According to Butler and Hodos (2005), in each primary vertebrate radiation—cartilaginous fishes, ray-finned fishes, and fringe-finned fishes with tetrapods—two types of brain organization can be distinguished, type I (“laminar”) and type II (“elaborated”). In type I brains, neurons either remain in the periventricular zone or migrate short distances from it. The brain ventricles are large, and the brain wall is relatively thin. In type II brains, neurons migrate extensively; therefore, these brains are larger and composed of numerous nuclei. The ventricles are of reduced size, whereas the brain wall has thickened. All amniotes, teleost fishes, batoids, and the advanced so-called “galeomorph” sharks have type II brains; the others have type I brains (Butler and Hodos, 2005). Note that some batoids exhibit a high range of cerebralization. Their brain weight/body weight ratios overlap the lower range of birds and mammals (Jerison, 1973), mainly due to the evolution of the telencephalon and cerebellum. GFAP-free areas occur only in “elaborated” brains, but not in each of them, see basal teleosts and galeomorph sharks.

During evolution, two types of astroglia appeared, tanycytes and true, stellate-like astrocytes. The tanycytes (Horstmann, 1954) are thin, elongated, fiber-like glial cells usually of ependymal origin (ependymoglia); they are best visible in Figures 4a, b, 8a, 10b. “Astrocyte,” in the strict sense, refers to stellate-shaped glial cells, which are independent from the ependyma. In contrast, the term “astroglia” comprises both astrocytes and tanycytes, according to the terminology proposed by Mugnaini (1986).

The appearance of astrocytes, at least their predominance, is phylogenetically younger than that of tanycytes (see also Verkhratsky et al., 2019; Falcone, 2022). In reptiles, ray-finned fishes, and sharks, the predominant GFAP-immunopositive elements are tanycytes. Astrocytes are absent in ray-finned fishes (Kálmán et al., 2021) and turtles (Kálmán et al., 1994, 1997; Lazzari and Franceschini, 2006; Lőrincz and Kálmán, 2020a); they are only complementary elements of astroglia in crocodilians (Kálmán and Pritz, 2001; Lőrincz and Kálmán, 2020b), snakes and lizards (Lőrincz and Kálmán, 2020a), see also Monzón-Mayor et al. (1990), Lazzari and Franceschini (2001, 2005a,b), Ahboucha et al. (2003), and Lőrincz and Kálmán (2020b). The predominance of astrocytes is characteristic of birds and mammals; it appeared independently in these two groups, as it is not seen in either turtles or crocodilians. During their separate evolution, astrocytes also became predominant in skates and rays (Horstmann, 1954; Wasowicz et al., 1999), but – and it is interesting—only in the prosencephalon and mesencephalon, but not in the rhombencephalon (Kálmán and Gould, 2001; Ari and Kálmán, 2008a)—in contrast to mammals and birds, where they dominate the whole brain and spinal cord.

The appearance of astrocytes during evolution can promote the formation of regional differences in the GFAP content. In tanycytes, which are elongated, the scaffold-forming function of GFAP fibrils is mandatory, and a single glial nucleus controls the GFAP expression throughout the full length of the fiber-like cell from the ventricle to the meningeal surface of the brain wall. Once multiple astrocytes have replaced the long tanycytes through the brain wall: “astrocytes demarcate … functional compartments … regulated by single glial cells” (Nedergaard et al., 2003); they can form a versatile glial network, each astrocyte has control over its individual GFAP expression, and the shorter processes may miss the GFAP. Therefore, the GFAP content can be better adapted to the local demands (Mugnaini, 1986), and the unnecessary expression of GFAP can be avoided. Horstmann (1954) and later Reichenbach et al. (1987) attributed the formation of astrocytes instead of tanycytes to the enlargement of brains and the thickening of the brain wall during both evolution and ontogenesis, which is especially conspicuous comparing sharks and rays (Horstmann, 1954) and following mammalian ontogenesis (Reichenbach et al., 1987). Not only astrocyte predominated brain areas are free of GFAP (exceptions are, e.g., euteleost tectum and cyprinid vagal lobe), and not each of these areas is free of GFAP (see, e.g., mammalian and avian pallidum); however, GFAP-free areas extend really only at astrocyte predominance (in mammals, birds, and batoids), and there is no vertebrate brain in which astrocytes predominate but there is no GFAP-free area.

Many of the GFAP-rich astrocytes belong to the “fibrous” type, and forms cordons and frames between and around the axon fascicles of white matter. In contrast, the astrocytes free of GFAP belong to the “protoplasmic” type characteristic of the gray matter (Connor and Berkowitz, 1985; Privat and Ratabul, 1986). Although GFAP is essential for process formation, even GFAP-deficient (GFAP–/–) astrocytes are capable of “stellation” (i.e., process formation) in the presence of neurons (Pekny et al., 1998a). Still, the processes are shorter and finer than those of the GFAP+/+ astrocytes. Although some populations of “protoplasmic” astrocytes do display GFAP immunoreactivity (Connor and Berkowitz, 1985; Walz, 2000), however, in these cells GFAP immunopositivity visualizes only the main branches, so the difference from the fibrous astrocytes is less conspicuous with GFAP immunostaining than following impregnations (Connor and Berkowitz, 1985). One could suppose that the “protoplasmic” type was formed during evolution as the GFAP production has been suspended (become facultative) completely or at least decreased below detectability.

A subdivision of those GFAP-immunonegative cells, which were formerly identified as “smooth protoplasmic astrocytes,” presents NG2 chondroitin sulfate proteoglycan on the surface (Levine and Card, 1987). These cells proved to be oligodendrocyte precursor cells (OPCs) as well, and have been considered a separate type, known as NG2 glia (Nishiyama et al., 1999) or “synantocytes.” Considering their large number persisting beyond the age of oligodendrocyte proliferation, it is likely that these cells also serve other roles (Kimelberg, 2004).

First of all, it is an evolutionary advantage to cease the unnecessary synthesis of a protein. However, it remains unclear whether the saving is significant enough to provide a considerable advantage during natural selection; therefore, other effects may also intoned to be considered. According to an early study by Pekny et al. (1995) in GFAP–/– rats, no abnormalities were observed in their behavior, motility, memory, and BBB function; although the lack of GFAP was not compensated for by the upregulation of other intermediate filament proteins, such as vimentin.

The glutamine synthetase activity is more intense in GFAP-free astrocytes. This enzyme neutralizes the toxic ammonium ions by binding them to glutamate (see, e.g., Norenberg and Martinez-Hernandez, 1979; Cooper, 2012). Glutamine levels appear to correlate inversely with GFAP expression. Astrocytes from GFAP–/– mice contained an elevated concentration of glutamine (Pekny et al., 1999a). The GFAP and glutamine synthetase contents are opposite in the “fibrous” and “protoplasmic” astrocytes (Patel et al., 1985; Didier et al., 1986; Ong et al., 1993; Sosunov et al., 2014), although Linser (1985) found glutamine synthetase in fibrous astrocytes as well.

Walz (2000) distinguishes “complex” and “passive astrocytes.” The “complex” ones, which contain no or hardly detectable GFAP, have voltage-gated K(+) and Na(+) channels. These channels can activate or inactivate astrocytes, helping to stabilize the ionic environment of neurons (McNeill et al., 2021). The disadvantageous extracellular K⁺ accumulation evoked by membrane depolarization was lower in the vicinity of GFAP–/– astrocytes than that of GFAP+/+ astrocytes (Anderova et al., 2001). In contrast, “passive” astrocytes, characterized by high GFAP content, lack voltage-gated ion channels (Walz, 2000).

Astrocyte processes contact synapses and may modulate synaptic function, synaptic efficacy, for example, long-term potentiation and depression. Deletion of GFAP increases the former one in the hippocampus (McCall et al., 1996), whereas it decreases the latter one in the cerebellum (Shibuki et al., 1996). Astrocyte processes lacking GFAP are remarkably mobile and therefore have significant effects on neuronal function (Theodosis et al., 2008).

The functional state of several neuron groups changes in parallel with the GFAP content of their astroglia. Retraction of astrocytic processes promotes interaction between neurons and the diffusion of transmitters, whereas expansion reduces neuronal excitability by wedging them apart (Theodosis et al., 2008; Wang and Purpura, 2018). Decrease or increase of GFAP content, redistribution of GFAP to or from the perikarya (Wang et al., 2017), and GFAP depolymerization or polymerization (Wang and Purpura, 2018) are found in the background of the retraction or expansion of processes.

The correlation between the expansion/retraction of astrocytic processes and the activity of the surrounded neurons is well-documented in the supraoptic nucleus in the case of lactation and suckling (Theodosis et al., 2008; Wang and Hatton, 2009; Wang and Hamilton, 2009; Wang et al., 2017), and dehidration/water overload (Wang and Hamilton, 2009; Wang and Purpura, 2018; Wang et al., 2021). Similar correlation was found in the arcuate and preoptic nuclei in the active phases of reproduction (Theodosis et al., 2008; Steinman et al., 2013), in the suprachiasmatic nucleus with the circadian rhythms (Fernandez-Galaz et al., 1999; Theodosis et al., 2008; Lawal et al., 2022), and in canaries in the song center in the period of song learning (Kafitz et al., 1999).

However, the causal relations remain to be elucidated: whether the alteration of astrocytes and GFAP staining is a primary phenomenon (Lawal et al., 2022), or only a consequence of the neural processes, a secondary reaction (Theodosis et al., 2008). GFAP expression is modulated by signaling molecules elicited by neuronal activity and hormones (Li et al., 2020). Neuronal activity, an increase in K⁺ or glutamate, may lead to an increase in GFAP fiber formation (Heimfarth et al., 2016; Wang et al., 2021). Afferent activity blockade evoked GFAP production in the chicken cochlear nucleus (Canady and Rubel, 1992). N-methyl-d-aspartate (NMDA) receptors, voltage-dependent calcium channels as well as metabotropic glutamate receptors induce hyperphosphorylation, thereby depolymerizing GFAP (Heimfarth et al., 2016; Wang et al., 2021). The depolymerized form is not detectable with immunohistochemical reaction (Hajós et al., 1993). However, I found no data indicating that neuronal activity in permanently GFAP-free areas (e.g., caudate-putamen complex in rats) surpasses that found in GFAP-rich areas (e.g., the globus pallidus).

Axon growth and regeneration are held to be inhibited by GFAP-containing glial processes, which appear in the post-lesion glial reaction, and demarcate the lesion (Reier, 1986; Hatten et al., 1991; Menet et al., 2001). Inactivation of the GFAP and vimentin genes, the absence of GFAP and vimentin prevents the post-lesion reactive hypertrophy of astrocytic processes and improves post-traumatic regeneration (Pekny et al., 1999b; Pekny and Pekna, 2004; Wilhelmsson et al., 2004; Holy and Pekny, 2015). It is worth noting that, in post-lesion glial reaction (but not in intact cerebral tissue), vimentin substitutes for GFAP in process formation; therefore, both are to be eliminated. The absence of GFAP and vimentin could induce a reorganization of the actin network, influenceing the composition of the extracellular matrix (ECM) and adhesion molecules, thus yielding an enhanced pearmeability of astrocytes to axon growth (Menet et al., 2001). However, it is usually not considered that the lack of regeneration in the mature central nervous system (CNS) is hardly a “mistake” of evolution. Possibly, adult regeneration could lead to maladaptive responses and mistaken connections (Aswendt et al., 2022). Note that post-lesion reactive GFAP expression also occurs in the areas free of GFAP immunoreactivity when they are intact (see, e.g., Figures 3c, d), and considerable regeneration does not occur either.

The presence or absence of GFAP, which may promote or inhibit, respectively, the synaptic plasticity, the rebuilding and re-arrangement of synapses to adapt to new situations (Missler et al., 1994; Finch, 2003; Wilhelmsson et al., 2004, 2019). The presence of astrocyte processes along the neuronal surface hampers the formation of synapses, whereas their absence allows it (Theodosis et al., 2008).

An increased neuronal differentiation was observed in vitro in the presence of GFAP–/–Vim–/– astrocytes, as more neurons survived and escaped apoptosis (Widestrand et al., 2007; Wilhelmsson et al., 2012). However, in vivo, in the adult mammalian brain, neither neurogenesis nor axon regeneration occurs in the GFAP-negative areas.

The ECM also influences synaptic plasticity (see, e.g., Song and Dityatev, 2018; McKeon and Silver, 1995). ECM accumulates around certain types of neurons, forming perineuronal nets (PNN, for reviews see Bosiacki et al., 2019; Duncan et al., 2019; Jakovljevic et al., 2021). PNN can be visualized by Wisteria floribunda agglutinin (WFA), which binds to N-acetylgalactosamine, a subunit of hyaluronan, the “backbone” of ECM; see, for example, Hilbig et al. (2000). The distribution of WFA has a similar pattern to that of GFAP in rats (Hilbig et al., 2000; Gáti et al., 2010); in other animals, the WFA distribution has not been studied enough.

The cytoskeletal system, including GFAP, is connected with the astrocyte membrane, for example, with the membrane-embedded glutamate-aspartate transporter (GLAST), the ezrin (Sullivan et al., 2007), the dystrophin-dystroglycan complex (Hendriksen et al., 2016; Li et al., 2020), the plectin of the plakin family, and the integrin (α6β4) (Potokar et al., 2020), which may form focal adhesions (Melrose, 2019). Hyaluronan synthetase and CD44 receptor bind the hyaluronan to the cell membrane (Roszkowska et al., 2016; Miyata and Kitagawa, 2017). In this way, GFAP can stabilize the ECM, and through the ECM, it also stabilizes the PNN. A stable PNN inhibits lateral diffusion (Valenzuela et al., 2014) and, therefore, stabilizes synapses and inhibits their plasticity (Bosiacki et al., 2019; Miyata and Kitagawa, 2017; Duncan et al., 2019; Yang, 2020; Jakovljevic et al., 2021). Taking these together, it is to be supposed that the absence of GFAP may promote plasticity and functional adaptation of astrocytes, and therefore that of neurons as well.

GFAP provides the maintenance of astrocyte processes and their resistance to mechanical stress. Fibrous astrocytes form “cordons” along the brain tracts; they are rich in both GFAP and CD44 (Sosunov et al., 2014). The hyaluronectin mesh, which surrounds the myelinated axons, is likely produced by GFAP-producing white matter astrocytes since hyaluronectin and GFAP showed similar distributions (Bignami and Dahl, 1986; Bignami et al., 1992). In this way, GFAP is necessary for the integrity of white matter architecture and the long-term maintenance of myelination (Liedtke et al., 1996; Li et al., 2020). In GFAP-free mice, the vulnerability of white matter increases to traumatic injury, ischemia (Galou et al., 1997; Nawashiro et al., 1998, 2000), and neurotoxicity (Otani et al., 2006).

In GFAP–/– Vim–/– mice, astrocytes do not form cytoplasmic intermediate filaments, the reactive gliosis is impaired (Pekny et al., 1999b; Menet et al., 2001; Pekny and Pekna, 2004; Wilhelmsson et al., 2004; Holy and Pekny, 2015), and the infarcts developing after ischemic strokes are larger (Li et al., 2008). The advantageous effect for axon regeneration was mentioned in the previous subchapter. However, neither of these effects manifests itself in the GFAP-free brain areas of “wild” animals, where reactive gliosis also develops following lesions. Note that in cultured astrocytes and post-lesion glial reaction, vimentin substitutes for GFAP in process formation; but it does not happen in the intact adult cerebral tissue (Pekny et al., 1999b; Menet et al., 2001).

According to Ding et al. (1998), Nawashiro et al. (2000), and Jurga et al. (2021), GFAP is necessary in astrocytes to help endothelial cells in the formation of the BBB. Although Pekny et al. (1995) found no extravasation of Evans blue in GFAP knockout mice, which implied the existence of an intact BBB, in a later study, however, they discovered that GFAP-positive astrocytes are necessary to restrict the passage for some hydrophilic molecules, for example, saccharose (Pekny et al., 1998b). According to our experience, perivascular astrocytes usually display GFAP immunopositivity, even in areas that are otherwise devoid of it (see, e.g., Figures 2c, 3b).

Extracellular water content, osmotic pressure, and sodium/potassium concentration are balanced by astrocytes (Anderova et al., 2001). GFAP is a key component in the cytoskeletal network implicated in cell volume regulation. Cytoskeleton makes astrocytes resistant to cell swelling in response to hypoosmotic stimulation (Ding et al., 1998; Wang et al., 2021), its impair increases their susceptibility to cytotoxic edema (Ding et al., 1998; Nawashiro et al., 2000; Anderova et al., 2001; Bélanger et al., 2002; Butterworth, 2010); its lack could facilitate cell swelling, and decrease the compensatory efflux or influx of ions and organic osmo-equivalents, mainly taurine (Ding et al., 1998), The interactions between GFAP and F-actin networks and membrane anchor proteins, for example, plectin, dystrophin squeezes water out and reduces volume (Wang and Hatton, 2009). The expression of AQP4 water-pore protein is in parallel with the levels and spatial distribution of GFAP (Wang and Hatton, 2009).

Astrocytes take up the excess of glutamate produced by neuronal activity (synaptic release), protecting the neurons against glutamate excitotoxicity. GFAP knockout mice exhibit a reduced glutamate clearance due to a decrease in glutamate transporters (Hughes et al., 2004; Sullivan et al., 2007; Butterworth, 2010; Kim et al., 2018). Glutamate is also the source of gamma-aminobutyric acid (GABA), which is produced by astrocytes and serves as a neurotransmitter for neurons (“GABA shunt,” Madsen et al., 2010).

The elimination of free oxidative radicals, which are formed during oxidative metabolism, is impaired. It appears that GFAP influences this process, although probably indirectly, by mediating other factors; the mechanisms involved remain to be determined (De Pablo et al., 2013).

Vesicle and enzyme trafficking are also impaired, as well as gliotransmitter release, since GFAP, as a component of the cytoskeletal system, has some role in transport functions. Several enzymes are associated with GFAP, e.g. vesicular GABA transporter (Potokar et al., 2007; Li et al., 2020; Wang et al., 2021). GFAP serves as a platform for interactions between different signals, and between enzymes and their substrates.

Elevated factors of plasticity may also render them vulnerable to abnormal structural changes, as seen in psychiatric diseases (García-Cabezas et al., 2017). Decreased GFAP was found in association with schizophrenia models (Kim et al., 2018; Li et al., 2020).