Address correspondence and reprint requests to Carmem Gottfried, Departamento de Bioquímica, Instituto de Ciências Básicas da Saúde, Universidade Federal do Rio Grande do Sul, Rua Ramiro Barcelos 2600 anexo, 90035, 003, Porto Alegre, RS, Brazil. E-mail: email@example.com
In the past two decades, there has been an explosion of research on the role of neuroglial interactions in the control of brain homeostasis in both physiological and pathological conditions. Astrocytes, a subtype of glia in the central nervous system, are dynamic signaling elements that regulate neurogenesis and development of brain circuits, displaying intimate dynamic relationships with neurons, especially at synaptic sites where they functionally integrate the tripartite synapse. When astrocytes are isolated from the brain and maintained in culture, they exhibit a polygonal shape unlike their precursors in vivo. However, cultured astrocytes can be induced to undergo morphological plasticity leading to process formation, either by interaction with neurons or by the influence of pharmacological agents. This review highlights studies on the molecular mechanisms underlying morphological plasticity in astrocyte cultures and intact brain tissue, both in situ and in vivo.
Richard Rodnight recently passed away (See (Dunkley 2012) for an Obituary). In the last years of his life, Richard wrote a complete draft of this review, but because of ill health was unable to finalize its submission. Richard gave the review to me on the understanding that I would attempt to have it published. Carmem Gottfried, Richard's former student and long-term colleague in the field of astrocyte morphology, agreed to update the review prior to its submission.
Peter R. Dunkley
Knowledge of the functions of the astroglia or astrocytes has undergone a revolution in the past two decades. Formerly considered to have a limited role in supporting neurons, they are now recognized as key players in virtually all aspects of brain function. Many research groups during the 1990s revealed the existence of bidirectional communication between neurons and astrocytes and proposed the term ‘tripartite synapse’ to indicate the astrocyte as the third element of the synapse (reviewed by (Araque et al. 1999; Halassa et al. 2009; Santello et al. 2012). This vision is supported by the fact that astrocytes are capable of releasing neuroactive substances termed gliotransmitters (e.g., ATP, glutamate, GABA, D-serine, and prostaglandin-E2). The release of gliotransmitters occurs in the millisecond time scale, integrating and processing synaptic information, which allows these cells to modulate synaptic transmission and plasticity (Navarrete et al. 2012). The revolution in astroglial plasticity and function has generated a number of excellent reviews and comments (Verkhratsky et al. 2002; Magistretti 2006; Barres 2008; Wang and Bordey 2008; Halassa et al. 2009; Eroglu and Barres 2010; Matyash and Kettenmann 2010; Kettenmann and Ransom 2012; Navarrete et al. 2012; Parpura et al. 2012).
One of the striking features of astroglial cells is their morphological profile. In vivo astrocytes are classified as either protoplasmic or fibrous and the number and size of astrocytes changes among species and increases with brain size (Eroglu and Barres 2010). The glial fibrillary acidic protein (GFAP), a class-III intermediate filament and the S100B protein are both astrocyte-specific markers. Based on astrocyte labeling methods comprising GFAP and S100B immunostaining and GFAP-GFP expressing mice, astrocytes have been referred to as tanycytes, ‘radial’ cells, Bergmann glia, protoplasmic astrocytes, fibrous astrocytes, velate glia, marginal glia, perivascular glia, and ependymal glia (Matyash and Kettenmann 2010).
The astrocyte intermediate filaments are modulated by GFAP phosphorylation/dephosphorylation state (Goncalves et al. 1994; Tasca et al. 1995; Wofchuk and Rodnight 1995; Leal et al. 1997; Lenz et al. 1997; Rodnight et al. 1997; Vinade et al. 1997; Gottfried et al. 1999) and this induces changes in shape. Specialized astrocytes have been identified by their complex morphological profile with multiple processes in particular brain regions, including Bergman, radial glia, and ventral glia limitans, subjacent to the hypothalamic supraoptic nucleus (SON). All astrocytes have the capacity for morphological plasticity and it is now known that plasticity plays a critical role in these cells in vivo.
The hypothalamic SON has been studied as a suitable model of CNS plasticity (Salm and Hawrylak 2004). In this region, morphological changes in astrocytes undergo a reversible thinning upon activation of the magnocellular neuroendocrine cells. Glutamatergic inputs to SON neurones may influence the transmission of peripheral information, such as in response to suckling during lactation and dehydration (Oliet and Piet 2004). In particular, changes in astrocytic environment of neurones appear to modify both synaptic and extrasynaptic transmission mediated by glutamate.
The mechanisms whereby morphological plasticity occurs in vivo as well as the corresponding functional consequences are far from being understood. When astrocytes are isolated from the brain and maintained in culture, they exhibit a polygonal shape unlike their precursors in vivo. However, cultured astrocytes can be induced to undergo morphological plasticity leading to process formation (Gottfried et al. 2003), either by interaction with neurons or by the influence of pharmacological agents. The molecular mechanisms underlying this morphological plasticity have been investigated in some detail and they are likely to contribute to part of the morphological plasticity seen in vivo.
The main purpose of this article is to review the molecular mechanisms underlying morphological plasticity in astrocyte cultures; then comparing the morphological plasticity of cultured astrocytes with the plasticity of astroglia seen in the intact brain tissue, either in situ preparations in vitro or in vivo.
Morphological plasticity of cultured astrocytes: stellation versus growth
Morphological plasticity is a striking feature of cultured astrocytes. In contrast to their process-bearing cells in vivo or in freshly derived in vitro cultures or preparations, mature monolayers of these cells from neonatal rodent brain exhibit few processes and present a polygonal fibroblast-like shape, which may be changed to a full process-bearing profile by addition of neurons or a variety of pharmacological stimuli. Depending on the nature of the stimulus, process formation occurs either by a mechanism known as stellation or by process growth (The term ‘stellation’ is used rather loosely in the literature and is sometimes applied to examples of process growth).
Stellation proceeds by diverse complex mechanisms and is typically complete in 1–3 h; it does not involve growth of new membrane and its physiological relevance is uncertain. Stellation is the most extensively studied aspect of morphological change in culture and illustrates dramatically the complex intracellular machinery that regulates cell shape. Most of the stimuli that induce this response are pharmacological and up to 20 examples exist in the literature, where information is available distinct intracellular pathways are involved in each stimulus (see Table 1 for examples of different stellation stimulators). Herein, we focus in the intracellular pathways involved in stellation induced by agents that increase intracellular cAMP (cAMPi) (Pinto et al. 2000) and stellation induced by exposure to bicarbonate-free saline (Cechin et al. 2002; Gottfried et al. 2003). These two examples are chosen because the stimuli are more physiological than pharmacological in nature and because they are recognized targets for morphological change in vivo.
Table 1. Astrocyte stellation can be induced by a range of factors in primary cell cultures
Stellation was most robust in cerebellar and brainstem astrocytes
In contrast, process growth in culture is relatively slow and once initiated continues over a period of days. Growth is defined as the extension of long processes beyond the original boundaries of the cell and requires protein synthesis. It may be initiated in astrocyte cultures by several growth factors, by thyroid hormones and by co-culturing with neurons (Matsutani and Yamamoto 1998). Furthermore, it mimics a physiological phenomenon, comparable to the growth of processes during development.
General features of process formation by stellation
Stellation in astrocytes occurs as a result of morphological change known as cavitation in which the cell membrane retracts and the cell body rounds up with the formation of thin branching processes, an event accompanied by depolymerization of the actin cytoskeleton (Fitch et al. 1999). Cavitation is equivalent to cell rounding in fibroblasts where processes are not formed when the cell body contracts. The area occupied by a stellate cell is approximately the same as the original polygonal cell, though in practice the stellate processes formed by cavitation occasionally appear longer, possibly because of the stretching of invaginations present in the rounded cell.
Stimuli resulting in stellation are modulated by the presence or absence of serum in the medium and serum withdrawal by itself eventually results in process formation (Safavi-Abbasi et al. 2001). Therefore, serum conditions the morphological response to stimuli: some stimuli only result in stellation in the absence of serum, while others are effective even in the presence of serum. Also, there are regional differences in cAMP-induced stellation and process formation in astrocytes (Won and Oh 2000). In this study, primary astrocyte cultures were prepared from six different regions of neonatal rat brains, including cerebral cortex, hippocampus, brainstem, midbrain, cerebellum, and hypothalamus. They found that 250 μM 8-CPT-cAMP produced a maximum effect causing 95% stellation in all regional astrocytes except hypothalamic astrocytes (56% stellation). At lower cAMP concentrations, cell stellation most effectively occurred in cerebellar astrocytes. Also, the effect of glutamate on cAMP-induced astrocyte stellation was investigated. Interestingly, glutamate was able to block cAMP-induced astrocyte stellation most effectively in cortical and hippocampal astrocytes but not in cerebellar, midbrain, brainstem, and hypothalamic astrocytes. These results demonstrate that primary astrocytes display distinct morphologic features even after isolation from their original tissue.
The intracellular machinery that regulates astrocyte shape consists of pathways controlled by several small GTPases, with RhoA together with Rac1 and Cdc42 being the most relevant (Burridge and Doughman 2006) (see Box 1 for background information to the RhoA pathway).
The RhoA pathway
RhoA belongs to a family of small GTPases that act as binary molecular switches cycling between an active GTP-bound and an inactive GDP-bound state, with 23 mammalian genes (Elias and Klimes 2012; Xing et al. 2012). The signaling functions of the family members RhoA, Rac and Cdc42 have been most extensively researched. In the GTP-bound state, the switch is in the ‘on’ position and signaling to downstream effectors occurs; reversion to the ‘off’ position occurs by hydrolysis of GTP to GDP. Cycling is regulated by exchange factors (GEFs), which catalyze the exchange of GDP for GTP and GTPase activating factors (GAPs) that in turn act to increase the rate of intrinsic GTPase activity. In its GTP-bound state, RhoA is inserted in the plasma membrane where it can signal to its effectors. Cycling between membranes and cytosol is controlled by guanine nucleotide dissociation inhibitors (GDIs), which can form stable complexes with both the GDP- and GTP-bound forms of Rho (Elias and Klimes 2012). The RhoA effectors involved in the assembly of the actin cytoskeleton include a family of protein kinases collectively referred here as “Rho-kinase”. Downstream targets of Rho-kinase include myosin light chains (MLCs), MLC phosphatase, the LIM-kinases, and the sodium/hydrogen exchange enzyme NHE1 (Elias and Klimes 2012).
Tonic activation of the RhoA pathway maintains the polygonal shape of cultured astrocytes. In this condition, the actin cytoskeleton is organized in the form of so-called ‘stress fibers’ (Abe and Misawa 2003). Cumulated evidence confirms that inactivation of the pathway leads to stellation; for example, phorbol, 12-myristate, 13-acetate, ADP-ribosylation of Rho-GTPase with the C3 transferase from Clostridium botulinum results in stellation. Moreover, astrocytes expressing constitutively active RhoA (V14RhoA) do not form processes when exposed to agents that induce stellation (Davis-Cox et al. 1994; Ramakers and Moolenaar 1998; Abe and Misawa 2003; Gottfried et al. 2003). Downstream of RhoA, the pharmacological inhibition of Rho-kinase with Y-27632 induces stellation in cortical astrocytes. This effect was prevented by the microtubule inhibitor colchicine, indicating that the response requires the rearrangement of cytoskeletal elements including microfilaments and microtubules (Abe and Misawa 2003). Lysophosphatidic acid (LPA), thrombin, and endothelin prevent and reverse stellation through mechanisms dependent on RhoA signaling (Ramakers and Moolenaar 1998; Bustelo et al. 2007). In fibroblasts, tyrosine phosphorylation downstream Gα13 is involved in the regulation of the actin cytoskeleton by LPA (Nobes and Hall 1995). Also, inhibition of tyrosine kinase activity results in astrocyte stellation (Padmanabhan et al. 1999; Cechin et al. 2002). Independently of the RhoA pathway, stellation occurs when filamentous actin is disrupted by cytochalasins (Baorto et al. 1992).
Stellation induced by increased intracellular cAMP
It is well known that increased cAMPi levels in cultured astrocytes can convert flat polygonal shaped cells into process-bearing (stellate) cells. Astrocyte stellation was first observed as a result of treatments that increase cAMPi levels, including forskolin, dibutyryl cAMP (dBcAMP) (Shapiro 1973; Moonen et al. 1975; Abe and Saito 1997), or ß-adrenergic receptor stimulation (Junker et al. 2002).
The response is not unique to astrocytes as several cells of epithelial lineage form processes after treatment with agents that increase cAMPi (Goldman and Abramson 1990). Stellation is initiated around 15 min after treatment and is complete in 45 min in the absence of serum, and even longer when serum is in the culture medium.
The loss of actin stress fibers, resulting in retraction of the cell membrane and the formation of processes, depends on microtubule reassembly (Ramakers and Moolenaar 1998; Holtje et al. 2005). In stellate cells, actin appears concentrated in a concentric ring in the contracted cell body and at the tips of processes (Baorto et al. 1992). Immunostaining for tyrosine phosphorylated proteins and vinculin shows that the morphological change is accompanied by the loss of focal adhesion in the periphery of the cell body and their appearance in the tips of the processes (Padmanabhan et al. 1999; Perez et al. 2005). This suggests that as the cell membrane retracts, the formation of processes is initiated at the sites of focal adhesion and may explain why cAMP-induced stellation requires adhesion to a substrate (Padmanabhan et al. 1999).
Consideration of mechanisms
In cerebellar astrocytes, stellation induced by dBcAMP is associated with a 70% decrease in the phosphorylation of myosin light chains (MLCs) and 57% decrease in that of an actin depolymerizing factor (Baorto et al. 1992; Padmanabhan and Shelanski 1998). Both these decreases would be expected to result in at least a partial disassembly of the actin cytoskeleton and process formation. The decrease in MLC phosphorylation may be related to the down-regulation of the Rho pathway and/or inhibition of the activity of MLC kinase (MLCK) through its phosphorylation by cAMP-dependent protein kinase (PKA), as has been demonstrated in fibroblasts (Lamb et al. 1988) and in astrocytes (Baorto et al. 1992). Interestingly, in SH-EP cells, a neuroblastoma-derived cell line that exhibits a stellation response to increased cAMPi very similar to astrocytes, the small GTPase Cdc42, as well as RhoA, has been shown to be involved in the regulation of the actin cytoskeleton through MLC phosphorylation (Dong et al. 2002). Thus, injection of constitutively active Cdc42V12 into cells, prior to the induction of cAMPi increase by forskolin, blocked stellation. This action of Cdc42V12 does not involve the RhoA pathway but rather the Cdc42 effector myotonic dystrophy kinase-related Cdc42-binding kinase α (myotonic dystrophy kinase-related Cdc42-binding kinase). Myotonic dystrophy kinase-related Cdc42-binding kinase α is known to phosphorylate and to inhibit MLC phosphatase and thus increase the phosphorylation state of MLCs (Tan et al. 2011). The effect of Cdc42V12 can be blocked by the pharmacological inhibition of MLCK and it was concluded that the action of activated GTPase in protecting against cAMP-induced stellation is primarily related to its ability to maintain the phosphorylation state of MLCs (Dong et al. 2002).
These observations pointed to the dephosphorylation of MLCs and actin depolymerizing factor as control points in cAMP-induced stellation, possibly in the case of MLCs related to inhibition of MLCK by PKA-dependent phosphorylation. However, recent work questions a role for PKA, as effective inhibition of the kinase (also in cerebellar astrocytes) with the specific blocker Rp-cAMP, did not prevent stellation induced by two agents known to increase cAMPi: forskolin or pituitary adenylate cyclase-activating polypeptide) (Perez et al. 2005). This observation suggests that the dephosphorylation of MLCs (Baorto et al. 1992) was because of cAMP-induced (and PKA-independent) down-regulation of the Rho pathway occurring upstream of RhoA by a pathway not yet characterized, rather than inhibition of MLCK by PKA phosphorylation. In addition, PKA could directly phosphorylate Limk1, thus enhancing the phosphorylation of cofilin. This indicates that PKA is crucial to cell shape and migration, through the modulation of LIM kinase (Nadella et al. 2009).
Despite the negative results by Perez et al. (2005) in astrocytes, evidence for a role for cAMP-induced PKA activity in down-regulating the Rho pathway and leading to morphological change has been obtained in studies with SH-EP cells, where phosphorylation by PKA of RhoA on Ser188 was shown to inhibit its coupling to Rho-kinase (Ellerbroek et al. 2003). The significance of this PKA target was demonstrated in experiments in which Ser188 of constitutively active RhoA (RhoV14) was mutated to alanine (RhoV14A188) and shown by microinjection to block cAMP-induced stellation in these cells. However, exactly why phosphorylation of Rho on Ser188 de-couples it from Rho-kinase and inhibits its ability to maintain an intact actin cytoskeleton remains unclear. Clues as to a possible mechanism come from an earlier study in cytotoxic lymphocytes (Lang et al. 1996). These authors were the first to show, using recombinant RhoA, that Ser188 in the C-terminal region of the protein is phosphorylated by PKA. Exposure of cytotoxic lymphocytes to dBcAMP modified cell shape and resulted in translocation of membrane-associated RhoAGTP to the cytosol. The translocation was mediated by RhoGDI (see Box 1), which binds preferentially to phosphorylated RhoA-GTP (Ellerbroek et al. 2003). Taken together with the study in SH-EP cells, these results suggest that increased cAMPi, apparently acting through PKA, does not affect GTP/GDP cycling in RhoA. This may be the case in astrocytes as forskolin treatment had no effect on GTP loading on RhoA in these cells over a period of 1 h (S.Cechin and R.Rodnight, unpublished results).
However, in the light of the results of Perez et al. (2005), reconsideration of the role of PKA is needed. One possibility is that, at least in cerebellar astrocytes, PKA activity is involved in the maintenance rather than the initiation of morphological changes. This is suggested by the competitive nature of the Rp-cAMP inhibitor and the persistence of forskolin-induced stellation after the termination of the transient inhibition of PI3-K activity toward Akt. Accordingly, the accumulation of cAMPi after forskolin may eventually displace the inhibitor and reactivate the enzyme. Thus, phosphorylation of Ser188 by PKA on RhoA, which leads to shape change in cytotoxic lymphocytes and SH-EP cells, in cerebellar astrocytes may be insufficient to down-regulate the Rho pathway on its own, but may contribute to the maintenance of the shape change initiated by PI3-K.
Stellation induced by saline media lacking bicarbonate
The stellation response described in this section results from a negative stimulus, namely the absence of bicarbonate buffering in the medium. The response was discovered accidentally by the exchange of the bicarbonate/CO2-buffered Dulbecco's modified Eagle medium culture medium for a saline medium buffered by N-[2-hydroxyethyl]piperazine-N’-[2-ethanesulfonic acid] (HEPES-saline) (Cechin et al. 2002). The resulting stellation was rapid with characteristics identical to those induced by increasing cAMPi. Confirming the importance of bicarbonate buffering, Stellation also resulted with salines buffered by PIPES or phosphate, but did not occur in saline buffered with bicarbonate/CO2.
As cytoplasmic acidification occurs when astrocytes are exposed to HEPES-buffered media (Deitmer and Rose 1996), an association between intracellular pH (pHi) and cellular shape was sought in the study of Cechin et al. (2002). In culture medium, astrocyte pHi was 7.26 ± 0.04. HEPES-induced stellation was associated with a transient acidification in 15 min to 6.81 ± 0.06 followed by a recovery to a plateau value of 7.13 ± 0.04 at 60 min, which was significantly less than the value in culture medium This partial recovery of pHi was prevented by ethylisopropylamiloride (EIPA), a specific inhibitor of the ubiquitous Na+/H+ exchange enzyme, subtype NHE1 (Cechin et al. 2002). The activity of this enzyme is minimal at neutral pH but increases proportionally as the pHi falls (Bevensee et al. 1997). Therefore, the transient phase of acidification and subsequent recovery suggests that as the protons produced by metabolism accumulated in the absence of bicarbonate/CO2 buffering, the exchanger was gradually turned on and after 10–15 min it was able partially to reverse the decline in pHi. Stellation in HEPES-saline was both prevented and reversed when complete, by LPA, but only in the presence of Ca2+. Other treatments that prevent HEPES-induced stellation include inhibition of protein tyrosine dephosphorylation by orthovanadate and activation of NHE1 with 75 mM LiCl (Kobayashi et al. 2000). Although these treatments were associated with varying degrees of transient acidification, in all cases the final pHi reached after 60 min was significantly higher than the pHi reached in HEPES-saline and equal or higher than the pHi of 7.26 in culture medium (Table 2). Acidification did not occur when astrocytes were transferred to saline buffered with bicarbonate/CO2, maintaining the polygonal shape. However, treatment of astrocytes in bicarbonate/CO2-saline with EIPA or induction of a mild acid load with a low concentration (1 mM) of NH4Cl resulted in rapid stellation and acidification to 6.8–6.9 in 5 min.
Table 2. pHi values in HEPES-saline in relation to final astrocyte cell shape
In contrast to cultures maintained on polylysine, astrocytes cultured on physiological extracellular matrices (ECM), presented a different picture (Gottfried et al. 2003). First, stellation in HEPES-saline did not occur on matrices of type I or type IV collagens, provided Ca2+ was present. Secondly, the intracellular acidification induced by HEPES-saline observed in astrocytes cultured on polylysine was abolished in cells cultured on the collagens, although inhibition of NHE1 with EIPA in these cells resulted in acidification. Astrocytes cultured on type I collagen also exhibited a higher constant pHi (7.33 ± 0.03) as compared with 7.26 ± 0.02 for polylysine (p < 0.001).
Two observations pointed to the involvement of integrin (Hynes 2002) and focal adhesions in the effects of collagens. First, treatment of astrocytes on collagens with a blocking antibody to ß1 integrin subunit abolished protection against HEPES-induced stellation as shown in Fig. 1. Secondly, compared with polylysine, astrocytes cultured on collagens expressed increased contents of phosphotyrosine proteins, the focal adhesion proteins vinculin and paxillin and the ß1 integrin subunit (Gottfried et al. 2003). Similar results have been reported in non-cerebral cells (Schwartz and Shattil 2000).
Consideration of mechanisms
With astrocytes cultured on polylysine, the association between shape change and intracellular acidification in cells exposed to HEPES-saline suggests that in some circumstances a transient decrease in pHi is sufficient to trigger the disruption of the actin cytoskeleton. This is supported by the observation showing that in bicarbonate-saline a mild acid load induced by a low concentration of NH4Cl was associated with stellation. However, several observations show that acidification is not a necessary factor in inducing stellation: thus, in bicarbonate-saline lacking Ca2+ stellation occurred in the absence of acidification (Cechin et al. 2002; Gottfried et al. 2003).
The enzymatic targets of the extra protons that accumulate following incubation in HEPES-saline are presumably related to the actin cytoskeleton and the RhoA pathway. A likely target is NHE1 as similarly its ion exchange function, this enzyme regulates astrocyte cell shape and motility by virtue of its localization with focal adhesion sites, actin cytoskeleton and integrins. These regulatory functions are both influenced by modifications in the molecular state of the C-terminal domain, consisting of protein phosphorylation and the binding of regulatory proteins and factors. For example, exposure of cells to serum and growth factors such as LPA results in phosphorylation of C-terminal serine residues in NHE1 by Rho-kinase and p90RSK (Takahashi et al. 1999), increased ion exchange activity at neutral pHi and stabilization of morphological profile. Moreover, a relationship between ion exchange activity and cell shape is supported by observations where NHE1 disrupt the actin cytoskeleton, for example, inhibition of ion exchange by EIPA in astrocytes cultured in bicarbonate-saline resulted in stellation (Cechin et al. 2002).
Taking into account the above observations, a plausible hypothesis to explain the results with HEPES-saline suggests that acidification down-regulates the phosphorylation state and/or protein binding state of the C-terminal region of NHE1, thus destabilizing astrocyte cell shape while leaving basal ion exchange activity intact. This would account for effect of LPA, orthovanadate and direct stimulation of NHE1 by Li+ in increasing pHi and stabilizing morphological profile, as shown in Table 2.
The stabilization of shape in astrocytes cultured on collagens in HEPES-saline containing Ca2+ was an unexpected result, as astrocytes in culture secrete extracellular proteins including fibronectin (Price and Hynes 1985) and type I collagen (Heck et al. 2007). It may be assumed therefore that mature astrocyte cultured on polylysine secrete and deposit a physiological matrix consisting of these and other proteins, although this ECM secretome is apparently either quantitatively or qualitatively insufficient to maintain polygonal shape or pHi homeostasis in bicarbonate-free media. Only when an artificial collagen matrix was supplied did the cells remain polygonal and express focal adhesion proteins, increased levels of phosphotyrosine proteins and a higher constant pHi when exposed to HEPES-saline. The integrin involvement (Box 2) shows that the cells adhere to the matrix by engagement with focal adhesions and the actin cytoskeleton (Schwartz 2001).
The integrins belong to a family of cell surface receptors that link the ECM to focal adhesions and the actin cytoskeleton (Hynes 2002; Ginsberg et al. 2005). Integrins are heterodimers composed of α and ß subunits in various combinations. Ten of these dimers use ß1 as a common subunit and blocking antibodies to this subunit are used to probe the involvement of integrins in adhesion. The major integrin-type collagen receptors are α1ß1 and α2ß1, but only α1ß1 is expressed by astrocytes (Tawil et al. 1994). Therefore, it is likely that a1ß1 is involved in the astrocyte response to HEPES-saline in cells cultured on collagen.
It is noteworthy that while type I collagen represents a minor component of the brain ECM in vivo (Ruoslahti 1996), type IV collagen is abundantly seen in the basal lamina of the endothelial cells that surround the cerebral capillaries where, together with laminin, forms a matrix for the astrocyte end-feet (Kacem et al. 1998). Taking into account the pervasive nature of the cerebral microvasculature, type IV collagen and laminin of the endothelial basal lamina may comprise a quantitatively significant proportion of the extracellular environment encountered by astrocytes.
Process growth in culture induced by growth factors
Process growth in cultured astrocytes can be induced by exposure to epithelial growth factor, nerve growth factor, and fibroblast growth factors (FGF). Most work has been carried out concerning the signal transduction pathways involved in the action of basic FGF (bFGF) as described below.
Process growth induced by bFGF
Astrocytes express FGF receptors and mRNA for both acidic and basic FGF (aFGF or FGF-1 and bFGF or FGF-2, as recently reviewed (Kettenmann and Ransom 2012). In an early study, both aFGF and bFGF applied to serum-free cortical astrocyte cultures promoted extensive process growth over a period of 6–21 days (Perraud et al. 1988). With both factors, very long processes were formed and membrane retraction appeared a little more marked in the case of bFGF. Ras family GTPases were shown to be involved in the action of bFGF in process growth in cultured rat polygonal hippocampal astrocytes (Kalman et al. 1999). Applied to these cells bFGF caused process growth after 1–2 days, extending significantly beyond the boundaries of the polygonal control cells and was associated with depolymerization of actin stress fibers in the soma and membrane retraction.
Retroviral infection of astrocytes with constitutively active RhoA (RhoA-V14) or constitutively active Rac1 (Rac1-V12) abolished or reversed the effect of bFGF; in contrast, infection with dominant negative RhoA (RhoA-N19) or dominant negative Rac1 (Rac1-N7) mimicked the effect of bFGF (Kalman et al. 1999): thus, 2–3 days after infection 80% of cells infected with Rac1-N17 had contracted and extended long processes. Infection with a dominant negative form of H-Ras had no effect on astrocyte shape, but prevented the effect of bFGF. However, dominant negative H-Ras did not block morphological changes induced by dominant negative Rac1 or RhoA. These results suggest that H-Ras mediates the effects of bFGF upstream of RhoA and Rac1 and from experiments in which various combinations of mutant Rac and Rho were examined it was concluded that Rac1 is upstream to RhoA. This cascade in which bFGF activates H-Ras, which in turn inhibits Rac1 and RhoA is at variance with the results from other cell types where H-Ras activates Rac1 (Ridley and Hall 1992; Rodriguez-Viciana et al. 1997).
Subsequent work demonstrated the involvement of the ERK and p38 MAPK pathways in the effect of bFGF upon process growth in mouse cortical astrocytes (Heffron and Mandell 2005). Exposure to bFGF led to a transient phosphorylation of ERK, and inhibition of MEK with UO126 and consequent down-regulation of ERK led to complete inhibition of bFGF-induced process growth. In contrast, inhibition of p38 MAPK with SB202190 stimulated basal process growth and greatly potentiated the effect of bFGF. The upstream pathway from p38 is not known but may be a Rac GTPase. A hypothetical and simplified scheme to illustrate these results is shown in Fig. 2.
Astroglial shape and plasticity in the intact brain
Astroglia in the brain are traditionally divided into two major morphological types named protoplasmic and fibrous astrocytes. Protoplasmic astrocytes are found in the cell-rich areas of the brain or gray matter and fibrous astrocytes are typical of white matter areas (Kettenmann and Ransom 2012). Included here with the protoplasmic astrocytes are two examples of specialized glia: the Bergmann glia of the cerebellum and radial glia in the ventral hypothalamus. Although morphological plasticity in the intact brain has been described in fibrous astrocytes, it has been most studied in protoplasmic astrocytes.
Compared with cultured astrocytes, protoplasmic astrocytes in in situ preparations such as acute and cultured slices and in vivo exist in 3-dimensional space and present an astonishing complex morphological profile. They are generally identified by immunostaining with anti-GFAP antibodies. GFAP staining of astrocytes in situ clearly reveals the primary stem processes, typically 5–10 in number in rodents and many more in humans (Oberheim et al. 2009). Beyond the main GFAP-positive processes, techniques such as high voltage electron microscopy and diffusion of dyes injected into the perikaryon demonstrate the existence of an extensive so-called ‘spongiform’ profile, consisting of a mass of very fine processes in which GFAP is often below the level of detection by conventional immunohistochemistry (Bushong et al. 2004). In the same study, intracellular injection with fluorescent tracers combined with immunohistochemistry for GFAP show that this protein delineates only approximately 15% of the total volume of the astrocyte. It indicates that the analyses of astrocyte shape and processes based on immunohistochemistry for GFAP are not conclusive considering the interaction of processes among cells.
Dye-filled astrocytes observed by optical imaging techniques and analyzed by 3D reconstruction programs show that the fine processes arise as lateral filopodium-like appendages to highly ramified branches of the stem processes (Wilhelmsson et al. 2006). Moreover, although the stem processes ramify into finer processes, their terminals do not give rise to ‘spongiform’ profile, but rather form endfeet in the cerebral vasculature and pia (Kacem et al. 1998). Serial electronmicroscopy has shown that the terminals of the filopodium-like appendages ensheath dendritic spines and other synaptic structures, hence forming the morphological basis of the reciprocal communication between astrocytes and neurons (Haydon et al. 2009). Ensheathment of spines varies considerably in different brain regions, from around 57% in the hippocampus (Ventura and Harris 1999) to 100% in the Bergmann glia of the cerebellum (Lippman et al. 2008).
Protoplasmic astrocytes also occupy separate anatomical domains in the neuropil with only limited interdigitation of the fine ‘spongiform’ processes amounting to 4–5% of the astrocyte volume of approximately 6.6 × 104 μm3 (Ogata and Kosaka 2002; Bushong et al. 2004). Interestingly, it has been estimated that the domain of an individual rodent astrocyte invests between 90 000 and 190 000 synapses (Bushong et al. 2002; Halassa et al. 2007).
In contrast, common astrocyte cultures exist in 2-dimensional monolayer space, lack the environment of the intact tissue and consequently possess a much simpler shape (Ramakers and Moolenaar 1998; Cechin et al. 2002). There is, however, morphological evidence of their derivation from protoplasmic astrocytes, as shown by observations indicating that the stellation response in cultured astrocytes is more complex than that revealed by phase contrast microscopy or immunostaining for GFAP (Gottfried et al. 2003; Holtje et al. 2005).
Dynamic plasticity of astroglial processes observed in situ and in vivo
Normal morphological plasticity of astrocytes in the brain has been mainly researched in perisynaptic regions where motile astroglial processes ensheath and contact synaptic structures (Theodosis et al. 2008). The following brief account gives the background necessary to discuss the contribution of astroglial plasticity in cell culture to our understanding regarding the molecular basis of plasticity in vivo.
Astroglial processes ensheathing dendritic spines
Astrocytes are able to extend and retract processes near dendritic spines over the course of minutes. Time-lapse confocal microscopy has been employed to observe process motility in organotypic slice cultures and in acutely isolated slices prepared from hippocampal, brainstem, or cerebellar tissue. In the central region of these in situ preparations astrocytes retain their complex 3-dimensional shape and exhibit motile filopodium-like and lamellipodium-like processes in close spatial relation to neuronal dendritic spines at excitatory synapses. These perisynaptic processes are specialized and stain to anti-ezrin (Derouiche and Frotscher 2001), an actin-binding protein, contributing to the analysis of the processes. The positioning of astrocytic processes near dendritic spines allows vesicular-mediated release of molecules, including D-serine in response to synaptic activity. D-serine can directly regulate channel function and modulate long-term potentiation, thus affecting physiological properties of synapses (Yang et al. 2003).
Process motility is characterized by extension and retraction of structures over a time period of minutes (Haber et al. 2006). However, the signal transduction mechanisms involved in directional process motility toward spines are poorly understood. Presumably, chemoattractants released by spines are involved and in excitatory synapses of the hippocampus one of these may be glutamate. Transfection of astrocytes with an inactive mutant of the small GTPase Rac1 (RacV17) decreased both the length and the motility of processes in the vicinity of spines in the hippocampus (Nishida and Okabe 2007) and length, but not motility, of processes of the Bergmann glia in the cerebellum (Lippman et al. 2008). In both conditions, the treatment was associated with modifications in dendritic spine morphological profile.
Considerably, more information is available on signal transduction mechanisms involved in the astroglial regulation of the shape of dendritic spines (Hotulainen et al. 2009). These structures, first observed by Santiago Ramon y Cajal, undergo rapid dynamic changes in shape dependent on actin dynamics regulated by small GTPases. The modulation of spine morphological profile by astroglial processes is developmentally regulated and involves transient interaction between the membrane-bound ligand ephrin-A3 situated in the processes and the EphA4 receptor tyrosine kinase located on spines (Nishida and Okabe 2007).
In excitatory synapses, the downstream consequences of this interaction include inhibition of complex integrin signaling pathways (Warren et al. 2012), the recruitment and activation of phospholipase C-γ1, resulting in the release of a pool of the actin depolymerizing factor cofilin (see Box 1) associated with the membrane (Zhou et al. 2007) and the activation of the proline-directed kinase Cdk5 and ephexin1 (Fu et al. 2007). Among the morphological consequences of the engagement of EphA4 receptors with astrocytic ephrin ligands are spine retraction and stabilization [see (Nishida and Okabe 2007) for a detailed appraisal], with consequent modulation of functional synaptic plasticity (Murai and Pasquale 2004; Nestor et al. 2007).
Regulating spine ensheathment of excitatory synapses by astroglial processes serves to regulate perisynaptic glutamate levels through the activity of the glutamate transporters GLT-1 and GLAST present in the membranes of the processes (Danbolt 2001). These transporters are 3-fold more expressed in astroglial processes enveloping the post-synaptic dendritic spines compared with pre-synaptic terminals (Lehre and Rusakov 2002).
A further point concerns the relation between glutamate uptake and the regulation of spine shape. Diffusion of glutamate from the synaptic cleft requires extracellular space between glial and neuronal membranes, whereas regulation of dendritic spine morphological profile involves transient adhesion between processes and spines (Rusakov et al. 2011). If these two functions are present in the same process, possibly the transient nature of ephrin-EphA4 association is sufficient to allow glutamate to access the transporters. However, it is not known whether the regulation of spine shape and uptake of glutamate occurs in the same process.
Relation to morphological change in culture
Astroglial processes and dendritic spines
The actin-rich filopodia observed by scanning electron microscopy in cultured hippocampal astrocytes appear to represent primitive precursors of the fine astroglial processes that ensheath dendritic spines (Cornell-Bell et al. 1990). This hypothesis is supported by two observations: firstly, that focal exposure of astrocyte cultures to glutamate resulted within seconds to a dramatic increase in the number of filopodia; secondly, numerous filopodia appeared upon addition of a suspension of pyramidal cell neurons to the cultures, but only in areas where neurite endings and growth cones made contact with the astrocytes.
Glutamate also produced filopodia on neuronal dendrites, but considerably less frequently than in astrocytes. Parallel to the formation of filopodia in astrocytes, focal glutamate also induced rapid transient increase in intracellular Ca2+ and the propagation of Ca2+ waves through the culture syncytium (Cornell-Bell et al. 1992). Although the exact relation between increased intracellular Ca2+ and the formation of filopodia remains unknown, Ca2+ is intimately involved in actin dynamics regulated by Rho family GTPases. In the intact tissue synaptically released glutamate is known to induce Ca2+ transients in local astrocyte processes, but the issue as to whether Ca2+ waves propagate any distance in in situ preparations or in vivo is contentious (Wang and Bordey 2008). Any propagation would be limited as astrocytes in the intact brain are organized in isolated gap junction-coupled networks of around 100 cells (Houades et al. 2006).
Lesion-induced gliosis: reactive astrocytes
Neural cell injury can cause astrogliosis, characterized by cell swelling and/or proliferation, followed by the secretion of a myriad of molecules involved in the neuroglial response, including cytokines, nitric oxide, metallothioneins, and chemokines (Chen and Swanson 2003; Leung et al. 2009; Li et al. 2009; L'Episcopo et al. 2010; Summers et al. 2010). Despite the growing number of studies related to lesion-induced gliosis, reactive astrogliosis stimulated by CNS infections and changes in blood–brain barrier properties remains less well known.
Increasing evidence points toward the potential of astrocytes to undergo polarization when they are near to blood vessels; they can establish contacts not only by endfeet but also by transitory contacts of the glial processes (Wolburg et al. 2009a). The establishment and maintenance of the polarity seems to be influenced by extracellular matrix of the basal lamina and by agrin, through the modulation of orthogonal arrays of intramembranous particles. One of the proteins present in the orthogonal arrays of intramembranous particles is water channel aquaporin 4 (Wolburg et al. 2009b).
In addition astrocytes display polarity with T cells. The astrocyte response to immune T-cell attack is followed by establishment of immunological synapses between T cells and target cells. It was observed that astrocytes display dramatic morphological changes from multipolar to unipolar cells, with process formation, becoming polarized toward the immunological synapse (Barcia et al. 2008). Therefore, the astroglial polarity which follows T-cell recognition deserves further studies to increase our understanding of immune responses in the brain.
Studies related to astrogliosis have been performed in a variety of animal (Norton et al. 1992) and cell culture (Wu and Schwartz 1998) models. Treatment of Wistar rats with clinically relevant doses of lithium chloride, used as a mood-stabilizing drug, resulted in a 35% increase in the astrocyte marker GFAP in the hippocampus, as defined by immunohistochemistry (Rocha et al. 1998). In this study, GFAP labeling revealed a mild gliosis in the CA1 area and the dentate gyrus, which was associated with changes in the orientation of astrocytic processes. The culture models range from astrocyte cultures (Gottfried et al. 2003), both neonatal and adult, to co-cultures of astrocytes with either neurons or microglia (Wu and Schwartz 1998), to organ cultures (Valentim et al. 2001).
Astrocytes comprise a complex cell population with several morphological variants and the mechanisms whereby morphological plasticity occurs, as well as the corresponding functional consequences are far from being understood. Accumulating evidence in the study of astrocyte-neuron communication emphasize the influence of astrocytes, via gliotransmission, in the modulation of synaptic plasticity and function. In addition, astrocytes are recently known to play a major role in the regulation of the immune/inflammatory response in several human CNS diseases and astrogliosis processes. It is now clear that reactive astrogliosis is finely tuned followed by a continuum of changes, influenced by cellular domains and tissue structure. Study of the molecular mechanisms related to astroglial shape, plasticity and function is essential to discover the molecules that control the processes. These molecules in turn have the potential to be targets where therapeutic agents can be developed for use as neuroprotectants.