The problem of astrocyte identity

Neurochemistry International 45 (2004) 191–202

The problem of astrocyte identity Harold K. Kimelberg∗ Neural and Vascular Biology Theme, Ordway Research Institute Inc., Center for Medical Science, 150 New Scotland Avenue, Albany, NY 12208, USA Received 12 June 2003; received in revised form 27 August 2003; accepted 27 August 2003

Abstract Astrocytes were the original neuroglia of Ramón y Cajal but after 100 years there is no satisfactory definition of what should comprise this class of cells. This essay takes a historical and philosophical approach to the question of astrocytic identity. The classic approach of identification by morphology and location are too limited to determine new members of the astrocyte population. I also critically evaluate the use of protein markers measured by immunoreactivity, as well as the newer technique of marking living cells by using promoters for these same proteins to drive reporter genes. These two latter approaches have yielded an expanded population of astrocytes with diverse functions, but also mark cells that traditionally would not be defined as astrocytes. Thus we need a combination of measures to define an astrocyte but it is not clear what this combination should be. The molecular approach, especially promoter driven fluorescent reporter genes, does have the advantage of pre marking living astrocytes for electrophysiological or imaging recordings. However, lack of sufficient understanding of the behavior of the inserted constructs has led to unclear results. This approach will no doubt be perfected with time but at present an acceptable, practical definition of what constitutes the class of astrocytes remains elusive. © 2004 Elsevier Ltd. All rights reserved. Keywords: Astrocytes; Astroglia; Glia; GFAP; Stem cells; Radial glia; Muller cells; Gene expression

1. History of neuroglia and astrocytes A critical part of science is definition, or as Ludwig Wittgenstein said of philosophy, the meanings and uses of words. Glia were first named by the famed neuropathologist Rudolf Virchow in 1856, but looking at the representative drawings of the stained sections he used (Somjen, 1988) they barely resemble anything we would now recognize as adequately fixed and stained brain tissue showing any recognizable members of the different sub-classes of glia. However, Virchow had clearly made a major advance in our concept of the constituents of nervous tissue. The next advance about 30 years later was in staining, the black chrome-silver reaction of Golgi (1885). Using this technique Golgi recognized both ependymoglial cells, now referred to as radial glia, and multipolar autonomous glia, both of which were clearly different from neurons. Andriezen in 1893, using the staining method of Weigert, was the first to distinguish fibrous glia in the white matter containing a greater number of intracellular fibrils from protoplasmic glia in the gray matter (Privat and Rataboul, 1986; Somjen, 1988). However, Andriezen erroneously speculated that the fibrous glia derived from ectoderm while the pro∗

Tel.: +1-518-641-6466; fax: +1-518-641-6304. E-mail address: [email protected] (H.K. Kimelberg).

0197-0186/$ – see front matter © 2004 Elsevier Ltd. All rights reserved. doi:10.1016/j.neuint.2003.08.015

toplasmic glia derived from mesoderm. It appears to have been Ramón y Cajal (1913), and perhaps earlier (Somjen, personal communication), who was the first to attach the name astrocyte to the fibrous and protoplasmic glia, now visualized with Cajal’s own astrocyte-specific gold sublimate stain which deposited on the glial intracellular fibrils (Duffy, 1983; Privat and Rataboul, 1986; Somjen, 1988). “My repeated inquiries upon the technique of coloring the neuroglia selectively, stimulated considerably by the interesting work of Achúcarro (carried on in my laboratory) upon the structure and connections of the human glia, led, in 1913, to my discovering the method of gold sublimate, a most simple procedure which allows one to impregnate specifically with a purple violet color the two types of neuroglia of the cerebral cortex, and especially the “protoplasmic” form, or that with short processes, which is so notoriously refractory to the tedious methods of Weigert, Fano, Alzheimer, and others currently used by pathological anatomists” (Craigie, 1989). Ramón y Cajal considered that both the protoplasmic and fibrous neuroglia were of ectodermal origin, and he mainly used the term neuroglia in referring to these cells rather than the term astrocytes. In a drawing in Ramón y Cajal’s 1913 publication of brain tissue stained with gold chloride and reproduced as Fig. 3 in Somjen’s review (Somjen, 1988), the astrocytes clearly resemble GFAP-stained or otherwise visualized astrocytes in current preparations. Later del R´ıo

192

H.K. Kimelberg / Neurochemistry International 45 (2004) 191–202

Hortega, Ramón y Cajal’s renowned pupil, showed in 1920 that Ramón y Cajal’s non-astrocytic, non-neuronal enigmatic third element consisted of the oligodendroglia and the microglia, the latter being unlike the other two forms of glia in that they were likely of mesodermal rather than ectodermal origin (Somjen, 1988). Thus the term neuroglia became redundant and was replaced by oligodendroglia (oligodendrocytes) and astroglia (astrocytes). These two groups comprise the macroglia, and the microglia are off in a class on their own as befits their properties as immune cells of mesodermal rather than ectodermal origin. The astrocytic fibrils that were the targets of both Weigert’s and Cajal’s stain are bundles of the 8–9 nm intermediate filaments (IMF) that we now know consist mainly of one kind of intermediate filament protein, glial fibrillary acidic protein (GFAP) (Eng et al., 1971). In the brain this is specific to developing, mature and reactive astrocytes, and astrocytes are increasingly being defined in this period of molecular biology as cells that synthesize this protein or even just express the mRNA of the GFAP gene (Nolte et al., 2001). It should be noted that the microscopic appearance of stained astrocytes is technique-dependent. Thus astrocytes stained for their filaments show much less of their total structure, while astrocytes stained with the Golgi stain or filled with dye have a much fuller complement of processes showing a much bushier appearance (Bushong et al., 2002; Connor and Berkowitz, 1985; Ogata and Kosaka, 2002). Nowadays we have many more histological techniques such as antibody labeling for many protein “markers” and the filling of individual cells by dyes using patch-clamp electrodes. We can combine filling of cells with dyes or driving a fluorescent gene product linked to astrocyte-specific promoters with function-related properties, usually electrophysiological, to define these cells and so we can match dynamic properties and histology. We can also determine the developmental position of these cells by permanently labeling cells early on and finding out what they become through their subsequent divisions. We can then describe which cells are terminally differentiated and what cells they derive from. We can approach their functions in the intact animal by deleting or adding genes linked to cell-specific promoter activities. It should be no surprise, therefore, that we now need to redefine what we mean by an astrocyte in light of newer data and the newer concepts that derive from such data. In this commentary I will consider all cells that have been considered as astrocytes from time to time and critically evaluate current methods used to identify astrocytes. In the 1880–1920 period, as discussed above, this term was applied to cells based on morphology, specific staining with Cajal’s gold sublimate, and exclusionary definitions such as those cells that were neither neurons, oligodendrocytes, microglia and ependymal cells of the CNS. I will try to relate these original concepts to the newer data just alluded to, but in my opinion an acceptable and useful definition of the term astrocyte is a work in progress.

2. The importance of lineage in naming (Oscar Fingal O’Flaherty Wills Wilde—the importance of being earnest) To sort out relationships in organismic biology, the traditional approach is to apply that most basic biological principle of all, the Principle of Evolution. In both the plant and animal kingdoms simpler organisms evolve into the more complex, and by determining the nearest kin you have a good basis for classifying them. The same principle applies to cells for, after Schwann and Schleiden first enunciated the cell theory in 1839 that all living organisms consist of cells, with larger and more complex organisms having a proportionally greater number and more diverse collection of cells, there immediately arose the problem of where all the cells come from. Rudolf Virchow in 1859 stated totally accurately omnis cellula e cellula (all cells from cells). So, in sorting out which cells belong together, their relationships during ontogeny can be applied. The totipotential fertilized egg (stem) cell transfers its genetic blueprint unchanged to all its progeny since all subsequent cell multiplication is clonal. Therefore cell differentiation must be via the complex mechanisms of differential gene expression controlled by cues within the developing organism. Studies on the passages of stem to precursors to differentiated cells are referred to as lineage studies. I illustrate this by showing one figure from a recent paper showing the results using markers to determine the lineage of all the different cells so far identified in the rat spinal cord (Fig. 1). The astrocyte lineage is identified with the purple violet colored cells to the right of the diagram and shows three different origins. One from radial glia, another directly from stem cells and a third from an A2B5(+), GFAP(−) precursor. The mature astrocyte is indicated as always being a GFAP(+) cell. We know that cells that would be classified morphologically as astrocytes do not always have readily detectable GFAP (see Walz and Lang, 1998; Walz, 2000 and the later section on GFAP(−) astrocytes). Also radial glia can be GFAP(+) (see later section). NG2(+) cells found in both the developing and adult mammalian CNS were until recently referred to as a smooth protoplasmic astrocytes, because of their astrocytic appearance but with less branched processes and a paucity of intracellular filaments (Levine and Card, 1987). They are always GFAP(−) (Nishiyama et al., 1999) and are now placed in the oligodendrocyte lineage and are commonly referred to as oligodendrocyte precursor cells (OPCs). However, as it is a numerous cell type in the adult it seems unlikely, though obviously not impossible, that these cells only serve as a source of oligodendrocytes in the adult CNS. Recently they have, using a CNP promoter linked marker, been shown to give rise to neurons as well (Gallo et al., 2003) Also they could be bipotential in the sense of performing both physiological and progenitor functions.

H.K. Kimelberg / Neurochemistry International 45 (2004) 191–202

193

Fig. 1. Possible lineage relationship between different glial and neuronal precursors. The antigenic characteristics and temporal appearance of the precursors are summarized. Solid lines indicate likely lineage relationships. Two oligodendrocyte–astrocyte precursors (GRP1 and GRP2) are shown based on the two domains of expression of glial markers. Astrocyte precursors may originate from uni- or bipotential glial precursors or from totipotential stem cells. Radial glia may also generate neurons and this is indicated by the bottom horizontal dotted line. The time line at the right indicates the approximate age at which the different cells are first detected in rodents. Precursor cells generated at an earlier stage may persist even while a subset matures (from Liu et al., 2002 with permission).

3. Evolutionary view of astrocytes One long held view is that mature stellate astrocytes in later vertebrates derive from or replace the radial glia (ependymoglia) that traverse the entire wall of the brain from ependyma to pial surface and persist through adulthood in the lower vertebrates (Chanas-Sacre et al., 2000; Kálmán and Pritz, 2001; Roots, 1986). Radial glia are the dominant form in reptiles, and comparative studies have revealed considerable similarity between these GFAP(+) structures in the closest living members (crocodiles) of the reptilian ancestors of the birds, and the turtles that are the closest to the extinct stem reptiles that led to both birds and mammals (Kálmán and Pritz, 2001). Birds and mammals resemble each other in that stellate astrocytes, rather than radial glia, are the dominant form in adults and may comprise both GFAP(−) and GFAP(+) cells. This leads to the view that astrocytes expanded independently in birds and mammals to fit the increased size and complexity of the brain, as would be expected from the

parallel evolution of birds and mammals (Kálmán and Pritz, 2001). An hypothesis has been advanced that astrocytes evolved from radial glia in response to K+ spatial buffering being progressively less efficient as radial glia become longer and thinner with the increasing thickness of the brain wall (Reichenbach, 1987). An alternative, but not mutually exclusionary explanation for this change, is that proteins needed at more distal parts of the radial glia would need to be transported longer and longer distances from the sites of syntheses close to the single nucleus in the radial glial cell body in the subventricular zone, or the mRNA would need to be transported to distant ERs. As Kálmán and Pritz (2001) point out the conversion of radial glia to more numerous stellate astrocytes does allow for greater variation of gene expression in each astrocyte. Kálmán and Pritz (2001) were specifically interested in why some mammalian astrocytes express GFAP while others do not, whereas all radial glia in the reptiles are GFAP(+). Since the effects of GFAP knockouts are quite subtle and often delayed and affect such

194

H.K. Kimelberg / Neurochemistry International 45 (2004) 191–202

diverse properties as long term potentiation and depression and blood–brain barrier integrity, and their structural intermediate filament roles can be replaced by other IMFs such as vimentin, there are no clear consequences of the lack of GFAP (Eng et al., 2000). Be that as it may, astrocytes under individual genetic control can express many different astrocytic phenotypes to accommodate different functions occasioned by increased brain size, synaptic complexity and a greater need to control substrate availability from the blood.

4. Radial glia Radial glia were termed by Ramón y Cajal and his contemporaries ependymoglia or neuroepithelial cells as they have their cell bodies close to the ependymal cells of the ventricle linings and project as thin, bipolar and columnar cells to the brain surface. Currently a distinction has been made, based on markers, between the earliest cells as neuroepithelial cells and the later cells as radial glia (Malatesta et al., 2000). The original proposals for their functions included giving rise to neurons, directing axonal growth or, as Ramón y Cajal favored, provide a support or provisional scaffolding to the developing nervous system (DeFelipe and Jones, 1988). But, as was done at that time, these were simply speculations based on cell morphology and location and could not be tested then by the techniques we now have at our disposal. Thus for “function”, the third arm of the scientific method; (1) observation; (2) a precise and testable hypothesis followed by (3) the testing of the hypothesis, could not be done. Ramón y Cajal’s speculations as to function could be right or wrong but there was no way at the time of determining the likelihood of any of the speculations, although his basic morphological observations still stand famously intact and a number of his functional proposals have turned out to be prescient. Rakic and his colleagues in the early 1970s used morphological observations at the ultrastructural level to show that neurons migrated from their origins in the SVZ along radial glia in primate brain, with the earliest neurons migrating the shortest distance and the later neurons migrating to the outer layers of the cortex (Rakic, 1971). At this time, markers for the different cell types and especially three dimensional light microscopy imaging were not quite available. This gave rise to the temporally inside-out, radial glia directed principle of neural migration. Also by the appropriate geometric arrangement of the radial glia, whereby their cross sectional territorial areas increased from the inner to outer layers, it provided a simple means whereby a small area of the SVZ could populate successively larger areas of the developing cortex (Rakic, 1988). The concurrent identification of GFAP as an astrocyte specific marker (Eng et al., 1971) led to radial glia being viewed as members of the astrocytic lineage, at least in the primate. However, this expression is species-dependent as rodent and chick radial glia do not express detectable GFAP but do have other astrocytic markers,

such as the astrocyte-specific glutamate transporter, GLAST (Campbell and Götz, 2002; Malatesta et al., 2000). One of the most enduring ideas regarding radial glia is that after the period of neuronal migration, at least some radial glia in most regions of the CNS retract and become stellate astrocytes (Schmechel and Rakic, 1979). In more primitive vertebrates such as the teleosts (bony fishes) and reptiles, radial glia persist in most regions (see preceding section on evolution). In mammals they only persist in a few regions such as the Bergmann fibers of the Golgi epithelial cells of the cerebellum and as the Muller cells of the retina and a few other regions such as the hilar region of the dentate gyrus. But although they have the elongate appearance of radial glia it is unclear whether they are the same cells, although the Bergmann fibers do support inward migration of granule cells. The Bergmann fibers in mammals are GFAP(+) but the radial appearing Muller cells of the rodent retina are not but do become GFAP(+) upon injury (Zhuo et al., 1997).

5. GFAP(+) stem cells and a revisionist view of radial glia Recent work has shown that cells that give rise to cells on one of the three lineage lines of neural cells, neurons or the two classes of macroglia (microglia are excluded because of their mesodermal origin), can stain for GFAP and may self-renew, although the latter is a more difficult property to establish. In adult rodents these stem cells have been described in the subventricular zone and the subgranular zone of the dentate gyrus (Doetsch et al., 1999). Since the behavior just described fits the current definition of stem cells (Seaberg and van der Kooy, 2003; Temple and Alvarez-Buylla, 1999) it follows logically that, if all GFAP(+) cells are astrocytes, then these stem cells are astrocytes. Here the use of the human GFAP promoter (Zhuo et al., 1997) linked to a reporter construct is used to identify the cells longitudinally and in this respect it is critical that the behavior of the linked reporter gene faithfully follows GFA protein expression for an unambiguous interpretation of the results. The reader interested further in this question could profitably consult many of the articles in the July 2003 special issue of Glia (Götz and Steindler, 2003), which appeared as I was just putting the finishing touches to this review. As well as stellate GFAP(+) stem cells, radial glia have also been reported not only to generate astrocytes but also neurons, but only rarely oligodendrocytes (Malatesta et al., 2000; Campbell and Götz, 2002; Heins et al., 2002). As a hypothesis this is not new as a bifunctional role for radial glia of providing a scaffold for cell migration and also dividing to give rise to other cell types was proposed by Magini in 1888. All possible variants were proposed thereafter until the current, and now experimentally supported hypothesis, recapitulates Magini (Fishell and Kriegstein, 2003). Again GFAP, although expressed in radial glia in primates, is not

H.K. Kimelberg / Neurochemistry International 45 (2004) 191–202

expressed in rodent and chick radial glia. The question here is if radial glia in some species express GFAP and are therefore considered astrocytes, does it follow that other cells that have the appearance of radial glia but do not stain for GFAP can also be put in the class of astrocytes? If astrocytes are defined as cells expressing GFAP then chick and rodent radial glia are not astrocytes. Also, if one defines an astrocyte as a cell type of the mature CNS then the radial glia, that are only present during early developmental stages would not be considered astrocytes, although a GFAP(+) one would presumably be considered in the astrocyte lineage. But then other terms such as “neuroepithelium-radial glia-astrocytes” (Seri et al., 2001), “astroblasts” or “astrocytic progenitors” or “astroblastoscaffolds” might be more appropriate. But all these are rather a mouthful and astrocyte is a lot easier. But then we are on the slippery slope of a simplistic terminology that is misleading to all but the cognoscenti.

6. GFAP(−) astrocytes As Walz has pointed out (Walz, 2000) cells that morphologically would be considered astrocytes when otherwise stained, such as for S100␤, are not always GFAP(+). The degree of disconnect between GFAP expression and astrocyte-type morphology depends on the brain region. In rat cerebral cortex around 40% of S100␤(+), stellate-looking astrocytes did not stain for GFAP while at least 80% of the morphologically recognizable astrocytes in the hippocampus, which also stain for S100␤, stain for GFAP (Bushong et al., 2002; Ogata and Kosaka, 2002). A cell type that has been considered astrocytic in the past but never stains for GFAP is an enigmatic cell that persists in the adult as a cell type that at one time was referred to as a “smooth protoplasmic astrocyte” (Levine and Card, 1987). It is now usually referred to as an oligodendrocyte precursor cell or OPC as it shares the chondroitin sulfate marker NG2(+) with cells that have been shown to give rise to oligodendrocytes. However, it is puzzling why so many of these NG2(+) cells persist in the adult, well beyond the age of oligodendrocyte proliferation, if their sole function is as oligodendrocyte precursors. It is quite possible, or even likely, that these cells serve under normal conditions several physiological roles but under other conditions function as OPCs (Nishiyama et al., 1999), and also neuronal precursors (Gallo et al., 2003). Thus when PDGF is overexpressed in mice there is a marked increase in OPCs (Temple and Alvarez-Bulla, 1999). These cells lack other characteristics of “astrocytes”. They are not linked by gap-junctions and some appear to make post-synaptic synapses with glutamatergic terminals (Bergles et al., 2000), a property never described so far for GFAP(+) astrocytes. They also do not immuno-stain for GLT-1 or GLAST (Schools et al., 2003). We have found that freshly isolated cells from P5-35 CA1, that had originally been termed “complex” cells because of their non-ohmic I-V curves (Steinhauser, 1993), are NG2(+)

195

(Schools et al., 2003). They do not stain for GFAP but ∼75% express mRNA for GFAP (Zhou et al., 2000), indicating some block to GFAP synthesis at the translational level (Zhou et al., 2000). They have also been reported to be S100␤(+) (Matthias et al., 2003).

7. Reactive astrocytes Astrocytes respond to neuronal injury by hypertrophy and some limited proliferation, although this latter response varies, and they become intensely GFAP(+) (Eng et al., 2000). Reactive astrocytes have long been considered as an impediment to neuronal and axonal regrowth because, in the CNS, reactive astrocytes surround axotomized and dying neurons (Ridet et al., 1997). However, it is not possible to distinguish between reactive astrocytes walling off the CNS neurons that are damaged and unable to regrow, and reactive astrocytes preventing the regrowth of otherwise competent neurons. This reactive astrogliosis could be a recapitulation of the development role of astrocytes in walling off functionally integrated regions of neurons or synapses or the CNS itself as represented by the glia limitans beneath the pia mater and the sub ependymal layer, and the astrocytic foot processes that surround the CNS blood vessels (Steindler, 1993). In terms of electrophysiology there is limited information but it appears reactive astrocytes express voltage sensitive K+ and Na+ channels that may indicate a less differentiated astrocyte. The pathological response could serve to protect still intact CNS tissue from secondary lesions by isolating the damaged region and preventing the spread of toxic materials out from the damaged area into normal tissue. Thus upregulation of the metal binding metallothioneins and antioxidants such as glutathione in the reactive astrocytes (Eng et al., 2000) would be beneficial. This is a detailed field and the reader should consult reviews for more information on the complexity of receptor expression on reactive astrocytes, especially for trophic factors, and their metabolic and other properties (Chen and Swanson, 2003; Ridet et al., 1997; Eng et al., 2000).

8. Functional diversity of GFAP(+) astrocytes Astrocytes, defined according to several different criteria in different studies but consistently within each study, have been shown since around 1990 to have complex as well as the classical ohmic current patterns (see refs. in Walz and Lang, 1998). Fig. 2B and C show the voltage-clamp current profiles of the two types of GFAP(+) “bushy” astrocytes we find in our freshly isolated cell preparations from rat hippocampus (Zhou and Kimelberg, 2000). The one shown in Fig. 2A and B, termed VRA for variably rectifying astrocyte (Zhou and Kimelberg, 2000), shows a close to ohmic pattern. The other shown in Fig. 2C (termed ORA for outwardly rectifying astrocyte) has a more complex pattern due

196

H.K. Kimelberg / Neurochemistry International 45 (2004) 191–202

Fig. 2. Fluorescence micrographs of a freshly isolated astrocyte (A) and a NG2(+) glia (D), respectively, filled with Lucifer yellow dye (0.3% in the pipette solution) and identified during recording (A is a VRA). B, C and E, represent typical membrane currents in the different cell types. In the figure these are induced by 50 ms duration voltage steps from −160 to +60 mV in 20 mV increments. GFAP(+) VRAs (B) are characterized by a largely + + symmetric expression of inward and outward K+ currents that consist predominantly of K+ OHM with small contributions by KA and KDR , but show no Na+ currents. GFAP(+) ORAs (C) are characterized by a dominant expression of outward IK + A and IK + DR plus small inward INa + currents (see inset below C). The ion current profile of NG2(+) glia (E) is qualitatively similar to ORAs in terms of K+ and Na+ current expression. However, the outward K+ currents show a slower desensitization than expected from the K+ A component obtained after current subtraction (not shown). Also the density of the K+ is significantly higher than in the ORAs (from Zhou and Kimelberg, 2000, 2001; Schools et al., 2003, with permission). A

to several voltage-gated K+ currents, and it also has a small, voltage-sensitive Na+ current. The GFAP(+) cells not only exhibit the different and diagnostic voltage-clamp profiles shown in Fig. 2B and C, but other differences including, surprisingly, absence of glutamate transporter currents for ORAs (Zhou and Kimelberg, 2001). It would be intriguing if the ORAs and VRAs represented the protoplasmic and intermediate GFAP(+) astrocytes, respectively, isolated from the gray matter of cerebral cortex, as described in situ some time ago based on morphology (Connor and Berkowitz, 1985). Such freshly isolated cells have not been exposed to artificial culture conditions at all-only the cutting of a hippocampal slice, maintenance in an oxygenated balanced salt solution (BSS) for 1 h, papain for 30 min, then 1 h in BSS and just before study trituration before the 1–2 h experimentation; these may be considered artificial enough! But recent studies have shown that cells in the mouse hippcampus in situ, that have been made to express green fluorescent protein (GFP) under the control of the human GFAP promoter (Zhuo et al., 1997) show an ohmic I–V curve in cells that had a “bushy” morphology and intense fluorescence, and complex electrophysiological profiles in cells that expressed a weaker fluorescence. These latter cells showed few processes by fluorescence but plentiful bushy processes when filled, through the recording electrode, with Texas red-conjugated dextran (Matthias et al., 2003). However, since the NG2(+), GFAP(−) cells have currents qualitatively similar to ORAs

(see Figs. 2D,E), the weakly GFP-expressing cells identified in situ by Matthias et al. (2003) as exhibiting complex currents, may have included both GFAP(+) ORAs and GFAP(−) NG2(+) complex cells if the latter cells also turn on the GFP gene, and indeed they mentioned that some of the weakly fluorescing cells expressed the AN2 protein, which is the mouse homologue of rat NG2. This could happen if the construct containing the GFAP promoter linked to the GFP gene results in its transcription but other elements, which in the wild type gene greatly decrease mRNA stability and would be found at the 3 UTR, are missing from the recombinant message causing a higher level of the mRNA to be transcribed (Malter, 2001). In that case some of the GFP(+) cells may not represent cells that normally express immuno-detectable GFAP.

9. The problem of protein markers Markers are usually proteins which are considered to be expressed specifically by different cell classes. But according to Locke, Hume and the other empiricists, and as assumed in the scientific method, things can only be known by the senses, either directly, or in the sciences indirectly through sensitive measuring devices. Hence, because something has always occurred in certain cell types it cannot safely be assumed it will always only be observed in those cell types and no others.

H.K. Kimelberg / Neurochemistry International 45 (2004) 191–202

Additionally we have the problem that protein expression is a contextual variable. At the turn of the last century it was speculated that differentiation of cells in the growing plant or animal was due to the differentiated cells losing different genes. We now know that cells differentiate because of differential expression or regulation of genes and that all cells in a single organism have the same genetic blue print, with only minor changes due to transposons. Indeed the total number of genes is only a few-fold greater in humans compared to much simpler organisms, although splice variants and editing can increase the number of different mRNAs far beyond the number of DNA genes. Differential gene expression or regulation is controlled by transcription and such like factors acting on gene regulators such as promoters and is, to a significant degree, altered by external factors acting via surface receptors. Thus protein expression varies normally in development controlled by changing contextual cues which can be mimicked unknowingly when cells are cultured under different conditions. Specificity should be improved when more than one protein marker is used. If we assume the independent probability of expression in an astrocyte of two proteins as 95% and the probability of expression in a non-astrocyte is 5%, then if both are found in a situation where 25% of the total population are astrocytes the probability that a stained cell is an astrocyte is 99.2%, as compared to 86.4% when only one of the markers is used. The differences between one or two tests increases as the percentage of the total cell population that are astrocytes decreases; when astrocytes are 10% the probability that a positive cell is an astrocyte is 67.8% with one marker versus 97.6% for two markers.1 So the smaller the percentage of the total sample cell population that the astrocytes represent the more important it is to stain for two or more proteins because of the increasing likelihood of false positives from the larger non-astrocyte population. We also have technical problems that contribute to the detectability of the staining so that from the logic of empiricism we can only refer to what is detectable by the conditions of a particular method. It is particularly difficult to be certain about negative results. This has been emphasized as one reason for immunocytochemical GFAP negativity because of sub optimal fixation conditions and antibody preparations (Eng et al., 2000), and is especially a problem for the protoplasmic astrocytes of gray matter with their lower amounts of IMFs and therefore GFAP. A claim of negative or positive staining is a completely operational one, dependent on methods. Even positive controls are not definitive for one has to consider differences in antigen content or epitope availability. At some level the antigen is not detectable and thus negativity means the cell may contain any content less than this limit, as well as zero. Nonetheless the specific expression of GFAP by cells generally accepted as “astrocytes” based on morphology, 1 Based on Bayes rule for revising probability for results of tests with known characteristics.

197

with only a few disconnects, has been quite resilient for 35 years (Eng et al., 2000). However, its surprising expression in stem and progenitor cells has now been demonstrated (Alvarez-Buylla et al., 2002). Also a continual problem has been that not all cells in the mature CNS that morphologically resemble astrocytes, or have some of the physiological properties of astrocytes such as a dominant K+ conductance and express the components of the glutamate-glutamine cycle such as EAA transporters and glutamine synthetase, have readily detectable GFAP (Walz, 2000; Walz and Lang, 1998). GFAP is also expressed in other cells outside the CNS, especially in cells that border blood vessels such as hepatic stellate cells or the podocytes of the kidney’s Bowman’s capsule. Podocytes also contain metallothionein, as do astrocytes (Buniatian et al., 2002). Thus the expression of GFAP may have something to do with the organization of components of cells which surround a vasculature. Recent studies (Simard et al., 2003) have shown that astrocytic end-feet around larger vessels of >8 ␮m diameter are predominantly GFAP(+) and aquaporin 4(+), while endfeet around <8 ␮m vessels are uniformly GFAP(−) and aquaporin 4(−). The latter are clearly capillaries and the former likely arterioles. It may be that the end-feet are too small for IFs or that there is an involvement of the GFAP-containing IFs in astrocyte-arteriolar signaling. Markers other than GFAP are demonstrably less specific; this also depends on the brain region. Thus some neurons, as well as ependymal cells express S100␤, and ependymal cells also express the astrocyte specific water channel, aquaporin 4 (AQP4). Again multiple markers or a marker combined with some other identifier are safer. The use of the GFAP gene is now being extended to its promoter action as well as its translated protein product. This is a powerful technique for one can use the GFAP promoter to label astrocytes with a fluorescent protein like green fluorescent protein to mark the cells in living tissue which can then be patched and studied electrophysiologically (Matthias et al., 2003). However, we have the question of the correlation of the promoter activity of the inserted construct with the normal translation of detectable GFAP by the endogenous gene. GFAP-promoter-based genetic engineering has also been used to study lineage. In a recent study the GFAP gene promoter was linked to Cre/loxP, which excises the stop signal for LacZ, which then permanently labels any progeny (Malatesta et al., 2000). This labeled virtually all cortical projection neurons, but surprisingly astrocytes, which should derive from radial glia, did not stain in the reporter mice in spite of their containing Cre. For another putative astrocytic marker, the Ca2+ -binding protein S100␤, it has recently been reported that, in the mouse, a wide variety of neurons and a number of neural cell types, other than astrocytes, are fluorescent for EGFP controlled by the murine S100␤ gene promoter sequence (Vives et al., 2003). The expression of the S100␤ promoter-linked EGFP and the S100␤ gene protein showed a high degree of colocalization,

198

H.K. Kimelberg / Neurochemistry International 45 (2004) 191–202

implying that expression of the protein is determined at the transcription level. In other cases a disconnect is seen between promoter-linked expression in a construct and the normally expressed protein. This was found for EGFP linked to the human GFAP promoter and GFAP (Nolte et al., 2001), and thus, in this case control appears to be at the level of translation of GFAP, so that for some reason ther cells make GFAP mRNA without transcribing it to detectable protein. In this case the human GFAP promoter linked to EGFP was transfected into mice. It is possible that certain key regulatory sequences that prevent expression in the non-GFAP expressing rat cells were missing in the human GFAP promoter/EGFP construct. However, a high degree of disconnect had been reported previously for the 2kb murine promoter linked to E. coli ␤-galactosidase activity (Galou et al., 1994). But, again, this is only the short promoter sequence leaving out many of the regulatory elements usually present on the 5 region upstream of the exons, or elements in the 3 untranslated region (3-UTR) that alter the half life of the mRNA (Malter, 2001). Expression of the transgene was found in granule, pyramidal and other neurons and also non-neural tissue, and the expression varied with age. A large number of possible reasons ranging from missing repressor and initiator sequences to differences in insertion sites were discussed (Galou et al., 1994). Thus it appears that promoter linked expression of fluorescent markers for in situ studies always needs to be independently validated. The newly sequenced promoter region of the astrocytespecific GLT-1 (EAAT2 in humans) glutamate transporter has also been used to drive the firefly luciferase gene (EAAT2-Prom-LUC) in primary human astrocyte cultures and several different cell lines (Su et al., 2003). Human primary astrocyte cultures gave a 6–10 fold greater expression than the other human cell lines but there was detectable expression in all the non glial lines. The weak expression of a reporter gene linked to promoters for EAAT2, GFAP or S100␤, which are presumed to define astrocytes in other cells, can lead to an erroneous interpretation of such cells as astrocytic. Thus some of the cells which show weak expression of GFP linked to the GFAP promoter in some cells in hippocampal slices which show electrophysiologically complex currents (Matthias et al., 2003) may actually be GFAP(−) cells, although some of these were S100␤(+).

10. The importance of location In intact tissue the location of a cell in relation to others is a traditional way of identifying it. It also has the advantage of providing clues as to function, as exemplified by the accuracy of many of the proposals of Golgi and Ramón y Cajal, and also Lugaro who first proposed in 1907 transmitter uptake by the finest velate tips of the neuroglia (astrocytes) that surrounded neuronal articulations (synapses) (Kimelberg, 1986). This is lost when we consider only the properties of isolated cells because we cannot be sure where they were

located unless we precisely excise them. If we study the whole population we get an average property. Astrocytes or their processes have long been known to surround essentially all CNS blood vessels, and some, or in some regions all, synapses of a certain type (Ventura and Harris, 1999). Their processes delimit regions such as the surface of the brain next to the pia mater as the glia limitans, and there is a corresponding glia limitans abutting the basal surfaces of the ependymal cells. Astrocytes also surround certain groups of neurons or synapses such as the somatosensory cortical barrels, and delimit synaptic areas (Steindler, 1993). The other putative astrocytes, the embryonic radial glia and the Golgi epithelial cells with their characteristic Bergmann fibers mediate migration of neurons. In the case of the radial glia this is from the SVZ to the cortex, and for the Bergmann fibers migration of the granule cells through the outermost molecular zone to the inner granule cell layer. The elongate shapes and location of radial glia in the developing brain and their traversing of the neural epithelium from the subventricular zone to the brain surface are clearly suited to this guidance task. The solution, to retain the advantages of location, appears to be to study astrocytes in situ by patch clamp and dye imaging (e.g. Simard et al., 2003). However, defining the cells studied still remains a question.

11. Focus on functionality Function is a surprisingly difficult word to define. To some it means any dynamic property such as a receptor activity measured by receptor-dependent changes in such things as intracellular Ca2+ levels or membrane potentials. At the other extreme, others might restrict function to the activity of a whole organ like the liver or brain, and consider the effects of receptor activation as being part of a catalogue of the properties of the constituent cells of the organ. Function would then be restricted to the control of some bodily state, or uniquely to the brain a change in behavior, emotion or cognition. Astrocytes would then be viewed as cells that have certain unique properties that allow them to perform certain roles as part of a complex web of cellular interactions and interdependencies within the brain that allows it to express its functions. Most workers use the term function rather loosely as something in between these two extremes, but it always has a dynamic component. We are currently quite uncertain as to the totality of these astrocytic functions or properties (see Fig. 3). There has been a long history of speculations on these functions starting with Golgi, Cajal and Lugaro (Somjen, 1988), with only work over the past 2–3 decades providing actual data. Further, until quite recently, most of the data derived from work in cultures which have serious problems on their applicability to in situ astrocytes (Kimelberg, 1999). An important question we need to resolve is which of these properties ranging from the static and structural to the dynamic, individually or more likely severally, serve to define an astrocyte. Thus, if expression of

H.K. Kimelberg / Neurochemistry International 45 (2004) 191–202

199

ASTROGLIA (Starlike or columnar, non-myelinatingglia)

Morphology of type:

Marker expression:

GFAP(+), S100 (+), GFAP mRNA (+)

Stellate Elongate Radial glia

Locations

General functions

Embryonic are throughout brain. Persist in the adult vertebrate brain as Bergmann glia and Muller glia and in hilar region of dentate gyrus.

Stem cells ORAs Embryonic are widespread. In adult limited to subventricular zone of lat. vent. and hippocampal dentate gyrus. Also radial glia.

Brain development and neural cell renewal.

Advanced functions

GFAP(-), S100(+) , GFAP mRNA(+)

Yet to be Reactive astrocytes identified?

NG2 (+) OPCs

GFAP(+), S100 (+)

Muller glia and radial glia in some species

Yet to be identified?

VRAs

So far studied only for CA1, 3. Other diversities in terms of glutamate receptors and transporter, etc, reported but still fragmentally. Also classified as receptor and transporter cells.

Brain metabolic functions including homeostasis and water and ion transport, and metabolic support to neurons.

React to loss of or damage to neurons. Intensively GFAP(+).

Widespread in brain gray and white matter. Previously termed smooth protoplasmic astrocytes. Recently OPC reported as ~75% GFAP mRNA (+). Possibly more general progenitors

Astroglia specific signaling processing (e.g., Ca2+ waves)?

Reported in cortex and hippocampus. May be a real population or due to low levels of GFAP below detection.

Delineates regions of brain from vasculature to groups of neurons such as somatic sensory barrels. Reactive astrocytes protect normal from damaged CNS.

Retina in some species, GFAP(-) rodent of chick.

Synaptic support and plasticity.

Learning, memory, behavior, etc.

Fig. 3. The variety of astroglia based on current knowledge, and future possibilities. Astroglia may be divided operationally into GFAP(+) and (−) subpopulations. The GFAP(+) astroglia can be further classified into ORAs, VRAs (see Fig. 2), stem cells and possibly unidentified GFAP(+) astroglia with other functions. For the GFAP(−) astroglia, one clearly distinctive yet controversial population in terms of astrocytic identity are the NG2(+) cells. S100␤ marks more astrocyte look alikes but also other cell types. Unidentified represents those GFAP(−) cells which morphologically resemble astrocytes and which may be identified as astrocytes in the future based on other markers such as GFAPmRNA or physiology. Location is rather specific, general and advanced functions purposely delocalized.

the IMF protein GFAP is sufficient, we have a number of different cells that would be termed astrocytes, but which have very different functions. To avoid this problem many workers refer to cells as being on the astrocytic lineage allowing many unrelated functional properties within the family of astrocytes. This is certainly not a problem. Diversity within related groups is characteristic in biology (e.g. Darwin’s finches). Are there cells that perform an important astrocyte type function such as glutamate or K+ uptake, morphologically resemble astrocytes but do not express detectable GFAP, or perhaps only the message? Such cells do not yet appear to have been described, and so far the existence of GFAP(–) astrocytes is based on cells that look like astrocytes when stained for the Golgi silver stain or Cajal’s gold sublimate or S100␤ which do not exhibit GFAP immunoreactivity. The technical question of GFAP detectability first needs to be resolved before this issue can be further addressed.

12. Confusing conclusions We see astrocyte complexity and heterogeneity as expected within and between different brain regions (Table 1 and Fig. 3). Even though differences are seen when primary astrocyte cultures are prepared from different brain regions (e.g. Amundson et al., 1992), it now seems unlikely that these systems will be useful for classifying astrocytes as

they are a population of strongly expressing GFAP(+) cells that appears to be a too restrictive definition for astrocytes, and they possibly most closely represent reactive astrocytes (Federoff et al., 1987). Also the functional protein expression by these cells can be modified by culture (Kimelberg et al., 1997). For astrocytic studies in the mammalian CNS we have a succession of questions. 1. What should be considered as the total class of astrocytes? 2. What biological preparations can be most usefully used to study them and by what methods? 3. What hypotheses as to function should we pursue? Pace Golgi, Ramón y Cajal and others of the tremendously fruitful 1880–1920s period it has for some time not been considered adequate to obtain number 3 from insights derived from morphological data and location, but we have to be able to reliably test hypotheses as to function. However, we still need much more reliable data to propose hypotheses for that peculiar pas de deux of data and hypothesis that in a craft like manner is the scientific method. According to Karl Popper hypotheses need to be falsifiable. This means in practice that, although we spend all our time trying to do experiments to “prove” hypotheses, the only conclusive experiment (i.e. test) is the one that disproves. Questions 1 and 3 are clearly interrelated as our hypotheses will determine

200

Table 1 Astrocyte-like cell types in the CNS Morphology

Representative markers

Location

Electrophysiological properties

Developmental properties

Radial glia

Elongate and columnar but with short processes surrounding synapses and blood vessels

Nestin, GFAP (absent in chick and rodents) RC1,2. GLAST, glycogen granules

Early in development in all regions across neural epithelium

Close to selective for K+ . Large negative membrane potentials in presumed counterparts in the adult CNS. i.e. Muller and Bergmann glia

Stem cells

GFAP(+) with processes. Do not show typical undifferentiated morphology of stem cells

GFAP, Nestin

Not yet defined

Stellate GFAP(+) astrocyte

The classic astrocyte of Cajal. Numerous radiating processes from a central or off center cell body giving them a spider like appearance

GFAP, glutamine synthetase. S100␤. GLAST and GLT-1 transporters. Glycogen granules

Early development; all subventricular zones (SVZ). Also occur in adults in the SVZs of lateral frontal cortex and dentate gyrus Uniformly in gray (protoplasmic) and white (fibrous) matter throughout the mammalian CNS

Present very early as developmental scaffolds for migration of neurons. Also thought to be origin of mature stellate astrocytes and more recently to give rise to neurons. Present in adult mammals only as specialized Muller or Bergmann glia Give rise to neurons and macroglia

Reactive astrocytes

Larger than normal stellate astrocytes

Stain intensely for GFAP

Stellate GFAP(−) astrocytes

In appearance same as GFAP(+)

GFAP(−) but tentatively +ve for S100␤

Smooth protoplasmic astrocytes

Name given to them because of their stellate astrocyte appearance but with thinner and smoother processes containing few filaments. Now referred to as OPCs

NG2(+) and GFAP(−). Do not have GLAST or GLT-1 so lack these astrocytic features

Throughout CNS in reaction to neuronal damage or death, occurring within hours or days Numerous in some regions such as cortex but less GFAP(−) astrocytes in other regions such as hippocampus In gray matter in most regions. Persist after main wave of oligodendrogenesis and are quite numerous (5–10% of all cells in adult)

Quite variable as opposed to classic view. Some only have largely passive K+ currents. Others different voltage gated K+ currents and small Na+ currents. Functions of such current complexity in non excitable cells is unclear. Ionotropic glutamate receptors also present All show voltage sensitive K+ and Na+ currents

Stellate astrocytes are mature astrocytes present from a few days before birth developing to a peak at around 15 days and then persisting. Surround synapses and blood vessels and form glia limitans. Numerous ionotropic and metabotropic receptors, ion channels and transporters present. Secrete many trophic and growth factors Show some reversion to earlier, developing astrocytes

Unclear. Not studied in detail for strict comparison with GFAP(+) astrocytes

Unclear

Show predominantly voltage gated K+ currents and small Na+ currents. AMPA ionotropic receptors

Present at E14 in the rodent and do give rise to oligodendrocytes in the young animal. Few astrocytic-like physiological properties in the adult. Workers now consider them as on the oligodendrocyte lineage

H.K. Kimelberg / Neurochemistry International 45 (2004) 191–202

Astrocyte type

H.K. Kimelberg / Neurochemistry International 45 (2004) 191–202

what experiments we will do and vice versa. The answer to 2 is clear; preparations that are the least perturbed from the in vivo state yet still allow controlled and informative experiments. An emerging picture of the astrocyte is that perhaps its identity is unclear because it is unusually multifunctional. Perhaps that is why we have problems in defining this class because this multifunctionality leads to plasticity of expression of protein markers. These are, after all, only the major cell constituents that allow it to perform its specific functions and therefore their levels will change as the cells functions change. Thus GFAP may organize astrocytes for their perivascular and perisynaptic locations (see above). The problem with the astrocyte class as presently conceived is that following the pioneering histologists, who defined these cells by morphology, we now are trying to squeeze all our new findings into what may be a taxonomically overly restrictive mold. Perhaps redefining what is an astrocyte will give us the intellectual freedom to define astrocytes in a less restrictive way. We do, however, have to start somewhere. One could take the approach of a classification “lumper” and include all CNS cells that express GFAP by an agreed on staining procedure with a specified antibody and/or expresses messenger RNA for GFAP as determined by RT-PCR, which is experimentally more consistent and replicable. A “splitter” approach might well restrict the term astrocyte to stellate GFAP(+) cells that have always been agreed on to be astrocytes and use different terms for all the other cells. Based on the well-tested classification of living organisms, lineage relationships might well be an important consideration. Thus although seals resemble fish in that they live in the water they are classified as mammals based on their anatomy and physiology. So there you have it. What is an astrocyte is a search that continues and it is all about definition; what the practitioners of the art define as an astrocyte at different times which, in hindsight, and it can be argued currently, never seems to adequately define these protean cells. There is still little clarity here even though this field can be considered 150, but more accurately is around 100 years old. But the techniques applied in the last decade are far different to those of 100 to even 10 years ago. Additionally because astrocytes have not been studied extensively the definition is still a work in progress, because our data base on the properties of these non neuronal cells of the CNS is too limited. This cannot reasonably be considered a criticism but an inevitable result of the scientific method, which in spite of the current emphasis on hypotheses, is basically empirical and data-based. Hypotheses are illustrative and helpful in suggesting what data may be useful. As Newton wrote; “and therefore because I have observed the heads of some great virtuoso’s to run much upon Hypotheses. . . that some when I could not make them take my meaning, when I spake of the nature of light & colors abstractly, have readily apprehended it when I illustrated my Discourse by an Hypothesis”. However, he

201

considered hypotheses “should not be the test of the truth and reality of things” and were useful only “so far as they may furnish experiments.” (Christianson, 1984). Thus hypotheses are clarifying, determine the next experiments and sometimes are on the right track. This essay on the problem of astrocyte identity, I think, illustrates that we need a bigger and reliable data base of astrocyte properties before we can propose reliable hypotheses as to why they have such properties. However, one can never know when we have a sufficient amount of data to prompt the correct hypothesis in a prepared mind. This is the intuitive and creative part of science that is quite unpredictable.

Acknowledgements Work on the ORAs and VRAs reported in this article were supported by former grant NS 19492 to HKK. I thank Min Zhou help with Fig. 3 and Paul Feustel for alerting me to and calculating the probabilities for the use of more than one marker according to Bayes’s rule and Christian Steinhauser for reading the ms and for helpful comments. References Alvarez-Buylla, A., Seri, B., Doetsch, F., 2002. Identification of neural stem cells in the adult vertebrate brain. Brain Res. Bull. 57, 751–758. Amundson, R.H., Goderie, S.K., Kimelberg, H.K., 1992. Uptake of [3H]serotonin and [3H]glutamate by primary astrocyte cultures II. Differences in cultures prepared from different brain regions. Glia 6, 9–18. Bergles, D.E., Roberts, J.D.B., Somogyi, P., Jahr, C.E., 2000. Glutamatergic synapses on oligodendrocyte precursor cells in the hippocampus. Nature 405, 187–191. Buniatian, G.H., Hartman, H.J., Traub, P., Wiesinger, H., Albinus, M., Nagel, W., Shoeman, R., Mecke, D., Weser, U., 2002. Glial fibrillary acidic protein-positive cells of the kidney are capable of raising a protective biochemical barrier similar to astrocytes: Expression of metallothionein in podocytes. Anatom. Rec. 267, 296–306. Bushong, E.A., Martone, M.E., Jones, Y.Z., Ellisman, M.H., 2002. Protoplasmic astrocytes in CA1 stratum radiatum occupy separate anatomical domains. J. Neurosci. 22, 183–192. Campbell, K., Götz, M., 2002. Radial glia: multi-purpose cells for vertebrate brain development. Trends Neurosci. 25, 235–238. Chanas-Sacre, G., Rogister, B., Moonen, G., Leprince, P., 2000. Radial glia phenotype: origin, regulation, and transdifferentiation. J. Neurosci. Res. 61, 357–363. Chen, Y., Swanson, R.A., 2003. Astrocytes and brain injury. J. Cereb. Blood Flow Metab. 23, 137–149. Christianson, G.E., 1984. In the Presence of the Creator: Isaac Newton and his Times. The Free Press, MacMillan Inc., New York/London, pp. 165, 188 and 532. Connor, J.R., Berkowitz, E.M., 1985. A demonstration of glial filament distribution in astrocytes isolated from rat cerebral cortex. Neuroscience 16, 33–44. Craigie, E.H., (Translator) 1989. Recollections of My Life. Santiago Ramón Cajal. The MIT Press, Cambridge/London, pp. 573–575. DeFelipe, J., Jones, E.G., 1988. Cajal on the Cerebral Cortex. Oxford University Press, pp. 48–50. Doetsch, F., Caille, I., Lim, D.A., Garcia-Verdugo, J.M., Alvarez-Buylla, A., 1999. Subventricular zone astrocytes are neural stem cells in the adult mammalian brain. Cell 97, 703–716.

202

H.K. Kimelberg / Neurochemistry International 45 (2004) 191–202

Duffy, P.E., 1983. Astrocytes: Normal, Reactive and Neoplastic. Raven Press, New York. Eng, L.F., Ghirnikar, R.S., Lee, Y.L., 2000. Glial fibrillary acidic protein: GFAP-thirty-one years (1969–2000). Neurochem. Res. 25, 1439–1451. Eng, L.F., Vanderhaeghen, J.J., Bignami, A., Gerstl, B., 1971. An acidic protein isolated from fibrous astrocytes. Brain Res. 28, 351–354. Federoff, S., Ahmed, I., Opas, M., Kalnins, V.I., 1987. Organization of microfilaments in astrocytes that form in the presence of dibutyryl cyclic amp in cultures, and which are similar to reactive astrocytes in vivo. Neuroscience 22, 255–266. Fishell, G., Kriegstein, A.R., 2003. Neurons from radial glia: the consequences of asymmetric inheritance. Curr. Opin. Neurobiol. 13, 34–41. Gallo, V., Belachew, S., Chittajallu, R., Aguirre, A., Yuan, X., 2003. Postnatal NG2 proteoglycan-expressing progenitor cells are intrinsically multipotent and generate functional neurons. Glia Suppl. 2, 6. Galou, M., Pournin, S., Ensergueix, D., Ridet, J.-L., Tchélingérian, J.L., Lossouarn, L., Privat, A., Babinet, C., Dupouey, P., 1994. Normal and pathological expression of GFAP promoter elements in transgenic mice. Glia 12, 28l–293. Golgi, C., 1885. Sulla fina anatomia degli organi centrali del sisterma nervoso. Riv. Sper. Fremiat. Med. Leg. Alienazione Ment. 11, 72–123. Götz, D.M., Steindler, D.D., 2003. Glia as stem cells in development and adulthood. Glia 43, 1–93. Heins, N., Malatesta, P., Cecconi, F., Nakafuku, M., Tucker, K.L., Hack, M.A., Chapouton, P., Barde, Y.A., Götz, M., 2002. Glial cells generate neurons: the role of the transcription factor Pax6. Nature Neurosci. 5, 308–315. Kálmán, M., Pritz, M.B., 2001. Glial fibrillary acidic protein-immunopositive structures in the brain of a crocodilian, Caiman crocodilus, and its bearing on the evolution of astroglia. J. Comparat. Neurol. 431, 460–480. Kimelberg, H.K., 1999. Current methods and approaches to studying astrocytes: A forum position paper. NeuroTox. 20, 703–712. Kimelberg, H.K., Cai, Z., Rastogi, P., Charniga, C., Goderie, S., Dave, V., Jalonen, T., 1997. Transmitter-induced calcium responses differ in astrocytes acutely isolated from rat brain and in culture. J. Neurochem. 68, 1088–1098. Kimelberg, H.K., 1986. Catecholamine and serotonin uptake in astrocytes. In: Fedoroff, S., Vernadakis, A. (Eds.), Astrocytes: Biochemistry, Physiology, and Pharmacology of Astrocytes, vol. 2. Academic Press, Orlando, FL, pp. 107–127. Levine, J.M., Card, J.P., 1987. Light and electron microscopic localization of a cell surface antigen (NG2) in the rat cerebellum: association with smooth protoplasmic astrocytes. J. Neurosci. 7, 2711–2720. Liu, Y., Wu, Y., Lee, J.C., Xue, H., Pevny, L.H., Kaprielian, Z., Rao, M.S., 2002. Oligodendrocyte and astrocyte development in rodents: an in situ and immunohistological analysis during embryonic development. Glia 40, 25–43. Malatesta, P., Hartfuss, E., Götz, M., 2000. Isolation of radial glial cells by fluorescent-activated cell sorting reveals a neuronal lineage. Development 127, 5253–5263. Malter, J.S., 2001. Regulation of mRNA stability in the nervous system and beyond. J. Neurosci. Res. 66, 311–316. Matthias, K., Kirchhoff, F., Seifert, G., Huttmann, K., Matyash, M., Kettenmann, H., Steinhauser, C., 2003. Segregated expression of AMPA-type glutamate receptors and glutamate transporters defines distinct astrocyte populations in the mouse hippocampus. J. Neurosci. 23, 1750–1758. Nishiyama, A., Chang, A.S., Trapp, B.D., 1999. NG2+glial cells: a novel glial cell population in the adult brain. J. Neuropathol. Exp. Neurol. 58, 1113–1124. Nolte, C., Matyash, M., Pivneva, T., Schipke, C.G., Ohlemeyer, C., Hanisch, U.K., Kirchhoff, F., Kettenmann, H., 2001. GFAP promotercontrolled EGFP-expressing transgenic mice: a tool to visualize astrocytes and astrogliosis in living brain tissue. Glia 33, 72– 86.

Ogata, K., Kosaka, T., 2002. Structural and quantitative analysis of astrocytes in the mouse hippocampus. Neuroscience 113, 221–233. Privat, A., Rataboul, P., 1986. Fibrous and protoplasmic astrocytes. In: Federoff, S., Vernadakis, V. (Eds.), Astrocytes: Development, Morphology, and Regional Specialization of Astrocytes. Academic Press, Orlando, pp. 105–129. Rakic, P., 1971. Neuron-glia relationship during granule cell migration in developing cerebellar cortex. A Golgi and electronmicroscopic study in Macacus rhesus. J. Comparat. Neurol. 141, 283–312. Rakic, P., 1988. Specification of cerebral cortical areas. Science 241, 170–176. Ramón y Cajal, S., 1913. Contribucion al conocimento de la neuroglia del cerebro humano. Trab. Lab. Invest. Biol. Univ. Madrid 11, 255–315. Reichenbach, A., 1987. Quantitative and qualitative morphology of rabbit retinal glia. A light microscopical study on cells both in situ and isolated by papaine. J. Hirnforsch. 28, 213–220. Ridet, J.L., Malhotra, S.K., Privat, A., Gage, F.H., 1997. Reactive astrocytes: cellular and molecular cues to biological function. TINS 20, 570–577. Roots, B.I., 1986. Phylogenetic development of astrocytes. In: Federoff, S., Vernadakis, A. (Eds.), Astrocytes. Academic Press, New York, pp. 1–34. Schmechel, D.E., Rakic, P., 1979. A Golgi study of radial glial cells in developing monkey telencephalon: morphogenesis and transformation into astrocytes. Anat. Embryol. (Berl) 156, 115–152. Schools, G., Zhou, M., Kimelberg, H.K., 2003. Electrophysiologically “Complex” Glial Cells Freshly Isolated from the Hippocampus are Immunopositive for the Chondroitin Sulfate Proteoglycan NG2. J. Neurosci. Res. 73 (6), 765–777. Seaberg, R.M., van der Kooy, D., 2003. Stem and progenitor cells: the premature desertion of rigorous definitions. Trends Neurosci. 26, 125– 131. Seri, B., Garcia-Verdugo, J.M., McEwen, B.S., Alvarez-Buylla, A., 2001. Astrocytes give rise to new neurons in the adult mammalian hippocampus 1. J. Neurosci. 21, 7153–7160. Simard, M., Arcuino, G., Takano, T., Liu, Q.S., Nedergaard, M., 2003. Signaling at the gliovascular interface. J. Neurosci. 23 (27), 9254–9262. Somjen, G.G., 1988. Nervenkitt: notes on the history of the concept of neuroglia. Glia 1, 2–9. Steindler, D.A., 1993. Glial boundaries in the developing nervous system. Annu. Rev. Neurosci. 16, 445–470. Steinhauser, C., 1993. Electrophysiologic characteristics of glial cells. Hippocampus 3, 113–124. Temple, S., Alvarez-Buylla, A., 1999. Stem cells in the adult mammalian central nervous system. Curr. Opin. Neurobiol. 135–141. Ventura, R., Harris, K.M., 1999. Three-dimensional relationships between hippocampal synapses and astrocytes. J. Neurosci. 19, 6897–6906. Vives, V., Alonso, G.R., Solal, A.C., Joubert, D., Legraverend, C., 2003. Visualization of S100B-positive neurons and glia in the central nervous system of EGFP transgenic mice. J. Comp. Neurol. 457, 404–419. Walz, W., 2000. Controversy surrounding the existence of discrete functional classes of astrocytes in adult gray matter. Glia 31, 95–103. Walz, W., Lang, M.K., 1998. Immunocytochemical evidence for a distinct GFAP-negative subpopulation of astrocytes in the adult rat hippocampus. Neurosci. Lett. 257, 127–130. Zhou, M., Kimelberg, H.K., 2000. Freshly isolated astrocytes from rat hippocampus show two distinct current patterns and different [K+ ]o uptake capabilities. J. Neurophysiol. 84, 2746–2757. Zhou, M., Kimelberg, H.K., 2001. Freshly isolated hippocampal CA1 astrocytes comprise two populations differing in glutamate transporter and AMPA receptor expression. J. Neurosci. 21, 7901–7908. Zhou, M., Schools, G.P., Kimelberg, H.K., 2000. GFAP mRNA positive glia acutely isolated from rat hippocampus predominantly show complex current patterns. Mol. Brain Res. 76, 121–131. Zhuo, L., Sun, B., Zhang, C.L., Fine, A., Chiu, S.Y., Messing, A., 1997. Live astrocytes visualized by green fluorescent protein in transgenic mice. Dev. Biol. 187, 36–42.