Astrocytes have traditionally been viewed as passive supportive cells, which were primarily responsible for maintaining an optimal environment for electrical neuronal activity. Recent studies have, however, demonstrated that the activity of nerve cells can be modulated by astrocytes, in that neurons are recruited into astrocyte-initiated and propagated calcium waves, both in vitro and in situ. By this means, propagated shifts in cytosolic calcium within the astrocytic syncytium may regulate neuronal response and firing thresholds. In turn, astrocytes are actively modulated by neuronal activity, and the existence of astrocyte–neuron signaling loops has been established in several areas of the brain. As a result of these findings, it is now recognized that astrocytes play an active role in brain function, particularly within the highly coupled astrocytic syncytium of the neocortex and the hippocampus. The mechanisms by which calcium signaling is propagated and how it is evoked are the focus of intense research activity. It is known that gap junctions and the connexins, their constituent proteins, together with the local cytoskeleton, the calcium buffer capacity, and calcium waves triggered by purinergic transmitters, all cooperate to modulate astrocytic signaling to neighboring cells in young animals. What changes do astrocytes and their signaling machinery undergo during the aging process? This is a question of paramount importance; altered astrocytic dynamics in the aged brain may alter synaptic efficacy and neuronal survival and perhaps contribute to the cognitive decline observed during aging. In this review, we analyze our current understanding of astrocytic function during aging by reexamining the mechanisms by which astrocytes contribute to neuronal function and survival in normal brain and the changes they undergo in the aged brain. Astrocytes outnumber neurons by five- to tenfold in the adult brain (Bignami, 1991) and establish numerous small contacts with neurons, neighboring astrocytes, and all the other brain cell types, including the endothelial cells of blood vessels (Rohlman and Wolf, 1996). These physical interactions allow them to function as metabolic and passive supportive cells of the brain. First, astrocytes regulate the ionic environment, in particular, after intense synaptic activity where unbalanced ionic fluxes, especially of K+ ions, are built up in the extracellular space (Karwoski et al., 1989). Second, astrocytic end feet contacts with capillaries and arterioles contribute to the blood–brain barrier formation (Janzer and Raff, 1987) and regulate blood flow after local changes resulting from neuronal activity (Clark and Mobbs, 1992). Third, astrocytes respond to the metabolic needs of neurons by activating glycogen metabolism and releasing lactate for neural consumption (Poitry-Yamate et al., 1995). Astrocytes can also respond to neuronal activity by clearance of glutamate from the extracellular space. Specific astrocytic glutamate transporters that are predominantly coupled to Na+-dependent systems mediate astrocytic glutamate uptake (for review see Anderson and Swanson, 2000). Once it is taken up by astrocytes, glutamate is either transformed into glutamine or oxidized via the tricarboxylic acid cycle (Martinez-Hernandez et al., 1977; Yu et al., 1982). Both pathways will lead to the production of several intermediates that will be taken up again by neurons as energy substrates (Poitry et al., 2000). Finally, glutamate–glutamine cycles between astrocytes and neurons are associated with intercellular fluxes of ammonium and constitute the major route for nitrogen balance between astrocytes and neurons (Marcaggi and Coles, 2001; Fig. 1). Metabolic coupling between neurons and astrocytes. Glutamate released from neurons at sites of synaptic activity is taken up by astrocytes via an Na2+-dependent transporter that is coupled to the Na2+/K+ ATPase (1). Glutamate in astrocytes is either converted to glutamine or oxidized in the tricarboxylic acid cycle (2). Glutamine is released by astrocytes and taken up by neurons, where it is converted back to glutamate to complete the cycle (3). This glutamate/glutamine cycle (4) also constitutes a major pathway for the flux of ammonium between the two cell types (5). Although not electrically excitable, astrocytes express a variety of ion channels and neurotransmitter receptors by which they actively respond to neuronal activity (Porter and McCarthy, 1997; Newman and Zahs, 1998). In addition, astrocytes can release a variety of modulatory substances, including neurotransmitters [adenosine triphosphate (ATP), glutamate], growth factors [nerve growth factor (NGF), neurotrophin-3 NT-3), basic fibroblast growth factor (bFGF)], and cytokines (ICAM), to which neurons respond (Frohman et al., 1989; Dani et al., 1992; Condorelli et al., 1995; Kuzis et al., 1995; Araque et al., 1998). Thus, although brain function has been traditionally thought of in terms of neuronal activity, glial cells can also generate signals that affect their neuronal counterparts (Fig. 2). As a result, astrocytes have gained serious consideration as active modulators of brain function (Smith, 1994). One of the mechanisms by which astrocytes can signal to neighboring cells is by propagating calcium increments that spread from cell to cell. These calcium waves can travel over long distances within a population of astrocytes and alter the calcium levels of neurons, microglia (Nedergaard, 1994; Parpura et al., 1994), and endothelial cells (Leybaert et al., 1998). A,B: Astrocytic calcium waves trigger neuronal calcium increases. Electrical or mechanical stimulation of astrocytes to increase their calcium levels causes concomitant calcium increses in adjacent neurons. Astrocytic calcium waves propagation requires the presence of gap junction proteins (Finkbeiner, 1992; Charles et al., 1992; Blomstrand et al., 1999). Gap junctions are intercellular channels that connect the interior of two neighboring cells and serve as direct conduits of ions and small signaling molecules up to 1 kDa in size (Sáez et al., 1986). They are a particular feature of adult astrocytes, with more than 50,000 gap junction channels interconnecting each astrocyte to its neighbors (Yamamoto et al., 1990), in contrast to neurons and other brain cells, which are poorly coupled. Connexin 43 (Cx43) is the major, although not the only, gap junction protein found in astrocytes (Dermietzel and Spray, 1998). Astrocytic gap junctions help to distribute metabolic substrates and products within the brain (Tabernero et al., 1996), help to redistribute potassium ions after neuronal electrical activity (Ransom, 1996), and contribute to the so-called cell-to-cell communication via calcium waves mentioned above (Smith, 1994). However, recent studies have shown that astrocytic connexins serve other roles that do not necessarily require the formation of gap junction channels, such as the regulation of cytoskeletal organization (Cotrina et al., 2000) and of ATP release from glia (Cotrina et al., 1998). ATP can serve as a neurotransmitter in brain and activates responses in both neurons (Edwards et al., 1992; Gu and McDermott, 1997) and glial cells (Salter and Hicks, 1994) via either ligand-gated cation channels or metabotropic receptors that promote release of calcium from intracellular stores by the activation of the inositol triphosphate (IP3) signaling cascade (Kastrikis et al., 1992; Salter and Hicks, 1994). In addition, ATP and its related metabolites have important trophic effects on several brain cell populations by regulating neurite outgrowth (Neary et al., 1996), astrocytic shape changes (Neary et al., 1994; Rathbone et al., 1998), or survival of motor neurons via intracellular elevation of cyclic adenosine monophosphate (cAMP; Hanson et al., 1998). The phenomenon of intercellular calcium signaling constitutes the foundation for the capacity of astrocytes to influence brain activity. For instance, direct stimulation of astrocytes can modulate the firing frequency of both inhibitory and excitatory neurons in dissected eyecup retinas (Newman and Zahs, 1998) and hippocampal cultures (Araque et al., 1998), and it can potentiate inhibitory synaptic transmission in hippocampal slices (Kang et al., 1998; Fig. 3). In addition, intracellular calcium oscillations in astrocytes can influence calcium dynamics of adjacent neurons (Pasti et al., 1997). Significantly, neurons can also affect the calcium levels of astrocytes in a reciprocal manner by the release of glutamate (Dani et al., 1992; Araque et al., 1998) or γ-aminobutyric acid (GABA; Kang et al., 1998). Thus, astrocytes and neurons establish feedback signaling loops that can affect a wide variety of synapses in the nervous system. Astrocyte stimulation modulates synaptic transmission in hippocampal slices. A: A pyramidal neuron (Pyr) exhibits an increase in miniature inhibitory postsynaptic currents (mIPSCs) after astrocyte (Ast) stimulation in hippocampus. B: Stimulation of a single astrocyte in a hippocampal slice loaded with the calcium indicator Fluo-3 triggers a calcium wave. BAPTA loading of the stimulated astrocyte (Ast) blocks the increase in mIPSCs of an adjacent pyramidal neuron (Pyr). The lower traces indicate five recordings before and after stimulation of a single astroyte. Grosche et al. (1999) have recently shown that small compartments of cerebellar glial processes appear to function as independent domains upon activation of neuronal inputs. Calcium increments in response to neuronal stimulation remained localized, and no spread of calcium signals within or between glial cells was observed. These observations call into question the existence of long-range calcium waves in the intact brain but provide evidence for the signaling importance of the spatial arrangement between neurons and astrocytes. Astrocytes, in conjunction with microglia, respond profoundly to neuronal injury and undergo a series of metabolic and morphological changes that are known as reactive gliosis or astrogliosis (also observed under a variety of conditions, including cerebral ischemia, Alzheimer's disease, Parkinson's disease, Huntington's disease, and amyotrophic lateral sclerosis; Schipper, 1996). Increased numbers of activated microglia and enlarged and phagocytic cells that express the cytokine interleukin-1 (IL-1) are prominent in reactive gliosis (Mrak et al., 1997). Concomitantly with the proliferation of microglial cells, there is hypertrophy of astrocytes and a marked variation in the expression of cytoplasmic antigens [glial fibrillary acidic protein (GFAP) and vimentin], surface proteins (PSA-NCAM), and growth factors (CNTF; Ridet et al., 1997). Metabolically, reactive astrocytes also exhibit an increase in oxidoreductive enzyme activities (Eddleston and Mucke, 1993). An early stage of reactive gliosis is what characterizes the astrocytes of the aging brain, even when no sign of disease is apparent. Aged astrocytes exhibit an elevated content of GFAP and of the calcium binding protein S100β (Sheng et al., 1996; Nichols, 1999). GFAP is not only increased at the single cell level but it is also detected in a higher proportion of total brain cells, suggesting that aging is associated with an increase in the relative number of glial cells. It has been estimated that the number of astrocytes and pericytes increases 20% in the aged cortex and other brain regions (Pilegaard and Ladefoged, 1996; Peinado et al., 1998; Rozovsky et al., 1998), whereas the number of oligodendrocytes and microglia does not change. Use of oligonucleotide arrays has yielded the first profile of gene expression from the aging brain of mice and evidence that aging seems to be associated with an inflammatory response and oxidative stress both in neocortex and in cerebellum (Lee et al., 2000), with parallels to human neurodegenerative disorders. GFAP is also one of the genes that undergoes a twofold increase in expression. Thus, the GFAP increases of the aged astrocytes may be the result of a response to the inflammatory and oxidative state of the aging brain. Indeed, better comprehension of the features that distinguish a normal, “healthy” old brain from a brain that is at an early stage of a neurodegenerative disease is a key aspect in developing treatments. This is important because inflammatory and oxidative responses do promote alterations in calcium signaling (for review see Squier and Bigelow, 2000), which is the primary signaling mechanism by which astrocytes modulate neuronal function and could thus be critical for the progression from the “aged” to the “diseased” brain. To our knowledge, alterations in the calcium regulation and activity of aged astrocytes remain totally unexplored. It has been shown only very recently that aged astrocytes continue to express high levels of gap junction proteins and, more importantly, that these gap junction channels are functional (Cotrina et al., 2001; Fig. 4). However, connexin-mediated calcium signaling among aged astrocytes and surrounding cells has not been evaluated. Likewise, cytoskeletal changes of aged astrocytes that could affect calcium signaling have not been reported. As mentioned above, the metabolic turnover of GFAP is increased in aged glia (Amenta et al., 1998), and accumulations of tau, although not neurofibrillary tangles (NFT), have been found in astrocytes and oligodendrocytes of aged baboons (Schultz et al., 2000). Glial tangles also contain hyperphosphorylated tau, as do NFT present in neurodegenerative neurons, but most of the proteins associated with NFT are absent (Ikeda et al., 1998). Therefore, glial tangles can be considered an earlier stage of the tangles observed in neurodegenerative disease. Alterations in cytosolic calcium can increase tau hyperphosphorylation (Mattson et al., 1991). However, at present we do not know how these changes might affect glial calcium signaling. Finally, nothing is known about the activity of extracellular purinergic compounds in glia of the aged brain. Measurements in cell suspensions of aged mouse cortex demonstrate alterations in intracellular ATP maintenance (Joo et al., 1999), which could be translated as changes in the production of extracellular ATP and its derived metabolites. Importantly, adenosine levels are increased under pathological conditions, including ischemia and epileptic activity (Phillis et al., 1991; Rudolphi et al., 1992). It has been hypothesized that, under these circumstances, purines may play a neuroprotective role by enhancing neuronal survival and decreasing excitotoxic transmitter release. Indeed, clinical trials with synthetic purine derivatives are currently being undertaken in patients with Alzheimer's disease (Rathbone et al., 1999). However, the actual adenosine levels in the aged brain are totally unknown. Functionality of astrocytic gap junction proteins is preserved in the aged brain. The fluorescence recovery after the photobleaching technique (FRAP) allows monitoring of gap junction function in vivo: hippocampal slices from 3-month-old (left) and 21-month-old (right) mice are loaded with the gap junction-permeable tracer CDCF (top panels); an area is selected for laser bleaching (rectangle; middle panels); refill of fluorescence is recorded 2 min after laser bleach (bottom panels). Another aspect that can contribute to altered generation of cellular ATP is the redox condition of the aged astrocytes. As mentioned above, alterations in oxidation and stress proteins seem to be a common feature of the aged brain and are likely to affect the calcium homeostasis and the generation of extracellular ATP. For instance, IL-1, a cytokine produced by activated microglia, induces S100β expression in astrocytes, which, in turn, increases intracellular free calcium levels (Mrak et al., 1995). IL-1 can differentially regulate calcium wave propagation by switching between a gap junction-mediated event to a purinergic-mediated pathway (John et al., 1999). Oxidation in the calcium signaling protein calmodulin (CaM) alters the maintenance of intracellular calcium levels and changes the transport activity of the plasma membrane Ca-ATPase of aged muscle and neuronal cells (Squier and Bigelow, 2000). Oxidation of the thiol groups in the IP3 receptor will lower the threshold IP3 concentration for calcium release (Peuchen et al., 1996), promoting amplified calcium responses (Fig. 5). Schematic representation of interastrocytic calcium signaling and some of the alterations that oxidation can induce in calcium of the an astrocyte is inositol is produced via The of an cascade calcium release from intracellular stores to the The calcium signal is propagated to neighboring cells by two the of IP3 gap junction intercellular channels a pathway that activates ATP release to the extracellular ATP in activates purinergic receptors in the membrane of the cells and the calcium calcium proteins can be upon oxidation by in Fig. 5). Oxidation of the IP3 receptor can calcium whereas changes in the oxidation state of the calcium signaling protein can the activation state of the in the plasma these changes are likely to affect not only the signaling of the astrocytes but also of all the other brain cells, including the neurons. Thus, on the of the of the aging brain to exhibit inflammatory and oxidative and the for metabolic and signaling proteins, we can that the calcium levels of aged astrocytes an This might be the for changes in the calcium signaling of the aging astrocytes. For increased calcium levels to the calcium signaling by the threshold for IP3 release. However, alterations in the ATP maintenance by aging will the of ATP and its derivatives in the extracellular space (for by the of from ATP into Although adenosine can promote neuronal survival it might also ATP signaling activity, a key for calcium wave Thus, on the one calcium signaling in the aging brain could be as a result of an increase in calcium levels of astrocytes, on the activity might be but for in by an increase in gap junction-mediated calcium waves by and other will these the activity and survival of the neighboring do not It is that we the in our of the modulatory role of calcium signaling. First, we have to how a gap junction-mediated wave from a purinergic-mediated wave in its to affect neurons and surrounding cells. Second, we have to the extracellular levels of ATP and its derivatives that we can the of each to calcium signaling during the aging Third, and most we have to the of calcium waves in the aging brain and the of signaling to neighboring cells. One to all of these be to establish an that the An might be the of intact brain slices from adult and aging animals. The more that are with this be for to most of the aging is associated with a number of that affect only brain Thus, that can the mechanisms in the brain could contribute to neuronal because normal aging is by brain function, particularly in and but not by neuronal et al., 1996; and 1996). a recent has an decline in the number and size of neurons in regions of aged et al., 1999). These to hippocampal and areas et al., and, when they are their to in and some of et al., et al., 1994). Importantly, the nerve growth factor to the effects of neuronal suggesting that neuronal activity. In the some of the changes that astrocytes undergo during aging in help the of the neurons. in the GFAP content of astrocytes have not been particular role other than the shape changes associated with the of this cell However, although we are not of how increased GFAP may alter astrocytic function, some of this gene in mice are to the of altered GFAP in neuronal First, GFAP astrocytes are a better for neuronal survival and neurite outgrowth than astrocytes et al., 2000), which in may high levels of GFAP are only in astrocytes, when the of has et al. have that of the GFAP gene in the which does not an increase in a of neuronal synaptic of cerebellar that of synaptic is in GFAP although motor is normal et al., 1996). mice GFAP are more to injury et al., 1998). the other of human GFAP in astrocytes of mice et al., 1998). It is that GFAP in cytoplasmic to observed in that appear in disease. Indeed, a recent has associated in GFAP with the of this which is by and et al., These observations that the changes by activated astrocytes, together with the associated in cell surface molecules and extracellular et al., are likely to be associated with alterations in interactions that affect synaptic activity and neuronal it to the increase in the GFAP content of aged This might be the an associated with increased in and and 2000), by more than the GFAP increase of old we do not know can be considered as a for An to some of the changes of aged astrocytes is astrocytes are to the of a variety of Both astrocytes and oligodendrocytes express a variety of particularly receptors for and the and and et al., 1999). receptors regulate astrocytic expression of growth factors and GFAP and glutamine et al., et al., 1994; Rozovsky et al., proteins that regulate ion function et al., 1994; et al., 1994) and proliferation et al., 1992; et al., 1997). has effects on calcium of cultures of astrocytes in a increase in the of calcium wave In addition, two and are also of astrocytic calcium waves, both the calcium levels and the of calcium wave propagation et al., Fig. ATP release is by whereas the functional coupling and the expression of remain in cultures et al., 1999). Thus, astrocytic calcium signaling decline in the aging brain, we could compounds that than interastrocytic calcium as to decline in astrocytic calcium signaling. A,B: Astrocytic calcium signaling can be by of astrocytes loaded with Fluo-3 astrocytic calcium signaling by both the of the calcium responses and the of wave the studies above, it is that our of the functional features of aged astrocytes is very do not know aged astrocytes their as as their signaling do not know how and oxidation glial function and, modulate synaptic importantly, no studies have been to the changes in aged astrocytes contribute to Finally, be taken in for the although seem to have a neuroprotective role the of Alzheimer's disease et al., presence of et al., 1998). is to how the changes of aged astrocytes astrocytic with on studies in in the aged brain than in aged astrocytes in