Glial Fibrillary acidic protein: From intermediate filament assembly and gliosis to neurobiomarker

Abstract

Glial fibrillary acidic protein (GFAP) is an intermediate filament-III protein uniquely found in astrocytes in the CNS, non-myelinating Schwann cells in the PNS and enteric glial cells. GFAP mRNA expressions are regulated by several nuclear-receptor hormones, growth factors and lipopolysaccharides. GFAP is also subjected to a number of post-translational modifications while GFAP mutations result in protein deposits known as Rosenthal fibers in Alexander disease. GFAP gene activation and protein induction appear to play a critical role in astroglia cell activation (astrogliosis) following CNS injuries and neurodegeneration. Emerging evidence also suggests that, following traumatic brain and spinal cord injuries and stroke, GFAP protein and its breakdown products are rapidly released into biofluids, making them strong candidate biomarkers for such neurological disorders.

GFAP overview and outline

Astrocytes are one type of glial cells in the CNS, a group that also includes resident and perivascular microglia, oligodendrocytes, radial glia and Muller cells. In fact, it is estimated that astroglia cells are the most abundant cell types in the brain, providing both structural and functional support for neurons (including neurotransmitter glutamate recycling and trophic factor release). Astrocytes (astroglia) are characterized by the presence of a unique structural protein, glial fibrillary acidic protein (GFAP) isolated and characterizated by Dr. Eng in 1969 [1]. GFAP is known to be present in non-myelinating Schwann cells in the PNS, and the enteric glia cells as part of the enteric nervous system (ENS) [2,3]. In this review article, we will first examine the structural layout of GFAP protein, its various splice variants and the pathological significance of GFAP mutations. We will discuss how GFAP is also regulated both at the transcriptional and the post-translational levels, and how such regulations might impact on GFAP's normal cytoskeletal functions and its involvement in maintaining the activated astroglial cell state (astrogliosis) following nervous system injury. We will further focus on the emerging evidence that GFAP and its modified forms as a promising protein biomarker for neurotrauma and stroke. Lastly, we will also discuss how GFAP was identified as a dominant autoantigen following traumatic brain injury (TBI) and its implications in terms of triggering a possible post-TBI and sustained autoimmune response towards the nervous system.

CNS-PNS-ENS specificity

GFAP and protein levels are highly expressed in the CNS (Figure 1)[4,5]- almost exclusively present in its astrocytes. GFAP protein is also present in the PNS along the peripheral nerve fiber track, such as the sciatic nerve. In this case, GFAP is localized to non-myelinating Schwann cells that are believed to be functionally similar to astrocytes [6,7]. In addition, GFAP can be found in the glia cells of the enteric nervous system (ENS) [2,3]. Such sub-epithelial glia cells have a trophic and supporting function of the intestinal epithelial cells and neurons. In addition, as GFAP-bearing glia cells are immediately associated with the enteric neurons and their nerve fibers in the submucosal and muscle layers of the gut, they are perfectly positioned to exert neuromodulation function (for example, by active uptake of extracellular neurotransmitters by glial cell surface neurotransmitter (NT) transporters and subsequent degradation. Enteric glia cell surface also has receptors for various NTs including purine P1 and P2 receptors that are responsive to neuron-released nucleotides and adenosine [8]. Since enteric neurons regulate gut motility and mucosal secretion, GFAP-bearing glial cells are thought to indirectly influence such processes as well. GFAP protein levels in enteric glia cells are also responsive to proinflammatory signals (such as IL-6 or bacterial endotoxin lipopolysaccharide LPS) [9]. Similarly, GFAP (and GDNF) levels are highly upregulated in mucosal plexus in the colon of patients with inflammatory bowel disease (IBD) [10]. In particular, GFAP might be an excellent index of enteric gliosis response to the more severe forms of IBD (such as ulcerative and infectious colitis) over Crohn's disease. Intriguingly, mouse enteric glial can undergo neurogenesis in response to injury [3]. In addition, enteric GFAP levels and phosphorylation are also increased in Parkinson's disease (PD) patients [11]. Further investigation is needed to understand the possible implications of this finding in PD patients.

Figure 1. GFAP tissue specificity.

(A) GFAP mRNA expression in human tissue and cells based on BioGPS database [4] (B) GFAP protein expression in various human and rat organs, tissues Human GFAP was extensively truncated to 48-38 kDa bands, likely due to postmortem proteolysis, while rat GFAP mainly exists in brain as 50-48 kDa form [5].

GFAP protein structure and Function

GFAP is a key intermediate filament (IF) III protein responsible for the cytoskeleton structure of glia cells and for maintaining their mechanical strength, as well as supporting neighboring neurons and the blood brain barrier (BBB) [1]. GFAP is structurally similar to other non-epithelial IF members (class III), including vimentin, desmin, and peripherin, and has a head, rod and tail domains. Activated astrocytes take on the morphology of thickened and elongated processes and GFAP-through its involvement in the IF network- is critical in the maintenance of such structure. In the most abundant isoform, GFAP-α (alpha), the head coil domain is followed by the rod domain that is composed of four coils (1A, 1B, 2A, 2B) which are flanked by three linker regions (1, 1,2 and 2, respectively) (Figure 2A). Such a structural organization is in fact highly conserved among class III IF proteins [12]. Crystal 3-D structure is currently not available, but through homology modeling using vimentin as a template, a 3-D structure of GFAP can be visualized (Figure 2B) [13]. Like other IF proteins, GFAP monomer transition to filamentous form by assembling first into dimers in parallel formation, followed by antiparallel assembly into tetramers, octamers and so on (Figure 2C) [14]. The N-terminal head domain is critically important for filament elongation (including the M₁ERRRITSARRSY₁₄ motif), while the C-terminal tail domain is also important in facilitating oligomerization [15].

Figure 2. Glial Fibrillary Acidic Protein (GFAP) structure and assembly.

(A) linear structure, functional domains and key modifications. (B) 3D-GFAP protein structure based on Ref. [13]; (C) Proposed GFAP dimer and tetramer assembly and oligomerization model, modified after Ref. [14].

Isoforms /splice variants

GFAP is encoded by a single gene mapped to human chromosome 17q21. To date, there are 10 isoforms/splice variants identified (Figure 3). GFAP-α (Isoform 1) is the predominant isoform in brain and spinal cord, but also presents in PNS [16] and has the classic 432 residues (protein accession # NP_002046.1) with the full usage of the 9 exons within the GFAP gene [17]. GFAP-δ, also called GFAP-ε (Isoform 2) (NP_001124491.1) is preferentially expressed by neurogenic astrocytes in the subventricular zone[18-22]. GFAP-δ includes the usage of intron before exon-8 and has alternate C-terminal and 431 residues. This unique C-terminal of GFAP-δ binds specifically to presenilins (PS1, PS2) proteins in yeast-two hybrid screening [20]. Presenilins are also involved in Notch-signaling and may play a critical role committing astrocyte into GFAP+ phenotype or neurogenic phenotype. It is tempting to suggest that presenilins–GFAP-δ interaction might have a regulatory mechanism of the Notch–singling pathway in astrocyte-cell fate determination [20]. Obviously further work is needed to clarify the physiologic and/or pathophysiologic significance of such interaction. Enhanced GFAP-δ/ε expression in human astrocytic tumor has been reported [23]. GFAP-κ (kappa) is the third isoform (isoform 3) (NP_001229305.1) including an intron before exon-8 and alternate shorten C-terminal and has 328 residues [24] (2007). GFAP-α, GFAP-ε, and GFAP-κ are the three most well characterized isoforms of GFAP. Of interest is that mRNA expression levels of all three isoforms gradually increase during development of the embryonic pig brain [24].

Figure 3. Known GFAP isoforms and features.

(Left) Description of key features of GFAP isoforms. (Right) Schematic comparison of key GFAP isoforms in linear models.

GFAP-β is highly expressed in non-myelinated Schwann cells in the PNS [25]. It includes the usage of sequence before exon-1; the exact start site is currently unknown and has at least 432 residues. mRNA and protein levels of GFAP-β are apparently induced following neural injury [19]. GFAP-γ also includes the sequence before exon-1, but with omitted exon-1 and includes an intron before exon-2. Again, its exact start site is currently unknown, but its length is predicted to be less than 432 residues. GFAP-γ mRNA is enriched in corpus callosum and is also present in bone marrow and spleen [26]. GFAP-ζ (zeta) is formed by the inclusion of the intron before exon-9 and has >438 residues [21] (Figure 3).

There are also four isoforms of GFAP collectively called GFAP+1, reflecting its single nucleotide frame-shift mediated variant formation. They are found in a subset of astrocytes throughout the brain. GFAP+1 includes the GFAPΔEx6 with a skipped exon-6 and lacking Coil 2B within the core region (347 residues), GFAPΔ164 which has shortened exon-6, 7and lacks Coil 2B (366 residues), GFAPΔ135 which has shortened exon-6 and lacks Coil 2B (374 residues), and GFAPΔEx7 which has skipped exon-7 (418 residues) (Figure 3). The detailed structures of these isoforms are also described by Middeldorp and Hol [17]. Interestingly, in AD brains, astrocytes near amyloid plaques have increased staining of both GFAP-α and GFAP-δ while GFAP+1 were found to be limited to a subset of astrocytes with long processes, with their number increased during the progression of AD [27]. Also, three GFAP+1 splice forms (GFAPΔ135, GFAPΔEx6 and GFAPΔ164) can be found in the pyramidal neurons of the hippocampus of AD and Down's syndrome patients [28]. The implication of such neuronal expression of GFAP is still unknown. Lastly, immunohistochemical evidence shows that GFAPΔEx6 and GFAPΔ164-positive astrocytes are found elevated and localized in focal lesions associated with chronic epilepsy; however, they appear far out-numbered by the dominant GFAP-α positive astrocytes in the same lesioned regions [29]. Lastly, as many of these isoforms of GFAP have alternative N-terminal (beta, gamma), or alternative (δ/ε, κ, ζ) or shortened C-terminal (GFAP+1), it is tempting to suggest that they might have significant effects on GFAP filament assembly (Figure 3).

GFAP Mutations/ SNP and Alexander Disease

GFAP is also a target for a single nucleotide polymorphism (SNP), resulting in Alexander Disease [30-33]. A number of mutations were found mainly in the coding regions of the GFAP gene (L47; C79; H79; E223, H239; A244, R258, C289; D295 & R416), but a few mutations are found in the promoter regions (Table 1, top). These mutations were suggested to be “gain-of-function” mutations [34], as GFAP knockout mice does not duplicate the Alexander Disease phenotype [35,36]. The mutant GFAP proteins show a range of competency in intermediate filament assembly - from (i) fully able to assemble into full length (10 nm) IF, (ii) capable of polymerization but not able to extend to full length IF, to (iii) not able to polymerize at all. In addition, mutated GFAP proteins are more prone to aggregate formation – producing astrocytic inclusions (called Rosenthal fibers) found in Alexander Disease brains [31]. Messing et al. recently reviewed how various GFAP mutations are linked to pathology of Alexander Disease [34]. Some researchers believe GFAP aggregates are toxic to astrocytes – and thus might contribute to the astroglial degeneration and the subsequent white matter degeneration pathology observed in Alexander Disease. Other researchers believe that Rosenthal fiber formation might simply be a reflection of excessive GFAP aggregate accumulation in dysfunctional astrocytes [34] (see Outstanding Questions). Lastly, caspase-cleaved N-terminal GFAP fragment might contribute to GFAP aggregate formation in Alexander Disease [37].

Table 1.

GFAP modifications and modulators

Known modifications of GFAP
Modification	Site	Effects	Key Refs
Phosphorylation	T7(PKA) S8 (PKA, PKC, cdk2 S13 (CAMPKII, PKA, PKC) S17 (CAMPKII) S38 (CAMPKII, PKA, PKC) S289 (CAMPKII)	Inhibition of GFAP polymerization	[38-42]
Citrullination	Arginine deimination to citrulline at R30, R36, R270, R406 & R416	Altered GFAP confirmation; increased autoimmune response	[43]
Acetylation	Acetylated lysine residues at K89, K153, K189, K218, K259 & K331	Found in ALS spinal cord (currently unknown effects of acetylation)	[46]
Proteolysis	Calpain sites (major): 69-70; 383-384 (BDP of 44K-38K) Caspase sites : 78-79, 266-267 (BDPs of 30K, 20K)	Disruption of IF elongation BDP appears glia-toxic	[47-54]
SNP (Single nucleotide polymorphism)	Promoter region; L47; C79; H79; E223, H239; A244, R258, C289; D295 & R416	Formation of GFAP mutant aggregates (Rosenthal fibers) and causing Alexander Disease	[32-34]

GFAP modulators
Modulating molecule	Chemical class	Mode of action & effects	Key Ref.
Inducers
Thyroid hormone, Glucocorticoids	Nuclear-receptor hormones	GFAP gene transcriptional activation	[62-63]
CNTF, (FGF and TGF-β1)	Growth factors	GFAP gene transcriptional activation	[60-61]
Lipopolysaccharide (LPS)	Bacteria endotoxins	GFAP gene transcriptional activation	[65,69]
GDNF and neurturin	glia-derived trophic factor	Promote GFAP gene transcriptional activation or stabilize GFAP protein levels	[74-75]
Suppressors/ Inhibitors
S100b	Glial protein	Binds and inhibits GFAP phosphorylation, promote disassembly of IF	[72-73]
Withaferin-A	Steroidal lactone (from winter cherry)	GFAP-covalent modification of Cys-294; inhibition of IF function	[79-80]
Prosaptide	N-terminal of prosaposin	Down-regulation of GFAP expression (precise mode of action unknown)	[76-78]
Ibudilast (AV411)	Pan PDE inhibitor	Down-regulation of GFAP expression (precise mode of action unknown)	[68, 70, 81]
Clomipramine (Clofranil)	Tricyclic antidepressant	Down-regulation of GFAP expression (precise mode of action unknown)	[82]
Aspirin/Acetylsalicylic acid	Cox-1 inhibitor	Down-regulation of GFAP expression (5 mM in vitro) (precise mode of action unknown)	[83]
Curcumin	Curcuminoid (from turmeric)	Down-regulation of GFAP expression (precise mode of action unknown)	[84]

GFAP post-translational modifications

GFAP protein is also subjected to a number of post-translational modifications (PTM) (Table 1, Figure 2). GFAP protein is highly regulated by protein kinases (such as protein kinase–A, calmodulin dependent kinase II and PK-C), with many phosphorylation sites mapped to the N-terminal domain: T7 (PKA), S8 (PKA, PKC, cdk2), S13 (CAMPKII, PK, PKC), S17 (CAMPKII), S38 (CAMPKII, PKA, PKC) and S289 (CAMPKII) (Figure 2; Table 1) [14,38-41]. One of the key phosphorylation pathways of GFAP appears to involve the G-protein-coupled mGluR receptor leading to calcium influx and CAMPKII activation [40,42]. As glutamate is a major excitatory neurotransmitter in the brain, such a mechanism might be involved in the cross-talk between neurons and glial cells [42]. In addition, since most of these phosphorylation sites reside at the N-terminal head domain, phosphorylation of GFAP has a negative effect on filament assembly. GFAP phosphorylation is elevated after hypoxic-ischemia in neonatal pig brain [38].

GFAP was recently identified as endogenously citrullinated, i.e. several arginine residues undergo deamination to citrulline (R30, R36, R270, R406 & R416) [43]. The exact extent of citrullinated GFAP is not fully understood and how it alters GFAP function is not elucidated. However, citrullinated protein epitopes are implicated in evoking autoimmune response [44,45] which could be relevant as GFAP is a dominant autoantigen in certain neurological disorders (see below) [5]. A number of lysine residues are also found to be subjected to differential acetylation (K89, K153, K189, K218, K259 & K331) in human spinal cord of amyotrophic lateral sclerosis patients (Figure 2, Table 1) [46]. Such acetylation is spread across the whole GFAP protein, but their effects on GFAP structure and functions are currently unknown and warrant further investigation.

GFAP proteolysis/fragmentation in cultured astrocytes was noted, in some early reports, as mediated by the calcium-activated protease calpain [47,48]. Increases in the fragmented form of GFAP has also be found in the spinal cord of patients with ALS [49]. We recently documented that GFAP is highly vulnerable to calpain-mediated truncation at both the C- and N-terminals. This results in a series of truncated GFAP breakdown product (BDP) (38 to 44 kDa) as compared to the 50 kDa intact protein during glial cell challenge in vitro [50,51] or in experimental brain injury in rodents as well as in human cerebrospinal fluid (CSF) from TBI patients [5,50,51]. These fragments can also be observed in post-mortem human brain preparations (Figure 1). Similarly, we also found that GFAP-BDP can be observed in rodent and human CSF after spinal cord injury [52,53]. The main calpain cleavage sites have been mapped out to be N59-A60 and T383-F384, although there are likely alternative cleavage sites near them. Regardless, the “limit fragment” (~38 kDa)with both N- and C-terminal truncation leaves essentially the rod domain intact but the head and tail domains removed. It is thus predicted that such GFAP-BDP would be unable to assemble into filaments. Caspase can also fragment GFAP in pro-apoptotic conditions, producing a shortened N-terminal fragment in AD brain (about 20 kDa) [54] at the DLTD₂₆₆-A₂₆₇ site, which is located at the beginning of coil 2B (Figure 2). Based on our own data, we also identified a second cleavage site at ELND₇₈*₇₉R between the head and rod domains (Figure 2). Chen et al recently demonstrated that GFAP might be cleaved by caspase-6 and that the C-GFAP (C-terminal GFAP) fragment (24 kDa) is unable to assemble into filaments [37]. The 26 kDa N-GFAP (N-terminal GFAP) can form filamentous structures with variable length but it is prone to form filamentous aggregation [37]. However, in experimental TBI and in human TBI CSF, the calpain mediated GFAP fragments (44-38 kDa) appear to dominate over those produced by caspase [5,50]

The role of GFAP in astrocyte activation (astrogliosis) and GFAP inducers/activators

Astroglial cells respond to brain injury and other neuro-perturbative conditions by undergoing “reactive astrogliosis”, a process whereby astroglial cells undergo cellular hypertrophy (increase of size and protein (GFAP) expression) and proliferation (increase number of glial cells) [1]. TBI itself, and its associated neuroinflammation, cause activation (and proliferation) of astroglia cells into damaged areas and a concomitant increase in GFAP levels [55]. Importantly, since GFAP, together with vimentin, is the key component responsible for the assembly and extension of the intermediate filament inside the astrocytic processes, it is believed that GFAP induction is critically important for the formation of extended and thickened astrocytic processes observed in reactive gliosis. In fact, an increase in GFAP protein is a prominent feature of TBI and degenerative diseases such as PD and AD [5,56-59]. Different locations around the lesion site may exhibit different severities of gliosis; e.g., a glial scar at the location of damaged tissue may be surrounded by areas with less severe astrocyte proliferation or hypertrophy. Diffused traumatic injury can result in diffuse or more moderate gliosis without scar formation. Activated astrocytes (with highly expressed GFAP) are found to surround amyloid and neurite plaques in AD. Astrocytes in culture can also be used to study glial activation and GFAP induction. It is generally regarded that certain levels of post-injury gliosis might be beneficial to the recovery process following brain injury while excessive gliosis and its associated neuroinflammatory responses will have a negative impact to brain structural and functional recovery. [58]. GFAP knockout mice were found to be essentially normal developmentally. In addition, post-spinal cord injury axonal sprouting and regeneration appear largely unaffected in GFAP-/- mice [35]. These results pointed to possible compensatory mechanism involving the related vimentin protein [35]. However, in a separate study with a peripheral nerve crush model, GFAP-/- mice did show defective Schwann cell differentiation and delayed nerve regeneration [36]. Again this is consistent with the concept that post-injury-GFAP induction and associated reactive gliosis might in fact promote neuroregeneration.

A number of growth factors such as CNTF, FGF and TGFβ can induce GFAP gene transcription activation [60,61], leading to increasing GFAP protein levels (Table 1, Figure 4A). Nuclear-receptor hormones (Thyroid hormone, Glucocorticoids) also can activate GFAP transcription [62]. The effect of thyroid hormone might also be mediated via the activation of ROCK pathway [63]. These hormone and growth-factor-based GFAP gene regulators are through to be potentially important in the induction of mature astroglia formation. On the other hand, LPS, via the production of nitric oxide is also a robust activation of cultured glial cells and induction of GFAP gene transcription [64] – it is likely that LPS activates toll-like receptor (TLR-4) and possibly CD14 in astrocytes [65]. Systemic administration or intracerebral injection of LPS also leads to gliosis and upregulation of GFAP protein and neuroinflammation in live animals [66-71]. Thus, LPS-based GFAP protein upregulation is more in line with astrogliosis formation.

Figure 4. GFAP patterns in injured and activated glia cells in vitro and in vivo.

(A) Rat GFAP staining of resting primary glia cells or Staurosporine (0.5 μM, 5 h)-injured glia or CNTF (200 nM for 24 h)- activated glial cells. (B) Naïve (control) rat cortex or 24 h after experimental TBI showing injured glia cells or 3 day after TBI showing activated glia cells. GFAP was stained red, while counterstain was DAPI for nuclear DNA (blue). (C) Anti-GFAP-immunblotting of control glia cell lysate shows intact GFAP of 50 kDa with two anti-GFAP antibodies. Upon induction of necrosis (induced by 10 μM A23187) or caspase-dominant apoptosis (induced by 5 mM EDTA) [51,112], core-directed antibody shows calpain-mediated 38 kDa limit GFAP-BDP after A23187 treatment, while C-terminal antibody detects two district caspase-mediated GFAP-BDPs of 44K and 20K after EDTA treatment (Results shown in both (A) and (B) are from a published study by us [5], but were not shown as representative examples in a primary publication).

GFAP is also modulated endogenously by calcium-dependent binding to EF-hand S100β. Upon binding, S100β promotes disassembly of IF [72]. S100β-binding leads to inhibition of GFAP phosphorylation, which might be the mode of action that led to disassembly of IF [73]. Glial derived neurotrophic factor (GDNF) and neurturin (another glia-derived trophic factor) have also been shown to be glia-protective [74,75]. Thus they might serve as auto-regulating factors that either promote GFAP gene transcriptional activation or stabilize GFAP protein levels (Table 1).

GFAP Suppressors and Glial Targeting Therapeutic Agents

A number of therapeutic agents can inhibit glia cells function or suppress GFAP expression (Table 1). An interesting compound is Prosaptide which is a 14-mer (Thr-D-Ala-Leu-Ile-Asp-Asn-Asn-Ala-Thr-Glu-Glu-Ile-Leu-Tyr) derived from the neurotrophic and glia-tropic N-terminal of human glycoprotein prosaposin [76-78]. Prosaptide is also known to cross BBB to exert its GFAP-suppression effects. Secondly, a number of drug-like agents have reported effects in suppressing either GFAP protein expression and/or gliosis induction. These include Withaferin-A (WF-A), ibudilast, clomipramine, aspirin and curcumin. Withaferin A is a steroidal lactone which binds to and thus inhibits GFAP and the related IF protein vimentin [79,80]. It was first isolated from Ayurvedic medicine Winter cherry and is a cancer drug due to its anti-angiogenesis effects. It was found that Withaferin-A covalently modifies the single Cys-294 of GFAP (and a homologous cysteine residue in vimentin protein) leading to inhibition of IF function of GFAP and filament disassembly [80] (Figure 2, Table 1). WF-A can in fact attenuate GFAP levels and glial cell activation and appears BBB-permeable as it exerts its central effects in a model of retinal gliosis [80]. Ibudilast is an anti-inflammatory drug currently used mainly in Japan as an asthma medicine. Ibudilast is board-spectrum phospodiesterase inhibitor that also inhibits methamphetamine-induced GFAP upregulation and gliosis and attenuates methamphetamine self-administration and relapses [68,70]. Ibudilast also appears to attenuate mechanical allodynia in rat models of neuropathic pain [81]. Tricyclic antidepressant clomipramine is also found to robustly suppress GFAP levels in vivo [82]. Aspirin/Acetylsalicylic acid (5 mM in vitro), although having various anti-inflammatory actions including inhibition of COX2, was unexpectedly found to also inhibit GFAP up-regulation and glial cell activation via an NF-kB pathway in vitro [83]. Curcumin might also have beneficial effects by down-regulation of GFAP expression in an in vitro model of Alexander disease [84]. Although the exact modes of action (direct binding or regulation of protein levels) for GFAP attenuation by prosaptide, ibudilast, clomipramine, aspirin and curcumin are presently unknown, they might still be useful in studying the role of GFAP in neuro-disease models. Screening efforts are underway in finding additional druggable molecules that can either specifically suppress or reverse GFAP aggregate formation (for Alexander disease) or suppress GFAP expression (for gliosis), including several novel chemical candidates [76,82]. Molecular modeling and compound docking in silico is also being applied for such drug discovery effort [13]. However, at present it is unclear if binding of small molecule to GFAP monomer is sufficient in (a) reducing the steady state levels of GFAP; (ii) inhibiting GFAP polymerization and thus astrogliosis, or (b) destabilizing intracellular GFAP aggregates (see Outstanding Questions).

GFAP-breakdown product as marker for glial cell injury

Neurotrauma conditions are often associated with neuronal injury or death. However, since astrocytes are a major cell type in the brain we proposed that they are also subjected to mechanical or chemical injury shortly after neurotrauma and in neurodegenerative disorders. Our recent work shows that “glial injury” is a key pathologic event during the acute / subacute phase of neurotrauma (in animal model of TBI, spinal cord injury (SCI), as well as in human TBI/ SCI CSF samples), as aided by our newly identified glial injury signature – i.e. the proteolytic conversion of intact glial protein GFAP (50 kDa) into breakdown products (GFAP-BDPs) of 44-38 kDa by calpain [5,50,52,53,85,86] (Figure 4B). Similarly, GFAP is degraded into GFAP-BDPs by calpain proteases in cultured glial cells challenged with cytotoxins (staurosporine or calcium ionophore) [51] (Figure 4A). Caspase protease also can contribute to GFAP proteolysis, to a lesser extent (see above) [5].

GFAP as biomarker protein for acute CNS injury and other neurological conditions

Increasing evidence also suggests that GFAP and GFAP-BDP might be a useful tool as biofluid-based marker for a number of neurological conditions. The overall concept is that brain injury causes the release of GFAP-BDP and to a lesser extent, full length GFAP from injured astrocytes to the interstitial fluid (ISF)/extracellular fluid, where these proteins equilibrate into the subarachnoid CSF compartment, then release to the circulating blood by direct venous drainage (glymphatic pathway) ([113] or continue to follow the CSF flow and eventually enter the circulation by diffusing pass the (possibly compromised) brain blood barrier (BBB) (Figure 5). One of the advantages of GFAP as a brain biomarker is that it shows strong brain-specificity and high expression levels to the brain (Figure 1). There are over 1.7 million of TBI cases in the US annually [87]. TBI is categorized by Glasgow comma scale or CT abnormality: severe (GCS 3-8, Cranial computerized tomography (CT) abnormal; ~10%), moderate (GCS 9-12, CT abnormal; ~5-10%) and mild (GCS 13-15, CT normal). Additionally, impact and blast overpressure-wave induced TBI are a signature injury of recent warfare among US military personnel [88,89]. In penetrating TBI as well as overpressure blast-wave induced brain injury in rats, GFAP levels are elevated in CSF and/or serum at 4 to 24 hour after injury [50,90]. Emerging evidence from others and us also shows a robust release of GFAP and its BDPs into CSF and blood compartments hours and even days after the initial severe TBI event [91,92]. GFAP levels are linked to CT pathology and outcome in these subjects. GFAP blood levels can also help predict secondary insults after severe TBI [93]. Since moderate TBI is often difficult to distinguish from mild TBI in emergency room, there are two important studies that show that the blood levels of GFAP and/or its BDPs can be used to predict CT abnormality and thus can potentially differentiate mild vs. moderate TBI [85,86]. Consistent with these reports, blood GFAP loads were also elevated in individuals undergoing breacher training and exposures to blast overpressure waves [94]. GFAP levels in both brain tissue and serum are also elevated in animal model of blast overpressure-wave induced TBI that is combined with psychological stress [95].

Figure 5. Release of GFAP and GFAP-BDP into biofluid as acute CNS injury biomarker.

(Left) List of neuro-diseases and disorders in which GFAP and/or GFAP-BDP is released into biofluid. (Right) Schematic showing how GFAP BDP is generated after TBI and how both GFAP-BDP and to a lesser extent intact GFAP are released into interstitial fluid (ISF)/extracellular fluid. From there, these proteins diffused into the subarachnoid CSF [113]. GFAP-BDP/GFAP then either drain directly into the veins (glymphatic pathway, dashed arrows) or continue to follow the CSF flow through the ventricles and then enter the circulation by diffusing through the BBB. Blood-based GFAP and GFAP-BDP can be detected as a biomarker, but they also can serve as autoantigen, triggering autoantibody response in a subset of patients.

There are two type of strokes; hemorrhagic stroke or ischemic stroke. The former is a result of intracerebral hemorrhage, while the later is due to a blockage of major blood vessel, resulting in brain ischemia. It is important to differentiate the two, as tissue plasminogen activator (TPA) is currently the only FDA approved treatment for ischemic stroke. TPA (usually given within 3-4 hours post-injury) works by dissolving the existing blood clots in the brain and reducing further blood clot formation. However, TPA is counter-indicated in cerebral hemorrhage conditions. Increasing evidence shows that GFAP might be a key marker that can distinguish the two. Several studies have shown that GFAP is released within 3-4 hours following hemorrhagic stroke, while its release is delayed to 24-48 h post injury in ischemic stroke [96-100]. Thus, early release of GFAP into blood is a clinically important differentiating indicator of hemorrhagic stroke over ischemic stroke. In addition, GFAP appears to be useful in tracking progression or outcome of ischemic stroke [97,98] (Figure 5).

Thus these data taken together, strongly suggest that once a robust GFAP sandwich ELISA test become universally available, it will help realize the potential of GFAP as a biomarker for various forms of CNS injury. This is considered an important discovery as to date there are no FDA-approved in vitro diagnostic biofluid tests to monitor brain injury. Thus, it can be envisioned that a GFAP biofluid test being used routinely in brain injury patient management and changing medical practice. In addition, based on others’ and our recent observations, GFAP-BDPs formation and their subsequent release (rather than intact GFAP) into biofluids appears intimately tied to astrocyte damage or cell death after brain injury (Figure 5). Thus we reason that the elevated levels of GFAPBDP in accessible biofluids (CSF, blood) could then be used to track the levels and kinetics of astrocyte cell damage following brain trauma – In fact this is an under-investigated area that could be exploited for novel therapy development. In addition, from a clinical standpoint, it will be important to define if there are any differentiating utilities in monitoring intact GFAP vs. GFAP-BDP in biofluid following brain injury (see Outstanding Questions).

GFAP biomarker tests can also be further developed into “theranostic” guide to help much needed drug development for CNS injury as FDA approved therapies in this area are almost non-existing. For instance, drug treatment given after brain injury that are beneficial in reducing brain injury or promoting brain recovery should also reduce the biomarker load [101-103]. GFAP assay might be useful as a neurotoxicity biomarker tool during drug development as well [104]. Additionally, since most glioma (including glioblastoma) also express GFAP, monitoring of biofluid levels of GFAP might reflect such tumor formation or expansion [105,106].

GFAP protein or PTM-form of it might also become an autoantigen in neurodegenerative conditions, autism, workers exposed to lead, or in chronic cerebrovascular disorders [107-110]. We recently reported that about 40% of severe TBI patients unexpectedly show an immunodominant autoantibody response to GFAP and its BDPs 5-6 days post-injury [5]. Our hypothesis is that TBI causes (i) protease-mediated GFAP-breakdown (BDP) product formation in injured glial cells and (ii) subsequent release of such GFAP-BDPs in substantive quantity through the compromised brain-blood barrier into circulation. (iii) The combined effect is that GFAP-BDP becomes accessible to and recognized by the immune system as a non-self-protein, triggering autoantibody response in these vulnerable individuals. (iv) Autoantibody specifically targeting a major brain protein such as GFAP might trigger a persistent autoimmune attack of the CNS and negatively affect TBI patient's long-term outcome. Our findings of GFAP being an post-neurotrauma autoantigen have also been extended to spinal cord injury [52]. It is also possible that citrullinated GFAP are preferentially recognized by the immune system as an autoantigen, as citrullinated proteins are identified as robust autoantigens in autoimmune diseases such as rheumatoid arthritis [44,45] (see Outstanding Questions Box). It is important to further examine if such autoantibody response is long lasting and whether it leads to autoimmune attack of the nervous system, similar to multiple sclerosis. Interestingly, Serum autoantibody to GFAP also might have utility in tracking glioma progression [111].

CONCULSION AND FUTURE DIRECTION

This review points to the fascinating roles that GFAP is playing in our nervous system. They include maintaining the structure and functions of the GFAP-bearing cells in the CNS, PNS and ENS, while also mediating astroglia cell activation in the event of nervous system injury. It also appears that the number of GFAP splice variants in human is likely to exceed the 10 isoforms identified to date, while the number of GFAP mutations that lead to Alexander disease is still growing. GFAP is also highly regulated both pre- and post-translationally, and is a target of a number of pharmacological agents. Mounting clinical evidence also suggests that GFAP is arguably one of the most promising biofluid-based biomarker with diagnostic, prognostic and even theranostic utilities in managing neuro-injuries and possibly other neuro-diseases.

Highlight.

GFAP is tightly regulated at mRNA level and by post-translational modifications

GFAP plays a critical role in astrogliosis after CNS injury and in neurodegeneration

GFAP as a potential drug target for Alexander's disease and neurodegeneration

GFAP and breakdown products are emerging biomarkers for TBI and neuro-injuries

Outstanding Questions Box.

Outstanding Questions

• Does GFAP citrullination and/or acetylation plays a role in evoking autoimmune response to GFAP after brain injury or in other neurological disorders?

• What is the exact mechanistic relationship between aggregates of GFAP (Rosenthal fibers), astrocytic degeneration and white matter degeneration in Alexander disease?

• How might the expression of different isoforms of GFAP produce different astrocyte responses to neuroinjury and/or other neuro-perturbation?

• What are the Differentiating utilities in monitoring intact GFAP vs. GFAP-BDP in biofluid following brain injury [e.g. Do the CSF or blood levels of GFAP-BDP better reflect the extent of astrocyte injury post-TBI than the levels of intact GFAP protein]?

• If specific GFAP-binding or GFAP- inhibiting pharmacological agents can be developed, will they be useful in treating neuroinjury conditions or other neurological disorders (e.g. Alexander's disease)?