal cord OPCs show that PPARγ activation promotes the commitment of OPCs toward differentiation into myelin basic protein-expressing (MBP⁺) mature oligodendrocytes (Roth et al., 2003). B12 cells express all three PPAR isoforms but only respond to PPARγ agonists. Activation of PPARγ in B12 cells arrests cell proliferation and increases process extension (Roth et al., 2003). Furthermore, treatment with PGZ drives primary rat OPCs to differentiate and mature into oligodendrocytes (Bernardo et al., 2009; De Nuccio et al., 2011). Our recent work demonstrates that RGZ enhances OPC proliferation and increases the number of newly-generated mature oligodendrocytes after MCAO (Han et al., 2015). Nonspecific PPARγ activators such as statins have also been studied for their effects on OPCs. Statins are 3-hydroxy-3-methylglutaryl coenzyme A reductase inhibitors and activate PPARγ through depletion of geranylgeranyl-pyrophosphate and subsequent RhoA-Rho kinase inactivation (Sim et al., 2008; Paintlia et al., 2010). Statins contribute to activation of PPARγ and directly promote survival and differentiation of rodent and human OPCs in vitro and increase myelin repair in the EAE model of multiple sclerosis in vivo (Paintlia et al., 2005; Miron et al., 2007; Paintlia et al., 2008).

As the ability of PPARγ to promote OPC maturation has been established, the mechanisms underlying this process are also being examined. First, PPARγ is heavily involved in cellular lipid metabolism, and lipid metabolism is of vital importance to oligodendrocyte process formation. PPARγ agonists can promote rat B12 cell maturation by enhancing the formation of myelin-rich lipid plasmalogen, as shown by an increase in alkyl-dihydroxyacetone phosphate synthase (Roth et al., 2003). Second, PPARγ activation blunts the oxidative stress arising from rapid lipid metabolism during OPC differentiation. As mentioned above, recent studies suggest that PPARγ activation upregulates expression of the antioxidant enzymes Cu/Zn SOD and catalase and maintains the equilibrium of the glutathione redox systems, which would mitigate the oxidative stress generated from OPC maturation-related metabolism (Bernardo et al., 2009). Third, PPARγ activation enhances the function of mitochondria during OPC maturation. Repeated administration of PPARγ agonists accelerates primary rat OPC maturation and expression of myelin basic protein by increasing activity of the mitochondrial respiratory chain complex IV and upregulating ADP-induced Ca²⁺ waves (De Nuccio et al., 2011). In sum, PPARγ appears to facilitate the survival and differentiation of OPC cells and enhance their antioxidant defenses and mitochondrial function (Bernardo et al., 2009).

Despite the abovementioned evidence that PPARγ favors remyelination after injury, negative effects of PPARγ agonists on OPCs have also been observed. For example, 15d-PGJ2 treatment can induce apoptosis in early-stage mouse OPCs (Xiang et al., 2007). Statins can also impede oligodendrocyte maturation and myelin formation in vitro (Miron et al., 2007; Klopfleisch et al., 2008). Statins may also hamper intrinsic myelin repair and maintenance in the cuprizone-induced animal model of demyelination (Klopfleisch et al., 2008; Miron et al., 2009). Furthermore, PPARβ, but not PPARγ, was shown to promote OPC differentiation in primary mixed glial cell cultures enriched in oligodendrocytes (Saluja et al., 2001; Jana et al., 2012). Activation of PPARβ also protects primary human oligodendrocytes against serum deprivation-induced cell death (Jana et al., 2012). However, crosstalk between PPARβ and PPARγ may enhance OPC differentiation in undifferentiated C6 glioma cells (Leisewitz et al., 2008). Therefore, further studies are needed to resolve these controversies and decipher the precise role of PPARγ in regulating developmental processes in cells of the oligodendrocyte lineage. Furthermore, as gene transcriptional activation is a subtle but dynamic process, there needs to be greater resolution of the temporal kinetics of gene expression after PPARγ activation. Other unresolved issues include identification of the stage of OPC differentiation during which agonist-mediated PPARγ activation might be most effective, as well as the pharmacokinetics and optimal dosages of PPARγ agonists.

5.5 BBB repair and angiogenesis/revascularization

Angiogenesis plays important roles in tissue repair and functional recovery after various CNS injuries. Extensive studies have shown that newly formed microvessels improve tissue microperfusion around the ischemia boundary zone and promote functional recovery after stroke (Zhang et al., 2000; Zhang and Chopp, 2002). PPARγ is highly expressed in vascular walls, including the endothelial lining, and is one of major transcription factors involved in the regulation of cerebrovascular structure and function (Culman et al., 2007; Beyer et al., 2008; Halabi et al., 2008; Hamblin et al., 2009). Beyer et al. examined vascular tone and function using heterozygous knock-in mice expressing dominant negative-PPARγ (Beyer et al., 2008). They reported that endothelium-dependent agonist-induced dilation in cerebral blood vessels was significantly impaired in dominant negetive-PPARγ mice. The authors also concluded that interruptions in PPARγ signaling result in endothelial dysfunction through oxidative stress and cause vascular hypertrophy and inward remodeling.

Chu et al. investigated the effects of the PPARγ agonist RGZ on angiogenesis and neurological recovery after cerebral ischemia in the rat MCAO model (Chu et al., 2006). The authors reported that RGZ-pretreated rats present not only with smaller infarct volumes but also with fewer initial neurologic deficits and superior recovery from day 1 through 5 weeks after cerebral ischemia. Notably, RGZ pretreatment also causes significant increases in BrdU⁺ proliferating endothelial cells in the lesioned hemispheres. Furthermore, RGZ increased cerebral microvessels density, vascular branch points, and vascular surface areas in the ipsilateral hemisphere. The authors conclude that eNOS upregulation may be the main mechanism of RGZ-promoted angiogenesis after focal cerebral ischemia. Another recent study evaluated the role of PPARγ in regulating human pulmonary microvascular endothelial cell migration and angiogenesis (Vattulainen-Collanus et al., 2016). In that study, loss of PPARγ attenuated angiogenesis and migration capacity of the cells through E2F transcription factor 1-mediated Wnt signaling. Using in vivo angiogenesis assays with subcutaneously placed matrigel plugs, Vattulainen-Collanus and colleagues also reported reductions in angiogenesis and mobilization of endothelial progenitor-like cells from the bone marrow in Tie2Cre-PPARγ^flox/flox mice with ablation of PPARγ in endothelial cells and osteoclasts (Vattulainen-Collanus et al., 2016).

Taken together, these findings suggest that PPARγ activation enhances angiogenesis and functional recovery in rodent ischemic stroke models. The genomic studies discussed in Section 2.4 are consistent with the findings presented here, as they reveal PPREs in genes involved in Wnt signaling, cell growth and maintenance, and BBB integrity (Lemay and Hwang, 2006; Heinaniemi et al., 2007).

6. PPARγ protects against comorbidities

Comorbidities are well known to exacerbate CNS injuries, either by precipitating their onset, or by directly accelerating the underlying pathologic process. Fortunately, many comorbidities, such as hyperglycemia, hypertension, atherosclerosis, and infection are clinically modifiable. Modification of these comorbidities confers protective effects against CNS injury and alleviates disease severity, and is therefore practiced as standard clinical management. Notably, PPARγ serves as a target of many of these comorbid diseases. In this section, we will discuss the effects of modifiable comorbidities commonly associated with CNS injuries, as well as clinical and preclinical evidence favoring their PPARγ-dependent modulation.

6.1 Hyperglycemia

Metabolic syndrome and diabetes mellitus (DM) are known risk factors for stroke. The prevalence of hyperglycemia is 20%–50% in acute stroke, even in the absence of DM (Yu et al., 2015). The second European Cooperative Acute Stroke Study (ECASS-II) followed 748 stroke patients and reported that blood glucose levels were associated with almost all outcomes, including 7-day neurological improvements, 30-day favorable functional outcomes, 90-day negligible dependence, all-cause mortality within 90 days, and hemorrhagic transformation within the first 7 days (Sugawara et al., 2001). Concurrent hyperglycemia is an independent predictor for stroke outcome (Ghajar, 2000; Sarafidis and Lasaridis, 2006). According to the American Stroke Association guidelines, a target of <300 mg/dL blood glucose should be met (Writing Group et al., 2016). Even non-cardiovascular, traumatic CNS injuries such as TBI and SCI are associated with hyperglycemia and severe TBI patients exhibit higher blood glucose levels compared with mild TBI patients (Elder et al., 2004; Jeremitsky et al., 2005). Hyperglycemia is positively correlated with mortality rate, and blood glucose levels are predictive of worse neurological outcomes (Jeremitsky et al., 2005; Uruno et al., 2011; Bevers et al., 2017). Similar detrimental effects of hyperglycemia in SCI have been reported in mouse models as well as in humans (Writing Group et al., 2016). Although impaired neurovascular coupling and reduced cerebral perfusion have been reported in DM patients (Duarte et al., 2015), it is not known with certitude if hyperglycemia is the cause of detrimental effects in the clinic, or if it is a stress response and/or biomarker of injury (Tsai et al., 2009).

TZD treatment is part of the standard of care for DM patients, as it significantly lowers blood glucose levels without elevating the risk of hypoglycemia (Nathan et al., 2009). Not surprisingly, stroke preventive effects of TZD have also been reported. For example, the “Insulin resistance intervention after stroke trial of PGZ” (IRIS trial) (Inzucchi and Furie, 2016) revealed significant secondary stroke prevention, and this effect was more pronounced for the lacunar subtype. It is worth noting that delivery of insulin (and not oral hypoglycemics) is the drug of choice in the acute phase of stroke (Nakayama et al., 1994). A metaanalysis suggests that administration of PPARγ agonists reduces the incidence of primary stroke and recurrent stroke (Liu and Wang, 2015). However, RGZ was reported to increase the risk of heart failure, according to the “Rosiglitazone evaluated for cardiovascular outcomes in oral agent combination therapy for type 2 diabetes” (RECORD) trial (Home et al., 2009), which led to dispensing restrictions by the U.S. Food and Drug Administration (FDA) in 2010. However, after further discussion at the joint meeting of the Endocrinologic and Metabolic Drugs Advisory Committee and the Drug Safety and Risk Management Advisory Committee in 2013 (FDA, 2013), the RECORD trial was deemed to suffer from bias in identification of cardiovascular events and the restrictions were lifted in 2015 (FDA, 2015). There have been no clinical trials on use of PPARγ agonists in TBI or SCI patients thus far.

6.2 Hypertension

Hypertension is the most frequent comorbidity associated with CNS injuries (Abraham et al., 2016; Qureshi and Qureshi, 2017). Hypertension is evident in up to 70% of stroke patients (Staessen et al., 2003). Hypertensive patients experience a two-fold increase in the lifetime risk of stroke compared to subjects with normal blood pressure (BP <120/80 mmHg) (Writing Group et al., 2016). PPARγ has been shown to downregulate BP. Aside from metabolic abnormalities, loss-of-function mutations of PPARγ in humans can lead to severe hypertension (Barroso et al., 1999; Agostini et al., 2006; Auclair et al., 2013). Consistent with these clinical findings, a two-fold increase in PPARγ expression levels can decrease BP by ~2.8 mmHg in rodents (Tsai et al., 2009). The angiotensin II pathway is known to result in hypertension and might be the target of PPARγ. TZDs have been shown to decrease the transcription of angiotensin II receptor type 1 in vascular smooth muscles in vivo (Sugawara et al., 2001), in agreement with in vitro findings that PGZ suppresses angiotensin II-induced aldosterone expression (Uruno et al., 2011). Accordingly, activation of PPARγ with PGZ can attenuate angiotensin II-induced hypertension (Yu et al., 2015). Other mechanisms underlying PPARγ-mediated modulation of BP may include maintaining calcium levels, inhibiting smooth muscle cell proliferation, stabilizing renal function, and attenuating sympathetic over-activation (Sarafidis and Lasaridis, 2006). These findings indicate that PPARγ activation might be a therapeutic approach to lowering BP to prevent the development of stroke. However, BP control during acute stroke is challenging, as both high BP and low BP are independent prognostic factors for poor stroke outcomes (Willmot et al., 2004). Furthermore, in TBI and SCI, intracranial pressure, rather than systemic BP, is one of the most important prognostic factors (Ghajar, 2000), and does not appear to be under direct regulation of systemic BP.

6.3 Dyslipidemia and atherosclerosis

Abnormality in lipid metabolism is an independent risk factor for stroke. Increases in triglycerides are associated with both ischemic stroke and ICH (Nakayama et al., 1994; Ott et al., 1999). In contrast, levels of high-density lipoprotein-cholesterol (HDL-C), a reverse cholesterol transporter, are negatively correlated with stroke risk, and this effect is even more pronounced in thromboembolic stroke (Curb et al., 2004; Huxley et al., 2011). In contrast, low-density lipoprotein (LDL) and peripheral macrophages may directly mediate atherosclerosis (AS) formation. Excessive oxidized LDL deposition induces accumulation of macrophages in the vascular wall. Macrophages engulf detrimental lipoproteins and convert into foam cells (Chistiakov et al., 2016). Activated foam cells within the sclerotic plaques secrete various cytokines, such as IL-6 and TNF-α, which lead to endothelial malfunction (Ait-Oufella et al., 2011), smooth muscle cell migration and proliferation, as well as extracellular matrix production, which may constrict the diameter of the vascular lumen (Wahl et al., 1997). In late stages of AS, excessively narrow arteries limit cerebral blood flow. In addition, macrophages produce extracellular proteases, such as MMPs, which can rupture plaques (Gough et al., 2006), thereby leading to thrombotic stroke (Mughal et al., 2011).

PPARγ expression is high in macrophages within AS plaques (Wilson, 2010). PPARγ may regulate dyslipidemia and AS by modulating macrophage behavior in a number of ways. First, PPARγ promotes clearance of lipid deposition. In apolipoprotein E-knockout mice fed a high-fat diet, TGZ attenuated fatty streak lesion formation and increased HDL-C levels, and this was associated with increased CD36, the scavenger receptor that mediates LDL phagocytosis (Chen et al., 2001). Second, PPARγ prevents excessive engulfment of LDL, thereby reducing macrophage activation. This view is supported by an elegant study showing that, although RGZ cannot inhibit foam-cell transformation induced by LDL in vitro, it can promote cholesterol efflux from macrophages via over-expression of the gene encoding ABCA1 (Chinetti et al., 2001). Third, PPARγ inhibits macrophage-mediated release of inflammatory cytokines in both rodents and humans (Li et al., 2000). PPARγ agonists RGZ and GW7845 strongly inhibit AS development in LDL receptor-deficient mice, and this is associated with decreased TNF-α (Li et al., 2000). Consistent with these findings, RGZ administration in non-diabetic patients was associated with decreased inflammatory markers and adipokines, although no changes in HDL-C and total cholesterol levels were observed (Samaha et al., 2006). Finally, PPARγ mediates growth arrest in vascular smooth muscle cells (Bruemmer et al., 2003a; Bruemmer et al., 2003b), which may serve as another mechanism whereby PPARγ inhibits AS, although this awaits confirmation.

6.4 Renal dysfunction

Renal dysfunction is a common comorbid disease associated with CNS injury and is a risk factor for both ischemic stroke and ICH. Renal dysfunction is apparent in up to one-third of stroke patients and significantly increases mortality (Rowat et al., 2014). In patients with ICH, renal dysfunction is associated with larger lobar hematomas and poorer outcomes (Molshatzki et al., 2011). In TBI, acute renal failure is one of the most common and serious complications, with incidence varying from 1.5% to 18.1% (Schirmer-Mikalsen et al., 2007; Bagshaw et al., 2008). In SCI, renal deterioration was the main cause of death until the introduction of methods that improve bladder emptying in the 1980s (Frankel et al., 1998). SCI-induced renal failure occurs mainly due to increased risk of vesicoureteral reflux and upper urinary tract dilation (McGuire et al., 1981). In a case-control long-term study conducted by Elmelund and colleagues, 26% of SCI patients eventually developed renal deterioration (Elmelund et al., 2017).

A number of studies demonstrate favorable kidney-protective effects of TZDs. Constitutive expression of PPARγ has been detected in human and rodent kidneys, predominantly in the medullary collecting duct (Guan et al., 1997; Yang et al., 1999). In various rodent models of type 2 diabetes, TZD treatment ameliorates diabetic nephropathy and improves renal function, aside from exerting glycemic control (Baylis et al., 2003; Okada et al., 2006). In accordance with preclinical studies, TZD treatment in diabetics significantly decreases urinary albumin excretion in patients with normoalbuminuria, microalbuminuria, or proteinuria (Sarafidis et al., 2010). Other than mitigating diabetic nephropathy, TZDs can also protect against renal dysfunction induced by cyclosporine (Chung et al., 2005), puromycin aminonucleoside (Zuo et al., 2012), aging (Yang et al., 2009a), and polycystic kidney disease (Yoshihara et al., 2011). Mechanisms underlying PPARγ-mediated renal protection include improved insulin resistance and glycemic control, cell-cycle-dependent actions, upregulation of adiponectin expression, anti-inflammatory effects, anti-oxidative effects, prevention of renal lipid accumulation, lowering of BP, etc. (Yang et al., 2009b). However, there are no reports showing that PPARγ protects against renal dysfunction in the setting of CNS injury.

6.5 Infection

Infection is an established risk factor for stroke. Exposure to minor infections in the previous month is an independent risk factor for childhood ischemic stroke, and is evident in one third of cases (Hills et al., 2012). Of these cases, 80% are respiratory infections (Hills et al., 2014). Chronic infections such as periodontitis, chronic bronchitis, and infection with Helicobacter pylori, Chlamydia pneumoniae, and CMV are associated with stroke, although the causal relationship remains unclear (Grau et al., 2010). Infection is also a common complication associated with ischemia, especially in the elderly, and is recognized as an independent predictor of poor outcomes (Wartenberg et al., 2011). Notably, stroke itself may induce immune depression, thereby increasing the incidence of systemic infection (Urra et al., 2009). In addition, infection is associated with excessive ROS generation, which may exacerbate neuronal loss and initiate a self-amplifying cascade of tissue destruction (Berg et al., 2011). Perhaps for this reason, the sepsis death cluster coincides with the so-called “Stroke Belt,” which refers to a region of increased stroke mortality across Mississippi, Alabama, Georgia, Tennessee, Kentucky, North Carolina, and South Carolina (Borhani, 1965; Howard et al., 1995; Wang et al., 2010). This overlap may also suggest shared, unidentified etiology of stroke and sepsis, but this remains to be confirmed.

The effects of PPARγ in sepsis have been examined. PPARγ expression is reduced in the lung in a mouse sepsis model induced by LPS intraperitoneal injection, and 15dPGJ2 improves survival 6-fold (9% to 55%) (Kaplan et al., 2005). These effects are associated with reduced NF-κB and increased HSP70. In another sepsis model induced by cecal ligation and puncture, curcumin treatment increased PPARγ expression in the liver, and reduced mortality and serum TNF-α levels, and these effects were reversed by the PPARγ inhibitor GW9662 (Siddiqui et al., 2006). As far as the underlying mechanism, the PPARγ agonists 15d-PGJ2, BRL 49659, and ciglitazone have been shown to desensitize macrophages in cultures, thereby suppressing the oxidative burst (Von Knethen and Brune, 2001). In addition, 15d-PGJ2 and TGZ can inhibit increased NO production and IκB degradation induced by LPS or heat-killed Escherichia coli and Staphylococcus aureus in rat peritoneal macrophages (Von Knethen and Brune, 2001). Aside from macrophages, T cells are also targeted by PPARγ, as GW9662 inhibits T cell expression of IFN-γ by interfering with c-Jun activation (Cunard et al., 2004).

Whether or not the findings outlined in this section can be translated into CNS injury remains unknown. Furthermore, there exists some controversy, in that PPARγ ligands can induce some proinflammatory responses. For example, macrophagic uptake of oxidized LDL is strongly associated with a proinflammatory phenotype polarization, and this may be enhanced by PPARγ ligands (Tontonoz et al., 1998) (Section 6.3). Second, sepsis refers to a severe infection that disturbs systemic homeostasis. How PPARγ affects mild to moderate infections remains unknown. Third, no reports have focused on the role of PPARγ against infections under conditions of CNS injury. Thus, there remain many gaps in the field.

7. Translational perspectives

A large body of preclinical and clinical work outlined above supports the view that PPARγ activation promotes functional recovery of the CNS by 1) reducing inflammation, 2) minimizing cell death in gray and white matter, and 3) engaging cell repair programs. A wide range of synthetic compounds functioning as PPAR ligands have recently been developed, with TZDs being the most prominent (Lalloyer and Staels, 2010). In addition, it seems likely that endogenous PPARγ activity is naturally increased after acute injuries to help restore homeostasis. Consistent with these views, a clinical study demonstrated that higher plasma concentrations of 15d-PGJ2, a natural PPARγ agonist, were negatively correlated with infarct size and positively correlated with better outcomes in patients with acute athero-thrombotic ischemic stroke (Blanco et al., 2005).

Although there are no clinical reports on the therapeutic effects of PPARγ agonists in the acute stages of cerebral ischemia, the potential effects of PPARγ agonists in stroke prevention and recurrent stroke risk reduction in high-risk populations have been analyzed. We have summarized clinical studies on the therapeutic efficacy and safety of two FDA-approved PPARγ agonists—PGZ and RGZ—in ischemic stroke, and found that most report beneficial outcomes (Table 1). The results of several large-scale clinical trials favor the administration of PPARγ agonists: 1) the prospective PGZ clinical trial in macro-vascular events (PROactive), 2) insulin resistance intervention after stroke trial (IRIS), 3) RGZ evaluated for cardiovascular outcomes in oral agent combination therapy for type 2 diabetes (RECORD), and 4) bypass angioplasty revascularization investigation in type 2 diabetes (BARI2D) (Table 1). Consistent with these observations, a meta-analysis also suggests that administration of PPARγ agonists reduces the incidence of primary and recurrent stroke (Liu and Wang, 2015). Compared with RGZ, PGZ seems to be more favorable, with fewer safety issues and superior therapeutic effects (Liu and Wang, 2015). On the other hand, some of the studies we reviewed failed to show significant effects of PPARγ agonists in stroke prevention (Table 1). Furthermore, RGZ was shown to increase the risk of cardiovascular events and ischemic stroke in some studies, leading to post-marketing monitoring of its safety (FDA, 2015). The adverse effects were mostly found among patients who were elderly or displayed a higher risk of developing macro-vascular diseases (Table 1). In addition, TZDs have been suggested to increase the risk of heart failure, particularly in type 2 diabetics (Michalik and Wahli, 2006). An exacerbation of heart failure following PPARγ activation has also been reported in animal studies (Lygate et al., 2003; Xu et al., 2003). Nevertheless, considering all the clinical studies collectively, it appears that PGZ or RGZ might reduce the incidence of stroke and recurrent stroke, although the target population may need to be carefully selected, as the elderly and those with higher risk of macrovascular diseases might not benefit from such treatment. This type of selective testing of therapies only on the appropriate patient population is gaining acceptance, and future clinical treatments will likely be individualized more carefully based on pharmacogenomic approaches and personalized medicine (Di Sanzo et al., 2017). In addition, BBB permeability and pharmacokinetic-pharmacodynamic properties will need to be carefully considered in order to improve the therapeutic efficacy of PPARγ agonists in CNS diseases. For example, women, diabetic patients, and subjects with lower body weight and higher serum creatinine were found to exhibit lower clearance of the TZD drug rivoglitazone (Rohatagi et al., 2008). For some PPARγ agonists, such as tesaglitazar, rapid absorption, high bioavailability and an elimination half-life of approximately 45 h were observed (Ericsson et al., 2004), suggesting negligible first-pass metabolism. Furthermore, brain penetration of PPARγ agonists such as RGZ may be impaired if they are substrates of P-glycoprotein, a major drug efflux transporter in the BBB (Gold et al., 2010). Given the structural similarities of RGZ and PGZ, the passage of PGZ through the BBB may also be limited (Chang et al., 2015b). Thus, modifications of drug formulation, structure, or dose might be needed to further increase the passage of PPARγ agonists through the BBB. For example, the stereoselectivity of PGZ has been shown to increase its distribution in the CNS and improve its therapeutic efficacy in Alzheimer’s disease (Chang et al., 2015b). Another alternative is to boost the activation of endogenous PPARγ. For example, endogenous prostaglandin J2 activates PPARγ and elicits anti-inflammatory effects (Finch et al., 2017).

Table 1.

Clinical studies of PPARγ agonists safety and therapeutic effects in ischemic stroke.

Study ID	Reagent	Indication	Patient	Adverse events	Effects	Center	Ref.
	PGZ or RGZ	Functional Independence Measure (FIM) of ischemic stroke	Stroke with T2DM	Not mentioned	PGZ and RGZ significantly improved FIM.	US	(Lee and Reding, 2007)
	PGZ or RGZ	Incidence of stroke	T2DM	Heart failure	RGZ imposed a higher risk of developing stroke or heart failure, PGZ did not influence risk of stroke	China	(Lu et al., 2013)
	PGZ or RGZ	Incidence of stroke	T2DM aged 65 or older	Not mentioned	RGZ increased incidence of stroke.	US	(Graham et al., 2010)
	PGZ or RGZ	Incidence of stroke	T2DM with high risk of cardiovascular diseases	Not mentioned	RGZ increased incidence of stroke.	US	(Shaya et al., 2009)
	PGZ or RGZ	Incidence of stroke	T2DM	Not mentioned	No significant effects	UK	(Azoulay et al., 2010)
	PGZ or RGZ	Incidence of TIA	T2DM	Not mentioned	RGZ increased incidence of TIA	Taiwan	(Hsiao et al., 2009)
	PGZ or RGZ	Incidence of stroke and TIA	DM	Not mentioned	No significant effects	US	(Habib et al., 2009)
	PGZ or RGZ	Incidence of stroke	DM patients aged 65 or older	Congestive heart failure	No significant effects	US	(Winkelmayer et al., 2008)
ISRCTNNCT00174993 (PROactive)	PGZ	Incidenc e of stroke	T2DM	Edema, heart failure, hypoglycemia, bladder tumors, prostate cancer, weight gain	PGZ reduced risk of stroke and recurrent stroke	Multicenter	(Charbonnel et al., 2004; Dormandy et al., 2005; Katzan, 2007; Wilcox et al., 2007; Erdmann et al., 2010; Scheen, 2012; Erdmann et al., 2014; Erdmann et al., 2016)
NCT00091949 (IRIS)	PGZ	Incidence of secondary stroke or TIA	Insulin-resistant, nondiabetic patients with recent ischemic stroke or TIA	Edema, shortness of breath, bone fractures and weight gain	PGZ prevented secondary ischemic stroke and myocardial infarction	Multicenter	(Viscoli et al., 2014; Bath et al., 2016; Inzucchi and Furie, 2016; Kernan et al., 2016)
UMIN000013499	PGZ	Incidence of secondary stroke	Insulin-resistant, nondiabetic patients with recent ischemic stroke or TIA	Edema, rashes, bladder cancer, cytopenia, heart failure	PGZ reduced incidence of recurrence of stroke	Japan	(Tanaka et al., 2015)
	PGZ	Incidence of stroke	T2DM with high risk of cardiovascular diseases	Edema, cancer	No significant effects	Japan	(Yoshii et al., 2014)
	PGZ	Incidence of stroke	T2DM	Bladder cancer and bone fracture	PGZ reduced risk of stroke compared with insulin in T2DM patients	US	(Vallarino et al., 2013)
	RGZ	Incidence of stroke	Metabolic syndrome undergoing percutaneous coronary intervention (PCI)	Not mentioned	RGZ displayed a statistical trend to decrease stroke incidence	Multicenter	(Bhatt et al., 2007)
NCT00379769 (RECORD)	RGZ	Incidence of stroke	T2DM	Heart failure, myocardial infarction, bone fracture	RGZ reduced incidence of stroke	Multicenter	(Home et al., 2009; Nissen and Wolski, 2010; Mahaffey et al., 2013)
NCT00006305 (BARI2D)	RGZ	Incidence of stroke	T2DM and coronary artery disease	Bone fracture	RGZ reduced incidence of stroke	Multicenter	(Bach et al., 2013)
	RGZ	Incidence of stroke	DM patients aged 65 or older	Not mentioned	No significant effects	Quebec	(Vanasse et al., 2009)

8. Summary

PPARγ is a widely-expressed nuclear receptor that regulates the transcription of genes involved in fatty acid storage, glucose homeostasis, insulin sensitivity, anti-inflammatory effects, redox balance, and stem cell differentiation, among other roles. PPARγ regulates PPREs as well as other signaling pathways to elicit neuroprotection and tissue repair. The products of the genes that are modified by PPARγ are found in almost every subcellular compartment as well as the extracellular space. The remarkable breadth of its effects makes PPARγ an attractive therapeutic target for many disorders, including acute brain injury. The activation of Nrf2 and the inhibition of NF-κB are central to the anti-oxidative and anti-inflammatory effects of PPARγ. PPARγ can reduce expression of pro-inflammatory mediators such as IL-6, COX-2, iNOS, and TNF and increase expression of anti-inflammatory mediators such as IL-10 (Chen et al., 2012), which collectively mitigate inflammation in a synergistic fashion. In the acute injury phase, PPARγ may protect against tissue damage by inhibiting neuronal death, demyelination, and BBB disruption. During the chronic recovery phase, PPARγ may facilitate CNS repair by promoting resolution of inflammation, debris clearance, neurogenesis, remyelination, and resealing the BBB. Figure 6 lists the hypothetical mechanisms underlying the protective and reparative effects of PPARγ. Furthermore, PPARγ appears to protect against systemic comorbidities, such as hyperglycemia, hypertension, AS, dyslipidemia, renal dysfunction, and infection, all of which may influence the progression of brain injuries. Many unresolved questions remain, however, including 1) how PPARγ balances the beneficial and destructive faces of neuroinflammation, 2) whether PPARγ directly inhibits demyelination, 3) how PPARγ affects OPC recruitment and differentiation, and 4) whether PPARγ promotes neurogenesis and neuroplasticity. Furthermore, there are serious challenges associated with the application of PPARγ agonists in CNS injury conditions, and some of the effects of PPARγ agonists (such as TZDs) may even be PPARγ-independent, which may cloud the interpretations of preclinical studies. Although preclinical and clinical work supports the therapeutic potential of PPARγ in ischemic stroke, whether PPARγ agonists will be effective in human TBI and SCI is not clear, and this gap in our knowledge is deepened by the poor predictive validity of some of the rodent injury models. Furthermore, PPARγ agonists such as rosiglitazone are associated with weight gain, edema, heart failure, and anemia (Chen et al., 2012). In this respect, however, PPARγ agonists are not unique—there are challenges and limitations of every drug available on the market today, in part due to a failure to account for the genotypic and phenotypic heterogeneity of human populations relative to inbred laboratory animals. At the least, our analysis of the clinical literature in Table 1 and the meta-analyses conducted by Liu and Wang specifically support the use of PPARγ agonists against primary and recurrent stroke in judiciously selected patient populations (Liu and Wang, 2015). Thus, we remain hopeful that identifying the mechanisms underlying the beneficial effects of PPARγ will reveal novel targets for future therapies, and that rational drug design based on PPARγ synthetic or endogenous ligands will lead to superior pharmacological agonists to treat diabetes, cardiovascular diseases, and acute brain disorders.

Figure 6.

Potential mechanisms underlying the CNS protective effects of PPARγ. Abbreviations: antioxidant response element (ARE), blood-brain barrier (BBB), bone marrow (BM), brain-derived neurotrophic factor (BDNF), Cu/Zn superoxide dismutase (Cu/Zn SOD), cyclooxygenase 2 (COX-2), endothelial precursor cells (EPCs), inducible nitric oxide synthase (iNOS), inhibitor of kappa B (IκB), insulin-like growth factor (IGF), matrix metalloproteinases (MMPs), neural stem cell (NSC), nicotinamide adenine dinucleotide phosphate (NADPH), nuclear factor (erythroid-derived 2)-like 2 (Nrf2), oligodendrocyte precursor cell (OPC), reactive oxygen species (ROS), tight junction (TJ), zona occludens (ZO).

Highlights.

• PPARγ is a master gatekeeper of cell fate decisions

• PPARγ controls inflammation and oxidative stress and limits acute brain injury

• PPARγ fosters tissue repair and facilitates long-term functional recovery

• PPARγ may reduce the risk for stroke in humans

• PPARγ should be further investigated for its therapeutic potential in acute brain injuries

Abbreviations

15d-PGJ2

15-deoxy-D12, 15 prostaglandin J2

15(S)-HETE

15(S)-hydroxyeicosatetraenoic acid

ADP

adenosine diphosphate

AGE

advanced glycan end product

AMPA

α-amino-3-hydroxy-5-methylisoxazole-4-propionic acid

ARE

antioxidant response element

AS

atherosclerosis

BBB

blood-brain barrier

BDNF

brain-derived neurotrophic factor

CCI

controlled cortical impact

CDK

cyclin-dependent kinase

CGZ

ciglitazone

CTL

cytotoxic T lymphocyte

CMV

cytomegalovirus

CNS

central nervous system

COX

cyclooxygenase

DC

dendritic cell

DM

diabetes mellitus

EAE

encephalomyelitis

FDA

Food and Drug Administration

Galectin

galactose-specific lectin

GFAP

glial fibrillary acidic protein

GFP

green fluorescent protein

GLT

glutamate transporter

HDAC

histone deacetylase

HDL-C

high-density lipoprotein-cholesterol

HSP

heat shock protein

ICH

intracerebral hemorrhage

IFN

interferon

IGF

insulin-like growth factor

IL

interleukin

iNOS

inducible nitric oxide synthase

LDL

low-density lipoprotein

LPS

lipopolysaccharide

MBP

myelin basic protein

MCAO

middle cerebral artery occlusion

MCP

monocyte chemotactic protein

NCoR

nuclear receptor corepressor

NF

nuclear factor

NGF

nerve growth factor

NMDA

N-methyl-D-aspartic acid

NOX

nicotinamide adenine dinucleotide phosphate oxidase

NPC

neural progenitor cell

NR1C

nuclear receptor 1C

Nrf2

nuclear factor erythroid 2-related factor 2

NSC

neural stem cell

OGD

oxygen-glucose deprivation

OPC

oligodendrocyte precursor cell

PGZ

pioglitazone

PMN

polymorphonuclear neutrophil

PPAR

peroxisome proliferator-activated receptor

PPRE

peroxisome proliferator response element

RGZ

rosiglitazone

ROS

reactive oxygen species

RXR

retinoid X receptor

SAH

subarachnoid hemorrhage

SCI

spinal cord injury

SGZ

subgranular zone

SOD

superoxide dismutase

SVZ

subventricular zone

TAI

traumatic axonal injury

TBI

traumatic brain injury

TGF

transforming growth factor

TGZ

triglitazone

TLR

Toll-like receptor

TNF

tumor necrosis factor

Treg

regulatory T cell

TZD

thiazolidinedione